The natural history of primary dysmenorrhoea: a longitudinal study

Transcription

1 BJOG: an International Journal of Obstetrics and Gynaecology April 2004, Vol. 111, pp DOI: /j x The natural history of primary dysmenorrhoea: a longitudinal study Alicia M. Weissman, Arthur J. Hartz*, Michael D. Hansen, Susan R. Johnson Objective To describe the prevalence, severity, course and predictive factors of primary dysmenorrhoea in women of all reproductive ages. Design Prospective mailed surveys in 1985 and Setting University of Iowa, College of Nursing. Population We began with a stratified sample of 996 nurses who graduated between 1963 and We analysed data from 404 women who responded to both surveys, but denied endometriosis, pelvic inflammatory disease or uterine fibroids. Methods Participants were surveyed twice at an interval of six years (response rates 73% and 78%) regarding menstrual cycle characteristics. For analysis, dysmenorrhoea was dichotomised as none/mild or moderate/ severe. We analysed predictive factors using m 2 tests and stepwise logistic regression. Main outcome measure Severity of dysmenorrhoea. Menstrual cramps as experienced when not taking medication to prevent discomfort were rated on a four-point scale: 0 ¼ no dysmenorrhoea, 1 ¼ minimal (can work, somewhat uncomfortable), 2 ¼ moderate (can work, but quite uncomfortable) or 3 ¼ severe dysmenorrhoea (miss work, have to be in bed). Results In 1985, 80% of respondents were >25 years old and 60% were parous. There were few changes over six years in the prevalence of mild (51% to 53%), moderate (22% to 20%) or severe dysmenorrhoea (4% to 2%). After adjusting for dysmenorrhoea in 1985, each live birth during follow up (OR ¼ 0.20, 95% CI ¼ 0.08 to 0.53) and older age (OR ¼ 0.92, 95% CI ¼ 0.86 to 0.98) were associated with less dysmenorrhoea in Conclusions Primary dysmenorrhoea affects most women throughout the menstrual years. Dysmenorrhoea severe enough to cause absence from work occurs in less than 5% of women. Although improvement and worsening are equally likely for all women, improvement is more likely in women who bear children. INTRODUCTION Dysmenorrhoea is a common condition that occurs in 52%, 1,2 72% 3 5 or even 90% of women. 6,7 Previous studies have found high rates of absenteeism from work and school due to dysmenorrhoea, with 13 51% of women ever absent 3 5,8 11 and 5 14% frequently absent. 3,4,8 Although the majority of women experience dysmenorrhoea at some time, data on the natural history of primary dysmenorrhoea over the reproductive life span are lacking. Only two prospective studies, both focussing on women in their teens and early 20s, have examined the natural history of dysmenorrhoea. 4,5 The purposes of the present study were to describe the prevalence and severity of primary dysmenorrhoea in a large cohort of women, over half of whom were over 30 years of age at the time of the baseline assessment; College of Medicine, University of Iowa, Iowa City, Iowa 52242, USA * Correspondence: Dr A. J. Hartz, Department of Family Medicine, College of Medicine, University of Iowa, 200 Hawkins Drive, D PFP, Iowa City, Iowa 52242, USA. D RCOG 2004 BJOG: an International Journal of Obstetrics and Gynaecology to examine the longitudinal course of this symptom using a second evaluation six years later; and to identify factors associated with a change in symptoms. METHODS We performed two mail surveys six years apart with a stratified sample of female students and graduates of the University of Iowa College of Nursing. The study was approved by the University of Iowa Human Subjects Committee, and all participants gave informed consent. The 1985 questionnaire consisted of a 46-item survey eliciting information on menstrual, gynaecological, obstetric and socio-demographic variables. The 1991 questionnaire consisted of a 12-page, 57-item survey of menstrual, gynaecological, obstetric and socio-demographic variables, as well as questions about emotional response to pain and illness. Both questionnaires used identical questions, which asked respondents to describe the menstrual cramps usually experienced during the past 12 months when not taking medication to prevent the discomfort. A four-point scale was used: 0 ¼ no dysmenorrhoea, 1 ¼ minimal (can work,

