Sleep–wake timing and chronotype in perinatal periods: longitudinal changes and associations with insomnia symptoms, sleep‐related impairment, and mood from pregnancy to 2 years postpartum

Across the perinatal transition, existing research focuses mainly on significant changes in sleep duration and quality, neglecting sleep timing. This study investigated change trajectories of sleep timing and chronotype from late pregnancy to 2 years postpartum and examined longitudinal associations of chronotype with symptoms of insomnia, daytime sleep‐related impairment, and mood. Data were from a two‐arm randomised controlled trial testing parent‐focused wellbeing interventions. Participants were a community sample of nullipara without severe sleep/mental health conditions. Participants self‐reported bedtime, rise‐time, chronotype, insomnia symptoms, sleep‐related impairment, depression, and anxiety at seven time points: gestation Weeks 30 and 35, and postpartum Months 1.5, 3, 6, 12 and 24. Trajectories were estimated using mixed‐effects models with continuous time, quadratic splines, and a knot at childbirth, controlling for age and group allocation. A total of 163 participants (mean [SD] age 33.35 [3.42] years) took part. Bedtime and rise‐times delayed during late pregnancy (~8 and ~20 min, respectively) but became progressively earlier (~20 and ~60 min, respectively) over the 2 postpartum years. Chronotype became more eveningness in late pregnancy, and more morningness after childbirth, however changes were small. Controlling for sleep duration and efficiency, greater morningness was associated with significantly less symptoms of insomnia and sleep‐related impairment over time (all p < 0.001); longitudinal associations between chronotype and symptoms of depression and anxiety were non‐significant (all p > 0.65). Sleep–wake timing and chronotype became progressively earlier from pregnancy to 2 years postpartum. Morningness chronotype may be sleep‐protective during the transition from pregnancy to parenthood. Mechanisms underlying these associations require further research.


