volume seventeen • number two March 2011

Peer-Reviewed Journal of the Academy of Managed Care Pharmacy

JMCP

■ Jo

urn

al o

f Ma

na

ged

Ca

re Ph

arm

acy ■

March

2011

■ Vo

l. 17, No

. 2 ■

Pages 9

7-176

Page

Discontinuation Rates and Health Care Costs in Adult Patients Starting Generic Versus Brand SSRI or SNRI Antidepressants in Commercial Health Plans

INR Goal Attainment and Oral Anticoagulation Knowledge of Patients Enrolled in an Anticoagulation Clinic in a Veterans Affairs Medical Center

123

133156

143 Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

104 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 7, No. 2 www.amcp.org

AMCP HeAdquArters

100 North Pitt St., Suite 400 Alexandria, VA 22314Tel: 703.683.8416  •  Fax: 703.683.8417

BoArd of direCtors

President: Brian Sweet, BS Pharm, MBA, Grand Island, NY

President-Elect: David L. Clark, RPh, MBA, RegenceRx/The Regence Group, Portland, OR

Past President: Shawn Burke, RPh, FAMCP, Coventry Health Care, Inc., Kansas City, MO

Treasurer: John D. Jones, RPh, JD, FAMCP, Prescription Solutions, Irvine, CA

Secretary: Judith A. Cahill, CEBS, Academy of Managed Care Pharmacy, Alexandria, VA (Executive Director, AMCP)

Director: Diana I. Brixner, RPh, PhD, University of Utah, Salt Lake City, UT

Director: Douglas S. Burgoyne, PharmD, RPh, VRx Pharmacy Services, Salt Lake City, UT

Director: Babette S. Edgar, PharmD, MBA, Edgar Consulting Group, Reisterstown, MD

Director: Raulo S. Frear, PharmD, RegenceRx/The Regence Group, Boise, ID

Director: William H. Francis, RPh, MBA, UPH Health Plans, Tucson, AZ

Advertising

Advertising for the Journal of Managed Care Pharmacy is accepted in accordance with the advertising policy of the Academy of Managed Care Pharmacy.

For advertising information, contact: Bob HeimanRH Media LLC1814 East Route 70, Suite 350Cherry Hill, NJ 08003 Tel.: 856.673.4000 Fax: 856.673.4001 E-mail: bob.rhmedia@comcast.net

editoriAl

Questions related to editorial content should be directed to JMCP Editor-in-Chief Frederic R. Curtiss: fcurtiss@amcp.org. Manuscripts should be submitted electronically at jmcp.msubmit.net. For questions about  submission, contact Peer Review Administrator Jennifer A. Booker: jmcpreview@amcp.org; 703.317.0725.

suBsCriPtions

Annual subscription rates: USA, individuals, institutions–$90; other countries–$120. Single copies cost $15. Missing issues are replaced free of charge up to 6 months after date of issue. Send requests to AMCP headquarters.

rePrints

For article reprints and Eprints, contact Tamara Smith at Sheridan Reprints, 800.352.2210, ext. 8230; tsmith@tsp.sheridan.com. Information for ordering reprints or Eprints and instructions for permission for use of JMCP content are available at www.amcp.org.

All articles published represent the opinions of the authors and do not reflect the official policy or views of the Academy of Managed Care Pharmacy or the authors’ institutions unless so specified. Copyright © 2011, Academy of Managed Care Pharmacy. All rights reserved. No part of this publication may be repro-duced or transmitted in any form or by any means, electronic or mechanical, without written permission from the Academy of Managed Care Pharmacy.

C o n t e n t s

■ reseArCH

123 Discontinuation Rates and Health Care Costs in Adult Patients Starting Generic Versus Brand SSRI or SNRI Antidepressants in Commercial Health Plans

Anna Vlahiotis, MA; Scott T. Devine, PhD; Jeff Eichholz, PharmD; and Adam Kautzner, PharmD

133 INR Goal Attainment and Oral Anticoagulation Knowledge of Patients Enrolled in an Anticoagulation Clinic in a Veterans Affairs Medical Center

Jennifer W. Baker, PharmD, BCPS; Kristi L. Pierce, PharmD; and Casey A. Ryals, PharmD

■ ConteMPorArY suBJeCt

143 Pharmaceutical Step-Therapy Interventions: A Critical Review of the LiteratureBrenda R. Motheral, RPh, MBA, PhD

■ dePArtMents

108 Cover ImpressionsSir Miles Davis (2005)David Lloyd GloverSheila Macho, Cover Editor

156 Editorial What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

Kathleen A. Fairman, MA, and Frederic R. Curtiss, PhD, RPh, CEBS

175 Letter Case Report of Specialty Pharmacy Management of Hemophilia

Aggie Gilbert, RN, and Brian Tonkovic, PharmD

176 Letter Online Drug Pricing Tools Deserve More Attention

Alexander Grunewald, PhD

volume 17, no. 2

106 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 7, No. 2 www.amcp.org

JMCP accepts for consideration manuscripts pre-pared according to the Uniform Requirements for Manu scripts Submitted to Biomedical Journals.1

■■ Manuscript PreparationManuscripts should include, in this order: title page, abstract, text, references, tables, and figures (see Manuscript Submis sion Checklist for details).

JMCP abstracts should be carefully written nar-ratives that contain all of the principal quantitative and qualita tive findings, with the outcomes of statistical tests of comparisons where appropriate. Abstracts are required for all manuscript submis-sions except Commentaries and Letters. The format for the abstract is Background, Objective, Methods, Results, Conclusion.

For descriptions of editorial content, see “JMCP Editorial Policy” in this Journal or at www.amcp.org (Manuscript/Supplement Submission). Please note:•  The JMCP Peer Review Checklist is the best guide

for authors to improve the likelihood of success in the JMCP peer-review process. It is available at: www.amcp.org (“JMCP Peer Review Checklist and Guidelines” tab).

•  A subsection in the Discussion labeled “Limitations” is generally appropriate for all articles except Commentaries and Letters.

•  Most articles should incorporate or at least acknowledge the relevant work of others pub-lished previously in JMCP (see “Article Index by Subject Category” at www.amcp.org).

•  Product trade names may be used only once for the purpose of providing clarity for readers, gen-erally at the first citation of the generic name in the article but not in the abstract.

•  Many articles involve research that may pose a threat to either patient safety or privacy. It is the responsibility of the principal author to ensure that the manuscript is submitted with either the result of review by the appropriate institutional review board (IRB) or a statement of why the research is exempt from IRB review (see “Policy for Protecting Patient Safety and Privacy” at www.amcp.org), Manuscript/Supplement Submission.

■■ Reference StyleReferences should be prepared following modi-fied AMA style. All reference numbers in the manuscript should be superscript (e.g., 1 ). Each unique reference should have only one reference number. If that reference is cited more than once in the manuscript, the same number should be used. Do not use ibid or op cit for JMCP references. Please provide Web (hyperlink) addresses for all free access references. An access date should be included for every URL except links to JMCP arti-cles. See examples 2 and 3 in the second column. Here are examples of the style format for common types of references:1. Journal article — (list up to 6 authors; if more,

REFEREncE1. International Committee of Medical Journal Editors. Uniform requirements for manuscripts submitted to  biomedical journals: writing and editing for biomedical publication. Updated April 2010. Available at: http://www.icmje.org/urm_full.pdf. Accessed January 4, 2011.

The Journal of Managed Care Pharmacy,

including supplements, is indexed by MEDLINE/

PubMed, the International Pharmaceutical

Abstracts (IPA), Science Citation Index

Expanded (SCIE), Current Contents/Clinical

Medicine (CC/CM), and Scopus.

JMCP Author Guidelines

list only the first 3 and add et al.): Kastelein JJ, Akdim F, Stroes ES, et al., for the ENHANCE Investigators. Simvastatin with or without ezetimibe in familial hypercholesterolemia. N Engl J Med. 2008;358(14):1431-43. Available at: http://content.nejm.org/cgi/reprint/358/14/1431.pdf. Accessed January 4, 2011.2. no author given — Anonymous. More patients leaving Rxs at pharmacy counter. Manag Care. 2010;19(4):49. Available at: http://www.managedca-remag.com/archives/1004/1004.formfiles.html. Accessed January 4, 2011.3. Journal or magazine paginated by issue — McKinney M. Alarm fatigue sets off bells. Mod Healthc. 2010;40(15):14.4. Book or monograph — Tootelian DH, Gaedeke RM. Essentials of Pharmacy Management. St. Louis, MO: C.V. Mosby; 1993.5. Book or monograph with editor, compiler, or chairperson as author — Chernow B, ed. Critical Care Pharmacotherapy. Baltimore, MD: Williams & Wilkins; 1995.6. chapter in a book — Kreter B, Michael KA, DiPiro JT. Antimicrobial prophylaxis in surgery. In: DiPiro JT, Talbert RL, Hayes PE, Yee GC, Matzke GR, Posey LM, eds. Pharmacotherapy: A Pathophysiologic Approach. Norwalk, CT: Appleton & Lange; 1992:1811-12.7. Government agency publication — National Heart, Lung, and Blood Institute. Expert panel report 3: guidelines for the diagnosis and management of asthma. Full report 2007. August 28, 2007. Available at: http://www.nhlbi.nih.gov/guidelines/asthma/asth-gdln.pdf. Accessed January 4, 2011.8. Website - other online publication — Minnesota Department of Health. Essential Benefit Set Work Group. Background paper. September 4, 2009. Available at: http://www.health.state.mn.us/healthre-form/essential/EBS_Background.pdf. Accessed January 4, 2011.9. newspaper article — Abelson R, Pollack A. Medicare widens drugs it accepts for cancer. NY Times. January 26, 2009. Available at: http://www.nytimes.com/2009/01/27/health/27cancer.html?_r=2&ref=health. Accessed January 4, 2011.10. Online news article — Edwards J. How Pfizer hid a $2.3 bill. Bextra settlement in plain sight. BNet. January 26, 2009. Available at: http://industry.bnet.com/pharma/1000656/how-pfizer-hid-a-23-bill-bextra-settlement-in-plain-sight/. Accessed January 4, 2011.11. Drug label – prescribing information — Xolair (omalizumab) for subcutaneous use. Genentech. July 2010. Available at: http://www.gene.com/gene/

products/information/pdf/xolair-prescribing.pdf. Accessed January 4, 2011.12. Dissertation or thesis — Youssef NM. School adjustment of children with congenital heart dis-ease [dissertation]. Pittsburgh, PA: University of Pittsburgh; 1988.13. Paper or poster presented at a meeting — Gleason PP, Starner CI, Hyland-Marciniak B. Erythropoiesis-stimulating agent trends and utiliza-tion management opportunity. Poster presented at: 2010 AMCP Annual Meeting; April 9, 2010; San Diego, CA. Available at. http://www.amcp.org/data/jmcp/141-168.pdf.14. Letter or editorial — Barbuto JP. Categorizing patients from medical claims data – the influence of GIGO [letter]. J Manag Care Pharm. 2004;10(6):559-60. Available at: http://www.amcp.org/data/jmcp/Letters_559-566.pdf.15. Journal supplement — Academy of Managed Care Pharmacy. AMCP guide to pharmaceutical pay-ment methods, 2009 update (version 2.0). J Manag Care Pharm. 2009;15(6 Suppl A):S1-S61. Available at: http://www.amcp.org/data/jmcp/1002.pdf.

■■ Manuscript SubmissionA complete list of documents needed for submission to JMCP appears on the Manuscript/Supplement Submission page at www.amcp.org. Prior to peer review, all manuscripts are reviewed by the editors and/or members of the Editorial Advisory Board for appropriateness of the topic for JMCP, methodologi-cal transparency, and compliance with submission requirements. See Author Guidelines for description of the “pre-review process.” Peer review generally requires 4-6 weeks but may extend as long as 12 weeks in unusual cases. Solicited manuscripts are subject to the same peer-review standards and edito-rial policy as unsolicited manuscripts.

Disclosures and conflicts of interest: Manuscript submissions should (a) include a statement that iden-tifies the nature and extent of any financial interest or affiliation that any author has with any company, product, or service discussed in the manuscript and clearly indicates the source(s) of funding and finan-cial support and (b) be accompanied by completed and signed author attestation forms for the principal author and each coauthor.

■■ Manuscript Submission ChecklistBefore submitting your manuscript to JMCP, please check to see that your package includes all items in the JMCP Manuscript Submission Checklist, available at www.amcp.org (JMCP/Manuscript/Supplement Submission).

108 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 7, No. 2 www.amcp.org

It seems that internationally acclaimed modern Impressionist David Lloyd Glover was destined to

become an artist. Born in 1949 in the city of Victoria on Vancouver Island in British Columbia, Canada, he grew up on a 600-acre seaside estate known as Hatley Castle, formerly the Scottish baronial manor of James Dunsmuir, Lieutenant Governor. It had been con-verted in 1948 to the Royal Roads Military College, where his father served as an administrator. The estate’s beautiful English, Japanese, and Italian gardens had an early influ-ence on Glover and his artwork, and inspired his lifelong interest in man’s relationship with nature. A couple of creative family members also influenced the budding young artist. His uncles, Guy Glover, a famous film producer, actor, and head of the National Film Board of Canada, and Norman McLaren, a world-renowned film artist, animator, and painter, both encouraged him in his artistic pursuits. Possessing remark-able talent for his age, Glover won a city-wide, juried painting exhibition in 1958 when he was only 9 years old. At 14, he was accepted as a student at the Vancouver School of Art despite its minimum age requirement of 18. 

By his late teens, Glover had developed remarkable skills in  pen  and  ink  by  studying  the  techniques  of  19th-century English illustrators. His mastery of this style led to an editorial illustrator position at the Victoria Times newspaper. Within 3 years, the newspaper selected Glover to be its political cartoon-ist. His professional accomplishments also include graphic design, art direction, and the founding of an award-winning advertising agency, Truman/Glover Advertising, Ltd., in 1974.

After selling his interest in the company in 1977, Glover returned to his first love of fine art. He began with watercolors executed in the style of the British masters, and his paint-ings were met with instant success. Glover soon developed a strong following and became one of the best-selling artists at Vancouver’s most respected gallery, Harrison Galleries.

In 1988, Glover’s art dealer in Los Angeles encouraged him to come to Southern California to work as an artist in residence. “After a few months, California got under my skin in a big way and I decided to make it my home,” he says. During the next few years, several art galleries in Los Angeles began exhibiting his paintings. Enthusiastic art collectors and Hollywood celebrities quickly snapped up Glover’s original works and serigraph edi-tions. Since then, he has exhibited his art at other noteworthy galleries in California, New York, Arizona, and Louisiana, as well as galleries in Canada, Mexico, and Japan, where he ranks among the country’s most widely collected original artists.

Inspired by the great French Impressionists Monet, Renoir, 

and Cézanne, Glover began painting in oil in the early 1990s. He promptly mastered this medium, too. In an article that appeared in the December 2005 issue of American Art Collector magazine, Glover was described as a “sublime  colorist,”  whether  working in watercolor or oil. He also paints in acrylic, a very versatile medium.

The Art Brillant gallery on Rodeo Drive in Beverly Hills held a special exhibition of Glover’s art in June 2005. According to the biography on his website, davidlglover.com,  “The  show featured Glover’s jazz portrait series of

the great icons in American jazz. Images of the greats like Ella Fitzgerald,  Louis Armstrong,  John Coltrane,  and Miles Davis were featured, along with portraits of some of the contempo-rary legends like Wayne Shorter and Herbie Hancock. Herbie also made a personal appearance at the show in honor of his friend David Lloyd Glover.” Wayne Shorter happens to be a friend of the artist as well. Both Hancock and Shorter played in the Miles Davis quartet at one time. Davis, the iconic American trumpeter, bandleader, and composer, was at the forefront of several major developments in jazz music, including bebop, cool jazz, and jazz fusion. Glover says that he titled this issue’s fabulous cover painting Sir Miles Davis because the musician was  knighted  by  the  French  in  1991. He  cropped  the  image to emphasize Davis’ upper body and chose bright hues for his jacket to reflect his colorful music and persona.

In addition to Glover’s jazz portraits, he has painted por-traits of other pop stars such as John Wayne, Marlon Brando, Frank  Sinatra,  Elvis,  and  Michael  Jackson.  “Capturing  the images of the icons of American popular culture has always been of great  interest to me,” he says. “I  try to tell  their story with colors alone.” His vivid portraits are extraordinary—they truly convey the essence of each individual.

To see the list of galleries and websites that carry Glover’s magnificent iconographic and landscape art, visit the “Dealers & Links” page on his website, davidlglover.com/9.html. His work  is  also  available  online  through  the  Fine  Art  America website, fineartamerica.com—simply enter the name David Lloyd Glover in the search box.

Sheila MachoCover Editor

cOVER cREDITDavid Lloyd Glover, Sir Miles Davis, acrylic on canvas. Los Angeles, California. Copyright © 2005.

SOURcESInterview with the artist.http://davidlglover.com.

C o v e ri M P r e s s i o n s

A b o u t o u r c o v e r a r t i s t

Sir Miles Davis (2005) ■ David Lloyd Glover

“I try to tell their story with colors alone.”

114 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 7, No. 2 www.amcp.org

editoriAl Mission

JMCP publishes peer-reviewed original research manuscripts, subject reviews, and other content intended to advance the use of the scientific method, including the interpretation of research findings in managed care pharmacy. JMCP is dedicated to improving the quality of care delivered to patients served by managed care pharmacy by providing its readers with the results of scientific investigation and evaluation of clinical, health, service, and economic outcomes of pharmacy services and pharmaceutical interventions, including formulary management. JMCP strives to engage and serve professionals in pharmacy, medicine, nursing, and related fields to optimize the value of pharmaceutical products and pharmacy services delivered to patients. JMCP employs extensive bias-management procedures intended to ensure the integrity and reliability of published work.

editoriAl AdvisorY BoArd

The JMCP Editorial Advisory Board is chaired by Marvin D. Shepherd, PhD, Center for Pharmacoeconomic Studies,College of Pharmacy, University of Texas at Austin; Thomas Delate, PhD, Kaiser Permanente of Colorado, Aurora, serves as vice chair.

They and the other advisers review manuscripts and assist in the determination of the value and accuracy of information provided to readers of JMCP.

J. Daniel Allen, PharmD, RegenceRx, Portland, OR John P. Barbuto, MD, Modify Motion, LLC, Salt Lake City, UTchristopher F. Bell, MS, Global Health Outcomes, GlaxoSmithKline

Research & Development, Research Triangle Park, NCJoshua Benner, PharmD, ScD, Brookings Institution, Washington, DCScott A. Bull, PharmD, ALZA Corporation, Mt. View, CAnorman V. carroll, PhD, School of Pharmacy, Virginia Commonwealth

University, Richmond, VAMark conklin, PharmD, MS, Highmark Blue Cross Blue Shield,

Pittsburgh, PAEric J. culley, PharmD, MBA, Highmark Blue Cross Blue Shield,

Pittsburgh, PATimothy cutler, PharmD, School of Pharmacy, University of California,

Sacramento, CAGregory W. Daniel, RPh, MPH, PhD, HealthCore, Inc., Wilmington, DEThomas Delate, MS, PhD, Kaiser Permanente of Colorado, Aurora, COQuan V. Doan, PharmD, MSHS, Outcomes Insights, Inc., Orange, CAPatrick P. Gleason, PharmD, BcPS, Prime Therapeutics, LLC, Eagan, MNcharnelda Gray Lewis, PharmD, BcPS, Kaiser Permanente, Atlanta, GA noelle Hasson, PharmD, Veterans Affairs Palo Alto Health Care System,

Palo Alto, CALillian Hsieh, MHS, PharmD, CVS Pharmacy, San Mateo, CAMark Jackson, BScPhm, Bcomm, RPh, Green Shield Canada,

Windsor, OntarioRichard A. Kipp, MAAA, Milliman USA, Wayne, PAStephen J. Kogut, PhD, MBA, College of Pharmacy, University of

Rhode Island, Kingston, RIDaniel c. Malone, PhD, RPh, College of Pharmacy, University of Arizona,

Tucson, AZBradley c. Martin, PharmD, PhD, College of Pharmacy, University of

Arkansas, Little Rock, AKRobert L. Ohsfeldt, PhD, School of Rural Public Health, Texas A&M

Health Science Center, College Station, TX

Daniel A. Ollendorf, MPH, Institute for Clinical and Economic Review, Boston, MA

Mary Jo V. Pugh, Rn, PhD, South Texas Veterans Healthcare System, San Antonio, TX

Brian J. Quilliam, PhD, RPh, College of Pharmacy, University of Rhode Island, Kingston, RI

Marsha Raebel, PharmD, Kaiser Permanente of Colorado, Denver, COElan Rubinstein, PharmD, MPH, EB Rubinstein Associates, Oak Park, CA Jeremy A. Schafer, PharmD, MBA, Prime Therapeutics, LLC, Eagan, MNJordana Schmier, MA, Exponent, Alexandria, VAMarvin D. Shepherd, PhD, College of Pharmacy, University of Texas,

Austin, TX Linda M. Spooner, PharmD, BcPS, Massachusetts College of Pharmacy,

Worcester, MAJoshua J. Spooner, PharmD, MS, School of Pharmacy, Western New

England College, Springfield, MAMarilyn Stebbins, PharmD, CHW Medical Foundation, University of 

California, Sacramento, CAKaren Stockl, PharmD, Prescription Solutions, Irvine, CABurgunda V. Sweet, PharmD, FASHP, University of Michigan Health

System, Ann Arbor, MIBrian Sweet, BS Pharm, MBA, Grand Island, NYconnie A. Valdez, PharmD, MSEd, BcPS, School of Pharmacy, Aurora,

CO Peter Whittaker, PhD, School of Medicine, Wayne State University,

Detroit, MIVincent Willey, PharmD, University of the Sciences in Philadelphia,

Philadelphia, PA Karen Worley, PhD, Competitive Health Analytics, Humana Inc.,

Louisville, KY Andrew Yu, PhD, Analysis Group, Boston, MA

Associate editor, Kathleen A. Fairman, MA 602.867.1343, kfairman@amcp.org

Cover editor, Sheila Macho, 952.431.5993, jmcpcoverart@amcp.orgCopy editor, carol Blumentritt

Peer review Administrator, Jennifer A. Booker 703.317.0725, jmcpreview@amcp.org

graphic designer, Margie Hunter, 703.297.9319, mhunter@amcp.org

editoriAl stAff

editor-in-ChiefFrederic R. curtiss, PhD, RPh, cEBS

830.935.4319, fcurtiss@amcp.org

Publisher Judith A. cahill, cEBS, Executive Director, Academy of Managed Care Pharmacy

Journal of Managed Care Pharmacy (ISSN 1083–4087) is published 9 times per year and is the official publication of the Academy of Managed Care Pharmacy (AMCP), 100 North Pitt St., Suite 400, Alexandria, VA 22314; 703.683.8416; 800.TAP.AMCP; 703.683.8417 (fax). The paper used by the Journal of Managed Care Pharmacy meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper) effective with Volume 7, Issue 5, 2001; prior to that issue, all paper was acid-free. Annual membership dues  for AMCP include $60 allocated for the Journal of Managed Care Pharmacy. POSTMASTER: Send address changes to JMCP, 100 North Pitt St., Suite 400, Alexandria, VA 22314.

118 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 7, No. 2 www.amcp.org

■■ Editorial Content and Peer ReviewAll articles, editorials, and commentary in JMCP undergo peer review; articles undergo blinded peer review. Letters may be peer reviewed to ensure accuracy. The funda mental departments for manu-script submission are:  •  Research  •  Subject Reviews  •  Formulary Management  •  Contemporary Subjects  •  Brief Communications  •  Commentary/Editorials  •  Letters All manuscript submissions except Commentaries and Letters should include an abstract and 1-3 takeaway bullet points in each of 2 sections that immediately follow the abstract for “what is already known about this subject” and “what this study adds.”  For manuscript preparation requirements, see “JMCP Author Guidelines” in this Journal or at www.amcp.org.

■■ ResearchThese are well-referenced articles based on original research that has not been published elsewhere and reflects use of the scientific method. The research is guided by explicit hypotheses that are stated clearly by the authors.

■■ Subject ReviewsThese are well-referenced, comprehensive reviews of subjects relevant to managed care pharmacy. The Methods section in the abstract and in the body of the manuscript should make clear to the reader the source of the material used in the review, including the specific criteria used for inclusion and exclusion of information and the number of articles included and excluded by each criterion. Narrative reviews, defined as noncompre-hensive reviews that cover only a portion of the lit-erature on a topic, are not considered for publica-tion by JMCP. However, articles of this type may be considered as Commentary.

■■ Formulary ManagementThese are well-referenced, comprehensive reviews of subjects relevant to formulary management meth-ods or procedures in the conduct of pharmacy and therapeutics (P&T) committees and generally include description and interpretation of clinical evidence and comparative cost information.

■■ Contemporary SubjectsThese are well-referenced submissions that are par-ticularly timely or describe research conducted in pilot projects. Contemporary Subjects, like all arti-cles in JMCP, must describe the hypothesis or hypotheses that guided the research, the principal methods, and results.

■■ Brief CommunicationsThe results of a small study or a descriptive analy-sis that does not fit in other JMCP departments may be submitted as a Brief Communication. Brief Communications may warrant an Abstract with the typical JMCP categories (Background, Objective, Methods, Results, Conclusion).

■■ CommentaryThese submissions should be relevant to managed care pharmacy and address a topic of contempo-rary interest: they do not require an abstract but should include references to support statements.

■■ LettersIf the letter addresses a previously published article, an author response may be appro priate. (See “Letter to the Editor” instruc tions at www.amcp.org.)

■■ Advertising and Disclosure PoliciesAll aspects of the advertising sales and solicitation process are completely independent of the editorial process. Advertising is positioned either at the front or back of the Journal, and advertising is not accept-ed for placement opposite or near subject-related editorial content. The Academy of Managed Care Pharmacy endorses the Uniform Requirements for Manuscript Submissions to Biomedical Journals, available at: http://www.icmje.org/ including editorial freedom and management of conflicts of interest. Financial disclosure, conflict-of-interest statements, and

JMCP Mission Statement and Editorial Policy

editorial Mission and policies

JMCP publishes peer-reviewed original research manuscripts, subject reviews, and other content intended to advance the use of the scientific method, including the interpretation of research findings in managed care pharmacy. JMCP is dedicated to improving the quality of care delivered to patients served by managed care pharmacy by providing its readers with the results of scientific investigation and evaluation of clinical, health, service, and economic outcomes of pharmacy ser-vices and pharmaceutical interventions, including formulary management. JMCP strives to engage and serve professionals in pharmacy, medicine, nursing, and related fields to optimize the value of pharmaceu tical products and pharmacy services delivered to patients.

JMCP employs extensive bias-management procedures that include (a) full disclosure of all sources of potential bias and conflicts of interest, nonfinancial as well as financial; (b) full disclosure of potential conflicts of interest by reviewers as well as authors; and (c) accurate attribution of each author’s contribution to the article. aggressive bias-management methods are necessary to ensure the integrity and reliability of published work.

editorial content is determined by the editor-in-chief with suggestions from the editorial advisory Board. the views and opinions expressed in JMCP do not necessarily reflect or represent official policy of the academy of Managed care pharmacy or the authors’ institutions unless specifically stated.

Editorial OfficeAcademy of Managed Care Pharmacy100 North Pitt St., Suite 400Alexandria, VA 22314Tel.: 703.683.8416Fax: 703.683.8417E-mail: jmcpreview@amcp.org

Advertising Sales OfficeRH Media LLC 1814 East Route 70, Suite 350Cherry Hill, NJ 08003Tel.: 856.673.4000 Fax: 856.673.4001E-mail: bob.rhmedia@comcast.net

author attestations are required when manu-scripts are submitted, and these disclosures accompany the article in abstracted form if the article is published. See JMCP Author Guidelines in this issue or online including complete description of disclosure policies and require-ments for author attestations at www.amcp.org. See JMCP advertising opportunities at: www.amcp.org. Contact the Advertising Sales Office to receive a Media Kit.

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 123

•Evidence suggests that generic selective serotonin reuptakeinhibitors(SSRIs)orselectivenorepinephrinereuptakeinhibitors(SNRIs)providecost savingsoverbrand-namemedications.Anobservational study byDunn et al. showed that a step-therapyedit requiring patients to use a generic antidepressant prior touseofabrand-namemedicationresultedindrugcostsavingsof$1,880,562($0.36permemberpermonth)in2005dollarsinthefirstyearofimplementationoftheprogram,buttheauthorsdidnotlookattheimpactonhealthoutcomesormedicalcosts.

•Despiteevidenceofcostsavings,somecasereportsandbioequiv-alencestudiessuggestadisadvantageinefficacyandtolerabilityofgenericmedicationscomparedwithbrand-nameequivalents.Further,observationalreportshaveprovidedmixedevidenceofsafety andefficacy amongbrandandgenericSSRIs andSNRIs.Thesestudiesareofvaryingquality,addinglackofclaritytothedebate.

What is already known about this subject

Discontinuation Rates and Health Care Costs in Adult Patients Starting Generic Versus Brand SSRI or SNRI

Antidepressants in Commercial Health Plans

Anna Vlahiotis, MA; Scott T. Devine, PhD; Jeff Eichholz, PharmD; and Adam Kautzner, PharmD

ABSTRACT

BACKGROUND: Generic antidepressants offer significant prescription drug cost savings compared with brand-name antidepressants, but critics of managed care interventions promoting generic medication use suggest that some generic antidepressants are not as safe or effective as the brand alternatives.

OBJECTIVE: To assess (a) rates of discontinuation of the initially dispensed medication and (b) disease-specific and total health care costs and phar-macy costs, comparing patients who initiated therapy with brand versus generic selective serotonin reuptake inhibitors (SSRI) or selective norepi-nephrine reuptake inhibitors (SNRI).

METHODS: Antidepressant users aged 18 to 64 years with no pharmacy claims for an SSRI/SNRI in the 180 days prior to the start of SSRI/SNRI therapy (baseline) were identified in the MarketScan database between July 1, 2005, and June 30, 2007, and were followed for 180 days (follow-up). All study patients met the following criteria: (a) continuously eligible from baseline through follow-up; (b) at least 1 medical claim with a primary or secondary diagnosis of major depressive disorder (ICD-9-CM codes 296.2 or 296.3) in either the baseline or follow-up period; and (c) no pharmacy claims for antipsychotic medications in the baseline period. For brand versus generic antidepressant initiators, logistic regression was used to determine the odds of 6-month therapy discontinuation, defined as no medication refills or absence of a refill for the initially dispensed medi-cation within 1.5 times the days supply dispensed, adjusted for important covariates. Costs were measured as total plan allowed charges including member cost share. Adjusted mean (least squares means holding covari-ates at mean values) all-cause medical costs, disease-specific (claims with a ICD-9-CM diagnosis code for major depressive disorder in the primary or secondary diagnosis field) medical costs, all-cause pharmacy costs, and SSRI/SNRI antidepressant costs were compared for brand versus generic initiators using generalized linear regression models, also adjusted for baseline covariates.

RESULTS: Of 16,659 new SSRI/SNRI users, 47.8% (n = 7,955) initiated a brand-name medication and 52.2% (n = 8,704) initiated a generic product. Of the 7,955 who initiated a brand-name antidepressant, 46.8% (n = 3,723) discontinued the initially dispensed drug within 180 days, compared with 44.2% (n = 3,843) of the 8,704 who initiated a generic. The adjusted odds of discontinuation among generic and brand drug users did not significantly differ (odds ratio [OR] = 1.09, 95% CI = 0.98-1.22). Adjusted all-cause 6-month average health care costs in patients initiating therapy on a gener-ic antidepressant were $3,660 (95% CI = $3,538-$3,787) compared with $4,587 (95% CI = $4,422-$4,757) for patients initiating on a brand-name antidepressant. Adjusted average 6-month SSRI/SNRI antidepressant costs were 43.7% lower in patients initiating on a generic drug ($174 vs. $309).

CONCLUSIONS: The likelihood of discontinuation was similar for patients who initiated therapy with brand or generic antidepressants, and short-term health care costs and pharmacy costs were lower in patients starting

RESEARCH

•Comparing patients initiating antidepressant therapy with abrand versus generic medication, there was no significant dif-ference in the likelihoodofdiscontinuationof the initiallydis-pensedmedication during the first 180 days of SSRI or SNRItherapy.HealthcarecostswereloweramongpatientsstartingagenericSSRI/SNRI,evenafteradjustmentformeasurablefactorsthatmayaffectcosts,suchasageandcomorbidity.

•The discontinuation rate within 180 days of initiating anti-depressant drug therapy among patients starting on a genericmedicationwas 44.2%, comparedwith 46.8% among patientsinitiating therapyon a brandmedication (P =0.006).However,the adjusted odds of discontinuation among generic users didnotsignificantlydifferfromthoseofbranddrugusers(OR=1.09,95%CI=0.98-1.22).

What this study adds

a generic SSRI/SNRI. The results suggest that the use of generic antide-pressants as first-line agents in the treatment of major depressive disorder is associated with continuation rates similar to initiation with brand antide-pressants but with lower health care costs.

J Manag Care Pharm. 2011;17(2):123-32

Copyright © 2011, Academy of Managed Care Pharmacy. All rights reserved.

124 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

Discontinuation Rates and Health Care Costs in Adult Patients Starting Generic Versus Brand SSRI or SNRI Antidepressants in Commercial Health Plans

forpharmaceuticalcompanies.”12Still,concernsaboutgenericsubstitutionofantidepressantshavepersisted.13Observationalcomparisons and bioequivalence studies of patients treatedwith brand versus generic second-generation antidepressantshaveproducedmixedfindings.13-15

Aspartofcontainingtherisingcostofhealthcareingen-eral, andprescriptionmedications specifically,managed careorganizations, employer groups, and other plan sponsors areincreasinglyadoptinginterventionprogramsthataredesignedto encourage efficient use of pharmaceuticals, including steptherapy.Step-therapyprogramsworkbyrequiringthatpatientsattempt the use of first-line medications prior to receivingpharmacybenefitcoverageforotherprescriptiondrugsinthesametherapyclass.16Thefirst-linemedicationsareoftenlower-costgenericdrugsthatcanofferthesamehealthbenefitsasthemore expensive brand-name drugs originally prescribed byphysicians.Becausethevalueofthebrand-nameSSRIs/SNRIshasnotbeenunequivocallydemonstrated,genericantidepres-sants may offer cost savings without an increase in adversehealthoutcomes,asisintendedinstep-therapyprograms.

Studies have examined the impact of obligatory genericantidepressantstep-therapyprogramsonpharmacycostsandutilization17 or have estimated the economic impact of costcontrolprogramsamongpatientswithasinglediagnosis,18butnopublishedresearchidentifiedthroughaMedlinesearchhasevaluated multiple outcomes associated with the initial pre-scription of different generic SSRIs or SNRIs comparedwithbrandmedicationsinpatientswithmajordepressivedisorder.Although theoutcomesofautilizationmanagementprogramrequiringfirst-lineuseofagenericmedicationwerenotdirectlymeasured,thepurposeofthisstudywastoanswerquestionsaboutpotentialtreatmentfailurebyassessingdiscontinuationratesandhealthcarecosts,comparingpatientswho initiatedtherapy with a generic versus a brand-name SSRI or SNRI.Thestudywasconductedbyapharmacybenefitsmanagement(PBM)company.

■■  MethodsStudy Cohort and Data SourceA cohort study design was used to compare discontinua-tion rates and health care utilization costs among patientsusing brand versus generic SSRI or SNRI antidepressanttherapy.Data fromtheMarketScanCommercialClaimsandEncounters dataset (Thomson Reuters, New York, NY) fortheperiodof January 1, 2005, toDecember 31, 2007,wereused. These data include commercial (e.g., notMedicare orMedicaid)health insuranceclaims (inpatientandoutpatientmedical, and outpatient pharmacy) as well as enrollmentdatafromlargeemployersandhealthplansacrosstheUnitedStatesthatprovideprivatehealthcarecoverageformorethan45million employees, their spouses, anddependents, all ofwhom are younger than age 65 years. This administrative

Brand-name selective serotonin reuptake inhibitors (SSRIs), such as Lexapro (escitalopram; ForestLaboratories)andPaxilCR(paroxetinecontrolledrelease;

GlaxoSmithKline),andbrand-nameSNRIs,suchasCymbalta(duloxetineHCl;Lilly),areoftenprescribedasfirst-linemedi-cationsbyphysiciansforthetreatmentofsomementalhealthdisorders,suchasmajordepressivedisorder.Decisionstopre-scribebrand-namemedicationsratherthangenerictherapeuticequivalents may be made on the basis of perceived clinicaleffectivenessortolerability1butalsomaybeinfluencedbythemarketing efforts of pharmaceutical companies.2,3 However,there is mixed evidence of treatment continuity and cost-effectivenesswhen comparing brand and generic antidepres-santtherapies.

Multiplemeta-analysesofrandomizedcontrolledtrialshavefoundthattreatmentofadultswithmajordepressivedisorderwithsecond-generationantidepressantsisgenerallyefficaciousandsafe,withmorerecentevidencesuggestinggreaterefficacyforsertralinecomparedwithothersecond-generationdrugs.4-8 A2009multiple-treatmentsmeta-analysisof12new-generationantidepressants,whichincludedbothsingle-sourcebrandandgenericallyavailableagents,concludedthatclinicallymeaning-fuldifferencesinefficacyandtoleranceamongthemedicationsfavoredescitalopramandsertraline;theauthorsnotedthatser-tralinehadthemostfavorablebalancebetweencosts,benefits,andacceptabilityamongpatients.8Nonetheless,acomparativeeffectiveness review sponsored by the Agency forHealthcareResearch and Quality cautioned that such studies were notintendedtotestvariationinindividualresponsestoindividualdrugs,4andcasereportsofsymptomrelapsefollowinggenericsubstitution have emerged in the literature.9,10 Investigationsby the U.S. Food and Drug Administration into reports ofthistype,comparingbrandversusgenericbuproprion, foundminordifferencesinpharmacokineticsthatdidnotfalloutsideofacceptedboundariesforbioequivalence.11ACarlat Psychiatry Report on the issue (2009) noted that the preponderance ofevidence suggesting that generic drug substitutions result infailure was found in “single cases or very small case series,virtuallyallwrittenbyauthorswhoarealsopaidconsultants

•Totaladjustedaverage6-monthhealthcarecostsinpatientsini-tiatingtherapyonagenericdrugwere$3,660(95%CI=$3,538-$3,787), 20% less than the average $4,587 (95% CI=$4,422-$4,757)forpatientsinitiatingonabrand-namedrug(P <0.001).Average SSRI/SNRI antidepressant costs in patients initiatingtherapyonagenericdrugwere$174(95%CI=$169-$180),44%lessthantheaverage$309(95%CI=$300-$319)forpatientsini-tiatingonabrand-namemedication(P <0.001).

What this study adds (continued)

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 125

claimsdatabaseincludesavarietyoffee-for-service,preferredproviderorganization,andcapitatedhealthplans.

Inclusioncriteriarequiredthatthepatient:(a)hadaphar-macy claim for an SSRI or SNRI during an identificationperiod from July 1, 2005, through June 30, 2007, identifiedby mapping National Drug Code (NDC) numbers providedin theMarketScandataset to generic product identifier (GPI,Medi-Span,Inc.,Indianapolis,IN)codesbeginningwith5816or5818;(b)wasanewuserofanSSRIorSNRI,definedastheabsenceofapharmacyclaimforanSSRIorSNRI inthe180dayspriortothedateofthefirstobservedSSRI/SNRIclaimintheidentificationperiod(indexdate);(c)didnothavepharmacyclaimsforantipsychoticmedications(GPIcodesbeginningwith59)inthe180dayspriortotheindexSSRI/SNRIclaim;(d)wasaged18yearsorolderasoftheindexpharmacyclaimdate;(e)wascontinuouslyenrolledinaprescriptiondrugbenefitplanfor180dayspriortotheindexpharmacyclaimand180daysaftertheindexpharmacyclaim;(f)didnothaveclaimswithanegativedayssupply,duplicateclaims,orreversedclaimswithnegativecostvaluesinthe180dayspriortooraftertheindexpharmacyclaim;and(g)hadat least1claimwithaprimaryor secondary diagnosis of single-episode or recurrent majordepressivedisorderasindicatedbyInternational Classification of Diseases, Ninth Revision, Clinical Modification(ICD-9-CM)codes296.2 and 296.3 in the 180 days prior to or after the indexpharmacyclaim,asthesearepatientswhomayrequirelongertermtherapy.19BrandversusgenericstatusforeachSSRI/SNRIpharmacyclaimwasdeterminedbymappingtheNDCnumberin theMarketScan file to an internal PBM file that indicatedbrandorgenericstatus.

