Lectures and seminars SfoEpi Seminar Series in Biostatistics: Associate Professor Ruth Keogh

Abstract:

Randomized controlled trials are the gold standard for estimating causal effects of treatments on health outcomes, but can be infeasible or unethical. Longitudinal observational data offer the possibility of estimating treatment effects over long periods of follow-up and in diverse populations. However, to do this we must tackle the challenge of time-dependent confounding.

Several methods have been described for estimating causal treatment effects on survival using longitudinal observational data, the most commonly implemented being marginal structural Cox models. This work focuses on the “sequential trials” approach, which involves creation of a sequence of artificial trials from new time origins within an observational cohort. Despite being intuitive and straightforward to implement, this approach has not previously been considered in detail or compared with alternative methods. I will outline the sequential trials approach and discuss what is being estimated, using the potential outcomes framework. It will be contrasted with alternative methods, both theoretically and using simulations.

This work is motivated by a question about the effect of the treatment DNase on the survival of people with cystic fibrosis. I will present some results from applying the sequential trials approach to data from the UK Cystic Fibrosis Registry to answer this question.

If you have any questions, please contact Erin Gabriel at Erin.Gabriel@ki.se