{"paper_id":"2fbca04b-ac49-4c80-bd02-0798f0e29987","body_text":"Pelvic pain experienced by women encompasses a number of recurrent or persistent pain conditions. Pelvic pain comprises dysmenorrhoea (menstrual pain), dyspareunia (pelvic pain associated with sexual intercourse), and other pain conditions occurring not with periods and intercourse. The latter is usually defined as chronic pelvic pain if it occurs for at least six months. 1  Dysmenorrhoea is considered to be primary dysmenorrhoea if it occurs soon after menstrual cycles become established, in the absence of pelvic pathology. 2\nSystematic reviews of pelvic pain show prevalence estimates are high, but vary substantially even between higher quality studies. Dysmenorrhoea in women of reproductive age ranges from 16–91%, 2  and dyspareunia from 8–26.2%. 3  The prevalence of chronic pelvic pain is also high in population-based studies, with estimates of 11–25.4%, 4 – 6  Although prevalence studies have these wide variations, a consistent observation is that dysmenorrhoea, dyspareunia and chronic pelvic pain co-occur. 4 – 6\nPelvic pain conditions are associated with increased pain sensitivity due to central changes in the nervous system, 7 – 11  and other factors that affect the neurobiology of pain (e.g. sleep fragmentation). 11 ,  12  This suggests that all may have common causal factors, especially when pain-sensitising conditions are present (e.g. endometriosis, surgical nerve injury). 13 ,  14  Nevertheless, established risk factors common to all three conditions are limited to psychosocial stressors. 7  Age, obstetric and gynaecological factors have been found to be associated with dysmenorrhoea in many studies, but there is only weak evidence for these factors with dyspareunia and chronic pelvic pain.\nResearch on pelvic pain is complicated by the heterogeneity in definitions of pain, perception of pain and likely complicated causal networks. Nevertheless, it is notable that evidence for causation for many potential determinants is limited to single studies, poorly designed studies, or relies on long-term retrospective recall of both biological and psychosocial events. 2 ,  7 ,  15  Furthermore, there is evidence of publication bias. 7 ,  15\nBy using a long running birth cohort study of women born in New Zealand in 1972/3 with excellent retention of participants, we aimed to reduce bias from selective participation and inaccurate recall, provide further evidence regarding previously identified explanatory factors, and examine combinations of these factors to both address confounding and consider causal networks. We estimated the prevalence of pelvic pain (dysmenorrhoea, dyspareunia, and other pelvic pain), and associations with demographic, obstetric, gynaecological, and psychosocial factors, at age 38 years.\n\nThe Dunedin Multidisciplinary Health and Development Study (DMHDS) is an ongoing longitudinal study of 535 men and 502 women born between April 1972 and March 1973 in Dunedin, New Zealand, with the most recent complete assessment at age 38 years. 16  The cohort represented the full range of socioeconomic status in New Zealand’s South Island and 92% of the women identified with New Zealand European ethnicity. Ethical approval was obtained from the Otago and Southern Regional Ethics Committees for all assessments (reference LRS/10/03/012/AM03, dated 13/04/2010). Funding was provided by the Health Research Council of New Zealand (PA209-09/086 and GA209-09/086D); the United States National Institute on Aging (R01AG032282), and the United Kingdom Medical Research Council (MR/K00381X), with support from the New Zealand Ministry of Business, Innovation and Employment.\nAt the age 38 assessment, women responded to computer-presented questions about lower abdominal pain in the past 12 months with the aid of an illustration of the lower abdomen to the top of the mons pubis. Questions included when the pain occurred (menstrual pain; pain associated with sexual intercourse; and/or pain not related to menstruation or intercourse), severity and frequency of pain, and the when this pain first occurred. In the absence of a core outcome set for pelvic pain, we defined our main outcomes as dysmenorrhoea (any report of menstrual pain), dyspareunia (any pain with intercourse) and, otherwise, other pelvic pain (OPP). We did not account for severity for these three outcomes, in keeping with most other publications on pelvic pain. OPP was not defined as chronic pelvic pain since our