Severity of premenstrual symptoms in a health maintenance organization population

Severity of premenstrual symptoms in a health maintenance organization population

Severity of Premenstrual Symptoms in a Health Maintenance Organization Population Barbara Sternfeld, PhD, Ralph Swindle, PhD, Anita Chawla, PhD, Stace...

170KB Sizes 0 Downloads 3 Views

Severity of Premenstrual Symptoms in a Health Maintenance Organization Population Barbara Sternfeld, PhD, Ralph Swindle, PhD, Anita Chawla, PhD, Stacey Long, MS, and Sean Kennedy OBJECTIVE: To describe severity of emotional and physical symptoms in a large diverse sample; to examine demographic, health status, and behavioral correlates of symptom severity; and to describe use of medications and alternative remedies for premenstrual symptoms. METHODS: A total of 1194 women, ages 21– 45, selected from members of a large northern California health maintenance organization, completed daily ratings of symptom severity for two menstrual cycles. An empirically derived algorithm defined symptom severity groups as minimal (n ⴝ 186), moderate (n ⴝ 801), severe (n ⴝ 151), or premenstrual dysphoric disorder (n ⴝ 56). Symptom severity as a continuous variable was defined by the two-cycle mean symptom ratings in the luteal phase. Demographic, health status, and behavioral factors and use of treatments for premenstrual symptoms were assessed by self-report. RESULTS: Luteal phase symptom-specific ratings were generally significantly greater in the premenstrual dysphoric disorder group than in the other groups (P < .001). Symptom severity score increased with each comorbidity and decreased with each year of age. Symptom severity was also inversely associated with oral contraceptive use (emotional symptoms) and better perceived health (physical symptoms). Hispanics reported greater severity of symptoms, and Asians less, relative to whites. Use of herbal and nutritional supplements for premenstrual symptoms steadily increased from 10.8% in the minimal group to 30.4% in the premenstrual dysphoric disorder group (P < .01). CONCLUSION: The degree of premenstrual symptom severity varies in the population, is relatively constant within each woman over two consecutive cycles, particularly for emotional symptoms, and is influenced by age, race/ethnicity, and health status. (Obstet Gynecol 2002;99:1014 –24. © 2002 by the American College of Obstetricians and Gynecologists.) From the Division of Research, Kaiser Permanente, Oakland, California; Eli Lilly and Co., Eli Lilly Corporate Center, Indianapolis, Indiana; The MEDSTAT Group, Washington, DC; and Genentech, South San Francisco, California. The authors gratefully acknowledge the efforts of CODA, Inc., and Freeman, Sullivan & Co. for their assistance in planning and implementing recruitment and data collection activities for this study.


Premenstrual symptoms, which include a heterogeneous array of physical and psychologic complaints that occur, or intensify, during the week before the onset of menstrual bleeding,1 may range in severity from mild to disabling. Although estimates of prevalence vary widely,1,2 as many as 75– 80% of women typically experience at least mild symptoms. Fewer women report severe or disabling symptoms; only 3– 8% have premenstrual symptoms that meet the criteria for premenstrual dysphoric disorder (PMDD),3,4 defined as a distinct affective disorder, characterized by the presence of at least five symptoms (one of which must be affective) that occur in the late luteal phase, are not a luteal exacerbation of an existing psychiatric condition, and that significantly interfere with social activities or relationships with others.5,6 Previous epidemiologic work has suggested few consistent biologic, social, or behavioral risk factors for premenstrual symptoms. Younger age1,2,4,7,8 is generally related to increased number and severity of symptoms, although older studies tended to find that symptom severity peaked in the mid-30s.9 Other demographic characteristics, including education, income, employment status, and marital status, have shown inconsistent or no relationships.3,10 Because most studies have been conducted in white women, little is known about racial/ethnic differences. Limited evidence has identified nulliparity,1 earlier age at menarche,4,11 higher body mass index,4,7 greater alcohol2,4,12 and caffeine13 consumption, and higher levels of perceived stress2,4 as risk factors for one or another specific symptom, and oral contraceptive use1,3 and regular physical activity7,14 as protective factors. Many of the inconsistencies in the epidemiologic literature on premenstrual symptoms may arise from a variety of methodologic weaknesses. Small and generally Financial Disclosure This study was funded by Eli Lilly and Company, Indianapolis, Indiana. The financial support from Eli Lilly was specifically used to conduct this study. It supported investigator and research staff salaries for work on the project.

VOL. 99, NO. 6, JUNE 2002 © 2002 by The American College of Obstetricians and Gynecologists. Published by Elsevier Science Inc.

0029-7844/02/$22.00 PII S0029-7844(02)01958-0

nonrepresentative samples, varying case definitions, lack of standardized methods of assessment, and reliance on retrospective reporting of premenstrual symptoms are all common limitations.2,4 Currently, prospective reporting of daily symptoms for two complete menstrual cycles is considered the gold standard for assessing premenstrual symptoms and is required for the clinical diagnosis of PMDD.6 However, only two small studies have used prospective diaries to address the issue of symptom stability from cycle to cycle,15,16 and one15 found considerable symptom stability, whereas the other did not.16 Given these contradictory findings and methodologic weaknesses, prospective studies from large populations are needed to address fundamental issues such as the patterns of and risk factors for differing levels of premenstrual symptom severity. The purpose of this study was: 1) to describe the patterns of emotional and physical symptom severity in a large, generally representative sample; and 2) to examine demographic, health status, and behavioral correlates of prospectively assessed premenstrual symptom severity. A secondary aim was to describe the correlates of use of various types of remedies for premenstrual symptoms. MATERIALS AND METHODS The sampling frame consisted of women members of a large health maintenance organization in northern California who, based on electronic membership files, met the following criteria: ages 21– 45, no psychiatric diagnoses in the past 2 years, no psychotropic medication in the past 6 months, no history of hysterectomy or bilateral oophorectomy, and no pregnancy within the previous year. The names of all women meeting these criteria were ordered randomly and sequentially divided into batches of approximately 5000 each. Further screening (English speaking, within the defined age range at the time of the screening, and willing to participate) was then conducted by telephone, beginning with the first batch and continuing until the desired sample size was attained. Women were excluded if they reported never experiencing premenstrual symptoms (asymptomatic), or if their symptoms did not subside within 2–3 days of menstrual bleeding. Women currently taking psychotropic medications or with probable major depressive disorder, as determined by the Primary Care Evaluation of Mental Disorders, a standardized instrument for diagnosing mental disorders in the primary care setting,17 were also excluded. Of the 17,963 women for whom an attempt to contact was initiated, 9713 (54.1%) were unreachable after ten attempts, 2999 (16.7%) were ineligible, and 2463 (13.7%) refused to participate. To maximize the sample size at the ends of the symptom continuum (ie, those with minimal

