Natural History of Menstrual Pain and Associated Risk Factors in Early Adolescence.

Participants were enrolled in A Girl’s First Period Study (established in December 2019), an ongoing, observational longitudinal study over the course of 2–3 years of healthy, early post-menarchal girls residing in the Triangle region of North Carolina. The broad goal of the Study is to better understand normal hormonal dynamics from menarche to the establishment of mature ovulatory cycles. Full details of this protocol have been previously published 24 (note that enrollment has continued since this publication as the protocol remains active, NCT02583646 ). The current analyses include all participants who were enrolled prior to December 2023 and who completed at least one Menstrual Symptom Questionnaire [MSQ] (n=43). All data used in these analyses were collected prior to March 2024. The study was approved by the NIH Institutional Review Board. Informed assent and consent were obtained from each participant and her guardian, respectively. Participants were ages 10–14 years old and no more than 6 months post-menarchal at enrollment. Inclusion criteria included a healthy weight (defined as having a body weight >85th% of expected body weight and a body mass index [BMI] <99th percentile) and normal levels of thyroid hormone, prolactin, and testosterone. Exclusion criteria included pregnancy, anemia, taking or planning to take medication that affects reproductive hormones, or having a chronic medical condition (e.g., type 1 diabetes, cancer, ulcerative colitis, and as determined by the Principal Investigator’s discretion), a first degree relative with a pubertal disorder, or an excessive exercise routine (defined as running over twenty miles per week or its equivalent). Over the course of the study, one participant took medication for depression, two participants took medication for anxiety, and four participants took medication for attention-deficit/hyperactivity disorder (ADHD). These participants were included in the study. Screening visit procedures included a medical history and physical exam, questionnaires (as detailed below), BOD POD ® exam (COSMED USA 50.26L, Chicago, Illinois) to determine body composition, blood and urine sampling for screening laboratory testing, and a transabdominal pelvic ultrasound to confirm normal anatomy of the ovaries and uterus. For the duration of follow-up, participants provided daily first-morning urine samples using filter paper strips to measure creatinine (Cr)-corrected luteinizing hormone (LH), estrone-3-glucuronide (E1G), and pregnanediol-3-glucuronide (PdG) via enzyme immunoassay (ZRT Laboratory, Beaverton, Oregon). Participants reported menstrual bleeding using an online calendar, as previously described 24 . Every 4–6 months, participants underwent research study visits to complete clinical assessments and questionnaires, including the Menstrual Symptom Questionnaire (MSQ). Participants completed the Menstrual Symptom Questionnaire (MSQ) at screening and once at the start of each research study visit. The MSQ is a 24-item self-report survey 25 developed by Chesney and Tasto to differentiate adult women with spasmodic versus congestive dysmenorrhea. Each item is scored on a scale of 1–5, for a possible score of 24–120. In the original study, 56 otherwise healthy female college students with a history of menstrual pain completed the MSQ at baseline and 2 weeks later. The median scores were 77 at baseline and 76 at 2 weeks. Note that the MSQ was re-examined and validated in adolescents to assess menstrual symptoms by Negriff et al. in 2009 26 . Negriff et al. also proposed the use of a three-factor structure with separate consideration of abdominal pain, back pain, and negative affect/somatic complaints 26 . Participants also completed the Godin Leisure-Time Exercise questionnaire 27 at the screening visit to quantify their physical activity levels. Total scores of 24 or more indicate an active participant, a score of 14–23 is defined as moderately active, and less than 14 is interpreted as insufficiently active or sedentary. Parents of participants completed the PROMIS Emotional Distress depression and anxiety questionnaires 28 , 29 semi-annually to determine the participant’s mood over the previous seven days. For both questionnaires, a T-score less than 55 implies none or slight depression or anxiety, a score of 55–59.9 signifies mild depression or anxiety, a score of 60–69.9 correlates to moderate depression or anxiety, and a score of 70 or higher is interpreted as severe depression or anxiety 30 . Overall MSQ scores and the three factor-specific MSQ scores proposed by Negriff et al. (abdominal pain, back pain, and negative affect) 20 were the outcome variables for primary and secondary analyses. Prior to analysis, MSQ scores were natural log-transformed to improve