2 346 A.M. WEISSMAN ET AL. Fig. 1. Study population. somewhat uncomfortable), 2 ¼ moderate (can work, but quite uncomfortable), or 3 ¼ severe (miss work, have to be in bed). Women were asked to report whether endometriosis, pelvic inflammatory disease, uterine fibroids, polycystic ovarian disease, any form of cancer or several other specific named medical conditions had ever been diagnosed. To indicate the ways they coped with their symptoms during the previous five years, women were asked to check any of several options (talking with relatives or friends, changed diet, exercise to help symptoms, vitamins, non-prescription medication, prescription medication, none of the above), or to describe what other coping method they used. They were asked to list any medication used for menstrual cramps during the previous five years, and any medication regularly taken for any reason during the previous six months. Information on smoking was collected in 1991 only. For the purposes of analysis, we assumed that women smoking in 1991 were also smoking in As previously described, 12 the original sample consisted of 996 women stratified by year of graduation and included all female graduates of the College of Nursing from the classes of 1963, 1964, 1965, 1969, 1974, and 1979, as well as all female sophomore and senior nursing students enrolled in the autumn of In 1991, women who had responded in 1985 were re-surveyed. For the current analysis, we studied women who responded to the questions about dysmenorrhoea in both 1985 and Because we intended to study primary

3 NATURAL HISTORY OF PRIMARY DYSMENORRHOEA 347 dysmenorrhoea, we excluded from this study the 14% of women who reported endometriosis, pelvic inflammatory disease or uterine fibroids (all possible causes of secondary dysmenorrhoea) in either 1985 or To make the analysis easier to interpret, the four categories of the severity of dysmenorrhoea were collapsed into two: none/mild or moderate/severe. Moderate and severe dysmenorrhoea would be expected to interfere with efficiency or attendance at work or school, while mild or no dysmenorrhoea would not. The statistical significance of variation in the prevalence of moderate/severe dysmenorrhoea across categories was determined by the m 2 test without correction for continuity. The association of ordinal characteristics with dysmenorrhoea (e.g. parity) was assessed with the Wilcoxon Rank Sum test. To identify the characteristics of the women that were independently associated with moderate/severe dysmenorrhoea, we used stepwise logistic regression analysis. Characteristics associated with worsening of dysmenorrhoea from none/mild to moderate/severe between 1985 and 1991 were identified with stepwise logistic regression analysis. Although age and parity are described discretely in some tables, they were modelled continuously in the logistic regression. RESULTS Response rates to the surveys were 73% in 1985 and 78% in 1991 (Fig. 1). The educational level of the group was uniform and predictably high: 92% of the initial 996 women surveyed had a Bachelor of Science degree in nursing (B.S.N.): of these, 20% had a degree beyond the B.S.N. The racial distribution reflected that of the state of Iowa, which is predominantly white. Our primary dysmenorrhoea cohort (Group B in Fig. 1) consisted of 404 women who provided information about dysmenorrhoea on both questionnaires, and did not report endometriosis, pelvic inflammatory