Longitudinal Follow-Up of Bipolar Disorder in Women With Premenstrual Exacerbation: Findings From STEP-BD

Predictors of relapse and poor outcome for bipolar disorders include age at illness onset, presence of psychosis (1), psychiatric comorbidities (2, 3), residual mood symptoms (4–8), history of frequent episodes (2, 9, 10), and use of antidepressants (11, 12). In women with bipolar disorder, the postpartum (13) and the menopause transition (14) are periods of increased vulnerability to illness relapse. Hormonal fluctuations during the menstrual cycle may also influence the clinical course of bipolar disorder.

Mild premenstrual symptoms occur in approximately 80% of premenopausal women (15), although only 6% fulfill criteria for premenstrual dysphoric disorder, while another 19% have subsyndromal premenstrual dysphoric disorder (13). The association between bipolar disorder and premenstrual symptoms has not been closely studied. Epidemiological studies suggest that 5.7% of women with premenstrual dysphoric disorder meet criteria for bipolar I disorder, and 4.9% for bipolar II disorder (16), while 16% of women with premenstrual symptoms but not premenstrual dysphoric disorder have bipolar disorder (17). Furthermore, premenstrual exacerbation of major depressive disorder or bipolar disorder is reported retrospectively by 25%–68% of women (18, 19), and prospectively by 55%–65% (17), but has not been detected in several small studies of women with bipolar disorder (20–23).

Despite evidence that premenstrual exacerbation occurs in a substantial proportion of women with bipolar disorder, little is known about the clinical implications of this association. Reproductive hormones fluctuate across the menstrual cycle, and estrogen and other reproductive hormones modulate affective symptoms through their action in the CNS (24–26). Furthermore, fluctuating levels of estradiol and progesterone induce mood disturbance in women with, but not those without, premenstrual dysphoric disorder, which suggests that a subgroup of women are particularly vulnerable to negative effects of hormonal fluctuation on mood (27). By analogy, women with bipolar disorder and premenstrual exacerbation may have more frequent mood episodes or rapid cycling relative to those without premenstrual exacerbation. The purpose of this study was to determine whether women with bipolar disorder who have premenstrual exacerbation have a worse course of illness, as indicated by more illness episodes, a shorter time to relapse, and more persistent subsyndromal symptoms of depression and mood elevation, during a 1-year prospective follow-up period.

Method

Participants

Participants were from the Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD), a multicenter study designed to evaluate outcomes of patients with bipolar disorder over time (28). Of 2,194 women enrolled in STEP-BD, 1,211 women who were in the range of 18–40 years old at study entry were defined as the premenopausal-age cohort. Of those, 932 (77.0%) responded to two items about premenstrual mood changes: “Currently, do you typically experience depression symptoms around your menstrual period?” and “Currently, do you experience mood swings around your menstrual period?” Women who provided a positive response to one or both of these questions were categorized as having premenstrual exacerbation.

After exclusion of women with previous unspecified gynecological surgeries (N=152) and women for whom information was lacking about whether they had menstrual cycles (N=72), 706 women remained in the premenopausal-age cohort. Of those, sufficient data for this study were available for 293 women (65.2% of them with premenstrual exacerbation) who were followed prospectively for at least 1 year. A survival analysis was carried out using data from a second, overlapping group of 129 women (51.2% of them with premenstrual exacerbation) who were in recovered status at baseline and had at least one additional assessment during the follow-up period.

Procedures

A diagnosis of bipolar disorder was established using two questionnaires: the Affective Disorder Evaluation and the Mini International Neuropsychiatric Interview Plus–Version 5.0 (29). The Affective Disorder Evaluation includes the mood and psychosis modules from the Structured Clinical Interview for DSM-IV (SCID). STEP-BD used a modified scoring system that allows both categorical diagnosis based on the DSM-IV criteria, consistent with the SCID, and continuous scores for depression and mood elevation (30). Further details about STEP-BD methods are available elsewhere (30).

Illness status was determined using the Affective Disorder Evaluation at baseline and the Clinical Monitoring Form for mood disorders (30) at subsequent visits. Like the Affective Disorder Evaluation, the Clinical Monitoring Form consists of the mood modules from the SCID. It has good interrater reliability (23, 28), and its depression and mood elevation subscales correlate well with other established clinician-rated continuous measures of depressive and manic symptoms (30).

