Exploring predictors of analgesic response in adult women regularly using NSAIDs: insights from a machine learning approach.

INTRODUCTION: Non-steroidal anti-inflammatory drugs (NSAIDs) are considered a first-line treatment for dysmenorrhea (menstrual pain). However, as much as 20% of women fail to achieve sufficient pain relief from NSAIDs, although factors associated with NSAID non-response are not well-known. The purpose of this study was to investigate factors associated with analgesic response to NSAIDs for dysmenorrhea. METHODS: Data were from a large survey study of adult women who reported regularly using NSAIDs for dysmenorrhea. Demographic information, menstrual characteristics, comorbid conditions, and psychosocial factors were collected. NSAID non-response was defined as self-reported minimal or no menstrual pain relief from NSAIDs. We repeatedly applied a group least absolute shrinkage and selection operator (LASSO) logistic regression 100 times, and averaged results to account for model uncertainty and to identify variables related to NSAID non-response. RESULTS: Of 512 women (mean age = 34.0 years, SD = 9.4) who met inclusion criteria, 182 (35.5%) were classified as NSAID non-responders. NSAID non-responders reported significantly higher average menstrual pain severity than responders (mean = 7.4 vs. 6.2 on a 0-10 scale, p < 0.001). The strongest factors associated with NSAID non-response included greater menstrual pain severity and more severe menstrual-related symptoms (e.g., bloating and dull pelvic pain), as well as greater sleep disturbance. Higher negative affect was modestly associated with better odds of NSAID response. Classical risk factors such as heavy menstrual bleeding and self-reported endometriosis were not significantly linked to NSAID response in this sample. However, the LASSO models achieved only modest discrimination (mean cross-validated AUC = 0.64), indicating considerable unexplained variability. DISCUSSION: In this large cross-sectional cohort, women reporting inadequate menstrual pain relief from NSAIDs had more severe menstrual pain and related symptoms, along with indicators of pain amplification (sleep disturbance). Most psychosocial factors (e.g., negative mood) did not correlate with NSAID non-response. These findings underscore the heterogeneity of dysmenorrhea and suggest that clinical factors alone may be insufficient to predict NSAID effectiveness. Further prospective studies are needed to confirm these associations and to explore whether interventions (such as optimizing NSAID timing or addressing sleep disturbances) could improve menstrual pain outcomes for those with NSAID non-response.