questionnaire could not establish for 19 of 74 women with OPP whether they had pain of at least six months’ duration. Pelvic pain was considered to be clinically significant if it included reporting pain to be ‘severe’, and also ‘moderate’ if this occurred always for dysmenorrhea and dyspareunia or at least monthly for OPP. Primary dysmenorrhoea at age 38 was differentiated from secondary dysmenorrhoea based on reporting pain when participants were assessed at age 13, or in the absence of menstruation at age 13, reporting pain at the age 15 assessment, or, in the absence of menstruation at age 15, based on reporting pain at age 38 that occurred from the age they reported first menstruating.\nSocioeconomic status, based on employment, with adjustment for homemakers, was defined as high, medium or low. 17  Highest education qualification was measured at the age 26 assessment and grouped as none or secondary school, post-secondary, and university degree. Body mass index (BMI) was measured by a trained interviewer at age 38 and categorised as not obese (<30kg/m 2 ), moderate obesity (30–39.9kg/m 2 ) or severe obesity (≥ 40kg/m 2 ). Being a current smoker was self-reported.\nObstetric events (pregnancy, miscarriage, birth and caesarean section) and when these occurred was first reported at the age 26 assessment, and subsequent events were reported at the age 32 and 38 assessments. Time since the most recent birth was also calculated and grouped as never, >5, 3–5, and <2 years.\nAge at menarche was grouped as <13, 13–14, and ≥15 years. At age 13 and 15 the participants were asked about experiencing pain or discomfort for the first few days of menstruation with options being ‘no, does not apply’, ‘yes, applies somewhat’ and ‘yes, certainly applies’. Either of the affirmative responses was considered to be dysmenorrhoea. Past chlamydia infection was defined as self-reporting a diagnosis of  Chlamydia trachomatis  at age 26, 32 or 38, or positive serology at any of these assessments’. At age 38, women were asked if they had ever been diagnosed with fibroids or endometriosis and any use of hormones since age 32 was recorded.\nInformation about childhood sexual abuse was collected at the age 26 assessment, and included being touched by or forced to touch their abuser, penetrative and attempted penetrative sex, oral sex and subjection to pornography under the age of 16 years. A history of ever having been raped, and any major depressive episode or anxiety disorder between ages 32 and 38 was also documented at age 38. Poor sleep quality (reported overall sleep quality in the past month as ‘very bad’ or ‘fairly bad’) was measured at age 38.\nAll analyses were conducted in Stata/SE 14.2 (StataCorp, Texas) with 95% confidence intervals (CI) presented and two-sided P<0.05 defining statistical significance. Women who had been pregnant in the previous 12 months were excluded from all analyses. The response rate at age 38 was calculated and the demographic characteristics and prevalence of potential explanatory factors described. The overall prevalence of pain, and pain by subtype, was calculated. For dysmenorrhoea, women who had not menstruated in the past 12 months were excluded, and for dyspareunia women who had not had reported intercourse in the past 12 months were excluded. To estimate the unadjusted relative risks (RR) of explanatory factors with each pain subtype we used Poisson regression with robust standard errors. To provide adjusted relative risks (ARR) a modified Hosmer and Lemeshow approach was used: 18  potential explanatory factors and confounders were included in adjusted models if they were of  a priori  interest, other factors were initially included if unadjusted P<0.25 and retained in the adjusted model if they had a significant effect. The statistical significance of variables was determined using Wald tests. The dysmenorrhoea model was repeated limiting the outcome to those who had clinically significant pain, then to those who had primary dysmenorrhoea and then to those with secondary. As a full ethnicity breakdown for the cohort is currently unavailable, ARRs for the three pain types were recalculated adjusting for reporting European ethnicity to gauge any ethnicity-specific effect in this cohort.