disease or uterine fibroids in either survey. Women who responded to the survey but did not complete the questions on dysmenorrhoea were not menstruating due to menopause, hysterectomy, lactation or other reasons. Only one menstruating respondent, who was taking oral contraceptives, did not provide information about dysmenorrhoea in The primary dysmenorrhoea cohort was similar to the initial group who responded to questions on dysmenorrhoea in 1985 (Group A in Fig. 1) in all respects except age (Table 1). The difference in age arose because women who were excluded from the primary dysmenorrhoea cohort due to endometriosis, pelvic inflammatory disease or uterine fibroids tended to be older. The prevalence of mild, moderate, severe or no dysmenorrhoea in 1985 and in 1991 is given in Table 2a. The prevalence of moderate/severe dysmenorrhoea by age is depicted in Fig. 2. There was no consistent change in the prevalence of moderate/severe dysmenorrhoea from 1985 Table 1. Characteristics of the sample of women in Values are expressed as n (%), unless otherwise indicated. Participant characteristic Initial respondents to dysmenorrhoea questions (Fig. 1, Group A), N ¼ 705 Primary dysmenorrhoea cohort (Fig. 1, Group B), N ¼ 404 Age < (19) 78 (19) (36) 175 (43) > (45) 151 (37) Race White 695 (98.6) 401 (99.5) Marital status Single 196 (28) 115 (28) Married/living with a 452 (64) 265 (66) male partner Separated/divorced/widowed 56 (8) 23 (6) Student status Not a student 469 (66) 266 (66) Student part- or full-time 236 (33) 138 (34) Employment status Employed part-time 216 (31) 126 (31) Employed full-time 301 (43) 170 (42) Unemployed 119 (17) 71 (18) Retired 8 (1) 3 (1) Other 58 (8) 31 (8) Age at menarche N Mean (SD) 12.6 (1.4) 12.6 (1.3) Median Min max Parity (39) 152 (40) (43) 156 (41) 3þ 115 (17) 73 (19) Oral contraceptives in last year Yes 109 (16) 66 (16) Severity of dysmenorrhoea None 165 (23) 103 (26) Mild 360 (51) 213 (53) Moderate 153 (22) 80 (20) Severe 27 (4) 8 (2) to Although in 1985 moderate/severe dysmenorrhoea was more prevalent in the cohort compared with the cohort, the younger cohort did not report a significant decrease in the occurrence of moderate/severe dysmenorrhoea over the next six years, as would be expected from the point-prevalence data in Figure 3 shows how the severity of dysmenorrhoea changed for individual women between 1985 and Eighty-eight percent of those who reported dysmenorrhoea in 1985 still reported dysmenorrhoea in Sixty-one percent of those with dysmenorrhoea in 1985 reported

4 348 A.M. WEISSMAN ET AL. Table 2. Reported severity of dysmenorrhoea in the primary dysmenorrhoea cohort 1985 by age and by parity. Values are expressed as n (%). (a) Age in 1985 Year surveyed 1985 (n ¼ 404) 1991 (n ¼ 404) None 9 (12) 9 (12) Mild 41 (53) 39 (50) Moderate 22 (28) 27 (35) Severe 6 (8) 3 (4) None 15 (20) 16 (21) Mild 44 (58) 48 (63) Moderate 16 (21) 12 (16) Severe 1 (1) None 23 (23) 20 (20) Mild 58 (59) 55 (56) Moderate 18 (18) 22 (22) Severe 0 2 (2) None 20 (35) 16 (29) Mild 29 (52) 31 (55) Moderate 7 (13) 8 (14) Severe 0 1 (2) None 33 (36) 34 (37) Mild 40 (44) 42 (46) Moderate 17 (19) 14 (15) Severe 1 (1) 1 (1) 45 and over None 3 (75) 3 (75) Mild 1 (25) 1 (25) Moderate 0 0 Severe 0 0 Total cohort None 103 (26) 98 (24) Mild 213 (53) 216 (53) Moderate 80 (20) 83 (21) Severe 8 (2) 7 (2) (b) Parity in 1985 Year surveyed 1985 (n ¼ 381) 1991 (n ¼ 400) 0 None 22 (14) 12 (12) Mild 83 (55) 48 (47) Moderate 40 (26) 37 (36) Severe 7 (5) 6 (6) Fig. 2. Prevalence of moderate/severe dysmenorrhoea by age, no change in severity in Improvement was reported by 26% of those who initially had mild, moderate or severe dysmenorrhoea, and worsening was reported by 27% of those who initially had mild, moderate or severe dysmenorrhoea. We examined the relationship between the severity of dysmenorrhoea and possible predictive