Delayed sleep timing and circadian rhythms in pregnancy and transdiagnostic symptoms associated with postpartum depression

Women’s sleep is affected by pregnancy as early as the first trimester (for a review of sleep patterns in the peripartum, see refs. ^1,2). As pregnancy progresses, self-reported sleep quality worsens, with decreases in reported sleep duration and sleep efficiency³. In addition, over 40% of pregnant women endorse insomnia in the first trimester, and this increases to over 60% by the third trimester⁴. Polysomnographic data from the third trimester has also shown increased wake after sleep onset and decreased rapid eye movement (REM) sleep compared to non-pregnant women⁵. There is acute sleep loss seen on the nights preceding and following the birth, regardless of time of day of the birth or manner of delivery⁶. Following the birth, sleep quality continues to deteriorate, with worsened sleep efficiency⁷ and further sleep loss at night, although some studies show relatively stable total sleep duration with the inclusion of scattered daytime naps⁸. Although the sleep duration may be intact, fragmentation of sleep can still cause significant daytime sleepiness in the postpartum⁹. By the end of the so-called “fourth trimester,” that is, first 3 months postpartum, the mothers’ sleep has generally begun to improve⁷, but does not necessarily return to pre-pregnancy levels¹⁰.

The association between sleep disturbances and psychiatric illness, including depression, anxiety, mania, and psychosis, in the non-peripartum population is well established (for reviews, see refs. ^11–14). Even in healthy individuals, acute sleep deprivation is associated with onset of mood disturbance and eventually psychotic features can emerge as sleep deprivation extends¹⁵. Circadian dysregulation also has been implicated in psychiatric disorders (for a review, see ref. ¹⁶), including associations with chronotype and later sleep timing. For instance, greater “eveningness,” that is, being a night owl, is prognostically worse from a psychiatric standpoint^17,18. The underlying mechanisms are not fully understood, but may include alterations in cognitive¹⁹, psychomotor²⁰, or emotional processing²¹, which in turn contribute to symptoms in a variety of psychiatric disorders.

There is overlap between the sleep alterations of pregnancy and the sleep alterations found in psychiatric disorders. Depression, mania, and obsessive-compulsive disorder (OCD) are particularly important disorders in perinatal women because they occur frequently in this population^22,23, commonly overlap²⁴, and are related to suicide in new mothers²⁵. For example, sleep disturbances found both in pregnancy and depression include decreased sleep quality and sleep efficiency and increased sleep onset latency²⁶. According to a 2015 Cochrane review, there is evidence for the association of self-reported poor sleep in the peripartum and postpartum depression (PPD), but the evidence for the association with peripartum anxiety and peripartum psychosis is not robust enough to be conclusive²⁷. Women with bipolar disorder who reported previous manic episodes triggered by sleep loss are more than twice as likely to experience postpartum psychosis than women with bipolar disorder who never had episodes triggered by sleep loss²⁸, suggesting that there may be a subgroup of women who are particularly vulnerable to perinatal sleep disturbance. The data on OCD and obsessive-compulsive (OC) symptoms and sleep disturbance in the peripartum is limited. In a non-peripartum population, sleep disturbance has been significantly associated with OCD^29,30. Additionally, there is an increased prevalence of delayed sleep-wake phase disorder in people with severe OCD than in the general population³¹. Research on sleep in pregnant and postpartum women and its relationship to psychiatric symptoms is limited and has focused mostly on self-reported sleep data and on sleep duration rather than sleep timing. Additionally, the literature is more robust regarding depressive symptoms; less information is available for transdiagnostic manic and OC symptoms.

