Psychosocial Predictors of Dysmenorrhea Stability and Change: A Two-Year Longitudinal Study

Data were analyzed from an in-person screening baseline visit and two follow-up annual questionnaires (Year 1 and Year 2) between August 2014 and December 2018. The data were part of a larger study investigating the relationship between bladder and uterine sensitivity. 13 The larger study included N = 354 females ages 18-45 recruited by flyers, online advertisements (Craigslist), and referral from local gynecology clinics. These participants were mostly white (66 %), single (76 %), and nulliparous (87%), and had a range of symptoms with an oversampling for chronic-pain free females with moderate-to-severe menstrual pain (hereafter referred to as dysmenorrhea). Participants were ineligible if they had current pelvic or abdominal malignancies, genitourinary infection (within past four weeks), amenorrhea (absence of regular menses), or declined to withdraw from oral contraceptives for two months before the assessment visit. Participants included women with dysmenorrhea (menstrual pain ≥5 on a 0-10 numerical rating scale; 0 = no pain; 10 = worst pain imaginable) or low menstrual pain (<3). We also separately enrolled as positive controls a) some women with bladder pain syndrome/interstitial cystitis according to American Urology Association criteria 14 b) other chronic pain diagnoses that included back pain and idiopathic chronic pelvic pain (≥5 on a 0-10 numerical rating scale > three consecutive months). A complete list of study inclusion and exclusion criteria has been previously published. 13 Only participants who completed at least two of three time points approximately 12 months apart (e.g., baseline screening visit, Year 1, and Year 2) were included in the present study, as a minimum of two time points is necessary for conducting linear trajectory analyses. Thus, 197 participants with no follow-up data were excluded from the analyses. The analytic sample ( n = 157) was similar demographically to the larger study sample ( Table 1 ). This study was approved by the NorthShore University HealthSystem (now Endeavor Health) Institutional Review Board. During the screening visit, participants reported their age, race, ethnicity, marital status, oral contraceptive use, and number of deliveries using a standardized questionnaire. Participants rated their menstrual pain at the baseline in-person visit and at two annual follow-ups using a 0–100 visual analog scale (VAS; 0 = no pain, 100 = worst pain imaginable). 12 In this study, participants were not asked to forgo analgesics. The prompt asked: “Please indicate the average amount of cramping or pain you have experienced during your menstrual period over the past 3 months when not taking painkillers, on the worst day of your period.” Validation analyses demonstrated stronger agreement between diary-based maximal pain ratings and retrospective reports of pain without NSAID use (r = 0.70) than with NSAID use (r = 0.48), consistent with diary data suggesting most participants undermedicated their pain. Therefore, retrospective ratings of menstrual pain in the absence of analgesics were used in trajectory modeling to provide a more reliable measure of underlying pain intensity. Participants rated the average intensity of pain experienced (1) with urination, (2) with bowel movements, and (3) generally in the pelvis during the past week using a 0-100 visual analog scale. Although participants were allowed to report either non-menstrual or menstrual pelvic pain, we retroactively confirmed that none were menstruating during this time, based on their self-reported cycle day. These three aspects of pelvic pain were averaged to create a composite measure of NMPP. Validation of the NMPP construct has been previously published. 15 The NIH Patient Reported Outcomes Measurement Information System (PROMIS) 7-item Anxiety and 8-item Depression short-forms were administered at the screening visit. 16 Respondents are asked to indicate on a 5-point Likert scale from 1 ( never ) through 5 ( always ) the extent to which they experienced symptoms during the past seven days. Internal consistency for both anxiety (α = .93) and depression (α = .95) scales was excellent. The 13-item Pain Catastrophizing Scale (PCS) was administered at the screening visit. 17 Respondents are asked to indicate on a 5-point Likert scale from 0 ( not at all ) through 4 ( all the time ) the extent to which they experienced specific thoughts or feelings when experiencing pain. In the present sample, internal consistency was excellent (α = .93). The 7-item somatization subscale of the Brief Symptom Inventory-18 was used to capture somatic sensitivity levels. 