Aims: To provide a clinical view and interpretation on the methods for analysis of incontinence in patients with overactive bladder.
Methods: Results are analyzed using the total number of incontinence episodes in a 3-day diary period, using fixed and random effect Poisson regression models to calculate ratio of event rates and 95% confidence interval (CI) together with P-values and are compared with the analysis of the mean number of incontinence episodes/24 hr using analysis of covariance models to calculate P-values and 95% CI for the difference between treatments.
Results: Using random effects Poisson regression models demonstrated that the number of incontinence episodes was reduced by 26% more with mirabegron 50 mg than with placebo. For solifenacin 5 and 10 mg, treatment resulted in a 43% (41%) greater decrease in the number of incontinence episodes compared with placebo.
Conclusion: Instead of providing a fixed number of incontinence episodes/24 hr that reflects the mean effect, the estimate using Poisson methodology provides an efficacy estimate that can be interpreted in the context of, and relative to, the patient's baseline (severity). Using the total number of incontinence episodes in the diary period, and expressing this as percent decrease in the number of episodes, may be easier to interpret; for example, because this results in a relative measure of effect that provides an alternative understanding of a patient's improvement at end of treatment compared with the comparator arm. Also, it is based on statistical methods that are more suitable for the analysis of count data. Neurourol. Urodynam. 35:728-732, 2016. © 2015 Wiley Periodicals, Inc.
Keywords: ANCOVA; Poisson regression; rate ratio; urinary incontinence.
© 2015 Wiley Periodicals, Inc.