You are on page 1of 9

Premenstrual Exacerbation of Depressive Disorders In a Community-Based Sample in the United States




Objective: No published epidemiologic study has examined premenstrual exacerbation of depressive disorders (PME-DD) in a representative sample. Knowledge gained should indicate the burden of illness, suggest whom to monitor, and facilitate diagnosis. The objectives were to 1) ascertain the prevalence and predictors of PME-DD; and 2) test competing hypotheses that PME-DD is related to a) severity or history of depression, b) menstrual cyclicity in females in general, or c) a methodological artifact. Methods: Menstruating females (N 900) from ages 13 to 53 living in urban or rural Illinois completed semi-structured psychiatric diagnostic interviews and rated symptoms of depression daily for two menstrual cycles; 58 had major depressive, dysthymic, or subclinical depressive disorders, and the remaining 842 were the non-depressed portion of the representative sample. Results: Depressed females had 1.34 (95% confidence interval, 1.02 - 1.66) symptoms exacerbated premenstrually. The best model for predicting exacerbation contained only age. Older women more often had symptoms worsen. Symptoms during the follicular phase were most severe for clinically depressed, intermediate for subclinically depressed, and least severe for non-depressed participants, ps 0.001. Consistent with the hypothesis that exacerbation is related to cyclicity in all females, the number of symptoms that became worse did not differ between groups, ps 0.46. Given no symptoms in one cycle, the odds of having symptoms in the next cycle were 0.91. Only 56% of non-depressed females taking antidepressants were asymptomatic all month long; the remaining 44% still had symptoms premenstrually. Conclusions: Premenstrual exacerbation of depressive disorders is associated with deteriorated functioning over and above that already experienced by depressed females. Patients may be susceptible regardless of severity of depression, number of episodes, or remission status. Key words: depression, premenstrual exacerbation, premenstrual magnification, epidemiology. ANOVA analyses of variance; DSM-IV Diagnostic and Statistical Manual of Mental Disorders, 4th Edition; DSMQ Daily Symptom Mood Questionnaire; K-SADS-E Schedule for Affective Disorders and Schizophrenia for School Age Children Epidemiological Version-5 Kiddie SADS-E; MDD major depressive disorder; PME-DD premenstrual exacerbation of depressive disorders; PMDD premenstrual dysphoric disorder; RDC research diagnostic criteria; SCID-I/NP Structured Clinical Interview for DSM-IV Axis I Disorders - Non-Patient Edition; SSRI selective serotonin reuptake inhibitor.

INTRODUCTION eports indicating that hospital admissions (1), calls to suicide prevention centers (2), suicide attempts (3), and completed suicides (4) increase premenstrually suggest that other symptoms of depressive disorders (DD) may become worse at this time. However, no study indicates the frequency with which this occurs in the general population of depressed girls or women. Without such information, the degree to which the phenomenon needs to be addressed in treatment is unclear. Premenstrual exacerbation (PME), magnification (5), or aggravation (6) of the symptoms of DD must be distinguished from premenstrual dysphoric disorder (PMDD). Symptoms of PMDD are absent or mild post-menstrually, and are marked or

From the Department of Psychiatry (S.A.H., H.M.K.), Department of Psychology (S.A.H.), and Department of Preventive Medicine (H.M.K.), Rush University Medical Center and Rush Medical College, Chicago, Illinois; and McLaren Regional Medical Center, College of Human Medicine, Michigan State University (D.L.B.). Address correspondence and reprint requests to S. Ann Hartlage, PhD, Department of Psychiatry, Rush University Medical Center, 1720 West Polk Street, Chicago, Illinois 60612. E-mail: Received for publication December 11, 2003; revision received April 20, 2004. We appreciate Sarah Gehlerts contributions to the study, which are reflected in grant MH055226. The current study is based on data from S. Ann Hartlage generated from grant MH055221 from the National Institute of Mental Health, Bethesda, Maryland. DOI: 10.1097/01.psy.0000138131.92408.b9 698
0033-3174/04/6605-0698 Copyright 2004 by the American Psychosomatic Society

severe premenstrually and interfere with functioning. Some people argue that PMDD is mistaken for depression that becomes worse premenstrually (cf., 7). Indeed, up to half of women seeking treatment for premenstrual disorders have PME-DD (8). Moreover, six symptoms of major depressive disorder (MDD) also are symptoms of PMDD (e.g., depressed mood). In such cases, the difference between PME-DD and PMDD is not the nature of symptoms, but their timing (9). Both groups experience more severe symptoms pre-versus post-menstrually. However, women with PME-DD also experience marked or severe symptoms during the follicular phase. Other symptoms of DD (e.g., suicidal ideation), which may become worse premenstrually, differ from those of PMDD. Alternative hypotheses about the causes of PME-DD yield alternative predictions regarding premenstrual worsening of symptoms in depressed and non-depressed groups. Repeated episodes of depression may sensitize depressed women to PME of their symptoms. For example, past episodes may act as a kindling mechanism in which increasingly severe depression occurs with little provocation (10). The provocation may be fluctuations of hormones and neurotransmitters related to the menstrual cycle (11,12). It also is possible that PME may be related to menstrual cyclicity occurring in girls and women in general, regardless of history of depression. Consistently, van den Akker et al. (13) found no evidence for a genetic contribution to depression and premenstrual symptoms (except that in a common personality factor). However, other investigators have differed (14,15). A third alternative is that apparent PME of depressive symptoms is influenced by a methodological artifact. Assuming that symptoms of depression already are rated as severe on the scale provided, that scale may not be able to reflect the degree of worsening of symptoms. That is, although symptoms may be rated as more severe premenstrually, the amount
Psychosomatic Medicine 66:698 706 (2004)


