Seminar: "Leveraging big data to study causal mechanisms in complex diseases"

AUEB STATISTICS SEMINAR SERIES 2023-24

Ioanna Tzoulaki (Research Professor Biomedical Research Institution| Academy of Athens & Professor of Chronic Disease Epidemiology Schools of Public Health| Imperial College London)

Leveraging big data to study causal mechanisms in complex diseases

Room: Troias Amphitheater

ABSTRACT

Technological advances have profoundly transformed epidemiological studies and epidemiological investigations by enabling generation of vast amount of molecular data (genetics and other omicron technologies) as well as access to vast amounts of data from electronic health records, wearable devices, and imaging. Although many investigations focus on using these data to built advanced predictive algorithms for disease prognosis and diagnosis, a separate body of work has evolved towards causal inference. I will present here two examples, Mendelian Randomisation (MR) as a method to support causality and emulation of clinical trials using observational data. MR estimates causal effects in observational studies by leveraging genetic variants as instrumental variables, mimicking the randomization process in randomized controlled trials. Two-stage least squares regression, inverse variance weighting, and robust regression techniques, are used to estimate the causal effect of the exposure on the outcome while accounting for the genetic instrument's properties and potential biases. The MR studies in the medical literature are increasing exponentially and new MR methods are suggested of multivariate effects. The second paradigm attempts to emulate clinical trials leveraging electronic health record data. To estimate the treatment effects avoiding common biases, common approaches include propensity score methods, instrumental variable analysis, average treatment effects as well as methods from other fields such as regression discontinuity. During the seminar, I will focus on examples on how MR and emulation of clinical trials in observational data can be used in combination to examine drug repurposing opportunities.

Presentation slides can be found here.