Exposure and CovariatesThe exposure of interestwas the occurrence of a new phar-macyclaimforagenericSSRIorSNRI.Pharmacyandmedicalclaims were evaluated for the 6-month time periods before(baseline)andafter(follow-up)theindexdate.ExposurecouldoccuratanypointonorafterJuly1,2005.

Patient age category (in years: 18 to 25, 26 to 40, 41 to55,56to64)asofthedateoftheindexpharmacyclaimwasincludedasapotentialcovariatebecauseagemaybeassociatedwithantidepressanttreatmentoutcomes.20,21Gendermayalsoplace patients at increased risk for developing certain typesofpsychiatricdisorders suchasdepression,22 anddifferencesin adverse effects and time-to-response to different therapieshave been reported betweenmale and female antidepressantusers.23,24Basedupontheapparentdifferences inresponsetotherapy, genderwas also included as a potential confounder,withfemalegenderasthereferencegroup.

The risk for therapy discontinuation may also increasewith some comorbidities.25 A Charlson Comorbidity Index(CCI)scorebasedonthe180dayspriortoandaftertheindexpharmacy claim (baseline and follow-up) was calculated for

allmembersofthesample.TheCCIisavalidatedmeasureoftheburdenofchronicillnessesthathasbeenadaptedforusewith ICD-9-CM codes found in administrative claims dataovera12-monthperiod.26TheCCIwasincludedasapotentialconfounder andmeasured as a categorical variable (scores of0,1to2,3to5,and6ormore).Inaddition,themedicalcostsincurredbypatientswithsomementalhealthillnessescanbean indication of symptom severity.27 The summed all-causemedical costs (totalplan allowedamounts foroutpatient andinpatient hospital and for physician services, not includingpharmacy costs) incurred by patients in the baseline periodwere also measured, and analyzed as a categorical variablebased on quartiles ($1-$178, $179-$657, $658-$2,402, and$2,403ormore).

Depressionhasbeensuccessfullymanaged in theprimarycaresetting,28buttherapydiscontinuationmaybemorelikelyamongpatientswhoarereceivingspecializedpsychiatriccare,such as psychotherapy.29 Thus, receipt of inpatient or outpa-tientpsychiatricmedicalcare,indicatedbyCurrentProceduralTerminology(CPT)codes90801-90829(psychiatricinterviewsand psychotherapy) and 90862-90899 (other psychiatricservices or procedures, such as pharmacologicmanagement,electroconvulsive therapy, andhypnotherapy) in thebaselineperiod,was examined as apotential confounder.Recent evi-dencealsosuggeststhatsomeSSRIsmaybeassociatedwithanincreasedriskofbleeding,30whichcouldtheoreticallysuggestthat patients beginning anticoagulant therapy may be morelikely to discontinue therapy. A binary variable indicating apharmacy claim for an anticoagulant (GPI codes beginning8310 [heparins], 8320 [coumarin anticoagulants, includingwarfarin],or8333[thrombininhibitors])inthebaselineperiodwasalsomeasured.

Finally,primaryorsecondarydiagnosisofacomorbidmooddisorderduringeitherthebaselineorfollow-upperiod,includ-ing anxiety state (ICD-9-CM codes 300.0x), bipolar disorder(ICD-9-CM codes 296.0, 296.4, 296.5, 296.6, 296.7, 296.8)or obsessive compulsive disorder (ICD-9-CM codes 300.3,301.4)was also assessed. Each comorbidmooddisorderwasmeasuredasaseparatebinaryvariable.BothtimeperiodswereincludedbecausediagnosispriortotheinitiationofanSSRIorSNRImayhaveaffectedwhichmedicationwasprescribed,anddiagnosisaftertheinitiationoftherapymayhaveaffectedthedecisiontodiscontinueoraugmenttherapy.

Outcomes Two primary outcomesweremeasured during the follow-upperiod: discontinuation of the initial GPI-10 coded medica-tion, including its brand or generic status, and health carecosts.Duringthe180daysafterinitiationoftherapy,patientscould discontinue the initially dispensed therapy (i.e., eitherswitch to a different antidepressant drug or terminate anti-depressant therapy altogether) or continue on the index

Discontinuation Rates and Health Care Costs in Adult Patients Starting Generic Versus Brand SSRI or SNRI Antidepressants in Commercial Health Plans

126 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

medication.A6-monthfollow-uptimeperiodwaschosenforthisstudy,whichisconsistentwiththeNationalCommitteeforQualityAssurancerecommendationsforeffectivemanagementof major depressive disorder treated with an antidepressantmedication.31 A discontinuation date was considered as thedateofthelastpharmacyclaimfortheinitialmedicationplusthe days supply dispensed for the last pharmacy claim, anda study completion datewas defined as the index pharmacyclaim date plus 180 days. Patients were considered to havediscontinued therapy if the index pharmacy claim was notrefilledwithin 1.5 times the days supply dispensed (e.g., 45days on a 30 day supply, 135days on a 90day supply).32 AswitchfromtheinitialmedicationtoanothermedicationwithadifferentGPI-10wasconsideredadiscontinuation,aswasaswitchbetweenthebrandandgenericformulationsofthesamemedication.Medicationswitcheswerecountedasdiscontinua-tionstomeasurechangetotheinitiallyprescribedmedication,addressingtheconcernthattheinitialprescriptionofagenericmedication is associatedwith greater proportions of patientswhodiscontinueorotherwisefailtherapy.Onceapatientwasdefinedasdiscontinued,re-initiationoftheindexmedicationwasnotconsidered.

Totalhealthcarecostsincludedtheplanallowedcharges,including member cost share, for all inpatient and outpa-tientmedicalservicesclaimsandforalloutpatientpharmacyclaims in the 180-day follow-up period. Costs were furthercategorized as disease-specific (charges incurred for medicalclaimswith anassociatedprimaryor secondarydiagnosisofmajordepressivedisorderasindicatedbyICD-9-CMcodesorpharmacyclaimsforanySSRIorSNRI)andall-cause(chargesincurredforanyclaims—notdependentonICD-9-CMdiagno-siscodesortherapeuticclasses).

Older antidepressants, such as tricyclic antidepressants,arestillusedaloneandinconjunctionwithnewerSSRIsandSNRIs because somepatients responddifferently to differentmechanismsofaction.33Asresponseornonresponsetootherantidepressants may affect continuation of SSRIs or SNRIs,andasfailuretorespondtoSSRIsorSNRIsmaybereflectedbynewtreatmentwithanolderantidepressant,augmentationof therapy with a non-SSRI/SNRI antidepressant (GPI codesbeginning 5812 [modified cyclics, including nefazadone andtrazadone],5820[tricyclicantidepressants],and5830[miscel-laneousantidepressants,includingbuproprion)inthe180-dayperiodaftertheindexpharmacyclaimwasalsoconsideredasasecondaryoutcome.

Statistical AnalysisEach variable that could potentially affect the relationshipbetween the exposure and discontinuation was entered intoa logistic regression model. Variables with a P value of lessthan0.1inbivariatescreeningwereenteredintofullmodels,andvariableswithaPvalueoflessthan0.05wereretainedin

the finalmodel.Thevariablesused in specifying the logisticregressionmodel included patient age category, gender, CCIcategory,baselinemedicalcostcategory,receiptofpsychiatricmedical care in the baseline period, indication of anticoagu-lant therapy in thebaselineperiod, indicationof a comorbidmood disorder diagnosis in either the baseline or follow-upperiod,andan indicatorofwhichSSRIorSNRIwas initiallyprescribed.

ToassesstheimpactofaugmentationofSSRI/SNRItherapywith a non-SSRI/SNRI antidepressant after initiation of ther-apy,astratifiedanalysiscomparedtherelationshipbetweentheinitialexposuretoagenericorbrandSSRI/SNRIandtherapydiscontinuation in a subsample of patients who filled a pre-scriptionforanon-SSRI/SNRIantidepressantinthefollow-upperiodversusasubsampleofpatientswhodidnot.Theoddsratiosofthefinallogisticregressionmodelswerecomparedinthesubsamples.

Generalizedlinearmodels(GLM)wereconstructedtomea-suretherelationshipbetweenexposurestatusandhealthcarecosts.Modelswerespecifiedwithagammadistributionandloglink function to satisfy the assumptions of homoskedasticityandanormaldistribution.34Thesummedall-causetotalhealthcare costs, disease-specific total health care costs, all-causepharmacycosts,andSSRIorSNRIantidepressantcostsmea-suredinthefollow-upperiodwereeachevaluated.Theassocia-tionsofhealthcarecostsandpharmacycostswithgenericdrugstatus were assessed, controlling for covariate effects with a Pvalueoflessthan0.1inbivariatescreening.VariableswithaPvalueoflessthan0.05wereretainedinthefinalmodels.

Leastsquaresmeancostsholdingcovariatesat theirmeanvalueswerecalculatedforpatients initiatingbrand-namever-sus genericmedications. For each least squaresmean, t-type95% confidence intervals (CIs) were constructed using theSAS (SAS Institute Inc.,Carey,NC) LSMEANSoption in theGENMODprocedure.All variableswere checked formissingdata. Statistical analyses were performed using SAS version9.2.Theapriorialphavaluewas0.05.

■■  ResultsThere were 2,545,696 individual patients identified with apharmacy claim for an SSRI or SNRI between July 1, 2005,andJune30,2007(Figure1).Fromthissample,951,605wereexcluded because they were not new to SSRI/SNRI therapy;148,600wereexcludedbecausetheyhadapharmacyclaimforanantipsychoticinthebaselineperiod;74,237wereexcludedbecause theywere younger than 18 years old; 742,271wereexcluded for lack of continuous enrollment in the 180 daysprior to and after the index pharmacy claim; 35,524 wereexcludedbecauseofnegativedayssupplyoftheindexmedica-tion, single claims being countedmultiple times, or negativecost values; and 576,800 were excluded because they didnot have a primary diagnosis of major depressive disorder,

Discontinuation Rates and Health Care Costs in Adult Patients Starting Generic Versus Brand SSRI or SNRI Antidepressants in Commercial Health Plans

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 127

leaving16,659patientsinthefinalsample.Thedeclineinsam-plesizeresultingfromexcludingpatientswithoutadepressiondiagnosis has occurred in other observational studies usingadministrativeclaimsdatabases.15Therewerenomissingdataintheanalyticsample.

Therapy DiscontinuationThe demographic and clinical characteristics of the patientswere stratified by whether or not they discontinued therapy(Table1).Within180daysofinitiatingtherapyonabrandorgeneric SSRI/SNRI, 7,566 patients (45.4%) discontinued theinitiallydispensedtherapy.Ofthese,2,916(38.5%ofthosedis-continuing,17.5%ofthesampleoverall)didnotrefilltheinitialprescriptionand2,423 (32.0%of thosediscontinuing,14.5%of the sample overall) switched to a different antidepressantmedication(datanotshown).Amongpatientswhodiscontin-uedtherapy,themean(median)timeuntildiscontinuationwas50.8 (30)days (range1 to158days);68.8%ofpatientswhodiscontinueddidsointhefirst60daysoftherapy;and84.0%discontinuedwithin the first 90days after initiating therapy(datanotshown).

Of 8,704 patients who started treatment with a genericSSRI/SNRI,3,843(44.2%)discontinuedtheinitiallydispensedtherapyduringfollow-up.Ofthose,1,509(39.3%ofthosedis-continuing,17.3%ofgenericusersoverall)didnotrefill theirinitialprescription,and1,199(31.2%of thosediscontinuing,13.8%ofgenericusersoverall)switchedtoadifferentantide-pressant medication (data not shown). Among patients whodiscontinued therapy after initiating therapy with a genericSSRIor SNRI, themean (median) timeuntil discontinuationwas 51.2 (30) days (range 1 to 157 days); 68.0% of patientswhodiscontinueddid so in the first 60days after the indexpharmacyclaim,and84.4%discontinuedtherapywithin thefirst90days(datanotshown).

Of 7,955 patients who initiated treatment with a brandSSRI/SNRI,3,723(46.8%)discontinuedtheinitiallydispensedtherapy during follow-up. Of those, 1,407 (37.8% of thosediscontinuing,17.7%ofbrandusersoverall)didnotrefilltheinitial prescription and1,224 (32.9%of thosediscontinuing,15.4%of brandusers overall) switched to a different antide-pressant medication (data not shown). Among patients whodiscontinued therapy after initiating therapywith a brandedmedication,themean(median)timeuntildiscontinuationwas50.3(30)days(range1to158days);69.9%ofpatientswhodis-continueddidsointhefirst60daysaftertheindexpharmacyclaim;and83.7%discontinuedtherapywithinthefirst90days(datanotshown).

Patients who discontinued therapy were slightly youngeronaveragethanpatientswhodidnotdiscontinue(meanages39.3yearsvs.41.9years,respectively,P <0.001)andweremorelikely tohave comorbidmooddisorderdiagnosesduring thestudyperiod.UseofescitalopramorparoxetineHClwasmore

commonamongpatientswhodiscontinuedtherapycomparedwiththosewhodidnotdiscontinue(23.2%vs.20.6%,respec-tively,and11.2%vs.7.9%,respectively,bothP <0.001).

PatientswhoinitiatedtherapyonagenericSSRI/SNRIhadsimilaroddsofdiscontinuingtheinitiallydispenseddruginthefirst180daysoftherapycomparedwithpatientswhoinitiated

Discontinuation Rates and Health Care Costs in Adult Patients Starting Generic Versus Brand SSRI or SNRI Antidepressants in Commercial Health Plans

FIGURE 1 Patient Selection Flowchart

SNRI = selective norepinephrine reuptake inhibitor; SSRI = selective serotonin reuptake inhibitor.

Patients continuously eligible during 180 days prior to and after index SSRI/SNRI claim

N = 628,983

Patients aged 18 years or olderN = 1,371,254

Patients with no pharmacy claims for antipsychotics during 180 days prior to index SSRI/SNRI claim

N = 1,445,491

Patients with no SSRI/SNRI claims during 180 days prior to index SSRI/SNRI claim

N = 1,594,091

Patients with at least 1 claim for SSRI/SNRI from July 1, 2005, through June 30, 2007

N = 2,545,696

MarketScan database patientsN = 45,973,928

Patients with no negative days supply, duplicate claims, or negative cost values in 180 days prior to and/or

after index SSRI/SNRI claimN = 593,459

Patients with at least 1 medical claim with a primary or secondary diagnosis of major depressive disorder in 180 days

prior to and/or after index SSRI/SNRI claimN = 16,659

Patients with brand index claim

n = 7,955

Patients with generic index claim

n = 8,704

128 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

analysesstratifiedbyaugmentation(i.e.,in2subsamplegroupsofpatientswhoaugmentedtherapywithanon-SSRI/SNRIanti-depressantduringfollow-upvs.thosewhodidnot),patientswhoinitiatedtherapyonagenericSSRI/SNRIwerenomorelikelyto discontinue therapy than patients who started a branded

therapyonabrandedalternative(oddsratio[OR]=1.09,95%CI=0.98-1.22;Table2).Variables inthefinalmodel includedgenericstatusoftheindexpharmacyclaim;age;gender;indica-torsforpsychiatricmedicaltreatment,comorbidanxietydisor-der,andbipolardisorder;andspecificSSRI/SNRIdrug.Inthe

Discontinuation Rates and Health Care Costs in Adult Patients Starting Generic Versus Brand SSRI or SNRI Antidepressants in Commercial Health Plans

TABLE 1 Profiles of 16,659 New SSRI/SNRI Users in the First 180 Days After Initiating Antidepressant Therapy in the 2005-2007 MarketScan Database

Continued Initially Dispensed Therapy

n (%)

Discontinued Initially Dispensed Therapy

n (%) P ValueaUnadjusted Odds Ratio

(95% CI)

9,093 (54.6) 7,566 (45.4)ExposureBrandindexpharmacyclaim 4,232 (46.5) 3,723 (49.2) ReferentCategoryGenericindexpharmacyclaim 4,861 (53.5) 3,843 (50.8) 0.006 0.90 (0.85-0.96)

GenderFemale 6,005 (66.0) 4,880 (64.5) ReferentCategoryMale 3,088 (34.0) 2,686 (35.5) 0.038 1.07 (1.00-1.14)

Age in years18to25 1,023 (11.3) 1,248 (16.5) ReferentCategory26to40 2,947 (32.4) 2,773 (36.7) 0.77 (0.70-0.85)41to55 3,822 (42.0) 2,770 (36.6) 0.59 (0.54-0.65)56to64 1,301 (14.3) 775 (10.2) < 0.001 0.49 (0.43-0.55)

Charlson Comorbidity Indexb

0 7,233 (79.5) 6,011 (79.4) ReferentCategory1to2 1,601 (17.6) 1,334 (17.6) 1.00 (0.93-1.08)3to5 179 (2.0) 171 (2.3) 1.15 (0.93-1.42)6ormore 80 (0.9) 50 (0.7) 0.239 0.75 (0.53-1.07)

Medical costs in U.S. dollarsc

$1-$178 2,239 (24.6) 1,928 (25.5) ReferentCategory$179-$657 2,316 (25.5) 1,849 (24.4) 0.93 (0.85-1.01)$658-$2,402 2,336 (25.7) 1,826 (24.1) 0.91 (0.83-0.99)$2,403ormore 2,202 (24.2) 1,963 (25.9) 0.008 1.04 (0.95-1.13)

Comorbid mood disorderAnxietydisorderb 1,303 (14.3) 1,230 (16.3) 0.001 1.09 (1.00-1.19)Bipolardisorderb 69 (0.8) 112 (1.5) < 0.001 1.77 (1.29-2.42)Obsessive-compulsivedisorderb 57 (0.6) 50 (0.7) 0.785 0.85 (0.58-1.25)

Medical carePsychiatricmedicalcarec 6,303 (69.3) 5,383 (71.1) 0.010 1.04 (0.97-1.10)

Anticoagulant usec 89 (1.0) 65 (0.9) 0.422 0.99 (0.72-1.37)Specific SSRI/SNRI drugFluoxetined 1,964 (21.6) 1,317 (17.4) < 0.001 ReferentCategoryCitalopramd 1,459 (16.0) 1,104 (14.6) 0.010 0.89 (0.82-0.97)Escitaloprame 1,873 (20.6) 1,756 (23.2) < 0.001 1.17 (1.08-1.25)Fluvoxamined 19 (0.2) 20 (0.3) 0.461 1.27 (0.68-2.38)ParoxetineHCld 721 (7.9) 844 (11.2) < 0.001 1.46 (1.31-1.62)Paroxetinemesylatee 8 (0.1) 9 (0.1) 0.533 1.36 (0.52-3.51)Sertralined 1,829 (20.1) 1,493 (19.7) 0.594 0.98 (0.91-1.05)Duloxetinee 541 (5.9) 473 (6.3) 0.417 1.05 (0.93-1.20)Venlafaxined 679 (7.5) 550 (7.3) 0.627 0.97 (0.86-1.09)

aFor categorical variables, P values were derived from Pearson chi-square tests. All tests compared patients who did versus those who did not discontinue therapy.bMeasured during a 12-month period including both the 180-day baseline and follow-up periods. cMeasured in the 180-day baseline period.dAvailable in branded and generic formulations at given points between January 1, 2005, and December 31, 2007.eAvailable only in a branded formulation between January 1, 2005, and December 31, 2007.CI = confidence interval; HCl = hydrochloride; SNRI = selective norepinephrine reuptake inhibitor; SSRI = selective serotonin reuptake inhibitor.

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 129

CI=$771-$836vs.$1,125,95%CI=$1,077-$1,175).All-causepharmacy costs and SSRI or SNRI antidepressant pharmacycostseachweresignificantly loweramongpatientswho initi-atedtherapywithagenericmedicationcomparedwithpatientswho initiated therapy with a brand medication ($761, 95%CI=$738-$785vs.$965,95%CI=$934-$998,respectively,forall-causepharmacycosts;$174,95%CI=$169-$180vs.$309,95%CI=$300-$319,respectively,forSSRIorSNRIantidepres-santpharmacycosts).

■■  DiscussionTheprincipalobjectiveofthisstudywastodetermineiftherewere differences in discontinuation rates and health carecosts between patients who initiated antidepressant therapyonagenericSSRIorSNRIcomparedwithpatientswhoiniti-atedtherapyonabrand-nameSSRIorSNRImedication.Theadjustedcomparisonsuggestedtherewasnosignificantdiffer-enceinthelikelihoodofdiscontinuationduringthefirst180daysoftherapy,andtheanalysisofhealthcarecostsindicated

medication(Table2).Thisfindingsuggeststhatwhileimportant, augmentationwithanonSSRI/SNRIantidepressantdoesnotaffecttherelationshipbetweenthegenericstatusoftheindexpharmacyclaimanddiscontinuationoftherapy.

Health Care CostsAfteradjustment forother factorsassociatedwithhealthcarecostsinpatientswithmajordepressivedisorder,patientswhoinitiatedtherapyonagenericdrughadlowerhealthcarecostsduringthefirst180daysaftertheindexSSRI/SNRIprescrip-tioncomparedwithbranddrugusers(Table3).Theadjustedaveragetotalhealthcarecosts(leastsquaresmeancosts)inthefirst180daysaftertheindexpharmacyclaimamongpatientswho initiated therapy on a generic SSRI/SNRI were $3,660(95% CI=$3,538-$3,787) versus $4,587 (95% CI=$4,422-$4,757)amongpatientswhoinitiatedtherapyonabrandmedi-cation. Least squaresmean disease-specific health care costswerealsolowerinpatientswhoinitiatedtherapyonagenericmedication compared to a brand medication ($803, 95%

Discontinuation Rates and Health Care Costs in Adult Patients Starting Generic Versus Brand SSRI or SNRI Antidepressants in Commercial Health Plans

TABLE 2 Logistic Regression Analysis of Discontinuing the Initially Dispensed Antidepressant Therapy in 16,659 New SSRI/SNRI Users in the First 180 Days After Initiating Therapy in 2005-2007a

Adjusted Odds Ratio (95% CI)

All Patients (C = 0.582)

Adjusted Odds Ratioa (95% CI)

New Non-SSRI/SNRI (C = 0.638)

Adjusted Odds Ratioa (95% CI)

No New Non-SSRI/SNRI (C = 0.579)

Brandindexpharmacyclaim Referentcategory Referentcategory ReferentcategoryGenericindexpharmacyclaim 1.09 (0.98-1.22) 1.45 (0.89-2.34) 1.08 (0.96-1.21)Male 1.08 (1.01-1.15) 1.16 (0.89-1.51) 1.07 (1.00-1.15)Aged18to25years Referentcategory Referentcategory ReferentcategoryAged26to40years 0.76 (0.69-0.84) 1.00 (0.69-1.45) 0.75 (0.67-0.83)Aged41to55years 0.58 (0.53-0.64) 0.63 (0.44-0.90) 0.58 (0.52-0.64)Aged56to64years 0.47 (0.42-0.53) 0.44 (0.26-0.75) 0.47 (0.42-0.54)Anxietydisorderb 1.10 (1.01-1.20) 1.22 (0.88-1.69) 1.09 (0.99-1.19)Bipolardisorderb 1.80 (1.33-2.45) 3.33 (0.94-11.73) 1.67 (1.22-2.30)Medicalcosts$1-$178b Referentcategory Referentcategory ReferentcategoryMedicalcosts$179-$657b 0.94 (0.87-1.03) 0.82 (0.57-1.17) 0.95 (0.87-1.04)Medicalcosts$658-$2,402b 0.95 (0.87-1.04) 0.70 (0.49-1.00) 0.97 (0.88-1.06)Medicalcosts$2,403ormoreb 1.09 (0.99-1.19) 0.83 (0.58-1.20) 1.10 (1.01-1.21)Psychiatricmedicalcareb 1.06 (0.99-1.14) 0.92 (0.68-1.25) 1.06 (0.99-1.14)Fluoxetine Referentcategory Referentcategory ReferentcategoryCitalopram 1.13 (1.01-1.25) 1.48 (1.00-2.20) 1.10 (0.99-1.23)Escitalopram 1.50 (1.29-1.74) 3.57 (1.93-6.61) 1.42 (1.22-1.66)Fluvoxamine 1.45 (0.77-2.75) NA 1.31 (0.68-2.53)ParoxetineHCl 1.79 (1.57-2.03) 2.67 (1.63-4.39) 1.74 (1.53-1.98)Paroxetinemesylate 1.82 (0.69-4.79) NA 1.40 (0.50-3.92)Sertraline 1.24 (1.11-1.39) 2.52 (1.59-4.02) 1.19 (1.06-1.33)Duloxetine 1.48 (1.24-1.78) 2.36 (1.11-5.04) 1.44 (1.19-1.73)Venlafaxine 1.32 (1.11-1.57) 2.70 (1.34-5.46) 1.26 (1.05-1.50)aMarketScan database; results stratified by whether patients initiated therapy on a non-SSRI/SNRI during the 180-day follow-up period.bComorbid anxiety disorder or bipolar disorder diagnosis measured in the 180-day baseline and follow-up periods; medical costs measured in the 180-day baseline period; psychiatric medical care measured in the 180-day baseline period.C = c-statistic (area under receiver operating characteristics curve); CI = confidence interval; HCl = hydrochloride; NA = not applicable; SNRI = selective norepinephrine reuptake inhibitor; SSRI = selective serotonin reuptake inhibitor.

130 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

thatcostswereloweramongpatientsstartingagenericSSRI/SNRIcomparedwithpatientsstartingabranddrug,evenafteradjustmentforotherfactorsthatmayaffectthosecosts.

Recent proposed legislation in at least 1 state (Missouri)has called for a limitation on pharmacy utilization manage-ment programs—such as step therapy—which promote theuseofgenericmedicationsasfirst-linetherapy.35Althoughtheexplicitoccurrenceofastep-therapyeditatthepointofservicewas not measured in the present study, most of the samplepatients could have been prescribed generic antidepressants,eventhoughsomepatientswereprescribedbrand-namedrugs,asalmostallof theSSRIshadgenericalternativesduringthestudyperiod.36Onereportestimatedthatin2005,80%ofanti-depressantpharmacyclaimscouldhavebeenfilledwithgenericdrugs;however, thegenericdispensingratio inthatyearwasonly50%.37Thepresent studyprovidedanapproximationoftheconditionsinwhichstep-therapyprogramsarecommonlyappliedtopromoteincreaseduseofgenericmedications.Thefindings provide evidence that first-line use of generic SSRIsorSNRIsinthetreatmentofmajordepressivedisorderisnotassociated with a greater likelihood of discontinuation andmaycontributetolowertotalhealthcarecosts,mostlikelybyloweringthecostofdrugtherapy.Theinitiationofantidepres-sant therapy with a generic SSRI or SNRI could reduce thepharmacycostsforhealthcarepayersbyalmost50%insomecases,withoutleadingtotreatmentfailureorincreasedmedi-calcostsintheshortterm.Thestudyfindings,therefore,havespecific implications for costmanagement strategies like steptherapy,andareimportantforhealthcarepayerswhohaveaninvestmentinpatienthealth.

LimitationsThisstudyhasseveral limitations.First,becauseof itsobser-vationaldesign,thestudydemonstratesassociationsbutdoesnot establish causality. As in any observational research, thestudy analyses could controlonly formeasured confounders,not for all factors potentially affecting the study outcomes.In particular, the study may have been confounded by thefailure to adequately control fordisease severity. Informationaboutprevioustreatmentordurationofdiseaseoutsideofthe180 days prior to treatment initiation is not included in theMarketScan database, and no tests of mental health diseaseseverityarecapturedinthedatabase.Medicalserviceutiliza-tionandthelikelihoodofdiscontinuingtherapymayberelatedtodisease severityor thepresenceof treatment-resistantdis-ease.Adjustmentsinregressionmodelsforfactorsthatmayberelatedtodiseaseseverity,suchaswhetherthepatientreceivedpsychiatricmedicalcare,andstratificationbyotherfactorsthatmayberelatedtodiseaseseverity,suchastherapeuticaugmen-tationwithothernon-SSRI/SNRIantidepressants,wereusedtomitigatethepotentialimpactofdiseaseseverity.

Second, drug-specific side-effect profiles were anotherpotential confounder. Side effects are potentially influentialin a patient’s decision to discontinue therapy, and may dif-ferentiallydisadvantagesomedrugs.Adjustmentinregressionmodels for the specific SSRI or SNRImedication potentiallycontrolledfortheimpactofdifferentialsideeffectsontheriskofdiscontinuationandhealthcarecosts.

Third, the study resultsmaybe limited ingeneralizabilitybecauseMarketScandataarerestrictedtopopulationsofcom-mercially insured beneficiaries less than 65 years old. It hasbeenpreviouslyshownthatdepressionandanxietydisorders

Discontinuation Rates and Health Care Costs in Adult Patients Starting Generic Versus Brand SSRI or SNRI Antidepressants in Commercial Health Plans

TABLE 3 Adjusted Least Squares Mean All-Cause and Disease-Specific Health Care Costs in 16,659 New SSRI/SNRI Users in the First 180 Days After Initiating Antidepressant Therapy in the 2005-2007 MarketScan Database

Adjusted Least Squares Mean

All-Cause Health Care Costsa

(95% CI)

Adjusted Least Squares Mean Disease-

Specific Health Care Costsb

(95% CI)

Adjusted Least Squares Mean

All-Cause Pharmacy Costsc

(95% CI)

Adjusted Least Squares Mean SSRI/SNRI Antidepressant

Pharmacy Costsd (95% CI)

Brandindexpharmacyclaim $4,587 ($4,422-$4,757) $1,125 ($1,077-$1,175) $965 ($934-$998) $309 ($300-$319)Genericindexpharmacyclaim $3,660 ($3,538-$3,787) $803 ($771-$836) $761 ($738-$785) $174 ($169-$180)aAdjusted for age; Charlson Comorbidity Index score measured in the 180-day baseline and follow-up periods; medical costs incurred in the 180-day baseline period; comorbid anxiety disorder or bipolar disorder in the 180-day baseline and/or follow-up periods; psychiatric medical care in the 180-day baseline period; and specific drug.bAdjusted for age; gender; Charlson Comorbidity Index score measured in the 180-day baseline and follow-up periods; medical costs incurred in the 180-day baseline period; comorbid anxiety disorder, obsessive-compulsive disorder, or bipolar disorder diagnosis in the 180-day baseline and/or follow-up periods; psychiatric medical care in the 180-day baseline period; and specific drug.cAdjusted for age; Charlson Comorbidity Index score measured in the 180-day baseline and follow-up periods; medical costs incurred in the 180-day baseline period; comorbid anxiety disorder or bipolar disorder diagnosis in the 180-day baseline and/or follow-up periods; psychiatric medical care in the 180-day baseline period; pre-scription claim for an anticoagulant in the 180-day baseline period; and specific drugdAdjusted for age; gender; medical costs incurred in the 180-day baseline period; comorbid anxiety disorder, bipolar disorder or obsessive-compulsive disorder diagnosis in the 180-day baseline and follow-up periods; psychiatric medical care in the 180-day baseline period; and specific drug.CI = confidence interval; SNRI = selective norepinephrine reuptake inhibitor; SSRI = selective serotonin reuptake inhibitor.

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 131

Suchabenefitdesignmayhaveaffectedtheuseofpsychiatricmedicalcare,andinformationaboutpsychiatricmedicalcarein health plans with a behavioral health carve-out may beincompleteifclaimsforsuchcarewerebilledthroughanotherbenefit.Thisinformationwasnotavailableintheadministra-tiveclaimsdatausedinthepresentstudy.

■■  ConclusionsInitiationoftherapyonagenericSSRIorSNRIdoesnotappeartobeassociatedwithagreaterlikelihoodoftherapydiscontin-uationandmaybeassociatedwithdecreasedtotalhealthcarecostsinthefirst180daysoftherapy.Thestudy’sresultsdonotsupportthecontentionsometimesmadebycriticsofpharmacyutilizationmanagement tools thatgenericantidepressantsarelesseffectiveorsafethanbranddrugs.

commonlyaffectelderlypatients,38andthatprescriptiondrugsmay differentially affect these patients.21 In addition, patientlocationoraccess toaproviderofmentalhealthcarearenotavailable in the MarketScan data. The population may berestricted to thosewhohaveadequategeographicaccess toaproviderofmentalhealthcare,andmaynotbegeneralizabletopatientsinallgeographiclocations.

Fourth,itisalsopossiblethatsomepatientstakingagenericdrug who appeared to have discontinued therapy actuallybeganusingadeeplydiscountedorlow-costgenericprogramoffered at local grocery stores and supermarkets during thestudy period.39 This problem potentially overestimated thediscontinuationrateamongpatientswhoinitiatedtherapyonagenericmedication.Inaddition,patientswhowerereceivingsamplesofbrandmedicationsorpatientswhohadbeentreatedforpreviousepisodesofmajordepressivedisorderpriortotheinitialpharmacyclaimorbaselineperiodmaynothavebeennew users at the time of their initial pharmacy claim. Newusersweredefined aspatientswithout apharmacy claim foranSSRI/SNRI inthe180daysprior to the indexmedication.Boththeuseofsamplesofbrandmedicationsandrecurrencesofpreviouslytreateddepressionhavethepotentialto(a)under-estimatetruediscontinuationamongpatientswhohadalreadyexperienced intoleranceor lackof effect and (b)preventdis-continuationamongpatientswhohadalreadyfoundthemedi-cationtobetolerableoreffective.However,informationaboutbrand samplesorprevioushistoryofSSRI/SNRIuseprior tothestudyperiodwasnotavailableintheadministrativeclaimsdatausedforthisstudy.

Fifth,ourdefinitionoftherapydiscontinuationdidnotdif-ferentiatepatientswhodiscontinuedmedicationfrompatientswhoswitchedtherapies.Asthestudyobjectivewastoevaluatetheimpactofthegenericorbrandstatusoftheinitialpharmacyclaimonly,whichwouldmostcloselysimulatethepracticeofa cost control program such as step therapy, any switchesbetweenSSRIsorSNRIswereintentionallylabeledasdiscon-tinuations.Patientswhoswitchedtherapyratherthanrefillingprescriptions for the indexmedicationmayhaveappeared tobe discontinuing therapy entirely. This definition may haveincludedpatientswhoswitchedbetweenthebrandandgenericformulationsofthesamemedication.Thisdecisionrepresentsalimitationespeciallyforpatientswhoinitiatedtherapyonthebranded formulation of sertraline because a generic versionwasmadeavailable in Juneof2006andmayhavepromptedmany patients to switch from the brand to generic formula-tionofsertraline.Althoughpatientsmayhaveswitchedfromabrandtoageneric,whichwouldpresumablybebeneficialandcost saving, a patientmayhave been just as likely to switchfromagenericmedicationtoabrand-nameformulation,whichwouldminimize any impact of differentiallymisclassifying aswitchasadiscontinuationbetweentheexposuregroups.

Finally, we were not able to determine whether a patientwasenrolledinahealthplanwithbehavioralhealthcarve-out.

Discontinuation Rates and Health Care Costs in Adult Patients Starting Generic Versus Brand SSRI or SNRI Antidepressants in Commercial Health Plans

ANNA VLAHIOTIS, MA, is Senior Manager, Health Services Research; JEFF EICHHOLZ, PharmD, is Director, Clinical Programs; and ADAM KAUTZNER, PharmD, is Director, Clinical Programs, Express Scripts, Inc., St. Louis, Missouri. At the time this study was conducted, SCOTT T. DEVINE, PhD, was Director, Health Services Research, Express Scripts, Inc., St. Louis, Missouri.

AUTHOR CORRESPONDENCE: Anna Vlahiotis, MA, Research & New Solutions, Express Scripts, Inc., 4600 N. Hanley Rd, St. Louis, MO 63121. Tel: 314.684.6170; Email: avlahiotis@express-scripts.com.

Authors

DISCLOSURES

No external funding supported this research. At the time this study wasconducted,allauthorswereemployeesofExpressScripts, Inc.,apharmacybenefitsmanagementcompany,andheldstockinExpressScriptsaspartoftheiremployeebenefits.ConceptanddesignwereperformedbyDevineandVlahiotis.DatawerecollectedbyVlahiotisandinterpretedbyall4authors.ThemanuscriptwaswrittenandrevisedbyVlahiotiswiththeassistanceofDevine,Eichholz,andKautzner.

ACkNOwLEDGEMENTS

Theauthors thank the followingExpressScriptsemployees: JimIndelicato,SeniorBusinessAnalyst,Research andNewSolutions, forhis assistance inthepreparationofthedatausedintheanalysis;andKrisKotoucek,CreativeManager,Marketing, forhis assistance in reviewing and editing themanu-script.TheauthorsalsothanktheJMCPeditorsfortheirassistance.

REFERENCES

1.ZetinM,HoepnerCT,BjornsonL.Rationalantidepressantselec-tion:applyingevidence-basedmedicinetocomplexreal-worldpatients.Psychopharmacol Bull.2006;39(1):38-104.

2.DonohueJM,BendtER.Effectsofdirect-to-consumeradvertisingonmedicationchoice:thecaseofantidepressants.J Public Policy Marketing. 2004;23(2):115-27.Availableat:http://www.kff.org/rxdrugs/upload/Effects-of-Direct-to-Consumer-Advertising-on-Medication-Choice-the-Case-of-Antidepressants.pdf.AccessedJanuary13,2011.

132 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

23.BigosKL,PollockBG,StankevichBA,BiesRR.Sexdifferencesinthepharmacokineticsandpharmacodynamicsofantidepressants:anupdatedreview.Gend Med.2009;6(4):522-43.

24.KhanA,BrodheadAE,SchwartzKA,KoltsRL,BrownWA.Sexdiffer-encesinantidepressantresponseinrecentantidepressantclinicaltrials.J Clin Psychopharmacol.2005;25(4):318-24.

25.RobinsonRL,LongSR,ChangS,etal.HighercostsandtherapeuticfactorsassociatedwithadherencetoNCQAHEDISantidepressantmedica-tionmanagementmeasures:analysisofadministrativeclaims.J Manag Care Pharm.2006;12(1):43-54.Availableat:http://www.amcp.org/data/jmcp/research_43-54.pdf.

26.CharlsonME,PompeiP,AlesKL,MacKenzieCR.Anewmethodofclas-sifyingprognosticcomorbidityinlongitudinalstudies:developmentandvalidation.J Chronic Dis.1987;40(5):373-83.

27.DrussBG,RohrbaughRM,RosenheckRA.Depressivesymptomsandhealthcostsinoldermedicalpatients.Am J Psychiatry.1999;156(3):477-79.Availableat:http://ajp.psychiatryonline.org/cgi/reprint/156/3/477.AccessedJanuary13,2011.

28.MeredithLS,MendelP,PearsonM,etal.Implementationandmain-tenanceofqualityimprovementfortreatingdepressioninprimarycare.Psychiatr Serv.2006;57(1):48-55.Availableat:http://ps.psychiatryonline.org/cgi/reprint/57/1/48.AccessedJanuary13,2011.

29.OlfsonM,MarcusSC,TedeschiM,WanGJ.ContinuityofantidepressanttreatmentforadultswithdepressionintheUnitedStates.Am J Psychiatry. 2006;163(1):101-08.Availableat:http://ajp.psychiatryonline.org/cgi/reprint/163/1/101.AccessedJanuary13,2011.

30.TurnerMS,MayDB,ArthurRR.XiongGL.Clinicalimpactofselec-tiveserotoninreuptakeinhibitorstherapywithbleedingrisks.J Intern Med. 2007;261(3):205-13.