of worsening may appear less than is actually the case due to limitations of the scale. The current study will be the first published to ascertain the prevalence of PME of symptoms of depression, including suicidal ideation, in a representative English- and Spanishspeaking sample diagnosed with DD. Inclusion of the remaining representative sample of non-depressed controls will enable testing competing hypotheses regarding possible causes of exacerbation. Secondary analyses of clinical relevance include whether increases in premenstrual symptomatology are

associated with functional impairment, and the stability of the phenomenon across cycles. Whether standard psychiatric medication regimens, which typically include selective serotonin reuptake inhibitors (SSRIs), alleviate premenstrual aggravation also is a focus. METHODS
Data for the current study are from the urban (Chicago) and rural (DeKalb County) Illinois sites of an epidemiologic prospective study of womens health. Participants completed the Structured Clinical Interview for DSM-IV

Bivariate Relationships Between Demographic or Clinical Characteristics, and Mean Number of Symptoms Exacerbated Premenstrually Nondepressed (N 842) M (SD) N % Subclinically Depressed (N 26) M (SD) N 33.3 (9.1) 117 13.9 104 12.4 604 71.7 17 2.0 34 4.0 77 9.1 129 15.4 293 34.8 187 22.2 38 4.5 75 8.9 9 1.1 82 9.7 100 11.9 609 72.3 16 1.9 35 4.2 20 42 75 83 100 107 93 230 92 2.4 5.0 8.9 9.9 11.9 12.7 11.0 27.3 10.9 2 7.8 1 3.8 22 84.6 1 3.8 1 3.8 2 7.8 1 3.8 14 53.8 4 15.4 3 11.6 0 0 1 3.8 4 15.4 3 11.6 17 65.4 1 3.8 1 3.8 1 3 2 2 3 4 5 2 4 3.8 11.5 7.7 7.7 11.5 15.4 19.3 7.7 15.4 % Clinically Depressed (N 32) M (SD) N 34.9 (8.8) 2 6 24 0 2 1 11 10 3 2 3 0 1 0 26 3 2 2 5 4 0 4 2 6 6 3 21 11 0 5 27 6 26 6.2 18.8 75.0 0 F 6.2 3.1 34.5 31.2 9.4 6.2 9.4 0

Sample Tested*

% F

Representative .86, p 1.83, p .36 .40 F


Depressed 11.48, p 2.18, p .001 .34

Age Race Hispanic Black, not Hispanic White, not Hispanic Other or not reported* Education Fewer than 9 years Some high school High school graduate Some college/technical school College graduate Some post graduate education Post graduate degree Not reported* Work category Student with/without employment Homemaker Employed Unemployed Other or not reported* Total family income Less than $3,000 $3,000 to $9,999 $10,000 to $19,999 $20,000 to $29,999 $30,000 to $39,999 $40,000 to $49,999 $50,000 to $59,999 $60,000 or more Not reported* Marital status Married, or living as if married Single, separated, divorced, widowed Other or not reported* Environment Urban Rural Oral contraceptives Taking Not taking

33.0 (9.2)

3.37, p


.35, p


.65, p


5.16, p


3.1 0 81.3 9.4 6.2 F 6.2 15.6 12.5 0 12.5 6.2 18.8 18.8 9.4

.25, p


3.75, p


.21, p


4.13, p


542 64.4 293 34.8 7 0.8 329 39.1 513 60.9 179 21.3 663 78.7

18 69.2 8 30.8 0 0 6 23.1 20 76.9 4 15.4 22 84.6

65.6 34.4 0

.69, p


1.65, p


15.6 84.4

4.98, p


.32, p


18.8 81.2

* Data from participants in other or not reported categories were not included in analyses. Any difference among nondepressed, subclinically depressed, and clinically depressed (i.e., major depressive and/or dysthymic disorder) groups. Difference between subclinically and clinically depressed groups. Psychosomatic Medicine 66:698 706 (2004) 699

S. A. HARTLAGE et al.
Axis I Disorders Non-Patient Edition (SCID-I/NP) (16) or the Schedule for Affective Disorders and Schizophrenia for School Age Children Epidemiologic Version-5 Kiddie SADS-E (K-SADS-E) (17) to diagnose DD. They rated symptoms of depression and functioning daily for two menstrual cycles. Clinically and subclinically depressed females filled out an additional questionnaire about symptoms of these disorders.

Participants in the Womens Wellness Study of the prevalence of PMDD took part in this study (see Table 1). They were menstruating females from ages 13 through 53 with current major depressive, dysthymic, or subclinical depressive disorders (n 58); and the remaining nondepressed portion of the representative sample (n 842). Although women through age 55 were screened, no woman older than 53 met inclusionary/exclusionary criteria and had two menstrual cycles within the 100-day maximum data collection interval. Twelve (1.4%) of the controls did not complete psychiatric testing. They were included in the study in order to represent the full sample, because their percentage was small, and because it cannot be assumed that they had a current depressive disorder. Exclusionary criteria were a) less than 85% of normal weight for age and height (18); b) hormone replacement therapy for symptoms of menopause; c) a clotting disorder; d) pregnant, lactating, or a dilatation and curettage within the last 3 months; e) history of a hysterectomy or oophorectomy; or f) history of chemotherapy or radiation therapy at a level thought to interfere with ovarian functioning. Females taking oral contraceptives or psychiatric medications were included in the study, because there is no consistent evidence (from other investigators (19) and our unpublished data) that these medications alleviate premenstrual symptoms.

