Abstract
INTRODUCTION: The conventional clinical criteria for diagnosing a periprosthetic joint infection (PJI) rely on acute inflammatory readouts such as erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), and synovial white blood cell counts. These metrics only detect actively septic joint replacements and may miss detecting biofilm-embedded bacteria that suppress neutrophil signaling and persist as a “hidden” subset of implants with a dormant infection. We hypothesize that previously infected joint replacements have a high prevalence of dormant infection that can be distinguished from aseptic revision joint replacements (replaced for instability, loosening, wear, and fracture) by the persistent inflammatory response within synovial fluid and/or circulating plasma, and that detecting a dormant infection indicates an increased risk of infection relapse. METHODS: This is an observational cohort study using synovial fluid and plasma proteomics of 96 immuno-oncology mediators (Olink Proteomics, Sweden) with three-year clinical follow-up from a single academic medical center (Stanford University, USA). Thirty patients undergoing revision joint replacement: culture-positive actively septic joint replacements (n = 7), aseptic revision joint replacements (n = 12), and re-implantations of joint replacements previously classified as infection-free by 2018 Musculoskeletal Infection Society (MSIS) criteria (n = 11). Differential expression, unsupervised clustering, Euclidean distance mapping, principal-component analysis, and gene-set variation analysis were used to define the inflammatory signature of dormant infection present in joint replacements with a prior infection. The identified biomarkers of dormant infection were correlated with the three-year incidence of infection relapse. RESULTS: Eight of eleven MSIS-cleared joint replacements (73%) clustered with culture-positive active infections despite normal ESR, CRP, and scant synovial neutrophils revealing the synovial inflammatory signature of dormant infection. A nine-analyte synovial panel consisting of PDGF-B, CXCL5, CXCL11, MCP-2 (CCL8), ANGPT1, TIE2, EGF, NOS3, and Gal-1 distinguished dormant infection from truly aseptic cases with 100% specificity and positive predictive value (sensitivity 22%, negative predictive value 74%). Synovial CXCL5 over-expression was a universal hallmark of both active and dormant infection, whereas matched plasma profiles showed no discriminatory power for all immuno-oncology mediators tested. Dormant infections exhibited downregulation of granulocyte activation and T-cell proliferation pathways (FDR < 0.001), mirroring immune evasion programs seen in cancer microenvironments. After a mean of 3 years follow-up, infection relapse occurred in 22% of the biomarker-positive dormant infections, but relapse did not occur in any of the biomarker-negative aseptic cases. DISCUSSION: Profiling of the persistent inflammatory response within the synovial fluid of two-stage re-implantations classified as “infection-free” by the MSIS criteria unmasked a clinically silent reservoir of biofilm-embedded bacteria that suppress clinical diagnostic criteria of active infection and define the novel clinical state of dormant infection. We used that profile to identify a novel culture-free panel of rule-out biomarkers for determining which re-implantations were safe from infection relapse. These findings challenge the use of conventional clinical diagnostic criteria, which are over-reliant on acute phase reactants and neutrophil recruitment, and creates a new prognostic clinical paradigm that now includes a future with precision, immune-guided management of dormant infections likely present in many implant-associated infections. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12967-025-07425-y.