31.AgencyforHealthcareResearchandQuality,NationalQualityMeasuresClearinghouse.Antidepressantmedicationmanagement(effectivecontinu-ationphasetreatment):percentageofmemberswhowerediagnosedwithanewepisodeofmajordepression,treatedwithantidepressantmedica-tion,andwhoremainedonanantidepressantmedicationforatleast180days(6months).ReleasedJanuary1999,revisedJuly2009.Availableat:http://www.qualitymeasures.ahrq.gov/content.aspx?id=14961#Section304. AccessedJanuary13,2011.

32.MullinsCD,ShayaFT,MengF,WangJ,HarrisonD.Persistence,switch-ing,anddiscontinuationratesamongpatientsreceivingsertraline,parox-etine,andcitalopram.Pharmacotherapy.2005;25(5):660-67.

33.StahlSM.Basicpsychopharmacologyofantidepressants,part1:Antidepressantshavesevendistinctmechanismsofaction.J Clin Psychiatry. 1998;59(Suppl4):5-14.

34.BloughDK,RamseySD.Usinggeneralizedlinearmodelstoassessmedi-calcarecosts.Health Serv Outcomes Res Methodol.2000;1(2):185-202.

35.MissouriHouseofRepresentatives.Summaryoftheintroducedbill.HB458—PharmacyBenefitManagers,Provision8.November17,2009.Availableat:http://house.mo.gov/content.aspx?info=/bills091/bilsum/intro/sHB458I.htm.AccessedJanuary13,2011.

36.CascadeEF,KalaliAH,SheehanDV.GenericconversionoftheSSRImarketandtheimpactonbrandedproducts.Psychiatry (Edgemont). 2006;3(12):34-35.Availableat:http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2945894/.AccessedJanuary13,2011.

37.CoxE,BehmA,MagerD.2005GenericDrugUsageReport.ExpressScripts,Inc.internalreport.2005.Availableat:http://www.express-scripts.com/research/studies/pharmacybenefitresearch/benefitdesign/docs/gener-icDrugUsageReport2005.pdf.AccessedJanuary13,2011.

38.FernandezF,LevyJK,LacharBL,SmallGW.Themanagementofdepres-sionandanxietyintheelderly.J Clin Psychiatry.1995;56(Suppl2):20-29.

39.CzechowskiJL,TjiaJ,TrillerDM.Deeplydiscountedmedications:implicationsofgenericprescriptiondrugwars.J Am Pharm Assoc (2003). 2010;50(6):752-57.

3.BerndtER,BhattacharjyaA,MisholDN,ArcelusA,LaskyT.Ananaly-sisofthediffusionofnewantidepressants:variety,quality,andmarketingefforts.J Ment Health Policy Econ.2002;5(1):3-19.

4.GartlehnerG,HansenRA,ThiedaP,etal.Comparative Effectiveness of Second-Generation Antidepressants in the Pharmacologic Treatment of Adult Depression.ComparativeEffectivenessReviewNo.7.Rockville,MD:AgencyforHealthcareResearchandQuality.January2007.Availableat:http://www.ncbi.nlm.nih.gov/books/NBK43023/pdf/TOC.pdf.AccessedJanuary13,2011.

5.GartlehnerG,GaynesBN,HansenRA,etal.Comparativebenefitsandharmsofsecond-generationantidepressants:backgroundpaperfortheAmericanCollegeofPhysicians.Ann Intern Med.2008;149(10):734-50.

6.GartlehnerG,ThiedaP,HansenRA,etal.Comparativeriskforharmsofsecond-generationantidepressants:asystematicreviewandmeta-analysis.Drug Saf.2008;31(10):851-65.

7.CiprianiA,LaFerlaT,FurukawaTA,etal.Sertralineversusotherantidepressiveagentsfordepression.Cochrane Database Syst Rev. 2010;14(4):CD006117.

8.CiprianiA,FurukawaTA,SalantiG,etal.Comparativeefficacyandacceptabilityof12new-generationantidepressants:amultiple-treatmentsmeta-analysis.Lancet.2009;373(9665):746-58.

9.VanAmeringenM,ManciniC,PattersonB,BennetM.SymptomrelapsefollowingswitchfromCelexatogenericcitalopram:ananxietydisorderscaseseries.J Psychopharmacol.2007;21(5):472-76.

10.RosenthalJ,KongB,JacobsL,KatzmanM.Didaswitchtoagenericantidepressantcauserelapse?J Fam Pract.2008;57(2):109-14.

11.U.S.FoodandDrugAdministration.Reviewofthetherapeuticequiva-lencegenericbuproprionXL300mgandWellbutrinXL300mg.September18,2009.Availableat:http://www.fda.gov/AboutFDA/CentersOffices/CDER/ucm153270.htm.AccessedJanuary13,2011.

12.CarlatDJ.Genericsinpsychiatry:sortingthroughtheevidence.The Carlat Psychiatry Report.2009;7(6):1-2.

13.BorgheiniG.Thebioequivalenceandtherapeuticefficacyofgenericver-susbrand-namepsychoactivedrugs.Clin Ther.2003;25(6):1578-92.

14.WadeAG,SaragoussiD,DespiégelN,FrançoisC,GuelfucciF,ToumiM.Healthcareexpenditureinseverelydepressedpatientstreatedwithescitalopram,genericSSRIsorvenlafaxineintheUK.Curr Med Res Opin. 2010;26(5):1161-70.

15.EspositoD,WahlP,DanielG,StotoMA,ErderMH,CroghanTW.Resultsofaretrospectiveclaimsdatabaseanalysisofdifferencesinanti-depressanttreatmentpersistenceassociatedwithescitalopramandotherselectiveserotoninreuptakeinhibitorsintheUnitedStates.Clin Ther. 2009;31(3):644-56.

16.GleasonPP.Assessingstep-therapyprograms:astepintherightdirec-tion.J Manag Care Pharm.2007;13(3):273-75.Availableat:http://www.amcp.org/data/jmcp/273-75.pdf.

17.DunnJD,CannonE,MitchellMP,CurtissFR.Utilizationanddrugcostoutcomesofastep-therapyeditforgenericantidepressantsinanHMOinanintegratedhealthsystem.J Manag Care Pharm.2006;12(4):294-302.Availableat:http://www.amcp.org/data/jmcp/research_294-302.pdf.

18.PanzerPE,ReganTS,ChaioE,SarnesMW.ImplicationsofanSSRIgenericsteptherapypharmacybenefitdesign:aneconomicmodelinanxi-etydisorders.Am J Manag Care.2005;11(12Suppl):S370-S379.Availableat:http://www.ajmc.com/media/pdf/A147_05Panzer_S370to79.pdf.AccessedJanuary13,2011.

19.AmericanPsychiatricAssociation.Practiceguidelineforthetreatmentofpatientswithmajordepressivedisorder,thirdedition.Availableat:http://psychiatryonline.com/content.aspx?aID=654142.AccessedJanuary13,2011.

20.NolanL,O’MalleyK.Adverseeffectsofantidepressantsintheelderly.Drugs Aging.1992;2(5):450-58.

21.LotrichFE,PollockBG.Agingandclinicalpharmacology:implicationsforantidepressants.J Clin Pharmacol.2005;45(10):1106-22.

22.WeissmanMM,KlermanGL.Sexdifferencesandtheepidemiologyofdepression. Arch Gen Psychiatry.1977;34(1):98-111.

Discontinuation Rates and Health Care Costs in Adult Patients Starting Generic Versus Brand SSRI or SNRI Antidepressants in Commercial Health Plans

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 133

•Numerous factors complicate the clinical outcomes ofwarfarintherapy, such as concurrent disease states, drug regimens, dietand alcohol intake, physical illnesses, compliance, and patientknowledgeoftheirresponsibilitiesintherapy.

•Inastudyofpatientsaged80yearsoroldertreatedwithwarfa-rin,Kaganskyetal. (2004) found that insufficienteducationasperceivedbythepatientorcaregiverwasassociatedwithahigherrate of major bleeding events (5.2 per 1,000 patient-months)compared with no education (1.1 per 1,000 patient-months)andeducationthatwaspatient-ratedassufficient(0.5per1,000patient-months;P <0.001).

•Tang et al. (2003) found a slightly positive Pearson correla-tion (r) between knowledge aboutwarfarin and the number ofinternationalnormalized ratio (INR)valueswithin target range(r=0.2;P =0.024),indicatingthat4%ofvarianceinINRcouldbeexplainedbythepatient’swarfarinknowledge.

•Todate,2questionnairesmeasuringpatientknowledgeofantico-agulationtherapyhavebeenvalidated:theOralAnticoagulationKnowledge (OAK) test, created and validated by Zeolla et al.(2006), and the AnticoagulationKnowledgeAssessment (AKA)questionnaire,designedandvalidatedbyBriggsetal.(2005).Noanalysisoftherelationshipbetweenpatientknowledgeofwarfa-rintherapyandtherapeuticgoalachievementhasusedavalidatedinstrument.

What is already known about this subject

INR Goal Attainment and Oral Anticoagulation Knowledge of Patients Enrolled in an Anticoagulation

Clinic in a Veterans Affairs Medical Center

Jennifer W. Baker, PharmD, BCPS; Kristi L. Pierce, PharmD; and Casey A. Ryals, PharmD

ABSTRACT

BACKGROUND: In January 2009, the Joint Commission implemented a National Patient Safety Goal (NPSG) for ambulatory care, NPSG 3E, intend-ed to reduce harm associated with the use of anticoagulation therapy. The 2011 NPSG 3E encompasses 8 elements of performance, including require-ments that each organization (a) provide education regarding anticoagula-tion therapy to staff, patients, and families and (b) evaluate its safety prac-tices and take appropriate action to improve its practices. The Alvin C. York (ACY) outpatient anticoagulation clinic provides education to new patients and their families at the initial clinic visit, with follow-up reinforcement of education as needed throughout their care.

OBJECTIVES: To (a) assess the knowledge level of patients receiving warfa-rin therapy in an anticoagulation clinic using the validated Anticoagulation Knowledge Assessment (AKA) questionnaire and (b) examine the relation-ship between patient anticoagulation knowledge and anticoagulation con-trol as measured by the international normalized ratio (INR).

METHODS: All ACY Veterans Affairs (VA) anticoagulation clinic patients seen during their routine visit within an 8-week recruitment period from February 2010 to April 2010 were asked to complete the AKA questionnaire. Upon voluntary consent, the questionnaire was completed by the patient either during the clinic visit or returned later by mail. Demographic and clinical data were manually extracted from the computerized patient record system and included age, gender, indication for and duration of anticoagu-lation therapy, goal INR range, and the 10 INR values preceding the date of consent. A passing score was defined as at least 21 correct responses on the 29-item AKA questionnaire (72.4% correct). Statistical analyses includ-ed comparisons of demographic and clinical characteristics for patients with passing versus failing scores, assessed with Pearson chi-square and Fisher’s exact test, and bivariate analyses of INR control with anticoagula-tion knowledge, assessed with Spearman’s rho correlation. INR control was defined by 3 outcome measures: number of INRs within therapeutic range, time in therapeutic range (TTR) calculated using the Rosendaal method, and standard deviation (SD) of INR values. Anticoagulation knowledge was assessed with 2 measures: total AKA score and count of correct answers to a subset of 15 AKA items deemed by the investigators to be relevant to INR control.

RESULTS: Of 447 patients enrolled in the anticoagulation clinic, 260 con-sented to participate in the survey, of whom 185 patients completed the AKA instrument (n = 171 [92.4%] by mail) and were successfully matched to patient record system data. 178 (96.2%) respondents were male with a mean (SD) age of 68 (10.1) years. The majority of patients were undergo-ing anticoagulation treatment for atrial fibrillation (n = 113, 61.1%) or deep venous/pulmonary thromboembolism (n = 48, 25.9%). The majority of patients had been treated with warfarin for at least 1 year (n = 162, 87.6%). Most patients had goal INR ranges of 2.0 to 3.0 (n = 166, 89.7%). Of the 185 patients who completed the questionnaire, 137 (74.1%) achieved a passing score. The mean (SD) AKA questionnaire score was 78.1% (12.1%).

RESEARCH

There were 8 questions that were answered correctly by less than 70% of patients and identified as potential deficiencies in patient education. For the 167 patients who had been on warfarin therapy for at least 6 months and had 10 previous INR values, there was no significant Spearman’s rho correlation between total number of correct questionnaire responses and INR control, defined as the count of the 10 previous INR values within goal range (rho = –0.022, P = 0.776), TTR (rho = 0.015, P = 0.848), and SD (rho = 0.047, P = 0.550). There was also no significant relationship between number of correct INR-relevant responses and INR control by any of the 3 outcome measures (count in range rho = 0.033, P = 0.676; TTR rho = 0.067, P = 0.388; and SD rho = –0.029, P = 0.708).

CONCLUSIONS: Although 74.1% of patients on long-term warfarin therapy achieved a passing score of at least 21 correct answers on the 29-ques-tion AKA instrument, there was no significant relationship between patient warfarin knowledge and INR control. Areas for improvement in patient edu-cation have been identified and procedures for educational modification are currently in development.

J Manag Care Pharm. 2011;17(2):133-42

Copyright © 2011, Academy of Managed Care Pharmacy. All rights reserved.

134 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

INR Goal Attainment and Oral Anticoagulation Knowledge of Patients Enrolled in an Anticoagulation Clinic in a Veterans Affairs Medical Center

warfarin therapyandthepercentageof internationalnormal-izedratio(INR)valueswithintherapeuticgoalrange(r=0.20,P =0.024).10Additionally,inastudyofpatientsaged80yearsorolder,Kaganskyetal.(2004)foundthatinsufficienteducationonoralanticoagulationtherapy(asperceivedbythepatientorcaregiver)was a significant predictive factor formajor bleed-ing events (5.2per1,000patient-months) comparedwithnoeducation (1.1 per 1,000 patient-months) and education thatwaspatient-ratedas sufficient (0.5per1,000patient-months,P <0.001).ThepercentageofINRswithinthetherapeuticrangewashighestamongpatientswithperceivedsatisfactoryeduca-tion(45.1%)comparedwiththosewhoperceivedtheireduca-tion as insufficient (34.9%)or receivedno education (20.0%,P <0.001).11

However, not all studies have found apositive correlationbetweenpatientknowledgeandoutcomesofwarfarintherapy.Davisetal.(2005)foundthatadherencetotherapywithwar-farinwassignificantlyassociatedwithanticoagulationcontrol,definedasthenumberofbloodtestsinappropriatetherapeuticrangedividedbythenumberofbloodtestsperformedduringthe 60-day period; however, knowledge of warfarin therapy,assessed with an 18-question multiple-choice test, was notassociatedwithanticoagulationcontrol.12Only14%ofpatientsin the study by Davis et al. achieved good anticoagulationcontrol,definedasmorethan70%ofINRvalueswithinthera-peuticrange.12

Themajor limitation of all of the aforementioned studies,regardlessoftheirresults,wastheuseofanonvalidatedques-tionnaireor survey to evaluatepatientknowledge.Validationindicatesthatthequestionnairehasbeenthoroughlytestedforcontent validity, measures of question difficulty, readability,anditem/personreliability.Todate,2questionnairesmeasur-ingpatientknowledgeofwarfarintherapyhavebeenvalidated:theOralAnticoagulationKnowledge (OAK) test, created andvalidated by Zeolla et al. (2006),13 and the AnticoagulationKnowledge Assessment (AKA) questionnaire, designed andvalidatedbyBriggsetal.(2005).14

Evaluationofcurrentpatientknowledgeisthefirststeptoimproving the quality of anticoagulation therapy andpatientcare.Deficiencies inpatientknowledgecanbe identifiedandaddressed,creatinganongoingsystemofqualityimprovementforanticoagulationmonitoringandpatientsafety.Thepresentstudyassessedthecurrentknowledgelevelofpatientsreceiv-ingwarfarintherapyintheoutpatientanticoagulationclinicattheAlvinC.York(ACY)campusof theVeteransAffairs (VA)TennesseeValleyHealthcareSystem(TVHS).Study investiga-torsconductedthisresearchtodetermineifcorrectiveactionswereneededtoimprovepatientknowledgeinaccordancewiththestandardsofNPSG3Eelementnumber8.Additionally,todate no published studies have used a validated instrumentto assess the association of patient anticoagulation knowl-edgewith INRcontrol.Therefore, thepresent studyassessed

A nticoagulationwithwarfarinisawell-establishedandcrucial intervention for thepreventionand treatmentof thromboembolic events.1,2,3 Numerous factors can

affecttheclinicalresponsetoanticoagulants,particularlywar-farin,suchasapatient’sconcurrentdiseasestates,drugregi-mens,dietandalcoholconsumption,physical illnesses,com-pliance, andoverall knowledgeof therapy.1-4For this reason,patientsrequireaskillfulheathcareproviderinadditiontoasystematicapproachtoensureproperpatientfollow-upduringtheinitiationandcontinuationofwarfarintherapy.

Reflectingtheseneeds,in2009theJointCommissioncre-ated aNational Patient Safety Goal (NPSG) 03.05.01 (or 3E)forambulatorycareto“reducethelikelihoodofpatientharmassociated with the use of anticoagulation therapy.”5,6 The2011NPSG3Eguidelinesdescribepatienteducationas“avitalcomponent”ofawarfarintherapyprogram,statingthatprovid-ersshouldmakesurepatientsunderstandrisksinvolvedwiththerapy,precautionstotake,andtheneedforclosefollow-up(Figure 1). Element number 7 ofNPSG 3E requires that theorganizationprovideeducationregardinganticoagulationther-apytostaff,patients,andfamilies.Elementnumber8requiresthat“theorganizationevaluateanticoagulationsafetypractices,takeappropriateactiontoimprovepractices,andmeasuretheeffectivenessofthoseactions.”6,7

Todate,a fewstudieshaveshownanassociationbetweenthe outcomes of anticoagulation therapy and health literacy,patienteducation,orknowledgeofwarfarintherapy;however,results are mixed.8-12 In a sample of 122 patients attendinga warfarin clinic, Tang et al. (2003) discovered a positivePearson correlation (r) between patient knowledge about

•Of185patientswhoattendedanoutpatientanticoagulationclinicand received education at the initial clinic visit and as neededthereafter,137(74.1%)achievedapassingscoreofatleast21cor-rectanswersonthe29-itemAKAquestionnaire,whichcompareswithanaveragepassrateof31.3%reportedinpreviousliteraturefordifferentanticoagulationknowledgeinstruments.

•This is the first study to use a validated knowledge assess-ment questionnaire tool to assess the relationship betweenwarfarin knowledge and therapeutic goal range achievement.Surprisingly, there was no significant correlation betweenanticoagulationknowledge,definedbythenumberofcorrectlyansweredquestions,andINRcontrol,definedasthenumberofINRvalueswithintherapeuticrange,timeintherapeuticrangecalculatedusingtheRosendaalmethod,orstandarddeviationofINRvalues.

•Assessment of patient warfarin knowledge might be used inquality improvement initiatives in anticoagulation monitoringandpatientsafetyincludingidentificationofareasofknowledgedeficiency.

What this study adds

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 135

the association of performance on the AKA questionnaire(Appendix)14 with anticoagulation control defined using 3methods: (a) number of INRs within therapeutic range, (b)time in therapeutic range (TTR), and (c) anticoagulation sta-bilitymeasuredas the standarddeviation (SD)of INRvalues(lowerSDvaluesindicategreaterstability).

■■  MethodsStudy Design and SettingThiswasasingle-centercross-sectionalanalysisofallcurrentpatientsintheVATVHSACYanticoagulationclinic.ThisstudywasapprovedbytheInstitutionalReviewBoardandResearchand Development committees within TVHS. The outpatientanticoagulationclinicattheACYcampusoftheVATVHSwasestablished in2007 and is currently staffedwith adoctor ofpharmacy and a licensed practical nurse. Since initiation oftheACYanticoagulationclinic,patientsand/or theirprimarycaregiversareeducatedbytheclinicalpharmacyspecialistattheirfirstvisitregardlessoflengthofwarfarintherapypriortotheirinitialclinicvisit.Atthetimeofthisstudy,therewere447patientsenrolledintheanticoagulationclinic.

Study SampleAll 447 ACY clinic patients seen during their routine visitswithin an8-week recruitmentperiod fromFebruary2010 toApril 2010were askedby the clinicalpharmacy specialist to

complete the AKA questionnaire. Patients were excluded iftheyrefusedparticipation,couldnotvoiceunderstandingafterreadingtheinformedconsent,orifthestudyinvestigatorsfeltintheirbestclinicaljudgmentthatthepatientdidnotunder-standconsent.Patientswithoutatleast10INRreadingsfromtheACYclinicpriortotheconsentdateandpatientswhohadbeen enrolled in the clinic less than 6months were eligibletocomplete theAKAquestionnairebutwerenot included inthedataanalysisoftherelationshipbetweenINRcontrolandwarfarinknowledge.

Study Procedures Onceconsentwasreceived,theAKAinstrumentwasgiventothepatient.Patientswereinstructedtocompletethequestion-naireindependentlyandwithoutassistance.Thepatientcouldeither fill out the questionnairewhile in the office or take ithome andmail it back to the clinic. If thepatient requestedtocompletethesurveyathome,apostage-paidself-addressedenvelopewasprovidedforthereturnofthecompletedsurvey.Surveys were coded sequentially upon order of consent andpaired with patient identifiers. This step was performed sothatdemographicdataandthepatient’s10INRmeasurementscould be extracted manually from the computerized patientrecordsystem(CPRS)ata laterdate.Missinganswers insur-veyresponseswerescoredasincorrectanswers.ResultsoftheAKA instrument as completed by the enrolled patients were

INR Goal Attainment and Oral Anticoagulation Knowledge of Patients Enrolled in an Anticoagulation Clinic in a Veterans Affairs Medical Center

FIGURE 1 The Joint Commission’s Elements of Performance for NPSG.03.05.01Effective January 1, 2011. To Reduce the Likelihood of Patient Harm Associated with the Use of Anticoagulant Therapya

1. Use only oral unit-dose products, prefilled syringes, or premixed infusion bags when these types of products are available. Note: For pediatric patients, prefilled syringe products should be used only if specifically designed for children.

2. Use approved protocols for the initiation and maintenance of anticoagulant therapy.b

3. Before starting a patient on warfarin, assess the patient’s baseline coagulation status; for all patients receiving warfarin therapy, use a current International Normalized Ratio (INR) to adjust this therapy. The baseline status and current INR are documented in the medical record.b

4. Use authoritative resources to manage potential food and drug interactions for patients receiving warfarin.

5. When heparin is administered intravenously and continuously, use programmable pumps in order to provide consistent and accurate dosing.

6. A written policy addresses baseline and ongoing laboratory tests that are required for anticoagulants.

7. Provide education regarding anticoagulant therapy to prescribers, staff, patients, and families. Patient/family education includes the following:b

- The importance of follow-up monitoring- Compliance- Drug-food interactions- The potential for adverse drug reactions and interactions

8. Evaluate anticoagulation safety practices, take action to improve practices, and measure the effectiveness of those actions in a time frame determined by the organization.b

aThe Joint Commission. Accreditation program: ambulatory care. National Patient Safety Goals;6 and the Joint Commission. Accreditation program: hospital. National Patient Safety Goals.7bOnly performance elements 2, 3, 7, and 8 apply to the ambulatory care setting.

136 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

compiledwithdemographicandclinical information, includ-ing indications for anticoagulation and duration of warfarintreatment,fromthepatients’medicalrecords.

Definitions of Study VariablesTheprimaryoutcomewasthelevelofknowledgeasdeterminedbythescoreontheAKAquestionnaire.TheAKAquestionnaireismadeupof29multiplechoicequestions.Eachquestionwasworth3.45points.Correctlyanswering21questions(72.4%)ormorewasneededfordeterminationofadequateknowledgeof anticoagulation therapy (passing score). The secondaryoutcomewasINRcontrol,definedusing3differentmethods:(a) thecountof the10most recent INRvalueswithin thera-peuticgoalrange(0,1,2,3,etc.),(b)TTRcalculatedusingtheRosendaalMethod,15and(c)anticoagulationstabilitymeasuredastheSDofINRvalues.Thesemeasureshavebeenusedinpre-viousstudiesofINRcontrol.8,10,16,17The10INRvaluesobtainedpriortothedateofpatientconsentwereassessedforevaluationof apatient’s recent INRcontrol at the timeof assessmentofwarfarinknowledgeusingtheAKAinstrument.

Two different scoring methods for the AKA instrumentwere used to assess correlation of warfarin knowledge with

INRcontrol.The firstwas the total countof correct answersfortheinstrument,withamaximumof29.Thesecondscoringmethodeliminatedquestionsthatweredeemedbytheinves-tigators to be of limited relevance to INR control, for a totalmaximumscoreof15(Appendix).

Statistical AnalysisAssociations between the independent variables and passingversus failing score, measured as a binomial, were assessedusingFisher’sexacttestorPearsonchi-squaretestforcategori-calvariables.Becausethedatawerenotnormallydistributed,Spearman’s rho correlation analysis was used instead of aPearson correlation to assess the relationships between eachofthe3measuresofINRcontrolandbothmeasuresofantico-agulationknowledge.Analyseswereperformedwithanapriorialphavalueof0.05usingSPSSversion17.0(SPSSInc.,Chicago,IL) and the GraphPad InStat statistical package (GraphPadSoftwareInc.,LaJolla,CA).

■■  ResultsOfthe260anticoagulationclinicpatientswhoconsenteddur-ing the 8-week enrollment period, 186 (71.5%) returned acompletedquestionnaire(Figure2).Onepatientwas lostdue

INR Goal Attainment and Oral Anticoagulation Knowledge of Patients Enrolled in an Anticoagulation Clinic in a Veterans Affairs Medical Center

FIGURE 2 Patient Selection Flowchart

447 total patients in the anticoagulation clinic

167 patients assessed for secondary outcomeb

260 (58.2%) patients consented

185 questionnaires assessed for primary outcomea

74 patients did not complete the questionnaire

186 (71.3%) completed questionnaires• 15 (8.1%) completed at facility• 171 (91.9%) mailed to clinic

1 patient lost due to unmatched questionnaire

7 patients excluded for less than 10 INR values

11 patients excluded for warfarin therapy of less than 6 months duration

aPrimary outcome: total score on the validated AKA. bSecondary outcome: correlation of warfarin knowledge with anticoagulation control. AKA = Anticoagulation Knowledge Assessment questionnaire (Appendix), INR = international normalized ratio.

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 137

totheinabilitytomatchhisorherquestionnairewithpatientidentifiers for data extraction purposes, leaving 185 patientsfordataanalysis.DemographicandclinicaldataareinTable1.Themean(SD)ageofpatientsinthesamplewas68(10.1)yearswith178malepatients(96.2%).Themajorityofpatientswerebeingtreatedforatrialfibrillation(n=113,61.1%).Otherindi-cations included deep venous thrombosis/pulmonary embo-lism(DVT/PE;n=48,25.9%),valvereplacement(n=18,9.7%),cerebralvascularaccident/transientischemicattack(CVA/TIA;n=12, 6.5%), and other indications (n=9, 4.9%). Indicationcategories are not mutually exclusive—that is, patients withmultipleindicationswereincludedinallapplicablecategories.Mostpatients (n=162,87.6%)hadbeen treated for at least1year,with11patients(5.9%)treatedforlessthan6monthsand12patients(6.5%)treatedfor6monthsto1year.Intheassess-mentof individual goal INRspredeterminedby theproviderreferringthepatient to theclinic,166patients(89.7%)hadagoalrangeof2-3,15patients(8.1%)hadagoalrangeof2.5-3.5,

and4patients(2.2%)hadauniquegoalrange.Of the 185 patients who completed the 29-item AKA

questionnaire,137(74.1%)achievedascoreofatleast21cor-rect items (72.4%). Themean (SD) AKA questionnaire scorewas 78.1% (12.1%), representing approximately 22.6 correctresponses. Eight questions were answered correctly by lessthan70%ofpatients(Table2).Nopatientachievedascoreof100%.Forthecorrelationanalyses,11patientswhohadbeentreatedfor6monthsorlessand7patientswithaninsufficientnumberofINRswereexcluded,leavingatotalof167patientsfordata analysis.Demographic and clinical characteristics ofpatientsincludedinthecorrelationanalysesareshowninTable3.ComparingsubgroupscategorizedbypassingversusfailingAKA score, no statistically significant between-group differ-enceswerefound.

Spearman’srhocorrelation(rho)analyses(Table4)showedno significant correlations between total number of correctAKA answers and any of the 3 measures of INR control,whether defined as the count of the 10 previous INR valueswithin goal range (rho=–0.022, P =0.776), TTR (rho=0.015,P =0.848),orSD(rho=0.047,P =0.550).Insensitivityanalyseslimited to 15 AKA items deemed by the study investigatorsto be relevant to INR control (Appendix), therewere no sig-nificantrelationshipsbetweennumberofcorrectINR-relevantresponsesandINRcontrol(rho=0.033,P =0.676;rho=0.067,P =0.388;rho=–0.029,P =0.708).

■■  DiscussionCurrently 2 anticoagulation knowledge questionnaires havebeenvalidatedforcontentvalidity,constructvalidity,andreli-ability.TheAKAquestionnaireusedareviewofpublishedliter-atureandanticoagulationclinicprotocols,aswellasinterviewswith anticoagulation pharmacists for content validity. AKAinvestigatorscreateda29-iteminstrumentcovering9contentareas,includingmedication,medicationadministration,medi-cationinteractions,activity,diet,sideeffects,informinghealth

INR Goal Attainment and Oral Anticoagulation Knowledge of Patients Enrolled in an Anticoagulation Clinic in a Veterans Affairs Medical Center

TABLE 1 Baseline Characteristics of Patients Who Completed the AKA Questionnaire (N = 185)

Parameter n (%)

Age(mean=68;SD=10.1years)Youngerthan50years 6 (3.2)50to59years 22 (11.9)60to69years 73 (39.5)70to79years 61 (33.0)80yearsorolder 23 (12.4)

Gender Male 178 (96.2)Female 7 (3.8)

INR goal 2.0-3.0 166 (89.7)2.5-3.5 15 (8.1)Other 4 (2.2)

Indicationa

Atrialfibrillation 113 (61.1)Valvereplacement 18 (9.7)DVT/PE 48 (25.9)CVA/TIA 12 (6.5)Other 9 (4.9)

Duration of warfarin therapy Lessthan6months 11 (5.9)6monthsto1year 12 (6.5)1to3years 53 (28.6)3to5years 38 (20.5)Morethan5years 71 (38.4)

aPatients with multiple indications for anticoagulation were included in all applicable categories. Therefore, indications sum to more than 100%.AKA = Anticoagulation Knowledge Assessment questionnaire (Appendix); CVA = cerebral vascular accident; DVT = deep venous thrombosis; INR = international normalized ratio; PE = pulmonary embolism; SD = standard deviation; TIA = transient ischemic attack.

TABLE 2 Frequently Missed Questionsa

Question Answered Correctly

#25–foodsthataffectwarfarin 33%#16–liquidsthataffectwarfarin 42%#26–genericandbrandwarfarin 45%#5–effectofspinachindiet 55%#6–effectofalcoholonINR 62%#21–symptomsofsideeffectsofwarfarin 65%#28–expecteddurationofwarfarintherapy 66%#15–sideeffectsofelevatedINR 67%aIndicates questions that were answered correctly by less than 70% of respondents. See Appendix for Anticoagulation Knowledge Assessment questionnaire for full question text. Of 186 respondents, 171 (91.9%) returned the survey by mail. INR = international normalized ratio.

138 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

72.4% (21 of 29) of questions answered correctly, we foundthat74.1%ofourpatientsachievedapassingscore.However,patienteducationmethodsandstudymethodologyforthepre-viousstudiesvariedamonganticoagulationclinicsandpatientsamples,makingdirectcomparisonofstudyresultsdifficult.

Eight questions were answered correctly by less than70% of patients. Four frequently missed questions covereddietarymodifications;2addressedmedication information;1addressedsideeffects;and1addressedinforminghealthcareproviders.The4mostfrequentlymissedquestionswereevalu-atedforpossiblecausesofpatients’incorrectanswers.Question5 addressed dietary modifications, and 59 (31.9%) patientspickedtheanswerindicatingthatgreenleafyvegetablesshouldbeavoidedcompletelyinsteadofthechoicefordietaryconsis-tencyfromweektoweek.Question16alsoaddresseddietaryconcerns with 63 (34.1%) patients choosing orange juiceinsteadofnutritionalsupplementastheagentthatcandecreaseeffectivenessofwarfarin.Thisproblemcouldhavebeenduetoconfusionoverdiscussionintheclinicregardingotherjuices,suchascranberryandgrapefruit,whichcanincreasetheanti-coagulanteffectofwarfarin.Onquestion25,anotherdietaryevaluation, 44 (23.8%) and73 (39.5%)patients, respectively,answered celery and green beans instead of cole slaw forhaving themost effecton theirwarfarin therapy.Finally, theoutcomes for question26 reflected a conflict in the teachingstylesoftheACYclinicstaffandtheinvestigatorswhocreatedtheAKAquestionnaire.Thequestionaskedpatientswhattheyshoulddoiftheyhadbothbrandandgenericwarfarinathome.Thecorrectanswerwastotakeoneortheotherbutnotboth.However,anotheranswerstatednottotakeeitheruntilspeak-ing with their health care professional. Patients in the ACYclinicaretaughttocalltheclinicwithanyquestionstheymay

care providers, procedures, and lab monitoring. For contentvalidity they used the Marzano’s Taxonomy. For reliabilityandoverallanalysisof their instrument, theyusedtheRaschdichotomousmodel.14TheAKAquestionnairewasemployedinourstudybecauseofitsrangeoftopicsimportantinwarfarineducation.Patientresponsesprovideobjectivedataaboutdif-ferentaspectsofpracticethataretaughtinpatienteducation.Thus,theinstrumentservedasagoodqualitycontrolmeasureofpatientcounselingeffectiveness.

Evaluationofanticoagulationknowledgehasyieldedunde-sirablepass rates inprevious literature.Usinganovel surveycontaining7open-endedquestions,Tangetal.foundthatonly18%ofpatientsachievedapassingscoreofatleast70%.10Davisetal.foundthat37%ofpatientsachievedapassingscoreofatleast70%onanovel18-questionmultiple-choicetest.12Usingapreviouslycreated20-questiontrue-or-falsequestionnaire,Huetal.(2006)founda39%passrate,definedasascoreofatleast80%.18InusingtheAKAquestionnaireintheACYVATVHSanticoagulationclinicanddefiningapassingscoreasat least

INR Goal Attainment and Oral Anticoagulation Knowledge of Patients Enrolled in an Anticoagulation Clinic in a Veterans Affairs Medical Center

Parameter

Patients with Passing Scorea

(n = 124)

Patients with Failing Scorea

(n = 43) P Valueb

Age–mean[SD]inyears 68.0 [9.9] 70.6 [10.9]GenderMale 119 (96.0) 42 (97.7) 1.000Female 5 (4.0) 1 (2.3) 1.000

INR Goal Range2.0-3.0 112 (90.3) 37 (86.0) 0.4082.5-3.5 9 (7.3) 5 (11.6) 0.356Other 3 (2.4) 1 (2.3) 1.000

Indicationc

Atrialfibrillation 79 (63.7) 27 (62.8) 1.000DVT/PE 31 (25.0) 11 (25.6) 0.843Valvereplacement 12 (9.7) 5 (11.6) 0.770CVA/TIA 8 (6.5) 2 (4.7) 1.000Other 6 (4.8) 1 (2.3) 0.681

Duration of warfarin therapy6monthsto1year 9 (7.3) 1 (2.3) 0.4561to3years 36 (29.0) 17 (39.5) 0.2543to5years 30 (24.2) 6 (14.0) 0.199Greaterthan5years 49 (39.5) 19 (44.2) 0.595

aPassing score was defined as at least 21 questions answered correctly or an AKA questionnaire score of 72.4%. bAssociations between variables were assessed using Fisher’s exact test or Pearson’s chi-square test.cPatients with multiple indications for anticoagulation were included in all appli-cable categories. Therefore, indications sum to more than 100%.AKA = Anticoagulation Knowledge Assessment questionnaire (Appendix); CVA = cerebral vascular accident; DVT=deep venous thrombosis; INR = inter-national normalized ratio; PE = pulmonary embolism; SD = standard deviation; TIA = transient ischemic attack.

TABLE 3 Demographics and Clinical Characteristics of Patients (n = 167) in Spearman’s Rho Correlation Analysis

TABLE 4 Spearman’s Rho Correlation Analysis of Anticoagulation Knowledge with INR Control

Number in Rangea

Time in Therapeutic

RangeaStandard

Deviationa

Spearman’sRho(P)

Spearman’sRho(P)

Spearman’sRho(P)

TotalAKAscore –0.022 (0.776) 0.015 (0.848) 0.047 (0.550)

INR-relevant AKAitemsb

0.033 (0.676) 0.067 (0.388) –0.029 (0.708)

aNumber in range is the count of INR values in therapeutic range, with a maximum of 10, measured prior to the consent date for participation in the study. Standard deviation was calculated for the same 10 INR values. Time in therapeutic range was calculated using the Rosendaal method.15 bCount of correct answers in a subset of AKA items deemed by the investigators to be relevant to INR control (Appendix).AKA = Anticoagulation Knowledge Assessment questionnaire; INR = international normalized ratio.

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 139

beenanswered.Ifasimilarstudyisconductedinthefuture,itwouldbebeneficialforpatientstocompletethequestionnairein a standardized,monitored environment andhave comple-tionof theentirequestionnaireverifiedbefore submission topreventtheselimitationsfromrecurring.Third,of447patientstreated in theclinic,260consented,butnotall187noncon-sentingpatientsrefusedparticipation.Somewereexcludedbythe investigators because of inability to voice understandingof theproject or because the investigators determined, usingtheirbestclinicaljudgment,thatthepatientwouldbeunabletounderstandconsent.These exclusion factors are related tomentalcapacity,apotentialconfoundingfactorinthisstudy’sresults;however,countsofpatientsexcludedforthesereasonswerenot trackedduringsampleselection,andweareunableto report them. Fourth, the limited patient population avail-ablewithintheVAanticoagulationclinicsettingofthisstudypresentsasignificantlimitation;thestudysampleconsistedofveterans,mostofwhomareoldermales.

■■  ConclusionsThe present study assessed anticoagulation knowledge ofpatients receiving warfarin therapy in a VA outpatient anti-coagulation clinic using a validated instrument and found apass rate of 74.1%,much higher than reported previously inthe literature. We found no statistically significant relation-shipsbetween3measuresofINRcontrolandanticoagulationknowledge. However, this is the first report of the relation-ship between INR control and anticoagulation knowledge asassessedbytheAKAquestionnaire,anditispossiblethatthisassessmentinstrument,althoughpreviouslyvalidated,wasnotsufficientlysensitiveindetectingclinicallyimportantwarfarinknowledge.Thefrequentlymissedquestionsindicatepotentialareas for improvement in patient education. Re-evaluation ofthecurrenteducationaltechniquesisunderwayforidentifica-tionofareaswhereimprovementsmaybemade.

haveorchanges inmedications/diet.Therefore,bothanswerscouldbeconsideredcorrect.

These frequentlymissedquestions indicatepotentialareasforimprovementinpatienteducation,includingreinforcementof dietary guidelines forwarfarin therapy aswell aswhen itis appropriate to contact the clinic for questions. Both areasrepresent potential starting points for re-education of cur-rentpatientsandprimaryeducation fornewpatients seen inthe clinic. At this point, re-evaluation of the current educa-tionaltechniquesisindicatedforidentificationofareaswhereimprovementsmaybemade.

Therewerenostatisticallysignificantdifferencesindemo-graphicorclinicalcharacteristicsbetweenpassandfailgroups,includingdurationoftreatmentorindicationoftherapy,age,orgender.DatafromtheSpearman’srhoanalysesrevealednosta-tisticallysignificantcorrelationsbetweenwarfarinknowledgeand INR control. These results add to the existing literaturethathas foundmixedresultswhenassessingtherelationshipbetweenpatientwarfarinknowledgeandINRcontrol.Tangetal.did finda smallpositive correlationbetweenanticoagula-tion knowledge and the number of INR valueswithin targetrange,showingthat4%ofvarianceinINRcouldbeexplainedbyanticoagulationknowledge.10However,Davisetal.showedno significant association between knowledge or educationandtheproportionofINRswithinthetherapeuticrange.12Ourstudy revealed ahigherpass rateof74.1%as comparedwithpass rates reportedpreviously in the literature, but it ispos-siblethattheAKAquestionnairewasnotsufficientlysensitiveindetectingwarfarinknowledgethatisclinicallyimportantinthe3measuresofINRcontrol.