To identify potential participants, The National Opinion Research Center selected community-based samples of 25 segments in each of the two geographic locations, listed each housing unit within these segments, and randomly selected 128 housing units from each listing. Study personnel sent letters about the study to the woman living at each address. Subsequently, field interviewers visited housing units to screen menstruating (i.e., a menstrual period within the past 100 days) females from the ages of 13 through 55 years. It was unlikely that females who had not menstruated in the past 100 days would have three periods (two cycles) during the subsequent 100-day data collection interval. Interviewers offered eligible females $80 to $120, depending on the length of participation, for being in the study of womens health, in an attempt to disguise the menstrual focus of the study (20). In order to decrease the possibility that familial or environmental factors would unduly influence results, if more than one female lived in a household, only the one with the most recent past birthday was invited to participate (21). If she refused, no other woman in the household was invited. Participants 18 years and older, and the parents or guardians of participants less than 18 signed informed consent. Adolescents gave written assent. An institutional review board at Rush University Medical Center approved the study protocol. Participants in the main study gave health histories, rated 50 symptom and mood items, and gave a urine sample daily for two menstrual cycles. They completed psychiatric diagnostic interviews. We strived to administer the psychiatric interviews during the follicular phase in order to minimize possible premenstrual effects on responses. If participants met criteria for a clinical or subclinical DD, they began rating daily additional items about symptoms of depression. We did not give these to everyone, because items about death or hurting oneself might be disturbing to nondepressed participants. Investigators determined whether an eligible female lived in 5,281 of the 6,359 housing units visited, yielding a screener response rate of 88.8%. Of 1,489 eligible females, 80.4% enrolled. During the study, 100 women (8.4%) became ineligible (e.g., they became pregnant, moved to another country, did not have three periods during the maximum 100-day data collection interval), 183 (15.3%) dropped out before completing two menstrual cycles, and data from 15 participants (1.3%) were unusable (e.g., they completed data collection but skipped filling out questionnaires for one premenstrual phase). 700

MATERIALS Daily Symptom and Mood Questionnaire (DSMQ) Investigators designed the DSMQ to measure changes in the severities of symptoms of PMDD and in functioning in relation to the menstrual cycle. A discussion of symptoms of PMDD that are not also symptoms of depression is beyond the scope of this paper. However, 13 items comprising 10 symptoms of DDs (see Table 2) also are on the DSMQ. Compound symptoms, (eg, hypersomnia or insomnia) in the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition (DSMIV) are split into sub-symptoms (items; eg, hypersomnia). Positive sub-symptoms are included in order to avoid response set. Participants rate items on a 0 to 5 scale indicating severity from I did not experience the symptom or emotion to I experienced the symptom or emotion very severely. Coefficient alpha based on data from one randomly selected day for each of the current 900 participants was 0.94, indicating excellent internal consistency. To ascertain whether there were menstrual-related changes in the severity of symptoms, investigators analyzed symptom severity data using the effect size method (22,23), which is sensitive to an individuals day-to-day variability. The effect size for each sub-symptom in each cycle is calculated by subtracting the mean rating for the sub-symptom across days 5 through 10 of the follicular phase (counting the first day of full flow as day 1) from the mean rating for the sub-symptom across the 7 days before menses and then dividing by the SD of the sub-symptom across the individuals cycle. If ovulation occurred before day 11, according to analyses of urine samples for luteinizing hormone using ovulation predictor kits, then only the days before ovulation were included in the follicular phase. If the effect size was equal to or greater than one, the sub-symptom was counted as present. If any of the subsymptoms making up a compound symptom was present, the compound symptom was counted as present. We could have treated the effect size as a continuous measure, which would have yielded greater statistical power. We elected to dichotomize it in terms of the presence or absence of exacerbated symptoms because a) this approach is more conservative and less likely to yield spurious significant results; b) only meaningful change (equal to or greater than one SD) is considered; c) our results can be compared with those in the PMS/PMDD literature in which there is a precedent for using the effect size method; and d) the approach is consistent with clinical tradition, which considers the presence or absence of symptoms. DSMQ items about functional impairment and symptom severity are separate. Response choices are on Likert scales, generally from not at all to extreme. We analyzed responses via the same effect size method used to analyze symptom severity data. For the current study, we used the General Functioning factor, which contains seven items that could apply to almost everyone (eg, productivity at home, avoiding social activities, arguments with people). Analyses
Psychosomatic Medicine 66:698 706 (2004)


TABLE 2. Disorder MD, Dysthymic MD Depressive Disorder by Symptom of the Disorder, Item Rated to Assess the Symptoms, and Instrument Containing the Item Symptom Depressed mood Diminished interest or pleasure Item Depressed mood Decreased interest in usual activities Lost interest in things I usually enjoyed Lost pleasure in things I usually enjoyed Poor appetite Change in appetite Overate Trouble sleeping Slept too much Agitated Moved more slowly than usual Talked more slowly than usual Unable to sit still Easily fatigued Lack of energy Felt guilty Felt worthless Trouble thinking Difficulty concentrating Difficulty making decisions Thought about death Thought about hurting myself Negative thoughts about myself Low self-esteem Feelings of hopelessness additional symptom questionnaire. Instrument DSMQ DSMQ Add-On Add-On Add-On DSMQ DSMQ DSMQ DSMQ DSMQ Add-On Add-On Add-On DSMQ DSMQ DSMQ, Add-On Add-On Add-On DSMQ Add-On Add-On Add-On DSMQ Add-On DSMQ