factors. First, we looked at the effect of predictors from on the severity of dysmenorrhoea. We included use of tobacco as a predictor variable: 10% of the respondents smoked in Older age and higher parity (see Table 2a and b) were associated with less severe dysmenorrhoea, while use of tobacco and diabetes were associated with more severe dysmenorrhoea (Table 3). We asked women whether they used non-steroidal antiinflammatory agents, exercise or vitamins to treat dysmenorrhoea. Non-steroidal use (OR ¼ 8.43, P < 0.001) and exercise (OR ¼ 2.53, P < 0.001) were associated with greater dysmenorrhoea severity in 1985, while vitamin use was not. In stepwise multivariate logistic regression (Table 4), tobacco use, parity and age remained significantly associated with dysmenorrhoea severity, but diabetes was eliminated from the model. Thyroid disease was now marginally associated with dysmenorrhoea severity. Nonsteroidal use and exercise were included in the multivariate analyses described below but did not remain statistically significant, indicating that they did not provide more information about dysmenorrhoea severity than was available from other sources. 1 2 None 50 (32) 49 (24) Mild 88 (56) 125 (61) Moderate 17 (11) 29 (14) Severe 1 (1) 1 (0) 3þ None 30 (41) 36 (39) Mild 32 (44) 41 (44) Moderate 11 (15) 16 (17) Severe 0 0 Fig. 3. Severity of dysmenorrhoea in 1991, according to severity in 1985.

5 NATURAL HISTORY OF PRIMARY DYSMENORRHOEA 349 Table 3. Univariate relationship of moderate/severe dysmenorrhoea in 1985 and Proportion with dysmenorrhoea, n/n (%) Odds ratio (95% CI) (a) Patient characteristics in 1985 Age y <25 28/78 (36) 2.24 (1.24 to 4.05) /175 (20) * /151(16) 0.76 (0.43 to 1.34) Parity z 0 47/152 (31) 2.52 (1.22 to 5.22) /156 (12) 0.74 (0.33 to 1.65) 3þ 11/73 (15) * Tobacco use in 1991** Yes 13/39 (33.3) 1.93 (0.95 to 3.94) No 75/365 (20.6) * Tubal ligation Yes 10/62 (16.1) 0.65 (0.31 to 1.34) No 78/341 (22.9) * Regular menstruation Yes 68/328 (20.7) 0.69 (0.39 to 1.24) No 20/73 (27.4) * Oral contraceptives in last year Yes 11/66 (16.7) 0.68 (0.34 to 1.35) No 77/337 (22.9) * Polycystic ovarian disease Yes 0/6 (0.0) NC No 80/365(21.9) * Non-cancerous cyst Yes 3/13 (23.1) 1.09 (0.29 to 4.08) No 77/358 (21.5) * Thyroid disease Yes 5/16 (31.3) 1.70 (0.57 to 5.03) No 75/355 (21.1) * Migraine Yes 5/28 (17.9) 0.78 (0.29 to 2.11) No 75/343 (21.9) * Diabetes y Yes 2/2 (100) NC No 78/369 (21.1) Asthma Yes 3/14 (21.4) 1.00 (0.27 to 3.69) No 76/356 (21.4) * Other current medical problems Yes 21/71 (29.9) 1.65 (0.93 to 2.94) No 67/331 (20.2) All participants 22 (b) Patient characteristics in 1991 Age y /135 (25.9) * /162 (21.0) 0.76 (0.44 to 1.30) /87 (16.1) 0.55 (0.28 to 1.09) Parity z 0 43/103 (41.8) 3.45 (1.77 to 6.71) /204 (14.7) 0.83 (0.43 to 1.61) 3þ 16/93 (17.2) * Tobacco use Yes 16/39 (41) 2.43 (1.23 to 4.79) No 90/404 (22.3) * Tubal ligation Yes 13/87 (14.9) 0.65 (0.34 to 1.24) No 75/354 (21.2) * Table 3. (continued) Proportion with dysmenorrhoea, n/n (%) Odds ratio (95% CI) Regular menstruation Yes 241/310 (77.7) 1.42 (0.85 to 2.45) No 63/89 (70.8) Oral contraceptives in last year Yes 12/62 (19.4) 0.84 (0.43 to 1.66) No 76/342 (22.2) * Polycystic ovarian disease Yes 1/6 (16.7) 0.70 (0.08 to 6.09) Non-cancerous cyst Yes 8/23 (34.8) 1.87 (0.77 to 4.56) Thyroid disease Yes 6/25 (24) 1.11 (0.43 to 2.86) Migraine Yes 10/43 (23.3) 1.06 (0.50 to 2.25) Diabetes y Yes 2/7 (28.6) 1.40 (0.27 to 7.36) Asthma Yes 4/22 (18.2) 0.78 (0.26 to 2.37) Other current medical problems * All participants * Tobacco used in 1991 was used as a proxy for tobacco use in NC ¼ could not be calculated. * Reference group for odds ratio calculations. ** P ¼ y P < z P < Because none of the women reporting severe dysmenorrhoea in 1985 improved to mild or no dysmenorrhoea in 1991, we could not examine risk factors associated with improvement in