18 Respondents are asked to indicate on a 5-point Likert scale from 0 ( not at all ) through 4 ( very much ) to what extent they are troubled by the complaints. The subscale demonstrated good internal consistency (α = .80). The 8-item PROMIS Short Form for sleep disturbance was administered to participants. 16 Respondents are asked to indicate on a 5-point Likert scale from 1 ( not at all/never ) through 5 ( very much/always ), the extent to which they experienced restoration associated with sleep, sleep depth, and sleep quality. Internal consistency of this scale was excellent (α = .91). Data were analyzed using R version 4.1.0. Descriptive analyses were run for study variables, including bivariate correlations (Pearson’s and Kendall’s Tau) and Shapiro-Wilk tests of normality. To identify subgroups with similar underlying menstrual pain trajectories, a two-through-five class latent class growth mixture models (LCGMM) with linear specifications of menstrual pain as a function of study time points were fit to see what number of classes best fit the data via the lcmm package in R. 19 Multiple fit indices related to parsimony and clustering (e.g., Bayesian Information Criterion-BIC, Sample Size Adjusted BIC-ssBIC, Akaike Information Criterion-AIC, Classification Likelihood Criterion-CLC, Entropy, etc.) were examined in tandem with prior research, theory, and clinical judgment regarding menstrual pain to identify the number of classes. 20 , 21 Nonparametric Kruskal-Wallis tests were used to see if menstrual pain trajectory groups had differing baseline levels of NMPP, anxiety, depression symptoms, pain catastrophizing, somatic symptoms, and sleep disturbance. Post-hoc pairwise Wilcoxon rank-sum tests determined which groups differed significantly from each other, and the Benjamini–Hochberg procedure was used to limit the false discovery rate to 5%. 22

To identify cross-sectional factors influencing trajectories, we conducted a correlation analysis between menstrual pain and related symptom domains. Baseline menstrual pain was strongly correlated with pain ratings at Year 1 and Year 2 ( Table 2 ). Higher baseline menstrual pain was also associated with greater NMPP, somatic sensitivity, pain catastrophizing, and sleep disturbance. These relationships remained evident at both follow-up assessments, suggesting that greater menstrual pain is consistently linked with broader comorbid symptom burden over time. Growth mixture modeling indicated that a four-class solution best fit the data ( Figure 1 ). Average menstrual pain showed a small but significant decrease over time (β = −3.82, SE = 1.07). The four classes are further described here following their individual patterns. The most common was the high stable group : 63% ( n = 103) reported consistently high menstrual pain across two years. On the opposite spectrum, was a low stable group 15% ( n = 25) reported consistently low pain, with 88% being pain-free controls. There were two dynamic groups: an improving pain subgroup : 11% ( n = 14) whose initial high pain level decreased, and a worsening pain subgroup: 11% ( n = 15) who began with low pain that increased during follow-up; notably 40% of the latter women had either a chronic pain or bladder pain syndrome diagnosis already. Significant differences emerged between pain trajectory groups for NMPP ( p < .001, η 2 = 0.14), somatic sensitivity ( p < .001, η 2 = 0.09), and sleep disturbance ( p < .05, η 2 = 0.03). These values suggest a large effect for NMPP, a moderate effect for somatic sensitivity, and a small effect for sleep disturbance (see Figure 2 ). Compared with the low-stable group, women in the high-stable group reported higher baseline levels of NMPP ( M = 18 vs. 5), somatic sensitivity ( M = 4.0 vs. 1.2), and sleep disturbance ( M = 20 vs. 11; all p < .05). The low-stable group also reported significantly lower NMPP ( M = 19 vs. 5; p < .001) and somatic sensitivity ( M = 4.5 vs. 1.2; p < .001) than the improving group. However, the worsening group could not be distinguished from any other group based on these baseline variables. Given that hormonal treatments and reproductive events alter uterine biology, we conducted post-hoc analyses to compare Year 1 and 2 oral contraceptive use and childbirth events on trajectory groups. Small but significant effects were demonstrated, suggesting that the improving pain group was more likely to use contraceptives in Year 1 ( p = 0.023) and 2 ( p = 0.015) as compared with the high-stable pain group (67–77% of participants [improving pain] vs. 27–32% [high-stable], see Figure 3 ). There were too few births (new childbirths Year 1= 3; Year 2 = 4) to conduct statistical analyses for trajectory groups. It was also not feasible to analyze the impact of other medications —only 9 participants acknowledged consistently taking analgesics other than NSAIDs.

