Statistical Modelling 15 (6) (2015), 619–636

Validating predictors of therapeutic success: A causal inference approach

Ariel Alonso Abad
I-BioStat,
KU Leuven,
B-3000 Leuven,
Belgium
e-mail: Ariel.AlonsoAbad@kuleuven.be

Wim Van der Elst
I-BioStat,
Universiteit Hasselt,
B-3590 Diepenbeek,
Belgium


Geert Molenberghs
I-BioStat,
KU Leuven,
B-3000 Leuven,
Belgium


and


I-BioStat,
Universiteit Hasselt,
B-3590 Diepenbeek,
Belgium


Abstract:

In personalized medicine medical decisions, practices and/or products are tailored to the individual patient. The idea is to provide the right patient with the right drug at the right dose at the right time. However, our current lack of ability to predict an individual patient's treatment success for most diseases and conditions is a major challenge to achieve the goal of personalized medicine. In the present work, we argue that many of the techniques often used to evaluate predictors of therapeutic success may not be able to answer the relevant scientific questions and we propose a new validation strategy based on causal inference. The methodology is illustrated using data from a clinical trial in opiate/heroin addiction. The user-friendly R library EffectTreat is provided to carry out the necessary calculations.

Keywords:

causal inference; Personalized medicine; prediction of therapeutic success.s

Downloads:

Code in zipped archive and example data here. These data are available for download for all interested parties, however the user first has to register on the website and agree with the conditions that are specified there (the NIDA Data Share Agreement).
back