Using ratings from the Affective Disorder Evaluation or Clinical Monitoring Form, participants were categorized into one of eight clinical states (26). These clinical categories have been used widely in previous STEP-BD studies. Four syndromal states correspond to the DSM-IV mood states (depression, mania, hypomania, mixed). Patients who were euthymic (≤2 DSM-IV symptoms) for at least 1 week and less than 8 weeks were assigned the clinical status “recovering,” and those who were euthymic for 8 weeks or more, “recovered.” Those who did not meet criteria for any of these states but had three or four depression and/or mania symptoms were classified as being in a subsyndromal state. The subsyndromal states are defined as “roughening” when mood worsened after a recovered state was achieved and “continued symptomatic” when symptoms persisted after criteria were met for a syndromal episode. Subsyndromal states do not specify whether symptoms are depressive or manic, or both (28). Use of psychotropic medications, hormonal contraceptives, and other prescribed medications was documented at each study visit. Follow-up assessments were conducted at least every 3 months.

Illness relapse was defined as occurring if a woman who had been designated as recovered at baseline met criteria for a syndromal mood episode (depression, mania, hypomania, or mixed episode) on the Clinical Monitoring Form at a follow-up visit. A broader definition of relapse that included both syndromal episodes and the subsyndromal clinical roughening state was also examined (6). Timing of clinical status change was defined as the median time point between the previous evaluation with the Clinical Monitoring Form (or Affective Disorder Evaluation) when a participant was in a recovered clinical status and the first Clinical Monitoring Form evaluation documenting a change in clinical status. Rapid cycling was defined as four or more mood episodes during a 12-month period, consistent with DSM-IV criteria.

Data on reproductive life events were obtained from a self-report questionnaire developed for STEP-BD. This questionnaire included items about reproductive life events, menstrual cycle patterns, use of hormonal contraceptives, and 14 premenstrual symptoms (crying easily, food craving, bloating, breast tenderness, feeling overwhelmed or out of control, abdominal pain, lack of energy, decreased interest in activities, irritability/anger, anxiety/tension, mood swings, depression, insomnia/hypersomnia, difficulties concentrating).

Analysis

The occurrence of syndromal and subsyndromal mood episodes on prospective follow-up, as defined by the Affective Disorder Evaluation and Clinical Monitoring Form, between women with and without premenstrual exacerbation was compared using the Wilcoxon Mann-Whitney test for non-normally distributed continuous variables and the Cochran-Armitage test of trend for ordinal variables.

A Kaplan-Meier survival curve and log-rank comparison were used to compare time to relapse between women with and without premenstrual exacerbation who were in recovered status at baseline and who completed at least one additional assessment during the follow-up period. Cox regression models were built to compare time to relapse between those with and those without premenstrual exacerbation after adjusting for current psychotropic medications, comorbidity, and the number of DSM-IV-defined syndromal mood episodes in the previous year. Models were also adjusted for reproductive characteristics hypothesized to be potential confounders (current use of hormonal contraceptive and irregular menstrual cycles) (13, 18, 19, 26, 31). The number of retrospectively reported bipolar episodes in the previous year was dummy coded as an ordinal variable (none, 1–3 episodes, and ≥4 episodes), consistent with conventional STEP-BD categorization (28), which includes a distinct category for rapid cyclers.

Symptoms of depression and mood elevation on the Clinical Monitoring Form were evaluated separately as response variables in repeated-measures mixed models to determine the effect of premenstrual exacerbation on mood symptoms in models that also included time (in months) and the interaction between premenstrual exacerbation and time. The final model was chosen after a structural analysis of variance and autoregressive (AR1) model covariance structure was performed. The repeated structure was defined by monthly measurements. Subjects were used as the random effect.

To adjust for use of psychotropic medications while restricting the number of covariates that could be supported by the model, lithium, anticonvulsants, and second-generation antipsychotics were grouped together, and all selective serotonin reuptake inhibitors (SSRIs) and serotonin-norepinephrine reuptake inhibitors (SNRIs) were grouped together.

Analyses were conducted using SPSS, version 15.0, and the R software package. All statistical analyses were conducted using an alpha of 0.05 (two-sided).