The present study suggests that a stable elevated menstrual pain trajectory is relatively common in moderate or greater dysmenorrhea sufferers over two years. The phenotype associated with stable symptoms was differentiated from women with minimal symptoms by levels of pelvic pain outside the uterus, overall somatic sensitivity, and sleep disturbance. In contrast, a relationship with psychosocial factors was identified. Discerning between central and peripheral pain processing dysregulation in women reporting persistent dysmenorrhea should be a priority for follow-up studies.

Menstrual pain trajectories were stable across a two-year period in 80% of this cohort of largely dysmenorrhea sufferers. Baseline differences in NMPP, somatic sensitivity, and sleep disturbance differentiated between stable trajectories of high and low levels of menstrual pain. However, baseline factors such as pain catastrophizing, anxiety, and depression symptoms did not significantly differ between trajectory groups. While this is an important and reassuring finding, our sample may have been underpowered to detect such relationships in the worsening group, where they might have been most evident. In a post-hoc analysis, we observed that twice as many participants in the improving pain group used oral contraceptives, compared with the high-stable participants. The impact of hormonal therapy on trajectory class may not be causal, given that this is an observational cohort. We similarly identified four distinct menstrual pain trajectory groups as described by Ju and colleagues over 13 years - low-stable, high-stable, improving, and worsening. Key differences result from our oversampling for women with moderate or greater menstrual pain, resulting in higher proportion in the high-stable group (16.5% vs. 63.0%), with only 22% of participants having a varying time course over two years. Our assessment of psychosocial and behavioral predictors of trajectory expands on prior work with these factors which has been largely limited to cross-sectional comparisons, 23 and neither Weissman et al nor Ju et al included such variables in their analysis. 4 , 6 As expected, NMPP in our study had a large effect size for group membership, with the low-stable pain group having significantly lower levels of NMPP than both the high-stable and improving pain groups. A high prevalence of dysmenorrhea is observed in CPP populations, and high menstrual pain conveys risk for CPP later in life. 2 , 11 , 24 Since NMPP is related to the severity of menstrual pain, differences across the trajectory groups could be related to the relative degree of uterine cross-organ sensitization, leading to both overt or occult pelvic visceral sensitivity as previously described. 11 Indeed, a prior study of this cohort has shown that experimentally-assessed bladder sensitivity is positively associated with both menstrual pain and NMPP. 13 Additional individual-level factors differentiated the trajectory groups. Somatic sensitivity reported at baseline was lower in the low-stable pain group vs. the high-stable and improving groups (higher menstrual pain at baseline), which is consistent with this group overall exhibiting little symptom burden. Somatic sensitivity expressed as symptoms generally appear higher in populations with dysmenorrhea. 25 Levels of baseline sleep disturbance produced a small effect size which only differentiated the high-stable from low-stable pain trajectory groups. Prior research suggests that sleep disruption is more predictive of pain in general than vice versa, and similarly these findings indicated that disrupted sleep is most prominent among those experiencing chronic heightened experiences of menstrual pain over two years. 26 , 27 Among the nearly 80% of the cohort that had stable pain trajectories (high or low), it is notable that having higher levels of factors linked to chronic pain--NMPP, somatic sensitivity, and sleep disturbance--predicted trajectory class. Targeted early intervention for high-stable pain patients might shift them toward a more favorable trajectory—an idea that warrants future investigation. As one example, greater baseline sleep dysregulation in the high-stable compared with low-stable pain groups suggest a potential area for behavioral health intervention. Several other studies have reported sleep disruption in dysmenorrhea both during and between menstruation events. 28 - 30 Unfortunately for these women, higher menstrual pain was five times more likely to be stable than improve during follow-up. Given the widespread availability of known treatments for dysmenorrhea such as NSAIDs and hormonal contraception, effective treatment of primary dysmenorrhea may require optimal dosing of NSAIDs at sufficient doses, active management of hormonal therapy to ensure amenorrhea, or finding new pathways to reduce uterine nerve activation. 31 Such efforts may then reduce a woman’s risk for CPP development. This study characterizes many of the key constructs that influence chronic pain vulnerability that heretofore have not been used to study the progression of menstrual pain, including mood profiles, catastrophizing, somatic sensitivity, sleep, and non-menstrual pelvic pain experience. A key strength of our study was that menstrual pain intensity was prospectively characterized, and the study attempted to minimize participants with secondary dysmenorrhea through a baseline gynecological exam with a fellowship-trained gynecological surgeon. 12 However, we did not continuously re-examine patients (physical exam or imaging), so a minority of the worsening trajectory groups could be from new secondary causes. Other limitations include representation largely of nulliparous young adults, with very few intervening pregnancies, limiting the generalizability of the results across the reproductive timeline. Nevertheless, the late adolescent to young adulthood window is a critical epoch. Menstrual pain is often severe, causing limitations to critical social and professional development that influence future career prospects. 32 In addition, although we queried oral contraceptive use during follow-up, we could not assess other potential treatments such as NSAIDs or even surgery for secondary dysmenorrhea causes, because participants asked to limit the burden of questionnaires for these longer term follow-up surveys, which were ancillary to the original study aims. Finally, two years of follow-up may be too short a time window to evaluate chronicity, but we are collecting additional years of follow-up data which should extend these initial analyses.

