C-reactive protein to albumin ratio (CAR) in predicting surgical site infection (SSI) following instrumented posterior lumbar interbody fusion (PLIF).

Abstract

Identification of novel markers would contribute to the individualised risk assessment and development of a risk prediction model. This study aimed to investigate the role of the C-reactive protein to albumin ratio (CAR) in predicting surgical site infection (SSI) following instrumented posterior lumbar interbody fusion (PLIF) of lumbar degenerative diseases. This study enrolled patients who underwent PLIF and instrumentation for treatment of lumbar degenerative diseases between 2015 and 2020. Electronic medical records were inquired for data collection, with follow-up register for identifying SSI cases. The optimal cut-off for CAR was determined by constructing the receiver operator characteristic (ROC) curve. Patients with high- or low-CAR value were compared using the univariate analyses, and the association between CAR and the risk of SSI was investigated using multivariate logistics regression analysis. A total of 905 patients were enrolled, twenty-nine (3.2%) had developed an SSI with 72.4% occurring during index hospitalisation, and 11 (1.2%) had deep and 18 (2.0%) superficial SSIs. An SSI was associated with additional 10.7 days of index total hospital stay (P=.001). The CAR was 0-5.43 (median, 0.05), and the optimal cut-off was 0.09 and area under the curve was 0.720 (P < .001). 336 (37.1%) patients had a CAR ≥0.09 and 22 (6.5%) developed an SSI, with a crude risk of 5.6 relative to those with a low CAR. The multivariate analyses showed CAR ≥0.09 was associated with 8.06-fold increased risk of SSI, together with diabetes (P=.018), while hypertension was identified as a protective factor (OR, 0.34; 95%CI, 0.11-1.00, P= .049). High CAR is found to significantly predict the incident SSI following instrumented PLIF of lumbar degenerative diseases, and can be considered as a useful index in practice only after it is verified by future high-level evidences.