normality. Exposures evaluated in the primary analyses were gynecologic age (years) as well as characteristics of the menstrual cycle preceding MSQ completion, including: 1) peak PdG value in the last 17 days of the cycle (continuous); and 2) presumed ovulation status, defined as a peak PdG value exceeding 2500 ng/mg Cr (binary). Gynecologic age was calculated as the time between reported date of first period and the current date for each completed MSQ. The 2500 ng/mg Cr PdG threshold in the last 17 days of the cycle represents the minimum peak PdG observed in a study of 20 healthy women with regular menstrual cycles [unpublished data, Dr. Janet Hall]. Participant data were included in the analyses for gynecologic age each time they completed the MSQ. Participants contributed to models for peak PdG and presumed ovulation status when the menstrual cycle prior to MSQ completion had sufficient urinary hormone data to determine peak PdG and ovulation status. These models exclude MSQs conducted at baseline, when the prior cycle had not been fully observed. Secondary analyses further evaluated physical activity, anxiety, and depression measures as exposure variables. In analyses of physical activity, scores were carried forward from baseline because the Godin Leisure-Time Exercise questionnaire was only completed at the screening visit. Raw scores from the PROMIS Emotional Distress questionnaire were used as measures of anxiety and depression. Participants with a completed baseline Godin Leisure-Time Exercise questionnaire contributed to models each time they completed an MSQ. Participants contributed observations to analyses for anxiety and depression each time they completed a MSQ and their parent or guardian completed the PROMIS questionnaire at the same study visit. To account for potential confounding, analyses were adjusted for participant characteristics that may be related to both the exposures of interest and the experience of menstrual pain. These characteristics included: age at menarche (years), baseline BMI Z-score, race (White (ref.) vs. non-White), ethnicity (Hispanic vs. non-Hispanic (ref.)), mother’s age at menarche (years), parental education (both parents with college degrees vs. one or both parents without a college degree) and employment (both parents employed vs. one or both parents unemployed). For primary analyses where gynecologic age was not the primary exposure and for all secondary analyses, gynecologic age (years) was also included as a covariate. Participant demographic characteristics and anthropometric measurements were obtained from data collected at study enrollment. Generalized estimating equations (GEE) were used to model the association between each exposure and outcome for both primary and secondary analyses. We used GEEs for these analyses in order to appropriately account for having multiple observations per participant over time, which tend to be correlated and therefore violate independence assumptions underlying traditional regression approaches. More specifically, for each primary and secondary analysis, we fit a GEE with a normal distribution and autoregressive covariance structure and adjusted for the covariates as described above. Given that MSQ scores were natural log-transformed for analyses, we back-transformed the beta coefficients from the GEE results for each association of interest by exponentiating them to ease interpretation. These back-transformed estimates represent the expected multiplicative change in MSQ score for a 1-unit increase in each exposure of interest; from here on, we refer to these back-transformed estimates as ‘MSQ score ratios’ for simplicity. To aid in visualizing the relationship between gynecologic age and MSQ score for an individual participant, we plotted the predicted MSQ scores for gynecologic age ranging from 0–2 years ( Figure 1 ). Calculation of the predicted scores required us to specify values for the covariates. We selected covariate values that represented the approximate average values observed among our participants: age at menarche: 12.4 years; BMI Z-score: 0.6; White; non-Hispanic; both parents employed and college-educated, and mother’s age at menarche=12.6 years. We also explored GEE models with the following three changes: 1) using the average proportion of ovulatory cycles over the previous 6 months rather than just the prior cycles ovulation status; 2) evaluating the relationship between physical activity and MSQ scores at baseline only; and 3) modeling the t-scores corresponding to the raw scores from the PROMIS questionnaire. Additionally, we evaluated whether presumed ovulation status modifies the relationship between gynecologic age and MSQ scores by estimating a GEE using overall MSQ score as the outcome and including an interaction term between gynecologic age and presumed ovulation status in the preceding cycle.