MD, Dysthymic

Weight loss or gain, or decrease or increase in appetite

MD, Dysthymic MD

Insomnia or hypersomnia Psychomotor agitation or retardation

MD, Dysthymic MD MD, Dysthymic

Fatigue or loss of energy Worthlessness or guilt Diminished ability to think or concentrate, or difficulty making decisions

MD Dysthymic Dysthymic Note. MD

Thoughts of death or suicide Low self-esteem Hopelessness

major depressive; DSMQ

Daily Symptom and Mood Questionnaire; Add-On

of results from this data set of 900 participants indicate that the factor is reliable, alpha 0.86. Additional Symptom Questionnaire The additional symptom questionnaire contained 2 items relevant to thoughts of death or suicide, which are not on the DSMQ. Eleven additional items relevant to 7 other symptoms of depression, increase the reliability with which the symptoms are measured (see Table 2). Response sets are the same as those for symptoms on the DSMQ. SCID-I/NP (16) To diagnose current and past Axis I disorders, women ages 18 and older completed the SCID-I/NP. The nonpatient version is for persons not identified as psychiatric patients. Participants with a) MDD, including major depression superimposed on dysthymic disorder; or b) bipolar disorder and currently in a major depressive episode qualified for the major depressive diagnostic group. Participants with major depressive or dysthymic disorder except impairment, DD not otherwise specified except PMDD, or minor DD were categorized as subclinically depressed. Various investigators have reported test-retest reliabilities similar to those for other diagnostic interviews (24) and testretest reliability coefficients from 0.70 to 1.0 (25). In the current study, interrater reliability was high (kappa 0.94) when 12 interviewers rated tapes of 14 interviews of adults
Psychosomatic Medicine 66:698 706 (2004)

completing the SCID or adolescents completing the K-SADS-E (see below) who did or did not have diagnoses of major depressive, dysthymic, alcohol or substance use, panic, or phobic disorder. K-SADS-E (17) Participants from 13 to 18 years old living with their parents completed the K-SADS-E. It is a well-validated semistructured interview designed to facilitate diagnoses of DSM-IV disorders in adolescents not identified as psychiatric patients. It is more similar to the SCID-I/NP than is any other valid diagnostic interview for youth. This interview is not used primarily for making Research Diagnostic Criteria (RDC) (26) diagnoses. As is typical, a parent- or guardian-informant was interviewed before the adolescent, and discrepancies were discussed during the adolescents interview. Adequate testretest reliability, with scores of r 0.73 for MDD and r 0.72 for dysthymic disorder, was obtained (27). Statistical Analyses Preliminary and Descriptive Analyses Of the 11 symptoms of depression (see Figure 1), only 10 could be included in analyses of data from the representative sample, because they did not rate thoughts of death or suicide. The mean number of symptoms exacerbated (i.e., the number of symptoms exacerbated divided by the total possible number of symptoms exacerbated premenstrually per partici701

S. A. HARTLAGE et al.

Figure 1. Percentage of 110 menstrual cycles during which symptoms of depression were exacerbated premenstrually in participants with clinical or subclinical depressive disorder (N 58).

pant per cycle) was not distributed normally. Hence, unless otherwise specified, the dependent variable is a dichotomous split of the mean number of symptoms exacerbated (0, 1 or more). Two complete cycles of data were available for 52 participants and one cycle of data was available for 6 clinically or subclinically depressed participants. Logistic regression did not yield significant cycle (first, second) differences in the number of symptoms exacerbated P2 (1, N 52) 0.004, p .95; there was no indication that order effected the way participents filled out questionnaires. Thus, data were collapsed across cycles. Bivariate Analyses To ascertain whether the number of symptoms exacerbated in the depressed sample was associated with demographic or clinical variables, we performed a series of analyses of variance (ANOVAs) with age, level of education, or total family income as the dependent variable; and a series of logistic regressions with race/ethnicity (Hispanic, black not Hispanic, white not Hispanic), work category (unemployed, student, employed, homemaker), marital status (married or living as if married; single, separated, divorced, or widowed), environment (urban, rural), oral contraceptives (taking, not taking), depression status (clinically depressed [i.e., MDD and/or dysthymic disorder] , subclinically depressed), history of episodes of depression (recurrent, single, none [except subclinical ones]), or antidepressant medication (taking, not taking) as the predictor variable. Similar analyses in the entire representative sample identified control variables for testing hypotheses.

Multivariate Model We performed a backward stepwise logistic regression to identify the best set of predictors of the dichotomized number of symptoms exacerbated in the depressed sample. Candidate predictor variables were those that bivariate analyses indicated were associated with the dependent variable; p .25 was the screening criterion for candidate variables, because the more traditional value of p .05 often fails to identify variables known to be important (28). The criterion for removal was p .05; terms were returned to the model if they were significant at p .025. Testing Alternative Hypotheses Three possible causes for exacerbation are a) the severity of current or past depression, b) cyclicity around menses occurring in girls and women in general, or c) a ceiling effect (artifact) because symptoms rated severe all month long could not be rated higher premenstrually. We tested these competing hypotheses primarily in the complete representative sample. Regardless of the hypothesis, we predicted that the baseline level of depression indexed by the mean rating of depressive symptoms during the follicular phase would be from highest to lowest clinically, subclinically, and nondepressed participants, respectively after controlling for the effects of other predictors. We tested this postulate via logistic regression with depression status as the predictor variable, the dichotomized mean follicular phase rating (low severity, high severity) as the dependent variable, and variables identified in bivariate analyses as control variables. If the first hypothesis were true, the number of symptoms
Psychosomatic Medicine 66:698 706 (2004)