| INTRODUCTION
Existing literature has extensively documented changes in sleep quality and duration throughout pregnancy and the postpartum period (Bei et al., 2015).Studies using objective and self-report measures have consistently shown that by the third trimester, sleep becomes significantly disturbed (Beebe & Lee, 2007;Hedman et al., 2002), with 70% of women reporting poor sleep quality (Sedov et al., 2018).These changes in sleep-wake behaviour intensify around (Bei et al., 2012) and after childbirth (Lee et al., 2000;Matsumoto et al., 2003;Mindell et al., 2015).Whilst some aspects of sleep (e.g., duration) improve over the early postpartum period (Signal et al., 2007), other aspects (e.g., sleep efficiency) remain compromised (Matsumoto et al., 2003).
Further, symptoms of insomnia, characterised by difficulty in initiating and/or maintaining sleep, are common during perinatal periods (Quin et al., 2022;Román-Gálvez et al., 2018;Sivertsen et al., 2015) and have been associated with daytime impairment (American Psychiatric Association, 2013).In one large study, >40% of women continued to experience clinically significant insomnia symptoms even at 2 years postpartum (Sivertsen et al., 2015).Past studies collectively demonstrate significant, and in some cases, long-lasting changes in certain aspects of perinatal sleep.
Despite the plethora of perinatal research, studies of perinatal sleep-wake timing are limited.One longitudinal study explored changes in self-reported sleep-wake timing at four time points from late pregnancy to 12-15 months postpartum among American first-time female parents (n = 38; Wolfson et al., 2003).Results showed that bedtimes remained consistent across the third trimester, 2-4 weeks postpartum, 3-4 and 12-15 months postpartum (mean 10:37 p.m.); however, risetimes were variable.Compared to late pregnancy, rise-times became significantly later in the immediate postpartum period, before becoming earlier at the mid to latter postpartum time points.Despite a small sample size and long lapses between measurement time points, these results point to meaningful changes in sleep-wake timing across the peripartum that warrant future research.
Although scarce, there exists some evidence of circadian rhythm disruption during perinatal periods.Changes to the circadian rhythm in the postpartum period include reduced circadian amplitude and alterations to 24-h melatonin profiles (Parry et al., 2008;Thomas & Burr, 2006).Phase delays have also been observed in postpartum women with a history of major depressive disorder (MDD; Sharkey et al., 2013).Additionally, some research has shown significant changes in actigraphy-assessed diurnal activity rhythms after childbirth, which, despite stabilising around the fourth postpartum month, continue to differ significantly from non-pregnant controls (Gallaher et al., 2018).Potential contributors to changes in perinatal circadian rhythms include inconsistent sleep timing, increased overnight awakenings and daytime napping (Bei et al., 2012;Matsumoto et al., 2003), as well as low levels of ambient light exposure experienced during pregnancy (Sharkey et al., 2013) and the early postpartum (Lee & Kimble, 2009;Sharkey et al., 2013;Tsai et al., 2009).Chronotype (self-reported diurnal preference for alertness and activities), or morningness-eveningness, is usually measured via self-report (Wirz-Justice, 2007) and is correlated with biological markers of circadian rhythm (Kantermann et al., 2015).Chronotype could provide insight into circadian-related functioning without more burdensome biological assessment of circadian phase, yet trajectories of perinatal chronotype have seldom been described.
Perinatal sleep and circadian rhythm changes have been associated with a range of undesirable outcomes, including impaired daytime functioning (Lee & Kimble, 2009), fatigue (Insana et al., 2011), and most notably, symptoms and higher risk of perinatal depression and anxiety (Emamian et al., 2019;González-Mesa et al., 2019;Sharkey et al., 2013).Preliminary studies have also shown associations between chronotype and perinatal wellbeing.In one study of 12 women with a history of MDD, later objectively assessed circadian phase and self-reported eveningness chronotype in the third trimester were associated with higher depressive symptoms concurrently, and at the second and sixth week postpartum (Sharkey et al., 2013), mirroring findings observed in non-perinatal populations (Walker et al., 2020).Conversely, greater perinatal morningness chronotype has been associated with less mood disturbance (Obeysekare et al., 2020), better psychomotor functioning (McBean & Montgomery-Downs, 2013) and preferential patterns of diurnal fatigue (McBean & Montgomery-Downs, 2015).In non-perinatal populations, circadian rhythm disturbance has been associated with symptoms of insomnia (Flynn-Evans et al., 2017), thus potential changes in circadian functioning experienced in perinatal periods may represent an unexplored contributing factor to perinatal insomnia.
There are several gaps in existing literature.Firstly, longitudinal changes in sleep-wake timing and chronotype across perinatal periods, including during the transition from pregnancy to postpartum, are not well documented.Few studies have specifically assessed sleep-wake timing and chronotype, with some studies treating chronotype as a trait variable (Morales-Muñoz et al., 2019), whilst others have dichotomised women into 'early' or 'late' sleepers (Obeysekare et al., 2020), which does not allow for specific and continuous descriptions.Additionally, past studies are limited by small and highly selected samples (e.g., those with a mood disorder).Further, whether chronotype may be related to perinatal sleep, daytime functioning, and mood throughout perinatal periods are not well understood.
Using data from a large longitudinal intervention study from the third trimester of pregnancy to 2 years postpartum, the present study aimed to: (i) explore trajectories of self-reported sleep-wake timing and chronotype (exploratory aim), and (ii) investigate the associations between chronotype and symptoms of insomnia, daytime functioning, and mood, independent of sleep duration, quality, and relevant covariates; it was hypothesised that a more morningness chronotype would be associated with less self-reported symptoms of insomnia and lower sleep-related daytime impairment, depression and anxiety.

| METHODS
This study used data from a longitudinal randomised controlled trial undertaken in Melbourne, Australia from May 2016 to December 2019.Detailed methodology can be found in the published protocol (Bei et al., 2019).The trial was registered with Australian and New Zealand Clinical Trials Registry (ACTRN12616001462471) and received ethics approval from the Royal Women's Hospital (RWH) and Monash University Human Research Ethics Committees.