Approvals

All participants provided written informed consent after receiving a complete description of the study. Study procedures were approved by the institutional review board of each participating STEP-BD site (28).

Results

Premenstrual exacerbation was reported by 191 (65.2%) women who were followed for a 1-year period. Women in the premenstrual exacerbation group endorsed more individual premenstrual mood symptoms than did those without premenstrual exacerbation (median=9, interquartile range=6–11, compared with median=3, interquartile range=2–4; p<0.001) and were more likely to report that premenstrual symptoms interfered with work or school (78.1% compared with 23.7%, p<0.001), supporting the classification of women into our predefined categories of those with and without premenstrual exacerbation.

There were no significant differences in sociodemographic characteristics between those with and those without premenstrual exacerbation (Table 1). The median age was 30 years (interquartile range=25–36) for both groups together. Participants were mostly Caucasian, and most had at least a high school education; one-third were married, and half were employed full-time. Women with bipolar I disorder constituted approximately 60% of both groups, and approximately half of women in both groups met criteria for a comorbid psychiatric illness. However, other aspects of psychiatric illness history differed between groups. At study entry, women with premenstrual exacerbation were less likely to be in a recovered clinical status (14.1% compared with 34.3%, p<0.01) and were more likely to be in a depressive episode or subthreshold illness status. The retrospectively reported number of bipolar episodes during the previous year was higher in the group with premenstrual exacerbation than in the group without, such that women in the premenstrual exacerbation group were more likely to meet criteria for rapid cycling (46.1% compared with 32.4%, p=0.04). While the proportion of women using psychotropic medications did not differ between groups, those without premenstrual exacerbation were almost twice as likely to be using SSRIs or SNRIs (p<0.01).

The groups did not differ significantly in the proportion reporting irregular menstrual cycles (28.3% and 21.6% in the groups with and without premenstrual exacerbation, respectively) or using hormonal contraceptives (23.6% and 27.7% in the groups with and without premenstrual exacerbation, respectively). Among the women who reported regular menstrual cycles and provided information about the date of their last period (N=117 and N=60 among those with and without premenstrual exacerbation, respectively), there were no significant differences between groups in the proportion of those who completed the questionnaire during the premenstrual phase (44.4% and 46.7% among those with and without premenstrual exacerbation, respectively), indicating that the specific menstrual phase during which women completed study procedures could not explain differences in perceptions of symptoms.

Frequency of Mood Episodes During 1-Year Follow-Up

There was no difference between groups in the number of assessments conducted during follow-up (median=9.0, interquartile range=7.0–11.0, compared with median=9.0, interquartile range=6.75–11.0) (Table 1). Women with premenstrual exacerbation had more mood episodes overall than did those without (median=1, interquartile range=1–2, compared with median=1, interquartile range=0–2; p=0.02), including more depressive episodes (median=1, interquartile range=0–1, compared with median=0, interquartile range=0–1; p<0.001), but they did not have more episodes of mania or hypomania (median=0, interquartile range=0–1 for both groups). Differences remained when episode frequency was categorized (p=0.01); women reporting premenstrual exacerbation were more likely to have 1–3 episodes per year (73.8% compared with 55.9%) and less likely to have no episodes (24.1% compared with 42.2%) during follow-up. However, women with premenstrual exacerbation were not more likely than those without to have rapid cycling (2.1% compared with 2.0%) when rapid cycling was assessed prospectively.

Survival Analysis

Time to relapse was compared among 129 women (66 with premenstrual exacerbation and 63 without) who were in recovered status at baseline and had at least one additional assessment during the follow-up period. This group included 62 women from the 1-year prospective analysis who were followed for an entire year and in a recovered clinical status at baseline.

The sociodemographic, psychiatric, and reproductive characteristics of the group included in this analysis were similar to those of the 1-year follow-up group (Table 1). Women with premenstrual exacerbation endorsed more individual premenstrual symptoms than did those without (median=9, interquartile range=7–11, compared with median=3, interquartile range=2–5; p<0.001) and were more likely to report that premenstrual symptoms interfered with work or school (69.7% compared with 14.3%, p<0.001).