Limitations First,thecollectionofsurveydatafrompatientswasnotstan-dardized. Patients could complete the knowledge test in theclinic,takeithomeandreturnitbymail,orreturnitattheirnextscheduledappointment.Fifteenpatients(8.1%)completedthequestionnaireatthefacility;theother171patients(92.4%)completed the questionnaire outside of the facility.Althoughpatients were expected to complete the questionnaire inde-pendently regardless of the site of completion, there was nowaytopreventpatientsfromelicitingassistancefromvariousresourcestohelpthemcompletethequestionnaire.Thisprob-lemmayhavecontributedtoaninflatedpassrateinthisstudy.Second,66respondents(35.7%)leftatleast1questionunan-swered.Ofthe66respondentswithincompletequestionnaires,12 left more than 4 unanswered questions, 3 missed entirepages, and 2 failed to complete the questionnaire. Becauseunansweredquestionswereconsideredincorrectanswers,thelatter5patientsreceivedafailingscore.Thisdecisionmayhavecontributed to cases that were judged as knowledge failureswhowouldhaveachievedpassscoresifallofthequestionshad

INR Goal Attainment and Oral Anticoagulation Knowledge of Patients Enrolled in an Anticoagulation Clinic in a Veterans Affairs Medical Center

JENNIFER W. BAKER, PharmD, BCPS, is Clinical Pharmacy Specialist, Anticoagulation and Traveling Veteran Clinics; at the time of this study KRISTI L. PIERCE, PharmD, was a Pharmacy Practice Resident (PGY1); and CASEY A. RYALS, PharmD, was a Pharmacy Practice Resident (PGY1), VA Tennessee Valley Healthcare System, Murfreesboro, Tennessee.

AUTHOR CORRESPONDENCE: Jennifer W. Baker, PharmD, BCPS, VA Tennessee Valley Healthcare System, 3400 Lebanon Pike, Firm D, Murfreesboro, TN 37129. Tel.: 615.225.2912; Fax: 615.225.2918; E-mail: Jennifer.Baker4@va.gov.

Authors

140 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

DISCLOSURES

The authors report no financial or other conflicts of interest related to thesubjectsorproductsdiscussedinthisarticle.Aposterabstractofthisresearchwaspresentedatthe2010AnnualMeetingoftheAmericanCollegeofClinicalPharmacy,October19,2010,inAustin,Texas.

StudyconceptanddesignweresharedequallybyBaker,Pierce,andRyals.Data collection and interpretationwereperformedprimarilybyBakerwithassistancefromPierceandRyals.The3authorssharedinwritingthemanu-script,andtherevisionsweremadeprimarilybyBaker.

ACKNOWLEDGEMENTS

TheauthorswouldliketothankJenniferBean,PharmD,BCPS,andM.ShawnMcFarland, PharmD, BCPS, of the Residency Research Advisory Board fortheirassistanceandguidanceinthisresearchproject.

REFERENCES

1.AnsellJ,HirshJ,HylekE,JacobsonA,CrowtherM,PalaretiG.PharmacologyandmanagementofthevitaminKantagonists:AmericanCollegeofChestPhysiciansEvidence-BasedClinicalPracticeGuidelines(8thedition).Chest.2008;133(6Suppl):160S-198S.Availableat:http://chest-journal.chestpubs.org/content/133/6_suppl/160S.full.pdf+html.AccessedJanuary12,2011.

2.SchulmanS,BeythRJ,KearonC,LevineMN.Hemorrhagiccomplicationsofanticoagulantandthrombolytictreatment:AmericanCollegeofChestPhysiciansEvidence-BasedClinicalPracticeGuidelines(8thedition).Chest. 2008;133(6Suppl):257S-298S.Availableat:http://chestjournal.chestpubs.org/content/133/6_suppl/257S.full.pdf+html.AccessedJanuary12,2011.

3.BarcellonaD,ContuP,MarongiuF.Patienteducationandoralantico-agulationtherapy.Haematologica.2002;87(10):1081-86.Availableat:http://www.haematologica.org/cgi/reprint/87/10/1081.AccessedJanuary12,2011.

4.TaylorFC,RamsayME,TanG,GabbayJ,CohenH.Evaluationofpatients’knowledgeaboutanticoagulationtreatment.Qual Health Care.1994;3(2):79-85.Availableat:http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1055201/pdf/qualhc00011-0015.pdf.AccessedJanuary12,2011.

5.SipkoffM.JointCommissionsetsanticoagulanttherapysafetygoal.Drug Topics.August20,2007.Availableat:http://drugtopics.modernmedicine.com/drugtopics/Treatment+Areas/Joint-Commission-sets-anticoagulant-therapy-safety/ArticleStandard/Article/detail/450166.AccessedJanuary12,2011.

6.TheJointCommission.Accreditationprogram:ambulatorycare.NationalPatientSafetyGoals.EffectiveJanuary1,2011.Availableat:http://www.jointcommission.org/assets/1/6/2011_NPSGs_AHC.pdf.AccessedJanuary12, 2011.

7.TheJointCommission.Accreditationprogram:hospital.NationalPatientSafetyGoals.EffectiveJanuary1,2011.Availableat:http://www.jointcom-mission.org/assets/1/6/2011_NPSGs_HAP.pdf.AccessedJanuary12,2011.

8.EstradaCA,Martin-HryniewiczM,PeekBT,CollinsC,ByrdJC.Literacyandnumeracyskillsandanticoagulationcontrol.Am J Med Sci. 2004;328(2):88-93.

9.FangMC,MachtingerEL,WangF,SchillingerD.Healthliteracyandanti-coagulation-relatedoutcomesamongpatientstakingwarfarin.J Gen Intern Med.2006;21(8):841-46.Availableat:http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1831580/pdf/jgi0021-0841.pdf.AccessedJanuary12,2011.

10.TangEO,LaiCS,LeeKK,WongRS,ChengG,ChanTY.Relationshipbetweenpatients’warfarinknowledgeandanticoagulationcontrol.Ann Pharmacother.2003;37(1):34-39.

11.KaganskyN,KnoblerH,RimonE,OzerZ,LevyS.Safetyofanti-coagulationtherapyinwell-informedolderpatients.Arch Intern Med. 2004;164(18):2044-50.Availableat:http://archinte.ama-assn.org/cgi/reprint/164/18/2044.pdf.AccessedJanuary12,2011.

12.DavisNJ,BillettHH,CohenHW,ArnstenJH.Impactofadherence,knowledge,andqualityoflifeonanticoagulationcontrol.Ann Pharmacother. 2005;39(4):632-36.

13.ZeollaMM,BrodeurMR,DominelliA,HainesST,AllieND.Developmentandvalidationofaninstrumenttodeterminepatientknowledge:theoralanticoagulationknowledgetest.Ann Pharmacother. 2006;40(4):633-38.

14.BriggsAL,JacksonTR,BruceS,ShapiroNL.Thedevelopmentandper-formancevalidationofatooltoassesspatientanticoagulationknowledge.Res Social Adm Pharm.2005;1(1):40-59.Availableat:http://download.jour-nals.elsevierhealth.com/pdfs/journals/1551-7411/PIIS1551741104000038.pdf.AccessedJanuary12,2011.

15.RosendaalF,CannegieterS,VanDerMeerF,BrietE.Amethodtodeter-minetheoptimalintensityoforalanticoagulanttherapy.Thromb Haemost. 1993;69(3):236-39.

16.HirschlM,PluschnigU,KundiM,KatzenschlagerR.Oralanticoagula-tioninolderpatientswithvascularorcardiovasculardiseases.Agedover70years:samerisk?Samebenefit?Int Angiol.2003;22(4):370-75.

17.SconceE,AveryP,WynneH,KamaliF.VitaminKsupplementationcanimprovestabilityofanticoagulationforpatientswithunexplainedvariabilityinresponsetowarfarin.Blood.2007;109(6):2419-23.

18.HuA,ChowC,DaoD,ErrettL,KeithM.Factorsinfluencingpatientknowledgeofwarfarintherapyaftermechanicalheartvalvereplacement.J Cardiovasc Nurs.2006:21(3);169-75.

INR Goal Attainment and Oral Anticoagulation Knowledge of Patients Enrolled in an Anticoagulation Clinic in a Veterans Affairs Medical Center

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 141

INR Goal Attainment and Oral Anticoagulation Knowledge of Patients Enrolled in an Anticoagulation Clinic in a Veterans Affairs Medical Center

APPEnDIx Anticoagulation Knowledge Assessment (AKA) Questionnairea

1. WhichoneofthesemedicationsisrecommendedifyouaretakingCoumadin(warfarin)andwantrelieffromaheadache?a. Advilb. Motrinc. Aspirind. Tylenol

2. WhichofthefollowingfooditemswouldinterferewithyourCoumadin(warfarin)medication?a. Baconb. Broccolic. Bananasd. Peeledcucumbers

3. WhileonCoumadin(warfarin)medication,inwhichofthefollowingwouldyougodirectlytotheemergencyroom?a. Smallbruisesb. Yourappetitedramaticallyincreasesc. Nosebleedwhichwillnotstopbleedingd. Gumswhichbleedforafewsecondsafterbrushingteeth

4. YoujustrememberedthatyouforgottotakeyoureveningCoumadin(warfarin)medicationdoselastnight.Youwould—a. skipthedoseofCoumadin(warfarin)youmissedb. takethemissedCoumadin(warfarin)doserightnowc. waitandtake2dosesofCoumadin(warfarin)thiseveningd. takeone-halfofthemisseddoseofCoumadin(warfarin)rightnow

5. WhileonCoumadin(warfarin)you—a. shouldnoteatspinachb. caneatspinachonetimeamonthc. caneatasmuchspinachasyouwouldlikewheneveryouwouldliked. caneatspinachbutneedtoeatthesameamountregularlyeveryweek

6. Whileoutwithfriendsfordinner,youhavejustfinishedyourthirdglassofwine.Thisamountofalcoholconsumedinasingleeveningwill—a. causeadecreaseinyourINRb. causeanincreaseinyourINRc. notaffectyouoryourCoumadin(warfarin)inanywayd. makeyousickwhentakingCoumadin(warfarin)medication

7. Whileinyourpharmacy,younoticemultivitaminsareonsale.Aftersomethought,youdecidethatyoumayneedamultivitamin.Youwould—a. purchasethemultivitaminandbegintakingitregularlyb. nottakeamultivitaminbecauseitwillcauseabloodclotwhiletakingCoumadin(warfarin)c. starttakingitandbringthemultivitamintoyournextCoumadinClinicvisittoshowthepharmacistd. purchasethemultivitaminbutnotstarttakingituntilyoutalkedwiththepharmacistatyourCoumadinClinic

8. IfyouranoutofyourprescriptionforyourCoumadin(warfarin)youwould—a. borrowCoumadin(warfarin)fromafriend,aslongasitisthesamedoseasyoursb. callandaskforrefillsforthatdaysoyoudonotmissadoseofCoumadin(warfarin)c. waituntilyournextappointmentthatisjustafewdaysawaytogetanewprescriptiond. donothingbecauseyouhavetakenCoumadin(warfarin)longenough,otherwisetherewouldbemorerefillsonyourprescription

9. WhichofthefollowingisaneffectofCoumadin(warfarin)medicationthatwillmostlikelybeexperienced?a. Strokeb. LegClotc. Bruisingd. Bloodintheurine

10. Youhaveacold,whichincludesarunnynoseandcough.You—a. couldsafelytakeNyquiltohelpgetridoftherunnynoseandcoughb. takeyourfriend’smedicationthathe/sheusesforabadcoldbecausehe/sheisalsoonCoumadin(warfarin)medicationc. wouldcalltheCoumadinClinicandtellhim/heryouareonCoumadin(warfarin)medicationandaskwhatyoucantakeforyourcoldd. decideitissafertosufferthroughthecoldbecausemostcoldmedicationswillinteractwithyourCoumadin(warfarin)medication

11. WhenmakingadentalappointmentwhiletakingCoumadin(warfarin)medication,youneedtorememberyou—a. cannothaveproceduresdoneonyourteethwhiletakingCoumadin(warfarin)b. musttellyourdentistyouaretakingCoumadin(warfarin)wellinadvanceofhavinganyproceduredonec. canhaveproceduresdoneandthereisnotaneedtotellthedentistabouttheCoumadin(warfarin)d. canhavethedentalproceduredoneifwhenyouarriveatyourdentalappointmentyoutellthedentistyouaretakingCoumadin(warfarin)

12. Whentheneedarisestotakeanantibiotic(togetridofaninfection)whiletakingCoumadin(warfarin),youneedto—a. takehalfoftheprescribedlengthoftherapy,andthencalltheCoumadinclinicb. refusetotakeanynewmedicationbecauseyouaretakingCoumadin(warfarin)c. waituntilyournextCoumadinclinicvisitandthentellthepharmacistabouttheantibioticd. calltheCoumadinClinicrightawayandletthemknowyouarestartinganewmedication

13. Coumadin(warfarin)works—a. inmylivertomakemybloodthickerb. inmylivertomakemybloodthinnerc. inmykidneystomakemybloodthickerd. inmykidneystomakemybloodthinner

14. ThebesttimeofdayformetotakemyCoumadin(warfarin)is—a. atlunchtimeb. intheeveningc. inthemorningbeforebreakfastd. anytimeofdaywhenIremember

15. WhichofthefollowingisaneffectofmyCoumadin(warfarin)medicationthatIwillmostlikelyexperienceifmyINRistoohigh?a. Aclotinthelegb. Minorbleedingc. Clotinthelungd. Bleedinginthebrain

142 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

16. WhichofthefollowingdrinkscandecreasetheeffectivenessofyourCoumadin(warfarin)?a. Deans2%low-fatmilkb. Hershey’schocolateshakec. Tropicanaorangejuiced. Ensurenutritionalsupplement

17. WhiletakingCoumadin(warfarin),whichofthefollowingrepresentsasituationwhenyoushouldgototheemergencyroom?a. Youcoughupbloodb. Yournosebleedsslightlywhileblowingitc. Yourgumsbleedafterbrushingyourteeththenitstopsquicklyd. Youhavecutyourselfwhileshavingandyoucontrolthebleeding

18. Yourneighborbringsoverthisgreat“allnatural”herbalsupplementshejustboughtfromherchiropractor.Sheswearsthatthishelpsallherachesandpainsandrecom-mendsthatyoutakeitwhenyouache.Yourdecisionisto—

a. takeheradvice,realizingthatyoucouldusethisherbalsupplementb. starttakingtheherbalsupplementandtellyourpharmacistatthenextofficevisitc. askyourpharmacistiftheherbalsupplementwillinteractwithyourmedicationsbeforeyoutakeitd. avoidtakingherbalsupplementsaltogetherbecauseallmedicationsinteractwithCoumadin(warfarin)

19. OnceyouhavereachedastableCoumadin(warfarin)dose,aPT/INRbloodtest—a. shouldbecheckedonceayearb. shouldbecheckedonceevery3monthsc. shouldbecheckedatleastonceevery4weeksd. doesnotneedtobecheckedonceyouareonastableCoumadin(warfarin)dose

20. TheresultsofyourPT/INRtesttellsthepharmacist—a. howthickorthinyourbloodiswhiletakingCoumadin(warfarin)b. howwellyourkidneysareworkingsincetakingCoumadin(warfarin)c. whatyouraveragebloodsugarlevelwassincetakingCoumadin(warfarin)d. howmuchalcoholyouhavebeendrinkingsincetakingCoumadin(warfarin)

21. WhiletakingCoumadin(warfarin),youshouldcallyourCoumadinClinicwhenyouget:a. abackacheb. anupsetstomachc. atensionheadached. diarrheaformorethan1day

22. WhileonCoumadin(warfarin)youneedtoberoutinelymonitoredforwhichofthefollowing:a. PT/INRtestsb. Potassiumlevelsc. Bloodglucoselevelsd. Kidneyfunctiontests

23. WhichofthefollowingmayhaveasignificanteffectonhowwellyourCoumadin(warfarin)works?a. Changesinyourmoodb. Changesinsleephabitsc. Howmuchwateryourdrinkd. Usingoverthecountermedications

24. WhiletakingCoumadin(warfarin),whichofthefollowingshouldleadyoutotheemergencyroom?a. Lossofappetiteb. Brownloosestoolsc. Urinebecomesredincolord. Aquartersizebruiseonyourarm

25. WhichofthefollowingfoodscouldaffecthowwellyourCoumadin(warfarin)works?a. Celeryb. Carrotsc. Coleslawd. Greenbeans

26. YouhavegenericandbrandCoumadin(warfarin)tabletsathomethatareboththesamedose.Youshould—a. takebothbecausetheyworkdifferentlyb. takeonlybrandorgeneric,butnotbothc. nottakeeitheruntilyoucalltheCoumadinClinicd. alternatedaysbytakingbrandononedayandgenericonthenextday

27. OnceyourCoumadin(warfarin)isstopped,howlongdoesittaketogetthemedicationtogetoutofyoursystem?a. 5hoursb. 5daysc. 5weeksd. 5months

28. AfterstartingCoumadin(warfarin),howlong(inmonths/years)wouldyouexpecttobetakingCoumadin(warfarin)?a. 1yearb. 1monthc. Itdependsoneachperson’sneedsd. IfyoustartCoumadin(warfarin),youwillhavetobeonthemedicationfortherestofyourlife

29. WhichofthefollowingactivitiesaremoreriskywhiletakingCoumadin(warfarin)?a. Playingfootball,becauseyoucanhityourheadb. Takingabath,becausesoapinteractswithCoumadin(warfarin)c. Playingcardsbecauseusingyourhandsalotwillcauseabloodclotd. Walkingalot,becauseexerciseisnotgoodforyouwhiletakingCoumadin(warfarin)

APPEnDIx Anticoagulation Knowledge Assessment (AKA) Questionnairea (continued)

aCorrect answers are underlined. Items that were deemed relevant by the study investigators for inclusion in sensitivity analyses of knowledge with INR control are shaded in gray.INR = international normalized ratio; PT=prothrombin.

INR Goal Attainment and Oral Anticoagulation Knowledge of Patients Enrolled in an Anticoagulation Clinic in a Veterans Affairs Medical Center

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 143

•Nearly 60% of commercial payers reported having 1 or morestep-therapy (ST)programs in2010,making itoneof themostpopularpharmacybenefitmanagementtools.However,adoptionofSTisquicklyoutpacingunderstandingofitsclinical,human-istic,andeconomicoutcomes.

•Previousreviewsofdrugcostmanagement toolshavereporteddrug cost savings from ST, highlighted higher discontinuationrates, and identified gaps in the literature related to inclusionof quality of care metrics. Prior critical reviews have focusedprimarilyon the lackof inclusionof abroad rangeof relevantoutcomes rather than on study quality, particularly internalvalidity.

What is already known about this subject

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

Brenda R. Motheral, RPh, MBA, PhD

ABSTRACT

BACKGROUND: Adoption of step therapy (ST) is quickly outpacing the mar-ket’s understanding of its clinical, humanistic, and economic outcomes. The broad scope of previous reviews of drug management programs has prohibited an in-depth discussion of the ST literature specifically.

OBJECTIVE: To conduct a critical review of ST program evaluations, discuss their policy implications, and provide recommendations for future research.

METHODS: PubMed was searched for relevant English-language articles, and references of relevant articles were examined. The ST policy under evaluation had to require use of a first-line agent prior to coverage of a second-line agent.

RESULTS: Fourteen evaluations of ST programs have been published, 7 in commercial populations and 7 in Medicaid. Twelve of the studies empiri-cally examined claims data; 1 was a model; and 1 was limited to patient surveys. Five therapy classes, including antidepressants, antihyperten-sives, antipsychotics, nonsteroidal anti-inflammatory drugs (NSAIDs), and proton pump inhibitors (PPIs), have been evaluated. The research has con-sistently found statistically significant drug cost savings with the exception of antipsychotics, where rebates have frequently been excluded. Savings result from greater use of first-line medications and from reduced medica-tion initiation, with the magnitude of noninitiation varying across therapy classes. Three studies have examined medication adherence, producing mixed results. Five studies have empirically examined the effect of ST on hospitalization and emergency room utilization and costs, with none finding statistically significantly higher disease-related utilization or spend, outside of higher outpatient expenditures but not higher outpatient utilization in 1 study.

CONCLUSIONS: The research demonstrates that ST programs for therapy classes other than antipsychotics can provide significant drug savings through the greater use of lower-cost alternatives and, to a lesser extent, reduced drug utilization. The drug savings and clinical impact of ST for antipsychotics are unclear given the research conducted to date, but ST programs for NSAIDs and PPIs can provide significant drug savings without increasing use of other medical services. The research on ST shows gaps in the breadth of evaluation and methodological quality as well as possible study bias. Further research on ST is needed for other therapy classes and for the Medicare Part D population. Recommendations for other areas of research, needed methodological improvements, and reducing the potential for study bias are provided.

J Manag Care Pharm. 2011;17(2):143-55

Copyright © 2011, Academy of Managed Care Pharmacy. All rights reserved.

•TheresearchonSTshowsgapsinthebreadthofevaluationandmethodologicalqualityaswellaspotentialstudybias.

•Further researchonST isneeded fornumerous therapyclasseswhere ST is common and for theMedicare PartD population.ResearchisalsoneededtobetterunderstandtheimpactofSTontreatmentdiscontinuities,appropriatenessofuse,othermedicalcosts, andmember/provider satisfaction aswell as the effect ofalternativedesigns.

•Neededmethodological improvements include(a)useofappro-priate comparator groups, (b) examination of disease-relatedmedicalspending,(c)adjustmentformultiplestatisticaltests,and(d)bettercausallinkagebetweenprogramandoutcomes.

•Tohelpreducethepotentialforstudybias,independentlyfundedevaluationsandmandatorystudyregistrationprior to initiationshouldbeconsidered.

What this study adds

CONTEMPORARY SUBJECT

W ith the growing availability of generic and lower-costbrandalternatives,steptherapy(ST)hasgrownrapidlyinpopularityincommercial,Medicare,and

Medicaid settings as ameans to createbetter value from ris-ingpharmaceutical spend. ST requires amember to try1 ofthe first-linemedications,oftenagenericalternative,prior toreceivingcoverage for a second-lineagent,usually abrandedproduct.Nearly60%ofcommercialpayersreportedhaving1ormoreSTprogramsin2010,nowmakingitoneofthemostpopularmanagementtools.1Ina2007reportonMedicaredrugbenefits,7ofthetop10largestPartDplanshad1ormoreST

144 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

programs,withtheaveragebeing6programs.2However,adop-tionofST isquicklyoutpacingdecisionmakers’understand-ing of the clinical, humanistic, and economic value of theseprograms.Suchknowledgeisneededtoavoidpotentialunin-tended consequences, such as medication noncompliance.3,4

Conversely, concerns over unintended consequences canpreventplansponsors fromadoptingprograms thatevidencelater shows tobe robust, resulting inwasteddollarswithnoincrementalhealthbenefit.While systematic reviewsofdrugcostmanagement programs have been conducted previously,the broad scope of these reviews has prohibited an in-depthdiscussionof theST literature specifically.5-7Accordingly, thepurposeofthispaperistoconductacriticalreviewofSTpro-gramevaluationsand toprovide recommendations for futureresearch.

■■  MethodsTo identify potential studies for inclusion, PubMed wassearchedforrelevantEnglish-languagearticlesusingthesesub-jectterms:steptherapy,priorauthorization(PA),drugpolicy,andpharmaceuticalpolicy.Thereferencelistsofknownstud-iesandreviewarticleswereexaminedtofindarticlesaswell.Tobeconsideredeligible forreview, thepolicyunderevalua-tionhad tobe implemented in theUnitedStatesandrequireuseofafirst-lineagentpriortocoverageofasecond-lineagent,ahallmarkcharacteristicofSTthatdistinguishes it fromtra-ditionalPA.However,STprogramstypicallyallowpatientstoseek coverage for the second-line agentwithout having useda first-linemedication,utilizing amedical exceptionprocess;thus,STandPAprogramsoverlapintheirdesigns.Threestud-iesthatexaminedpolicychangesacrossmultipleMedicaidpro-gramsbutdidnotseparateSTfromPAprogramsintheresultswereexcludedfromthereview.8-10

■■  ResultsFourteenevaluationsofSTprogramshavebeenpublished todate.11-24 Seven of the studieswere conducted in commercialpopulations(Table1),while7assessedMedicaidSTprograms(Table2).AnanalysisbyPanzeretal.(2005)wasaneconomicmodelofanantidepressantSTprogramforpatientswithanxi-ety,15andCoxetal.(2004)surveyedmemberswithanSTedit/rejecttoassesstheirresponsestotheedit.17Theremaining12studies were retrospective, claim-based analyses, 7 of whichexamineddrugexpendituresonly, and5ofwhichexamineddrug and medical expenditures. Therapy classes examinedinclude antidepressants, antihypertensives, antipsychotics,nonsteroidal anti-inflammatory drugs (NSAIDs), and protonpumpinhibitors(PPIs).

Antipsychotic ST in Medicaid has been the most heavilystudied subject.Researchers atHarvardpublisheda seriesofcloselyrelatedpapersexaminingtheeffectofSTprogramsforantipsychoticsinMaine.18-20Althoughthetitlesofthese3stud-

iesreferredtothemasPAprograms,theyactuallymetthecri-teriaforSToutlinedpreviously.18Forpatientsnewtotherapywith selected second-generation antipsychotics, prescribershadtoprovideevidencethatapatienthadnotbeenadequatelycontrolled by preferred agents. Anticonvulsants were placedonPA.Luetal.(2010)foundthatdrugtreatmentinitiationforbipolar illnessdecreasedby32%after implementationof theST and PAprograms inMaine, primarily due to a reductionin the use of nonpreferred agents without a correspondingincrease in theuse of preferred agents.18 Zhang et al. (2009)foundthatthesameprogramresultedinanaverage$27reduc-tionperpatientinpharmacyreimbursementforbipolardisor-derduringthe8-monthpolicyperiod.19However,thehazardrate of treatment discontinuation was 2.28 times as high inthepost-policyperiodthaninthepre-policyperiod,suggest-ing that the savings resulted from treatment discontinuationrather than switching to lower-cost alternatives. Soumerai etal. (2008) examined the impact of the same policy on anti-psychotic userswith a diagnosis of schizophrenia.20 Patientsinitiating atypical antipsychotics during the STprogramhada29%greaterriskoftreatmentdiscontinuity(i.e.,gap,switch,or augmentation) than patients initiating before the ST pro-gram, whereas no change was observed in the comparisongroupoverthesametimeperiod.Theauthorsdidnotobserveantipsychoticdrugsavingsbutacknowledgedthat theycouldnotaccountforincreasedrebatesfrompharmaceuticalmanu-facturers.Lawetal.(2008)alsoexaminedtheeffectofanSTprogram for selected second-generation antipsychotic agentsinTexasandWestVirginia.21Bothstatesgrandfatheredpriorusersandrequiredatrialwithapreferredagentbeforecover-age of a nonpreferred agent. Both states observed reductionsinthemarketshareofnonpreferredantipsychoticsbutneitherstatedemonstratedsignificantsavingsinpharmacyspend,notfactoringinmanufacturerrebates.

Incontrast,Farleyetal. (2008) foundabout$7million indrugcostsavingsforGeorgiaMedicaidafterimplementingSTfor atypical antipsychoticmedications comparedwith a stateMedicaid program without ST. Georgia Medicaid requiredpatientstotry2typicalantipsychoticsbeforereceivingcover-ageforanatypicalantipsychotic.ThestudybyFarleyetal. isthe only analysis ofmedical expenditures following an anti-psychotic ST program.22 The authors did not find increasesin disease-related inpatient or long-term care expendituresbutdidfindanincreaseindisease-relatedoutpatientmedicalexpenditures($32permemberpermonth[PMPM]).Tofurtherunderstandtheoutpatientcostfinding,theauthorsexamineddisease-relatedoutpatientutilizationandfoundnosignificantincreaseintheSTgroupcomparedwiththenon-STgroup.22

Forothertherapyclasses,researchhas foundthatSTpro-ducesdrugsavings.Threestudieshaveexaminedantidepres-santST.11,14,15Dunnetal.(2006)founda9.0%lowerdrugcostper day ($0.36 PMPM) for patients in an antidepressant ST

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 145

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

TABLE 1 Published Evaluations of Step-Therapy Programs in Commercial Populations

Authors / Source / Intervention Study Groups Design and Outcomes Key Findings and Limitations Sponsor

AntidepressantsMarketal.(2010)11

ThomsonReutersGenericantidepressant requiredbeforecoverageofbrand(nootherdetailsprovided)Priorusersweregrandfathered

Antidepressantusersinemployerplans,youngerthan65yearsofageand3.75yearsofcontinuouseligibility

Pre/PostwithcomparisongroupofemployerswithnoSTPrimaryoutcomes:antidepressantprescriptionuseandall-causemedicalutilizationandspending

•23.6%greatergenericuseand~18%lowerbrandusebyfifthquarterafterimplementation

•Nodifferenceinantidepressantdayssupplyornumberofprescriptionsafter4quarters

•AntidepressantRxsavingsnotassessed•Increaseinall-causeoutpatientoffice,inpatient,andERvis-itsandspending

•Nodifferenceinmentalhealth-relatedutilizationThekeylimitationsinclude:•Lackofreportingofbaselineantidepressantutilizationorall-causespendingforthe2groups

•Reportingofnonsignificantdifferencesinmentalhealth-relatedutilizationassignificantintheabstract

•Noexaminationofmentalhealth-relatedexpenditures

Pfizer

Dunnetal.(2006)14

IntermountainHealthcareGenericantidepressant(excludingTCAs)requiredbeforecoverageofbrand;waiveroffirstgenericcopayment($5–$10)Priorusersweregrandfathered

Antidepressantusersina440,000- memberHMOcomparedwithPBMwithoutST

Pre/Postwithcomparison,compared12monthsbeforeandafterPrimaryoutcomes:antidepressantgenericdispensing,utilization,andspending

•Genericdispensingwas12.6percentagepointsgreaterinyearfollowingimplementation

•Nodifferenceindaysoftherapybetweeninterventionandcomparisongroup

•Antidepressantdrugcostperdaywas9.0%loweror$0.36PMPMforinterventiongroup

Thekeylimitationsinclude:•Exclusivefocusondrugutilizationandsavings•Implementationofadose-optimizationprogramduringthestudymayhaveoverestimatedsavings

No externalfunding

Panzeretal.(2005)15

BlueCrossBlueShieldofDelawareGenericSSRIrequiredbeforecoverageofbrandSSRIforpatientswithanxiety

Commercialhealthplanmembersusingantide-pressantsforanxiety

MarkovModelPrimaryoutcomes:lengthoftherapy,changeintherapy,antidepressantdrugcost,andall-causemedicalspending

•4.5%greaterfrequencyoftherapychangeinST•4.5%lowerfrequencyofcontinuoustherapyforatleast6months

•Antidepressantdrugsavingsof$0.26PMPM•All-causemedicalcostsincreased$0.32PMPMThekeylimitationsinclude:•Useofacross-sectionalstudyofcomplianceandmedicalcostsasamodelinputtomakecausalconclusionsaboutST

Glaxo-SmithKline

ARBs/ACE InhibitorsMarketal.(2009)12

ThomsonReuters130-daytrialwithpreferredACEinhibitororARBbeforecoverageofnonpreferredACEinhibitororARBPriorusersweregrandfathered

ARB/ACEinhibitorusersinemployerplans,youngerthan65yearsofageand3.75yearsofcontinuouseligibility

Pre/PostwithcomparisonPrimaryoutcomes:antihypertensiveprescriptionuseandall-causemedicalutilizationandspending

•Nodifferenceindayssupplyornumberofprescriptionsafter5quarters

•Higherdiscontinuationrate(13%vs.10%unadjusted)•Higherall-causeinpatient,ERandoutpatientvisitsandspending

•Didnotreportbaselineantihypertensiveutilizationorall-causespendingforthegroups

Thekeylimitationsinclude:•Noexaminationofcardiovascular-relatedutilizationandexpenditures

•Noreportingofbaselineantihypertensiveutilizationorall-causespendingforthe2groups

Pfizer

Yokoyamaetal.(2007)13

WellpointNextRxCoverageofARBifclaimforACEinhibitor,ACEinhibitor/HCTZ,ARB,orARB/HCTZinprior3monthsPriorusersweregrandfathered

ARB/ACEinhibitorusersin3healthplanswith15monthscontinuouseligibility

Pre/PostwithcomparisonPrimaryoutcomes:initiationofdrug,editresponse,andswitching

•AttemptedARBinitiationwas19.2%forinterventionand26%forcomparisongroup

•OfthosewithARBreject,45%receivedARB,49%receivedotherRx,7%receivednomedication

•24.0%ofpatientsintheinterventiongroupwhoappliedforanARBbutweregivenACEinhibitormonotherapyswitchedtooraddedanARBwithin12months,comparedwith6.1%ofthosestartedinitiallyonACEinhibitor

•Savingsof16.9%or$0.03PMPMThekeylimitationsinclude:•Questionablecomparabilityofthecontrolgroup,whichmayhaveledtounderestimationofthesentineleffectand,correspondingly,drugsavings

•Exclusivefocusondrugutilizationandsavings

Novartis

146 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

theGeorgiaMedicaidpopulation,Delateetal.(2005)founda$1.70PMPM,ornearly50%,decreaseinantisecretoryexpendi-turesfromimplementingSTwithoutgrandfatheringforPPIs.23 Smalleyetal.(1995)foundadecreaseof53%inexpendituresforNSAIDsafterimplementationofSTforbrandNSAIDswith-outgrandfatheringinTennesseeMedicaid.24Accordingly,otherthan studies of antipsychotics within Medicaid populations,the only studies that have not found sustained drug savingsfromSTarethosethatexaminedtotalprescriptiondrugcostsfor patients enrolled in ST rather than disease-specific drugcosts.11,12

Research shows thatdrug cost savings result fromgreateruse of first-line medications and also from less medicationuseasevidencedintheclaimsdata.Usingpatientself-report,Motheraletal.foundthat17%ofpatientsreceivednomedica-tion after the ST edit, and16%paid full price for the brandmedication.16 Using similarmethodology for PPI andNSAIDusers,Coxetal.foundthat11%ofpatientsreceivednomedi-cation, 8%purchased an over-the-counter (OTC) alternative,and11%receivedasampleforthebranddrugfromtheirphy-sicians.17 Inbothstudies, thepercentageofpatientsreceivingnomedicationwashighestforPPIs.Yokoyamaetal.foundthat7%ofpatientsdidnothaveanantihypertensiveclaimin the12monthsfollowingtheSTedit.13Delateetal.foundthat22%ofMedicaidpatientswithanSTeditforaPPIdidnothaveany

programversusacomparisongroup,14andPanzeretal.’smodelsimulateddrugsavingsof$0.26PMPMforantidepressantST.15 Alsolookingatantidepressants,Marketal.(2010)reportedaninitialsavingsof1.7%acrosstotalprescriptiondrugcosts(notjust antidepressant spend)but found that the savingsdimin-ishedovertime.11Theauthorsdidnotreportexpendituresforantidepressantmedications.

Two evaluations of antihypertensive ST have been pub-lished.12,13 Yokoyama et al. (2007) found a 12.8% lower costper day for patients in an angiotensin II receptor blockers(ARB) ST program versus a comparison group.13 An accom-panying editorial in JMCP questioned whether Yokoyama etal. underestimated the drug savings from the sentinel effect,since the rate of initiation of either angiotensin-convertingenzyme(ACE)inhibitororARBtherapywas2.4timeshigherinthecomparisongroup.25Marketal.(2009)examined total prescriptiondrugexpendituresforantihypertensiveusers(i.e.,notjustantihypertensivespend)andreportedthatoverallpre-scriptiondrugspendinginitiallydeclined3.1%afterprogramimplementation,butthesavingsdiddiminishsomewhatovertime.12 Theauthorsdidnotreportchangesinantihypertensivedrugspending.

Motheraletal. (2004) founddrugsavingsof$0.93PMPMfollowingimplementationofanSTprogramforPPIs,NSAIDs,and selective serotonin reuptake inhibitors (SSRIs).16 Within

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

Authors / Source / Intervention Study Groups Design and Outcomes Key Findings and Limitations Sponsor

Multiple Therapy ClassesMotheraletal.(2004)16

ExpressScriptsH2-blockerusebeforecoverageofaPPI;2trialsofgenericNSAIDSbeforecoverageofbrandNSAID;fluoxetineorfluvoxaminemaleateusebeforecoverageofbrandSSRIPriorusersweregrandfathered

1employergroup comparedwithrandomsampleofusersfromPBMwithoutSTSurveyincludedpatientswithSTedit

Pre/Postwithcompari-songroupforclaimsanalysisDescriptive,post-onlyforpatientsurveyPrimaryoutcomes:drugsavingsforPPIs,antidepressants,andNSAIDs,responsetotheedit,andpatientsatisfaction

•Savingsof$0.93PMPMacrossthe3classes•Frompatientself-report,29%switchedtogeneric;23%hadaPAforthebrand,16%paidfullpriceforthebrand,17%receivednomedication,and15%hadotherresponses,suchasOTCpurchase

•Patientsatisfactionlowerfornonusersandthosepayingfullprice

Thekeylimitationsinclude:•Only1employerrepresentedintreatmentarm•33%responseratetosurveywithsmallsamplesizes•Noexaminationofclinicalandothereconomicoutcomes

No externalfunding

Coxetal.(2004)17

ExpressScriptsH2-blockerusebeforecoverageofaPPI;2trialsofgenericNSAIDSbeforecoverageofCOX-2inhibitorPriorusersweregrandfathered

Membersfrom1HMOwithSTeditforPPIsorNSAIDs

Descriptive.Initialsurveysenttothosewitheditsandfollow-upsurveysenttononutilizersPrimaryoutcomes:responsetotheeditandreasonsfornon-utilization

•44%receiveddifferentdrugthanprescribed,15%receivedPAforbrand,11%receivednomedication,11%paidfullpriceforbrand,8%purchasedOTC,and11%hadotherresponses,suchasobtainingsamples

•Amongnonusers,32%saidtheyactuallyreceivedmedica-tionatalaterdate;16%citedlackofwillingnesstopayasthereasonfornonuse,and28%citedaffordability

Thekeylimitationsinclude:•23%responseratetosurveywithsmallsamplesizes•Potentialresponsebiasthatunderestimatednonuse•Noexaminationofclinicalandothereconomicoutcomes

No externalfunding

ACE = angiotensin-converting enzyme; ARB = angiotensin II receptor blocker; COX = cyclooxygenase; ER = emergency room; H2-blocker = histamine-2-receptor antagonist; HCTZ = hydrochlorothiazide; HMO = health maintenance organization; NSAID = nonsteroidal anti-inflammatory drug; OTC=over-the-counter; PA = prior authorization; PBM = pharmacy benefits management company; PMPM = per member per month; PPI = proton pump inhibitor; Rx = prescription; SSRI = selective serotonin reuptake inhibitor; ST = step therapy; TCA = tricyclic antidepressant.