these women. Therefore, these women were excluded when analysing the effect of 1985 predictors on the severity of dysmenorrhoea in In univariate analysis, we found that older age ( P < 0.05), higher parity ( P < 0.10) and live birth after 1985 ( P < 0.05) decreased the odds of worsening dysmenorrhoea, while tobacco use Table 4. Variables from 1985 associated with moderate/severe dysmenorrhoea in 1985 (multivariate analysis) Variable Odds of moderate or severe dysmenorrhoea in 1985 (95% CI), n ¼ 285 Tobacco use (1.00 to 5.51) Thyroid disease 3.24 (0.95 to 11.06) Parity y 0.70 (0.54 to 0.91) Age y 0.94 (0.90 to 0.99) y Parity and age are modeled in increments of 1 birth for parity and 1 year for age.

6 350 A.M. WEISSMAN ET AL. Table 5. Variables from 1985 independently associated with moderate/ severe dysmenorrhoea in 1991, after excluding women reporting severe dysmenorrhoea in 1985 (multivariate analysis) Variable Odds ratios (95% CI) of progression to moderate or severe dysmenorrhoea in 1991, n ¼ 218 Presence of dysmenorrhoea in (3.09 to 18.15) Live birth after (0.08 to 0.53) Age in (0.86 to 0.98) Parity and age are modeled in increments of 1 birth for parity and 1 year for age. (OR ¼ 2.17, P < 0.05) and diabetes (OR ¼ not calculable, P < 0.05) increased the odds of worsening dysmenorrhoea. We then examined the effect of the predictor variables on the severity of dysmenorrhoea in 1991 using stepwise logistic regression, adjusting for dysmenorrhoea severity in 1985 (Table 5). A live birth during the interval and older age were significantly associated with less severe dysmenorrhoea. Some factors that had been statistically significant in the univariate analysis (i.e. parity, tobacco use and thyroid disease) were not independently associated with the severity of dysmenorrhoea. For the 80 women with moderate dysmenorrhoea in 1985, an increase in parity of 1 in the logistic regression analysis was associated with an improvement in dysmenorrhoea (OR ¼ 1.383, P ¼.08, CI ¼ to 1.994). DISCUSSION In this group of women representing a broad age spectrum, primary dysmenorrhoea affected the majority of women and was found in all age groups. Dysmenorrhoea persisted over the six years between the surveys. Improvement and worsening were equally likely. However, severe dysmenorrhoea, requiring absence from work or school, was rare. Non-steroidal anti-inflammatory agents were the most commonly used treatment for dysmenorrhoea. Oral contraception was not significantly associated with less severe dysmenorrhea probably because women were asked to rate the severity of dysmenorrhoea when not using medication. Previous studies have found that dysmenorrhoea causes frequent absence from work in 5 14% 3,4,8 of women, and absence from school in 14 46% of adolescents. 11,13,14 Our participants reported a very low prevalence of severe dysmenorrhoea (2%), which was defined as dysmenorrhoea severe enough to warrant staying home or in bed. We asked women to average their symptoms over the past year and did not attempt to quantify absence from work, which may have reduced the reported frequency of absence. In additional, 74% or our sample of women was employed part- or full-time, which may have reduced the likelihood of staying in bed or missing work. The prevalence of severe dysmenorrhoea was not low because we excluded women with underlying gynaecological conditions that could cause secondary dysmenorrhoea: note that women in Group A and in Group B had a similar prevalence of severe dysmenorrhoea. Another possible explanation is that these women were able to obtain appropriate treatment and avoid absence from work due to dysmenorrhoea. However, our survey asked women to rate their dysmenorrhoea when not using medication such as non-steroidal anti-inflammatory agents, and so other factors may be at work. In addition, use of nonsteroidal drugs was entered into the multivariate models, which should have eliminated confounding from this source. Absence from work and decreased productivity due to dysmenorrhoea remains poorly quantified throughout life, and should be the subject of future research. Some studies have found that dysmenorrhoea decreases in severity after bearing a child, 2,4 but others found that this relationship disappeared after controlling for age 15 or age and use of tobacco. 