There were no significant differences between groups in sociodemographic characteristics or in number of follow-up assessments (Table 1). Women with premenstrual exacerbation reported more illness episodes in the year prior to study entry (p=0.03) and were more likely to meet criteria for rapid cycling on retrospective reporting (28.8% compared with 10.5%, p=0.05). However, neither group had any participants who met criteria for rapid cycling on prospective follow-up. Use of SSRIs or SNRIs did not differ between groups. There was no difference between groups in the proportion of women who had irregular menstrual cycles (28.8% and 25.4% in the groups with and without premenstrual exacerbation, respectively) or were using hormonal contraceptives (18.2% and 23.8% in the groups with and without premenstrual exacerbation, respectively).

Time to Relapse as a Function of Premenstrual Exacerbation

No significant difference was seen between groups in time to relapse to a syndromal episode (p=0.36; Figure 1A). However, when the survival analysis included subsyndromal episodes as part of a broader definition of relapse (6), women with premenstrual exacerbation relapsed sooner on average than did those without (p=0.02; Figure 1B). Using this broader definition of relapse, the median time to relapse was 4.5 months in the group with premenstrual exacerbation, compared with 8.5 months in the group without. There was a statistical trend (p=0.05) in the type of episode into which women relapsed, with those with premenstrual exacerbation more likely to relapse to a subsyndromal episode (60.9% compared with 36.7%), less likely to relapse to one of the four syndromal states in general (10.9% compared with 30.0%), and equally likely to relapse to a syndromal depressive episode (28.3% compared with 33.3%). No significant differences between groups were seen in the proportion of women changing medications before they relapsed (16.9% compared with 11.1%).

In Cox regression models comparing the hazard of relapse to a syndromal or subsyndromal episode, premenstrual exacerbation predicted a shorter time to relapse during follow-up (hazard ratio=1.7, 95% confidence interval [CI]=1.04–2.8, p=0.04) after adjusting for key covariates (hormonal contraceptive use, SSRI or SNRI use, mood-stabilizer or second-generation antipsychotic use, psychiatric comorbidity, and irregular menstrual cycles). When the number of illness episodes (none, 1–3, or ≥4) reported retrospectively during the previous year was added to the model, the significant effect of premenstrual exacerbation on time to relapse was lost (hazard ratio=1.4, 95% CI=0.8–2.4, p=0.19) and the number of retrospectively reported episodes during the previous year became the only significant predictor of a greater hazard for relapse (hazard ratio=3.9, 95% CI=1.7–8.7, p=0.001). SSRI or SNRI use was not a predictor of relapse in these models.

Symptom Severity During Follow-Up as a Function of Premenstrual Exacerbation

Because of the preponderance of subsyndromal episodes in the premenstrual exacerbation group, we conducted a post hoc random-effects repeated-measures analysis to determine overall symptom severity during follow-up. For depressive symptoms (Figure 2A), there was a significant effect of premenstrual exacerbation and time, but no significant effect of the time-by-premenstrual exacerbation interaction (F=1.06, p=0.49). For mood elevation (Figure 2B), there was a significant effect of premenstrual exacerbation, but not of time or the time-by-premenstrual exacerbation interaction. These results indicate that premenstrual exacerbation adversely affects the severity of both depression and mood elevation symptoms.

Discussion

In this longitudinal study, we found that premenopausal women with bipolar disorder who reported premenstrual exacerbation of their mood disorder had more frequent mood episodes, but not specifically a rapid-cycling course, during a 1-year follow-up period. Although time to relapse to a syndromal episode did not differ between women with premenstrual exacerbation and those without during follow-up, time to relapse in those with premenstrual exacerbation was half that of those without when relapse was defined more broadly as a syndromal or subsyndromal episode. Women with premenstrual exacerbation also had more severe depressive and mood elevation symptoms during follow-up, reflecting greater symptom burden. Taken together, these findings suggest that premenstrual exacerbation may be a marker for a more recurrent and symptomatic course of illness among premenopausal women with bipolar disorder.

This is the first large study to prospectively investigate the impact of premenstrual exacerbation on the course of bipolar disorder. Our results suggest that a large proportion of women with bipolar disorder experience mood destabilization, possibly because of fluctuations in reproductive hormones across the menstrual cycle (24–26, 32). While direct causal evidence is not available, it is possible that the susceptibility of mood to fluctuating hormone levels results in greater mood instability in general. Sensitivity to hormonal fluctuations could result in more frequent episodes, but in this study it did not appear to influence the occurrence of rapid-cycling bipolar disorder per se. Vulnerability of mood to hormonal fluctuation may also result in a more rapid subsyndromal relapse and higher levels of depressive and mood elevation symptoms.

