The effect of study design and analysis methods on recovery rates in Bell's palsy

We investigated how study design affects the rate of recovery in Bell's palsy.

Prospective, randomized, double-blind, placebo-controlled, multicenter trial.

Data were extracted from the Scandinavian Bell's palsy study, which included 829 patients. The study design was factorial; 416 patients given prednisolone, 413 not given prednisolone, 413 patients given valacyclovir, 416 not given valacyclovir. Data were analyzed with intention-to-treat principle and complete-case analysis methods and recovery was defined as Sunnybrook score 100, House-Brackmann grade I or < or =grade II at 12 months.

With the intention-to-treat principle and last-observation-carried-forward method (n = 829) and recovery defined as Sunnybrook 100, 300 of the 416 patients (72%) receiving prednisolone had recovered compared with 237 of the 413 (57%) who did not receive prednisolone (P < .0001). With recovery defined as House-Brackmann grade I, the corresponding recovery rates were 324 of 416 (78%) and 266 of 413 (64%) (P < .0001). With complete-case analysis and recovery defined House-Brackmann grade I (n = 782), 335 of 389 patients (86%) given prednisolone recovered compared with 277 of 393 (70%) in the group not given prednisolone (P < .0001). With recovery defined as House-Brackmann < or =grade II (n = 797), the corresponding recovery rates were 380 of 396 (96%) and 353 of 401 (88%) (P < .0001). The analysis method affected the recovery rates in the valacyclovir and no-valacyclovir groups in a similar way as in the prednisolone and no-prednisolone groups.

Recovery rates in a Bell's palsy study are substantially affected by the choice of analysis method and definition of recovery.