Develop new methodology for the analysis and reporting of meta-analyses in the presence of Heterogeneity
Prof P Sasieni
Dr J Waller
No more applications being accepted
Funded PhD Project (European/UK Students Only)
We have two studentships available across a possible four projects, which includes this project. The other three are available to view on Find a PHD or at www.kcl.ac.uk/health/study/studentships/div-studentships/cps/cps-studentship-projects
Most meta-analyses use random-effects models in the presence of heterogeneity, but many analyses do not consider the likely sources of heterogeneity and their implications. Non-compliance and contamination in trials of cancer screening, and poor adherence to protocol in delivering a complex intervention will tend to attenuate the magnitude of “treatment effects”. If the treatment effect is heterogeneous then one should be interested in what features are associated with larger treatment effects. If we consider each trial to have a random treatment effect, then we should be interested in the proportion of trials (random effects) that are clinically important (e.g. what proportion (with a 95% confidence interval) of screening effects result in at least a 20% reduction in cancer-specific mortality). This is related to tolerance intervals (which are used in engineering but rarely in medical statistics). The student will review the literature regarding random effects meta-analysis and tolerance intervals and derive new methodology to address these questions. The methodology will be applied to RCTs of lung cancer screening and to observation studies of the association between excision of CIN and subsequent pre-term delivery.
£21,000 annually. The Studentship is open to UK & EU candidates.