Prediction of nonrecovery in Bell's palsy using Sunnybrook grading.

Objectives/hypothesis: To develop a clinical prognostic model to identify Bell's palsy patients with risk for nonrecovery at 12 months.

Study Design: Data from a prospective, randomized, double-blind, placebo-controlled, multicenter study.

Methods: There were 829 patients with Bell's palsy randomized in a factorial fashion to treatment with prednisolone or no prednisolone. Facial function was assessed with the Sunnybrook grading scale. Univariate and multivariate logistic regression analyses at different time points were used to identify factors predicting nonrecovery, defined as Sunnybrook <70 at 12 months. Variables studied were age, gender, time to inclusion, prednisolone treatment, side of palsy, pain at inclusion, and Sunnybrook scores. Factors of predictable significance were used to construct prognostic models at baseline, days 11 to 17, and at 1 month. Receiver operating characteristics curves were created to test the predictive capacity of the models.

Results: At baseline, treatment with prednisolone or no prednisolone (P = .0005), age (P = .04) and the Sunnybrook score (P = .0002) were significant factors for predicting nonrecovery. The receiver operating characteristics area under the curve at baseline for these three variables was 0.74 (sensitivity 0.83, specificity 0.57). At days 11 to 17 and at 1 month, the Sunnybrook score was the only significant predictive variable. The respective areas under the curves for the Sunnybrook score at these time points were 0.83 (sensitivity 0.81, specificity 0.75) and 0.94 (sensitivity 0.91, specificity 0.85).

Conclusions: Sunnybrook grading at 1 month most accurately predicts nonrecovery at 12 months in Bell's palsy.