Dysmenorrhea, cyclical menstrual pelvic pain of moderate or greater intensity, affects 40% of reproductive-age women and is a leading cause of missed work, school, and social activities. 1 In addition, dysmenorrhea is a risk factor for general chronic pain and frequently comorbid with chronic pelvic pain disorders. 2 In general, the contributing factors underlying progression are poorly understood, in part due to variations in menstrual pain experienced over time from hormonal contraceptive use and intervening pregnancies. 3 There are only a few prospective studies of menstrual pain symptoms. One study identified four distinct menstrual pain trajectories among a nationally representative sample of young adult Australians (N=9671). 4 Over a 13-year span, “normative” (no or few symptoms; 38.3%), recovering (decreasing intensity; 17.2%), low (moderately increasing intensity; 28.0%) and chronic (stable high intensity; 16.5%) trajectories emerged. Unemployment, tobacco use, and earlier menarche were linked to more persistent symptoms. Conversely, the normative group was more likely to be married and have used hormonal contraception. 4 Two other longitudinal studies over five and six-year periods found that use of oral or intrauterine contraceptives, prior childbirth, and advancing age predicted reductions in pain. 5 , 6 These studies, however, did not examine other modifiable factors associated with pain states including psychosocial factors, pain catastrophizing, somatic sensitivity, or sleep disturbance. 7 - 9 In addition, these cohorts being from more general populations may not be as informative about chronic pain risk as a group enriched for dysmenorrhea. In the present study, we conducted a secondary analysis of a prospective two-year natural history study of menstrual and bladder pain mechanisms to characterize menstrual pain trajectories. Prior research in this cohort of young women, which oversampled for dysmenorrhea, has documented a clear association between degree of menstrual pain intensity and bladder sensitivity even among chronic pain-free patients, which may reflect early cross-organ sensitization. 10 Further justification for performing multifocal pelvic pain assessment when investigating menstrual pain comes from evidence that treating one pelvic pain condition can reduce pain in adjacent pelvic organs. 2 , 11 Furthermore, we have previously shown that chronic-pain free women with comorbid dysmenorrhea and subclinical bladder pain sensitivity score worse than pure dysmenorrhea sufferers on multiple psychosocial domains, suggesting their pain vulnerability may be closer to that seen in full-blown chronic pain patients. 12 Therefore, we characterized levels of non-menstrual pelvic pain (NMPP), psychosocial distress, overall somatic sensitivity, and sleep disturbance at baseline as potential factors that could distinguish trajectories of menstrual pain over time. Based on prior research, we hypothesized that four distinct menstrual pain trajectory groups would emerge: (1) worsening pain; (2) improving pain; (3) low-stable pain; and (4) high-stable pain over two years. Next, we hypothesized that baseline factors linked to chronic pain (e.g., NMPP, anxiety and depression symptoms, pain catastrophizing, somatic sensitivity, and sleep disturbances) would differentiate these trajectory groups. Baseline oral contraceptive use, age, and history of childbirth were evaluated as potential covariates to include in the trajectory analysis. Given that hormonal treatments and reproductive events alter uterine biology, we also examined hormonal contraceptive use and pregnancy history serially to consider whether these differed between trajectory groups as a separate post-hoc analysis.