confidence interval, 1.02 - 1.66). In a given cycle, 58% of depressed participants had one or more symptoms become worse before menses (see Figure 2). The percentage of cycles during which each symptom worsened premenstrually ranged from 1.8% for thoughts of death or suicide to 20.2% for insomnia or hypersomnia (see Figure 1); 15.6% of participants experienced insomnia and 7.3% experienced hypersomnia. Bivariate analyses of data from the depressed sample showed that age and marital status predicted the number of symptoms exacerbated (see Table 1). Older women, or those married or living as if married, were more likely to have symptoms become worse before menses. No other demographic or clinical variable significantly predicted at the p .05 level number of symptoms exacerbated in the depressed sample. (A secondary analysis with depressive diagnosis [subclinical, dysthymic disorder, or MDD] as the predictor variable also did not yield significant differences.) Results of a backward stepwise logistic regression indicated that age alone was the best predictor of the number of symptoms exacerbated per cycle. A test of the model with age vs. intercept only was significant, 2 (1, N 110) 10.93, p .001. The standardized regression coefficient equaled 0.075, Wald 2 9.74, accounting for 13% of the variance. Neither income, marital status, work category, nor environment remained in the regression equation. Examination of the correlation matrix for candidate variables (see Table 3) suggests that this is because older age was correlated with higher income; being employed and not a student (mean 35.41 8.08) vs. unemployed, a student, or a homemaker (mean 27.78 9.66); and being or living as if married (mean 35.35 8.25) vs. single, separated, divorced, or widowed (mean 31.81 9.94). The odds of a woman ages 34 to 53 vs. 13 to 33 having one or more vs. no symptoms were 2.59 (95% confidence interval: 1.19 - 5.65). Testing Alternative Hypotheses As expected, the mean rating of the severity of symptoms of depression during the follicular phase on the 0 to 5 scale was high for clinically depressed (mean 1.43 0.91), intermediate for subclinically depressed (mean 1.08 0.65), and low for nondepressed (mean 0.62 0.59) participants after effects of other predictors (ie, age, marital status, work category, income, education, antidepressant) were taken into account, 2 (2, N 1793) 33.49, p .001. Moreover, results were consistent with the hypothesis that PME-DD is a function of cyclicity occurring in all girls and women. Analyses of data from 900 participants yielded no evidence that the number of symptoms exacerbated differed as a function of depression status. This was true whether 2 (2, N 1793) 1.59, p .46; or not, 2 (2, N 1793) 1.44, p .49, effects of other predictors (at the p .25 level; ie, oral contraceptives, education) were controlled. Similarly, the mean rating of the severity of symptoms during the follicular phase failed to predict the number of symptoms exacerbated before 2 (1, N 1793) 1.31, p .26; or after 2 (1, N 1793) 1.90, p .17, other effects were removed. Consis703

Figure 2. Percentage of 110 menstrual cycles during which various numbers of symptoms were exacerbated premenstrually in participants with clinical or subclinical depressive disorder (N 58).

exacerbated would be higher for clinically, subclinically, and non-depressed participants, in that order, after the effects of other predictors were taken into account. If the second hypothesis were true, the same analysis in the large representative sample should fail to yield significant differences according to depression status in the number of symptoms exacerbated. If the third hypotheses were true, there should be an inverse relationship between depression and exacerbation, presumably due to a ceiling effect. The number of symptoms exacerbated would be fewest for clinically depressed, intermediate for subclinically depressed, and greatest for nondepressed participants. Another prediction based on the first hypothesis is that the number of symptoms exacerbated is greater for participants with more severe depression as indexed by the mean follicular phase rating after the effects of other predictors are taken into account. Alternatively, the number of symptoms exacerbated would be highest for depressed participants with a history of recurrent, single, and no episodes of depression (except subclinical ones), respectively. Clinically Relevant Secondary Analyses To investigate the clinical significance of premenstrual worsening of symptoms, we performed logistic regressions with the dichotomized number of 7 general functioning items (no items, one or more items; see DSMQ) as the dependent variable, both in the depressed and representative samples. As an indication of how long to monitor subclinically or clinically depressed patients for possible PME, we calculated the odds of having one or more symptoms in the second cycle given no symptoms in the first cycle. RESULTS The mean and SD of the number of symptoms exacerbated per cycle in the depressed sample was 1.34 ( 1.69; 95%
Psychosomatic Medicine 66:698 706 (2004)

S. A. HARTLAGE et al.
TABLE 3. Means, Standard Deviations, and Intercorrelations for Number of Symptoms Exacerbated Premenstrually per Cycle by Demographic or Clinical Predictor Variable Among 26 Subclinically* and 32 Clinically Depressed Participants Variable Number of symptoms exacerbated Predictor variable 1. Age 2. Income 3. Marital status 4. Work 5. Environment** N 110 110 98 110 104 110 M 1.3 34.2 31.4 SD 1.7 9.0 4.4 1 0.34c 2 0.21a 0.57c 3 0.16 0.19 0.48c 4 0.07 0.35c 0.39c 0.31c 5 0.12 0.03 0.40c 0.38c 0.15