\n\nOf the 502 women in the original cohort, 12 were deceased at the age 38 assessment. Of the remaining 490 women, one woman was excluded from this interview due to severe intellectual disability, and 466 (95.3%) participated in the age 38 assessment. Our analyses excluded 37 women who had been pregnant in the 12 months prior to the age 38 assessment. At age 13, 139 of the 429 women included in this analysis reported menstruating and 80 (57.6%) reported dysmenorrhoea; at age 15, 237 (68.9%) of 344 reported dysmenorrhoea. The prevalence of other relevant characteristics of the study population are shown in  Table 1 .\nOf the 429 eligible participants at the age 38 assessment, 234 (54.5%, 95% CI 49.7–59.3%) had experienced pelvic pain in the past 12 months ( Table 2 ). Dysmenorrhoea was most common, being reported by 184 of 398 women (46.2%, 41.3–51.3%) who had menstruated. This was significantly lower than the prevalence at the age 15 assessment (68.9%) (McNamara’s  χ 2  P<0.001). More than half (112) of these women currently had, or had previously experienced, primary dysmenorrhoea. Of all women reporting dysmenorrhoea, 59 (32.1%) reported having frequent or severe pain and 67 (36.4%) felt their pain was getting worse with age. Of the 31 women who did not menstruate, 24 reported this was because of menstrual suppression with hormones or the levonorgestrel-releasing intra-uterine system, five because of hysterectomy, one was in menopause and one did not know the reason. Among women who had intercourse in the previous 12 months, 11.6% (8.7–15.2%) had dyspareunia. Four (8.5%) of these 47 women reported having frequent or severe pain and 15 (31.3%) reported their pain was getting worse with age. Among all 429 women, 17.3% (13.8–21.2%) had OPP; 22 (29.7%) of these 74 women reported having frequent and/or severe pain and 22 (29.7%) reported their pain was getting worse with age. Twenty-four (32.4%) had a cyclical component (either before or after a period or at ‘mid-cycle’).\nDyspareunia (ARR 1.77, 95% CI 1.44–2.16), fibroids (1.70, 1.20–2.40), and endometriosis (1.60, 1.27–2.01) were significantly (P<0.05) associated with dysmenorrhoea at age 38 after adjusting for potential confounding factors ( Table 3 ). These confounding factors (ever giving birth, miscarriage, and caesarean section) were of interest and, thus, retained in the adjusted model, but were not significantly associated with dysmenorrhoea. However, data were suggestive of a protective effect for those with a more recent birth in the past two years (unadjusted RR 0.73, 0.47–1.13) compared with never having a birth; this was not significant and had no impact on the final model, so was not retained. All other factors listed in  Table 1  and OPP were tested for association, none of these had any evidence of an effect (all unadjusted RRs >0.85 and <1.20, data not shown), except for BMI. Those who were severely obese had an unadjusted RR of 1.58 (1.07–2.31) compared with not obese, but the ARR was reduced (1.39) and no longer significant; as BMI also had no impact on the adjusted model it was not retained.\nClinically significant dysmenorrhoea was not significantly associated with dyspareunia (ARR 0.89, 0.53–1.49) or fibroids (1.56, 0.67–3.66), but was strongly associated with endometriosis (2.58, 1.71–3.88). For primary dysmenorrhoea the effect of dyspareunia was moderately reduced (1.45, 0.99–2.14) compared with the models for any dysmenorrhoea. For secondary dysmenorrhoea, fibroids (1.57, 0.54–4.55), and particularly, endometriosis (1.09, 0.54–2.17) had reduced ARRs; whereas dyspareunia increased (2.36, 1.45–3.84). Other patterns of association were similar to the model for any dysmenorrhoea in the past 12 months at age 38. All of the dysmenorrhoea models also controlled for hormone use since age 32; this was not associated with pain, ARRs were >0.95 and <1.05 for all models apart from significant pain (0.71, 0.47–1.07) (data not shown).\nPast pregnancy was protective for dyspareunia (unadjusted RR 0.54, 0.30–0.98), but this effect was borderline and in the adjusted model the slightly increased ARR was non-significant ( Table 4 ). Dysmenorrhoea at age 38 (ARR 4.54, 2.18–9.48) and rape (1.87, 1.08–3.22) were the only significantly associated factors in the adjusted model. However, endometriosis (1.81, 0.95–3.43), anxiety (1.68, 0.96–2.94) and OPP (1.62, 0.73–2.97) all had substantial positive associations, these associations were statistically significant in unadjusted models. There were also elevated unadjusted RRs for severe obesity (2.51, 0.93–6.79), past chlamydia infection (1.37, 0.79–2.37), major depressive episode (1.43, 0.79–2.59) and poor sleep quality (1.60, 0.93–2.73). However, in the adjusted model these factors remained non-significant, with the ARRs moving much closed to the null for depressive episode (1.08, 0.58–2.01) and sleep quality (1.18, 0.68–2.05), and they did not impact on associations shown in  Table 4 . No other factors showed any substantial deviation from a null effect (data not shown).