| Participants
A community sample of expectant first-time female parents enrolled in RWH childbirth education were invited via email to participate in the study.Inclusion criteria were: nulliparity, age ≥ 18 years, singleton pregnancy, ability to communicate in English, and regular access to email and internet.Exclusion criteria were: sleep-affecting pharmacotherapy and medical conditions, regular night shift work, severe sleep disorders (i.e., sleep apnea, restless legs syndrome, periodic limb movement disorder, circadian rhythm sleep-wake disorders) and severe psychiatric disorders (i.e., MDD, post-traumatic stress disorder, bipolar disorder, panic disorder if associated with frequent nocturnal panic attacks, psychotic disorders, substance use disorders).Sleep and psychiatric disorders were assessed via structured clinical interview via telephone as detailed in the protocol (Bei et al., 2019).Nullipara were recruited in the original study to reduce confounders (e.g., the impact of having older children on sleep).

| Procedure
Participants provided informed consent online and were screened for eligibility through a structured telephone interview.Those eligible were randomised 1:1 to a 'healthy sleep' or 'healthy diet' group.
Those in the healthy sleep group received cognitive, behavioural, and psychoeducational strategies for improving their own sleep and supporting the sleep of their infant; those in the healthy diet condition received education on healthy eating habits, weight management, and nutritional suggestions specific to different perinatal milestones.

Perinatal sleep questions. A modified version of the Consensus Sleep
Diary (Carney et al., 2012) was used to assess self-reported sleepwake behaviour over the past week including bedtime (i.e., getting into bed intending to sleep), rise-time (i.e., get out of bed for the day), and overnight awakenings (infant-related or spontaneous) to ascertain sleep-wake timing, duration (i.e., total sleep time) and efficiency (i.e., proportion of total sleep time to time spent in bed).nitely evening' (4-7; Chelminski et al., 2000).Utility (Tonetti & Natale, 2019) and validity (Natale et al., 2006) of the rMEQ have previously been demonstrated.Scores on rMEQ were included as a continuous variable in analyses.
Symptoms of insomnia.The Insomnia Severity Index (ISI) is a seven-item measure of insomnia symptom severity (Bastien et al., 2001).Items (e.g., 'difficulty falling asleep', 'difficulty staying asleep') are rated on a 5-point Likert-type scale for the past weeks.
Total scores range between 0 and 28, with higher scores indicating more severe insomnia symptoms (Morin & Barlow, 1993).
Sleep-related impairment.Sleep-related impairment was measured using the Patient-Reported Outcomes Measurement Information System-Sleep Related Impairment (PROMIS-SRI) -Short Form (Yu et al., 2012).
The PROMIS-SRI is an eight-item measure of impairment in daytime functioning due to sleep disturbance (e.g., daytime sleepiness and functioning).Respondents rate the frequency of eight items (e.g., 'I had problems during the day because of poor sleep') in the past days on a 5-point Likert-type scale ranging from 1 ('not at all') to 5 ('very much').

Mood. Mood was measured through PROMIS-Depression -Short
Form -8a and PROMIS-Anxiety -Short Form -8a (Pilkonis et al., 2011).Both PROMIS-Depression and PROMIS-Anxiety are eight-item measures of depressive and anxious symptoms respectively.Respondents are asked to rate the frequency of their experiences in the past days (e.g., 'I felt worthless', PROMIS-Depression; 'I felt anxious', PROMIS Anxiety) on a 5-point Likert-type frequency scale ranging from 1 ('never') to 5 ('always').
Social support.PROMIS Instrumental Support -Short Form -4a and Emotional Support -Short Form -4a (Hahn et al., 2014)  as well as PROMIS Emotional Support (e.g., 'I have someone to talk with when I have a bad day').All items are rated on a 5-point Likerttype scale, ranging from 1 ('never') to 5 ('always').
All PROMIS scales above result in a T-score, which has a population mean of 50 and standard deviation (SD) of 10.