TABLE 1 Published Evaluations of Step-Therapy Programs in Commercial Populations (continued)

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 147

TABLE 2 Published Evaluations of Step-Therapy Programs in Medicaid Populations

Authors / Source / Intervention Study Groups Design and Outcomes Key Findings Sponsor

Antipsychotics (and Anticonvulsants for Selected Studies)

Luetal.(2010)18

HarvardMedicalSchoolCoverageofnonpreferreddrug(olanzapineoraripiprazole)aftertrialofrisperidoneandsubsequenttrialofeitherziprasidoneorquetiapineused atfulldosefor2weeksNonpreferredanticonvulsants(lamotrigine,topiramate,gabapentin,brandcarbamazepine,brandvalproicacid,oxcarbazepine,levetiracetam)requiredaPAPriorusersweregrandfathered

MaineandNewJersey(comparison)Medicaidmembersaged18yearsorolderandcontinuouslyenrolledfrom2001to2004withbipolardisorder

Pre/Postwithcomparisonstate withnoSTPrimaryoutcomes:bipolarRx(antipsychoticsandanticonvulsants)initiationandswitching

•DecreaseinRxinitiationof32%by4monthsafterSTimplementation,representing50fewerinitiationsper10,000patientspermonth

•Reductionininitiationofnonpreferredagentsbutnocorrespondingincreaseinuseofpreferredagents

•Nodifferenceinswitchingpre-andpost-changeorbetweentreatmentandcomparisongroups

Thekeylimitationsinclude:•Noexaminationofclinicalandeconomicoutcomes(duetoshortpolicyduration)

•SignificantoverlapwithZhangetal.—unclearwhyitwaspublishedasasecondmanuscript

RobertWoodJohnson

Zhangetal.(2009)19

HarvardMedicalSchoolSamepolicyinterventionas Luetal.(2010)

MaineandNewJersey(comparison)Medicaidmembers,aged18yearsorolderandcontinuouslyenrolledfrom2001to2004withbipolardisorder

Pre/Postwithcomparisonstate withnoSTPrimaryoutcomes:bipolarRxuse(antipsychoticsandanticonvulsants)andspending,discontinuation,andswitching

•8percentagepointreductioninuseofnonpreferredagentsover8-monthpolicyperiod

•Noincreaseinuseofpreferredagents•Noincreaseinratesofswitching•Hazardratiooftreatmentdiscontinuationwas2.28timesashighaftertheSTimplementation,adjustingforcomparisongrouptrends

•Drugsavingsof$27perpatientover8monthsresultedfromtreatmentdiscontinuationsratherthanswitchestopreferredagentsforbipolarpatients

Thekeylimitationsinclude:•Lackofinclusionofrebatesindrugsavings•Noexaminationofclinicalandeconomicoutcomes(duetoshortpolicyduration)

RobertWoodJohnsonandothers

Soumeraietal.(2008)20

HarvardMedicalSchoolSamepolicyinterventionas Luetal.(2010)

MaineandNewJersey(comparison)Medicaidmembers,aged18-63yearsandcontinu-ouslyenrolledfrom2001to2004withschizophrenia

Pre/Postwithcomparisonstate withnoSTPrimaryoutcomes:schizophreniaRx(antipsychotics)spendinganddiscontinuities

•Maineantipsychoticuserswithschizophreniahada29%greaterriskofagap,switch,oraugmentationintherapy,inpost-changeperiod(morethantwo-thirdsweregaps)

•Nochangeindrugspendobserved,relativetocomparisongroup,withoutfactoringinrebates

Thekeylimitationsinclude:•Lackofinclusionofrebatesindrugsavings•Noexaminationofclinicalandeconomicoutcomes(duetoshortpolicyduration)

AHRQ andothers

Lawetal.(2008)21

HarvardMedicalSchool14-daytrialofpreferredagentbeforecoverageofaripiprazole,brandclozapine,fluoxetine-olanzapine,orolanzapine (WestVirginia)Trialofpreferredagentbeforecoverageofbrandclozapine,fluoxetine-olanzapine,or olanazpine(Texas)

WestVirginiaandTexasMedicaidcomparedwith38otherstateMedicaidprograms

Pre/Postwithcom-parisontostateswithnoSTPrimaryoutcomes:antipsychoticutilizationandspend

•Marketshareofnonpreferredantipsychoticsdecreased3.5%immediatelyand13.9%after2yearsinWestVirginia.NostatisticallysignificantdifferencewasobservedforTexasorthecomparisonstates

•Neitherstateshowedsignificancedecreasesinoverallantipsychoticdrugspend,withoutincludingrebates,relativetothecomparisongroups

Thekeylimitationsinclude:•Lackofinclusionofrebatesindrugsavings•Noexaminationofclinicalandeconomicoutcomes

Multiple

Farleyetal.(2008)22

UniversityofNorthCarolinaNonresponseto2typicalantipsychoticsbeforecoverage ofanatypicalantipsychotic

GeorgiaandMississippi(control)Medicaidmemberswhowerenotdual-eligibleandsubsetofcontinu-ouslyenrolledschizophreniapatients

Pre/PostwithcomparisontostatewithnoSTPrimaryoutcomes:antipsychoticandotherhealthservicespending

•Savingsof$7millioninantipsychoticdrugspendingover11months

•Patientswithschizophreniahaddrugsavingsof$17.42PMPMandincreasedoutpatientexpendituresof$31.59,butfollow-upanalysisdidnotfinddifferencesinoutpatientutilizationforschizophreniapatients

•NodifferencesininpatientorERexpenditureswereobservedforschizophreniapatients

Thekeylimitationsinclude:•Noexaminationofmedicationinitiationordiscontinuation•Noexaminationofpotentialcausallinkbetweendrugutilizationandoutpatientexpendituresatthepatientlevel

Pfizer

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

148 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

augmentation with an ARB was 6.1%, similar to that of thecomparisongroup.13

Otherthanstudiesofantipsychotics,4studieshaveempiri-callyexaminedtheeffectofSToninpatient,emergencyroom(ER),andoutpatientmeasures.Marketal.reportedhigherall-cause inpatient, ER, and outpatient expenditures after imple-mentationofanSTprogramforantidepressants,adjustingforbaselinedifferencesincostsanddemographics.11However,theauthorsfailedtoreportbaselinemedicalutilizationorspendingtoallow forassessmentof thecomparabilityof the2groups.Calculationofunadjustedaverageall-causemedicalandphar-macycostsshowedanaverage$1,967annualcostperpatientforSTplansintheyearafterimplementationofST(2006),6.9%lower than the average for the comparison planswithout STfor antidepressants. The authors also reported nonsignificantincreases asmeaningful differences formental health-relatedutilization.Specifically,mentalhealth-relatedinpatientadmis-sionshadageneralizedestimatingequation(GEE)coefficientof0.184andPvalueof0.15,andmentalhealth-relatedERvis-itshadaGEEcoefficientof0.185andPvalueof0.13(Table4inthestudyreport),buttheabstractreportedthat“overallandmental health-specific inpatient and ER utilization and costsincreased.”11Astocosts,despitethementionofincreasedmen-talhealthcostsinthestudyabstract,nomultivariatemodelofmentalhealth-relatedmedical expenditureswas found in thearticle, only amodel for all-cause expenditures (Table 5 andFigure1 inthestudyreport).Finally,duringthetimeperiodofthestudybyMarketal.,11manyemployersandhealthplans

subsequentantisecretoryclaim.23Inbothstudies,theextenttowhichthepatientsreceivedsamples/OTCsorpaidfullpriceforthesecond-linebrandmedicationisunknown.

Other than studies of the atypical antipsychotics, only3 studies have empirically assessed the impact of ST onlonger-term medication adherence. Mark et al. found a 3percentage point higher rate of antihypertensive medication discontinuation in an ST group (13%) versus a comparisongroup(10%).12Amongantihypertensiveusers,days supplyofantihypertensiveswas 7.9% lower for the ST group than thecomparison group immediately after ST implementation, butthis difference disappeared by 5 quarters after implementa-tion.12 Interestingly,thesamepatternoccurredfortotalnumberofprescriptionsperantihypertensiveuser.Forantidepressants,Marketal.found3.9%lowerdayssupplyimmediatelyfollow-ingSTimplementation,butby4quartersafterimplementation,the antidepressant days supply was higher in the ST groupthaninthecomparisonplanswithoutSTforantidepressants.11

Dunnetal.alsofoundnosignificantdifferencesindayssupplyofantidepressantsfollowingimplementationofST.14

In terms of switching or augmentation, Yokoyama et al.found that 24.0% of patients in the intervention group whoapplied for an ARB but were given ACE inhibitor mono-therapy switched to or added an ARB within 12 months, ahigher switchpercentage thanwasobserved forpatients ini-tially started on an ACE inhibitor in the comparison group(7.2%).However, for all the intervention group patientswhoinitiatedanACEinhibitor, the12-monthrateofswitchingor

Authors / Source / Intervention Study Groups Design and Outcomes Key Findings Sponsor

Delateetal.(2005)23

ExpressScripts2-weektrialofH2-blocker(notautomated)beforecoverageofpreferredPPIforpatientswithnonerosivehypersecretoryconditionsNonresponsetopreferredPPI(lansoprazole,omeprazole,pantoprazole)beforecoverageofnonpreferredPPI(esomeprazole,rabeprazole)

AllGeorgiaMedicaidenrolleesandsubsetofantisecretoryusers

Pre/PostPrimaryoutcomes: drugexpendituresandGI-relatedmedicalexpenditures

•Antisecretoryspendingdecreased49.9%or$1.70PMPMin12monthsafterimplementation

•48%ofnonusersafterstepedithadaGI-relateddiagnosisversus81%ofPPIuserswhoreceivedpriorauthorization

•NodifferenceinGI-relatedmedicalexpendituresforPPI,H2-blocker,andnon-users

Thekeylimitationsinclude:•Noexternalcontrolgroupfordrugcostcomparison

No externalfunding

Smalleyetal.(1995)24

VanderbiltUniversitySchoolofMedicineArthriticorotherconnective-tissuedisorderandrecenttrialofgenericNSAIDforatleast2months(notautomated)beforecoverageofanybrandNSAID

TennesseeMedicaidenrolleesandsubsetofNSAIDuserscontinuouslyenrolledfor36months

Pre/PostPrimaryoutcomes: drugspendingandothermedicalutilizationformusculoskeletaldisorders

•MeanNSAIDdrugexpendituresdecreased53%duringthe2yearsfollowingthepolicychange

•SavingsresultedfromgreatergenericNSAIDuseandareductioninoverallNSAIDuse(19%)

•NoincreaseinmusculoskeletalexpendituresforothermedicalcarewasobservedacrosstheentirepopulationorforregularusersofbrandNSAIDs

Thekeylimitationsinclude:•Noexternalcontrolgroupfordrugcostcomparison

AgencyforHealthCarePolicyResearchandFDA

AHRQ = Agency for Healthcare Research and Quality; ER = emergency room; FDA = U.S. Food and Drug Administration; GI = gastrointestinal; H2-blocker = histamine-2-receptor antagonist (H2RA); NSAID = nonsteroidal anti-inflammatory drug; PA = prior authorization; PMPM = per member per month; PPI = proton pump inhibitor; Rx = prescription; ST = step therapy.

TABLE 2 Published Evaluations of Step-Therapy Programs in Medicaid Populations (continued)

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 149

wereactivelyadoptingnewstrategies to improveHealthPlanEmployerDataandInformationSet(HEDIS)scoresfordepres-siontreatment,includinginitiationofantidepressanttherapy.26 Theauthorsdidnotindicatewhetherthetreatmentorcompari-songrouphadsuchprogramsinplaceduringthistimeperiodthatcouldhaveaffectedthestudyresults.

InastudywithsimilardesignforantihypertensiveST,Marketal.didnotfinddifferencesinall-causeinpatientoroutpatientexpenditures but did find higher all-cause inpatient, outpa-tient,andERutilization.12Theauthorsdidnotexamineeitherdisease-related (cardiovascular)utilizationor expenditures inthisearlierstudy.

In Georgia Medicaid, Delate et al. found that enrolleeswhoreceivedahistamine-2receptorantagonist(H2RA)ornoantisecretory drug following an ST edit were nomore likelyto incur greater gastrointestinal (GI)-related or total medicalcare expenditures than were enrollees who received a PPI.23 NonusersfollowingaPPIPAeditwerelesslikelythanpatientswhoreceivedasecond-linePPImedicationtohaveanyGIdiag-nosis(48%vs.81%,respectively,P <0.001),suggestingatleastsome channeling of patients to the appropriate alternative.23 Smalley et al. found no increase in musculoskeletal-relatedMedicaidexpenditures fornondrug services following imple-mentationofanSTprogramforbrandNSAIDs.24Neitherstudyhad an external control group, but both used within-groupcomparatorstoaccountforunderlyingmarkettrends.

Panzer et al. used a model to simulate medical costs forantidepressant ST for patients with anxiety, with or withoutdepression.15Theprobability that eachpatientwould experi-encecontinuoustreatmentfor180daysorlonger,discontinuetreatment early, or have a therapy change (defined as eithera switch or augmentation) was determined from previousdescriptive literature (which the authors couldnot access forreview—1studybeingpublishedinamagazinenotMedline-indexed and the other as an abstract only). Drug use wasthenassignedtoeachpatientbasedonapreviousstudy,withpatientshaving less than180daysof therapybeingassigned$5,223 inmedical costs, thosewith180days ormore beingassigned $4,102 in costs, and those with a therapy changebeingassigned$6,741inmedicalcosts.

The model simulated a net $0.06 PMPM increase in all-cause totalmedical costs.Thekey limitationof thismodel isthe choice of study used for the medical cost assumptions,which was a cross-sectional examination of the associationbetween drug use patterns for antidepressants and medicalcosts. Results from such cross-sectional studies cannot betranslated into causal conclusions about interventions dueto thepresenceof thehealthyadherereffect, the tendencyofpeoplewhoareadherenttotheirmedicationstoalsoengageinotherhealthybehaviors,suchasexercisingregularlyandeat-ingahealthydiet.27Thepresenceofthehealthyadherereffectleads to cross-sectional studies showing that higher rates of

medicationadherenceareassociatedwithbetteroutcomesandlowerhealthcarecostswitheffectsizes farbeyondwhatevi-dencefromrandomizedcontrolledtrials(RCTs)wouldsuggest.Thus, theattempttomakecausal inferencesaboutSTfromacross-sectionalstudyofcomplianceandmedicalcostsisafatalflawandprohibitsmakinganyconclusionsfromthemodelbyPanzeretal.

Finally, the number of studies of humanistic outcomes,suchaspatientsatisfaction,isalsoquitelimited.AcrossPPIs,SSRIs,andNSAIDs,Motheraletal.foundthatcomparedwithreceiptof ageneric,payingoutofpocket for thebranddrug(oddsratioof0.25;P <0.05)andreceivingnomedication(oddsratioof0.12;P <0.01)wereassociatedwithsignificantlylowersatisfaction with the pharmacy benefit.16 Cox et al. did notfind a statistically significant relationship between outcomeof theSTeditandpharmacybenefit satisfactionbutdid findthatpatientswhoreceivedacoveredmedicationotherthanthebrandwere less satisfiedwithmedication they received thanthosewhoreceivedthebranddrug(P <0.001).17

■■  DiscussionThe research demonstrates that ST programs for therapyclassesotherthanantipsychoticscanprovidesignificantdrugsavingsthroughthegreateruseoflower-costalternativesandtoalesserextent,reduceddrugutilization.Asexpected,pro-gramsthatdonotgrandfatherhaveshownthelargestsavings,buteliminationofgrandfatheringwillnotbeappropriateforalltherapyclassesandwill risk increasedmemberandproviderdissatisfaction. While substitution with lower-cost alterna-tivesistheprimaryobjectiveofanSTprogram,reduceddrugutilization isnotan intendedgoal, theexceptionbeing thosetherapy classeswhere appropriateOTC substitutes are avail-able.Dependingonthetherapyclass,researchhasfoundthatbetween7%and22%ofpatientshavenoprescription claimsubmittedtotheirinsuranceproviderfollowinganSTedit.13,23

SomenonutilizationintheprescriptionclaimsactuallyreflectsOTC purchases or other responses that may be clinicallyappropriate.16,17 However, survey research has shown that asmallpercentageofnonutilizationremainsevenafteraccount-ing for these other behaviors, at least for the therapy classesstudiedtodate,includingNSAIDs,PPIs,andantidepressants.Inthefirst2classes,nonutilizationmayhavelittleifanyclini-cal impactonthepatientgiventhelesssevereindicationsforwhichthesedrugsaresometimesused,andtheextentoftruenonutilizationhasnotbeenexaminedfortherapyclasses,theexception being antipsychotics, where the potential clinicalimpactmaybemoresignificant.

At least 1 pharmacy benefitmanagement company (PBM)hasdeveloped aprogram to reduce the likelihoodof nonuseafter ST intervention. This program identifies patients whohave not filled a prescription claim within 2 days follow-ing their ST edit and then notifies the patient and provider

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

150 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

abouttheSTprogram,explainingmedicationalternativesandthe steps for obtaining a PA.28 Based on unpublished RCTs,the addition of a follow-up letter has improved medication initiationrates,increaseduseofgenerics,andreducedPAsforbrand medications.29 Notably, the Centers for Medicare andMedicaidServices(CMS)requiresasimilarnoticetobemailedwithin 3 business days to eachMedicare Part D beneficiarywhoreceivesa1-timetransitionfillforadrugsubjecttoST,PA,ornonformularyedits.30Greateradoptionofthistypeoftool,forpatientswithandwithouttransitionfills,wouldlikelyhelptoimproveappropriateutilizationassociatedwithSTprogramsandincreasemembersatisfaction.

Forantipsychotics, the inconsistent findings fordrugsav-ingsacrossstudieslikelyreflectsthedifferentprogramdesignsandthelackofaccountingforrebateswhererelevant.GeorgiaMedicaidfoundsignificantsavingsforaprogramthatrequireduseof2 typicalantipsychotics,whichareavailable ingener-icsandarefarlessexpensivethanatypicals.Theotherstud-iesallhad1ormoreatypicalantipsychoticon thepreferreddrug list, reducing the potential savings, particularly whensavings from rebates are not included. In addition, unlikeother antipsychotic programs evaluated, Georgia Medicaiddidnotgrandfatherpriorusers,whichalsowouldhavecon-tributedtothegreaterobservedsavings.AntipsychoticSTinMedicaid enrollees with bipolar disorder or schizophreniahas been associated with reductions in treatment initiationand continuation, unintended consequences that may havecontributedtothewithdrawalofthesepoliciesinMaineandGeorgia.Inthesestudies,risperidonewasthepreferredatypi-cal antipsychotic, except for Georgia which had no atypicalon the preferred list.18-22 Each atypical medication’s efficacyandsideeffectprofilemayormaynotmakeitagoodchoicefor initial therapy in anyonepatient,31perhaps contributingto the reduced initiation observed in Maine. However, theextenttowhichtheobservedtreatmentdiscontinuitiesresultin increased use of other medical services is unclear, sinceFarleyetal. foundno increase in inpatientexpenditures,ERexpenditures,oroutpatientutilizationforanSTprogramthatconsidered typicals as first-line and appeared not to grand-father prior users.22 Furthermore, in the Vermont Medicaidprogram,rescissionofaPAexemptionfornewusersofanti-psychotics,antidepressants,andanxiolytics/sedativeswasnotfollowedbydecreasedutilizationofmedicationsonthePAlist,andmentalhealth-relatedhospitalizationsdeclinedfollowingremoval of the PA exemption.32 As none of the studies thatfound increased treatment discontinuities examined othermedicalutilizationorclinicaloutcomes,itisunclearwhetherthepotentiallyinconsistentfindingsareduetodifferencesinprogramdesignoralackofshort-termlinkbetweendecreasedmedicationuseanduseofothermedicalservices.Ifthelatter,aretheobservedtreatmentdiscontinuitiesleadingtonegativeeffectsonimportantclinicaloutcomesthatdonotnecessarily

manifestinclaims-basedmeasuresofutilization?Given the conflicting findings to date, future research

shouldassesswhetherSTforantipsychoticscreatesunintendedadverseclinicalconsequencesby increasingtherapydisconti-nuitiesearlyintreatment,examiningvariationsacrossspecificdisease populations and programdesigns.33No direct impli-cationscanbedrawn fromthis research for theuseofST toaddressthebroaderuseofantipsychoticsinthecommerciallyinsuredpopulationgiventhatratesofoff-labelusearehigh,34 andcommercially insuredpatientsmaybe lessvulnerable totheadministrationrequirementsofST.Aspartofitscompara-tiveeffectivenessresearchprogram,theAgencyforHealthcareResearch andQuality (AHRQ) concluded in 2007 that therewas “insufficient high-grade evidence to reach conclusionsabout the efficacy” of atypical antipsychotics (olanzapine,quetiapine,risperidone,ziprasidone)foroff-labelusessuchasbehavioralproblems indementia,depression,obsessive-com-pulsivedisorder,posttraumaticstressdisorder,andpersonalitydisorders.35

ForPPIsandNSAIDs,theresearchindicatesthatST,evenwithoutgrandfathering,canreducedrugexpenditureswithoutincreasinguseofotherrelatedmedicalservices,makingthesetherapy classes low-hanging fruit for increased managementforplansponsorswhohaveyettoimplementST.Beyondtheselimitedcomments,nootherdefinitiveconclusionscanbemadeaboutSTprogramsrelatedtoqualityofcare,humanisticout-comes,ormedicalexpenditureoffsetsduetothelackofdataon key outcomes and the critical limitations of some of thepublishedwork.Recommendationsforfutureresearchfallinto3areas:(a)researchtopics,(b)methodologicalconsiderations,and(c)publicationbias.

Research TopicsSTresearchpublishedtodatehasexaminedonlyasmallpor-tionofthetherapyclasses,outcomedomains,andpopulationsthatwarrantevaluation(Table3).Specifically, there isaneedfor further researchonSTprograms ineachof the followingareas:

Evaluation of ST programs in other populations and therapy classes, such as antihyperlipidemics, antiasthmatics, antidi-abetics, andmultiple specialty therapy classes. ST programsnowexistfordozensoftherapyclasseswithadoptioncontinu-ing to grow. For example, in 2010, nearly 40%of employershad implemented ST for antihyperlipidemics, yet no evalua-tionsof this therapy classhavebeenpublished.1Research intheMedicarepopulationisalsoneededaspricesensitivityandindicationsforusecanvary.

Improved understanding of the prescription drug savings from step therapy, includinganexaminationof thenet sav-ings of ST after program fees and rebates (when applicable),an understanding of how the savings change over time as

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 151

physicians become familiar with the program (i.e., sentineleffect),and theextent towhichsavingsaredrivenbygreateruseofgenericsversusOTCuseornonutilization.

Evaluations of the effect of step therapy on treatment dis-continuities, includingmedication switching and adherencein clinically relevant therapy categories. To date, treatmentdiscontinuitieshavebeen examined in aminorityof studies,producingmixedresults.Arepatientswhoswitchtothefirst-lineagentmorelikelytoprematurelydiscontinuetherapyortoswitch toanalternativemedication? If so,are thesedisconti-nuitiesdrivenbyrealorperceiveddifferencesineffectiveness?Alternatively, do lower copayments for generic drugs lead tohighermedicationadherenceinST?

Examination of the effect of step-therapy programs on the appropriateness of use.Theextenttowhichpatientsarebeingchanneled to the appropriate alternative has received littleattentionoutsideoftheworkofDelateetal.23Forstatins,are

patientsathigherriskforcardiovasculareventsmorelikelytoreceiveaPAforahigherdosebrandmedicationandlesslikelytonot initiate therapy?Thedefinitionofappropriateusewillvarybytherapyclassandcouldincludecomparisonsofusebyindication,severity,and/orbasedonclinicalguidelines.

Evaluations of the effect of step-therapy programs on clini-cal outcomes and use of other medical services, includinghospitalizationsandERvisits.Thecurrent researchonothermedicalutilizationislimitedinquantityandplaguedbyques-tionablemethodology. This research should also include rel-evantsafety-relatedhospitalizations,sinceresearchinCanadafound that more restrictive ST for cyclooxygenase (COX)-2inhibitorswasassociatedwithalowerrateofhospitaladmis-sionduetoGIbleed.36WhileSTevaluationshaveoftenbeenconductedbyPBMsthatmaylackmedicaldata,healthplansroutinelyhaveaccesstothemedicaldata.Inthefuture,exami-nationof clinicaloutcomesbeyond thoseobtainedbyclaims

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

TABLE 3 Summary of Outcomes Assessed for Step Therapy Across Different Drug Classes and Populationsa

Drug Cost Savings

MedicationAppropriate-ness of Use Medical Spend

Clinical Outcomes

Patient/Provider Satisfaction/

UnderstandingInitiation Adherence

CommercialAnticonvulsantsAntidepressants X X X X XAntidiabeticsAntihypertensives X X X XAntipsychoticsHypnoticsLeukotrienemodifiersMultiplesclerosistreatmentsNasalsteroidsNSAIDs/COX-2inhibitors X X XProtonpumpinhibitors X X XStatinsStimulantstotreatADHDTNF-blockersTriptans

MedicaidAnticonvulsantsAntidepressantsAntidiabeticsAntihypertensivesAntipsychotics X X X XHypnoticsLeukotrienemodifiersNasalsteroidsNSAIDs/COX-2inhibitors X XProtonpumpinhibitors X X X XStatinsStimulantstotreatADHDaThis list represents only a subset of the most commonly adopted step-therapy programs. None of these outcomes has been studied in Medicare.ADHD = attention deficit hyperactivity disorder; COX = cyclooxygenase; NSAID = nonsteroidal anti-inflammatory drug; TNF = tumor necrosis factor.

152 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

placed on it, creating additional administrative burden andhaving questionable impact on quality of care.40,41 WhetherphysicianswillviewtheseautomatedtoolsandeditsforSTasatimelyandefficientadditiontotheirpracticestosupportpro-visionofhighqualityandcost-effectivecareor, alternatively,distortthesystembylearningandreportingtheclinicalcriteriathatproduceanapprovalisanopenquestion.

Methodological ImprovementsThe suggestions for methodological improvements are notuniquetoSTevaluationsnordotheyrepresentanexhaustivelist of good research practices. Rather, the recommendationshighlight important areas for improvement based on the STevaluationspublishedtodate.

Inclusion of appropriate comparator groups.GiventhattheeffectofanSTprogramondrugcostsormedicalexpenditureis likelytobesmallasapercentageof totalexpenditure(i.e.,smalleffectsize),itiscriticaltousecomparisongroupsthataresimilaratbaselineonthekeyoutcomes,keydriversofutiliza-tionandexpense(e.g.,age),andbenefitdesign(e.g.,copaymentamounts).InbothstudiesbyMarketal.,baselinecomparisonson key variables, such as drug spending and disease-relatedmedical expenditures, were not reported.11,12 Although theauthorscontrolledforbaselinemeasuresinthestatisticalanal-yses, statistical controls arenot always sufficient asnotedbyFairmanandCurtiss(2008),andbaselinedifferencesinknownmeasuresmaysignaldifferencesinunknownfactors.42

Examination of disease-related drug and medical spend. Mark et al.’s (2009, 2010) use of all-cause drug or medicalspend as an outcomemeasure is a concerningmethodologi-cal approach because examination of all-cause expenditurescreates greater risk for spurious findings and confoundingduetoothervariables.Thisconcern ismagnifiedbythe lackof reporting of relevant baseline utilization and spending. Itis critical to examinedisease-relatedmedical spendingwhenassessing medical expenditure offsets to establish a logicalcausal pathway between ST implementation and medicaloutcomes.43While sensitivity analysis can examine all-causemedical spendbecauseofuncertainties indiagnostic coding,the primary endpoint should be disease-related drug andmedicalspending.

Adjustment for multiple statistical tests. ST evaluationshave historically conducted dozens of statistical compari-sons, increasing the risk of a significant findingmerely dueto chance.Thisproblem is compoundedby the large samplesizes,whichcanmake themost trivialofdifferences statisti-callysignificant.Accordingly,itisappropriatetoadjustforthemultiple statistical tests bymodifying the alpha required forstatistical significance,44,45whichgenerallyhasnotbeendoneinpublishedSTevaluations.

Better causal linkage between program and outcomes. In

data will be important as research expands into conditionswhere claims-basedmeasuresmay not exist and/or lack thesensitivity todetectclinicallymeaningfulchanges inselectedconditions,suchasrheumatoidarthritis.

Insights into patients’ and physicians’ understanding and satisfaction with step therapy. Early research found thatcommercially insured patients perceived ST edits to be thesame as PAs andwere not aware that lower cost therapeuticalternativesexisted, leadingtounnecessarilyhighratesofPAandnonutilization.29SuchmisperceptionsmaybeevengreateramongMedicaidandMedicarepopulations,whichonlyservesto reduce program performance and increase dissatisfactionamong beneficiaries. Given that concerns over member dis-ruptionanddissatisfactionareperhapsthebiggestbarrierstoadoptionofSTprogramsandthatresearchisverylimited,thisisavaluableareaforfurtherresearch.

Examination of alternative step-therapy designs. ST pro-gramscanvaryintermsoftheirextentofcoverageandapprovalprocess.Twocoveragevariationsthatwarrantevaluationare(a)programsthatdonotgrandfatherpriorusersofthesecond-linealternative,afeaturethatmaygrowforselectedtherapyclasseswhereclinicallyquestionableuseisrampant,and(b)programsthatcoverOTCalternativeswhenavailable.Suchdesignvaria-tions have the potential to significantly affect not only drugsavings,as research inMedicaidhasalreadyshown,butalsotreatmentdiscontinuities,patientsatisfaction,andadministra-tiveburden forpatients andproviders.AsCurtisspreviouslyhighlighted,therestrictivenessofthePAprocesscanalsovarybased on the PA exception criteria and the approval process(e.g.,phoneversusfax),andmoreresearchisneededtounder-stand how these variations affect program outcomes.25 Eventhemostcommoncriterion,prioruseofafirst-lineagent,maynotalwaysbeautomatedintheclaimsdata,37butevaluationsofnonautomatedinterventionshavebeenlimitedtoMedicaidpopulations.

Inaddition,atleast1PBMhasreporteduseof“smartedits”for ST that link to the patient’smedical history to check fordiagnoses in PA criteria.38 This process allows for automaticcoverage of second-line medications for patients with clini-cal indications that meet the PA criteria, such as automatedcoverage of a statin for patients with evidence of a previousmyocardialinfarctionintheclaimsdata,39orautomatedcover-ageofanantipsychotic forpatientswithapreviousdiagnosisofschizophreniaorbipolardisorder.Anunderstandingoftheincremental cost, savings, and quality of care from an inte-gratedprogramisanimportantareaofinquiry.Lastly,asphy-sicianadoptionofe-prescribingandelectronicmedicalrecords(EMRs)grows—allowingforinstantawarenessofandresponsetoanSTedit—evaluationsofreal-timeSTwillbeneeded.Todate, research has found that health information technology,suchasEMRs,doesnotalwaysliveuptothehighexpectations

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 153

assessmentsandhasbeensuggestedforstudiesofprescription cost-sharing.48-50Whilepharmaceuticalfundingdoesnotnec-essarilyindicatebias,itisimportanttoconductindependentlyfundedassessmentsofcommercialSTprograms.Ofcourse,thepotentialforbiasisnotlimitedtopharmaceuticalmanufactur-ers.PBMsandhealthplanscouldbebiasedtodemonstratethattheseprogramssavemoneywithoutaffectingqualityofcare.However, no research has formally assessed the presence ofbiasamongthesehealthcarestakeholders.

Study registration prior to initiation and mandatory publi-cation.Registrationofstudyprotocolspriortostudyinitiationand mandatory publication of full study results after studycompletioncanhelpmitigate thepotential forreportingbias.This idea was discussed for decades with regard to clinicaltrials, and in 2005, the International Committee of MedicalJournal Editors implemented a requirement for clinical trialregistration before patient enrollment.While the policy wascriticizedasburdensomeandstiflingofcompetition,thenum-berofregisteredtrialsatClinicalTrials.gov,thelargesttrialreg-istryatthattime,grewfrom13,153beforethepolicytoinclude67,000 trials as of January 2009.51 Currently, observationalstudies, such as ST evaluations, do not require registration.Whilethisapproachisnotwithoutitschallenges,mandatorystudyregistrationisbeginningtoreceiveseriousconsiderationforthebroadersetofobservationalresearch.52

LimitationsFirst, this study was limited to ST evaluations published intheUnited States. Second,while every attemptwasmade toidentifySTevaluationseveniftheywereidentifiedasPApro-grams in the title, relevant researchcouldhavebeenmissed.Similarly,STandPAarenotentirelydistinctbenefittools,andwhile this study made every attempt to include only thoseprogramsthatwereprimarilydesignedasST,programscouldhavebeenmisclassified.Third, the studydidnotuse formalor quantitative approaches to assess publication bias; rather,conclusionsreflectinferencesmadebythestudyauthor.

■■  ConclusionsThepopularityofstep therapyamongcommercial,Medicaid,andMedicareplansisnodoubtduetothewideavailabilityofgeneric alternatives that offer significant savings, the strongclinical evidence that typically underlies these programs,andtheirabilitytoaffectonlynewusers,therebyminimizingmemberdisruption.However,evaluationsofSTprogramshavenotkeptpacewiththeirgrowinguseinthepublicandprivatesectors. An expanded research agenda is needed to betterunderstandtheeconomic,clinical,andhumanisticoutcomesofSTprograms.Inparallel,improvedmethodologicalapproachesandgreateruseofestablishedstrategiesforreducingstudybiasarewarranted.

thefewstudiesthatexaminedmedicalspending,theresearchhas been challenged by the lack of linkage betweenmedicalexpendituresandmedicationusepatterns.Onewouldhypoth-esizethatothermedicalexpenditureswouldincreasethemostamongnoncompliers and, inparticular, noncomplierswith adiagnosisindicatinggreaterseverityorneed(e.g.,usersofanti-depressantsformajordepressionversusnonspecificpain),yetpriorresearchhasnotbeenconductedatthatlevelofspecific-ity.Giventhequasi-experimentalandsometimesobservationalnatureoftheseevaluationsandthemultitudeofpotentialcon-founders, such specificity isnecessary to establish the causallinkagebetweentheprogramandoutcomesobserved.

Study Bias Study bias can take many forms, including publication bias(i.e., selectivepublicationof research findings, dependingonthe nature and direction of the results),multiple publicationbias, or outcomes reporting bias.46 Outcomes reporting biasoccurswhenoutcomesareselectivelyreported,whennegativeresults are reported in a positivemanner, andwhen conclu-sionsarenotsupportedbytheresults.

Examplesofoutcomereportingbiasarewidespreadinhealthcare.OnestudyofRCTsfoundthatprimaryoutcomesdatahadbeennewlyintroduced,omitted,orchangedinmorethan60%ofcomparisonsbetweenpublicationsandstudyprotocols.47AsSTevaluationsdonotrequireregistration,inferencesmustbedrawnfromexaminationofthepublishedstudies.Inthecaseof the 2pharmaceutical industry-sponsored studies byMarket al., it is unclear why the authors chose all-cause medicalutilizationandexpendituresastheexclusiveendpointsinonestudyandtheprimaryendpointsinanother.11,12Theinclusionoftotalprescriptiondrugspendingratherthandisease-relateddrugspendingastheprimarymeasureofprogramsavings isequallypuzzling,sincethereisnocompellingexplanationforwhySTwouldaffectdrugexpenditures thatareunrelated tothedisease.

Another potential indication of outcome reporting bias isthatintheiranalysisofanSTforantidepressantdrugs,Marketal.reportedthatmentalhealth-relatedinpatientadmissionsandERvisitswerehigherintheSTgrouprelativetothecom-parison group following implementation, but they failed tomentionthattheseresultswerenotstatisticallysignificant.11 Infact, the resultsdidnotevenapproachstatistical significancedespitethelargesamplesize.Againstthisbackdrop,2recom-mendationsareprovidedtoreducethepotentialforbiasinSTevaluations.

Independently funded evaluations.WhilestudiesofMedicaidpopulations have received significant public funding, com-mercial program studies examining medical spending havebeen funded exclusively by pharmaceutical manufacturers.Bias in studies funded by pharmaceuticalmanufacturers hasbeen reported for clinical trial efficacy and cost-effectiveness

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

154 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

DISCLOSURES

Therewasnoexternal funding for this research.Theauthor isPresidentofCareScientific,whichprovidesconsultingservicesinpharmacyandmedicalbenefitsmanagement,includingdesignandevaluationofpharmaceuticalcareanddiseasemanagementinterventions.

REFERENCES

1.PharmacyBenefitManagementInstitute.Prescriptiondrugbenefitcostandplandesignonlinereport.Utilizationmanagement.2010-2011ed.Availableat:http://www.benefitdesignreport.com/UtilizationManagement/UtilizationManagementTools/tabid/110/Default.aspx.AccessedFebruary9,2011.

2.JoadleyJ,HargraveE,MerrellK,CubanskiJ,NeumanT;fortheHenryJ.KaiserFamilyFoundation.TheMedicaredrugbenefit.BenefitdesignandformulariesofMedicaredrugplans:acomparisonof2006and2007offer-ings.Afirstlook.November2006.Availableat:http://www.kff.org/medi-care/upload/7589.pdf.AccessedFebruary9,2011.

3.GibsonTB,OzminkowskiRJ,GoetzelRZ.Theeffectsofprescriptiondrugcostsharing:areviewoftheevidence.Am J Manag Care.2005;11(11):730-40.Availableat:http://www.ajmc.com/media/pdf/AJMC05Nov_Gibson730to740.pdf.AccessedFebruary9,2011.

4.SoumeraiSB,Ross-DegnanD,AvornJ,McLaughlinT,ChoodnovskiyI.EffectsofMedicaiddrug-paymentlimitsonadmissiontohospitalsandnurs-inghomes.N Engl J Med.1991;325(15):1072-77.Availableat:http://www.nejm.org/doi/full/10.1056/NEJM199110103251505.AccessedFebruary9,2011.

5.HoltorfAP,McAdam-MarxC,SchaafD,EngB,OderdaG.Systematicreviewonqualitycontrolfordrugmanagementprograms:isqualityreport-edintheliterature?BMC Health Serv Res.2009;9:38.Availableat:http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2653499/pdf/1472-6963-9-38.pdf. AccessedFebruary9,2011.

6.McAdam-MarxC,SchaafDT,HoltorfA,EngB,OderdaGM.Systematicanalysisofoutcomesevaluationsappliedtodrugmanagementprograms.Am J Manag Care.2008;14(11Suppl):S36-S45.Availableat:http://www.ajmc.com/media/pdf/AJMC_08SpclMcAdamSP36.pdf.AccessedFebruary9,2011.

7.ShoemakerSJ,PozniakA,SubramanianR,MauchD.Effectof6man-agedcarepharmacytools:areviewoftheliterature.J Manag Care Pharm. 2010;16(6Suppl):S3-S20.Availableat:http://www.amcp.org/data/jmcp/JulSuppA.pdf.

8.FischerMA,ChoudhryNK,WinkelmayerWC.ImpactofMedicaidpriorauthorizationonangiotensin-receptorblockers:canpolicypromoterationalprescribing?Health Aff (Millwood).2007;26(3):800-07.

9.RougheadEE,ZhangF,Ross-DegnanD,SoumeraiS.DifferentialeffectofearlyorlateimplementationofpriorauthorizationpoliciesontheuseofCoxIIInhibitors.Med Care.2006;44(4):378-82.

10.FischerMA,SchneeweissS,AvornJ,SolomonDJ.Medicaidprior-authorizationprogramsandtheuseofcyclooxygenase-2inhibitors.New Engl J Med. 2004;351(21):2187-94.Availableat:http://www.nejm.org/doi/pdf/10.1056/NEJMsa042770.AccessedFebruary9,2011.

11.MarkTL,GibsonTM,McGuiganK,ChuBC.Theeffectsofantidepres-santsteptherapyprotocolsonpharmaceuticalandmedicalutilizationandexpenditures.Am J Psychiatry. 2010;167(10):1202-09.

12.MarkTL,GibsonTB,McGuiganKA.Theeffectsofantihypertensivestep-therapyprotocolsonpharmaceuticalandmedicalutilizationandexpenditures.Am J Manag Care.2009;15(2):123-31.Availableat:http://www.ajmc.com/media/pdf/AJMC_09Feb_Mark123to131.pdf.AccessedFebruary9,2011.

13.YokoyamaK,YangW,PreblickR,Frech-TamasF.Effectsofastep-ther-apyprogramforangiotensinreceptorblockersonantihypertensivemedi-cationutilizationpatternsandcostofdrugtherapy.J Manag Care Pharm. 2007;13(3):235-44.Availableat:http://www.amcp.org/data/jmcp/235-44.pdf.

