A comparison of individual patient data (IPD) over meta-analyses of published data, implications for clinical trial design and the utility of endpoints for early breast cancer trials. CRUK STUDENTSHIP AVAILABLE
Meta-analyses of data from all relevant randomised clinical trials are considered to be the most reliable form of summarising the evidence on particular therapeutic questions. Meta-analyses enhance statistical precision by including larger numbers of patients and events and allow more reliable investigation of interactions between patient characteristics and treatment effect.
There are typically two different approaches to collecting data for a meta-analysis: either using the results as published (published data meta-analysis) or collecting individual patient-level data (IPD) from all trials in a collaborative meta-analysis. The two approaches require differing amounts of resources, and yield data of differing richness. In particular, IPD meta-analyses can require a great deal of time and resource and, for this reason, remain rare compared to published data meta-analyses. The Early Breast Cancer Trialists’ Collaborative Group (EBCTCG) have, over the past 35 years, performed several IPD meta-analyses. The EBCTCG database of over 700,000 women in randomised trials allows an investigation of the comparative benefits of IPD and published data meta-analyses for specific features.
The primary aim of this DPhil proposal will be to investigate where the benefits of IPD meta-analyses over published data meta-analyses lie, with particular emphasis on breast cancer trials and EBCTCG data. It is proposed to compare the results obtained from the published data with those obtained from the EBCTCG IPD meta-analyses to quantify the added value of IPD over published data meta-analyses in determining risk or benefit of treatment on long-term clinical outcomes and how this depends on various patient and tumour characteristics.
Ancillary studies will be undertaken to investigate what cross-sectional analyses of all the data at different points in the past 30 years would have shown, a comparison of the findings of patient-level subgroup analyses with meta-regressions based on study-level characteristics and an investigation of the utility of composite outcomes such as disease-free survival (DFS) as an outcome measure compared to use of recurrence and breast cancer mortality. The dependence of measures such as DFS on length of follow-up will also be investigated using the methods outlined above.
RESEARCH EXPERIENCE, RESEARCH METHODS AND TRAINING
The student will work within a multidisciplinary team and will gain research experience in systematic literature reviews, meta-analysis of patient-level data, epidemiological and statistical methods, programming, data analysis, scientific writing and presentation of findings at scientific meetings.
FIELD WORK, SECONDMENTS, INDUSTRY PLACEMENTS AND TRAINING
The student will be part of the EBCTCG secretariat and will have all support required for undertaking the work. In addition, the student will have access to the range of training programmes offered by the University for postgraduate students including teaching and personal development.
This project involves statistical analysis of large scale data to improve breast cancer treatment. It requires the ability to acquire skills in data analysis and statistical programming or previous training/experience with and an interest in applying these skills to medical research.
Students should have a good degree in mathematical/science subject or medicine, and preferably an MSc in epidemiology, statistics, public health, or bioinformatics.