Post on 13-Dec-2015
69th meeting of EWG-MCDA, Brussels
Multi-criteria decision analysis in drug benefit-risk assessment
T. Tervonen(1), D. Postmus(2), H.L. Hillege(3)
1 Faculty of Economics and Business, RUG.nl2 Department of Epidemiology, UMCG.nl3 Department of Cardiology/Epidemiology, UMCG.nl
69th meeting of EWG-MCDA, Brussels
>Introduction
Drug Benefit-Risk (BR) analysis aims to systemically compare the benefits and risks of drugs within a therapeutic group
BR analysis has multiple possible applications- Support prescription decisions- One criterion for drug marketing
authorization decision (in Europe, FDA in USA doesn’t give incorporate BR analysis in clinical assessment)
Benefit-risk assessment
Data and evidence
Regulatory Logic
69th meeting of EWG-MCDA, Brussels
>Two ways to approach BR analysis
>Universal model Becomes too general Explicitly requires
qualitative measurements
Hard for MD’s to accept
Doesn’t show the potential of MCDA
>Therapeutic group specific model Allows to take into account
quantitative clinical data The model can be discussed
with leading experts of the therapeutic area
Separates qualitative judgments from clinical data
69th meeting of EWG-MCDA, Brussels
>Clinical data
Drug 1 Drug 2 Drug 3
Therapeuticgroup
Endpoint A Endpoint b
Study 1 Study 2 Study 3 Study 4
Endpoint c
69th meeting of EWG-MCDA, Brussels
> SMAA approach to BR analysis Step 1: Analyze without preference
information to characterize the drugs Step 2: Analyze through common scenarios
including ordinal preferences obtained from expert MD’s
> Justification for SMAA:1. Allows missing/incomplete preferences2. Gaussian distributed criteria values3. is based on MAUT
69th meeting of EWG-MCDA, Brussels
>Example Therapeutic group: Second-generation anti-
depressants Drugs:- Fluoxetine (Prozac)- Paroxetine (Seroxat)- Sertraline (Zoloft)- Venlafaxine (Effexor)
Purpose: Analyze trade-offs based on clinical data to support prescription decision for two scenarios:- Mild depression- Severe depression
69th meeting of EWG-MCDA, Brussels
> 1 benefit criterion (efficacy), a primary endpoint in studies of the 4 drugs
> 5 risk criteria corresponding to the 5 most frequent adverse drug events
> Measurements from meta-analysis that pooled results of compatible studies
Name Measurements Pref. dir. Scale
EfficacyRelative value compared
with Fluoxetine↑ [0.97, 1.23]
Diarrhea ADE’s
Absolute % ↓ [1, 20.6]
Dizziness ADE’s
Absolute % ↓ [2.9, 24.4]
Headache ADE’s
Absolute % ↓ [8, 31.3]
Insomnia ADE’s
Absolute % ↓ [3.4, 21.3]
Nausea ADE’s
Absolute % ↓ [22.1, 34]
69th meeting of EWG-MCDA, Brussels
>Measurements (mean, stdev)
Drug Efficacy Diarrhea Dizziness Headache Insomnia Nausea
Fluoxetine 1, 0 11.7, 2.5 7.2, 1.45 16.6, 3.27 13.7, 1.89 18.6, 1.79
Paroxetine 1.09, 0.06 9.2, 1.86 10.6, 1.58 21.2, 5.15 14.3, 2.93 18.3, 3.7
Sertraline 1.1, 0.05 15.4, 2.65 7.5, 1.48 20.2, 3.78 15, 3.21 19.5, 2.6
Venlafaxine 1.12, 0.05 5.5, 2.32 15.7, 4.44 12.8, 2.45 11.2, 3.98 31, 1.68
69th meeting of EWG-MCDA, Brussels
>Measurements (mean, stdev)
Drug Efficacy Diarrhea Dizziness Headache Insomnia Nausea
Fluoxetine 1, 0 11.7, 2.5 7.2, 1.45 16.6, 3.27 13.7, 1.89 18.6, 1.79
Paroxetine 1.09, 0.06 9.2, 1.86 10.6, 1.58 21.2, 5.15 14.3, 2.93 18.3, 3.7
Sertraline 1.1, 0.05 15.4, 2.65 7.5, 1.48 20.2, 3.78 15, 3.21 19.5, 2.6
Venlafaxine 1.12, 0.05 5.5, 2.32 15.7, 4.44 12.8, 2.45 11.2, 3.98 31, 1.68
Not asignificantdifference!
69th meeting of EWG-MCDA, Brussels
> SMAA analysis without preferences: central weights and confidence factors
> Can be used in describing the most preferred drug taking into account the patient history
0
5
10
15
20
25
Efficacy Diarrhea Dizziness Headache Insomnia Nausea
%
Fluoxetine
Paroxetine
Sertraline
Venlafaxine
CF
46%
53%
34%
68%
69th meeting of EWG-MCDA, Brussels
>Ordinal preferences Expert in the field of anti-depressants could
understand the model and rank the criteria swings during a short teleconference (30min)
Two rankings for the two scenarios:- Mild depression: Diarrhea > Nausea > Dizziness
> Insomnia > Headache > Efficacy- Severe depression: Similar ranking, except
efficacy the most important criterion Ranking took into account swings, and was
justified through clinical practice
69th meeting of EWG-MCDA, Brussels
>SMAA analyses with preferences: rank acceptabilities
>Can be used for scenario-based prescription
Mild depression
Severe depression
69th meeting of EWG-MCDA, Brussels
>Conclusions We constructed a therapeutic group specific
SMAA model for benefit-risk assessment of second-generation anti-depressants
Separation of clinical data from preferences gives “credibility” to the model
The problem statement is not “choice” or “ranking”, but “risk assessment”
Merci !