In this study, we collected longitudinal, objectively measured sleep and circadian measures, including actigraphically estimated sleep onset and sleep offset times, sleep efficiency and sleep duration, circadian phase (measured with salivary dim light melatonin onset, DLMO), and relationship between sleep and circadian rhythms (defined as phase angle between salivary DLMO and sleep onset,) as well as transdiagnostic psychiatric symptoms. We conceptualize circadian rhythms and sleep as a foundation for biobehavioral regulation, with subsequent influences on many higher-order symptoms and disorders. Given the preponderance of sleep disturbance in a wide variety of psychiatric disorders³², it is worthwhile to study this topic in a transdiagnostic fashion. Furthermore, as the peripartum is a time of high frequency and amplitude of psychiatric symptoms^33,34, a transdiagnostic approach to studying psychiatric symptoms in postpartum women is particularly salient. From the perspective of the Research Domain Criteria (RDoC)³⁵, we are examining intersections among the following domains: Arousal and Regulatory Systems, Negative Valence Systems, and Cognitive Systems. Our goal is for this approach to begin to refine the phenotypes associated with sleep disturbance and manifestations of psychiatric illness, particularly in the peripartum. Based on the rich literature demonstrating associations between evening chronotype and poorer mental health outcomes^17,18, we hypothesized that later sleep timing and circadian phase position during pregnancy—operationalized by sleep onset time and time of salivary melatonin onset—would be associated with greater manifestation of psychiatric and psychologic symptoms.

The present analyses are derived from the final data set of a study with two previous, interim publications^36,37. We analyzed data from 51 participants in whom actigraphy data were available from pregnancy. This sample size was selected based on our ability to enroll and retain participants from this difficult-to-recruit population. Post hoc power calculations showed that we maintained ~73% power to detect large effect sizes at a two-tailed α = 0.05. We recruited pregnant women with a past history of major depression or bipolar disorder; they underwent an informed consent process with signed consent, and received monetary compensation for participating. To be included, women had to be between the ages of 18 and 40 years and meet the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition (DSM-IV) criteria for history of major depressive disorder (MDD) or bipolar disorder (I or II), diagnosed by the Structural Clinical Interview of DSM-IV Disorders³⁸. We used the following exclusion criteria: current mood episode at 33 weeks of gestation (calculated by last menstrual period), diagnosis of an Axis I disorder other than an anxiety disorder (e.g., current substance dependence), a primary sleep disorder, employment as a shift worker, use of sedatives for insomnia, high-risk pregnancy, and/or plan for the infant to have a nighttime caregiver other than the participant and her partner. Antidepressant use, parity, or plan to breast or bottle feed did not exclude women from participating. The study was approved by institutional review boards at Rhode Island Hospital and Women & Infants Hospital and was carried out in accordance with the Declaration of Helsinki.

Fifty-one women were included in these analyses with a mean age of 28.2 ± 5.2 years (range: 18–38 years). Forty-four had a past history of MDD (86.3%) and seven had a past diagnosis of bipolar disorder. Eighteen (35.3%) were first-time mothers and the median number of children among experienced mothers was 1 with a range of 1–4. Fifteen women reported taking at least one medication with the potential to affect sleep or mood during the third trimester, including antidepressants (n = 6), benzodiazepines (n = 1), levothyroxine (n = 1), pain relievers, specifically acetaminophen (n = 5), β-blockers (1), inhaled β-agonists (1), anti-histamines, specifically ranitidine and loratadine (n = 4), and/or anti-emetics (n = 1). The sample was ethnically, racially, and socioeconomically diverse. The majority of women (70.6%) were “neither” types on Horne–Östberg Morningness–Eveningness Questionnaire. Six (11.8%) were “moderate morning” types and nine (17.9%) were “moderate evening” types. See Table 1 for additional participant demographics.

Average actigraphically estimated sleep onset time during the third trimester (pregnancy week 33) ranged from 9:01 p.m. to 1:58 a.m., with a median of 11:27 p.m. To test our hypotheses about associations between delayed sleep timing and transdiagnostic psychiatric symptoms, we used the median sleep onset time of 11:27 p.m. during the third trimester of pregnancy to split the sample into two groups: “early sleep” and “late sleep.” Repeated-measures ANOVA revealed that, as expected, there was a main effect of group in sleep onset time (F = 43.214, p < 0.001). Sleep offset time also differed significantly between groups (F = 17.163, p < 0.001), with group differences averaging more than 1 h at all four study weeks. Although the average Horne–Östberg Morningness–Eveningness scores were in the “neither” range for both groups, they differed significantly between groups (53.9 vs. 44.9, t = 5.4, p < 0.001), with the late group showing more evening preference. Furthermore, all six moderate morning types were in the “early sleep” group and all nine moderate evening types were in the “late sleep” group. Most women fell into the same group at all three postpartum time points (week 2: 68.6%; week 6: 76.4%; week 16: 72.9%). Antidepressant use was divided evenly between groups; three women in the “early sleep” group reported taking sertraline. In the “late sleep” group, one woman took sertraline, one took nortriptyline, and one took bupropion and clonazepam.