* Subclinically depressed participants met all criteria for major depressive or dysthymic disorder except impairment, depressive disorder not otherwise specified except premenstrual dysphoric disorder, or minor depressive disorder. Clinically depressed participants met criteria for (a) major depressive disorder, including major depression superimposed on dysthymic disorder; (b) bipolar disorder and currently in a major depressive episode, or (c) dysthymic disorder. The natural log of one plus the mean number of symptoms exacerbated premenstrually was the dependent variable in analyses. Total family income in thousands of dollars. 2 Married or living as if married; or single, separated, divorced, or widowed. Married or living as if married; or single, separated, divorced, or widowed. Employed and not a student; or unemployed, student, or homemaker. ** Urban or rural. a p .05; b p .01; c p .001.

tently, in the depressed sample, there was no evidence for a relationship between the number of symptoms exacerbated and the number of lifetime episodes of depression experienced (recurrent [mean 1.24 1.70] , single [mean 1.48 1.71] , none except subclinical [mean 1.30 1.81]), before 2 (2, N 110) 3.27, p .20; or after 2 (2, N 110) 1.75, p .42, controlling for effects of other predictors. Clinically Relevant Secondary Analyses Among depressed participants, having one or more vs. no symptoms exacerbated was associated with deterioration in General Functioning, 2 (1, N 110) 10.21, p .001. Given no symptoms in one cycle, the odds of having one or more symptoms in the next cycle was 0.91. Among participants who took antidepressant medication, 75.7% of non-depressed and 64.3% of depressed took SSRIs. The remainder took tricyclics (9.8%) or other medications with some action on the serotonin system (17.6%). There was no significant evidence that taking versus not taking antidepressants was associated with fewer symptoms exacerbated after the effects of age were removed in either the depressed, 2 (1, N 110) 1.26, p .26 (mean 1.43 1.79 symptoms for not taking, 1.01 1.20 for taking); or the representative, 2 (1, N 1793) 0.80, p .37 (mean 1.17 1.79 symptoms for not taking, 1.00 1.54 for taking) samples. Only 56.5% of non-depressed participants taking antidepressants were asymptomatic all month long; the remaining 43.5% had one or more symptoms exacerbated premenstrually. DISCUSSION Data from a community-based sample of 900 girls and women provide a unique opportunity to understand PME-DD. This cohort minimizes selection biases associated with recruiting participants in clinical settings. Semi-structured interviews were used to diagnose DD. Depressed participants met criteria for MDD, dysthymic disorder, or subclinical forms of these

disorders. Participants rated the severity of symptoms of depression daily throughout the menstrual cycle. Knowledge about prevalence rates of PME-DD, suicidal ideation, and functional impairment should indicate the burden of illness and facilitate accurate diagnosis (29). Understanding whether premenstrual worsening of symptoms is characteristic of depressed patients or of females in general who may seek treatment indicates whom to monitor for possible worsening of symptoms. Depressed females had 1.34 (95% confidence interval, 1.02 - 1.66) symptoms exacerbate premenstrually. In any given menstrual cycle, 58% of girls and women with DD experience worsening of from one to eight symptoms of depression before menses (see Figure 2). Sleep disturbance is most often exacerbated (see Figure 1), 15.6% of depressed females had difficulty sleeping and 7.3% slept too much before menses. This is consistent with the work of other investigators who have noted underlying circadian rhythm disturbances during sleep in women with PMDD (30), though these investigators propose sleep deprivation as an intervention for PMDD. Although increased thoughts of death or suicide was the least frequent symptom exacerbated, the fact that 1.8% of depressed girls or women experience this symptom is of particular concern due to its serious nature (3). The best model for predicting exacerbation of depression included age alone. Older versus younger girls or women were 2.6 times more likely to have one or more symptoms of depression exacerbated. This finding raised the possibility that sleep difficulty associated with the menopausal transition underlies the association between PME and age (31). However, when we removed sleep difficulty as a possible symptom of PME, the association between exacerbation and age remained. There also was no evidence for a relationship between the number of symptoms exacerbated, and short or long menstrual cycles, both typical in perimenopausal women. We tested three competing hypotheses with regard to posPsychosomatic Medicine 66:698 706 (2004)


sible causes of PME-DD. Results were consistent with the second hypothesis: PME appears to be a function of menstrual cyclicity occurring in all girls or women, regardless of depression status or history. In the representative sample, the baseline severity of depression, according to the average daily rating of symptoms of depression during the follicular phase, was more likely to be high for clinically, intermediate for subclinically, and low for non-depressed participants. However, we found no evidence for differences among these groups in the likelihood that symptoms worsened before menses, ps 0.46. This was despite the large sample size of 900 participants and data from 1,793 cycles. Thus, there was no evidence to support a hypothesis that PME-DD is a function of an artifact or of the depressive condition. In the depressed group, baseline severity of depression during the follicular phase also was unrelated to number of symptoms exacerbated, p .17. Moreover, in the depressed group, there was no evidence for an effect of history of depression as indexed by number of lifetime clinical episodes (recurrent, single, none except subclinical) on the average number of symptoms becoming worse premenstrually. Even though the extent to which symptoms of depression worsen premenstrually may be similar in depressed and nondepressed girls and women, PME is of particular concern among those who are depressed for two reasons. First, the baseline follicular phase severity of symptoms is greater in depressed than non-depressed females, so the severity of symptoms premenstrually in those affected will be higher still. Second, our findings indicate that the number of symptoms worsening before menses is highly related to deterioration in functioning. This leads us to consider whether typical medication regimens ameliorate exacerbation of depression before menses. Although our data are limited in this regard and a detailed examination of the question is beyond the scope of this study, results failed to provide support for a premise that the potential problem currently is being addressed. Existing research supports the efficacy of only antidepressant medication that acts on the serotonin system for the treatment of PMDD (3235). Among participants who took antidepressant medication, 74.3% of non-depressed and 64.3% of depressed took SSRIs. The remainder took tricyclic or other medications with some action on the serotonin system. We found no significant evidence for differences between those who did and did not take antidepressants in the number of symptoms that became worse premenstrually. Even among girls and women for whom the current medication regimen appeared to be effective, that is, they were not currently depressed according to results of semi-structured diagnostic interviews, only 56% were asymptomatic all month long; the remaining 44% had at least one symptom become worse premenstrually. In viewing these data, it is important to keep in mind that the statistical power of analyses was limited by the small sample size of those taking antidepressants. We also do not know the length of treatment, the dosages of antidepressants, nor the levels of compliance, all of which could affect effecPsychosomatic Medicine 66:698 706 (2004)