| Data analysis
Changes in bedtime, rise-time, and rMEQ scores were each explored using mixed-effects models.Each model included random intercept ).These covariates were included so that we could examine the effects of rMEQ over and above other factors that may influence the outcome.The interaction term between group allocation and rMEQ was included as a predictor in all models and dropped if non-significant.Sensitivity analyses removing the six participants who had a second child between 1 and 2 years postpartum showed that all findings were held; these participants were included in final analyses for generalisability.
All analyses were conducted in R version 4.0.5 (R Core Team, 2021), and statistical significance was set at p < 0.05.Participants from the two groups did not differ significantly on baseline characteristics (all p ≥ 0.12).Post-baseline group differences on ISI, PROMIS-SRI, PROMIS-Depression, and PROMIS-Anxiety were reported in the trial's primary outcome paper (Bei et al., 2021).Briefly, compared to the healthy diet group, participants who received the sleep intervention had significantly lower ISI and PROMIS-SRI at 35 weeks' gestation (T2; all p ≤ 0.001) and 2 years postpartum (T7; all p < 0.05), but not other time points.Group differences on PROMIS-Depression and -Anxiety were non-significant at all time points (all p > 0.20) except at T5 where the healthy sleep group had somewhat higher scores on Depression ( p = 0.046).

Participants
Time-by-group allocation was a non-significant predictor for trajectories of bedtime, rise-time, and rMEQ (all p > 0.41); further, associations between rMEQ and ISI, PROMIS-SRI, -Depression, and -Anxiety also did not differ between group allocations (all p > 0.21).
Therefore, group allocation as a moderator was dropped, and the main effect of group allocation was included as a covariate in all following models.Chronotype initially became more evening during the late pregnancy period and prior to childbirth.After childbirth, there was a progressive shift towards more morning tendency that continued to 2 years postpartum.These changes in rMEQ scores were rather small in magnitude (<0.5 point), and both models estimated mean and 95% CI band fell in 'neither morning nor evening' types throughout all time points.Further, the random effects for time were significant ( p < 0.001), suggesting significant individual differences in the chronotype change trajectory.

