Two-stage meta-analysis methods and the validity of the fixed-effects approach

Meta-analysis is the analysis of multiple results from individual studies (often clinical trials) to synthesise the findings. It is increasingly used as the basis for health technology assessment and clinical guidelines.

Despite the popularity of meta-analysis, there are still important debates about the best methods to apply. Some of these debates concern the optimal methods to combine results from multiple trials into a weighted average. Decisions are usually based on the variability (‘heterogeneity’) among studies, which may reflect differences in patients in the trials, or aspects of the trial itself. If heterogeneity is detected, analysts will attempt to model it using a random effects model. If variability estimates are low, they may use a fixed-effects model.

Simulation work at the Health Sciences Research Group suggests that current methods for detecting heterogeneity can be flawed, which raises questions about current procedures and the potential for bias.

This PhD research will build on our previous work to explore these critical issues further. The successful candidate will:

- Test performance of variance estimators in the presence of non-normality
- Research a new random-effects model that will be more flexible in detecting heterogeneity
- Generate software to aid meta-analysts

This project would suit a person with a background in quantitative methods, who wishes to develop advanced quantitative skills for applied health care applications. Although the focus of the PhD is on meta-analysis, the candidate will develop a solid grounding in longitudinal and multilevel modelling, design and analysis of clinical trials, and advanced casual modelling. We expect the PhD to provide an excellent platform for further post-doctoral applications to the National Institute of Health Research, which has identified advanced methodological skills as a strategic priority for capacity development.

This 3-year project is open to UK/EU and non-EU nationals but no funding is provided. Applicants must therefore be able to evidence their ability to provide self-arranged funding. Annual tuition fees for this project are currently:

*UK/EU nationals: £6, 600
*Non-EU nationals: £17, 800

Enquiries from suitable candidates without existing funding are welcomed as it may be possible to apply for alternative external/internal funding at a later stage.

Applicants should hold a minimum upper-second honours degree (or equivalent) in statistics, mathematics, psychology, social sciences or a related area. A Masters qualification in a similar field with a significant quantitative component or extensive relevant professional experience is also essential.

Please direct applications in the following format to Dr Evan Kontopantelis ([Email Address Removed]):

• Academic CV
• Official academic transcripts
• Contact details for two suitable referees
• A personal statement (750 words maximum) outlining your suitability for the study, what you hope to achieve from the PhD and your research experience to date.
• Evidence of funding.

Any enquiries relating to the project and/or suitability should be directed to Dr Kontopantelis at the address above. Applications are invited on an ongoing basis with the project commencing October 2012, or potentially early into 2013.

http://www.medicine.manchester.ac.uk/staff/EvangelosKontopantelis
http://www.medicine.manchester.ac.uk/healthsciences/