Do overactive bladder symptoms and their treatment-associated changes exhibit a normal distribution? Implications for analysis and reporting

Abstract

AIMS To explore the use of means vs medians (assuming or not the presence of normal distribution) in studies reporting overactive bladder syndrome symptoms and to test for normal distribution of basal values and treatment-associated changes thereof in two large noninterventional studies.

METHODS Systematic review of all original studies reporting on at least one overactive bladder syndrome symptom published in four leading urology journals in 2016 to 2017. Testing of the normal distribution of urgency, incontinence, frequency, and nocturia in two large noninterventional studies (n = 1335 and 745).

RESULTS Among 48 eligible articles, 86% reported means (assuming a normal distribution), 6% medians (not making this assumption), and 8% a combination thereof. Baseline values for all four symptoms and treatment-associated alterations thereof deviated from a normal distribution (P < .0001 in all cases). Means overestimated basal value and absolute changes thereof as compared with medians, for example, basal number of incontinence episodes in study 1 5.1 vs 4. Differences between means and medians for percentage changes of symptoms were small and did not consistently favor means over medians.

CONCLUSIONS Dominant reporting of means implies the assumption of a normal distribution of overactive bladder syndrome symptoms but our data from two noninterventional studies do not support this assumption. We recommend that basal values and absolute symptom changes should be reported as medians and subjected to nonparametric analysis; means may be appropriate for the reporting of percentage changes of symptoms.