VOL. 99, NO. 6, JUNE 2002

symptoms and those with probable PMDD), recruitment of women with moderate-severe symptoms, as determined by a retrospective Daily Record of Severity of Problems18 for the last menstrual cycle, was closed before recruitment of women with either lesser or greater symptom severity. This resulted in exclusion of 472 (2.6%) additional women. The current report is based on the 1194 women who enrolled in the study and completed the study questionnaires, including daily record of severity of problems for two complete menstrual cycles (51.6% of the 2316 enrolled). Approval for this study was obtained from the Northern California Kaiser Permanente Institutional Review Board. The Daily Record of Severity of Problems, developed and validated by Yonkers et al18 consists of 12 emotional symptoms, such as feeling depressed, sad, down or blue, feeling anxious, tense, keyed up or on edge, and feeling angry or irritable, and nine physical symptoms, such as breast tenderness, joint or muscle pain, and increased appetite. The symptoms included in this scale correspond to those specified by the Diagnostic and Statistical Manual of Mental Disorders, fourth edition, for diagnosis of PMDD. We excluded three items that assess functional impairment because they overlapped with study outcomes. On each day, respondents were asked to rate each symptom on a scale of 1 (“not at all”) to 6 (“extreme”). Symptom severity was categorized as minimal, moderate, severe, or PMDD, based on the daily ratings during the fifth through the tenth day of the follicular phase and the first day of menstrual bleeding plus the 5 preceding days in the luteal phase, using the average of the 3 highest days of ratings in each phase in each of the two menstrual cycles. The specific classification of symptom severity groups is described in Table 1. In addition, daily emotional and physical symptom severity scores were defined as continuous variables by computing the two-cycle mean value of the symptom ratings for all relevant symptoms for each day of the cycle. Phasespecific emotional and physical symptom severity scores were defined similarly for each of the 6-day luteal and follicular phase time frames specified above. Age, race/ethnicity, education, marital status, and number of children under age 5 were reported during the telephone screening interview. Standard questions on the baseline survey provided information on employment status, perceived health status, number of comorbidities, current use of oral contraceptives, current smoking status, and participation in strenuous exercise three or more times a week. The baseline questionnaire also asked about remedies for premenstrual symptoms used in the past 12 months, including prescription and over-the-counter medications, complementary and alternative therapies, and

Sternfeld et al

Severity of Premenstrual Symptoms


Table 1. Classification of Symptom Severity Groups Minimal:

No symptoms with ratings of 5 or 6 (severe or extreme) in luteal phase AND average ratings greater in luteal phase than the follicular phase Moderate: At least one symptom rated as 5 or 6 (severe or extreme) in luteal phase in at least one cycle AND average rating for qualifying symptom greater in luteal than follicular phase BUT criteria for PMDD not met for either cycle OR criteria for PMDD met in one cycle BUT no symptom rated as 5 or 6 (severe or extreme) in luteal phase in other cycle Severe: Criteria for PMDD met in one cycle AND in other cycle at least one symptom rated as 5 or 6 (severe or extreme) during luteal phase with average rating for qualifying symptom greater in luteal than follicular phase PMDD: At least five symptoms with average rating ⬎3.0 (mild) in luteal phase, and at least one being an emotional symptom AND for at least five qualifying symptoms, average score in luteal phase 30% greater than average score for corresponding symptom in follicular phase with at least one being an emotional symptom AND average rating for all symptoms in follicular phase ⬍3.0

PMDD ⫽ premenstrual dysphoric disorder.

herbal or nutritional supplements. Use of each of these categories of remedies was dichotomized into any or none. The mean daily emotional symptom score for each symptom severity group was plotted by day of the cycle, with day 0 indicating the first day of menstrual bleeding and assuming a 28-day cycle (ie, plotting only from day ⫺14 to day 14). The same procedure was used to plot mean daily physical symptom scores. Differences among the symptom severity groups in the two mean daily symptom scores on day 0, when symptoms would be high in all groups, and day 8, when symptoms would be low, were compared using analysis of variance with the Tukey test for multiple comparisons. All pair-wise differences among the symptom severity groups in symptomspecific, phase-specific ratings were assessed with nonparametric tests (Wilcoxon rank sum and median test). Based on the Bonferroni adjustment for multiple comparisons, a difference was considered statistically significant if at least one of the two nonparametric tests was significant at P ⬍ .001. Differences in symptom- and phase-specific severity ratings between cycles 1 and 2 were compared in the sample as a whole using paired t tests with the significance level again adjusted by the Bonferroni method and set at P ⬍ .001. All demographic, health status, and lifestyle factors were treated as categoric variables and described by proportions with differences among symptom severity groups evaluated


Sternfeld et al

Severity of Premenstrual Symptoms

by the ␹2 statistic. Separate multivariable linear regression models were fit to examine the independent association of each demographic, health status, and behavioral factor with either emotional or physical luteal phase symptom score. Finally, to explore the relationship between use of remedies for premenstrual symptoms and demographic, health status, and behavioral factors, independently of each other and of symptom severity, three different multivariable logistic regression models were fit with prescription and over-the-counter medication, alternative therapies and herbal or nutritional remedies, each dichotomized as any or none, as the respective outcomes.

