Abstract
Aim
In the diagnosis of prosthetic joint infection (PJI), many biomarkers have shown a sound performance in terms of accuracy, sensitivity and specificity. In this study we aimed to test the frequently used serum biomarkers C-reactive Protein (CRP), Fibrinogen, Leukocytes, Interleukin-6 (IL-6), Interferon alpha (IF-alpha) and Procalcitonin (PCT) regarding these qualities. Following that, the optimal multi-biomarker combination was calculated to further improve the diagnostic performance.
Method
124 knee or hip revision arthroplasty procedures were prospectively investigated focusing on preoperative serum blood levels of CRP, Fibrinogen, Leukocytes, IL-6, IF-alpha and PCT. The presence of PJI was determined by a blinded researcher. Logistic regression with lasso-regularization was used for the biomarkers and all their ratios. Following cross-validation on a training sample set to get optimal performance estimates, we performed the final model on a test set (25% of all samples).
Results
Out of all evaluated biomarkers, CRP (AUC 0.91, p-value 0.03) and Fibrinogen (AUC 0.93, p-value 0.02) had the best performances. The optimal combination when testing multiple biomarkers in 32 cross-validation runs was calculated including Fibrinogen, CRP, the ratio of Fibrinogen to CRP and the ratio of serum Thrombocytes to CRP (AUC 0.92, accuracy 0.77, specificity 0.92, sensitivity 0.68, cut-off 0.63, p-value 0.04).
Conclusions
It was not possible to increase the diagnostic performance by combining multiple biomarkers using sophisticated statistical methods. The calculated Multi-biomarker models did not improve the AUC, accuracy, sensitivity and specificity when compared to single biomarkers.
