Background: The accuracy of intraoperative periprosthetic frozen section histologic evaluation in predicting a diagnosis of periprosthetic joint infection prior to microbiologic culture results is unknown. Methods: We performed a systematic review and meta-analysis of all longitudinal studies that compared frozen section histologic results with simultaneously obtained microbiologic culture at the time of revision total hip or total knee arthroplasty. The data sources were Ovid MEDLINE, Ovid EMBASE, the Cochrane Library, ISI Web of Science, and SCOPUS, from the inception of each database to January 2010. Results: Twenty-six studies involving 3269 patients undergoing revision hip or knee arthroplasty met the inclusion criteria. A culture-positive periprosthetic joint infection was confirmed in 796 (24.3%) of the patients. Frozen section results, using any of the diagnostic criteria chosen by the investigating pathologist, had a pooled diagnostic odds ratio of 54.7 (95% confidence interval [CI], 31.2 to 95.7), a likelihood ratio of a positive test of 12.0 (95% CI, 8.4 to 17.2), and a likelihood ratio of a negative test of 0.23 (95% CI, 0.15 to 0.35) for the diagnosis of periprosthetic joint infection. Fifteen studies utilizing a threshold of five polymorphonuclear leukocytes (PMNs) per high-power field to define a positive frozen section had a diagnostic odds ratio of 52.6 (95% CI, 23.7 to 116.2), and six studies utilizing a diagnostic threshold of ten PMNs per high-power field had a diagnostic odds ratio of 69.8 (95% CI, 33.6 to 145.0). There was no significant difference between the diagnostic odds ratio or likelihood ratios associated with these thresholds. The moderate to high heterogeneity among the included studies was unexplained by variability in the study design, diagnostic criteria for acute inflammation, reference standard for periprosthetic joint infection, or prevalence of infection. This heterogeneity could be due to differences in the inclusion and exclusion criteria, tissue sampling error, experience or technique of the pathologists, number of microscopic fields visualized, and field diameter examined. Conclusions: Intraoperative frozen sections of periprosthetic tissues performed well in predicting a diagnosis of culturepositive periprosthetic joint infection but had moderate accuracy in ruling out this diagnosis. Frozen section histopathology should therefore be considered a valuable part of the diagnostic work-up for patients undergoing revision arthroplasty, especially when the potential for infection remains after a thorough preoperative evaluation. The optimum diagnostic threshold (number of PMNs per high-power field) required to distinguish periprosthetic joint infection from aseptic failure could not be discerned from the included studies. There was no significant difference between the diagnostic accuracy of frozen section histopathology utilizing the most common thresholds of five or ten PMNs per high-power field. Copyright © 2012 by the Journal of Bone and Joint Surgery, Incorporated.