Abstract. The diagnosis of prosthetic joint infection (PJI)following total joint arthroplasty is difficult for clinicians to make decisions due to the similar symptoms presented by aseptic loosening and infection. Gram staining (GS) is a widely used test but its value remains controversial due to conflicting results in the diagnosis of PJI. The aim of the present study was therefore to evaluate the value of GS in the diagnosis of PJI. Searches using MEDLINE, EMBASE and OVID databases were conducted for data published between January 1990 and December 2013. Meta-analysis was used to pool the sensitivity, specificity, diagnostic odd ratios (DORs), area under the receiveroperating characteristic curve (AUC), positive-likelihood ratios (PLRs), negative-likelihood ratios (NLRs) and post-test probability. The heterogeneity and publication bias were assessed, and subgroup and meta-regression analyses were conducted. A total of 18 studies, including a total of 4,647 patients, were selected for analysis. The pooled sensitivity and specificity values for the diagnosis of PJI were 0.19 and 1.00, respectively. The AUC, PLR and NLR were 0.89, 41.6 and 0.82, respectively. Subgroup analyses indicated that the sensitivity/specificity for total hip arthroplasty was 0.14/0.99, whereas that for total knee arthroplasty was 0.14/1.00. Synovial fluid best reflected accurate GS-based diagnoses, with the highest DOR of 242, whereas tissue had the highest AUC of 0.96 (95% CI, 0.94-0.97). GS had a poor clinically acceptable diagnostic value for detecting PJI. These data do not support the routine use of GS, without additional proof of infection, for diagnosing PJI; instead, GS could be used as an adjuvant tool to support the results of other investigations.