RESULTS The sample as a whole experienced a mean increase in daily emotional and physical symptom scores that tended to peak on the day before or the first day of menstrual bleeding (Figures 1 and 2). However, by definition, the values of the symptom scores and the magnitude of change in scores throughout the cycle generally differed by symptom severity group. For instance, on day 0, the mean emotional symptom score in the PMDD group (36.5 ⫾ 18.1) was significantly greater than the mean scores in the other three groups (29.1 ⫾ 16.6, 21.0 ⫾ 13.3, and 15.7 ⫾ 11.4 in the severe, moderate, and minimal groups, respectively, P ⬍ .001). On day 8, the severe group had the highest mean emotional symptom score (19.8 ⫾ 14.3), and only the minimal group differed significantly from the PMDD group (13.4 ⫾ 7.2 versus 18.5 ⫾ 11.1, P ⬍ .001). Similar differences were observed in the mean daily physical symptom scores. The groups also differed, again partly by definition, in terms of the distribution of symptom-specific ratings in the 6-day follicular phase and 6-day luteal phase. As shown in Table 2, during the follicular phase, the median was significantly lower and interquartile range narrower in the minimal group (P ⬍ .001) than all other groups for all symptoms except breast tenderness. The moderate group differed significantly from the severe group for all emotional symptoms, and for all physical symptoms, except increased appetite, sleeping more, and trouble sleeping, but differed from the PMDD group only in feelings of hopelessness. There were no significant differences between the severe and PMDD groups. In the luteal phase, all of the groups differed significantly from each other for all symptoms except for the severe and PMDD groups in which the ratings did not differ for the following symptoms: having conflicts or problems with people, having less interest in usual activities, having


Figure 1. Mean daily emotion symptom score by symptom severity group. PMDD ⫽ premenstrual dysphoric disorder. Sternfeld. Severity of Premenstrual Symptoms. Obstet Gynecol 2002.

trouble getting to sleep, breast tenderness, headaches, and joint or muscle pain. For the sample as a whole, none of the symptomspecific follicular or luteal phase emotional symptom ratings or follicular phase physical symptom ratings differed significantly (P ⬍ .001) between the first and second menstrual cycles. In contrast, four of the nine luteal phase physical symptoms were rated significantly lower in the second cycle: increased appetite or overeating (mean per woman difference, cycle 2-cycle 1 ⫽ ⫺0.03 ⫾ 0.87, P ⬍ .001), food cravings (⫺0.05 ⫾ 0.86, P ⬍ .001), breast tenderness (⫺0.10 ⫾ 1.00, P ⬍ .001), and breast swelling, bloating, weight gain (⫺0.10 ⫾ 1.04, P ⬍ .001). Table 3 presents the demographic, health status, and behavioral characteristics of the sample, overall and by symptom severity group. The sample ranged in age from 21 to 45 and was racially/ethnically diverse. More than half had not graduated from college, and a fifth were not employed. Overall, this was a sample of healthy women. Fewer than 10% rated their health as fair or poor, and more than 50% reported no comorbidity. Only 15% were current smokers. Several of the demographic, health status, and behavioral factors varied by symptom severity group (Table 3). Relative to the age distribution of the sample as a whole, women over 40 were over-represented in the

VOL. 99, NO. 6, JUNE 2002

minimal group; conversely, women between 21 and 30 were over-represented in the group with severe symptoms. The PMDD group had a greater proportion of white, non-Hispanic women, whereas the severe group had a greater proportion of Hispanic women. Asian women were under-represented in both of those groups and over-represented in the minimal group. Also, a greater proportion of women in the severe and PMDD groups reported at least one comorbidity, compared with the other two groups, and although not statistically significant, a greater proportion in these groups rated their health as poor or fair. The proportions of women using over-the-counter and prescription medications for premenstrual symptoms, alternative therapies, and herbal and nutritional supplements were consistently lowest in the minimal group and consistently highest in the PMDD group. The proportions of women in the minimal and moderate groups who currently used oral contraceptives were higher than the proportions in the other symptom groups, but this difference was not statistically significant. Finally, the proportion of women in the minimal group who were current smokers was the lowest of all the groups, whereas the proportion in the severe group was the highest (20.5%), but this was also not statistically significant. There were no differences among the groups in terms of education, employment or marital

Sternfeld et al

Severity of Premenstrual Symptoms


Figure 2. Mean daily physical symptom score by symptom severity group. PMDD ⫽ premenstrual dysphoric disorder. Sternfeld. Severity of Premenstrual Symptoms. Obstet Gynecol 2002.

status, number of children under 5 living in the house, or participation in strenuous exercise. To examine whether any of these associations was confounded by each other, separate multivariable linear regressions were performed with the luteal phase emotional or physical symptom scores as the outcome. As Table 4 shows, age and oral contraceptive use were inversely related to number and severity of emotional symptoms, wheras having a comorbidity and being Hispanic, relative to being white, were directly related. Being Asian, compared with white, was associated with fewer symptoms (P ⫽ .051), but neither smoking nor exercise showed a significant association. For physical symptoms, better health status was inversely related, and the magnitude of the direct association with comorbidity was greater than that observed for emotional symptoms. Because oral contraceptives can either exacerbate or ameliorate premenstrual symptoms, the analysis presented in Table 4 was also conducted excluding the oral contraceptive users (n ⫽ 318). The results were largely unchanged. Age still showed an inverse association with both the emotional (␤ ⫽ ⫺0.118, standard error [SE] ⫽ 0.038, P ⫽ .002) and physical symptom (␤ ⫽ ⫺0.062, standard error [SE] ⫽ 0.029, P ⫽ .04) scores,


