Methods

Our research programme addresses a variety of different methods for compiling, assessing and combining existing evidence from studies in genetic epidemiology. The main emphasis is on combining evidence from multiple, sometimes disparate, sources; this is the process that is often termed ‘meta-analysis’.

Our methodological projects include:

  • Developing methods for examining departure from Hardy-Weinberg equilibrium (HWE) and its empirical relationship with magnitude of association. In a study of 239 published associations we found that reporting of HWE is often neglected and that departures are rarely admitted. We have developed methods for studying the dependence of magnitude of association on degree of departure from HWE, and have observed in a review of 42 meta-analyses that statistically significance of results can be sensitive to inclusion studies that deviate from HWE.
  • Examining tools for assessing susceptibility to bias in observational epidemiological studies. We have reviewed existing tools with a view to developing one for use in systematic reviews of genetic association studies.
  • Developing methods for integrating sources of ‘partial’ evidence. We have applied this approach to systematic review and meta-analysis of the interaction between NAT1 and NAT2 polymorphisms and smoking in susceptibility to bladder cancer. Although only one published study incorporates all three factors (both genes plus smoking), we have shown that it is possible to devise valid ways of combining results from studies that measured only some of these factors, thus making optimum use of the partial evidence that is available.