The 43 participants were 12.6 ± 1.1 years old (mean ± SD) at enrollment. Most participants were White (69.8%), non-Latina (90.7%), had a healthy weight (BMI between 5 th and 85 th percentile, 72.1%), with Tanner V breast development (95.2%) ( Table 1 ). On average, mother’s age at menarche was 12.6 ± 1.1 years. The majority (69.8%) of participants had parents who both attended college and 48.8% had parents who were both employed. Participants completed 166 MSQs (mean: 3.9 ± 1.8, range of 1–9 completed questionnaires across participants) over the course of an average of 1.8 years (SD=1.2, range: 0–4.0) of follow-up. At the first completed MSQ, the average overall score across participants was 39.7 (SD=12.7), with average factor scores of 13.8 (SD=5.4) for abdominal pain, 4.6 (SD=2.4) for back pain, and 18.9 (SD=1.1) for negative affect ( Table 1 ). By the time of the last completed MSQ, average overall participant scores had increased to 51.5 (SD=16.4), with average factor scores of 18.3 (SD=6.9) for abdominal pain, 5.8 (SD=3.3) for back pain, and 23.8 (SD=8.0) for negative affect. At the time of enrollment, the average gynecologic age was 0.3 ± 0.2 years ( Table 1 ). During the first full menstrual cycle observed after enrollment, the average peak PdG (ng/mg Cr) was 1456.3 (SD=1043.7) and 15% of participants’ cycles were considered ovulatory. Average baseline physical activity scores were 55.4 (SD 28.4), which is interpreted as active. Over the course of follow-up, the majority of PROMIS questionnaires scored in the none or slight depression (93.4%) and anxiety (94.6%) categories. Total MSQ scores were higher for every 1 year increase in gynecologic age (MSQ score ratio: 1.12; 95% CI: 1.08, 1.17; p<0.0001) as were all factor-specific MSQ scores (abdominal pain: 1.14; 95% CI: 1.08, 1.19; p<0.0001; back pain: 1.08; 95% CI; 1.02, 1.14, p=0.0066; negative affect: 1.12; 95% CI: 1.06, 1.17; p<0.0001) ( Table 2 ). Predicted MSQ scores over gynecologic age values ranging from 0–2 years for a participant with otherwise average covariate values are displayed in Figure 1 to illustrate how the estimated MSQ score ratio corresponds to increasing predicted scores over time. Total MSQ scores were 1.04-fold higher (95% CI: 1.02, 1.06; p=0.0002) for every 1000 ng/mg Cr increase in peak PdG in the preceding cycle, and the abdominal pain factor score was 1.04-fold higher (95% CI: 1.01, 1.07; p=0.0041) for every 1000 ng/mg Cr increase in peak PdG ( Table 2 ). Total MSQ scores were 1.26-fold higher (95% CI: 1.11, 1.44; p=0.0002) when the preceding cycle was presumed ovulatory. Similarly, the abdominal pain factor score (MSQ score ratio: 1.32, 95% CI: 1.12, 1.57; p=0.0012) and the negative affect factor score (MSQ score ratio: 1.16, 95% CI: 1.02, 1.32; p=0.02) were positively associated with presumed ovulation in the prior cycle ( Table 2 ). Total MSQ scores were not significantly associated with physical activity, anxiety raw score, or depression raw score ( Table 3 ). Exploratory GEE models that either: 1) used the average proportion of ovulatory cycles over the previous 6 months, 2) evaluated the relationship between physical activity and MSQ scores at baseline only; or 3) modeled PROMIS questionnaire t-scores (instead of raw scores) had comparable results to those presented in the primary and secondary analyses above and are not included here. When an interaction term for gynecologic age and presumed ovulation status in the preceding cycle was added to GEE for overall MSQ score, the interaction term was not significant (p=0.17). This suggests that ovulation status did not modify the relationship between gynecologic age and MSQ score (data not shown).

