Blue pills on a table

What is Biostatistics?

The Biostatistics Theme develops statistical solutions for research in epidemiology, population biology, and medicine

The theme includes researchers working on:

  • Bayesian methods for infectious disease modelling
  • Improved statistical efficiency in adaptive clinical trials
  • The statistics of event history data in health
  • Methods for improving the value of healthcare in the population.

Our research is typically inspired by our close relationship with health agencies in the UK, including NHS and UKHSA, as well as application-focused academic and government collaborations world-wide. We collaborate closely with many other parts of the Lancaster campus, including Lancaster Medical School, the Department of Health Research, and Biomedical Life Sciences, and support early career researchers including NIHR pre-doctoral fellows.

Infectious disease modelling

Infectious disease models are a popular method for investigating how interactions between individuals within a population give rise to large-scale disease outbreaks, such as Covid19 in humans, foot and mouth disease in livestock, and influenza in all animal species. To be useful in practice, these models must first be calibrated by fitting to existing disease data, with the challenge that important events such when individuals get infected are unobserved: you know when you start feeling ill, but you don't know when you actually got infected!

Our researchers work alongside collaborators in health agencies and population health research organisations, developing Bayesian methods to fit infectious disease models in the presence of unobserved events. This brings together statistical innovations and cutting-edge high-performance computing to provide solutions to aid real-time outbreak management and future public health policy. Through this work, we have advised many high-impact organisations, such as the SPI-M-O subcommittee of SAGE during the Covid19 pandemic, the Animal and Plant Health Agency, and World Health Organisation.

Statistical methods for clinical trials

Adaptive designs are increasingly used in clinical research to improve trial efficiency and to account for heterogeneity in patient's responses to treatment. Recent developments in this area include the introduction of master protocol trials (such as platform, basket and umbrella trial designs), adaptive enrichment designs, and response adaptive trials for small populations. A particular methodological interest of the group is in inference following an adaptive design, such as testing for response adaptive trials or parameter and interval estimation.

Dose-finding is a crucial part of early-phase clinical trials and principally involves determining the maximum tolerated dose of a new drug or combinations of drugs. Recent work in collaboration with pharmaceutical companies includes determining the most appropriate dose-finding strategies for therapies with late-onset toxicities.

Modelling of event history data

Event history data arises from studies which follow patients over time and either record the times of clinical events, or record health status, or health-related quality of life at a series of discrete examination times. Methodological interest lies in assessing, or relaxing key assumptions used in fitting multi-state models to such data, and accounting for non-standard observation schemes, such as dual censoring, informative observation times, state misclassification, or dealing with missing covariate data.

An important application is health economic evaluations, where the main statistical challenge is the estimation of expected quality-adjusted life years (QALY) from event history data, usually also requiring survival extrapolation. The group is involved in the Horizon-Europe PREVENTABLE project to assess the cost-effectiveness of preventative treatment for rare tumour risk syndrome patients, where the primary methodological challenge is in estimating transition probabilities from small, and possibly, distinct populations.