1 Among women with no or mild dysmenorrhoea in 1985, subsequent live birth was most protective against progression to moderate or severe dysmenorrhea. For women with moderate dysmenorrhea in 1985, subsequent live births were associated with reduced dysmenorrhea in 1991 ( P ¼.08). Because longitudinal data eliminate the confounding factor of age and provide much stronger evidence for cause and effect than cross sectional data, we believe that our study strongly supports the association between parity and decreased severity of dysmenorrhoea. In our analysis, gravidity was consistently less significant than parity (data not shown), confirming previous work that showed no effect from pregnancies ending in miscarriage or abortion. 4 A longitudinal study of young women by Sundell et al. 4 showed a decrease in the prevalence of dysmenorrhoea between the ages of 19 and 24, from 72% to 67%. The severity of dysmenorrhoea also decreased over time. However, these decreases occurred only in women who had children during those five years, and was unchanged in women who remained nulliparous, had a miscarriage or had an abortion. 4 After adjusting for the severity of dysmenorrhoea in 1985 and change in parity, we still found a significant effect of age on the severity of dysmenorrhoea in 1991 (Table 5), indicating that older women are more likely to experience a decrease in the severity of primary dysmenorrhoea, independent of childbearing. However, childbearing was clearly the more influential factor in our analysis. Previous studies found that earlier age at menarche was associated with a higher incidence of dysmenorrhoea. 5,16 We did not find a significant relationship between dysmenorrhoea and age at menarche, possibly because the age of menarche was very similar across our sample. Although one previous study found less dysmenorrhoea in smokers, 3 others have found that smoking increases the incidence of dysmenorrhoea, 17 the severity of dysmenorrhoea, 4 the heaviness of menstrual bleeding and the duration

7 NATURAL HISTORY OF PRIMARY DYSMENORRHOEA 351 of dysmenorrhoea. 18 Environmental smoke exposure also increases the severity of dysmenorrhoea. 19 Although Harlow and Park 5 did not find an increase in the incidence or severity of dysmenorrhoea with smoking, smoking did increase the likelihood of having dysmenorrhoea lasting more than two days. In multivariate analysis, we found that use of tobacco in 1991 (used as a proxy for use of tobacco in 1985) was associated with a greater severity of dysmenorrhoea in 1985, but did not affect the subsequent progression of dysmenorrhoea from none/mild to moderate/severe after taking into account additional births. We did not collect data on the number of cigarettes/day smoked, and were unable to assess any dose/response effect of smoking on the severity of dysmenorrhoea. Previous work has found that tubal sterilisation probably does not increase the prevalence of dysmenorrhoea. 