tiveness. The study was not designed to answer questions about the effectiveness of antidepressant regimens; analyses were exploratory. Despite these limitations, however, findings at the very least suggest the need for further study of PME in relation to treatment. It may be that women with PME need a different medication regimen that addresses premenstrual worsening of symptoms as well as baseline level of depression. Our results and those of other investigators suggest the importance of closely monitoring patients for residual depressive symptoms occurring only premenstrually after their DDs remit (36,37). We found that if no symptoms worsened in one menstrual cycle, the odds of one or more symptoms worsening in the next cycle was 0.91; the absence of premenstrual worsening in one cycle has no effect on worsening in a second cycle. This consistency across cycles seems low. However, Ekholm et al. (22) reported that 35% of PMS patients changed diagnostic categories across two cycles. These findings suggest that patients should rate symptoms daily across more than one cycle in order to fully assess the phenomenon. Investigators have begun to explore the effects that psychotropic medications may have on PME of symptoms of depression (38,39). However, we are aware of only one published placebo-controlled efficacy study for the treatment of premenstrual worsening of DD (40). Other approaches that might be considered include cycling antidepressant medication so that a higher dose is administered before menses (39). Research clearly seems needed in this regard. REFERENCES
1. Abramowitz ES, Baker AH, Freischer SP. Onset of depressive psychiatric crises and the menstrual cycle. Am J Psychiatry 1982;139:475 8. 2. Mandell AJ, Mandell MP. Suicide and the menstrual cycle. JAMA 1967;200:1323. 3. Baca-Garcia E, Diaz-Sastre C, Ceverino A, Saiz-Ruiz J, Diaz FJ, de Leon J. Association between the menses and suicide attempts: a replication study. Psychosom Med 2003;65:237 44. 4. MacKinnon PC, MacKinnon IL. Hazards of the menstrual cycle. BMJ 1956;1:555. 5. Woods NF, Mitchell ES, Lentz M. Premenstrual symptoms: delineating symptom clusters. J of Womens Health Gend Based Med 1999;8: 1053 62. 6. Backstrom T, Hammarback S. Premenstrual syndromepsychiatric or gynaecological disorder? Ann Med 1991;23:62533. 7. Endicott J, Amsterdam J, Eriksson E, Frank E, Freeman E, Hirschfeld R, Ling F, Parry B, Pearlstein T, Rosenbaum J, Rubinow D, Schmidt P, Severino S, Steiner M, Stewart DE, Thys-Jacobs S. Is premenstrual dysphoric disorder a distinct entity? J of Womens Health Gend Based Med 1999;8:66379. 8. Plouffe LJ, Stewart K, Craft KS, Maddox MS, Rausch JL. Diagnostic and treatment results from a southeastern academic center-based premenstrual syndrome clinic: the first year. Am J Obstet Gynecol 1993;169:295303. 9. Hartlage S, Gehlert S. Differentiating premenstrual dysphoric disorder from premenstrual exacerbations of other disorders: A methods dilemma. Clinical Psychology: Science and Practice 2001;8:24253. 10. Post RM, Ballenger JC. Kindling models for the progressive development of psychopathology: Sensitization to electrical, pharmacological, and psychological stimuli. In: van Praag HM, series editor; van Praag HM, Lader MH, Rafaelsen OJ, Sachar EJ, volume editors. Experimental and Clinical Psychiatry. New York: Marcel Dekker; 1981;1(4):609 51. 11. Breaux C, Hartlage S, Gehlert S. Relationships of premenstrual dysphoric disorder to major depression and anxiety disorders: a re-examination. J Psychosom Obstet Gynaecol 2000;21:1724. 12. Kendler KS, Thornton LM, Gardner CO. Genetic risk, number of previ705