Sternfeld et al

Severity of Premenstrual Symptoms

whereas the number of comorbidities continued to be directly associated (␤ ⫽ 1.006, standard error [SE] ⫽ 0.207, P ⬍ .001 for emotional symptoms; ␤ ⫽ 1.179, standard error [SE] ⫽ 0.159, P ⬍ .001 for physical symptoms). Racial/ethnic differences were still evident, with Hispanics having higher symptom scores and Asians lower symptom scores (not significant) relative to whites (data not shown). However, perceived health, which was inversely associated with the physical symptom score in the full sample, failed to show any significant association when oral contraceptive users were excluded. The associations between use of various categories of premenstrual symptom remedies and demographic, health status, and behavioral variables, independently of each other and of symptom severity, are presented in Table 5. For all types of remedies, likelihood of use increased as severity of symptoms increased. Having a comorbidity was associated with increased likelihood of use of prescription and over-the-counter medication, whereas being Asian and participating in regular strenuous physical activity were associated with decreased likelihood. Use of alternative therapies was also associated with having a comorbidity, but regular strenuous


Table 2. Median and Interquartile Range of Symptom-Specific Severity Ratings* Across Symptom Severity Groups by Phase of Menstrual Cycle† Follicular phase

Luteal phase

Minimal Moderate Severe PMDD Minimal Moderate Severe PMDD (n ⫽ 186) (n ⫽ 801) (n ⫽ 151) (n ⫽ 56) (n ⫽ 186) (n ⫽ 801) (n ⫽ 151) (n ⫽ 56) Median Median Median Median Median Median Median Median (IQ range) (IQ range) (IQ range) (IQ range) (IQ range) (IQ range) (IQ range) (IQ range) Emotional symptoms Felt depressed, sad, “down” or “blue” Felt hopeless Felt worthless or guilty Felt anxious, tense, “keyed up” or “on edge” Had mood swings (eg, suddenly felt sad or tearful) Was more sensitive to rejection, hurt feeling Felt angry, irritable Had conflicts or problems with people Had less interest in usual activities Had difficulty concentrating Felt overwhelmed or that could not cope Felt out of control Physical symptoms Felt lethargic, tired, fatigued, or lack of energy Had increased appetite or overate Had food cravings Slept more, took naps, found it hard to get up Had trouble getting to sleep or staying asleep Had breast tenderness Had breast swelling, felt “bloated,” weight gain Had headache Had joint or muscle pain

1.0 (0.1)‡§㛳 1.2 (0.6)§

1.5 (1.1)

1.6 (1.0)

1.0 (0.3)‡§㛳 1.4 (0.9)§㛳

2.3 (1.2)㛳

2.9 (1.2)

1.0 (0.0)‡§㛳 1.0 (0.1)§㛳 1.0 (0.0)‡§㛳 1.0 (0.2)§ 1.0 (0.2)‡§㛳 1.3 (0.8)§

1.0 (0.4) 1.1 (0.4) 1.5 (1.0)

1.2 (0.4) 1.1 (0.5) 1.7 (1.0)

1.0 (0.0)‡§㛳 1.0 (0.3)§㛳 1.0 (0.0)‡§㛳 1.0 (0.3)§㛳 1.1 (0.3)‡§㛳 1.6 (1.0)§㛳

1.5 (1.0)㛳 1.5 (1.1)㛳 2.5 (1.3)㛳

2.2 (1.3) 2.3 (0.9) 3.3 (1.2)

1.0 (0.2)‡§㛳 1.2 (0.6)§

1.5 (1.0)

1.4 (0.9)

1.1 (0.3)‡§㛳 1.5 (1.0)§㛳

2.7 (1.3)㛳

3.3 (1.2)

1.0 (0.2)‡§㛳 1.1 (0.4)§

1.5 (1.0)

1.5 (0.8)

1.0 (0.2)‡§㛳 1.3 (0.8)§㛳

2.5 (1.5)㛳

3.1 (1.6)

1.0 (0.2)‡§㛳 1.3 (0.7)§ 1.0 (0.2)‡§㛳 1.1 (0.4)§

1.6 (0.9) 1.4 (0.7)

1.6 (0.9) 1.3 (0.7)

1.2 (0.4)‡§㛳 1.8 (0.9)§㛳 1.0 (0.2)‡§㛳 1.3 (0.7)§㛳

2.7 (1.2)㛳 2.2 (1.0)

3.6 (1.4) 2.5 (1.2)

1.0 (0.2)‡§㛳 1.2 (0.5)§

1.5 (0.9)

1.4 (0.7)

1.0 (0.2)‡§㛳 1.3 (0.8)§㛳

2.2 (1.0)

2.8 (1.5)

1.0 (0.2)‡§㛳 1.2 (0.5)§

1.4 (1.1)

1.5 (0.8)

1.0 (0.2)‡§㛳 1.3 (0.8)§㛳

2.2 (1.1)㛳

2.6 (1.6)

1.0 (0.0)‡§㛳 1.0 (0.3)§

1.2 (0.7)

1.3 (0.5)

1.0 (0.0)‡§㛳 1.1 (0.5)§㛳

1.8 (1.2)㛳

2.6 (1.4)

1.0 (0.0)‡§㛳 1.0 (0.1)§

1.0 (0.4)

1.1 (0.3)

1.0 (0.0)‡§㛳 1.0 (0.3)§㛳

1.7 (1.1)㛳

2.4 (1.5)

1.1 (0.3)‡§㛳 1.5 (1.1)§

1.8 (1.3)

1.9 (1.0)

1.3 (0.5)‡§㛳 2.0 (1.2)§㛳

2.9 (1.0)㛳

3.8 (1.5)

1.0 (0.1)‡§㛳 1.1 (0.5)

1.3 (0.9)

1.3 (0.9)