20 Our results confirm this finding. An interesting finding in this sample is the possible association between thyroid disease and dysmenorrhoea, which we observed in the 1985 multivariate analysis. However, the type of thyroid disease was not specified on the questionnaire. Future studies of dysmenorrhoea should include hyper- or hypothyroidism as possible predictor variables. The association between diabetes mellitus and dysmenorrhoea that we observed in univariate analysis was based on only two women. Therefore, no conclusions can be drawn from this association. Diabetes did not enter into the multivariate logistic regression model. The study by Harlow and Park 5 of university students age lasted only one year and excluded women who were pregnant or had children. The only previous work similar to ours is the report by Sundell et al. 4 which used a similar pain scale to assess four hundred and sixty 19 year old women over five years. They observed moderate dysmenorrhoea in 22 23% of women, similar to our sample, and severe dysmenorrhoea in 10 15%, a much higher prevalence than in our sample. They also found that dysmenorrhoea decreased after live birth but was unchanged in women who remained nulliparous or did not carry a pregnancy to term. Our study extends this finding to older women and shows that the proportion of women aged with moderate or severe dysmenorrhoea remains relatively consistent over time, both longitudinally and between age cohorts (Fig. 2). Since women were asked to describe their symptoms over the previous year, there may be errors in recall. The intensity of dysmenorrhoea is under-estimated after some time has passed since experiencing the pain. 21 However, our definition of dysmenorrhoea included its effect on concrete activities (attendance and performance at work or school) which would tend to improve recall. Diaries would be a more accurate way to quantify dysmenorrhoea over time. Our survey did not evaluate additional factors that appear to ameliorate dysmenorrhoea, (such as calcium intake, 22 physical activity 23 or dietary factors 16 ) or intensify dysmenorrhoea (such as affective symptoms and social support 24 ). It is possible that some of these factors could be unmeasured confounders. Although we excluded women with potential causes of secondary dysmenorrhoea, it was not possible to discriminate definitively between primary and secondary dysmenorrhoea in our sample. Traditionally, it is thought that women over 30 years of age are more likely to have secondary dysmenorrhoea. It may be that women reporting worsening symptoms over time had developed undiagnosed secondary dysmenorrhoea. These nurses were presumably reliable reporters, with greater access to the latest information on treatment compared with women in the general population. However, it is possible that even the nurses were not adequately investigated for underlying pathology. The results may be affected by selection bias. We cannot evaluate the non-response bias in However, we did determine that the severity of dysmennorrhea did not differ between the non-respondents and respondents in the 1991 survey (Table 1). The primary purpose of the study was not to estimate the rates of dysmenorrhoea but to evaluate the association between dysmenorrhoea and certain characteristics of the women. These associations should be little influenced by the non-response rate in this study. Our sample was very homogeneous with respect to education, race and socio-economic status, which limits cultural confounding factors, but also limits the external validity of our results. Other studies have found that socioeconomic status is inversely associated with dysmenorrhoea. 25 We did not study adolescents: The prevalence of dysmenorrhoea 25,26 and absenteeism 6,25 may well be higher in that age group. Dysmenorrhoea is a significant symptom for women throughout the menstrual years, rather than a self-limiting disorder confined to teenagers. We have confirmed some traditional beliefs about dysmenorrhoea, such as improvement with age and childbearing. However, this improvement is by no means universal, and many women experience unchanged or worse symptoms over time. We found that dysmenorrhoea severe enough to cause absence from work occurred in about 2% of women, a much smaller proportion than previously reported. References 