We start by setting the scene, with a brief reminder of the randomised controlled trial, the gold standard study design to evaluate the effectiveness of interventions in healthcare, and the role of statistics within such studies.

Clinical trials collect a huge amount of data. During this activity, we will learn about the different types of data commonly collected in trials and how we can best summarise these data to make sense of them.

Next, we consider ways to measure the impact of different interventions on health outcomes, and estimate how effective one intervention is compared to another.

An estimate of treatment effect comes from a single sample drawn from the population. A different sample would lead to a different estimate. So what range of values could the true treatment effect plausibly take?

We now consider hypothesis testing, used to decide whether the data observed in our trial supports a particular hypothesis. For example, are two interventions equally effective, or is one superior to the other?

A lot of the statistical methods we have covered so far assume that data follows a normal distribution. What can we do when this is not the case?

How do we determine how many participants to recruit to a trial and why does sample size matter?

Interest often lies in the time taken until a specific event for each participant, such as resolution of symptoms, death, or recurrence of disease. Such data requires alternative statistical methods.

Common problems that affect virtually all clinical trials are missing data and non-compliance with planned interventions. What are the implications of such problems and how are they best dealt with?

Statistical analyses need to be planned in advance and reported transparently and accurately. Published guidelines exist for analysis plans, protocol and result papers.

The course ends with an introduction to subgroup analyses and Bayesian statistical methods, and common problems seen in published trial analyses.