- To provide a foundational analysis on statistical/causal inference with a focus on the critical assessment of current practices in drug approval and pharmacosurveillance.
- To build a unified epistemic framework within which different kinds of evidence for pharmaceutical harm can be combined and used for decision: evidence amalgamation.
- To provide a theoretical framework for the development of new standards of drug evaluation.
The project addresses these issues by advancing a system for evidence amalgamation for the purpose of probabilistic causal assessment; i.e. where rather than grading the evidence, we grade the hypothesis of the causal link. Although the system integrates various pieces of evidence through an epistemic Bayesian network, it is intended to incorporate any kind of statistical data (e.g. also results from classical Fisherian or Neyman-Pearson tests), or any kind of studies (from laboratory experiments on certain cell cultures, to animal studies, up to machine learning analysis of “big data”).
The theoretical foundation of the system are presented in Landes et al. 2017 (see below); additional publications are in progress where each layer of the framework, from the more abstract level of “causal indicators” (derived from Bradfrod-Hill guidelines) to the lower level of concrete data, are spelled out in detail.
The project “template” can be represented by the following picture:
Fig. 1: A Multilayer Approach to Modelling Probabilistic Causal Inference through Evidence Synthesis: a. issues related to the evidential support provided by the evidence to the hypothesis at hand as discussed both in the statistical literature as well as in the philosophy of science; b. issues related to meta-evidential dimensions: consistency of studies, structure of the body of evidence in terms of mutual dependence of observations, reliability of the pieces of evidence and their relevance with respect to the target group; c. social dimensions of the pharmaceutical ecosystem (funding structures, reputational concerns, regulatory constraints), generally discussed in the social epistemology literature.
The three layers represent three levels of the inferential problem at hand: the overall assessment of the causal link under consideration. This depends on:
a. the basic level of evidential support to the hypothesis provided by the available evidence (and various evidence synthesis techniques);
b. a higher order level of “meta-evidential” (or second order) dimensions of evidence related to the individual items of evidence and to the body of evidence in its entirety: consistency of study results, topological structure of the evidence – whether (some) of the items of the observations are dependent – reliability of the evidence (accuracy and internal validity), relevance (external validity);
c. an additional level, related to the information/evidence concerning these meta-epistemic dimensions (e.g. financial interests, reputation concerns, legal constraints).
These levels have been working relatively independently so far, especially in the standard Evidence Based Medicine approach. Our project goes in the “meta-analytic” direction advocated by Andrew Gelman (“Working through some issues.” Significance 12.3 (2015): 33-35.).
Recently published work:
- Landes, J., B. Osimani, R. Poellinger (2017): Epistemology of Causal Inference in Pharmacology. Towards a Framework for the Assessment of Harms. EJPS