Method - Analysis

Clinical trial data is riddled with gaps—patients drop out, miss appointments, or switch treatments—and this research programme builds the statistical tools to handle those gaps without distorting the results. Why this matters: When a trial loses track of what happened to a patient, the entire analysis can become unreliable. The standard fix—simply ignoring missing data—often introduces bias, making a treatment look better or worse than it really is. This programme tackles that head-on, developing practical methods for making defensible assumptions about missing outcomes and correctly incorporating them into trial interpretation. It also addresses deeper design problems: defining the right patient population, choosing meaningful outcome measures, and accounting for unplanned events like treatment intolerance or switching. Potential impact: If these methods become standard practice, future trials—especially in tuberculosis and cancer—will produce more trustworthy results. That means doctors and regulators can make better decisions about which treatments actually work. The programme also uses computer simulation to test how resilient a proposed trial would be to real-world disruptions, and collaborates with the Alan Turing Institute to apply machine learning for identifying patient subgroups that benefit most from specific treatments. The result is not a single breakthrough, but a quieter, systemic improvement in how clinical evidence is generated and interpreted.