Aims
To evaluate how surgical strategy, and JS-BACH complexity are associated with treatment failure in hip and knee prosthetic joint infection (PJI).

Methods
A multi-national, retrospective cohort study of patients undergoing surgery for EBJIS-confirmed PJI was undertaken at 7 non-specialist hospitals and 3 specialist musculoskeletal infection hospitals. Cases were classified by joint involved, centre type, and JS-BACH complexity group (uncomplicated, complex, limited options). Index procedures were one-stage revision, debridement, antibiotics and implant retention (DAIR), open washout, or removal of prosthesis as the first stage of a two-stage strategy. Treatment failure was defined using a predefined composite outcome including recurrence at re-operation according to EBJIS criteria, long-term antibiotic suppression, death within 3 months of surgery, or amputation. In addition, failure to undergo reimplantation was also recorded. Multivariable logistic regression assessed the association between surgical procedure and treatment failure after adjustment for JS-BACH complexity and joint.

Results
756 cases were included, with 388 hip PJIs and 368 knee PJIs. Mean age was 70.5 years (SD 12.4), and mean follow-up 3.2 years (SD 1.5). Overall, 432 cases were managed at specialist centres and 324 at non-specialist hospitals. Treatment strategy differed significantly by centre type (p<0.001), with specialist hospitals favouring two-stage treatment (215/432, 49.8%) and non-specialist hospitals favouring DAIR (177/324, 54.6%). Across the whole cohort, there were 83 one-stage revisions, 351 DAIRs, 58 open washouts, and 264 removals of prosthesis, for 2-stage revision. 

Failure was most frequent after open washout (44/58, 75.9%), followed by removal of prosthesis (133/264, 50.4%), DAIR (148/351, 42.2%), and one-stage revision (24/83, 28.9%). Failure increased markedly with JS-BACH complexity. After adjustment, compared with uncomplicated PJI, complex PJI was associated with a higher failure rate (OR 6.28,95% CI 3.93-10.03; p<0.001) and limited-option PJI (OR 19.52, 10.72-35.56; p<0.001). By comparison, procedure had a smaller effect; compared with one-stage revision, DAIR (OR 1.88, 1.07-3.29; p=0.029) and removal of prosthesis (OR 2.29, 1.28-4.09; p=0.005) were associated with more modest increases in failure, whereas open washout had consistently poor outcomes (OR 11.08, 4.71-26.04; p<0.001).

Conclusions
JS-BACH complexity was a stronger predictor of treatment failure than operative strategy in PJI. Open washout consistently performed poorly but, PJI complexity was more associated with outcome than procedure choice. Overall, one-stage revision gave the best results in all groups.