Maternal sleep measures for “early sleep” and “late sleep” groups at 33 weeks of gestation, and 2, 6, and 16 weeks postpartum are shown in Table 2. As would be expected across the perinatal period, a main effect of study week was observed for sleep offset time (F = 9.784, p < 0.001), time in bed (F = 6.007, p = 0.002), sleep duration (F = 7.727, p < 0.001), and sleep efficiency (F = 28.605, p < 0.001). Within-subject contrasts revealed significant differences between sleep duration and sleep efficiency at postpartum weeks 2 and 6 compared with pregnancy, and significant differences in sleep offset time and time in bed between postpartum weeks 2 and 16 compared with pregnancy. Finally, we observed significant week by group interactions for sleep onset time (F = 4.316, p = 0.009), time in bed (F = 3.451, p = 0.024) and sleep duration (F = 3.34, p = 0.028). Planned post hoc t tests showed that sleep onset differed significantly between the “early sleep” and “late sleep” groups at all four study weeks, whereas the interaction effects for time in bed and sleep duration between groups were driven by significant differences only during week 33 of pregnancy. During pregnancy, the “late sleep” group spent less time in bed (8.0 vs. 8.7 h, t = 2.920, p = 0.005) and had shorter estimated sleep durations (6.4 vs. 7.3 h, t = 2.693, p = 0.01) than the “early sleep” group.

Melatonin onset data were available for 45 participants at 33 weeks of pregnancy and 37 participants at 6 weeks postpartum. Clock time of DLMO did not differ significantly between “early sleep” and “late sleep” groups (see Table 2), but pregnant women in the “late sleep” group had significantly longer phase angles between DLMO and sleep onset time during pregnancy than the early group (2.7 ± 1.2 vs. 1.5 ± 1 h, t = −3.487, p = 0.001), indicating that they were going to sleep later with respect to their internal circadian clock. Although phase angle did not differ between groups at 6 weeks postpartum, women in the “late sleep” group had a significant shortening of their phase angle from the third trimester to 6 weeks postpartum (t = 2.371, p = 0.024), whereas phase angle did not change in the “early sleep” group. On average, “late sleep” mothers’ sleep onset time occurred 55 min earlier with respect to circadian phase at 6 weeks postpartum than during pregnancy.

The results of the transdiagnostic psychiatric symptoms are shown by study week in Table 2.

There was an overall effect of group (F = 5.494, p = 0.024), but no significant effect of time and no time × group interaction on the Highs scale. Although Highs scores were higher in the “late sleep” group at all weeks (Fig. 1a), this difference was statistically significant only at postpartum week 2 (t = −2.786, p = 0.008), with the “late sleep” group showing significantly more manic symptoms. Highs score at 2 weeks also was correlated significantly with longer phase angle at 33 weeks (r = 0.355, p = 0.018; Fig. 2a), consistent with the notion that delayed sleep timing with respect to circadian phase in pregnancy is associated with symptoms such as racing thoughts, difficulty concentrating, and feeling more active than usual in the early postpartum period.

OCI score did not differ significantly between the early and late sleep groups at 33 weeks of pregnancy or 6 weeks postpartum (Fig. 1b), however, OCI score at 6 weeks postpartum was correlated with longer phase angle at 33 weeks of pregnancy (r = 0.327, p = 0.032, Fig. 2b), indicating that women who went to sleep later with respect to circadian phase during third trimester showed more OC symptoms at 6 weeks postpartum.

ANOVA showed no effect of time and no group × time interaction on depressive symptoms measured by the HAM-D17, but the main effect of group approached significance (F = 4.033, p = 0.051). Post hoc t tests showed that compared with the “early sleep” group, the “late sleep” group had significantly higher HAM-D17 scores at postpartum week 2 (t = −2.576, p = 0.013, Fig. 1c). Longer phase angle at 33 weeks did not correlate significantly with HAM-D17 scores.