1.1 (0.3)‡§㛳 1.6 (1.0)§㛳

2.5 (1.4)㛳

3.3 (1.6)

1.0 (0.1)‡§㛳 1.1 (0.4)§ 1.0 (0.3)‡§㛳 1.3 (0.9)

1.3 (0.7) 1.7 (1.3)

1.3 (0.7) 1.5 (1.0)

1.1 (0.4)‡§㛳 1.5 (1.0)§㛳 1.1 (0.3)‡§㛳 1.6 (1.0)§㛳

2.4 (1.4)㛳 2.3 (1.4)㛳

3.2 (1.8) 3.1 (1.8)

1.0 (0.1)‡§㛳 1.2 (0.6)

1.3 (0.9)

1.3 (0.8)

1.1 (0.3)‡§㛳 1.4 (0.9)§㛳

2.2 (1.5)

2.8 (1.4)

1.0 (0.1)‡§ 1.0 (0.1)§ 1.0 (0.0)‡§㛳 1.0 (0.3)§

1.0 (0.2) 1.1 (0.5)

1.0 (0.1) 1.1 (0.5)

1.3 (0.7)‡§㛳 1.8 (1.3)§㛳 1.3 (0.6)‡§㛳 2.0 (1.4)§㛳

2.5 (1.9) 3.1 (1.5)㛳

3.1 (1.7) 4.0 (1.4)

1.1 (0.3)‡§㛳 1.3 (0.7)§ 1.0 (0.3)‡§㛳 1.2 (0.9)§

1.4 (1.2) 1.3 (1.0)

1.5 (0.8) 1.5 (1.0)

1.2 (0.3)‡§㛳 1.5 (0.8)§㛳 1.2 (0.4)‡§㛳 1.7 (1.3)§㛳

2.1 (1.3) 2.4 (1.9)

2.4 (1.7) 2.7 (2.2)

IQ ⫽ interquartile; other abbreviation as in Table 1. * Scale ranged from 1 (not at all) to 6 (extreme); rating for each woman ⫽ average from two menstrual cycles; group ratings compared with Wilcoxon rank sum and median tests, with significance level set at P ⬍ .001 by the Bonferonni adjustment for multiple comparisons. † Follicular phase ⫽ days 5–10; luteal phase ⫽ first day of bleeding plus 5 preceding days. ‡ Significantly different from moderate. § Significantly different from severe. 㛳 Significantly different from PMDD.

exercise was associated with increased, rather than decreased, likelihood of use, whereas better health status was related to decreased likelihood (not significant). The factors associated with increased use of herbal or nutritional therapies were older age and having a comorbidity; decreased use was associated with being Asian, relative to whites (P ⫽ .056).

VOL. 99, NO. 6, JUNE 2002

DISCUSSION This study of a large, diverse population of women of reproductive age found that, for the sample as a whole, mean daily ratings of physical and emotional symptoms increased steadily from the follicular to the luteal phase, peaking around the first day of menstrual bleeding.

Sternfeld et al

Severity of Premenstrual Symptoms


Table 3. Demographic, Health Status, and Lifestyle Characteristics of the Total Sample and by Symptom Severity Classification Percent All (N ⫽ 1194)

Minimal (n ⫽ 186)

Moderate (n ⫽ 801)

Severe (n ⫽ 151)

PMDD (n ⫽ 56)

26.5 46.7 26.7

22.0 43.0 35.0

26.1 48.1 25.8

34.4 45.0 20.5

26.8 44.6 28.6

65.8 6.0 10.8 9.5 7.9

61.3 7.0 7.5 12.9 113

66.9 5.7 11.0 9.7 3.8

62.9 6.0 14.6 6.0 10.6

73.2 7.1 8.9 3.6 7.2

16.4 41.7 41.9

21.0 35.5 43.5

15.6 45.5 41.0

13.2 44.4 42.4

21.4 30.4 48.2

61.6 18.4 20.0 66.8 25.3

62.4 18.3 19.4 67.6 25.3

62.1 18.2 19.7 67.0 25.0

57.6 19.9 22.6 63.6 26.5

62.5 17.9 19.7 69.6 26.8

5.1 33.1 42.2 19.6 47.0 82.3

3.3 30.1 47.5 19.1 32.3 71.0

4.4 34.1 40.9 20.6 47.3 82.7

9.4 31.3 41.3 18.0 57.6 90.1

9.3 33.3 46.3 11.1 62.5 94.6

79.7 15.7 10.1 7.1 9.5 20.1

68.3 8.6 4.8 3.2 5.4 10.8

79.8 14.5 9.5 7.2 9.2 19.4

88.1 27.2 16.6 9.9 13.3 31.8

94.6 25.0 19.6 10.7 16.1 30.4

16.4 6.0 12.8 8.6 26.6 39.9 15.5

10.8 3.2 5.9 2.2 25.8 44.1 11.3

15.6 5.2 12.4 7.9 28.7 37.7 15.5

24.5 9.9 19.9 14.6 19.9 45.7 20.5

25.0 16.1 23.2 25.0 17.9 42.9 16.1

Age (y)* 21–30 31–40 ⬎40 Race/ethnicity White, non-Hispanic Black, non-Hispanic Hispanic Asian Other Education ⱕHigh school graduate Some college ⱖCollege graduate Employment status Full time Part time Not employed Married, living as married Children ⬍5 y at home Self-reported health status Poor/fair Good Very good Excellent One or more comorbidities† Any use of OTC or prescription medication for PMS† OTC pain drug† Prescribed pain drug† Any use of alternative therapies of PMS† Massage* Meditation, imagery, relaxation* Any use of herbal remedies or nutritional supplements for PMS† Calcium† Magnesium† Vitamin B6† St. John’s wort† Current OC use Any strenuous exercise Current smokers

OTC ⫽ over the counter; PMS ⫽ premenstrual syndrome; OC ⫽ oral contraceptive; other abbreviation as in Table 1. * Proportions in groups significantly different, ␹2 test, P ⬍ .05. † Proportions in groups significantly different, ␹2 test, P ⬍ .01.

