PhD defence A. (Alexandros) Rekkas

Below is a brief summary about the dissertation:

In order to provide the most optimal medical care, doctors are advised to align their clinical treatments with the results of well-conducted clinical trials, or the aggregated results from multiple such trials, assuming that all patients experience the same treatment effects—benefits and harms—as the reference trial population. However, the estimated treatment effect is often an average of heterogeneous treatment effects and, as such, may not be applicable to most patient subgroups, let alone individual patients. A patient’s baseline risk—her or his probability of experiencing an outcome of interest—is an important determinant of treatment effect. Low-risk patients can only experience minimal treatment benefit before their risk is reduced to zero, while high-risk patients can benefit much more. Hence, baseline risk is a crucial component of medical decision making. For that reason, many treatment guidelines are based on risk prediction models, i.e., mathematical functions relating the occurence of the outcome of interest to a set of measured predictors, developed on data available from clinical trials or observational studies.

Healthcare data is routinely collected by general practitioners, hospitals, insurance companies, and many other private or public bodies and is becoming increasingly available, giving researchers access to large amounts of patient data that can be used to support medical decisions. However, as prescribing physicians do not suggest treatments at random, analyses of observational healthcare data often suffer from confounding, i.e., patients receiving a treatment under study are systematically different from patients treated with the comparator, therefore complicating comparisons. In addition, the very diverse body of data collectors has resulted in a plethora of data structures, very often incompatible with each other, which further complicates the incorporation of multiple data sources in statistical analyses. Providing a common data structure like the Observational Medical Outcomes Partnership Common Data Model (OMOP-CDM) has the potential to enable large multi-cohort observational studies.

The overall aim of this thesis was to explore the use of baseline risk prediction models as the basis for medical decision making. We studied and applied methods for the evaluation of treatment effect heterogeneity in both clinical trial data and observational data. More specifically, we aimed to systematically review the existing literature on predictive approaches to the evaluation of heterogeneity of treatment effect, develop scalable and reproducible risk-based predictive approaches to the assessment of treatment effect heterogeneity, and apply risk-based methods to better guide medical decisions.