The association of premenstrual exacerbation with subsyndromal episodes and more severe symptom levels suggests that hormonal sensitivity may result in chronic and persistent mood instability even in the absence of a full-blown illness episode. We included subsyndromal mood episodes in the definition of relapse because residual symptoms have been associated with reduced functional impairment and higher risk for relapse (4–6). Our results suggest that women who self-identify as having premenstrual exacerbation may therefore have a worse course of illness, including more frequent relapse to subsyndromal episodes, greater symptom burden, and functional impairment, in the absence of a rapid-cycling pattern. Rapid cycling is a complex mood condition that may be better explained by other factors, such as genetic predisposition, environmental exposures, and treatment (1, 33, 34).

Several studies have shown that a prior history of frequent episodes is a predictor of subsequent episode of bipolar disorder (1, 35). Our survival analysis indicates that the significant effect of premenstrual exacerbation on time to relapse was no longer observed after adjustment for the number of illness episodes reported during the previous year. There are several interpretations of this finding. First, it is plausible that women reporting premenstrual exacerbation retrospectively endorse more frequent episodes because they consider premenstrual breakthrough of symptoms to be an illness episode rather than premenstrual exacerbation. Supporting this notion is that rapid cycling was reported more frequently by women with premenstrual exacerbation than by those without on retrospective assessment; however, prospective evaluation did not support this difference and instead revealed that rapid cycling was not more common among women with premenstrual exacerbation. Other researchers have similarly found that rapid cycling occurs less commonly when evaluated prospectively (21, 36) and that premenstrual exacerbation is not associated with rapid cycling (37, 38).

Women with premenstrual exacerbation were not less likely to use psychotropic medications in general, but they were less likely than those without to receive treatment with an SSRI or SNRI. These agents are highly effective in the treatment of premenstrual dysphoric disorder (15), and although their role in treating premenstrual exacerbation symptoms in women with bipolar disorder remains unclear (11, 12, 23), it is possible that the more widespread use of SSRIs and SNRIs in recent years protected women without premenstrual exacerbation in our study from developing premenstrual exacerbation. In addition, women with premenstrual exacerbation may have been less likely to receive prescriptions for SSRIs or SNRIs because of concern about inducing rapid cycling, given that they already had more frequent episodes (11, 39). Despite the complex relationship between SSRI or SNRI use and premenstrual exacerbation, it is notable that SSRI or SNRI use was not a significant predictor of relapse, and adjusting for use of these agents did not alter the impact of premenstrual exacerbation on time to relapse.

This study is limited by the use of retrospective reporting of mood symptoms in relation to the menstrual cycle to classify participants into those with and those without premenstrual exacerbation. The validity of retrospective reports of premenstrual exacerbation in women with mood disorders has not been investigated, but studies of women with premenstrual dysphoric disorder demonstrate that retrospective reporting typically overestimates the prevalence of the disorder (15). Nonetheless, the proportion of women who reported premenstrual exacerbation in our study is consistent with what has been observed in several other studies of bipolar disorder (18, 19, 40), including studies that document premenstrual exacerbation with prospective daily monitoring (19, 40). As a result of the retrospective determination of premenstrual exacerbation, we cannot definitively establish whether some women who are prone to multiple relapses have been misclassified as having premenstrual exacerbation because they have attributed relapses to premenstrual exacerbation rather than an illness episode. However, the premenstrual exacerbation group endorsed many more classic physical and psychological premenstrual symptoms than did the group without premenstrual exacerbation, providing further supporting evidence that they experience mood symptoms as part of a pattern of premenstrual exacerbation.

In summary, the results of this longitudinal study indicate that women reporting premenstrual exacerbation have more frequent illness episodes, relapse more rapidly, and have greater symptom severity. Our findings suggest that reports of premenstrual exacerbation may be a potential marker for a more severe clinical phenotype in reproductive-age women with bipolar disorder. The results of this study reinforce the importance of characterizing the fluctuations of mood symptoms across the menstrual cycle in women with bipolar disorder.