14.DunnJD,CannonE,MitchellMP,CurtissFR.Utilizationanddrugcostoutcomesofastep-therapyeditforgenericantidepressantsinanHMOinanintegratedhealthsystem.J Manag Care Pharm.2006;12(4):294-302.Availableat:http://amcp.org/data/jmcp/research_294-302.pdf.

15.PanzerPE,ReganTS,ChiaoE,SarnesMW.ImplicationsofanSSRIgenericsteptherapypharmacybenefitdesign:aneconomicmodelinanxi-etydisorders.Am J Manag Care.2005;11(12Supp):S370-S379.Availableat:http://www.ajmc.com/media/pdf/A147_05Panzer_S370to79.pdf.AccessedFebruary9,2011.

16.MotheralBR,HendersonR,CoxER.Plan-sponsorsavingsandmemberexperiencewithpoint-of-serviceprescriptionsteptherapy.Am J Manag Care.2004;10(7Pt1):457-64.Availableat:http://www.ajmc.com/media/pdf/AJMC_04JulMotheral457to64.pdf.AccessedFebruary9,2011.

17.CoxER,HendersonR,MotheralBR.Healthplanmemberexperi-encewithpoint-of-serviceprescriptionsteptherapy.J Manag Care Pharm.2004;10(4):291-98.Availableat:http://www.amcp.org/data/jmcp/Research-291-298.pdf.

18.LuCY,SoumeraiSB,Ross-DegnanD,ZhangF,AdamsAS.UnintendedimpactsofaMedicaidpriorauthorizationpolicyonaccesstomedicationsforbipolarillness.Med Care.2010;48(1):4-9.

19.ZhangY,AdamsAS,Ross-DegnanD,ZhangF,SoumeraiSB.EffectsofpriorauthorizationonmedicationdiscontinuationamongMedicaidbenefi-ciarieswithbipolardisorder.Psychiatr Serv.2009;60(4):520-27.Availableat:http://ps.psychiatryonline.org/cgi/reprint/60/4/520.AccessedFebruary9,2011.

20.SoumeraiSB,ZhangF,Ross-DegnanD,etal.Useofatypicalanti-psychoticdrugsforschizophreniainMaineMedicaidfollowingapolicychange.Health Aff (Millwood). 2008;27(3):w195-w195.

21.LawMR,Ross-DegnanD,SoumeraiSB.Effectofpriorauthorizationofsecond-generationantipsychoticagentsonpharmacyutilizationandreimbursements.Psychiatr Serv.2008;59(5):540-46.Availableat:http://ps.psychiatryonline.org/cgi/reprint/59/5/540.AccessedFebruary9,2011.

22.FarleyJF,ClineRR,SchommerJC,HadsallRS,NymanJA.RetrospectiveassessmentofMedicaidstep-therapypriorauthorizationpolicyforatypicalantipsychoticmedications. Clin Ther.2008;30(8):1524-39.

23.DelateT,MagerDE,ShethJ,MotheralBR.Clinicalandfinancialout-comesassociatedwithaprotonpumpinhibitorprior-authorizationprograminaMedicaidpopulation.Am J Manag Care.2005;11(1):29-36.Availableat:http://www.ajmc.com/media/pdf/AJMC05jan_Delate_p29to36.pdf.AccessedFebruary9,2011.

24.SmalleyWE,GriffinMR,FoughtRL,SullivanL,RayWA.Effectofapri-or-authorizationrequirementontheuseofnonsteroidalanti-inflammatorydrugsbyMedicaidpatients.N Engl J Med.1995;332(24):1612-17.Availableat:http://www.nejm.org/doi/full/10.1056/NEJM199506153322406.AccessedFebruary9,2011.

25.CurtissFR.Outcomesofswordswallowingandpharmaceuticalstep-therapyinterventions. J Manag Care Pharm.2007;13(3):284-86.Availableat:http://www.amcp.org/data/jmcp/284-86.pdf.

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

BRENDA R. MOTHERAL, RPh, MBA, PhD, is Associate Professor, University of Kentucky College of Pharmacy, Lexington, Kentucky.

AUTHOR CORRESPONDENCE: Brenda R. Motheral, RPh, MBA, PhD, Associate Professor, University of Kentucky College of Pharmacy, 789 S. Limestone, Lexington, KY 40536-0596. Tel.: 859.323.3849; E-mail: brenda.motheral@uky.edu.

Author

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 155

39.BlueCrossBlueShieldofIllinois.Lipidmanagementsteptherapycrite-riawithmedicaldiagnosesoption.June2010.Availableat:http://www.bcb-sil.com/provider/pdf/lipid_mgmt.pdf.AccessedFebruary9,2011.

40.BlackAD,CarJ,PagliariC,etal.TheimpactofeHealthonthequalityandsafetyofhealthcare:asystematicoverview.PLoS Med. 2011;8(1):e1000387.Availableat:http://www.plosmedicine.org/article/info%3Adoi%2F10.1371%2Fjournal.pmed.1000387.AccessedFebruary9,2011.

41.RomanoMJ,StaffordRS.Electronichealthrecordsandclinicaldecisionsupportsystems:impactonnationalambulatorycarequalityArch Intern Med.2011Jan24[Epubaheadofprint].

42.FairmanKA,CurtissFR.Makingtheworldsafeforevidence-basedpolicy:let’sslaythebiasesinresearchonvalue-basedinsurancedesign.J Manag Care Pharm.2008;14(2):198-204.Availableat:http://www.amcp.org/data/jmcp/JMCPMaga_March%2008_198-204.pdf.

43.MotheralBR,FairmanKA.Theuseofclaimsdatabasesforoutcomesresearch:rationale,challenges,andstrategies. Clin Ther.1997;19(2):346-66.

44.LagakosSW.Thechallengeofsubgroupanalyses—reportingwithoutdistorting.N Engl J Med.2006;354(16):1667-69.

45.KaulS,DiamondGA.Trialanderror.Howtoavoidcommonlyencounteredlimitationsofpublishedclinicaltrials.J Am Coll Cardiol. 2010;55(5):415-27.

46.McGauranN,WieselerB,KreisJ,SchulerY,KolschH,KaiserT.Reportingbiasinmedicalresearch—anarrativereview.Trials.2010;11:37.Availableat:http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2867979/pdf/1745-6215-11-37.pdf.AccessedFebruary9,2011.

47.ChanAW,HróbjartssonA,HaahrMT,GøtzschePC,AltmanDG.Empiricalevidenceforselectivereportingofoutcomesinrandomizedtrials:comparisonofprotocolstopublishedarticles.JAMA.2004;26;291(20):2457-65.Availableat:http://jama.ama-assn.org/content/291/20/2457.full.pdf+html.AccessedFebruary9,2011.

48.SismondS.Pharmaceuticalcompanyfundinganditsconsequences:aqualitativesystematicreview.Contemp Clin Trials.2008;29(2):109-13.

49.BellCM,UrbachDR,RayJG,etal.Biasinpublishedcosteffectivenessstudies:systematicreview.BMJ.2006;332(7543):699-703.Availableat:http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1410902/pdf/bmj33200699.pdf.AccessedFebruary9,2011.

50.FairmanK.Thefutureofprescriptiondrugcost-sharing:realprogressordroppedopportunity?J Manag Care Pharm.2008;14(1):70-82.Availableat:http://www.amcp.org/data/jmcp/JMCPMaga_JanFeb%2008_070-082.pdf.

51.WoodAJ.Progressanddeficienciesintheregistrationofclinicaltrials.N Engl J Med.2009;360(8):824-30.Availableat:http://www.nejm.org/doi/full/10.1056/NEJMsr0806582.AccessedFebruary9,2011.

52.LoderE.GrovesT,MacAuleyD.Registrationofobservationalstudies:thenextsteptowardsresearchtransparency.BMJ.2010;340:375-76.

26.AndersonB.HEDISantidepressantmedicationmanagementmeasuresandperformance-basedmeasures:anopportunityforimprovementindepressioncare.Am J Manag Care.2007;13(4Suppl):S98-S102.Availableat:http://www.ajmc.com/supplement/managed-care/2007/2007-11-vol13-n4Suppl/Nov07-2639pS098-S102.AccessedFebruary9,2011.

27.DormuthCR,PatrickAR,ShrankWH,etal.Statinadherenceandriskofaccidents:acautionarytale.Circulation.2009;119(15):2051-57.Availableat:http://circ.ahajournals.org/cgi/reprint/119/15/2051.AccessedFebruary9,2011.

28.ExpressScripts,Inc.Drugtrendreport2005.Pharmacybenefitguide:190.Availableat:http://www.expressscripts.com/research/studies/drugtren-dreport/2005/.AccessedFebruary9,2011.

29.Commentsbasedontheauthor’sexperienceinconductingarandomizedcontrolledevaluationoftheprogramthathasnotbeenpublished.

30.CentersforMedicareandMedicaidServices.Medicareprescrip-tiondrugbenefitmanual.Chapter6–PartDdrugsandformularyrequirements.February19,2010.Availableat:https://www.cms.gov/PrescriptionDrugCovContra/Downloads/Chapter6.pdf.AccessedFebruary9,2011.

31.KomossaK,Rummel-KlugeC,SchwarzS,etal.Risperidoneversusotheratypicalantipsychoticsforschizophrenia. Cochrane Database Syst Rev. 2011 Jan19;1:CD006626.

32.SimeoneJC,MarcouxRM,QuilliamBJ.Costandutilizationofbehav-ioralhealthmedicationsassociatedwithrescissionofanexemptionforpriorauthorizationforsevereandpersistentmentalillnessintheVermontMedicaidprogram.J Manag Care Pharm.2010;16(5):317-28.Availableat:http://www.amcp.org/data/jmcp/317-328.pdf.

33.FarleyJ.Containingcostdespitethecost.Med Care.2010;48(1):1-3.

34.AlexanderGC,GallagherSA,MascolaA,MoloneyRM,StaffordRS.Increasingoff-labeluseofantipsychoticmedicationsintheUnitedStates,1995-2008.Pharmacoepidemiol Drug Saf.2011;20(2):177-84.

35.ShekelleP,MaglioneM,BagleyS,etal.Efficacyandcomparativeeffec-tivenessofoff-labeluseofatypicalantipsychotics:comparativeeffective-nessreviewno.6.PreparedbytheSouthernCalifornia/RANDEvidence-basedPracticeCenterundercontractno290-02-0003fortheAgencyforHealthcareResearchandQuality.AHRQpublicationno.07-EHC003-EF.January2007.Availableathttp://www.effectivehealthcare.ahrq.gov/repFiles/Atypical_Antipsychotics_Final_Report.pdf.AccessedFebruary10,2011.

36.MamdaniM,WarrenL,KoppA,etal.Changesinratesofuppergastrointestinalhemorrhageaftertheintroductionofcyclooxygenase-2inhibitorsinBritishColumbiaandOntario.CMAJ.2006;175(12):1535-38.Availableat:http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1660595/pdf/20061205s00016p1535.pdf.AccessedFebruary9,2011.

37.WellpointNextRx.Clinicalsystemedits.Availableat:https://www.well-pointnextrx.com/wps/portal/wpo/client/clinicalservices/clinicalinterven-tions/clinicalSystemEdits.AccessedFebruary9,2011.

38.GleasonPP.Assessingstep-therapyprograms:astepintherightdirec-tion.J Managed Care Pharm.2007;13(3):273-75.Availableat:http://www.amcp.org/data/jmcp/273-75.pdf.

Pharmaceutical Step-Therapy Interventions: A Critical Review of the Literature

156 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

What Do We Really Know About VBID? Quality of the Evidence and

Ethical Considerations for Health Plan Sponsors

Kathleen A. Fairman, MA, and Frederic R. Curtiss, PhD, RPh, CEBS

As health care proposals are debated and specific reforms are imple-mented, there will be a vital need for solid, credible, actionable infor-mation about the impact of current policies. … To meet the needs of health care reform, research will need to be significantly more respon-sive to demands for findings that are timely and actionable.1

In March 2009, the Agency for Healthcare Research and Quality (AHRQ) convened a meeting of “policymakers, researchers, and producers of health care data” to “begin

developing a strategy to optimize the availability of informa-tion and data for enactment and implementation of health care reform.”1 The needs described by participants were as expected: timely information about “the likely concrete effects of unique, specific policy proposals,” provided with an eye to “the specific levers available to policymakers,” developed with a “clear and transparent” methodology, and assessing “the impact on the public sector and consumer.” It is telling that the need for “solid data” was emphasized by “veterans of past health care reform efforts,” who also observed that current data needs “are much broader.”1

Data Promulgated Versus Data Used : Is There a Credibility Gap?The AHRQ’s envisioned role for accurate and transparent data in developing and monitoring a redesigned health care system represents an optimal scenario—if health care research “builds it,” policymakers “will come.” Yet, the ways in which health plan sponsors actually obtain and use information in making benefit design decisions may not match this expectation. A Diamond Management/Willis North America survey of large and small employers in a variety of industry sectors, conducted after passage of the Patient Protection and Affordable Care Act (PPACA) using an unreported sampling methodology, found that a slight majority (736 of 1,400, 52.6%) of respondents cited “in-house personnel” as their primary source of infor-mation about “the health care reform statute or regulations.”2 Insurance brokers and carriers were cited as a primary infor-mation source by 25.1% and 6.6%, respectively. Only 5.2% said that their primary information source was “free resources and tools as published by government agencies.”2 Recent news reports also suggest that employers currently rely primarily on consultants to guide their planning for implementation of PPACA, including choices of cost-sharing levels and benefit

designs.3 Similarly, only 6% of human resource and benefits managers responding to a 2010 Kaiser Family Foundation/Health Research & Educational Trust health benefits survey (N = 2,046, response rate 47%) reported that they “review performance indicators for their health plan’s service or clini-cal quality,” and of that small group, only 18% reported using Healthcare Effectiveness Data and Information Set (HEDIS) measures promulgated by the National Committee for Quality Assurance.4

Various factors may explain employers’ apparent reluctance or inability to use externally promulgated information about health care policy and quality. One is organizational size; large and small employers generally exhibit markedly different responses to external change.2,4 However, as we have observed previously, another possible reason for what appears to be inexplicable indifference to the guidance of external “experts” is appropriate wariness about recommendations that have not yet been tested in rigorous research.5 That is, the most skeptical decision makers may be the best informed about the limita-tions of available evidence. A growing body of work suggests that even widely used standards, such as ambulatory care met-rics that are endorsed by key quality measurement organiza-tions6 and infectious disease treatment guidelines promulgated by the Infectious Diseases Society of America,7 are commonly supported primarily by expert opinion or a methodologically suboptimal evidence base. Moreover, guidelines supported solely by observational evidence and/or expert opinion are commonly refuted and replaced when subjected to the more rigorous test of a randomized controlled trial.8

Slow Uptake of VBID: Information Gap or Sensible Decision Making by Plan Sponsors?The Rx Outcomes Adviser, a recently launched blog intended to provide health plan sponsors with actionable and evidence-based information about pharmacy benefits, highlighted in an early posting the costs and benefits of value-based insurance design (VBID), the lowering or waiving of copayments for “high-value” medications in an effort to increase their use.9 Because the blog’s purpose is to help decision makers navi-gate the challenge of reviewing and applying health outcomes research to make well-informed policy decisions,10 a focus on the cost-effectiveness of VBID is particularly appropriate.

EDITORIAL

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 157

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

and implementing or enhancing clinical disease management programs (DMPs).23-28 The annual drug trend reports (DTRs) of PBMs, which reflect their marketplace strategies and benefit design recommendations, suggest a recognition that nonad-herence is a multifaceted phenomenon encompassing negative perceptions about medications and side effects, patient char-acteristics (e.g., low literacy, forgetfulness), and therapeutic complexity requiring pharmacist intervention.24-28

However, the topic of VBID has not been avoided altogether by PBMs. A few PBMs have reported observational investiga-tions of the association between out-of-pocket cost and medica-tion use27,29,30 or between medication adherence and all-cause health care costs.31 In a balanced approach to VBID, a health plan-owned PBM advises against zero-dollar copayments because “even as little as $5 per prescription gives the benefit more perceived value to members;” suggests that plan sponsors “carefully consider whether value-based pharmacy benefits are a good fit” for members with adherence of less than 80% for medications to treat diabetes, high blood pressure, or dys-lipidemia; and recommends that to improve cost-effectiveness, sponsors consider copayment reductions for generic drugs only.28 These observations, and the advice of most PBMs to use a variety of tools and messaging strategies to improve adher-ence, are consistent with behavioral economics research sug-gesting that consumer purchasing behavior is neither entirely rational nor based solely on price.32

Peer-reviewed evidence about the reasons for payers’ reluc-tance to adopt copayment reduction strategies is currently unavailable. However, popular press and industry reports suggest that major factors in the slow uptake for VBID include (a) the “potential for short-term increase in utilization and cost” with “uncertain” health benefits,21,33 (b) the possibility of “unintended incentives” that could reduce generic drug utiliza-tion if brand drug copayments are reduced too much,33 and (c) synergies between business interests and the use of alternative adherence promotion strategies.23

Payer Perspectives on High Cost and Uncertain Evidence for VBIDA PBM medical officer interviewed for an October 2009 report on novel business models in the PBM industry cited a focus on improving medication adherence as an important market niche. Although noting that medication nonadherence repre-sents a “tremendous opportunity ... to lower medical costs, like [emergency room] visits for asthmatics,” the PBM executive argued against using copayment waivers to promote adher-ence, largely because of the high cost of VBID relative to other strategies that the PBM advocates as more effective: “Benefit designs that eliminate copays increase compliance by 1 to 5 percent, but a company takes on a huge additional expense by absorbing the copay. We found that moving patients to [mail order pharmacy] is the single most important thing you can

Section 2713(a) of the PPACA calls for the creation of guide-lines to “permit a group health plan and a health insurance issuer offering group or individual health insurance coverage to utilize [VBID],”11 suggesting that whether to adopt a VBID for the drug benefit may become an increasingly important question for payers.

Yet, despite being launched more than 9 years ago as “a concept that stands on science and equity”12 and sustained by financial support from the pharmaceutical industry13,14 and enthusiastic popular press coverage,15-17 VBID (initially termed “benefit-based copay”) has not caught on among the majority of health plan sponsors. Recent VBID use rates are estimated at approximately 20%, and plan deductibles over the past few years have generally trended higher, not lower.18-22 For example, when 507 employers with at least 1,000 employees were asked in the 2010 National Business Group on Health/Towers Watson (NBGH/TW) survey to indicate benefit design strategies that they had used during 2010, reductions in “pharmacy copays or coinsurance for those with chronic conditions” were cited by 19% of employers. Of 37 strategies mentioned in the 2010 survey, only 2 were used less often than copayment reductions: increasing the use of selective provider networks and providing incentives to use personal health records. In contrast, 23% of employers reported having “significantly” increased “pharmacy copays, deductibles or coinsurance” in 2010; 54% reported offering a consumer-directed (high-deductible) health plan (CDHP); and the percentage of CDHPs covering “preventive prescription drugs” at 100% (i.e., no patient cost-sharing) was just 20%.18 When asked about their plans for 2011, 28% of employers reported planning to significantly increase phar-macy cost-sharing requirements, and total estimated CDHP availability was expected to grow to 61%.18 Although reluctance to expend funds on health care in a declining economy was no doubt partly responsible for these trends, it is notable that employers were willing to make investments in other areas. For example, the percentages offering programs for lifestyle behavioral change, health coaching, and smoking cessation were 50%, 56%, and 76%, respectively.18

Another indication of lukewarm interest in VBID is the less-than-enthusiastic treatment it receives from most large pharmacy benefits management (PBM) companies. This phe-nomenon is curious and worthy of attention because PBMs can be exceptionally customer-focused and would presumably embrace customer demands that are popular in the market-place, since they typically do not bear financial risk (e.g., capitation) for the cost of pharmacy benefits. Yet, instead of pharmacy copayment reduction intended to improve medica-tion adherence, PBMs focus more attention on alternative strat-egies that include dispensing medication in 90-day supplies, increasing direct contact between pharmacists and patients, providing targeted refill reminders to nonadherent patients,

158 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

do to improve adherence.”23 The DTR of a health plan-owned PBM similarly acknowledges that “quantifying the [return-on-investment] for [VBID] is challenging” because “medical costs avoided, improved quality of life, reduced absenteeism and other positive outcomes are often cited but can be lost when incentives are applied too broadly.”28

Popular press interviews with benefit design consultants reported in May 2009 also suggested that cost relative to uncertain benefit is an important consideration for employers in making decisions about VBID, with one consultant noting that because of the economic downturn “it’s a little bit harder to get approval for reducing copays, especially when results are a little bit squishy” and another suggesting that growth in the copayment waiver trend is limited because “there’s not a ton of data” about the effects of VBID on overall medical cost.21

Nonetheless, a November 2010 press report on adopters of VBID suggested that the estimated 18% of employers with a chronic medication copayment waiver or reduction in place as of 2009 “are convinced this [approach] works, despite an absence of rigorous research to back them up.”20

Still, lack of reliable evidence about VBID may not be the only reason for tepid response to it by most PBMs. First, PBMs that own mail order pharmacies have an interest in using copayment incentives to encourage use of mail order by refill customers and might be expected for business reasons to rec-ommend 90-day supplies, although some PBMs recommend 90-day dispensing for community pharmacies as well as mail order.28 Second, PBMs may be reluctant to recommend copay-ment reductions for brand medications for reasons that include client preference and higher profit margins on generic than brand drugs because of “spread” pricing.23

Notably, however, research evidence suggests that both of these PBM business interests are congruent with the needs of health care consumers. That is, consistent with PBM recom-mendations, mail order use is cost-effective for plan spon-sors34 and positively associated with adherence to chronic medication therapy,35,36 although randomized trials of the effects of mail order pharmacy use on patient outcomes are clearly needed.37 Also supporting PBM recommendations are the findings of AHRQ-funded comparative effectiveness analy-ses suggesting that generic medications for high-prevalence chronic conditions, such as hypertension and diabetes,38,39 on average provide equivalent therapeutic effect at equivalent or better safety—a remarkable bargain for patients considering that 90-day supplies of generic drugs to treat high-prevalence chronic conditions are commonly available for $10-$12 in large community pharmacies.40

Proposal to Offset VBID Costs: Higher Cost Sharing for “Low-Value” ServicesVBID proponents rightly point out that the net total cost of a VBID program consisting only of copayment reductions

“critically depends on whether the incremental spending on the targeted services, such as hypertension medication, can be offset through a decrease in adverse events, such as hos-pitalizations.”41 In other words, does a health plan sponsor that pays for both medical and pharmacy benefits and incurs higher pharmacy benefit costs because of reducing or waiving copayments for medications to treat high-prevalence chronic conditions (e.g., diabetes, hypertension, dyslipidemia) reap the benefit of healthier enrollees who incur fewer catastrophic medical events (e.g., myocardial infarctions [MIs], strokes) as a result of increased medication use? More importantly, do health plan members—who ultimately pay a portion of the cost of health policy decisions in their share of health plan premiums—experience increased, decreased, or unchanged health benefits when the health plan sponsor implements VBID copayment reduction?

A transparent “plausibility calculator” analysis, designed and reported by Melnick and Motheral in the March 2010 issue of JMCP, suggests that “health plan sponsors are highly unlikely to experience net savings by implementing VBID programs, even under generous assumptions, for 2 reasons.”42 First, price elasticity for medications is generally low in commercially insured groups, meaning that most copayment reductions go to patients who would have been equally compliant without the copayment savings. Second, in a population of commer-cially insured enrollees “with varying risk levels,” the baseline risk of avoidable expensive medical events (e.g., emergency room [ER] visits and hospitalizations) is low, meaning that it is mathematically nearly impossible for increased medication compliance to prevent a sufficient number of expensive events to offset the costs of copayment reductions.42

Advocates of VBID have argued based on “break-even” economic modeling that VBID “can be effective”43 but acknowl-edge that “direct medical savings from increased use of services with strong evidence of clinical benefit are unlikely to finance the entire [VBID] investment in the short term.”41 Thus, these VBID proponents support “investments in processes to define low-value care” and a benefit design that “couples cost-sharing reductions for high-value services with cost-sharing increases for services not identified as high value.”41 This suggestion has led to an important emerging issue for employers considering VBID adoption—provider and patient acceptance of higher lev-els of cost-sharing for services deemed “low-value.”19,41,44

The Payer’s Dilemma: Value to Whom?Methods for defining a “low-value” service are in their nascence and are far from straightforward. Fendrick et al. (2010) have suggested that services should be defined as “low-value” if they “result in harm—for example, services with D designation that are discouraged by the [U.S.] Preventive Services Task Force” (USPSTF)—or, more broadly, if their cost is “deemed too expensive for the health benefits produced.”41

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 159

Health economist James Robinson has argued that “low-value” services should be subsets of “high-cost health services,” such as “biopharmaceuticals, implantable medical devices, advanced imaging modalities, and specialized surgical procedures.”44 Various methods would be used to determine which high-cost services would fall into the “low-value” subset, depend-ing on the clinical circumstances. For biopharmaceuticals, for example, Robinson proposes that value might be defined based on the patient’s “clinical indication, disease severity, and comorbidities,” perhaps in conjunction with practice setting characteristics including “physician specialty, the presence of care coordination, and patient education services.” He suggests that U.S. Food and Drug Administration (FDA) approval might also be used as an indication of value (i.e., drugs used off-label would be subject to higher patient cost sharing), perhaps in conjunction with a drug compendium or “an authoritative, evidence-based care pathway.”44 For implantable devices (e.g., stents, pacemakers), he proposes that payers “cover the basic-function device, leaving the patient to buy up to a higher-function alternative—unless the higher-function device were known to offer a clinically better outcome to this particular type of patient” as determined by “the health plan’s medical management professionals in consultation with the patient’s physician.”44 And, for common surgeries, such as knee replace-ments, “the insurer could specify a contribution that it would make, with the contribution based on the payment rate for the most efficient provider team in the local market. The patient would be assigned responsibility for paying the extra costs incurred if a higher-price provider team is chosen.”44 Putting aside the complexity of designing, maintaining, and adminis-tering a system like that envisioned by Robinson, 2 important and potentially controversial issues are apparent.

Whose “Evidence-Based Conclusion” Prevails in Determin- ing “Value?” First, key organizations, viewing the same evi-dence, often reach different conclusions about the value of medical services. For example, the suggestion by Fendrick et al.41 to assign a “low-value” designation to USPSTF “D” services is reasonable, but the USPSTF currently has a D designation for teaching women to perform monthly breast self-examinations (BSE)45 and, at this writing, appears poised to give prostate cancer screenings for men in all age groups a D designation in an upcoming meeting.46 In contrast, screening guidelines pro-mulgated by the American Cancer Society (ACS) describe BSE training and prostate cancer screening as options that should be made available to patients and providers, depending on the patient’s level of risk and personal preferences.47-49

Similarly, the USPSTF currently has a C designation for mammography before the age of 50 years, meaning that the USPSTF “recommends against routinely providing the ser-vice” because “there is at least moderate certainty that the net benefit [of the service] is small” after accounting for both

benefits and harms associated with the screening process; these harms include primarily “psychological harms, unnec-essary imaging tests and biopsies in women without cancer, and inconvenience due to false-positive screening results” and “overdiagnosis,” defined as the detection and treatment “of cancer that would never have become clinically apparent.”50 Nonetheless, the USPSTF recognizes the possibility of “con-siderations that support providing the service in an individual patient,” including “the patient’s values regarding specific ben-efits and harms.”50 In contrast, the ACS recommends annual mammograms beginning at 40 years of age “and continuing for as long as a woman is in good health.”51 The American College of Radiology (ACR) goes a step further in describing the value of screening mammography for women aged 40 to 49 years, calling the USPSTF recommendations “ill advised and danger-ous” and “unconscionable,” describing annual mammography screening for women aged 40 years or older as “one of the major health care advances of the past 40 years,” and arguing that the USPSTF “selectively reviewed the literature, ignor-ing hundreds of well-regarded studies on the subject.”52 This controversy leaves a payer that is trying to define cost-sharing levels based on “value” with an obvious and seemingly irrecon-cilable conundrum: which evidence-based conclusion should the payer adopt in determining the value of mammography for women aged 40 to 49 years—that of the USPSTF, the ACS, or the ACR?

Ethical Considerations in Deeming Services “Low-Value” When Evidence is Uncertain. Second and more important, a policy of redirecting scarce resources from one patient group (users of high-cost medical services, often for catastrophic ill-ness) to another patient group (users of chronic medication therapy for high-prevalence conditions, often for primary prevention) has profound ethical implications that require thoughtful consideration of the evidence about the proposed policy change. In other words, consistent with the principles of comparative effectiveness research, appropriate and ethical resource allocation requires reliable and valid evidence about the costs and benefits of available alternatives. Yet, as Choudhry et al. (2010) observed in a recent review, “there is a paucity of data to indicate for which services cost sharing should be increased,” as well as a “lack of evidence” about whether drug copayment reductions “will yield better health outcomes and lead to reductions in other health care costs.”11 This lack of evi-dence makes the resource allocation shifts suggested by VBID proponents ethically questionable at the present time.

These ethical implications become even more cogent when one considers the possibility that reduced brand drug copay-ments could potentially incentivize patients to use brand medications for therapeutic purposes that could be served with generics, a phenomenon of “unintended incentives.”33 Such incentives reduce the affordability of the benefit, both overall

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

160 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

and—a point that often goes unnoticed—for the individual patient who pays an unnecessarily high, albeit reduced, copay-ment for the brand drug.

The clinical implications of potentially encouraging greater first-line brand drug use should also be thoughtfully consid-ered because the risks and benefits of newer medications are sometimes not as fully understood prior to FDA approval as they are in the postmarketing phase. The history of rosiglita-zone—a top-selling diabetes drug that was initially approved in 1999 but “significantly” restricted by the FDA in September 2010 because new studies suggested that it elevated cardiovas-cular event risk—provides a case in point.53,54

Early Responses to Defining “Low-Value” Services. Given these uncertainties, it is perhaps not surprising that early efforts to implement VBID programs encompassing both copayment reductions for “high-value” services and copayment increases for “low-value” services have been met with skepticism and controversy. In March 2010, Kaiser Health News reported that 5 insurers in Oregon had begun offering a benefit design that provided free or low-cost physician visits and prescription drugs for certain chronic conditions (asthma, congestive heart failure, diabetes, depression, heart disease, chronic bronchitis, and emphysema) but required much higher patient cost shar-ing for “treatments deemed overused.”55 Patients using targeted “overused” treatments—such as knee or hip replacement, coronary artery bypass surgery, stent placement, hysterectomy, certain imaging examinations, and ER visits—would pay double the usual plan deductible, double the office visit copay-ment, and up to one-half of hospitalization and ER cost, up to an annual out-of-pocket maximum of $1,500 for individuals and $3,000 for families. Only 1 employer, a steel manufacturer, adopted the plan offering, and its head of benefits estimated that only 7% of workers would select it.55

Detractors of this and similar VBIDs argue that the designs fall into a “danger zone of limiting access to medical care” and that they represent “too much of a blunt instrument.”55 For example, not every patient with heart disease can be treated medically instead of surgically, and “while researchers say that, overall, too many hysterectomies are performed, women with uterine cancer have little choice.”55 When the state of Oregon implemented a similar VBID for its public employees in 2010, its planning group “ruled out including cardiac treatments and hysterectomy in the higher-cost [category of] coverage because doing so was considered too contentious.”56

Inaccurate Descriptions Hamper Efforts to Evaluate VBIDThe important task of finding answers to the question of VBID’s effectiveness in improving patient health has been complicated by deficiencies in the base of evidence about pharmacy benefit copayment change. We noted in a 2008 editorial that evidence suggesting an effect of VBID on medication adherence was

scant and methodologically nontransparent.57 We have also noted that the “case” for the negative effect of typical copay-ment increases on adherence has been made primarily with research employing observational designs in which association was sometimes erroneously equated with causation; quasi-experimental (pre vs. post with comparison group) studies of typical copayment increases in commercially insured groups have generally found positive outcomes for patients and plan sponsors.8,58-61

Unfortunately, much of the literature on cost sharing has tended to overstate or misstate previous research findings, suggesting that cost sharing had a greater influence on medica-tion adherence than was actually documented in the original work. For example, in discussions of cross-sectional analyses comparing higher versus lower cost sharing levels, one some-times sees references to “price responsiveness” to “[copayment] changes” when no copayment change was measured58 or to “adherence” outcomes from studies that included no measures of adherence;57 and seemingly small but often substantively important discrepancies are common. In a recent example, a 2008 study by Sedjo and Cox62 of changes in statin utilization following simvastatin (Zocor) patent expiration was among the studies cited in the introduction to a research article to support a statement that medication adherence increases by 2%-5% in the first year after VBID implementation.63 Sedjo and Cox did find that following simvastatin patent expiration and the resulting change from a brand to generic copayment (not a VBID program implementation), the medication pos-session ratio (MPR) for simvastatin users increased by a mean adjusted 0.52%, compared with an MPR decline of 2.02% for users of brand statins other than simvastatin, who experienced no copayment decrease. However, they also found that among patients who experienced the greatest copayment decreases measured in the study, more than $15, price elasticity was –0.02—that is, no meaningful price responsiveness; and, they found that the percentages of patients whose MPR increased from less than 80% before patent expiration to at least 80% after patent expiration were 10.5% in the simvastatin group and 10.0% in the group using other brand statins—that is, no meaningful difference.62

A more serious discrepancy appeared in the introduction to a recently published research article, in which a study by Huskamp et al. (2003) was cited to support the statement that “prescription drug benefits that shift greater costs to consum-ers or increase barriers to access can result in lower rates of drug treatment, poorer adherence, and increased rates of treatment discontinuation.”29 This description is inconsistent with the actual finding of Huskamp et al. that a change from a single-tier $7 copayment to a 3-tier $8/$15/$30 benefit, but not a change from a 2-tier $6/$12 to a 3-tier $6/$12/$24 benefit, was associated with a reduced probability of using medication in 3 therapy classes (proton pump inhibitors,

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 161

care costs (Appendix). Although all the VBID studies provided evidence that implementation of VBID might have a small favorable effect on medication adherence, none provided the information that payers need to make an informed decision about VBID because of serious problems in reporting, program design, and effect calculation.

No Information About Generic Utilization. None of the new studies of VBID reported its effect on generic utiliza-tion,29,63,68-71 an especially important omission for programs in which copayment reductions either included brand drugs29,68,70 or were limited to brand drugs.69,71 Failure to report the effects of brand drug copayment decreases on generic drug utiliza-tion is especially problematic because previous observational investigations have found an association between prescription drug cost-sharing increases and increased use of lower-cost medications (i.e., generic and preferred brand drugs) in the same therapeutic class.58,61

No Information About Payer Cost. None of the new VBID studies reported the cost of the intervention to the payer.29,63,68-71 A report by Gibson et al. (2011a) of a program in which coin-surance rates for brand antidiabetic drugs were reduced—from 20% for tier 2 and 35% for tier 3 to the 10% level previously in place only for generic medications—is especially notable because it purported to show a favorable return-on-investment (ROI) attributable to medical cost offsets but actually measured total costs, not payer costs.71 In other words, the calculation ignored the cost to the employer for the copayment waivers (i.e., the cost shift from patient to payer); thus, its results do not represent ROI from the payer perspective. The report also indicated significant effects for VBID only for a subgroup of patients enrolled in a DMP but did not report the cost of the DMP. This omission of DMP costs is especially important because the program was intensive, including nurse telephone outreach, coaching, and monitoring, in addition to numerous written educational components.71 Thus, despite the indication in the article’s title that VBID coupled with a DMP “produced savings,” the study did not address that research question.

In another report by Gibson et al. (2011b) of a copay-ment reduction intervention for medications to treat diabetes, asthma, and hypertension, the authors “raise the prospect that this program may have saved the company money by reducing other medical costs” because “clinical effects such as changes in glucose levels, blood pressure, and lung function-ing might have occurred,” but no assessment of these measures was made, and the report’s quantitative analysis indicated no significant association between VBID and any of 3 measures of spending (pharmaceutical, medical, or total) during the 3-year post-implementation period.70 The study report also transparently disclosed 2 important confounding factors—a DMP was phased in for the VBID employer during the post-implementation period, and the VBID employer was a phar-maceutical manufacturer that provided its brand drugs free of

angiotensin-converting enzyme [ACE] inhibitors, and statins).61 More importantly, Huskamp et al. found that treatment discontinuation rates were significantly higher for patients who experienced a $23 copayment increase (from $7 to $30) but not for patients who experienced a $12 copayment increase (from $12 to $24). And, perhaps because of type 1 error or an unmeasured confounder, among patients taking ACE inhibitors, treatment discontinuation rates were lower (not higher) for the patients who experienced a copayment increase from $12 to $24 than for patients whose copayments remained unchanged at $12—that is, persistence with ACE inhibitor therapy was better for patients with a higher copayment.61

The Huskamp et al. study has been inaccurately described elsewhere including a 2008 commentary, which indicated that “when an employer increased cost-sharing requirements by about $10 to $20 per prescription … 21% of patients stopped taking their medication for high cholesterol (compared with 11% in a control group);”64 the actual research finding by Huskamp et al. for a group experiencing a copayment increase of approximately $10 was that 3 of 33 (9.1%) statin users whose copayment increased by $12 versus 1 of 25 (4.0%) statin users without a copayment increase discontinued statin use, a non-significant difference (P = 0.45).61

The same commentary, an assessment of policy implica-tions from the Rand Health Insurance Experiment (RHIE) and selected observational studies, indicated that by discouraging use of appropriate health care services, cost sharing “could lead to additional hospitalization, emergency department visits, and even death.”64 The RHIE did find that higher levels of cost shar-ing led to reduction in use of health care services, including both those deemed essential and nonessential by RHIE inves-tigators; yet, the RHIE found that these service use reductions had no effects on health except for a few minor outcomes (e.g., vision, dental, and increase of an average 3 millimeters mer-cury [mm Hg] in diastolic blood pressure) that were limited to low-income RHIE enrollees only.65,66 In contrast to the com-mentary’s portrayal, RHIE authors even argued in 1987 and 1992 that their finding of “enormous potential savings” from cost sharing, with “little apparent health impact on the kind of people who typically are covered under employer health insur-ance,” may have prompted plan sponsors to increase coinsur-ance and deductibles in the years immediately following the RHIE’s publication, resulting in large health care cost savings nationwide.65,67

New Observational Evidence About VBID Copayment ReductionSince publication of our earlier observations about deficien-cies in research on VBID,8,57 new copayment reduction stud-ies using nonrandomized comparison groups have been published,29,62,63,68-71 and an additional study31 measured the association between medication adherence and all-cause health

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

162 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

not contain the patient subgroup counts necessary to calculate unbiased discontinuation rates for the study groups.

Less serious but still important is a problem with attrib-uted causal linkage that appears in a transparent and clearly reported analysis by Roebuck et al. (2011) of the association between medication adherence and all-cause health care costs.31 In that analysis, the authors made no assessment of whether the observed all-cause utilization could reasonably be clinically attributed to medication nonadherence (e.g., cardiovascular events vs. automobile accidents); the authors chose instead to rely solely on fixed-effects statistical modeling techniques to establish causal linkage, a suboptimal method for observational analysis.72 Additionally, the authors assumed that a patient with a chronic disease diagnosis (heart failure, diabetes, hypertension, or dyslipidemia) but without pharmacy claims had medication adherence of zero, regardless of whether any medication was prescribed. This decision excludes the pos-sibility of nonpharmacologic interventions based on lifestyle modification and is, although appropriate for heart failure, inconsistent or only partly consistent with treatment guide-lines for the other 3 disease states.73-75

Another problem common to varying degrees in the new VBID studies is lack of specificity in the descriptions of sample construction. For example, only 1 included a sample selec-tion flowchart to enable the reader to determine the effect of each sample inclusion and exclusion criterion,69 although this information is considered a standard for research reporting.72 Moreover, it was difficult to determine key details, such as the specific criteria used to define a “user” of a particular drug class or to qualify for DMP entry, in many reports. Similarly, in many VBID reports it was difficult or impossible to identify the pre-intervention values and/or the absolute change amounts for the outcome measures.

