Event report: Perspectives on statistics in medicine annual joint LSHTM-RSS lecture

Perspectives on statistics in medicine annual joint LSHTM-RSS lecture
It’s about time - asking better questions for causal inference with time-dependent data
Speaker: Professor Vanessa Didelez
Date: Thursday 19 June 2025, 5.30PM (held online)
 
The 34th Joint LSHTM-RSS annual lecture on perspectives on statistics in medicine (formerly known as the Bradford Hill Memorial Lecture) took place online this year. The speaker was Professor Vanessa Didelez, the Professor of Statistics and Causal Inference at the Leibniz-Institute of Prevention Research and Epidemiology – BIPS, Bremen, Germany, in a joint appointment with the Department of Mathematics and Computer Science of the University of Bremen.

Professor Didelez began her talk by discussing crucial statistical considerations often overlooked in research: what she called “structural” and “wrong question” biases. She identified the importance of eliciting clear causal questions and why this is relevant in practice . This leads to the question of estimands (or simply put, what is it we want). These translate these causal questions into precise, quantifiable targets, guiding the choice of methodology and ensuring that the statistical analysis directly answers the scientific question.

Professor Didelez then focused on design biases in observational studies and randomized controlled trials (RCTs). Using the example of screening colonoscopy for colorectal cancer, she showed how prevalent cases (pre-existing cancers detected by screening) can inflate perceived benefits, leading to an "exaggerated effect." Professor Didelez advocated for "target trial emulation" to clarify the precise research question and account for these complexities, emphasizing that simply observing a reduction in incidence due to early detection is distinct from demonstrating a true preventive effect.

The discussion then extends to the issue of "competing events" in time-to-event analysis, using dementia research as an illustration. When a competing event, such as death, occurs before the event of interest (dementia onset), it significantly complicates interpretation. A seemingly counter-intuitive finding that smoking "prevents" dementia is presented as a demonstration of how a "total causal effect" can be misleading if the underlying competing risk (higher mortality in smokers) is not explicitly considered. Finally, Professor Didelez presented an example about prediction modelling, discussing two types of “what if”  and decision support.
In summary, the lecture stressed the paramount importance for researchers to clearly define their research questions and consider competing events to avoid misinterpreting findings and ensure robust statistical inference. Questions followed, particularly interested in competing risks in medical applications. The event concluded with thanks for an entertaining and insightful lecture.
 
Report written by Nathan Green on behalf of the RSS Medical Section
 
Load more