Rationale and Background

CCB and cancer risk is one study among a larger studyto, develop, test and disseminate methodologicalstandards for the design, conduct and analysis of,Phamacoepidemiological (PE) studies applicable todifferent safety issues and using different data sourcesby (Pharmacoepidemiological Research on Outcomes ofTherapeutics by a European ConsorTium- PROTECT).

Selection criteria Adverse events that caused regulatory

decisionsPublic health impact (seriousness of the

event, prevalence of drug exposure,etiologic fraction)Feasibility Range of relevant methodological issues

The five pairs

Antidepressants (incl.Benzodiazepines) ‐ Hip FractureAntibiotics ‐ Acute liver injury Beta2 Agonists ‐ Myocardial infarction Antiepileptics ‐ Suicide Calcium Channel Blockers ‐ Cancer

Rationale and background….

Emphasis will be on themethodological aspects of thestudies in this protocol and noton the clinical consequences ofthe association underinvestigation.

Primary Objective

Investigating the possibleassociation between uses of CCBsand risk of all forms of cancercombined, among adult patients (18to 79 years of age during the studyperiod, 1 January 1996 and 31December 2009)

Specific Objectives of PROTECT Assess the feasibility to ascertain key adverse events

(AEs) with sufficient validity using standard definitionalalgorithms in data sources used for PE studies;

Develop study protocols for pilot PE studies ofassociations between selected drugs and key adverseevents (AEs);

Provide guidelines on how to identify and assess thevalidity and use of national drug utilisation data toestimate public health impact of adverse events (AEs);

Identify and optimise methods to control for confoundingin simulation studies;

Specific objectives… Investigate discrepancies in results from pilot PE studies

between databases, and explore differences with otherdata sources and paradigms as appropriate;

Identify and evaluate statistical methods for conductingmulti-database studies;

Evaluate identified signal from signal detection detectionstrategies applied by WP3 in electronic healthcaredatabases;

Test guidelines and standards in validation studiesperformed with WP6, regarding different safety endpointsin the same databases, the same safety endpoints indifferent databases, and same safety endpoints in samedatabases among patients with different indications.

Hypothesis

CCBs modify the risk of cancer (all forms of cancercombined and various groups of cancers).

It has been proposed that CCBs may interfere withapoptosis, leading to an increased potential for abnormalcell proliferation and tumor growth (Daling JR. Calcium channelblockers and cancer: is an association biologically plausible? Am J Hypertens1996;9:713‐714.)

CCBs have also been shown to inhibit apoptosis in certainnon‐transformed cell lines but promote apoptosis in othernon‐transformed and transformed cell lines. The results fromnon‐human genotoxicity studies have shown no linkbetween CCB use and tumor development (Mason RP. Effects ofcalcium channel blockers on cellular apoptosis: implications for carcinogenicpotential. Cancer 1999;85:2093‐2102).

Data sources for investigating hypothesis

Cohort and Case control studies will beconducted from data from the following datasources

UK• General Practice Research Database (GPRD)• The Health Improvement Network (THIN) Danish national database

Descriptive studies will be conducted in thefollowing

Mondriaan (The Netherlands) BIFAP (Spain) Bavarian Claims Database (Germany)

Study design

Study designs• Descriptive• Cohort• Population based cohort• Nested case control

General Practice Research Database (UnitedKingdom) largest ongoing health care database available in

the UK since 1987 more than 4.8 million active research quality patient

data from more than 590 general practices in the UK The validity of a wide range of drug exposure data is

routinely tested. Cancer registration in England is conducted by eight

regional registries, which also submit a standarddataset of information to the Office for NationalStatistics (ONS), for collation of national cancerincidence data.

The Health Improvement Network(United Kingdom)

collaboration between two companies; In Practice SystemsLtd. (INPS), developer of Vision software used by generalpractices in the UK, and EPIC, provider of access to data foruse in medical research.

collected during routine medical practice and regularlydelivered to a central database.

database currently contains the electronic medical records ofalmost 8 million patients (more than 3 million active patients)collected from over 386 general practices in the UK.

Patient data are arranged in five standardized files perpractice: patient, medical, therapy, additional health data anda file to enable data linkage containing postcodes

Danish National Database (Denmark) The Danish national databases can provide

computerized records of: sociodemographic factors,visits to general practitioners and to hospital, thedispensing of medication by pharmacies, andmortality for the Danish population (5.5 millionpeople).

Mondriaan (The Netherlands)

The Dutch Mondriaan project is a private‐publiccollaboration

Under the umbrella of Mondriaan, theparticipating databases currently include: theDutch General Practitioner (LINH) database, TheAlmere Health Care (ZGA) database, TheGeneral Practitioners of Utrecht (HNU) databaseand The Leidsche Rijn Julius Health Centre(LRJG) database.

The four databases within Mondriaan havedifferent starting dates and scope of data.

BIFAP (Spain) A computerized database of medical records of

Primary Care) The project has started in 2003 having the goal to

achieve a pool of collaborators in the range of 1000general practitioners and pediatricians. Currently,1190 physicians (995 GPs and 195 pediatricians)from 9 different autonomous communities in Spaincollaborate with BIFAP and send their data to BIFAPevery 6 months.

Bavarian Claims Database (Germany) This German database includes a

population‐based data on diagnosisand medical services, covering 10.5million people. It is a pharmacy (claims) database

linked to outpatient treatment datathrough general practitioners andspecialists.

Data collectionInclusion criteria CCB first‐time users and non‐users from 1 January

1996 to 31 December 2009, aged 18 to 79 years. Concerning GPRD and THIN/BIFAP patients are

included if they have Primary Care Practice historyof at least two years and at least one year’scomputerized GPRD/GP prescription history.