Significant between-group differences in symptom-specific ratings were most pronounced in the luteal phase, with the greatest severity reported by those who met the diagnostic criteria for PMDD. Differences in symptomspecific ratings between cycles were observed only for four luteal phase physical symptoms. The stability of emotional symptoms from cycle to cycle is in contrast to the findings of Gallant et al16 and suggests that two cycles


Sternfeld et al

Severity of Premenstrual Symptoms

of diaries may not be necessary to differentially diagnose PMDD from other premenstrual syndromes. In multivariable models, symptom severity was inversely associated with age, oral contraceptive use (emotional symptoms only), and better perceived health status (physical symptoms only) and directly associated with having any comorbidity. Relative to whites, Hispanics reported greater severity of symptoms, and


Table 4. Adjusted Association* of Demographic, Health Status, and Behavioral Factors With Luteal Phase Emotional and Physical Symptom Severity Scores Emotional symptom score Age, per 1 y Race Black Asian Hispanic Other White Children ⬍5 y at home Yes No Comorbidity, per one condition Health status Excellent, very good, good Fair, poor OC use Yes No Smoker Yes No Vigorous activity three or more times per week Yes No

Physical symptom score

␤ coefficient



␤ coefficient









⫺0.93 ⫺1.40 1.66 ⫺0.15

0.88 0.72 0.69 0.78

.29 .05 .02 .85

⫺1.12 ⫺0.92 0.95 ⫺0.25

0.68 0.55 0.53 0.60

.10 .10 .07 .68




















0.50 0.00

















SE ⫽ standard error; other abbreviation as in Table 3. * From multivariable linear regression models with each variable adjusted for all other variables.

Asians less. Asians also reported less use of prescription or over-the-counter medicines or herbal or nutritional supplements for premenstrual symptoms compared with whites, independently of symptom severity, whereas having a comorbidity was associated with more use of any type of premenstrual symptom remedy. Regular participation in vigorous activity was related to less use of prescription and over-the-counter medicines, but more use of alternative therapies. These findings are generally consistent with previous studies. In particular, the inverse association between age and symptom severity observed in the present investigation reinforces the conclusion of others that symptoms are more pronounced earlier in reproductive life and do not worsen as menopause approaches.8,11,19 Although premenstrual symptoms are not explained by a simple hormonal deficiency or excess, the dramatic decrease in physical and emotional symptoms that occurs with “medical ovariectomy” from use of long-acting gonadotropin-releasing hormone agonist clearly suggests that premenstrual symptoms are linked to the ovarian hormonal cycle.20 Thus, as women approach the end of their reproductive lives, the regular cyclic fluctuation of ovarian hormones may, perhaps, become less pronounced, resulting in decreased severity of premenstrual symptoms, despite the increase in menopausal

VOL. 99, NO. 6, JUNE 2002

symptoms that may occur in association with declining estrogen. The finding in this study of an inverse association between oral contraceptive use and emotional symptom severity, which has also been noted by others1,3 may provide further support for the importance of fluctuating ovarian hormones as a trigger for premenstrual symptoms. On the other hand, these findings may contradict the experience of clinical practice; only a longitudinal study, which follows women for premenstrual symptoms as they age, will conclusively resolve this issue. Although a few previous investigators have examined the prevalence of premenstrual symptoms in Mexican11,21 or African22 women, or differences in prevalence between whites and blacks in the United States1,4,19,23,24 or the United Kingdom,25 the findings regarding racial/ethnic differences in prevalence of symptoms have been inconsistent. Premenstrual symptoms appear to be reported by a substantial proportion of women in all cultures, but some symptoms may be more common in one population group than another. The extent to which education or geography (ie, urban versus rural) confounds these differences is not clear.11 In the present study, whites and blacks reported no differences in symptom severity, but Hispanic women reported greater severity of symptoms and Asian women reported

Sternfeld et al

Severity of Premenstrual Symptoms


Table 5. Adjusted Association* of Demographic, Health Status, and Behavioral Factors With Reported Use of Medical, Alternative, and Herbal PMS Remedies Prescription/OTC medicine† Age, per 1 y Race Black Asian Hispanic Other White Children ⬍5 y at home Yes No Comorbidity, per one condition Health status Excellent, very good, good Fair, poor OC use Yes No Smoker Yes No Vigorous activity three or more times per week Yes No Symptom severity group PMDD Severe Moderate Minimal

