Applications are invited for a 3-year fully-funded PhD Studentship starting summer or autumn 2020. This is a PhD opportunity working in the biostatistics group in the Institute of Applied Health Research, University of Birmingham with supervision provided additionally from Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, Australia and the School of Dentistry, University of Birmingham.
Details of project: Routinely used medical treatments are often unevaluated and used heterogeneously across settings. At the level of the individual treatment differences may be small but aggregated across large populations differences become very important. Pragmatic trial designs are thus urgently needed to allow detection of small but clinically relevant differences in treatments in common use, for which no head-to-head evidence is available. Sufficiently powered trials to detect such small differences need to be very large. This calls for trials with large sample size, broad eligibility criteria, simple logistics, routinely collected outcome data and cost efficient designs. The multiple-period cluster randomised cross-over design, randomises clusters (e.g., hospitals) to alternating exposure to differing intervention conditions. This is a highly efficient trial design and with many logistical advantages it has real potential to bridge the gap to allow large scale assessment of the comparative effectiveness of competing interventions.
There have been several prominent examples of this trial design in the literature in recent years (including the NEJM). Yet, there is a dearth of methodological literature on how best to design these trials to ensure they provide robust (internally and externally valid) evidence. Some issues raised include: how to how to determine the frequency and number of cross-overs; the importance of a time-balanced design; and how to determine the required sample size; and how to analyse the subsequent data to ensure unbiased estimates of effects and confidence intervals.
In this PhD we propose a package of applied and methodological research projects that are important to making this trial design a success. Case studies will be chosen throughout to ensure practical value from this research and will align with other ongoing mpCRXO studies run in dentistry. Specifically:
WP1: Review of reporting and design properties of the mpCRXO studies. Whilst use of this trial design appears to be on the increase, there are a couple of very early examples. This project will systematically review and document all mpCRXOs conducted to date. This will inform WPs 1 and 2.
WP2: Statistical analytical project to determine the required number of cross-overs. One of the main appeals of the mpCRXO design is its statistical efficiency over the parallel CRT. This increase in statistical efficiency arises because each cluster receives both treatment conditions. These within-cluster comparisons have other advantages: in studies with small number of clusters, they eliminate chance imbalance between intervention and control on any cluster-constant characteristics. Furthermore, increasing the number of periods can help eliminate chance imbalance on any time-varying cluster characteristics. Work package 2 will determine the frequency of cross-overs to balance expected secular trends. This will be a theoretical work-package but tested on empirical data (real trends).
WP3: Analysis needs to account for cluster effects and any residual confounding due to period effects. Cluster effects are needed to allow for the non-independence of observations within clusters. Period effects refer to how the outcome changes over time (in absence of treatment). Analysis thus requires an amalgamation of time-series based approached and cluster analysis (mixed models). A series of simulation studies will identify optimal models for specific scenarios.
Applicants should have a strong background in statistics. They should have a commitment to medical statistics research and hold or realistically expect to obtain at least an Upper Second Class Honours Degree in statistics or a closely related subject, or (preferably) a Master’s degree in statistics or medical statistics.
How to apply
Applications should be directed to [email protected]
. To apply, please send:
• A Detailed CV, including your nationality and country of birth;
• Names and addresses of two referees;
• A covering letter highlighting your research experience/capabilities;
• Copies of your degree transcripts;
• Evidence of your proficiency in the English language, if applicable.
Applicants will be required to attend an interview. This can be conducted face –to –face, or by video conference