| Associations between chronotype and selfreported sleep, daytime impairment, and mood
Mixed-effects models with chronotype predicting self-reported symptoms of insomnia, sleep-related daytime impairment, depression and anxiety are summarised in Table 2. Greater morningness was associated with significantly less symptoms of ISI and lower PROMIS-SRI over time (all p < 0.001).Associations between chronotype and symptoms of depression ( p = 0.97) and anxiety ( p = 0.66) were nonsignificant.
Among the covariates, longer total sleep time and higher emotional support stood out, being associated with significantly more positive scores on all four outcome measures (all p < 0.01), with one exception, whereby total sleep time was not a significant predictor of anxiety (p = 0.06).Higher sleep efficiency was associated with significantly lower ISI and PROMIS-SRI (all p < 0.001) but was not significantly associated with symptoms of depression ( p = 0.85) or anxiety ( p = 0.40).Having a current or past mental health condition and having lower instrumental social support was associated with significantly higher depression and anxiety symptoms (all p < 0.05); but these two covariates were not independently associated with ISI or PROMIS-SRI ( p = 0.14-0.88).In a large sample of relatively healthy nulliparous female parents, our longitudinal study found changes in sleep-wake timing and chronotype from the third trimester of pregnancy to 2 years postpartum.These changes included a delay in sleep timing and a change towards eveningness during late pregnancy, followed by advances in sleep timing and a tendency towards morningness chronotype after childbirth.Additionally, across the seven observational time points, greater morningness chronotype was associated with less symptoms of insomnia and lower daytime impairment but not with symptoms of depression or anxiety.
The pattern of change in sleep-wake timing observed in this study are generally consistent with a previous observation of progressively earlier rise-times from pregnancy to 12-15 months postpartum (Wolfson et al., 2003).In the present study, the delayed sleep timing and shift towards evening chronotype during late pregnancy may have been contributed to by beginning of parental leave and absence of morning commutes to work.We noted that during this period, risetime delayed more than bedtime ($20 versus 8 min).This could contribute to a longer sleep opportunity ahead of childbirth.Although rise-time became progressively earlier after childbirth, it remained later than pre-childbirth levels for most of the first postpartum year.This, along with a progressively earlier bedtime, allowed for longer sleep opportunities during the early postpartum period when nighttime sleep disruption was common.
At 2 years postpartum, bedtime showed a trajectory of returning to the pre-childbirth level (although still somewhat earlier than late pregnancy), whilst rise-time had advanced to be $60 min earlier than late pregnancy.In Australia, most parents start to return to employment at 1 year postpartum.The combined responsibilities of childcare and employment may contribute to earlier rise-time during latter part of the observational period in this study.Future research could provide insight to whether observed sleep-timing changes are related to changing life experiences such as employment.
Similar to past research assessing chronotype during pregnancy (Morales-Muñoz et al., 2019) and the early postpartum (Yamazaki, 2007), participants in our study mostly reported being 'neither morningness nor eveningness' chronotypes.Repeated assessment of chronotype allowed this study to longitudinally observe how chronotype changed over time, which was consistent with changes in both bedtime and rise-time.It is possible that later rise-time during late pregnancy delayed morning light exposure, leading to the shift towards more eveningness.During the postpartum period, childcare demands necessitating earlier rise-times, and thus earlier exposure to morning light, which may have advanced circadian phase and chronotype.Recent research has also found advances in chronotype related to increasing age (Druiven et al., 2021).However, without the assessment of objectively measured circadian timing, this remains simply a speculation.It is worth noting that the magnitude of average change in chronotype was small.Further, changes in chronotype ought to be interpreted in the context of significant individual differences in change trajectories across individuals (i.e., significant random effects).It is possible that some individuals may have become more eveningness, whilst others more morningness, thus future research is needed to better understand predictors of individual differences over time.
Over the observation period, more eveningness chronotype was associated with greater symptoms of insomnia and sleep-related impairment, and these effects were independent of sleep duration and quality.This adds to previous findings of poorer outcomes in the context of greater eveningness among non-perinatal populations, including greater insomnia symptoms (Flynn-Evans et al., 2017), and in perinatal populations, poorer psychomotor performance (McBean & Montgomery-Downs, 2013).Having a more eveningness tendency may make it more challenging to adjust to changing wake schedules in the context of perinatal demands such as variable infant wake-time and childcare as described earlier.It may also be possible that those with more eveningness chronotype may attempt to initiate night-time sleep at times earlier than preferred, which may lead to symptoms of insomnia (American Psychiatric Association, 2013).A T A B L E 2 Summary of mixed effects models with chronotype predicting outcomes controlling for covariates.potential mismatch between chronotype and sleep-wake timing may contribute to the associations between greater eveningness chronotype and greater sleep-related daytime impairment observed in our sample; future research is encouraged to investigate this further.
Chronotype was not associated with symptoms of depression or anxiety after adjusting for covariates.In this study, average scores of depression and anxiety symptoms were below the population mean, which may have caused a floor effect.Another explanation may lie in the complexity of perinatal mood.Anxiety and depression in the peripartum are multifaceted experiences with many associated factors (Buist et al., 2006).The factors adjusted for in analysis (i.e., sleep quality and duration, social support, mental health history, age) are known to impact perinatal mood (Bottino et al., 2012;González-Mesa et al., 2019;Kahn et al., 2014;Leahy-Warren et al., 2012).It is possible that chronotype was not significantly associated with mood beyond the contributions of these such factors.

| Limitations
First, sleep-wake timing was assessed via self-report and may have differences from objectively assessed bedtimes and rise-times.Similarly, we assessed chronotype rather than biological markers of circadian timing or light exposure and therefore were unable to assess changes in endogenous circadian timing.Second, the sample was relatively healthy, highly educated, and the effects of pregnancy/delivery complications or premature birth (in two individuals) were not examined, therefore, findings may not generalise to other populations such as those with sleep, mental health difficulties, or pregnancy/birth complications.Third, although frequent assessments were carried out during the first postpartum year, no repeated measures were taken in the first two trimesters of pregnancy, or between the first and second postpartum year; this precludes more detailed assessment of potential changes in sleep-wake timing and chronotype throughout pregnancy, and during the second postpartum year.As our sample consisted of a relatively small sample of first-time parents, results may not generalise to parents with more than one child.This study only collected sleep timing information for the primary nocturnal sleep periods, and timing for daytime naps were not examined.Although data collection was carried out across all seasons, we did not assess the potential impact of season on reported findings.Finally, although intervention group allocation was not associated with changes in chronotype or sleep timing, the sleep-and diet-related interventions delivered may impact on aspects of these changes described in the study.

