Abstract
We recently demonstrated improved sensitivity of prosthetic joint infection (PJI) diagnosis using an automated blood culture bottle system for periprosthetic tissue culture [T. N. Peel et al., mBio 7(1):e01776-15, 2016, https://doi.org/10.1128/mBio.01776-15]. This study builds on the prior research by examining the optimal number of periprosthetic tissue specimens required for accurate PJI diagnosis. Current guidelines recommend five to six, which is impractical. We applied Bayesian latent class modeling techniques for estimating diagnostic test properties of conventional culture techniques (aerobic and anaerobic agars and thioglycolate broth) compared to inoculation into blood culture bottles. Conventional, frequentist receiver operating characteristic curve analysis was conducted as a sensitivity analysis. The study was conducted at Mayo Clinic, Rochester, MN, from August 2013 through April 2014 and included 499 consecutive patients undergoing revision arthroplasty from whom 1,437 periprosthetic tissue samples were collected and processed. For conventional periprosthetic tissue culture techniques, the greatest accuracy was observed when four specimens were obtained (91%; 95% credible interval, 77 to 100%), whereas when using inoculation of periprosthetic tissues into blood culture bottles, the greatest accuracy of diagnosis was observed when three specimens were cultured (92%; 95% credible intervals, 79 to 100%). Results of this study show that the greatest accuracy of PJI diagnosis is obtained when three periprosthetic tissue specimens are obtained and inoculated into blood culture bottles or four periprosthetic tissue specimens are obtained and cultured using standard plate and broth cultures. Increasing the number of specimens to five or more, per current recommendations, does not improve accuracy of PJI diagnosis.