Cadth 2015 c2 panel.mohsen

15
How to integrate external evidence into within trial economic evaluations? Mohsen Sadatsafavi MD, PhD University of British Columbia 2015.04.13

Transcript of Cadth 2015 c2 panel.mohsen

Page 1: Cadth 2015 c2 panel.mohsen

How to integrate external evidence into within trial economic evaluations?

Mohsen Sadatsafavi MD, PhDUniversity of British Columbia

2015.04.13

Page 2: Cadth 2015 c2 panel.mohsen

2

Conflict of Interest

No apparent or perceived conflict of interest

Received CIHR fellow-ship award for this work

All interpretations are my own

Our TeamCarlo Marra, Shawn Aaron, Stirling Bryan

Page 3: Cadth 2015 c2 panel.mohsen

3

Outline

• A novel method for incorporating external evidence in economic evaluations

• Context in which it is applicable• What it does• A running example• Pros + Cons + conclusions

Page 4: Cadth 2015 c2 panel.mohsen

4

Economic trialsas vehicles for evaluations

• Model-based evaluations• Evidence is ‘parameterized’ and fed into a computer

program (mathematical function)

• simulation

• Trial-based evaluations• Individuals are assigned to competing treatments and

their experiences is represented through cost and effectiveness values

• (bivariate) statistical inference• bootstrap as s popular paradigm

Page 5: Cadth 2015 c2 panel.mohsen

5

Issues with trial-based CEAs

Failure to compare all available options

A truncated time horizon

Lack of relevance to the jurisdiction of interest

Failure to incorporate all evidence the most damning criticism

Page 6: Cadth 2015 c2 panel.mohsen

6

ContextAn example: OPTIMAL trial in COPD

• Mono, double, and triple inhaler therapy in COPD (T1, T2, T3)

• N=442• Follow-up 1 year• Primary outcome: exacerbation rate• Prospective economic evaluation component• Results

– RR of T2 v. T1: 1.01 (0.59 – 0.73)– RR of T3 v. T1: 0.84 (0.47 – 1.49)

Page 7: Cadth 2015 c2 panel.mohsen

7

ContextOPTIMAL CEA

• Sequence of imputation, regression, outcome calculations within a bootstrap envelope

Page 8: Cadth 2015 c2 panel.mohsen

8

Trialist versus economist

Treatment effects (RR=0.84, NS) ICER=243K

Pooled estimate

Regression-based methods Bootstrap methods

OPTIMAL trial

Trialist Economist

Decision maker

Welte et. al. T3 v. T1

RR=0.38 (0.28 – 0.57)

Meta-analyst

Page 9: Cadth 2015 c2 panel.mohsen

9

CONTEXTReconciling external evidence in RCT-based CEAs

Desist RCT-based CEA and resort to model-based evaluation Paradigm shift

Switch to parametric Bayesian evidence synthesisMCMC using WinBUGS (technical, lots of programing, model

convergence etc) Off-putting

Stay with the bootstrap-based methods, incorporate external evidence

Page 10: Cadth 2015 c2 panel.mohsen

10

MethodsRCT-CEA without external evidence

Page 11: Cadth 2015 c2 panel.mohsen

11

MethodsRCT-CEA with external evidence

Page 12: Cadth 2015 c2 panel.mohsen

12

ResultsCase study in COPD

Page 13: Cadth 2015 c2 panel.mohsen

13

ResultsCase study in COPD

Page 14: Cadth 2015 c2 panel.mohsen

14

Conclusions

• Theoretically, possible to use trials as ‘vehicle’ for evidence synthesis, yet avoid the ‘most damning criticism’

• Practically, with too many parameters, it can be computationally prohibitive

• Best use is in sensitivity and secondary analyses• What if we incorporate the results of the study by xyz?

• The experts in our team believe it is almost impossible that chemo+radiation increases the risk of metastasis ->half-flat priors

Page 15: Cadth 2015 c2 panel.mohsen

FacultyMohsen SadatsafaviJ Mark FitzGeraldStirling BryanLarry Lynd

Research StaffHamid TavakoliTania ConteRoxanne Rousseau

StudentsZafar ZafariWenjia Chen

Thank You!

[email protected]

Respiratory Evaluation Sciences Program (resp.med.ubc.ca)