S. A. HARTLAGE et al.
ous depressive episodes, and stressful life events in predicting onset of major depression. Am J Psychiatry 2001;158:582 6. van den Akker OB, Eves FF, Stein GS, Murry RM. Genetic and environmental factors in premenstrual symptom reporting and its relationship to depression and a general neuroticism trait. J Psychosom Res 1995;39: 477 87. Treloar SA, Heath AC, Martin NG. Genetic and environmental influences on premenstrual symptoms in an Australian twin sample. Psychol Med 2002;32:2538. Kendler KS, Silberg JL, Neale MC, et al. Genetic and enviromental factors in aetiology of menstrual, premenstrual, and neurotic symptoms: a population-based twin study. Psychol Med 1992;22:85100. First MB, Spitzer RL, Gibbon M, Williams JBW(1996, February): Structured Clinical Interview for DSM-IV Axis I Disorders - Non-patient Edition (SCID-I/NP, Version 2.0). (Available from the Biometrics Research Department, New York State Psychiatric Institute, 722 West 168th Street, New York, NY 10032) Orvaschel H (1995, January): Schedule for Affective Disorder and Schizophrenia for School-Age Children Epidemiologic Version-5 Kiddie-SADS-E (K-SADS-E). (Available from Dr. Helen Orvaschel, Nova Southeastern University, Center for Psychological Studies, 3301 College Avenue, Ft. Lauderdale FL 33314) Pai MP, Paloucek FP. The origin of the ideal body weight equations. Ann Pharmacother 2000;34:1066 9. Warner P, Bancroft J. Factors related to self-reporting of the premenstrual syndrome. Br J Psychiatry 1990;157:249 60. Marvan ML, Escobedo C. Premenstrual symptomatology: role of prior knowledge about premenstrual syndrome. Psychosom Med 1999;61: 1637. Kish L. Survey sampling. New York: Wiley & Sons; 1965. Ekholm UB, Ekholm NO, Backstrom T. Premenstrual syndrome: comparison between different methods to diagnose cyclicity using daily symptom ratings. Acta Obstet Gynecol Scand 1998;77:5517. Schnurr PP. Measuring amount of symptom change in the diagnosis of premenstrual syndrome: psychological assessment. J Consult Clin Psychol 1989;1:277 83. Williams JBW, Gibbon M, First MB, Spitzer RL, Davies M, Borus J, Howes MJ, Kane J, Pope HG, Rounsaville B, Wittchen HU. The Structured Clinical Interview for DSM-III-R (SCID): multisite test-retest reliability. Arch Gen Psychiatry 1992;49:630 6. Segal DL, Hersen M, Van Hasselt VB. Reliability of the Structured Clinical Interview for DSM-III-R: an evaluative review. Compr Psychiatry 1994;35:316 27. Spitzer RL, Endicott J, Robins E. Research diagnostic criteria. Arch Gen Psychiatry 1978;35:773 82. Orvaschel H, Beeferman D, Kabacoff R. Depression, self-esteem, sex, and age in a child and adolescent clinical sample. J Clin Child Psychology 1997;26:285 89. Hosmer DW, Lemeshow S. Applied logistic regression. New York: John Wiley and Sons. Weber EU, Bockenholt U, Hilton DJ, Wallace B. Determinants of diagnostic hypothesis generation: effects of information, base rates, and experience. J Exp Psychol Learn Mem Cogn 1993;19:1151 64. Parry BL, LeVeau B, Mostofi N, Naham HC, Loving R, Clopton P, Gillin JC. Temperature circadian rhythms during the menstrual cycle and sleep deprivation in premenstrual dysphoric disorder and normal comparison subjects. J Biol Rhythms 1997;12:34 46. Kravitz HM, Ganz PA, Bromberger J, Powell LH, Sutton-Tyrrell K, Meyer P. Sleep difficulty in women at midlife: a community survey of sleep and the menopausal transition. Menopause 2003;10:19 28. Cohen LS, Miner C, Brown EW, Freeman E, Halbreich U, Sundell K, McCray S. Premenstrual daily fluoxetine for premenstrual dysphoric disorder: a placebo-controlled, clinical trial using computerized diaries. Obstet Gynecol 2002;100:435 44. Freeman EW, Rickels K, Sondheimer SJ, Polansky M. Differential response to antidepressants in women with premenstrual syndrome/ premenstrual dysphoric disorder: a randomized controlled trial. Arch Gen Psychiatry 1999;56:9329. Halbreich U, Bergeron R, Yonkers KA, Freeman E, Stout AL, Cohen L. Efficacy of intermittent, luteal phase sertraline treatment of premenstrual dysphoric disorder. Obstet Gynecol 2002;100:1219 29. Pearlstein T. Selective serotonin reuptake inhibitors for premenstrual dysphoric disorder: the emerging gold standard? Drugs 2002;62: 1869 85. Glick R, Harrison W, Endicott J, McGrath P, Quitkin FM. Treatment of premenstrual dysphoric symptoms in depressed women. J Am Med Womens Assoc 1991;46:1825. Yonkers KA, White K. Premenstrual exacerbation of depression: one process or two? J Clin Psychol 1992;53:289 92. Jensvold MF. Psychological aspects of womens health care: the interface between psychiatry and obstetrics and gynecology. In: Stewart DE, Stotland NL, editors. Psychological aspects of womens health care: the interface between psychiatry and obstetrics and gynecology. Washington, D.C.: American Psychiatric Press; 1993. p. 16592. Miller MN, Miller BE, Chinouth R, Coyle BR, Brown GR. Increased premenstrual dosing of nefazodone relieves premenstrual magnification of depression. Depress Anxiety 2002;15:48 51. Freeman EW, Sondheimer SJ, Rickels K. Gonadotropin-releasing hormone agonist in the treatment of premenstrual symptoms with and without ongoing dysphoria: a controlled study. Psychopharmacol Bull 1997; 33:3039.


28. 29.

14. 15. 16.






18. 19. 20. 21. 22. 23. 24.




37. 38.


25. 26. 27.



Psychosomatic Medicine 66:698 706 (2004)