Individual patient data meta-analysis in cancer

Abstract

Summary As in many areas of health care, treatments for cancer may differ only moderately in their effects on major end points, such as death. But, such differences are worth knowing about, particularly in common diseases in which they could represent a substantial benefit to public health. Large-scale randomized evidence allows moderate differences to be investigated reliably, and one way to achieve this is by metaanalyses of updated and centrally collected individual patient data from all relevant trials. This paper illustrates why this form of research can often be important in cancer. It also offers the first list of such projects, as a source of information on current and past research in this area.Keywords: individual patient data; meta-analyses; collaborative overview; randomized controlled trialThe conduct of meta-analyses of updated and centrally collected individual patient data (IPD) from all relevant randomized controlled trials has been discussed previously (Stewart and Clarke, 1995). This paper illustrates why this form of research is often particularly important in cancer. It also publishes the first list of IPD meta-analyses in cancer (see Appendix), thereby providing an important source of information on what research has been, or is being, conducted in this area. It may also help avoid duplication of effort as IPD meta-analyses generally involve considerable work, particularly for those who organize them. THE NEED FOR SYSTEMATIC REVIEWS TO ASSESS TREATMENTS FOR CANCERAs in many areas of health care, a fundamental principle underlying the need for large-scale randomized evidence in cancer is that, for major end points, the difference between the treatments under investigation is unlikely to be large. But, if a widely practicable treatment produced a moderate improvement for a common disease, this could represent a substantial benefit to public health. Similarly, clear and reliable evidence that there is no such difference could avoid much unnecessary treatment, along with its associated toxicity and cost (Collins et al, 1996).Large-scale randomized evidence can be obtained by suitably large randomized controlled trials (RCTs) that will accrue future patients, systematic reviews of past trials or ideally a combination of the two. At present, most trials in cancer are of limited size and so this disease is particularly well suited to systematic review. In addition, some treatments have been investigated in numerous RCTs over many years. For example, more than 40 000 women world-wide have, since 1974, been randomized into at least 50 separate trials of tamoxifen vs no tamoxifen for operable breast cancer. Data from approximately 30 000 women in 40 of the trials (Stewart and Clarke, 1995). The finalized data for each trial are then analysed separately to obtain summary statistics that are then combined to give an overall estimate of the effect of treatment. In this way, patients are compared directly only with others in the same trial. This paper concentrates on IPD meta-analyses in cancer and a ...

Outcome of axillary staging in early breast cancer: a meta-analysis

“…This study has some limitations; it is based on available published results and the use of updated individual patient data may further enhance the accuracy and reduce the uncertainty of the estimates [25,26]. Another potential limitation is that the results of recently launched randomized studies were not available to include in the meta-analysis, and data from some published studies could not be extracted.…”

Synthesizing medical evidence: systematic reviews and metaanalyses

“…55 An example of neurosurgical interest is the metaanalysis of chemotherapy treatment for malignant glioma reported by the Glioma Meta-analysis Trialists Group in 2002. 327 The metaanalysis of individual patient data is much more demanding of resources than one based solely on published data, and it requires cooperation from the principal investigators of the primary trials.…”

“…327 The metaanalysis of individual patient data is much more demanding of resources than one based solely on published data, and it requires cooperation from the principal investigators of the primary trials. 137,328 Nevertheless, the benefits can be great; 55,329 many published trial reports lack information critical for metaanalysis due to poor reporting, so that they cannot even be included in a literature-based metaanalysis. Metaanalyses of individual patient data can often use additional follow-up data that have accrued since trials were published as well as information on patients excluded from the published analyses for inappropriate reasons.…”

“…Metaanalyses of individual patient data can be used to compare survival experiences throughout the follow-up period, rather than using a single time point that may be "optimally" chosen (post hoc) to illustrate differences between treatments that could be spurious. 55,329 Finally, subgroup analyses allow testing for consistency of treatment effect in varied patient populations, so that conclusions can be confidently applied to new patients. 55 Some of these advantages apply to binary outcome variable trial metaanalysis as well, and metaanalyses of individual patient data have been performed in this setting.…”

“…55,329 Finally, subgroup analyses allow testing for consistency of treatment effect in varied patient populations, so that conclusions can be confidently applied to new patients. 55 Some of these advantages apply to binary outcome variable trial metaanalysis as well, and metaanalyses of individual patient data have been performed in this setting. Despite their advantages, however, the time and expense required for analyses of individual patient data has restricted their use, and most comparisons of individual patient data and literature-based analyses on the same topic have shown summary effect estimates that are in general agreement.…”