| Implications
Findings from this longitudinal study add to existing literature by demonstrating that beyond sleep duration and quality, sleep-wake timing and chronotype also change from pregnancy to 2 years postpartum.
Future research could incorporate these additional dimensions when assessing sleep during the perinatal periods.Our findings suggest that gestational parents with more eveningness chronotype may be more vulnerable to symptoms of insomnia and sleep-related daytime impairments.Including the assessment of sleep-wake timing and chronotype in addition to sleep disturbance in perinatal care could help identify individuals who may have a higher risk of perinatal insomnia.
Future research is needed to examine whether experimentally promoting earlier sleep schedules via intervention such as bright light therapy (Terman et al., 2001) may lead to reduced insomnia symptoms and improved daytime functioning during the perinatal period.
).The rMEQ is a five-item self-report measure in which items (e.g., 'Considering only your own "feeling best" rhythm, at what time would you get up if you were entirely free to plan your day?') are responded to and scored individually, combining to give a total rMEQ score, with higher scores indicating greater morningness chronotype, whilst lower scores indicate more eveningness chronotype.Scores can be interpreted as: 'definitely morning'(22)(23)(24)(25), 'moderately morning' (18-21), 'neither' (12-17), 'moderately evening' (8-11), or 'defi- and time slopes (Aim 1).Continuous time was used and was defined as the duration, in weeks, between the date of questionnaire completion and the date of childbirth.Time for childbirth is therefore coded as '0', times for pregnancy are negative, and times for postpartum are positive values.To ensure non-linear trajectories were captured, model fit based on Akaike information criteria and Bayesian information criteria were compared for the following three configurations in each outcome (i.e., bedtime, rise-time, rMEQ): (a) models with both linear and quadratic time slopes, with the quadratic time slope dropped if nonsignificant.This configuration explores whether there was an overall quadratic or linear trajectory; (b) models with quadratic splines with one knot placed at childbirth (time = 0).This knot was chosen based on significant changes to sleep patterns immediately around delivery(Bei et al., 2012); and (c) models with cubit splines and five knots.This configuration allows for greater degrees of freedom, and captures complex trajectories, if needed.For all outcomes, configuration (b) with quadratic splines and one knot at childbirth provided the best fit, and it was used for final analyses.Each model also included age, group allocation, and the interaction between intervention group allocation and time; if the interaction term was not significant, it was dropped, leaving the main effect of group as a covariate, along with age.Model estimated means and 95% confidence intervals (CIs) are visualised.To investigate the associations between chronotype with insomnia severity, sleep-related impairment, and mood across all observational time points (Aim 2), a series of four mixed-effects models were undertaken.Four separate models were carried out, with scores of ISI, PROMIS-SRI, PROMIS-Depression, and PROMIS-Anxiety as respective outcomes.Each model included random intercepts and rMEQ as the predictor, adjusting for the following covariates: age, total sleep time, sleep efficiency, history of past or current mental health at baseline (0 = 'no history'; 1 = 'past or current'), PROMIS Instrumental and Emotional Social support scores, and group allocation (0 = 'healthy diet'; 1 = 'healthy sleep' Mixed-effects models estimated change trajectories for bedtime, risetime, and rMEQ with continuous time, quadratic splines, and one knot at week 0 are shown in Figure1; descriptive statistics of these variables for each discrete time point are shown in TableS1in the Supplement.From gestation Weeks 30-35, both bedtime and rise-time delayed by $8 and $20 min, respectively.After childbirth, both bedtime and rise-time became progressively earlier, and speed of these advances slowed in the second, compared to the first postpartum year, as indicated in the quadratic trend (Figure1).Compared to the time around childbirth, bedtime and rise-time were $20 and $60 min earlier, respectively.The random effects of time for bedtime were non-significant ( p = 0.21), suggesting small individual differences in this change.For rise-time, only the fixed effect of time was estimated as the model did not converge when including random effects.
Model estimated change trajectories in sleep-wake timing and chronotype from pregnancy to 2 years postpartum.Model predicted means and 95% confidence intervals are shown.Estimates are based on mixed-effects models controlling for age and group allocation.Zero for weeks since delivery indicates delivery, with negative values indicating time during pregnancy.All models included quadratic splines and one knot placed at week 0. Morningness-Eveningness was measured using the reduced Morningness-Eveningness Questionnaire.