Accuracy and heterogeneity of histological analysis for detecting periprosthetic joint infections

BACKGROUND: Intraoperative frozen section histology is a cornerstone for diagnosing periprosthetic joint infection (PJI), yet its reported accuracy varies significantly. Previous studies have primarily focused on the accuracy of histopathological diagnosis for PJI, with analysis of heterogeneous sources remaining insufficient. This study aimed to systematically evaluate the diagnostic accuracy of frozen sections for PJI and, specifically, to analyze where the discrepancies may exist. METHODS: We retrospectively reviewed 171 patients (89 PJI, 82 Aseptic) undergoing revision hip or knee arthroplasty (2015-2023). PJI was defined strictly by International Consensus Meeting (ICM) major criteria to avoid incorporation bias. Aseptic failures were confirmed by a lack of infection signs and > 2 years of survival without antibiotic usage. Diagnostic performance of frozen section in the diagnosis of PJI was evaluated. Histological heterogeneity was quantified by analyzing the concordance rate across three distinct tissue specimens per case. We further compared diagnostic accuracy using different interpretation strategies: requiring either one, two, or three positive samples to confirm the diagnosis. Additionally, inter-observer agreement was assessed by two independent pathologists re-evaluating 81 sections. RESULTS: The overall sensitivity was 71.9% (95% CI 61.4-80.9%) and specificity was 75.6% (95% CI 64.9-84.4%), with an area under the curve (AUC) of 0.738 (95% CI 0.672-0.803). Diagnostic performance was consistent between hip and knee revisions (p = 0.829). Notably, histological heterogeneity was prevalent: among confirmed PJI cases, only 44.0% (37/84) exhibited consistent positive results across all three specimens, whereas 73.7% of aseptic cases were consistently negative. Among the three specimens, if consider at least one specimen had >5 neutrophils per high power field in 5 high power fields (×400) as histology positive, the sensitivity and specificity was 70.2% (59.3-79.7%) and 73.7% (62.3-83.1%), respectively. Furthermore, inter-observer agreement was moderate (Kappa = 0.581), suggesting that diagnostic interpretation varies between pathologists. CONCLUSION: Intraoperative frozen section analysis provides reliable diagnostic accuracy for patients meeting ICM Major Criteria, with consistent performance across hip and knee revisions. However, the significant heterogeneity existed in both surgical sampling and pathological interpretation. Standardization from the surgeon's intraoperative sampling to the pathologist's result interpretation stage may offer a potential solution to this problem.