Still No Randomized Trials. As Roebuck et al. candidly acknowledge, even a well-done multivariate statistical analy-sis controlling for fixed effects (e.g., as a proxy for “healthy adherer” effects) is “not as good as a randomized controlled trial in establishing causality.”31 The report by Gibson et al. of the brand antidiabetic drug copayment reduction program is

charge to employees. Unfortunately, the brand drugs were not named and the DMP was not described, making it difficult for decision makers to interpret or use the study results in esti-mating the costs and benefits of VBID. Finally, as in the other recent report by Gibson et al., the authors indicated that “the program was mostly cost-neutral to the company;” however, the measure of expenditures was “eligible charges,” which was not defined specifically but appears to be a total (rather than a net after copayment) that does not represent the health plan sponsor’s costs.70

Unfortunately, in addition to failing to report the cost of the intervention to the health plan sponsor, none of the new VBID study reports contained even enough quantitative information for readers to calculate the intervention cost. Although most provided some information about mean copayment amounts paid per claim by the intervention and/or comparison groups before and/or after the intervention, only a few provided quan-titative information for both time periods and groups, and none provided information about utilization (number of claims), which is essential to translating cost per claim into total cost. Additionally, tier-specific utilization data were not included in any of the new VBID study reports but are necessary to deter-mine whether utilization shifts to drugs in higher or lower tiers affected payers’ ingredient cost expenditures.

Problems in Study Design and/or Reporting. Several prob-lems in recent VBID research reports, including discrepancies between study methods and either methodological or clinical guidelines, are apparent (Appendix). For example, in an analy-sis of therapy discontinuation rates for a VBID program for antidiabetic medications, Chang et al. (2010) removed from the denominator patients who made a medication switch from one antidiabetic subclass to another (e.g., from a thiazolidinedione [TZD] to metformin or to a sulfonylurea).29 This decision had the mathematical effect of inflating the discontinuation rate overall, and especially in the presence of even appropriate and desired switches from brand to therapeutically equivalent generic drugs, thereby systematically biasing the analysis to find higher “discontinuation” rates for the comparison group than the VBID group (Table 1). Unfortunately, the report does

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

Hypothetical Plan Number of UsersNumber Switching to Generic Medication

Number Discontinuing Therapy

Actual Discontinuation Rate

Discontinuation Rate After Removing Switching Patients from Denominator

Plan A — VBIDa 100 0 10 10.0% 10.0%

Plan B — tiered benefita 100 20 10 10.0% 12.5%aThis example assumes, for purposes of illustrating the mathematical effect, that patients in the VBID had no incentive to switch from brand to generic medication, in con-trast to patients in the traditional tiered benefit plan.VBID = value-based insurance design.

Illustration of the Mathematical Effect of Removing Patients Making Medication Switches Before Calculating Discontinuation Rates

TABLE 1

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 163

the approximate payer cost per member per year (PMPY) for that VBID program after full phase-in was $18.60 ($1.86 mil-lion per year in a drug plan with 100,000 members) for statins to achieve a 3.4 percentage point increase in MPR, or about 12 added days of therapy each year on average.57 Although the VBID for statin drugs reported recently by Choudhry et al.68 was limited to patients using prescription medications for diabetes or cardiovascular disease, one could reasonably assess the cost of providing free statins to all statin users in a health plan by using the mean pre-intervention statin copayment of $24 per claim reported by Choudhry et al., coupled with the use prevalence rate of 12.1% and 10.0 claims per user per year reported by a PBM for 200926 to calculate an estimated VBID cost of approximately $29 PMPY (12.1% × 10.0 × $24). For the VBID group in the study by Maciejewski et al., the cost would be lower because the intervention consisted of a generic copayment waiver; however, the estimate of VBID cost would be somewhat misleading because in both the VBID and com-parison groups, all tier 3 brand drugs in the study drug classes were set to a tier 2 copayment. Putting aside that limitation, and assuming that the pre-intervention generic dispensing ratio for statins (38%) was unchanged in the post-intervention period, the cost of waiving the approximately $11 generic copayment in the study by Maciejewski et al. would be $5.06 PMPY (38% × 12.1% × 10.0 × $11).

Three key points about these calculations are noteworthy for payers. First, these figures illustrate the conclusion reached by Melnick and Motheral based on their VBID plausibility cal-culator that mathematically, it is nearly impossible for a health plan sponsor to achieve cost neutrality in VBID programs using medical cost offsets, with the possible exception of programs applied only to generic medications.42 Assuming that a private payer incurs a cost of approximately $60,000 per patient for a major cardiovascular event (in 2006, costs per hospital stay billed to private insurance were $44,733 for coronary artery dis-ease, $54,697 for MI, and $44,239 for stroke),79 achieving cost neutrality for a VBID program that costs $29 PMPY in a plan with 100,000 members would require prevention of an addi-tional 48 cardiovascular events per year ($2.9 million divided by $60,000). Thus, using clinical trial data as indicators of the baseline effectiveness of statins in preventing major cardiovas-cular events (Table 2)80-82 for the health plan’s assumed 12,100 statin users (use prevalence of 12.1% × 100,000) and assuming a baseline MPR of 50%-80%, a VBID copayment reduction program that costs $29 PMPY would have to increase the effectiveness of statin treatment by approximately 41%-49% in secondary prevention (e.g., an increase of 48 over a baseline of 97 to 118 patients in whom events are prevented annually) and 68%-79% in primary prevention—despite increasing medica-tion adherence by only about 4%-6% (3 percentage points improvement over a baseline MPR of 50%-80%) to break even. As VBID proponents have observed, additional cost offsets

especially notable in this regard because its findings strongly suggest that the “effects” of the program were actually due to confounding. Specifically, the study found that among DMP participants, compliance with recommended screening and monitoring (i.e., primary care provider visits, hemoglobin A1c tests, lipid tests, and urinalysis) was higher for VBID enroll-ees than for comparison group patients.71 However, the study report indicated that the VBID program changed pharmacy copayments only. There is no logically apparent causal linkage between pharmacy copayments and receipt of guideline-based medical care, and the study authors offered none in the report. Like the finding of Dormuth et al. (2009) that patients com-pliant with statin therapy were less likely than noncompliant patients to have automobile and workplace accidents,76 the findings of Gibson et al. sound a cautionary note for those who attempt to equate association with causation; the resulting con-clusions may be biologically or logically implausible.

In contrast, an observational analysis by Maciejewski et al. (2010) is commendable for its inclusion of a comparison group of patients treated with therapy classes in which copayments were reduced in both the VBID intervention and comparison groups. The authors made the reasonable case that the lack of VBID “effect” in these classes, coupled with the positive asso-ciation of VBID with adherence in the other study drug classes, strengthened the internal validity of their findings.63 A study by Choudhry et al. of copayment elimination for statins and copayment reduction for clopidogrel, effective January 1, 2007, appears to have good internal validity;68 however, the interven-tion was conducted among enrollees whose employer coupled widely publicized copayment reductions with intensive educa-tional interventions in 2002,77 making the context and external validity of the 2007 intervention unclear.

Costs and Benefits of VBID InterventionsThe 2 new VBID studies with the strongest internal validity, one by Choudhry et al. and the other by Maciejewski et al., suggest a VBID effect size of approximately 1.5 to 4 percentage points in adherence measured by MPR or proportion of days covered (PDC).63,68 Effect sizes of about 2.6 to 4.0 percentage points annually were reported previously by Chernew et al. (2008) for copayment reductions (for tier 1/tier 2/tier 3) from $5/$25/$45 to $0/$12.50/$22.50 for diabetes drugs, statins, and antihypertensive drugs (ACE inhibitors and angiotensin II receptor blockers [ARBs]).78 Despite the statistical significance of MPR or PDC changes of 2 to 4 percentage points measured using large samples, these changes represent only 7 to 15 added days of therapy per year on average (e.g., 0.02 X 365 = 7.3 days).

The costs and benefits of these small adherence improve-ments have not yet been reported. In a previous editorial on the study by Chernew et al., we used national data to compensate for missing information in the study report and estimated that

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

164 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

changes of 2.6 to 4.0 percentage points, Chernew et al. predicted that “because clinical evidence supports adherence to these medications, we expect health improvements, although we do not quantify them in this study.”78 In an accompany-ing press release from the sponsoring institution, the study’s lead author speculated that “while future studies need to be done to actually quantify this specifically, there is considerable evidence that use of the classes of medication in this study will reduce the frequency of adverse clinical events and associated hospitalizations and ER visits.”15 However, about 2 years later, the research team reported that “preliminary statistical analysis of the spending data” for the VBID had “indicated considerable uncertainty surrounding estimates of the impact of the [VBID] intervention on aggregate spending,” preventing econometric analyses of the data and prompting the investigators to develop a pharmacoeconomic model instead of analyzing the actual observed outcomes for the sample.43 Unfortunately, no data on hospitalizations or ER visits—consequences of the “adverse clinical events” that the investigators had suggested would be prevented by the VBID—were reported.

Similarly, an analysis plan for a study of the MHealthy ini-tiative, a VBID program that was implemented for University of Michigan employees and their dependents with diabetes on July 1, 2006, was published in April 2009.83 The analysis plan indicated that study follow-up would end 30 months

from improved worker productivity (e.g., reduced absenteeism and disability days), although not easily measured, are possible for active employee groups;43 however, it is difficult to imagine that these effects would be substantial given the small size of the medication adherence effects.

Second, the results of the Maciejewski et al. analysis suggest, consistent with PBM observations, that the adherence effects of dispensing medication in 90-day supplies may overwhelm those of VBID. In the Maciejewski et al. analysis, enrollees with at least 1 pharmacy claim for a 90-day supply had adherence rates 17 to 22 percentage points higher than those of patients without any 90-day supplies.63 Notably and unfortunately, none of the other new VBID studies reported utilization rates of 90-day supplies or mail order pharmacy or analyzed the effect of 90-day supplies on medication adherence.

Third, the need to calculate these “back-of-the-envelope” estimates from published research articles highlights deficien-cies in the base of evidence about copayment reduction pro-grams, especially for payers that need actionable, quantitative estimates of the impact of adherence promotion strategies on economic and clinical outcomes.

What to Expect When You’re Expecting Information About VBID In their 2008 report of the association of VBID with MPR

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

TABLE 2 Effectiveness of Statin Treatment in Preventing Major Cardiovascular Events in a Hypothetical 100,000-Member Health Plan

Study

Description of Events Avoided

by Statin Treatment

Per-Patient Probability of

Having at Least 1 Avoided Event During Follow-

Up Period

Length of Follow-Up

Period (Years)

Annual Probability

of Event Avoidancea

Number of Treated Patientsb

Annual Number of

Patients with at Least 1

Avoided Eventc

LIPID Secondary Prevention Trial81

Events (30 deaths, 28 nonfatal MIs, and 9 nonfatal strokes) were avoided in an unduplicated 48 patients for every 1,000 treated.

0.048 6.1 0.0080 12,100 97

Meta-analysis of all statin users, secondary prevention82

Events were avoided in 48 patients per 1,000 treated.

0.048 5 0.0098 12,100 118

Meta-analysis of all statin users, primary prevention82

Events were avoided in 25 patients per 1,000 treated.

0.025 5 0.0051 12,100 61

JUPITER, primary prevention80

0.59 composite endpoint events per 100 person-years of follow-up.d

NA NA 0.0059 12,100 71d

aAnnual rate is derived algebraically from the cumulative probability formula: C = 1–(1–a)n, where C = cumulative probability, a = annual probability, and n =number of years84—that is, a = 1–(1–C)1/n. bNumber of treated patients is derived from 12.1% annual statin use prevalence26 times assumed health plan population of 100,000 enrollees.cNumber of treated patients × annual avoided event rate. Annual probabilities shown in table are rounded to the fourth decimal.dFor the first 3 rows of the table (the meta-analysis and the LIPID trial), the final outcome reflects the estimated number of patients in whom at least 1 event is avoided. For the last row of the table ( JUPITER), the final outcome reflects the estimated number of avoided events.JUPITER = Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin; LIPID = Long-Term Intervention with Pravastatin in Ischaemic Disease; MI = myocardial infarction; NA=not applicable.

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 165

(an effect that is captured in the total cost measure), and (b) the reduction in copayment (an effect that is not captured in the total cost measure). Because of the inelasticity (lack of price responsiveness) in drug purchasing behavior in commercially insured groups, a calculation of total cost change in a VBID program will produce a small estimated effect that understates the actual cost of VBID to the payer. Yet, only 1 of the 3 reports, the economic “break-even” analysis by Chernew et al.,43 trans-parently disclosed in the study report the implications of the analytic perspective for plan sponsors. Thus, when provided with information about VBID costs and benefits, health plan sponsors should check carefully to determine if payer costs were measured.

Overextension. Completion of the first randomized trial of a copayment reduction intervention is expected in early 2011.11 That study, the Post-MI FREEE (Free Rx and Economic Evaluation), is a test of eliminating copayments for secondary prevention medications (statins, beta-blockers, ACE inhibitors, and ARBs) for patients aged 64 years or younger following hospitalization for MI.85 Because the baseline probability of adverse cardiovascular events is higher in a patient population with a previous MI than in the population of cardiovascular medication users as a whole, the likelihood of finding clini-cal benefits is higher in this patient sample than in a typical patient population in which medications are used both for primary and secondary prevention. Yet, an unfortunately com-mon occurrence in health policy research is the overextension of findings from a study sample manifesting one set of charac-teristics to a larger population with completely different clinical or demographic characteristics. We have previously pointed out the frequent inappropriate extensions of a study conducted in a small group of low-income elderly with numerous chronic illnesses to healthier patient populations and to different out-comes—even to outcomes that were not studied in the original research.86 In assessing the information presented in popular press accounts and literature reviews, payers should compare the assertions made about the Post-MI FREEE findings with the study report—or the study abstract for decision makers who lack the time to read the entire report—to determine if findings are represented accurately or misstated.

VBID: What Should Health Plan Sponsors Do Now?We predicted in 2008 that managed care was approaching a crossroads at which policymakers would have to decide whether to reject cost-sharing policies based on high-quality evidence in favor of weaker, but more vigorously promoted, evidence.58 That issue has become more acute and ethi-cally important with the recent suggestions that the costs of drug copayment reductions—often for primary prevention using brand medications in therapeutic classes with generic alternatives—should be offset by increasing cost-sharing

after implementation, or approximately January 2009, and that the analysis would assess the primary outcomes of medication utilization and adherence, as well as secondary outcomes including health care costs and utilization rates for outpatient visits, ER visits, and hospitalizations.83 Although the MHealthy study follow-up was scheduled to end more than 2 years ago at this writing, to date the only reported findings of which we are aware include a brief mention, reported in a 2010 review article, of “preliminary results” indicating that the MHealthy intervention was associated with “a 7 percent increase in adherence to blood pressure-lowering medication and a nonsignificant 4 percent increase in adherence to statins.”11

Against this shortfall of evidence promised but not yet reported, several reasons for a “caveat emptor” approach to VBID research are apparent, including the following key points.

Isolated Significant Findings. Because of cumulative type 1 (false positive) error, it is mathematically expected that at least 1 finding in a series of numerous statistical tests will be statisti-cally significant solely because of chance. For example, in 20 statistical tests at an alpha of 0.05, the probability of at least 1 false positive finding is 64%.84 In the presence of publication bias, these multiple tests are not necessarily apparent to readers of a single research article. We encourage health plan sponsors to be mindful of the history of VBID research reporting (or nonreporting) in considering the results reported in any single publication, especially for hospitalizations and ER visits.

Causal Linkages. In viewing published results of observational studies of VBID programs, payers should consider whether the findings are plausible. For example, one would not expect large medical cost offsets from tiny medication adherence gains, large adherence gains from small cost-sharing changes, or changes in a measure of quality of care that is not affected by pharmacy copayments (e.g., receipt of guideline-based medical care, such as examinations and laboratory tests). Although understanding complex nuances of methodological and statistical techniques may be difficult for managed care decision makers who lack research training, consideration of basic clinical and practi-cal causal logic is not difficult and will go a long way toward assessing the validity of VBID studies.

Total Cost Versus Health Plan Sponsor Cost: A Major Difference for Payers. To date, 3 assessments of costs associ-ated with VBID have either measured70,71 or estimated43 change in total cost, including both the payer cost and the patient cost share. Although this method is not inherently erroneous because it is common for health policy research to examine cost outcomes from a societal perspective, it produces a result that is uninformative for health plan sponsors making a decision about VBID. The financial effects of VBID on plan sponsors are the sum of (a) increased medication adherence

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

166 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

requirements for patients using expensive treatments for seri-ous or catastrophic illness. Assessment both of the evidence and, more importantly, the limitations of the evidence regard-ing VBID suggests 4 important steps going forward.

First, as Choudhry et al. observed in their recent review, important questions about the effects of drug copayment reductions on clinical and economic outcomes “should be answered before [VBID] is used more widely.”11 Because VBID has been associated with only minimal medication adherence increases documented only in observational research, and because no health or medical utilization outcomes (e.g., ER or hospital use) have yet been reported for VBID programs, the evidence is insufficient to support expanding its use at the present time.

Second, managed care decision makers should refuse to accept studies that lack transparent reporting necessary to make even basic estimates of the costs and benefits of VBID.Nontransparent study reports do not provide useful informa-tion and should not be used for policymaking. For promul-gators of research information, the logical corollary is that continuing to publish suboptimal work will eventually cause decision makers to be indifferent to study findings—exactly the opposite of the goal that the AHRQ has set for evidence-based decision making in health care policy.

Third and related, although it would seem that health care decision makers should not have to be responsible for critical appraisal of the information provided in peer-reviewed articles, it appears that such critique is unfortunately necessary in cost-sharing research because of a pattern of inaccurate statements about research findings. In reading literature reports about VBID, health care decision makers will find it helpful to read the source materials—at least the study abstract and the results tables if time is short—instead of relying solely on secondary descriptions of what previous research has shown. Summary study tables including critical appraisal of the evidence, such as those that we have reported here and in previous JMCP issues,58 as well as that reported by Motheral in the current issue87 are also helpful, and we encourage other health policy researchers to provide summaries of this type for health care decision makers.

Fourth, and most important from a policy perspective, deliberations about whether to fund drug copayment reduc-tions, especially for primary prevention, by directing scarce resources away from a subset of prescribed medical and phar-maceutical treatments should demand high-quality evidence. Yet at this writing, nearly a decade after pharmacy benefit copayment reduction was initially proposed as a means to improve the outcomes of patients with chronic health condi-tions, health plan sponsors have little of the information that they need to assess the costs and benefits of VBID. Available evidence is of suboptimal quality, much of it so opaque that it is uninterpretable; outcomes critically important to enrollees and

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

KATHLEEN A. FAIRMAN, MA, is Associate Editor and Senior Methodology Reviewer, and FREDERIC R. CURTISS, PhD, RPh, CEBS, is Editor-in-Chief of the Journal of Managed Care Pharmacy.

AUTHOR CORRESPONDENCE: Kathleen A. Fairman, MA, Academy of Managed Care Pharmacy, 100 North Pitt St., Suite 400, Alexandria, VA 22314. Tel.: 602.867.1343; E-mail: kfairman@amcp.org.

Authors

health plan sponsors, such as cost to the payer and effects on generic drug utilization and patient health, remain either unex-amined or unreported; and the only randomized trial of VBID of which we are aware, the Post-MI FREEE, is laudable for its design and transparency but limited to secondary prevention. An evidence base that falls so far short of the AHRQ’s ideal for health care policy research—timely, clear, transparent, solid, and actionable—should be viewed as an example of how not to inform health care system change.

DISCLOSURES

The authors report no financial conflicts of interest related to the subjects or products discussed in this article. Fairman reports that she works and vol-unteers a few hours each month in nursing homes, primarily with bedbound patients who have catastrophic illnesses.

REFERENCES

1. Agency for Healthcare Research and Quality. Filling the information needs for healthcare reform: expert meeting summary and identification of next steps. Available at: http://www.ahrq.gov/data/hinfosum.htm. Accessed January 24, 2011.

2. Diamond Management & Technology Consultants. The Health Care Reform Survey 2010. Willis. November 10, 2010. Available at: http://www.willis.com/documents/publications/Services/Employee_Benefits/Health_Care_Reform_Survey_2010_V9.pdf. Accessed January 24, 2011.

3. Mattioli D. Firms feel pain from health law. WSJ. December 13, 2010. Available at: http://online.wsj.com/article/SB10001424052748703727804576011791786531256.html. Accessed December 13, 2010.

4. The Kaiser Family Foundation and Health Research & Educational Trust. Employer Health Benefits 2010 Annual Survey. Available at: http://ehbs.kff.org/pdf/2010/8085.pdf. Accessed November 23, 2010.

5. Fairman KA, Curtiss FR. Selling real-world health care research to reluc-tant buyers: evidence-based education or marketing a defective product? J Manag Care Pharm. 2009;15(3):294-96. Available at: http://www.amcp.org/data/jmcp/294-296.pdf. Accessed November 23, 2010.

6. De Brantes F, Wickland PS, Williams JP. The value of ambulatory care measures: a review of clinical and financial impact from an employer/payer perspective. Am J Manag Care. 2008;14(6):360-68. Available at: http://www.ajmc.com/media/pdf/AJMC_08jun_Brantes360to68.pdf. Accessed November 23, 2010.

7. Lee DH, Vielemeyer O. Analysis of overall level of evidence behind Infectious Diseases Society of America practice guidelines. Arch Intern Med. 2011;171(1):18-22.

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 167

25. Walgreens Health Initiatives. Pharmacy benefit solutions: 2010 trend report. Available at: http://www.walgreenshealth.com/pdf/newsletterreport/PBS12843_TrendReport_web2.pdf. Accessed December 14, 2010.

26. Miller S, Cox E, Nease B, et al. 2009 Drug trend report: solving for America’s $163 billion in pharmacy-related waste. April 2010. Available at: http://www.express-scripts.com/research/studies/drugtrendreport/2009/dtrFinal.pdf. Accessed January 24, 2011.

27. Lofberg P, Brennan T, Greer L, et al. Insights 2010: evolving pharmacy care. Available at: https://www.caremark.com/portal/asset/CVSCaremark_Insights_2010.pdf. Accessed January 24, 2011.

28. Prime Therapeutics. 2010 drug trend insights: where smart pharmacy benefit management meets results. Available at: http://www.primetherapeu-tics.com/pdf/2010primedrugtrends.pdf. Accessed January 24, 2011.

29. Chang A, Liberman JN, Coulen C, Berger JE, Brennan TA. Value-based insurance design and antidiabetic medication adherence. The American Journal of Pharmacy Benefits. 2010;2(1):39-45.

30. Gleason PP, Starner CI, Gunderson BW, Schafer JA, Sarran HS. Association of prescription abandonment with cost share for high-cost spe-cialty pharmacy medications. J Manag Care Pharm. 2009;15(8). Available at: http://www.amcp.org/data/jmcp/648-658.pdf.

31. Roebuck MC, Liberman JN, Gemmill-Toyama M, Brennan TA. Medication adherence leads to lower health care use and costs despite increased drug spending. Health Aff (Millwood). 2011;30(1):91-98.

32. Fairman KA. What pharmacy benefit designers need to know about per-ception and reality: never forget the elephant in the pharmacy. J Manag Care Pharm. 2008;14(4):387-94. Available at: http://www.amcp.org/data/jmcp/JMCPMaga_387-394.pdf.

33. Fitch K, Iwasaki K, Pyenson B. Value-based insurance designs for dia-betes drug therapy: actuarial and implementation considerations. Prepared for Takeda Pharmaceuticals, North America, Inc., by Milliman. December 1, 2008. Available at: http://publications.milliman.com/research/health-rr/pdfs/vbid-diabetes-drug-therapy-RR12-01-08.pdf. Accessed January 24, 2011.

34. Federal Trade Commission. FTC issues report on PBM ownership of mail-order pharmacies. September 6, 2005. Available at: http://www.ftc.gov/opa/2005/09/pharmbenefit.shtm. Accessed February 7, 2011.

35. Doru OK, Schmittdiel JA, Dyer WT, et al. Mail-order pharmacy use and adherence to diabetes-related medications. Am J Manag Care. 2010;16(1):33-40. Available at: http://www.ajmc.com/media/pdf/AJMC_2010Jan_Duru_33to40.pdf. Accessed December 16, 2010.

36. Devine S, Vlahiotis A, Sundar H. A comparison of diabetes medication adherence and healthcare costs in patients using mail order pharmacy and retail pharmacy. J Med Econ. 2010;13(2):203-11.

37. Brown TA, Rickles NM. Mail-order versus local pharmacies on adher-ence: study methods make for unfair comparison [letter and author reply]. Am J Manag Care. 2010;16(3):226-28. Available at: http://www.ajmc.com/media/pdf/AJMC_2010marLtrToEd_226to228.pdf. Accessed December 16, 2010.

38. Matchar DB, McCrory DC, Orlando LA, et al. Comparative effectiveness of angiotensin-converting enzyme inhibitors (ACEIs) and angiotensin II receptor antagonists (ARBs) for treating essential hypertension. November 2007. Available at: http://www.effectivehealthcare.ahrq.gov/ehc/prod-ucts/12/45/ACEI_ARBFullReport.pdf. Accessed January 13, 2011.

39. Bolen S, Wilson L, Vassy J, et al. Comparative effectiveness and safety of oral diabetes medications for adults with type 2 diabetes. Agency for Healthcare Research and Quality. July 2007. Available at: http://www.effec-tivehealthcare.ahrq.gov/repFiles/OralFullReport.pdf. Accessed January 13, 2011.

40. Choudhry NK, Shrank WH. Four dollar generics—increased acces-sibility, impaired quality assurance. N Engl J Med. 2010;363(20):1885-87. Available at: http://www.nejm.org/doi/full/10.1056/NEJMp1011625. Accessed January 25, 2011.

8. Fairman KA, Curtiss FR. Still looking for health outcomes in all the wrong places? Misinterpreted observational evidence, medication adher-ence promotion, and value-based insurance design. J Manag Care Pharm. 2009;15(6):501-07. Available at: http://www.amcp.org/data/jmcp/501-507.pdf.

9. Motheral B. Value-based insurance design: a bargain at $10 per day? Rx Outcomes Adviser. December 15, 2010. Available at: http://rxoutcomesad-viser.wordpress.com/2010/12/15/value-based-insurance-design-a-bargain-at-10-per-day/. Accessed January 24, 2011.

10. Motheral B. About Rx Outcomes Adviser. Available at: http://rxoutcome-sadviser.wordpress.com/about/. Accessed January 24, 2011.

11. Choudhry NK, Rosenthal MB, Milstein A. Assessing the evidence for value-based insurance design. Health Aff (Millwood). 2010;29(11):1988-94.

12. Fendrick AM, Smith DG, Chernew ME, Shah SN. A benefit-based copay for prescription drugs: patient contribution based on total benefits, not drug acquisition cost. Am J Manag Care. 2001;7(9):861-67. Available at: http://www.ajmc.com/media/pdf/AJMC2001sepFendrick861_867.pdf. Accessed January 24, 2011.

13. University of Michigan, Center for Value-Based Insurance Design. Funding. Available at: http://www.sph.umich.edu/vbidcenter/about/funding.html. Accessed January 24, 2011.

14. Fendrick AM, Edlin ML. Value-based insurance design landscape digest. National Pharmaceutical Council. July 2009. Available at: http://www.vbhealth.org/wp-content/uploads/2009/08/NPC_VBIDreport_7-22-09.pdf. Accessed December 10, 2010.

15. University of Michigan Department of Public Relations and Marketing Communications. The co-pay connection: lowering drug co-pays for chronic disease patients increases use of important preventive medicines, rigorous study shows. January 8, 2008. Available at: http://www.med.umich.edu/opm/newspage/2008/drugcopay.htm. Accessed January 23, 2011.

16. Fuhrmans V. New tack on copays: cutting them. WSJ. May 8, 2007:D1. Available at: http://www.sph.umich.edu/vbidcenter/news/pdfs/WSJ%20Copays%20May%208%20(2).pdf. Accessed January 25, 2011.

17. Hensley S. Chopping copays makes medicines more attrac-tive. NPR. November 2, 2010. Available at: http://www.npr.org/blogs/health/2010/11/02/130998493/chopping-copays-makes-statins-plavix-more-attractive. Accessed January 24, 2011.

18. National Business Group on Health/Towers Watson. Raising the bar on health care: moving beyond incremental change. 2010. Available at: http://www.towerswatson.com/assets/pdf/1345/TW_15565_NBGH.pdf. Accessed January 24, 2011.

19. Andrews M. doctors play crucial role in tying health care choices to value. Shots—NPR’s Health Blog. November 30, 2010. Available at: http://www.npr.org/blogs/health/2010/11/30/131690967/doctors-play-crucial-role-in-tying-health-care-choices-to-value. Accessed January 24, 2011.

20. Blanton K. Health care: spend more, save more. CFO.com. November 18, 2010. Available at: http://www.cfo.com/article.cfm/14540665/c_14540734. Accessed December 10, 2010.

21. Kertesz L. Economic problems put brakes on Rx copay waiver trend: senior management wait for more evidence of health care savings. Business Insurance. May 11, 2009. Available at: http://www.businessinsurance.com/apps/pbcs.dll/article?AID=9999100027607&template=printart. Accessed February 12, 2011.

22. Anonymous. Health benefit cost growth accelerates to 6.9% in 2010. Mercer. November 17, 2010. Available at: http://www.mercer.com/press-releases/1400235. Accessed December 10, 2010.

23. Reinke T. Large PBMs transform old business models. Manag Care. 2009;18(10):20-21, 25-28. Available at: http://www.managedcaremag.com/archives/0910/0910.pbms.html. Accessed December 14, 2010.

24. Medco. 2010 drug trend report: Solving the healthcare cost/quality equation. Available at: http://www.drugtrend.com/art/drug_trend/pdf/DT_Report_2010.pdf. Accessed December 14, 2010.

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

168 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

57. Fairman KA, Curtiss FR. Making the world safe for evidence-based policy: let’s slay the biases in research on value-based insurance design. J Manag Care Pharm. 2008;14(2):198-204. Available at: http://www.amcp.org/data/jmcp/JMCPMaga_March%2008_198-204.pdf.

58. Fairman KA. The future of prescription drug cost-sharing: real progress or dropped opportunity? J Manag Care Pharm. 2008;14(1):70-82. Available at: http://www.amcp.org/data/jmcp/JMCPMaga_JanFeb%2008_070-082.pdf.

59. Fairman KA, Motheral BR, Henderson RR. Retrospective, long-term follow-up study of the effect of a three-tier prescription drug copayment system on pharmaceutical and other medical utilization and costs. Clin Ther. 2003;25(12):3147-66.

60. Motheral B, Fairman KA. Effect of a three-tier prescription copay on pharmaceutical and other medical utilization. Med Care. 2001;39(12):1293-304.

61. Huskamp HA, Deverka PA, Epstein AM, Epstein RS, McGuigan KA, Frank RG. The effect of incentive based formularies on prescription-drug utilization and spending. N Engl J Med. 2003;349(23):2224-32. Available at: http://www.nejm.org/doi/full/10.1056/NEJMsa030954.

62. Sedjo RL, Cox ER. Lowering copayments: impact of simvastatin pat-ent expiration on patient adherence. Am J Manag Care. 2008;14(12):813-18. Available at: http://www.ajmc.com/media/pdf/AJMC_08dec_Sedjo_813to818.pdf. Accessed December 28, 2010.

63. Maciejewski ML, Farley JF, Parker J, Wansink D. Copayment reduc-tions generate greater medication adherence in targeted patients. Health Aff (Millwood). 2010;29(11):2002-08.

64. Chernew ME, Newhouse JP. What does the RAND Health Insurance Experiment tell us about the impact of patient cost sharing on health out-comes? Am J Manag Care. 2008;14(7):412-14. Available at: http://www.ajmc.com/media/pdf/AJMC_08jul_Chernew_412to414_.pdf. Accessed January 23, 2011.

65. Manning WG, Newhouse JP, Duan N, Keeler EB, Liebowitz A, Marquis MS. Health insurance and the demand for medical care: evidence from a randomized experiment. Am Econ Rev. 1987;77(3):251-77.

66. Brook RH, Ware JE, Rogers WH, et al. The effect of coinsurance on the health of adults: results from the RAND Health Insurance Experiment. The RAND Corporation. R-3055-HHS. 1984. Available at: http://www.rand.org/pubs/reports/R3055.html. Accessed February 9, 2011.

67. Keeler EB. Effects on cost sharing on use of medical services and health. J Med Pract Manage. 1992;8:317-21. Available at: http://www.rand.org/con-tent/dam/rand/pubs/reprints/2005/RP1114.pdf. Accessed February 7, 2011.

68. Choudhry NK, Fischer MA, Avorn J, et al. At Pitney Bowes, value-based insurance design cut copayments and increased drug adherence. Health Aff (Millwood). 2010;29(11):1995-2001.

69. Zeng F, An JJ, Scully R, Barrington C, Patel BV, Nichol MB. The impact of value-based benefit design on adherence to diabetes medications: a pro-pensity score-weighted difference in difference evaluation. Value Health. 2010;13(6):846-52.

70. Gibson TB, Wang S, Kelly E, et al. A value-based insurance design pro-gram at a large company boosted medication adherence for employees with chronic illnesses. Health Aff (Millwood). 2011;30(1):109-17.

71. Gibson TB, Mahoney J, Ranghell K, Cherney BJ, McElwee N. Value-based insurance plus disease management increased medication use and produced savings. Health Aff (Millwood). 2011;30(1):100-08.

72. Vandenbroucke JP, von Elm E, Altman DG, et al. Strengthening the reporting of observational studies in epidemiology (STROBE): explanation and elaboration. Ann Intern Med. 2007;147:W-163-W-194. Available at: http://www.annals.org/cgi/reprint/147/8/W-163.pdf. Accessed January 15, 2011.

73. National Cholesterol Education Program. ATP III guidelines at-a-glance quick desk reference. Available at: http://www.nhlbi.nih.gov/guidelines/cho-lesterol/atglance.pdf. Accessed January 15, 2011.

41. Fendrick AM, Smith DG, Chernew ME. Applying value-based insurance design to low-value health services. Health Aff (Millwood). 2010;29(11):2017-21.

42. Melnick SJ, Motheral BR. Is “value-based” value wasted? Examining val-ue-based insurance designs through the lens of cost-effectiveness. J Manag Care Pharm. 2010;16(2):130-33. Available at: http://www.amcp.org/data/jmcp/130-133.pdf.

43. Chernew ME, Juster IA, Shah M, et al. Evidence that value-based insur-ance can be effective. Health Aff (Millwood). 2010;29(3):530-36.

44. Robinson JC. Applying value-based insurance design to high-cost health services. Health Aff (Millwood). 2010;29(11):2009-16.

45. U.S. Preventive Services Task Force. Screening for breast cancer. November 2009, updated July 2010. Available at: http://www.uspreventi-veservicestaskforce.org/uspstf/uspsbrca.htm. Accessed January 25, 2011.

46. Hobson K. Prevention task force cancels Nov. meeting; would have included prostate screening vote. WSJ Health Blog. October 26, 2010. Available at: http://blogs.wsj.com/health/2010/10/26/prevention-task-force-cancels-nov-meeting-would-have-included-prostate-screening-vote/. Accessed January 25, 2011.

47. American Cancer Society. Breast awareness and self-exam. September 15, 2010, revised January 4, 2011. Available at: http://www.cancer.org/Cancer/BreastCancer/MoreInformation/BreastCancerEarlyDetection/breast-cancer-early-detection-acs-recs-bse. Accessed January 25, 2011.

48. American Cancer Society. American Cancer Society recommenda-tions for prostate cancer early detection. December 1, 2010. Available at: http://www.cancer.org/Cancer/ProstateCancer/MoreInformation/ProstateCancerEarlyDetection/prostate-cancer-early-detection-acs-recom-mendations. Accessed January 25, 2011.

49. Brawley OW, Gansler T. Introducing the 2010 American Cancer Society prostate cancer screening guideline. CA Cancer J Clin. 2010;60(2):68-69. Available at: http://caonline.amcancersoc.org/cgi/reprint/60/2/68. Accessed January 25, 2011.

50. U.S. Preventive Services Task Force. Screening for breast cancer: U.S. Preventive Services Task Force recommendation statement. Ann Intern Med. 2009;151(10):716-26. Available at: http://www.uspreventiveservicestaskforce.org/uspstf09/breastcancer/brcanrs.pdf. Accessed January 25, 2011.

51. American Cancer Society. American Cancer Society guidelines for the early detection of cancer. Available at: http://www.cancer.org/Healthy/FindCancerEarly/CancerScreeningGuidelines/american-cancer-society-guidelines-for-the-early-detection-of-cancer. Accessed January 25, 2011.

52. American College of Radiology. Detailed ACR statement on ill advised and dangerous USPSTF mammography recommendations. Undated docu-ment. Available at: http://www.acr.org/MainMenuCategories/media_room/FeaturedCategories/PressReleases/UPSTFDetails.aspx. Accessed January 6, 2011.

53. Harris G. F.D.A to restrict Avandia, citing heart risk. NY Times. September 23, 2010. Available at: http://www.nytimes.com/2010/09/24/health/policy/24avandia.html?_r=2. Accessed January 25, 2011.

54. U.S. Food and Drug Administration. FDA significantly restricts access to the diabetes drug Avandia. September 23, 2010. Available at: http://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm226975.htm. Accessed January 25, 2011.

55. Appleby J. Carrot-and-stick health plans aim to cut costs: potentially controversial policies are part of trend toward ‘value-based design.’ Kaiser Health News. March 11, 2010. Available at: http://www.kaiserhealthnews.org/Stories/2010/March/11/value-based-health-insurance.aspx. Accessed January 25, 2011.

56. Kapowich JM. Oregon’s test of value-based insurance design in coverage for state workers. Health Aff (Millwood). 2010;29(11):2028-32.

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 169

81. Anonymous. Prevention of cardiovascular events and death with pravas-tatin in patients with coronary heart disease and a broad range of initial cholesterol levels. The Long-Term Intervention with Pravastatin in Ischaemic Disease (LIPID) Study Group. N Engl J Med. 1998;339(19):1349-57. Available at: http://www.nejm.org/doi/full/10.1056/NEJM199811053391902. Accessed January 24, 2011.

82. Baigent C, Keech A, Kearney PM, et al. Efficacy and safety of cholesterol-lowering treatment: prospective meta-analysis of data from 90,056 partici-pants in 14 randomised trials of statins. Lancet. 2005;366(9493):1267-78.

83. Spaulding A, Fendrick AM, Herman WH, et al. A controlled trial of val-ue-based insurance design—the MHealthy: Focus on Diabetes (FOD) trial. Implement Sci. 2009;4:19. Available at: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2673203/pdf/1748-5908-4-19.pdf. Accessed January 18, 2011.

84. The formula for calculating the cumulative probability of type 1 error is: C=1–(1–a)n, where C=cumulative probability, a = single-test alpha, and n=number of tests. Kaul S, Diamond GA. Trial and error. How to avoid com-monly encountered limitations of published clinical trials. J Am Coll Cardiol. 2010;55(5):415-27.

85. Choudhry NK, Brennan T, Toscano M, et al. Rationale and design of the Post-MI FREEE trial: a randomized evaluation of first-dollar drug coverage for post-myocardial infarction secondary preventive therapies. Am Heart J. 2008;156(1):31-36. Available at: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2697130/pdf/nihms78751.pdf. Accessed January 18, 2011.

86. Fairman KA. Accuracy in pharmacoeconomic literature review: lessons learned from the Navajo Code Talkers. J Manag Care Pharm. 2008;14(9):886-91. Available at: http://www.amcp.org/data/jmcp/886-891.pdf.

87. Motheral BR. Pharmaceutical step-therapy interventions: a critical review of the literature. J Manag Care Pharm. 2011;17(2):143-55. Available at: http://www.amcp.org/data/jmcp/143-155.pdf.

74. U.S. Department of Health and Human Services, National High Blood Pressure Education Program. JNC 7 express: the seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure. December 2003. Available at: http://www.nhlbi.nih.gov/guidelines/hypertension/express.pdf. Accessed January 15, 2011.

