Autor:
Andres Vaher

Teadusseminar "Disentangling causal relationships in epidemiology: some recent trends" (Krista Fischer)

 

Research Seminar "Disentangling causal relationships in epidemiology: some recent trends"

 "Põhjuslike seoste lahtiharutamine epidemioloogias: uuemad trendid"

Kolmapäeval, 15. novembril 2023 kell 11.00-12.00  

Krista Fischer (Tartu Ülikool)

As we all know, detecting a statistical association in data is much easier than to establish causality. However, in epidemiology (similarly to many other fields), estimating the causal effects is the main purpose of most studies, as understanding the causal mechanisms leading to disease is a necessary prerequisite for any evidence-based recommendations and interventions. 

A gold standard in evidence-based medicine is to rely on the results of randomized controlled trials (RCTs) - however, there are many occasions where conducting an RCT would be either impossible or unethical. Therefore there is a lot of ongoing work on developing and improving the methodology for causal inference in observational studies. The expansion of biobanks has brought the immense popularity of Mendelian Randomization - the application of Instrumental Variable modelling, where genetic variants are used as instruments.  Although this approach has helped to answer some burning questions (such as the question on the effect of moderate alcohol-drinking on heart disease), it is not always possible to find appropriate genetic instruments. A quickly developing approach is "Target trial emulation", where the idea is to obtain a result that is as close as possible to a RCT. 

In the seminar these ideas are explained with some examples, where we have applied them on the Estonian data