Skip to the content.

Novel statistical methodology to assess the reliability of observational data for heterogeneous treatment effect estimation

Project ID: 2531ac1475

(You will need this ID for your application)

Research Theme: Healthcare Technologies

UCL Lead department: Statistical Science

Department Website

Lead Supervisor: Brieuc Lehmann

Partner Organisation: Novo Nordisk A/S

Stipend enhancement: £2,500 per year

Project Summary:

Observational or “real-world” data (RWD) is increasingly used to assess treatment effect heterogeneity in drug evaluations. While randomized controlled trials (RCTs) remain the regulatory “gold standard” for their internal validity, they are often underpowered for estimating heterogeneous treatment effects. In contrast, RWD can provide insights into treatment effects across diverse populations but may suffer from confounding, missing data, and measurement error, which bias treatment effect estimates. Additionally, differences between participants in RCTs and those in observational datasets, known as covariate shift, complicate comparisons between the two.

This PhD project aims to develop rigorous statistical methods to evaluate the reliability of RWD for treatment effect estimation. Leveraging target trial emulation, we will benchmark observational studies against equivalent RCTs, identifying key factors (e.g., confounding, missingness, measurement error, covariate shift) driving disparities in treatment effect estimates between RWD and RCTs. Additionally, we will create methods to enable valid inference for heterogeneous treatment effects (HTEs), facilitating comparisons despite larger sample sizes in RWD. Our objective is to develop model-independent techniques and best practices that can be applied broadly across various settings, offering more robust evaluations of drug safety and efficacy. These advancements will ultimately support a more efficient and reliable drug approval process.

The awarded PhD student will be supported by a team of experts across UCL Statistical Science (including Dr Brieuc Lehmann and Professor Karla Diaz-Ordaz) and Novo Nordisk. The student will also undertake a minimum of 3 months placement with Novo Nordisk during the PhD.

Requirements: A strong background (e.g. Master’s degree) in statistics or a closely related field such as machine learning Research interests in one or more of the following: Clinical trials; Epidemiology; Causal inference; Generalizability

Desirable: Strong statistical programming skills (e.g. R, Python, Julia)