Meta-analysis of cancer trials: a new approach to the assessment of treatment.

Meta-analysis of clinical trials offers the opportunity to pool results from a number of randomised studies so increasing the statistical ability to detect the value of treatment. More accurate estimates of the likely size of such effects can also be obtained. Subset analysis, which is seldom reliable in individual clinical trials, can be made more trustworthy. Meta-analysis of randomised trials of adjuvant therapy in early breast cancer illustrates the value of such analyses. These have helped not only routine clinical practice but also thrown light upon biological mechanisms and directed future research. Meta-analyses or overviews are playing an increasingly prominent role in cancer research. They offer the opportunity to maximise the use of information about a given treatment by combining the results from multiple randomised trials in a meaningful way. Whilst data from studies examining the same issue cannot simply be combined (due to trial heterogeneity), summation of the treatment effect across trials can be performed. It is reasonably assumed that differences between studies are differences in magnitude rather than differences in direction. The net result of such a process is to produce a result which is more accurate in its estimate of treatment benefit than its component parts.