The prognostic accuracy of three classification systems used for periprosthetic joint infection

Aims: The aim of this study was to evaluate and compare the prognostic value of three classification systems for periprosthetic joint infection (PJI) - McPherson, PJI-TNM, and JS-BACH - by analyzing their association with three key clinical outcomes: recurrence of infection after the initial treatment, final PJI status, and PJI-related mortality. Methods: This was a retrospective cohort study using a prospectively maintained institutional database. All patients who underwent surgery for PJI of the hip or knee between January 2011 and December 2022 were included, provided that they had a minimum of two years of follow-up, unless they had earlier recurrence of infection or had died. Each case was classified according to the three systems and grouped into three levels of severity (A: less severe; B: intermediate; C: most severe) to allow direct comparison. Associations between the variables of the classification systems and outcomes were examined using chi-squared and Fisher's exact tests. Discriminative ability was assessed using receiver operating characteristic (ROC) curve analysis and the area under the curve was calculated for each outcome. Results: Of the 186 patients who were included, recurrence after initial treatment occurred in 50 (26.9%) and PJI-related mortality in 19 patients (10.2%). Among the sub-categories of the classifications, host-related variables consistently correlated with most outcomes across all systems. Local extremity grade/soft-tissues also correlated with recurrence after initial treatment throughout. The JS-BACH classification was significantly associated with all three outcomes: recurrence (p = 0.002), final PJI status (p = 0.008), and mortality (p < 0.001). It also showed the best discriminative ability on ROC analysis. PJI-TNM was significantly associated with the final status (p = 0.004) and mortality (p = 0.002), but not with recurrence of infection. The McPherson classification was only significantly associated with the final PJI status (p = 0.039). Patients classified as 'limited options' in JS-BACH had a 28.6% mortality rate, while none in the 'uncomplicated' group died or required suppressive antibiotic treatment. Conclusion: The JS-BACH classification system showed the best prognostic performance, correlating with all the outcomes which were evaluated and offering a balanced combination of simplicity, clinical use, and predictive value. The host status remained a critical predictor regardless of which classification system was used.