75. Nathan DM, Buse JB, Davidson MB, et al. Medical management of hyperglycemia in type 2 diabetes: a consensus algorithm for the initiation and adjustment of therapy. Diabetes Care. 2009;32(1):193-203. Available at: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2606813/pdf/193.pdf. Accessed January 15, 2011.

76. Dormuth CR, Patrick AR, Shrank WH, et al. Statin adherence and risk of accidents: a cautionary tale. Circulation. 2009;119(15):2051-57. Available at: http://circ.ahajournals.org/cgi/reprint/119/15/2051. Accessed January 18, 2011.

77. Mahoney JJ. Reducing patient drug acquisition costs can lower diabe-tes health claims. Am J Manag Care. 2005;11(5 Suppl):S170-S176. Available at: http://www.ajmc.com/media/pdf/A128_05augMahoneyS170to176.pdf. Accessed December 10, 2010.

78. Chernew ME, Shah MR, Wegh A, et al. Impact of decreasing copayments on medication adherence within a disease management environment. Health Aff (Millwood). 2008;27(1):103-12.

79. Andrews RM. The national hospital bill: the most expensive conditions by payer, 2006. HCUP statistical brief #59. September 2008. Available at: http://www.hcup-us.ahrq.gov/reports/statbriefs/sb59.pdf. Accessed January 17, 2011.

80. Ridker PM, Danielson E, Fonseca FA, et al.; JUPITER Study Group. Rosuvastatin to prevent vascular events in men and women with elevated c-reactive protein. N Engl J Med. 2008;359(21):2195-207. Published online ahead of print on Nov 9, 2008. Available at: http://www.nejm.org/doi/full/10.1056/NEJMoa0807646. Accessed January 23, 2011.

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

Editors’ note to online readers: All JMCP articles contain hyperlinks to the source documents for free-access references. These hyperlinks are embedded in the reference numbers cited in the text as well as in the list of references at the end of each article.

170 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

Summary of Recent Observational Studies of Copayment Reductions and Medication Adherencea

Chang (2010)29 Overview: Study of copayment reductions for antidiabetic medications—free generics and insulin, reductions in copayments for tier 2 brand drugs, no change for tier 3, effective January 1, 2007; study period was January 1, 2006, through December 31, 2007.

Sample Enrollees with at least 1 claim for an antidiabetic medication during the study period, classified as initiators (no claims in 2006) and continuers/discontinuers (use in 2006 including final 45 days of the year); all were continuously enrolled throughout the study period.

Groups Intervention: 3 PBM clients without concurrent educational or DSM programs; n = 20,173 (however, outcomes were reported for only 2,538 patients). Copayments changed from $15 to $0 for generic; $30 to a range of $10-$15 for tier 2; a range of $30-$35 for tier 3 in both periods; mean copayments changed from $28 to $0 for insulin and from $21 to $10 for OADs.Comparison: PBM clients without VBID; n = 190,889 (however, outcomes were reported for only 8,912 patients), matched on “adju-dication platform,” client type, benefit characteristics including “mail or retail coverage,” and “mean out-of-pocket costs at the group level;” and by demographic characteristics and geographic region at the individual level. Copayments were $10 for tiers 1 and 2 in both periods,b and ranged from $30-$35 for tier 3 in both periods; insulin mean changed from $23 to $28 and OAD mean was $18 in both periods.

Analysis Comparisons of MPR, treatment initiation and DC rates pre-intervention vs. post-intervention for intervention and comparison groups; DC rate calculations excluded patients switching to a different therapeutic antidiabetic subclass. Logistic regression models of discontinuation and linear regression models of MPR adjusted for age, sex, and mail order use.

Results •Treatmentinitiationrateswere2.3%interventionvs.1.6%comparisongroup.•Amonginitiators,DCrateswere16.0%interventionvs.24.3%comparison;MPRwas0.857interventionvs.0.858comparison.•Amongcontinuers,DCrateswere26.0%interventionvs.29.8%comparison;MPRchangewas+0.049interventionvs.–0.023

comparison.Comments •Calculationsdifficulttointerpretbecauseofmissinginformationinstudydatatables,especiallypresentationofdataforonlya

small fraction of study cases.•Pre-interventionvaluesofmostoutcomevariablesnotreported,makingitdifficulttointerpretthepractical/clinicalsignificanceof

study results.•Priortotheintervention,useprevalencerateswereabout1-2percentagepointshigher,andthepercentagewithMPRmorethan

80% was much lower, in the VBID group than the comparison group; MPR change results could represent RTM, but pre-interven-tion MPRs were not documented, making RTM effects difficult to ascertain.

•Theeffectof$4genericprogramsamonglargecommunitypharmaciesisunknown;itispossiblethathighertreatment“initia-tion” rates reflect switch from $4 generics (not in database and paid by patient) to $0 generics (in database and paid by health plan sponsor).

•ThedecisiontoexcludefromtheDCcalculationpatientswhomadeasubclassswitch(a)maskedthepotentiallypositive result of switching to lower cost medication and (b) artificially inflated DC rates in presence of elevated switch rates (Table 1), thereby bias-ing the analysis to find higher DC rates in the comparison group.

•Insulincopaymentsincreasedby22%inthecomparisongroupfrompre-topost-intervention.Choudhry (2010)68 Overview: Study of elimination of all statin copayments for patients with vascular disease or diabetes and reduction of copayments

for clopidogrel, effective January 1, 2007.Sample Patients who from January 2006 through December 2007 either (a) had at least 1 statin claim and proxy measure for diabetes and

vascular disease (at least 1 claim for diabetes medication or supply, beta blocker, or platelet inhibitor) or (b) had at least 1 claim for clopidogrel; there was no continuous enrollment requirement for sample inclusion.

Groups Intervention: Employees and dependents of Pitney Bowes; n = 2,051 statins and 779 clopidogrel; mean copayments changed from $24.18 to $0.60 for statins and from $17.22 to $8.86 for clopidogrel.Comparison: Commercially insured enrollees of BCBS-NJ who used the same PBM; n = 38,174 statins and 11,627 clopidogrel; mean copayments changed from $11.80 to $11.95 for statins and from $10.65 to $14.43 for clopidogrel.

Analysis Interrupted time series of monthly PDC from January 2006 through December 2007; GEE models controlled for demographics, ZIP level data, and comorbidities.

Results •Unadjusted:StatinPDC2.8percentagepointshigherforVBIDthancomparison“immediatelyafter”implementation;clopidogrelPDC 4.0 percentage points higher for VBID than comparison 12 months post-implementation.

•Adjusted:3.1percentagepoint“immediate”increaseinPDClevelforstatins;4.2percentagepoint“immediate”increaseinPDClevel for clopidogrel; no significant changes in slope in either group.

Comments •PitneyBowesimplementedwidelypublicizedcopaymentreductionsandeducationbeginningin2002,5yearspriortothiscopay-ment reduction;c context of present study not clear. Analysis “unable to account” for the extent of participation in DMPs that were available to both study groups.

•Meancopaymentincomparisongroupwasessentiallyunchangedforstatinsinpre-andpost-interventionperiodsbutincreasedby35% (from $10.65 to $14.43) for clopidogrel.

•GenericsimvastatinbecameavailableinJune2006,midwaythroughthepre-interventionperiod;itisnotclearwhetherplansencouraged switches to generic simvastatin during post-implementation period.

•Positivemethodologicaldecisionsincludedthefollowing:(a)daysspentinahospitalornursinghomeweresubtractedfromthedenominator; (b) statin PDC accounted for all drugs (i.e., switching did not count against PDC); and (c) numerous sensitivity anal-yses were performed (producing similar findings).

APPEnDIx

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 171

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

Summary of Recent Observational Studies of Copayment Reductions and Medication Adherencea (continued)

Gibson (2011a)71 Overview: Study of reduction in copayment to 10% for all diabetes medications, including both generic and brand, effective January 1, 2006; a DMP also became available to both intervention and comparison group employees on January 1, 2006.

Sample Enrollees aged 64 years or younger who either participated in DMP or opted out of DMP and were continuously enrolled in at least 4 consecutive quarters from 2005 through 2008; use of antidiabetic medication was not required for sample inclusion, and the authors note that “our enrollee pool may have included patients who were using diet and exercise to manage their condition.”

Groups Intervention: Employees and dependents of 2 units “of a large multi-industry firm” (n = 33,160); n = 1,876 in DMP and n = 328 who opted out of DMP. Copayments for diabetes medications changed from 10% generic, 20% tier 2, and 35% for tier 3 to 10% for all diabetes medications (generic and brand).Comparison: Employees and dependents in the rest of the firm’s units (n = 59,038); enrollees were propensity matched for probabil-ity of being in the intervention group based on demographic characteristics, health status (CCI and psychiatric diagnosis groups), employment characteristics (e.g., salary vs. hourly, active vs. retiree), relationship to employee, and ZIP-level income and education; n = 1,876 in DMP and n = 328 who opted out of DMP. Copayments remained constant at 10% generic, 20% tier 2, and 35% tier 3.

Analysis Time-series panel data analysis of 1 baseline year (2005) and 3 post-implementation years (2006 through 2008), with calendar quar-ter as unit of time; analyses were stratified by participation in DMP. Costs and utilization were analyzed using GEE, and costs were measured as total payments (not payer costs).

Results •AmongDMPparticipants,MPRforVBID(copaymentreduction)groupcomparedwithnon-VBIDgroupwas3.7percentagepointshigher in 2006, 5.1 percentage points higher in 2007, 6.5 percentage points higher in 2008; similar trends were observed for per-centage of patients with MPR at least 80%.

•AmongDMPparticipants,percentagescomplyingwithrecommendedtestsandserviceswerehigherinVBIDgroupthaninnon-VBID group (e.g., percentage point differences in 2008: 4.5 for HbA1c tests, 5.0 for lipid tests, 7.7 for PCP visits, 4.0 for urinalysis).

•Innon-DMPgroup,fewsignificantdifferencesbetweenVBIDandnon-VBIDgroups,but(a)OADMPRwas3.8-4.7percentagepoints higher with VBID in 2006 and 2007 (but not 2008) and (b) compliance in the first year was significantly lower with VBID for HbA1c (5.2 percentage points) and urinalysis (3.1 percentage points).

•InDMPgroup,diabetes-relatedmedicalcostswerelowerandmedicationcostshigherinall3yearsforVBIDthannon-VBIDgroup; authors estimated an ROI of 1.33.

•NoconsistentcostpatternforVBIDvs.noVBIDinnon-DMPgroup.Comments •Enhancedcompliancewithscreening/monitoringtestsandmedicalexaminationswasdescribedasan“effect”ofVBIDalthough

the VBID intervention changed drug copayments only.•Reportmeasured“ROI”ontotal costs paid by all sources (including the patient cost share) instead of payer costs. Thus, cost to

payer was not reported; results do not represent the ROI that should be expected by the payer; and description of VBID program as “cost-neutral” does not accurately describe the payers’ cost outlays.

•CostofDMPwasnotreportedalthoughtheDMPwasextensiveandthepositiveresultsforVBIDwerelimitedtoDMPparticipants.Study report acknowledged that “to obtain a full cost estimate, some form of assignment into disease management, to measure [DMP] effects and costs, would be necessary.”

•CriteriaforinvitationintotheDMPwerenotspecified,makingtheclinicalcharacteristicsofthesampleunclear(e.g.,anydiabe-tes diagnosis, use of particular services, such as hospitalizations, for diabetes, or what, specifically?); proportion of patients who opted-out not reported for the comparison group.

•TheauthorsapparentlymeasuredMPRaszero(0)forpatientsusingdietandexercisealone,regardlessofwhetheranymedicationwas prescribed—for example, pre-intervention insulin MPR was reported as 10% or less for all 4 groups.

Gibson (2011b)70 Overview: Study of a “large, global pharmaceutical firm” that implemented a VBID (copayment reduction) on January 1, 2005, for drugs to treat asthma (SABAs, LABAs, leukotriene modifiers, inhaled corticosteroids, methylxanthines, mast cell stabilizers, and combinations); HTN (ACE inhibitors, ARBs, diuretics, alpha-2 agonists, aldosterone receptor blockers); and diabetes (insulin and OADs). In addition to the VBID, DMPs for “asthma, cardiac conditions, and diabetes were also implemented for enrollees in the com-pany’s indemnity and point-of-service plans in 2005 and … across all of the company’s self-insured plans in 2007,” with educational materials about the programs “communicated to all employees … starting in the fourth quarter of 2004.”

Sample Enrollees aged 18-64 years who were enrolled for a minimum of 1 year prior to the first quarter of enrollment post-implementation and for at least 2 quarters after implementation; no other sampling requirements were imposed, and the sample included “enrollees who did not use any medical services.”

Groups Intervention: n = 25,784 before matching, 25,065 after matching; pre-intervention copayments not clear but appear to be 20% coinsurance in community pharmacies and 10% in mail order pharmacy with minimum $10 and maximum $40; post-intervention copayments were 10% coinsurance in community pharmacies and 7.5% in mail order pharmacies with the same minimum and maximum. No quantitative information about cost-sharing was provided except that multivariate models estimated that in the first post-implementation year, “enrollees paid $4 less on average for VBID medications with the VBID ($11.16) than without the VBID ($15.37);” the difference increased to about $5 per enrollee in the third post-implementation year; and, “overall cost-sharing amounts declined 7.2% for VBID enrollees” from Q1 2004 to Q4 2007. Brand drugs manufactured by the employer were provided free of charge to all enrollees in both periods.Comparison: n = 154,444 before matching, 25,065 after matching; enrollees insured in 4 “empirically similar” firms with data in the Thomson Reuters MarketScan database were propensity matched for probability of being in the intervention group based on demographic characteristics, health status (CCI and psychiatric diagnosis groups), employment characteristics (e.g., salary vs. hourly, active vs. retiree), relationship to employee, ZIP-level income, and number of quarters of enrollment. No quantitative information about copayments paid in either period was provided except that: (a) “patient cost-sharing did not change in the pre-period” and (b) “cost-sharing increased 12%” from Q1 2004 to Q4 2007.

APPEnDIx

172 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

Summary of Recent Observational Studies of Copayment Reductions and Medication Adherencea (continued)

Analysis Pre-post design with matched comparison group for 2004 (year prior to implementation) and each of 3 subsequent years—2005 through 2007; data file contained 1 observation per enrollee per calendar quarter, analyzed using GEEs with VBID treated as a time-varying effect (trend) set to zero in Q1 2005 and 11 in Q4 2007. Outcomes were (a) use rate (percentage with at least 1 claim in therapeutic class); (b) adherence based on PDC (at least 50% for asthma, at least 80% for HTN and diabetes); (c) number of 30-day equivalent fills both overall and for VBID classes; and (d) total “eligible charges” for drugs and medical services.

Results •AveragedrugspendingonVBIDmedicationsinQ42007was$68forVBIDgroupand$51forcomparisongroup.•Fordiabetesdrugs,useratesandnumberoffillsdidnotsignificantlydifferinanyyear;adherencewaslower for the VBID group in

the first 2 post-implementation years and not significantly different in the third year.•Forasthmadrugs,nosignificantdifferencesonmostmeasuresinmostyears;onlyexceptionwasthatadherencetoasthmamedi-

cations (but not use rates or number of fills) was significantly higher in the third year only.•ForHTNdrugs,adherence,userpercentage,andnumberoffillsweresignificantlyhigherintheVBIDgroupinallyears.•GEEsshowednosignificantdifferencesonanymeasureoftotalcharges(prescriptiondrug,medical,ortotal)inanyofthe3years.

Comments •Resultsareimpossibletointerpretbecauseoftheabsenceofspecificcost-sharinginformationandbecauseof2unusualconfound-ing factors—provision of coincident DMP and of unnamed brand drugs free-of-charge in the intervention group.

•AbstractimpliesthatadherencewashigherwithVBIDinallclasses,butdatatablesindicatethatadherenceandmedicationuseincreased only for HTN medications.

•Authors“raisetheprospectthatthisprogrammayhavesavedthecompanymoneybyreducingothermedicalcosts”because“clini-cal effects such as changes in glucose levels, blood pressure, and lung functioning might have occurred,” but these effects were not measured.

•Cost-sharingchangefortheVBIDprogramappearstobesmall(onlya7.2%decrease),makingitlikelythatresultsfortheinter-vention group are at least partly attributable to the DMP or to another confounding factor.

•“Eligiblecharges”werenotspecificallydefined(perhapstotalprovider-billedcharges)butdonotappeartorepresentthepayer’soutlays; thus, the authors’ description of the intervention as “mostly cost-neutral to the company” appears to be incorrect.

•Discussiondescribesnonsignificantcostdifferencesasiftheyweresignificant.•Authorsindicatethatresultsweresimilarinasubsetofenrolleeswithpre-interventiondruguse;however,descriptionofresultsin

text does not match exhibit table, and the table names what appears to be the wrong pharmaceutical company (employer). Thus, these results cannot be interpreted.

Maciejewski (2010)63 Overview: Study of a VBID (copayment reduction) in which (a) for both the intervention and comparison groups, all tier 3 brands were made tier 2 in 8 classes, including 2 without generic drugs (ARBs and cholesterol absorption inhibitors) and (b) for the inter-vention group, generic copayments were waived in 6 classes, effective January 1, 2008.

Sample Enrollees of BCBS-NC who were continuously enrolled from January 2007 through December 2008, and who “were taking a medica-tion” from at least 1 of the study classes (time period for criterion not clear).

Groups Intervention: Enrollees whose employers opted into the VBID (n = 638,796); n = 5,077 metformin, 15,605 diuretics, 14,250 ACE inhibitors, 11,137 beta-blockers, 18,346 statins, 7,191 CCBs, 7,445 ARBs, 4,019 cholesterol absorption inhibitors. Pre-intervention copayments (reported as means per claim) were $10.74-$11.38 generic, $33.79-$34.39 brand, and overall: $13-$17 for ACE inhibitors, beta-blockers, diuretics, and metformin; $22-$25 for statins and CCBs; $36-$37 for ARBs and cholesterol absorption inhibitors.d Post-intervention copayments were $0 generic, $30.50-$30.75 brand, and overall not reported.Comparison: Enrollees of BCBS-NC whose employers did not opt in (n = 638,091); n = 2,826 metformin, 9,137 diuretics, 7,668 ACE inhibitors, 6,343 beta-blockers, 10,162 statins, 4,099 CCBs, 4,514 ARBs, 2,291 cholesterol absorption inhibitors. Pre-intervention copayments were not reported for generics, were same as intervention group for brands, and overall: $14-$18 for ACE inhibitors, beta-blockers, diuretics, and metformin; $24-$27 for statins and CCBs; $39-$40 for ARBs and cholesterol absorption inhibitors.d Post-intervention copayments were not reported except that brand copayments were the same as in the intervention group.

Analysis DID (GLM) model of MPR with person-year as unit of analysis, controlling for age, sex, ERG comorbidity burden measure, and pre-intervention medication burden (count of medications, mean copayments, at least 1 90-day supply, GDR).e

Results •Pre-to-postchangesinMPR(percentagepoints)wereslightlyhigherforinterventionthancomparisongroupinclasseswithgener-ic copayment waiver: metformin 3.8, diuretics 3.3, ACE inhibitors 2.9, beta-blockers 2.5, statins 1.8, CCBs 1.5.

•In2classesconsistingentirelyofbranddrugs(i.e.,copaymentreductionsinbothgroups),groupsdidnotsignificantlydiffer.•Fillingatleast1prescriptionwitha90-daysupplywasassociatedwithMPRincreasesof17to22percentagepoints.•Authorsstatedthat“significantlyhighermedicationcostsgiventhenumberofpeopleparticipatingintheprogram”posedapoten-

tial threat to “the long-term viability of this innovative policy change.”Comments •Methodologicallystrengthenedbycomparisonof6VBID(genericdrugcopaymentwaiver)classeswith2drugclasses(ARBsand

cholesterol absorption inhibitors) consisting entirely of brand drugs (copayments reduced in both groups).•Titleincludestheterm“targetedpatients,”butnopatientswereexcludedfromcopaymentreductioninthe6drugclassesstudied

(i.e., there was no targeting for secondary prevention or specific risk factors). •AbstractanddiscussionerroneouslyindicatethatthestudyVBIDincludedbothgenericandbrandcopaymentreductions;actually,

brand copayments were reduced in both the intervention and comparison groups.•GDRwasreportedpre-interventionbutnotpost-intervention.

APPEnDIx

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 173

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

Summary of Recent Observational Studies of Copayment Reductions and Medication Adherencea (continued)

Sedjo (2008)62 Naturalistic study of patent expiration for simvastatin effective June 23, 2006; there was no intervention—enrollees who used brand simvastatin prior to expiration were compared with those who used other brand statin drugs (no copayment decrease). Study patients were aged 18 years or older and continuously enrolled from June 2005 through May 2007.

Groups Copayment reduction group: Commercially insured health plan enrollees with at least 1 claim for brand simvastatin from June through August 2005 and at least 1 generic simvastatin claim from June through August 2006 (n = 13,319). Pre-expiration copay-ment was a mean $14.60; post-expiration copayments not reported but patients were subgrouped by amount of reduction (subgroup counts not specified).Comparison group: Commercially insured health plan enrollees with at least 1 claim for a brand statin (not simvastatin) from June through August 2005 and at least 1 claim for any statin (not simvastatin) from June through August 2006 (n = 26,569); matched to intervention group patients on incident use (binary indicator) and pre-expiration copayment ± $2. Pre-expiration copayment was a mean $14.57; post-expiration copayments not reported.

Analysis Bivariate and multivariate (linear regression modeling) by-group comparisons of change in MPR from pre-expiration to post-expira-tion periods; MPR represented any statin drug “to allow for switching within the statin class” and was measured in each period from the first fill date through the subsequent 270 days. Secondary outcomes included percentage adherent (MPR at least 80%) and elas-ticity (percentage change in MPR divided by percentage change in copayment). Multivariate analyses controlled for age, sex, incident statin use, chronic disease score, and pre-expiration MPR and copayments.

Results •MPRdeclinedinbothgroups:–0.17%incopaymentreductiongroupand–1.67%incomparisongroup;multivariateanalysessug-gested 0.52% adjusted mean MPR increase in copayment reduction group and 2.02% adjusted mean MPR decrease in comparison group.

•IncreaseinMPRtoatleast80%occurredin10.5%ofreductionand10.0%ofcomparisonpatients.•DecreaseinMPRtobelow80%occurredin12.1%ofreductionand11.3%ofcomparisonpatients.•Elasticity0.02forreductionsof$0to$5and–0.02forreductionsofmorethan$15(i.e.,a10%decreaseincopaymentwasassoci-

ated with a 0.2 percentage point increase in MPR).Comments •Discontinuationratesnotreported;resultsreflectonlypatientswithstatinuseinbothperiods.

•Neitherpost-changecopaymentsnorcountsforcopaymentchangegroupswerereported,makingitdifficulttoassesstheoverallmagnitude of the cost-sharing change.

•Meanpre-expirationMPRswerehigh(85%-89%)inbothgroups;resultsmaynotgeneralizetolessadherentpatientpopulations.Zeng (2010)69 Study of the movement of “a comprehensive list of diabetes medications” into tier 1 effective January 1, 2007; tier 1 drugs included

SSB (e.g., pioglitazone, rosiglitazone/metformin, exenatide) and MSB (e.g., Amaryl as well as glimepiride).f “Most diabetes medica-tions and supplies in tier 2” and “a few in tier 3” were “moved … into [VBID].”

Sample Commercially insured HMO enrollees aged 18 years or older on January 1, 2006, with at least 1 diabetes medication claim in 2005, 2006, and 2007; continuously enrolled from January 1, 2005, through December 31, 2007.

Groups Intervention: Employees and dependents of the clinic that owned the HMO (n = 71); pre-intervention copayments were $10 for generic, 30% for tier 2, and 50% for tier 3, for an overall mean of $18.80 per 30-day claim. Post-intervention copayments were $10 for tier 1 and not reported for tier 2 and tier 3, for a mean of $10.40 per 30-day claim.Comparison: Enrollees of the same HMO who were not VBID enrolled (n = 5,037, of whom 639 were selected at random for analy-sis); pre-intervention copayments are not entirely clear but appear to be the same structure as the intervention group’s, with a mean of $14.30 per 30-day claim. Post-intervention copayments were “unchanged” and a mean of $18.80 per 30-day claim.

Analysis Propensity-score weighted DID analysis in which each enrollee had 2 observations, one for 2006 and the other for 2007; outcomes were PDC and percentage adherent, defined as PDC of at least 80%. Propensity score was based on age, sex, mean diabetes medica-tion copayment, insulin use, and comorbidities in 2005 measured using RxRisk categories.

Results •PDC(afterpropensity-scoreweighting)0.88pre-interventioninbothgroups;changedto0.90ininterventiongroupandunchanged in comparison group; nonsignificant in multivariate analysis.

•Percentadherent(afterpropensity-scoreweighting)changedfrom75.3%to82.5%ininterventionandfrom79.1%to78.5%incomparison group; OR of adherence in logistic regression controlling for demographics, insulin use, and RxRisk score = 1.56, 95% confidence interval = 1.04-2.34.

•PercentagepointchangeinuseofTZDs+1.3inintervention,–0.6incomparisongroup;authorsnoted“moderate”shifttowardincreased use of brand drugs in VBID group.

Comments •Difficulttointerpretwithoutmorespecificinformationaboutthecost-sharingamountsandtierassignments.•RxRiskcategoryassignmentusedincomorbiditymeasurementandlogisticregressionmaycontainerrors;studyreportindicates

that 10% of study sample (mean age approximately 58 years) had cystic fibrosis.•Studyabstractdoesnotmentionthefindingthattheby-groupdifferencesinPDCchangewerenotsignificant.•Althoughcomparisongroupcopaymentsweredescribedas“unchanged,”theyincreasedfrompre-topost-interventionby31%;

however, the increase was less in the propensity-weighted analysis.•Nofcasesininterventiongroupwassmall,andtheexternalvalidityisunclear.•Meanpre-interventionPDCswerehigh(86%-88%)inbothgroups;resultsmaynotgeneralizetolessadherentpatientpopulations.•Transparentreportingofpre-andpost-interventionvaluesforoutcomemeasures.

APPEnDIx

174 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

What Do We Really Know About VBID? Quality of the Evidence and Ethical Considerations for Health Plan Sponsors

Summary of Recent Observational Studies of Copayment Reductions and Medication Adherencea (continued)

Roebuck (2011)31 Study measured association between MPR and all-cause health care expense; not a study of copayment reduction, and there was no intervention.

Sample Commercially insured members of 9 health plans with primary, secondary, or tertiary diagnoses on at least 2 outpatient visit claims or 1 inpatient stay or ER visit claim for any of 4 chronic conditions: CHF (n = 16,353), HTN (n = 112,757), diabetes (type 1 or type 2, n = 42,080), and dyslipidemia (n = 53,041); all were enrolled continuously from January 1, 2005, through June 30, 2008.

Analysis Linear fixed-effects modeling of panel data in which each patient contributed 3 (yearly, July through June) observations—2005/2006 through 2007/2008; results represent marginal effects. Models controlled for demographics, CCI, and time trends. MPR for each condition represented “the average of the [MPRs] for all therapeutic classes for each chronic disease, weighted by the days’ supply in each therapeutic class,” with zero (0) adherence assumed for patients with a diagnosis but no drug treatment. Adherence was defined as MPR of at least 80%, and there was no assessment of continuous MPR. Outcomes were 3 measures of annual all-cause service use: inpatient days, ER visits, outpatient physician visits; and 3 measures of all-cause cost: pharmacy, medical, and total.

Results •Adherencewasassociatedwithfewerall-causeinpatientdaysannually(1.18forpatientswithdyslipidemia,approximately2forpatients with HTN or diabetes, 5.72 for patients with CHF).

•Marginaleffectforinpatientdayswasgreaterforthoseaged65yearsorolder(1.88dyslipidemia,3.41diabetes,3.14HTN,5.87CHF).

•Associationsofadherencewithall-causeERvisituse(reductionof0.01-0.04visitsannually)andoutpatientphysicianvisits(increase of about 1 visit annually for patients with CHF and less than 0.5 visits for the remaining patients) were small.

•All-causemedicalservicescostswerelowerandpharmacycostshigherforadherentpatients;netannualsavingsestimatedat$1,258 dyslipidemia, $3,576 diabetes, $3,908 HTN, $7,823 CHF.

Comments •Transparentandclearreportofstudymethodsandlimitations.•DecisiontoassumezeroadherenceforpatientsnottreatedwithmedicationisinconsistentwithtreatmentguidelinesforHTNand

dyslipidemia and partially inconsistent with guidelines for type 2 diabetes.•Resultsrepresent all-cause (not disease-specific) utilization, and there was no assessment of whether health care utilization could

reasonably be clinically attributable to medication adherence or nonadherence; sole reliance on fixed-effects modeling to establish causality.

•Astheauthorsacknowledge,fixed-effectsmodelingdoesnotadjustforconfoundersthatchangeduringthestudyperiod,suchasa patient’s decision to improve both medication adherence and other health-related behaviors simultaneously (i.e., new “healthy adherence” behaviors).

•MPRdenominatorwasapparentlynotadjustedforinpatientdayswhenpatientswouldbereceivingdrugsfromthehospitalornursing home (not measured with pharmacy claims).

•Analysesandinterpretationdidnotaccountforpossibilitythatmedicationscouldhavebeenobtainedthroughcommunityphar-macy generic drug discount programs during the study period (July 2005 through June 2008).

•AlthoughauthorssuggestthatfindingshavepotentialimplicationsforVBID,associationofcopaymentwithadherenceorutiliza-tion was not assessed.

aAll information reported in this table reflects both the study reports and any online appendices, which can be accessed using hyperlinks in the study reports. No study of VBID assessed changes in GDR, generic drug utilization PMPM, or payers’ costs.bThis assessment may be an error because it sounds implausible; however, text indicates: “Users of tier 1 and tier 2 medications in the control group … had no changes in their copayment of $10 during the study period.”29

cSee Mahoney (2005).77

dFor ARBs and cholesterol absorption inhibitors (i.e., containing ezetimibe), which were 100% brand, the average pre-intervention copayment reported in the methodologi-cal appendix exceeds the top of the pre-intervention brand copayment range reported in the text for reasons that are not clear from the study report.eThe time period for measurement of the GDR covariate was not clear in the study report. The description appears in the methodological appendix in a section describ-ing “pre-period medication burden;” however, the description of the variable states that the authors “controlled for the [GDR] … for the immediate financial benefit of the [VBID] program by switching from brand to generic medications.”63

fList of drugs for Zeng et al. VBID program is in a study appendix at http://www.ispor.org/Publications/value/ViHsupplementary/ViH13i6_Zeng.asp. Our assessment that all drugs in the study appendix were tier 1 may be inaccurate, but the text refers to them as tier 1 medications and the study appendix does not specify tier status. Mean copayments shown in this Appendix are per 30-day supply before propensity score weighting. Propensity-score weighted copayments for the intervention and comparison groups were, respectively, $15.30 and $14.60 pre-intervention and $10.10 and $15.10 post-intervention.69

ACE = angiotensin-converting enzyme inhibitor; ARB = angiotenson II receptor blocker; BCBS-NC = Blue Cross Blue Shield of North Carolina; CCB = calcium channel blocker; CCI = Charlson Comorbidity Index; CHF = congestive heart failure; DC = discontinuation; DID = difference-in-difference; DMP = disease management program; ER = emergency room; ERG = Episode Risk Groups classification model, GDR = generic dispensing ratio; GEE = generalized estimating equation; GLM = generalized linear model; Hb = hemoglobin; HMO = health maintenance organization; HTN = hypertension; LABA = long-acting inhaled beta agonist; MPR = medication possession ratio; NA = not applicable; OAD = oral antidiabetic drug; OR = odds ratio; PDC = proportion (or percentage) of days covered; PBM = pharmacy benefits management company; PMPM = per member per month; Q = quarter; ROI = return-on-investment; RTM = regression to the mean; SABA = short-acting beta agonist; SSB = single-source brand; TZD = thiazolidinedione; VBID = value-based insurance design (implemented as drug copayment reduction in these studies); vs = versus.

APPEnDIx

www.amcp.org Vol. 17, No. 2 March 2011 JMCP Journal of Managed Care Pharmacy 175

■■  Case Report of Specialty Pharmacy Management of HemophiliaToday it is generally understood that a disproportionally high share of medical claims are attributable to a small minority of patients with rare chronic disease states. Hemophilia is consid-ered to be one of the most expensive chronic conditions cur-rently being treated by any health care system. Antihemophilic medications have been estimated to account for greater than 90% of the total costs of hemophilia care.1 The annual cost of medications to treat a patient with severe hemophilia may exceed $300,000 per year.2

Adherence to antihemophilic medications has been defined as infusing 75%-80% of prescribed doses.3 Noncompliance with these medications can contribute not only to poor out-comes, but also to significant costs associated with excess inventory. An average adult with severe hemophilia that is infusing 3,000 units 3 times per week as prophylaxis can accumulate 72,000 units of unused factor over the course of 1 year by missing 2 doses per month, or possible waste valued at $96,480 in antihemophilia drug cost (priced at 80% of average wholesale price [AWP] in January 2011).

Our specialty pharmacy developed a disease management program for hemophilia modeled after the results from the Haemophilia Utilization Group Study (HUGS).2 The program is designed to provide intensive patient management to improve clinical outcomes and to lower hemophilia-related health care costs. Our program was adopted by a 1.4-million member West Coast health plan in October of 2009 to serve its hemophilia population. At the initiation of this outcomes program, 54 unique patients were identified as potential candidates for the program. Each patient case was carefully evaluated to ensure appropriate utilization of health care resources and to achieve maximum benefits for the health plan. All patients with severe disease were selected as candidates. Other variables included: number of bleeds per year, age, comorbid conditions, annual factor consumption, and number of target joints. At the end of this screening period, 37 of the 54 patients were identified as candidates based on disease severity. Patient participation was optional, and 24 of the 37 patients (65%) elected to participate in our disease management program.

One of the 24 enrolled patients was a 21-year-old male with severe hemophilia A. His disease was complicated, with mul-tiple target joints, obesity, noncompliance, and a general lack of knowledge regarding his disease and treatment. He had a treatment regimen consisting of clotting factor Kogenate FS as primary prophylaxis and as needed for bleeding episodes. His social situation was problematic including a dual residence, unemployment, and a sedentary lifestyle. He did not use alco-hol or illicit drugs. On admission to the program, the patient completed a detailed telephonic assessment by a hemophilia specialist and hemophilia self management questionnaire. We then began the disease benchmarking period where it became apparent that his disease severity, bleeding history and utili-

zation were not congruent. This patient averaged 1.75 bleeds per month into his target joints which included both knees and elbows. A home visit by a hemophilia nurse specialist was scheduled to assess the patient.

On the day of the home visit, the patient presented with a notably swollen and painful right knee, bruised and swollen knuckles on both hands, and a recovering bleed in his left forearm. He also presented with limited range of motion in his right knee and left elbow. The patient admitted to discontinu-ing his primary prophylaxis and treating only the most severe bleeds; he believed that he could over time desensitize his body to the exogenously administered clotting factor, reducing his requirements and possibly the number of bleeds. Additionally, the patient treated his severe bleeds with a single dose of dou-ble the prescribed units. He assumed this would stop the bleed quickly and prevent its reoccurrence. This patient interview solidified our initial assessment of a lack of disease knowledge and noncompliance. The patient had an on-hand inventory of over 45,000 units and another 30,000 units at his alternate residence; this was nearly a 2-month supply of clotting factor at an estimated drug cost of $100,500 (priced at 80% of AWP).

The patient was receptive to the hemophilia nurse specialist and was eager to ask questions about his condition. He received education focusing on his disease severity, the mechanisms of joint destruction from bleeds, and the consequences of treat-ment delays and noncompliance to his prophylaxis regimen. Visual aids were used throughout this process. A treatment plan was developed and discussed with the patient, and he agreed to resume his prophylaxis treatment. He verbalized satisfaction with the therapy management program and agreed to continue enrollment. As part of the disease management program, the patient receives monthly telephonic assessment calls to capture self-reported bleeding episodes and other lifestyle indicators. Three months after the nurse visit, the patient demonstrated compliance with his treatment plan. The patient’s telephonic self-assessments indicated an increased level of compliance with his prophylaxis that is also confirmed by his factor utili-zation history. Bleeding into his target joints was reduced to 1 episode in the 3-month post intervention period. The patient also reported wearing his medical alert identifier. In addition to improved care management, the intervention with the nurse visit reduced the average monthly hemophilia drug cost by $6,490, from $225,832 in the 5 months before the intervention to $193,382 in the 5 months after the intervention.

Aggie Gilbert, RNHemophilia Clinical Nurse Specialist

Brian Tonkovic, PharmDClinical Manager, EyeOn Therapy Management Coram Specialty Infusion Services Therapy Managementtonkovicb@coramhc.com

LET TERS

176 Journal of Managed Care Pharmacy JMCP March 2011 Vol. 17, No. 2 www.amcp.org

DISCLOSURES

Gilbert and Tonkovic are employees of Coram Specialty Infusion Services Therapy Management.

REFERENCES

1. Globe DR, Curtis RG, Koerper MA; HUGS Steering Committee. Utilization of care in haemophilia: a resource-based method for cost analy-sis from the haemophilia utilization group study (HUGS). Haemophilia. 2004;10(Suppl. 1):63-70.

2. Manco-Johson MJ, Abshire TC, Shaprio AD, et al. Prophylaxis versus episodic treatment to prevent joint disease in boys with severe hemophilia. N Engl J Med. 2007;357(6):535-44. Available at: http://www.nejm.org/doi/pdf/10.1056/NEJMoa067659. Accessed January 14, 2011.

3. Thornburg CD. Prophylactic factor infusions for patients with hemophilia: challenges with treatment adherence. J Coagulation Disorders. March 2009.

■■  Online Drug Pricing Tools Deserve More AttentionWe read with great interest the article “Evaluation of health plan member use of an online prescription drug price com-parison tool” by Carroll and colleagues.1 The research shows how drug information tools are being used by consumers. We (DestinationRx) have been creating innovative drug and plan comparison tools for more than a decade. In the spirit of disclosure, DestinationRx’s Drug Compare and its predecessor Rxaminer are tools that compete with RxEOB, the online tool that was the subject of the study by Carroll and colleagues. We make available drug pricing tools to more than 100 million Americans, and as such we have comments on 3 points that your readers may find helpful.

First, while we do have a consumer version of Drug Compare that shows drug costs based on estimated community pharmacy prices, the vast majority of our users are accessing plan- and/or PBM-sponsored sites. These sites come loaded with member-specific benefit information including formulary, utilization management restrictions, drug tiers, member copay and pharmacy-specific price comparisons. Second, Carroll

and colleagues found that approximately 5.2% of families in a large integrated health plan who used the pharmacy ben-efit throughout the year used the MyPharmacyTools (RxEOB) online pricing tool in a 12-month period from July 2007 through June 2008. We found that 5.8% (n = 4,087) of approxi-mately 71,000 employees enrolled in health savings accounts used our Drug Compare online tool in a 9-month study period through October 2010.2

Third, we agree that drug pricing is of paramount con-cern to end-users. We believe that filtering the alternatives displayed based on comparative efficacy leads to even greater adoption and behavior change. It is for this reason that we always emphasize that beyond pricing it is imperative to dis-play therapeutic alternatives of similar efficacy.

Our experience shows that drug costs are still one of the most controllable of all health care expenditures, and we are glad to see drug information tools getting the visibility they deserve for allowing members to better manage their health care and drug spend.

Alexander Grunewald, PhDDestinationRxLos Angeles, Californiaalexander.grunewald@drx.com

DISCLOSURES

Grunewald is employed by DestinationRx.

REFERENCES

1. Carroll NV, Mitchell MP, Cannon HE, York BW, Oscar RS. Evaluation of health plan member use of an online prescription drug price comparison tool. J Manag Care Pharm. 2010;16(9):680-92. Available at: http://www.amcp.org/data/jmcp/680-692.pdf.

2. Grunewald A, Parker ME, Yaniga Y, West W, Nelleman S. Potential impact of a therapeutic alternative tool on drug spending. International J Medical Banking. 2011 (manuscript in review).

Letters