Alternative therapies‡

Herbal therapies§


95% CI


95% CI


95% CI


0.92, 0.98


0.96, 1.02


1.01, 1.06

0.98 0.30 0.75 0.73 1.00

0.46, 2.11 0.19, 0.48 0.43, 1.29 0.39, 1.36

1.04 0.69 1.11 1.19 1.00

0.46, 2.31 0.31, 1.57 0.59, 2.10 0.59, 2.40

0.96 0.52 0.67 1.41 1.00

0.50, 1.85 0.27, 1.02 0.38, 1.18 0.83, 2.41

0.74 1.00 1.28

0.51, 1.08

0.90 1.00 1.25

0.56, 1.43

0.74 1.00 1.23

0.51, 1.07

1.00 1.00

0.70, 1.44

0.67 1.00

0.44, 1.02

1.35 1.00

0.97, 1.89

1.50 1.00

0.97, 2.34

0.78 1.00

0.47, 1.28

1.22 1.00

0.84, 1.77

1.29 1.00

0.77, 2.17

0.81 1.00

0.46, 1.42

1.18 1.00

0.78, 1.78

0.89 1.00

0.81, 0.98

1.13 1.00

1.02, 1.27

1.05 1.00

0.96, 1.15

5.07 3.66 1.73 1.00

1.48, 17.32 1.81, 7.39 1.16, 2.58

3.53 3.10 1.80 1.00

1.32, 9.43 1.37, 7.03 0.87, 3.70

3.22 3.39 1.83 1.00

1.49, 6.87 1.85, 6.24 1.10, 3.06

1.08, 1.53

1.07, 1.46

1.08, 1.40

OR ⫽ odds ratio; CI ⫽ confidence interval; other abbreviations as in Tables 1 and 3. * From multivariable logistic regression models with each variable adjusted for all variables. † Includes antidepressants, antianxiety drugs, prescribed pain drugs, OTC pain drugs, muscle relaxants, hormonal therapy. ‡ Includes acupuncture, biofeedback, chiropractic care, homeopathic treatments, hypnosis, massage therapy, meditation, imagery or relation, reflexology, spiritual healing or prayer, other traditional medicine. § Includes calcium, magnesium, St. John’s wort, vitamin B6, vitamin E, zinc, evening primrose, other herbs and supplements.

less. Although others have noted that Asian women report fewer menopausal symptoms,26 it is unknown whether this reflects biologic and/or cultural differences or reporting differences in response to symptom scales. Future research, perhaps using an ethnographic, more qualitative approach may be needed to understand better the meaning of premenstrual symptoms in different cultures. The identification, in a nonclinic sample of women, of comorbidities and self-perceived poor health as risk factors for more severe premenstrual symptoms is in accord with a population-based cross-sectional survey of 2650 women in Calgary, Canada.2 Given that the sample in the current study was screened for psychologic morbidity, this finding may indicate: 1) that comorbid conditions exacerbate premenstrual symptoms through some unknown mechanism; or 2) that there are subgroups of


Sternfeld et al

Severity of Premenstrual Symptoms

the population with greater than average susceptibility to a range of disparate symptoms and disorders or with greater than average sensitivity to symptoms and pain. Better understanding of the connection between comorbidities, perceived health status, and premenstrual symptoms could prove beneficial for identifying and treating women with severe premenstrual symptoms. In contrast with several studies that suggest that exercise may prevent or minimize premenstrual symptoms,3,7,14,27 this study found no association between regular participation in vigorous activity and symptom severity. Two earlier studies found exercise to be more common in women with more severe symptoms,2,4 suggesting that exercise may be used as a treatment for symptoms in those women most severely affected. The decreased likelihood of prescription or over-the-counter drug use for premenstrual symptoms in women who


exercised, along with the increased likelihood of use of alternative therapies, reinforces the conclusion that women who exercise may be a heterogeneous group, including those who are more symptom free who exercise for a variety of reasons along with those who have more severe symptoms who exercise because they believe it will improve their symptoms. This study has several limitations that need to be noted. First, data for several important covariates, most notably body weight and dietary intake, were not available. More importantly, even with the prospective assessment of symptoms, the observational study design does not allow for determination of causality. This is particularly relevant in terms of the observed associations of symptoms with comorbidity and poorer perceived health and the absence of any observed association with exercise. Longitudinal studies or intervention trials may be necessary to address these questions adequately. Another limitation is that this study excluded women using psychotropics or with a previous PMDD diagnosis. As a result, the sample is probably most representative of women least likely to seek formal medical treatment for severe premenstrual symptoms. These exclusion criteria were imposed explicitly to ensure an untreated sample. However, because women in the severe group did not meet the formal criteria for a PMDD diagnosis, they might have been less likely to be treated, and, therefore, more likely to have been eligible for the study, relative to the PMDD group. Thus, any conclusions regarding differences between women in the severe and PMDD groups should be considered tentative. Also, only 51.6% of women initially enrolled completed two cycles of daily diaries, raising a concern about response bias. Specifically, differential study attrition by symptom severity group was observed. The completion rates for the three a priori preliminary premenstrual classification groups were 56.1%, 53.5%, and 49.0% for minimal, moderate-severe, and probable PMDD, respectively. Finally, inclusion of oral contraceptive users may have resulted in misclassification bias that could account for the observed inverse association with emotional symptoms, if women with more severe symptoms avoid oral contraceptives because they exacerbate their symptoms. The fact that the proportion of oral contraceptive users was higher in the minimal and moderate groups than the severe and PMDD groups lends support for this argument. However, the fact that the association between symptom severity and other risk factors was changed little by exclusion of oral contraceptive users suggests that oral contraceptive use did not confound these relations. Despite its limitations, the current study overcomes

VOL. 99, NO. 6, JUNE 2002

many of the limitations of previous studies of premenstrual symptoms. First, the defined sampling frame may have eliminated the bias that arises from studies of clinical populations. Although a large proportion of sampled names could not be contacted, and only slightly more than half of all the women who enrolled in the study completed the entire protocol, a comparison of the characteristics of the actual study sample with those of the privately insured population in Northern California, using the 1996 Behavioral Risk Factor Surveillance survey (personal communication, N. Gordon, December 1997), suggests that selection bias was minimal. For instance, although Hispanics were slightly under-represented in this study (11.1% versus 16.3%), the proportions of blacks and Asians were almost identical (6.2% versus 6.4% and 9.7% versus 8.8%, respectively). Secondly, standardized, validated instruments for assessing premenstrual symptoms were used, allowing for construction of a category of symptom severity that corresponded to diagnostic criteria for PMDD. Finally, prospective, daily recording of symptom severity over two complete menstrual cycles avoided recall bias and the general tendency for over-estimation of symptom severity that usually results from retrospective reporting.28 In conclusion, as this study demonstrates, premenstrual symptoms are a significant problem for a large proportion of women, and one that is not confined to any particular cultural, racial, or socioeconomic group. It is also a problem for which many women seek both medical/pharmacologic treatments and alternative remedies, including lifestyle modification. The challenge that remains, for researchers, medical practitioners, and for women themselves, is determining who is most likely to benefit from any specific treatment or combination of treatments, given each woman’s general state of health and level of symptom severity.