1. Pullon S, Reinken J, Sparrow M. Prevalence of dysmenorrhea in Wellington women. N Z Med J 1988;101(839): Ng T, Tan N, Wansaicheong G. A prevalence study of dysmenorrhea in female residents aged years in Clementi Town, Singapore. Ann Acad Med Singap 1992;21(3): Andersch B, Milsom I. An epidemiologic study of young women with dysmenorrhea. Am J Obstet Gynecol 1982;144(6): Sundell G, Milsom I, Andersch B. Factors influencing the prevalence and severity of dysmenorrhea in young women. Br J Obstet Gynaecol 1990;97(7): Harlow S, Park M. A longitudinal study of risk factors for the occurrence, duration and severity of menstrual cramps in a cohort of college women. Br J Obstet Gynaecol 1996;103(11):

8 352 A.M. WEISSMAN ET AL. 6. Wilson C, Keye WJ. A survey of adolescent dysmenorrhea and premenstrual symptom frequency. A model program for prevention, detection, and treatment. J Adolesc Health Care 1989;10(4): Jamieson D, Steege J. The prevalence of dysmenorrhea, dyspareunia, pelvic pain, and irritable bowel syndrome in primary care practices. Obstet Gynecol 1996;87(1): Busch CM, Costa PT, Whitehead WE, Heller BR. Severe perimenstrual symptoms: prevalence and effects on absenteeism and health care seeking in a non-clinical sample. Women Health 1988;14(1): Gruber VA, Wildman BG. The impact of dysmenorrhea on daily activities. Behav Res Ther 1987;25(2): Mergler D, Vezina N. Dysmenorrhea and cold exposure. J Reprod Med 1985;30(2): Banikarim C, Chacko MR, Kelder SH. Prevalence and impact of dysmenorrhea on Hispanic female adolescents. Arch Pediatr Adolesc Med 2000;154(12): Johnson SR, McChesney C, Bean JA. Epidemiology of premenstrual symptoms in a nonclinical sample. J Reprod Med 1988;33(4): Johnson J. Level of knowledge among adolescent girls regarding effective treatment for dysmenorrhea. J Adolesc Health Care 1988;9(5): Klein JR, Litt IF. Epidemiology of adolescent dysmenorrhea. Pediatrics 1981;68(5): Messing K, Saurel-Cubizolles M-J, Bourgine M, Kaminski M. Factors associated with dysmenorrhea among workers in French poultry slaughterhouses and canneries. J Occup Med 1993;35(5): Balbi C, Musone R, Menditto A, et al. Influence of menstrual factors and dietary habits on menstrual pain in adolescence age. Eur J Obstet Gynaecol Reprod Biol 2000;91(2): Parazzini F, Tozzi L, Mezzopane R, Luchini L, Marchini M, Fedele L. Cigarette smoking, alcohol consumption, and risk of primary dysmenorrhea. Epidemiology 1994;5(4): Hornsby P, Wilcox A, Weinberg C. Cigarette smoking and disturbance of menstrual function. Epidemiology 1998;9(2): Chen C, Cho SI, Damokosh AI, et al. Prospective study of exposure to environmental tobacco smoke and dysmenorrhea. Environ Health Perspect 2000;108(11): Rulin M, Davidson A, Philliber S, Graves W, Cushman L. Long-term effect of tubal sterilization on menstrual indices and pelvic pain. Am J Obstet Gynecol 1993;82: Brodie EE, Niven CA. Remembering an everyday pain: the role of knowledge and experience in the recall of the quality of dysmenorrhoea. Pain 2000;84(1): Thys-Jacobs S, Starkey P, Bernstein D, Tian J. Calcium carbonate and the premenstrual syndrome effects on premenstrual and menstrual symptoms. Am J Obstet Gynecol 1998;179(2): Izzo I, Labriola D. Dysmenorrhea and sports activities in adolescents. Clin Exp Obstet Gynecol 1991;18(2): Alonso C, Coe CL. Disruptions of social relationships accentuate the association between emotional distress and menstrual pain in young women. Health Psychol 2001;20(6): Hillen T, Grbavac S, Johnston P, Straton J, Keogh J. Primary dysmenorrhea in young western Australian women: prevalence, impact, and knowledge of treatment. J Adolesc Health 1999;25(1): Robinson J, Plichta S, Weisman C, Nathanson C, Ensminger M. Dysmenorrhea and use of oral contraceptives in adolescent women attending a family planning clinic. Am J Obstet Gynecol 1992;166(2): Accepted 15 October 2003