Development and Validation of an Evidence Based Algorithm for Diagnosing Periprosthetic Joint Infection.

Background: The guidelines for diagnosis of periprosthetic joint infection (PJI) introduced by 5 the AAOS served the orthopedic community extremely well over the last few years. The 6 introduction of new diagnostic modalities, including biomarkers, and the need for validation of 7 the guidelines prompted us to expand on the prior guidelines and develop an evidence-based, 8 validated diagnostic algorithm. 9

Methods: This multi-institutional study examined revision total joint arthroplasty patients from 10 three academic institutions. For development of the algorithm, an infected and aseptic cohort was 11 defined. PJI cases were defined using only the major criteria from the Musculoskeletal Infection 12 Society (MSIS) definition (n=684). Aseptic cases underwent one-stage revision for a non-13 infective indication and did not fail within 2-years (n=820). Risk factors, clinical findings, serum 14 and synovial markers as well as intraoperative findings were assessed. A stepwise approach 15 using random forest analysis and multivariate regression was used to generate relative weights 16 for each of the various variables assessed in each stage and create an algorithm for diagnosing 17 PJI using the 3 most important tests from each step. The algorithm was formally validated on a 18 separate cohort of patients with PJI (n=222) who failed with re-infection and aseptic patients 19 (n=200). 20

Results: The first step in evaluating PJI should include serum testing for C-reactive protein, D-21 dimer and erythrocyte sedimentation rate in that order of importance. If at least one of these are 22 elevated, or if there is a high clinical suspicion, clinicians should proceed with synovial fluid 23 testing using synovial white blood-cell, leukocyte esterase and polymorphonuclear %. Alpha 24 defensin did not show added benefit as a routine diagnostic test. Intraoperative findings including 25 purulence, histology and next generation sequencing (NGS) or single positive culture can aid in 26 inconclusive cases or when the aspiration does not yield fluid for analysis (dry tap). The 27 proposed algorithm demonstrated a high overall sensitivity (96.9%, 95% [Confidence Interval] 28 93.8-98.8) and specificity (99.5%, 95% [CI] 97.2-100) 29

Conclusions: This multi-institutional study offers an evidence-based algorithm for diagnosing 30 PJI which has shown an excellent performance on formal external validation.