REFERENCES 1. Woods NF, Most A, Dery GK. Prevalence of perimenstrual symptoms. Am J Public Health 1982;72:1257– 64. 2. Ramcharan S, Love EJ, Fick GH, Goldfien A. The epidemiology of premenstrual symptoms in a population-based sample of 2650 urban women: Attributable risk and risk factors. J Clin Epidemiol 1992;45:377–92. 3. Johnson SR, McChesney C, Bean JA. Epidemiology of premenstrual symptoms in a nonclinical sample. I. Prevalence, natural history and help-seeking behavior. J Reprod Med 1988;33:340 – 6. 4. Deuster PA, Adera T, South-Paul J. Biological, social, and behavioral factors associated with premenstrual syndrome. Arch Fam Med 1999;8:122– 8.

Sternfeld et al

Severity of Premenstrual Symptoms


5. Steiner M. Premenstrual syndromes. Annu Rev Med 1997;48:447–55. 6. Endicott J, Amsterdam J, Eriksson E, Frank E, Freeman E, Hirschfeld R, et al. Is premenstrual dysphoric disorder a distinct clinical entity? J Womens Health Gend Based Med 1999;8:663–79. 7. Gannon L, Luchetta T, Pardie L, Rhodes K. Perimenstrual symptoms: Relationships with chronic stress and selected lifestyle variables. Behav Med 1989;15:149 –59. 8. Freeman EW, Rickels K, Schweizer E, Ting T. Relationships between age and symptom severity among women seeking medical treatment for premenstrual symptoms. Psychol Med 1995;25:309 –15. 9. Logue CM, Moos RH. Perimenstrual symptoms: Prevalence and risk factors. Psychosom Med 1986;48:388 – 414. 10. Harrison WM, Endicott J, Nee J, Glick H, Rabkin JG. Characteristics of women seeking treatment for premenstrual syndrome. Psychosomatics 1989;30:405–11. 11. Huerta-Franco MR, Malacara JM. Association of physical and emotional symptoms with the menstrual cycle and life-style. J Reprod Med 1993;38:448 –54. 12. Tobin MB, Schmidt PJ, Rubinow DR. Reported alcohol use in women with premenstrual syndrome (see comments). Am J Psych 1994;151:1503– 4. 13. Caan B, Duncan D, Hiatt RA, Lewis J, Chapman J, Armstrong MA. Association between alcoholic and caffeinated beverages and premenstrual syndrome. J Reprod Med 1993;38:630 – 6. 14. Choi PYL, Salmon P. Symptom changes across the menstrual cycle in competitive sportswomen, exercisers and sedentary women. Br J Clin Psychol 1995;34:447– 60. 15. Bloch M, Schmidt PJ, Rubinow DR. Premenstrual syndrome: Evidence for symptom stability across cycles. Am J Psych 1997;154:1741– 6. 16. Gallant SJ, Popiel DA, Hoffman DM, Chakraborty PK, Hamilton JA. Using daily ratings to confirm premenstrual syndrome/late luteal phase dysphoric disorder. Part I. Effects of demand characteristics and expectations. Psychosom Med 1992;54:149 – 66. 17. Spitzer RL, Williams JB, Kroenke K, Linzer M, deGruy FV, Hahn SR, et al. Utility of a new procedure for diagnosing mental disorders in primary care. The PRIME-MD 1000 study (see comments). JAMA 1994;272:1749 –56.


Sternfeld et al

Severity of Premenstrual Symptoms

18. Yonkers KA, Halbreich U, Freeman E, Brown C, Endicott J, Frank E, et al. Symptomatic improvement of premenstrual dysphoric disorder with sertraline treatment. A randomized controlled trial. Sertraline Premenstrual Dysphoric Collaborative Study Group. JAMA 1997;278:983– 8. 19. Lee KA, Rittenhouse CA. Prevalence of perimenstrual symptoms in employed women. Women Health 1991;17: 17–32. 20. Muse KN, Cetel NS, Futterman LA, Yen SC. The premenstrual syndrome. Effects of “medical ovariectomy.” N Engl J Med 1984;311:1345–9. 21. Marvan ML, Diaz-Erosa M, Montesinos A. Premenstrual symptoms in Mexican women with different educational levels. J Psychol 1998;132:517–26. 22. McMaster J, Cormie K, Pitts M. Menstrual and premenstrual experiences of women in a developing country. Health Care Women Int 1997;18:533– 41. 23. Boyle CA, Berkowitz GS, Kelsey JL. Epidemiology of premenstrual symptoms. Am J Public Health 1987;77: 349 –50. 24. Stout AL, Grady TA, Steege JF, Blazer DG, George LK, Melville ML. Premenstrual symptoms in black and white community samples. Am J Psych 1986;143:1436 –9. 25. van den Akker OB, Eves FF, Service S, Lennon B. Menstrual cycle symptom reporting in three British ethnic groups. Soc Sci Med 1995;40:1417–23. 26. Gold EB, Sternfeld B, Kelsey JL, Brown C, Mouton C, Reame N, et al. Relation of demographic and lifestyle factors to symptoms in a multi-racial/ethnic population of women 40 –55 years of age. Am J Epidemiol 2000;152: 463–73. 27. Prior JC, Vigna Y. Conditioning exercise and premenstrual symptoms. J Reprod Med 1987;32:423– 8. 28. Van Den Brink M, Bandell-Hoekstra E, Abu-Saad H. The occurrence of recall bias in pediatric headache: A comparison of questionnaire and diary data. Headache 2001;41: 11–20. Address reprint requests to: Barbara Sternfeld, PhD, Kaiser Permanente, Division of Research, 3505 Broadway, Oakland, CA 94611; E-mail: [email protected] Received July 5, 2001. Received in revised form December 20, 2001. Accepted January 17, 2002.