Objectives: To determine whether unpublished dissertations are an important source of trials for systematic reviews.

Methods: In a review of infant massage, we identified 17 dissertations. We recorded whether each dissertation was included in the review and, if so, whether it contributed data to any analyses.

Results: Thirteen dissertations were excluded, 11 because of poor reporting of methodology. Three dissertations were also published as journal papers. The data from the remaining dissertation were not included in any analysis. Reviewing data on the Cochrane Library, only one of 878 reviews included data from a dissertation that might have changed a review's conclusions.

Conclusions: Searching for and retrieving unpublished dissertations involves considerable time and effort and appears to influence the conclusions of a review only rarely.