\nFor OPP, ever giving birth was protective (unadjusted RR 0.52, 0.2 5–1.09), and there was no evidence this was confounded by other gynaecological and obstetric factors ( Table 5 ). Adjusted models suggested moderately increased OPP prevalence (ARRs all close to 1.45) with past chlamydia infection, endometriosis, poor sleep quality and dyspareunia; however, none of these associations were statistically significant. There was no evidence of an association with caesarean section (ARR 0.85, 0.45–1.60), but this was retained in the model because of an  a priori  hypothesis of its influence. While a modest association between dyspareunia (1.41, 0.79–2.54) and OPP was found, we did not detect any between dysmenorrhoea and OPP (unadjusted RR 0.97, 0.62–1.51).\nModels for the three pain types showed no effect when adjusted for European ethnicity (ARRs varied by <5% compared with models without ethnicity, data not shown).\n\nPelvic pain was common (54.5%, 95% CI 49.7–59.3%) and often had a substantial impact, being severe or frequent. Despite age-related decline in dysmenorrhoea, almost half (46.2%) of women had been affected in the past 12 months at age 38. Dyspareunia was experienced by 11.6% and OPP by 17.3% at age 38. While there was a high prevalence of pain in this cohort of women and detailed histories available, attempts to identify associations from the potential explanatory factors were mostly inconclusive and CIs often very wide. Past pregnancy was protective for dyspareunia (ARR 0.60) as was ever giving birth for OPP (0.52). Endometriosis was associated with both dysmenorrhoea and dyspareunia (1.60 and 1.8 1, respectively). Psychosocial factors were only associated with dyspareunia. Dysmenorrhoea was strongly associated with dyspareunia (4.54), but not with OPP.\nThe DMHDS cohort is well supported by the participants, with continually high retention of survivors and little likelihood of a selection bias related to pelvic pain. Compared with cross-sectional studies, this study also has the advantage of having regularly interviewed participants, minimising recall issues. However, a larger sample size would be required to detect more modest associations, provide more confidence in the findings, and include more explanatory factors in adjusted models, especially for dyspareunia and OPP which were less common. This also limited our ability to directly investigate potential determinants of pain that were uncommon in this cohort of women, such as pelvic inflammatory disease (PID). As 29 women excluded from analysis of dysmenorrhoea had not menstruated due to menstrual suppression with hormones or due to a hysterectomy, the prevalence of dysmenorrhoea may have been slightly under-estimated. While the duration of OPP was not directly measured, the impact is not likely to be substantial; we established that three-quarters of women with OPP met the duration criteria for chronic pelvic pain. However, for all forms of pelvic pain, it is important to consider when comparing variability between studies that the measurement tools for are diverse and not validated. 7  Endometriosis and fibroids were self-reported diagnoses, and this could have introduced a diagnostic bias. Information about back pain was not sought, and since pelvic pain may present solely as back pain, 19  the prevalence of dysmenorrhoea and OPP may have been underestimated.\nThe 12-month prevalence estimates for dysmenorrhoea (46.2%), dyspareunia (11.6%) and OPP (17.3%) in this cohort of women were lower than those reported in a New Zealand survey which found three-month prevalence among 34–40 year olds to be 57.5%, 22.3% and 28.3% (for chronic pelvic pain), respectively. 5  That survey’s questionnaire was based on the Oxfordshire study, 4  and its prevalence estimates for the United Kingdom and New Zealand were very similar. An Australian-based telephone survey found higher rates of dysmenorrhoea (73.7%) and more moderate rates of dyspareunia (16.5%) and chronic pelvic pain (22.5%) among 30–39 year-olds. 20  These surveys had response rates from 57–74%; thus, selective participation may account for some differences between these studies and our cohort study. However, our study reinforces that all types of pelvic pain are common in women in their late thirties.