Normality analysis of numeric rating scale scores in patients with chronic axial spine pain before and after medial branch blocks: a multicenter study

Abstract

Abstract Objective The statistical analysis typically used to compare pain before and after interventions assumes that scores are normally distributed. The present study evaluates whether numeric rating scale (NRS) scores, specifically NRS-11 scores, are indeed normally distributed in a clinically relevant cohort of adults with chronic axial spine pain before and after analgesic intervention. Methods Retrospective review from 4 academic medical centers of prospectively collected data from a uniform pain diary administered to consecutive patients after they had undergone medial branch blocks. The pain diary assessed NRS-11 scores immediately before injection and at 12 different time points after injection up to 48 hours. D’Agostino–Pearson tests were used to test normality at all time points. Results One hundred fifty pain diaries were reviewed, and despite normally distributed pre-injection NRS-11 scores (K2 = 0.655, P = .72), all post-injection NRS-11 data were not normally distributed (K2 = 9.70– 17.62, P = .0001–.008). Conclusions Although the results of parametric analyses of NRS-11 scores are commonly reported in pain research, some properties of the NRS-11 do not satisfy the assumptions required for these analyses. The data demonstrate non-normal distributions in post-intervention NRS-11 scores, thereby violating a key requisite for parametric analysis. We urge pain researchers to consider appropriate statistical analysis and reporting for non-normally distributed NRS-11 scores to ensure accurate interpretation and communication of these data. Practicing pain physicians should similarly recognize that parametric post-intervention pain score statistics might not accurately describe the data and should expect articles to utilize measures of normality to justify the selected statistical methods.