Which Definition Is the Gold Standard? Reference-Standard Choice and the Reported Accuracy of Periprosthetic Joint Infection Diagnostic Tests: A Meta-Epidemiological Systematic Review

This is the authors' abstract. We don't add key points for this paper.

Background/Objectives: Periprosthetic joint infection (PJI) has no gold standard. Accuracy studies borrow one of several composite definitions instead, and these definitions do not label the same patients as infected. We asked how often accuracy studies say which definition they used, and how far the choice moves the numbers they report. Methods: In this systematic review, we searched PubMed on 20 October 2026 for primary accuracy studies of PJI tests and classified each abstract by the reference definition it named. Studies that applied two or more definitions (MSIS 2011/2013, ICM 2018, IDSA 2013, EBJIS proposed or 2021) to the same cohort were assessed further, and accuracy was extracted per definition. Ratios of PJI cases and the change in sensitivity of the alpha-defensin test were pooled with random-effects models. Results: Of 1744 accuracy studies, 437 (25%) named a reference definition in the abstract. The share rose from none before 2012 to 35% from 2021 onward. Seventeen studies reported accuracy under more than one definition in the same patients. EBJIS classified 1.47 times as many patients as infected as MSIS (95% CI 1.33–1.62; 6 cohorts; I2 = 4%) and 1.26 times as many as ICM 2018 (1.08–1.46; 6 cohorts; I2 = 58%). The pooled sensitivity of the alpha-defensin test was 85.6% (76.9–91.4) against MSIS and 56.3% (48.6–63.7) against EBJIS in the same three cohorts, while specificity stayed between 89% and 100%. Across 13 test–cohort pairs, the median change in sensitivity when the reference moved to EBJIS was −18.5 percentage points (IQR −29.0 to −4.9); sensitivity fell in 10 pairs and changed by less than 3 points in three. Conclusions: For several PJI tests, the reported sensitivity depends substantially on the reference definition chosen. A definition that labels more patients as infected is not, for that reason, a more accurate one. Accuracy figures obtained under different definitions should not be pooled or compared as if they measured the same thing, and every accuracy study should name its reference definition in the abstract.