In this investigation, we utilized intensive reproductive phenotyping and repeated measures of menstrual pain in a community-based observational cohort of healthy girls. We demonstrated that menstrual pain increases as early as the first two years after menarche and that higher cycle PdG levels, and hence presumed ovulatory cycles, are associated with higher menstrual pain scores. Presumed ovulation status did not modify the relationship between gynecologic age and menstrual pain. Thus, there may be a pathophysiologic role for unknown factors related to increasing gynecologic age other than the development of ovulatory cycles in the development of menstrual pain among adolescents. There have been very few longitudinal studies to investigate the natural history of menstrual pain in adolescents and adults. In a study of 148 girls with severe dysmenorrhea initially referred to a tertiary care center in Australia, 70 girls were recontacted, on average 10 years (range 4–14 years) later, at an average age of 26 years (range 17–37 years) to determine the course of dysmenorrhea 8 . Using structured telephone interviews and questionnaires, the investigators determined that dysmenorrhea had resolved in approximately one-quarter of participants. No single adolescent characteristic predicted the persistence of severe dysmenorrhea into adulthood 8 . A second observational prospective study included 262 healthy girls recruited from an urban teen health center and the surrounding community of Cincinnati, Ohio 9 . The participants represented four different age cohorts (ages 11, 13, 15, and 17 years) with mean gynecologic ages of −0.5, 1.2, 3.2, and 5.0 years, respectively. Participants completed annual MSQs (as well as the Children’s Depression Inventory and the somatic complaints subscale of the Youth Self Report) for 3 years. These data demonstrated, akin to our findings, that menstrual pain scores increased over time and began to decelerate at the end of adolescence (≈ 18 years). Importantly, however, MSQs were specifically analyzed in relation to chronologic age, rather than to gynecologic age, and nearly 20% of subjects were using oral contraceptives (an unreported number were also using depot medroxyprogesterone acetate, a transdermal patch, or an intravaginal ring). Baseline somatic complaints and baseline depressive symptoms (but only in girls with lower MSQ scores at baseline) predicted increases in menstrual symptoms over time. Similar studies in adult women have shown either a slight decrease 31 or no change 32 in the prevalence of menstrual pain over time, however, parity appears to be protective 31 , 32 . Our study provides the first longitudinal data on menstrual pain during the early gynecologic years (in girls NOT on any hormonal contraceptives). We demonstrate that, compared to available data on late adolescence and early adulthood 33 , 34 , there is a modest increase in menstrual pain at a time when regular, ovulatory cycles have yet to be established. Traditional dogma has held that menstrual pain is confined to ovulatory cycles 11 . This is based on the premise that the fall in progesterone levels during the luteal phase is the requisite trigger for uterine prostaglandin release which stimulates uterine contractions and vasoconstriction. However, combined oral contraceptives (OCPs), which block ovulation and would therefore be expected to eliminate menstrual pain, have only a modest effect. For example, in a recent Cochrane review of 21 randomized controlled trials, Schroll et al. found that women with primary dysmenorrhea had a 28% chance of improved pain scores on placebo versus a 37–60% chance of improving on OCPs 35 . The results of several recent studies in adolescents and young women have also not supported the hypothesis that ovulation is required for menstrual pain to manifest. For example, in three independent studies conducted in the US 15 , Turkey 14 , and Spain 13 , investigators found that in participants with primary dysmenorrhea (age 11–24 years, gynecological age 4–9 years), a substantial fraction (17–37%) of cycles were anovulatory, and menstrual pain scores were not higher during ovulatory than anovulatory cycles. Similar findings were reported in a one-year longitudinal study of healthy, regularly cycling premenopausal Canadian women (age 21–41 years): menstrual cramp frequency, duration, and intensity were all similar in ovulatory and anovulatory cycles 16 . Of note, tests chosen to define ovulatory status differed between these studies. In our study of healthy early post-menarchal girls, recent presumed ovulation, as defined by peak cycle PdG levels >2500 ng/mL Cr, was associated with only a modest increase in menstrual pain scores. This suggests that factors aside from recent ovulation may contribute to the occurrence of menstrual pain. For example, it is possible that menstrual pain increases over time because