Risk Factors and Predictive Models for Early Cerebrospinal Fluid Leak after Anterior Surgery for Degenerative Cervical Myelopathy

Abstract OBJECTIVE To investigate the incidence of and risk factors for early cerebrospinal fluid (CSF) leak after anterior surgery for degenerative cervical myelopathy (DCM) and construct risk predictive models.METHODS Patients diagnosed with DCM and treated with anterior surgery from January 2007 to August 2018 were retrospectively selected. The basic data of preoperative and intraoperative records were extracted. Subsequently, we analyzed the relationship between early postoperative CSF leak and variables by performing restrictive cubic splines (RCS) for continuous variables and univariate and multivariate logistic regression for possible variables. The selected independent variables and covariates were modeled based on the Akaike information criterion (AIC). We generated receiver operating characteristic (ROC) curves and calibration curves to evaluate the nomogram. RESULTS Seventeen patients had CSF leak, with an incidence of 3.66%. There was a linear relationship between CSF leak and age (P=0.003) as well as bleeding loss (P<0.001). Variables were selected by logistic regression analysis, including age (per SD) (OR=2.04, P=0.004), revision surgery (OR=1.13, P=0.032), ossification of the posterior longitudinal ligament (OPLL) (OR=4.07, P=0.027), and blood loss (per SD) (OR=1.75, P=0.002). The predictive model included age (per SD), sex, revision surgery, OPLL, and blood loss (per SD), with c-statistic=0.821. CONCLUSIONS Older age, revision surgery, OPLL and more intraoperative blood loss were independent risk factors for early postoperative CSF leak. The predictive model can be used to guide the prevention of CSF leak in DCM patients.