Monitoring Drug Safety in Clinical Trials

Identifying and collating statistical methods suitable for analysis of AEs

Team: Rachel Phillips, Odile Sauzet, Victoria Cornelius

An evaluation of current statistical research methods available to analyse adverse events and detect signals of adverse drug reactions in clinical trials. 

The protocol for this project is available here.


Exploring and developing model-based approaches to identify signals of adverse drug reactions (ADRs) in RCTs

Team: Rachel Phillips, Odile Sauzet, Victoria Cornelius

The causal mechanism of an adverse drug reaction (ADR) means that the occurrence is often time related, as a result time-to-event models can be used to detect a non-proportional hazard to signal an ADR on an exposed cohort only. In randomised controlled trials, where there is a valid control group but smaller sample size, we are currently exploring statistical models with the aim of developing signal detection tests that utilise this control group under a time-to-event framework. The tests will provide a means by which to analyse non-specific emerging AEs in a RCT with greater accuracy than current approaches which predominantly rely on subjective assessment or the use a Chi-squared test/Fisher's exact test. We are also exploring incorporation of Bayesian analysis for application with prespecified AEs of interest in order to incorporate prior trial information.

visuals image v2.PNG

A consensus to support researchers in their choice of visualisations for RCT publications

Team: Rachel Phillips, Victoria Cornelius

A well-designed graphic is an effective means by which to summarise and communicate important messages to a range of audiences. In clinical trials, where there is an abundance of complex harm data, visualisations could offer a distinct advantage over presenting tables of AEs. A methodology review identified many plots proposed specifically for AE analysis in RCTs, few of which are used in practice. You can find examples of such visualisations here. With such a variety of visualisation options available, we are seeking a national consensus to support researchers in their choice of visualisations for RCT publications.

If you are working in RCTs with a strong interest in AE analysis, and would like to take part in this research, please contact Rachel Phillips by April 1st 2020. You can find more information about the day here