of maturational-related changes in pain sensation. Several studies have suggested that recurrent menstrual pain may lead to greater CNS sensitivity to pain via both structural and functional reorganization of the brain 36 . Increased pain may also reflect changes in uterine sympathetic innervation. Uterine innervation is known to undergo remodeling in response to fluctuations in estrogen levels during the estrus cycle 37 and during pregnancy 38 . Alternatively, changes in the vaginal microbiome 39 and/or changes in sleep preference 40 during the early gynecologic years may also explain the increase in menstrual pain during this developmental stage. In contrast to previous studies, we did not observe poor mental health or lower physical activity levels to be risk factors for menstrual pain. In a larger cross-sectional study of >200 US adolescents, who were of a slightly older gynecologic age (average 3.1 years) than the current study’s participants, Dorn et al. found a positive association between depression, anxiety, and MSQ scores. However, in the study by Dorn et al., anxiety and depression symptoms were determined by self-report rather than parental report, using the State-Trait Anxiety Inventory and Children’s Depression Inventory, respectively 41 . Similar positive associations have been reported in other cross-sectional studies 42 , 43 . However, a recent large (n=1038) longitudinal study of a community-based cohort in Amsterdam found no association between anxiety and depression (assessed at age 11–12 years) and menstrual pain (assessed at age 15–16 years) 44 . Lifestyle factors (including physical activity) were not predictors of menstrual pain in the Amsterdam cohort study 44 . The key strengths of our study include the longitudinal design, daily urinary reproductive hormone measurements, and unique focus on the early post-menarchal period in a community-based cohort. However, there are limitations. While subjects were studied for up to 4 years, physical activity was determined by self-report at screening rather than by an objective assessment such as actigraphy throughout the study. MSQ and mental health questionnaires were only completed semiannually, whereas others have monitored dysmenorrhea in adolescents on a daily basis using a mobile app 45 . There was not a wide variation in the mental health questionnaire scores, which may have limited our ability to detect associations. Mental health questionnaires were also completed by participants’ parents rather than by the participants themselves. We did not include subjects with eating disorders (using an expected body weight criterion), subjects with severe obesity, and those who exercised excessively which may have also limited our ability to detect associations. Sample size was relatively small, and participants had variable amounts of follow-up. Furthermore, the cohort was predominantly non-Hispanic White, of normal weight, and of higher socioeconomic status, which may limit generalizability. Maternal history of menstrual pain was not obtained, and we cannot exclude the possibility of early endometriosis in participants. One additional limitation is that it is difficult to definitively determine if the observed change in MSQ scores over time is clinically significant because: 1) we did not collect additional clinical information from our participants, 2) the MSQ is not typically used in the clinical setting in treating adolescents with menstrual pain, and 3) to our knowledge, there are no studies demonstrating a change in MSQ scores in response to treatment in adolescents with menstrual pain (note that in studies by Jay et al. 46 in adolescent girls with primary dysmenorrhea there was a statistically significant decrease in MSQ scores in response to naproxen but the post-intervention scores were not reported). Nevertheless, in a small study of Turkish women with primary dysmenorrhea (mean age 23 yrs) randomized to placebo or 8-weeks of exercise, the women in the intervention group demonstrated a 31% decrease in MSQ scores (from 77 to 53), which represents a large clinical effect size, as determined by the authors (r=0.6) 47 . This change is comparable to that observed in our participants over time (baseline total MSQ 39.7, last completed MSQ 51.5, representing a 23% increase), suggesting that this change is indeed clinically relevant. In summary, menstrual pain may be present at a young gynecological age and tends to increase over time in healthy girls with normal pelvic anatomy. This increase is partly associated with increased progesterone levels and, hence, presumed ovulatory status. Future studies are necessary to identify additional factors which contribute to menstrual pain during the early gynecological years so that more targeted therapies may be developed for young girls with menstrual pain whose symptoms negatively impact their daily functioning and quality of life.

