Integrating patient preferences and clinical trial data
in a Bayesian model for benefit-risk assessment
H Broekhuizen1, CGM Groothuis1, AB Hauber2 and MJ IJzerman1
(1) University of Twente, the Netherlands
Patient preferences in benefit-risk assessment
Patient preferences matter
They experience the benefits and risks
Preferences can differ between regulators and patients Example: Natalizumab case
Growing interest (FDA, EMA)
The MCDA model
Clinical trial data Approximation Patient preferences Uncertainty Uncertainty Preferencestudies Clinical trials
Simulation
Case study: antidepressants
Patient preferences Uncertainty Preference studies Preferences: AnalyticalHierarchy Process study by Danner et al. (2011)
Respondents: 12 MDD patients
Benefit criteria: response and remission
Risk criterion: adverse events (low and high severity)
Benefit and risk outcomes assumed to be independent Approximated by a bootstrap
How is preference information approximated?
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1 2 3 4 5 6 7 8 9 10 11 12 W eigh t Respondent Adverse Events Remission Respons Respons weight F re q u e n cy (t o ta l= 1 0 0 0 ) 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0 50 100 150 Approximated by a bootstrap resampling methodCase study: antidepressants
Clinical trial data Uncertainty Clinical trials Performance: Systematicreview by German Institute for Quality and Efficiency in
Health Care (IQWIG) Drugs: Duloxetine,
Venlafaxine and Bupropion Odds ratio compared to
How is clinical data approximated?
Dulo x e tin e V e n laf a x ine B u p rop io n 0.5 1.0 2.0 5.0 0 2 4 6 8 OR compared to placebo Approximated by a normal distribution in the log domainBupropion remission performance
OR compared to placebo F re q u e n cy (t o ta l= 1 0 0 0 ) 1.0 1.2 1.4 1.6 1.8 2.0 0 20 40 60 80 100 120
For a particular drug i in simulation run t, 𝐵𝑒𝑛𝑒𝑓𝑖𝑡𝑠𝑖,𝑡 = 𝑊𝑊𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑡 𝑟𝑒𝑚𝑖𝑠𝑠𝑖𝑜𝑛𝑡 ⋅ 𝑂𝑅𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑖,𝑡 𝑂𝑅𝑟𝑒𝑚𝑖𝑠𝑠𝑖𝑜𝑛𝑡,𝑖 𝑅𝑖𝑠𝑘𝑠𝑖,𝑡 = 𝑊𝐴𝐸𝑡 ⋅ 𝑂𝑅𝐴𝐸𝑖,𝑡 𝐵𝑒𝑛𝑒𝑓𝑖𝑡𝑅𝑖𝑠𝑘𝑅𝑎𝑡𝑖𝑜𝑖,𝑡 = 𝐵𝑒𝑛𝑒𝑓𝑖𝑡𝑠𝑖,𝑡 𝑅𝑖𝑠𝑘𝑠𝑖,𝑡
Benefits and risks plotted in risk-benefit plane
Risk-benefit plane
μ=1 line W ei gh ted risks (OR)Weighted benefits (OR)
μ is decision threshold, μ=1 requires
drugs to have >1 weighted benefit for each weighted risk to be acceptable, i.e:
Benefit-risk-ratio>μ benefits outweigh risks
Percentage points under line approximates P(acceptable)
Risk-benefit plane
0.0 0.5 1.0 1.5 2.0 0 .0 0 .2 0 .4 0 .6 0 .8 1 .0 W e ig h te d r isks (o d d s ra ti o co m p a re d t o p la ce b o ) Bupropion Venlafaxine Duloxetine μ=1 line W ei gh ted risks (OR)Sensitivity
What is the impact of uncertainty surrounding model parameters Important distinction (Felli 1998)
Value sensitivity (change in expected value)
Decision sensitivity (change in decision, i.e. other drug chosen) Ranking sensitivity (change in rank order of drugs)
Why would we want to know? Robustness
Heterogeneity
0.15 0.20 0.25 0.30 0 1 2 3 4 5 6 7 D u lo xe ti n e B e n e fi t-ri sk ra ti o
Value sensitivity
95% CI Du lo x etin e be ne fit -risk ratioAdverse events weight
μ=1 μ=3
0.10 0.15 0.20 0.25 0.30 0.35 0 .0 0 .2 0 .4 0 .6 0 .8 1 .0
Decision sensitivity
95% CIAdverse events weight
P( Du lo x etin e acce pta bl e ) at μ =3
1.5 2.0 2.5 3.0 0 .0 0 .2 0 .4 0 .6 0 .8 1 .0 V e n la fa xi n e ra n k p ro b a b il it y
Ranking sensitivity
Rank reversal P(rank=1) P(rank=2) P(rank=3)Venlafaxine response performance (OR compared to placebo)
Prob
ab
A B Parameter 1 Parameter 2 Parameter 4 Parameter 5 Parameter 3 Parameter 6 80% 20%
Discussion
Integration preferences and performance Impact uncertainty can be assessed
Visual representations can help regulators and enrich the discussion during the benefit-risk assessment proces. Assumptions in antidepressants case
Simplified structure Independence
What probability is convincing?
Other methods needed to check external validity
Thank you
Email: h.broekhuizen@utwente.nl
References
M. Danner, J. M. Hummel, F. Volz, J. G. van Manen, B. Wiegard, C.-M. Dintsios, H. Bastian, A. Gerber, and M. J. Ijzerman, “Integrating patients’ views into health technology assessment: Analytic hierarchy process (AHP) as a method to elicit patient preferences.,” International journal of
technology assessment in health care, vol. 27, no. 4, pp. 369–75, Oct.
2011.
“Selective serotonin and noradrenalin reuptake inhibitors (SNRI) with
depression patients [Selektive Serotonin- und Wiederaufnahmehemmer (SNRI) bei Patienten mit Depressionen],” Cologne, 2010.
J. Felli, “Sensitivity analysis and the expected value of perfect