\nPotential causes of dysmenorrhoea identified in previous studies include smoking, earlier menarche, PID, and psychosocial factors. 7 ,  10  We did not have sufficient data on PID, but had robust data on diagnosed and undiagnosed chlamydia which we used as a marker of PID. Despite longitudinally collected data on these factors, and also on dysmenorrhoea history at ages 13 and 15 years, we did not find any of these factors to be significantly associated with dysmenorrhoea in women at age 38. Nor did we find evidence for an overall protective effect of childbirth as has been found in a number of previous studies. 2 ,  7  However, our data suggested a protective effect for more recent childbirth; it is plausible that any protective effect is not long-lived, or not as important as the effect of ageing in women in their late thirties. It has also been observed that there has been significant publication bias towards studies showing a protective effect for childbirth. 7 ,  15  Jansen  et al.  (2013) found about two-thirds of adolescent girls with chronic pelvic pain or dysmenorrhea had laparoscopic evidence of endometriosis, 21  and we found a strong association between diagnosed endometriosis and dysmenorrhea at age 38.\nPID, childhood sexual abuse, and depression have been implicated in dyspareunia. 7  While, we were unable to confirm any association with these factors, we did find, as described in previous studies, rape and anxiety were strongly associated with increased dyspareunia prevalence. The data were suggestive of an association with PID in that past chlamydia infection had an elevated, but non-significant, ARR. We found past pregnancy was associated with less dyspareunia, which, to our knowledge, has not been examined in previous studies. The direction or mechanism of the association is unknown. The strong association between dysmenorrhea and dyspareunia is not surprising since they share the same central afferent neurons and will be affected by any condition that increases pain sensitivity. 11 ,  13\nPossible factors associated with the development of chronic pelvic pain such as psychosocial factors and dysmenorrhoea were not found to be associated with OPP in our study despite frequent discussion in the literature. 6 ,  7 ,  22  Mental health disorders other than anxiety and depression may be associated with OPP (and other pain types), but these were not explored in this current study. However, an association, although non-significant, with poor sleep supports its potential contribution to the neurobiology of pain. 12  We found modestly elevated non-significant ARRs for OPP with past chlamydia infection, endometriosis and dyspareunia. Miscarriage, caesarean section and other pelvic pathology have also been associated with chronic pelvic pain in previous studies; 7  but the only strong association we found for OPP was a protective effect for childbirth. We are not aware of any long-term studies of chronic pelvic pain after childbirth. Post-surgical pain is prevalent in 10–22% of women up to one year after caesarean section. 23 ,  24  We excluded women who had been pregnant in the past 12 months, however, our findings suggest that nerve injury following caesarean section is not a significant long-term risk for OPP.\n\nThis longitudinal study provides robust prevalence estimates of dysmenorrhoea and other types of pelvic pain experienced by women in their late thirties, and is less affected by bias from selective participation than cross-sectional studies. The study confirms that pelvic pain is common; however, it is a complex condition which requires further study in order to elucidate the causes, especially in the absence of recognised organic pathology. Despite the high prevalence of pain and detailed longitudinal follow up, we did not find a substantial protective effect of childbirth for dysmenorrhoea. However, obstetric factors were unexpectedly protective for other types of pain; further research is needed to understand these mechanisms, as obstetric events and pelvic pain are both important and common experiences. Future research needs to be carefully designed to minimise selection bias from non-representative samples and information bias due to women with pain being more likely to be investigated and receive a diagnosis of one of the conditions being examined as a potential determinant of pain.","source_license":"CC0","license_restricted":false}