Menstrual pain, or dysmenorrhea, is one of the most common gynecological complaints among adolescent girls. Most adolescents with menstrual pain have primary dysmenorrhea, or menstrual pain of uterine origin that occurs in the absence of pelvic pathology. Menstrual pain is an important public health problem that affects up to 93% of adolescents 1 . In addition to pelvic pain, menstrual pain frequently includes other physical symptoms, such as headaches, gastrointestinal symptoms, back pain and/or psychological symptoms, such as anxiety, depression, or irritability 2 . This broad constellation of symptoms likely explains why menstrual pain is one of the leading causes of school and work absenteeism among adolescents 3 , 4 . Menstrual pain may also represent a risk factor for developing chronic pelvic and nonpelvic pain in adulthood 5 , 6 . While the onset of menstrual pain is anecdotally thought to occur 6 to 24 months after menarche 2 , 7 , there is limited data on the natural history of menstrual pain beginning in the early post-menarchal period and extending into adulthood 8 , 9 . The pathogenesis of menstrual pain (in the absence of pelvic pathology) is not fully understood, but it is believed to be related to prostaglandins, vasopressin, pro-inflammatory cytokines, and/or increased pain sensitivity 10 . The most widely accepted theory is that in patients with menstrual pain, the rapid fall in progesterone levels during the luteal phase triggers the excessive production of uterine prostaglandins 11 . This theory posits that the decrease in progesterone destabilizes endometrial lysosomes, leading to the release of the lysosomal enzyme phospholipase A2. This enzyme acts on cell membrane phospholipids to generate arachidonic acid, the main precursor of prostaglandins. Prostaglandins such as prostaglandin F 2α cause uterine contractions and vasoconstriction, ultimately leading to uterine ischemia and pelvic pain 12 . While the theory that menstrual pain only occurs in ovulatory cycles has dominated the literature since it was first contended in the 1930s 12 , a number of recent studies have identified menstrual pain among subjects with anovulatory cycles 13 – 16 , calling this notion into question. A number of risk factors for menstrual pain 17 have been identified, but effect sizes have been small, findings have been inconsistent, and with rare exception 4 , 18 , 19 , most studies have focused on adult women rather than on adolescents. Risk factors include those related to socio-demographics (younger age, lower socio-economic status, obesity), reproductive traits (younger gynecologic age, longer cycles, irregular cycles, heavy bleeding, no history of oral contraceptive use, higher parity), lifestyle (smoking, less physical activity), and mental health (increased stress, depression, and exposure to adverse childhood experiences) 20 – 22 . Patients with menstrual pain also frequently endorse a positive family history, suggesting a genetic component. Indeed, recent genome-wide association studies of dysmenorrhea identified several loci that achieved genome-wide significance, such as the nerve growth factor ( NGF ) gene locus 23 . In the current study, we utilized data from an ongoing longitudinal study of early post-menarchal girls to determine the following: 1) the natural history of menstrual pain, 2) the potential role of progesterone in its pathophysiology, and 3) risk factors for menstrual pain. Our primary aims were to investigate the relationship between menstrual pain, as measured by the overall and factor-specific Menstrual Symptom Questionnaire (MSQ) scores, and: 1) gynecologic age; 2) peak pregnanediol-3-glucuronide (PdG) observed in the prior menstrual cycle; and 3) PdG-based ovulatory status. We hypothesized that reported menstrual pain scores would be higher for increased gynecologic age and for cycles that had higher PdG values or that were presumed ovulatory. As secondary analyses, we also investigated the relationship between menstrual pain and physical activity, anxiety, and depression. We conducted these analyses to test the hypotheses that menstrual pain scores would be inversely associated with physical activity level and positively associated with mental health symptoms.