Preoperative prediction of spinal cord ischemia after thoracic endovascular aortic repair

Objective: Spinal cord ischemia (SCI) is a devastating but potentially preventable complication of thoracic endovascular aortic repair (TEVAR). The purpose of this analysis was to determine what factors predict SCI after TEVAR. Methods: All TEVAR procedures at a single institution were reviewed for patient characteristics, prior aortic repair history, aortic centerline of flow analysis, and procedural characteristics. SCI was defined as any lower extremity neurologic deficit that was not attributable to an intracranial process or peripheral neuropathy. Forty-three patient and procedural variables were evaluated individually for association with SCI. Those with the strongest relationships to SCI (P <.1) were included in a multivariable logistic regression model, and a stepwise variable elimination algorithm was bootstrapped to derive a best subset of predictors from this model. Results: From 2002 to 2013, 741 patients underwent TEVAR for various indications, and 68 (9.2%) developed SCI (permanent: n = 38; 5.1%). Because of the lack of adequate imaging for centerline analysis, 586 patients (any SCI, n = 43; 7.4%) were subsequently analyzed. Patients experiencing SCI after TEVAR were older (SCI, 72 +/- 11 years; no SCI, 65 +/- 15 years; P <.0001) and had significantly higher rates of multiple cardiovascular risk factors. The stepwise selection procedure identified five variables as the most important predictors of SCI: age (odds ratio [OR] multiplies by 1.3 per 10 years; 95% confidence interval [CI], 0.9-1.8, P = . 06), aortic coverage length [OR] multiplies by 1.3 per 5 cm; CI, 1.11.6; P = .002), chronic obstructive pulmonary disease (OR, 1.9; CI, 0.9-4.1; P = .1), chronic renal insufficiency (creatinine concentration >= 1.6 mg/dL; OR, 1.9; CI, 0.8-4.2; P = .1), and hypertension (defined as chart history or medication; OR, 6.4; CI, 2.6-18; P <.0001). A logistic regression model with just these five covariates had excellent discrimination (area under the receiver operating characteristic curve = .83) and calibration (x(2) = 9.8; P = .28). Conclusions: This analysis generated a simple model that reliably predicts SCI after TEVAR. This clinical tool can assist decision-making about when to proceed with TEVAR, guide discussions about intervention risk, and help determine when maneuvers to mitigate SCI risk should be implemented.