Objective: There is increasing empirical evidence for theexistence of bias in the publication of primary clinical research,with statistically significant results being published more readily,more quickly, and in higher impact journals. Meta-analysis ofindividual patient data (IPD) may represent a gold standard of“secondary” clinical research, giving the best possible summaryof current evidence for a particular question, but publication ofthese may also be subject to bias. This study aimed to explore whichfactors might be associated with publication of IPD meta-analysesand to identify potential sources of bias.

Methods: For all known IPD meta-analysis projects in cancer, theresponsible investigator was surveyed by means of a questionnaire todetermine descriptive characteristics of the meta-analysis, thenature of the results, and details of the publication history.

Results: There is no good evidence that overall publication status ofmeta-analyses in cancer is dependent on the statistical or clinicalsignificance of the results. However, those meta-analyses withnonsignificant results did seem to take longer to publish and werepublished in lower impact journals compared with those with morestriking results.

Conclusions: Based on the current data, there seems to be no strongassociation between the results of IPD meta-analyses in cancer andpublication.