Exclusion criteria Patients with any cancer recorded in the GPRD or

The Danish National registration of patients prior tocohort entry will also be excluded from the analysis

Patients less than 18 years and over 70 years

Definition of exposureExposure in the cohort study and the nested

case‐control study will be defined as cumulative useof CCB. The cumulated exposure will be derivedfrom the total amount of DDD, the number ofprescriptions and the period the prescriptions covers(package size).

The descriptive study will require one prescription todefine exposure.

Definition of outcomes

Outcomes will include all forms ofcancer combined, as well asbreast cancer in women, prostatecancer and colon cancer analyzedseparately.

Descriptive study The period prevalence and point prevalence of CCB use

(defined as ≥ 1 prescription for a CCB) will be describedby age group (0‐9,10‐19,20‐29, 30‐39, 40‐49, 50‐59,60‐79) and sex or by indication for the period 2000 to2009 (ten years)

Period prevalence and point prevalence will also bepresented for the years 2000, 2004 and 2009.

Period prevalence of patients having ever used a CCBwill be presented by number of prescriptions (1, 2‐4,5‐11, 12‐23, >23) for 2000 to 2009.

Descriptive study… The period prevalence of cancer will be presented

by age group (10 year categories) and sex for theperiod from Jan 1999 to Dec 2001 inclusive.

The yearly incidence of cancer will be presented byage group (10 year categories) and sex from 2000 to2009.

Incidence and prevalence will be presented for allforms of cancer combined, as well as for breastcancer.

Prostate cancer and colon cancer will be describedseparately.

Databases where descriptive studies will beheld

Cohort study

CCB first‐time users and CCBs non‐users between 1January 1996 and 31 December 2009.

Index date will be the date of the first registration ofcancer

CCB non‐users will be drawn as a random samplematched on age and gender.

Members of each cohort will be followed from entry tothe study until the earliest of: diagnosis date of studyoutcome (cancer), date of disenrollment from database,reaching 80 years of age, date of death, or end of studyperiod.

Cohort study…. The cohort study will employ various methods to assess

the risk of cancer outcomes by CCB user status. CCBusers will be compared with CCB non‐users in terms oftheir incidence of cancer outcomes. Initial analysis will bemade comparing cancer occurrences from a period of sixmonths to one year of cohort entry.

Also analysis for duration and dose of CCB treatment inrelation to cancer outcomes as well as the cumulativenumber of prescriptions will be done.

Comparison of cancer outcomes occurring six months –1 year, 1‐4 years and >5 years following initiation of CCBtherapy will be done. This will take into account thelag‐time for clinical onset of cancer and the diagnosedelay.

Cohort study Preliminary data analysis will employ standard

methods such as non‐parametric univariate andbivariate analyses. Further analyses will include Coxproportional hazard models with time dependentcovariates.

the relative risk of cancer

Population based Cohort study

The population‐based study will describeprevalence of CCB use and prevalence andincidence of cancer outcomes within the studypopulation defined above, over the study period.

In order to be more representative of the population,all patients with a history of cancer before the startof the study will be included in the analysis.

Cox proportional hazard analysis with timedependent covariates will be used to calculate RRand 95 % Cl to compare cases and controls.

Nested case‐control study A nested case control study design identifying cases of cancer

outcomes between 1January 1996 and 31 December 2009, andmatching them on a ratio of 1:4 with controls that had not yetdeveloped any form of cancer before the date of cancer diagnosis ofthe case.

Cases will be matched to controls on same age (years) and gender. Analysis will be made regarding length of use as well as number of

prescriptions. Cancer outcome will be defined as six months – 1year, 1‐4 years and >5 years.

Conditional logistic regression analyses will be used to calculateodds ratios (ORs) and 95% CIs to compare cases and controls interms of exposure to CCBs (<1‐year, 1‐4 years and >5 years beforethe date of cancer diagnosis of the case) and duration of CCB use,as well as other potential risk factors for cancer outcomes

Covariates in regression analysis Key variables in the Cox proportional hazards

regression and conditional logistic regressionanalyses will include antihypertensive treatmentother than CCB, cardiovascular medication otherthan antihypertensives,

co‐prescription data (e.g. post‐menopausal hormonetreatment, NSAID, immunosuppressants)

the data specification document for description ofcardiovascular medication), Body Mass Index (BMI),smoking, alcohol consumption and

Comorbidities such as pulmonary disease

Other statistical methods

Propensity scoremodelingInstrumental variableanalysis

Limitations of study designs, Datasources andanalyticalMethods

The most important (potential) limitation to this study relates to thechallenge of obtaining quality information on important confounder data(such as smoking, BMI, and alcohol consumption).

Pharmaco‐epidemiological studies involving cancer as an outcomealways have the potential to be confounded by smoking in particular.

breast cancer, analysis will be restricted to women; it is unlikely that wewill be able to obtain data on age at menopause and information on oralcontraceptive use may be limited.

For colon cancer, we are unlikely to obtain information concerningdietary habits (relating to the consumption of fiber, fruit, vegetables andred meat) or physical activity.

Anticipated results of studies

Creation of a European database network and furtherdevelopment of methodological standards for the conduct of(multi-) national PE studies.

Methodological standards will be included when appropriatein the EMA (European Medicines Agency)-basedENCePP(European Network of Centres forpharmacoepidemiology and Pharmacovigilance) guidanceon methodological standards

Increasing methodological standards and registration ofstudy protocols to decrease discrepancies in results fromthese studies and increase transparency and therebyincreasing the usefulness and reliability of these studies forbenefit-risk assessment and decision- making of marketeddrugs in Europe and beyond.

Thank you