Sleep alterations as a predictor of bipolar disorder among offspring of parents with bipolar disorder: a systematic review and meta-analysis

Abstract Introduction Bipolar disorder (BD) has a high heritability rate. Current studies have been dedicated to identifying prodromes of BD in the offspring of parents with BD (BO) and the sleep patterns of these individuals have been considered important factors. Objective To describe changes in sleep parameters among offspring of parents with BD when compared to offspring of controls and to identify if changes in parameters and quality of sleep predict the onset of BD among these individuals. Methods PubMed, PsycINFO, and Embase were systematically searched with no year or language restrictions, up to August 18, 2020. We searched for a combination of the following search items (“sleep*”) AND (“bipolar disorder*” OR “mania” OR “hypomania” OR “bipolar depression”) AND (“ultra-high risk” OR “high risk” OR “offspring” OR “first degree relatives”). Results A total of 10 studies were included in the systematic review and 4 studies were included in the meta-analysis. Our meta-analysis showed that the BO had greater daytime sleepiness as compared to the offspring of control parents. The systematic review indicated that shorter sleep duration, sleep disorders, and other related features can differentiate the two groups. Finally, some sleep patterns such as decreased sleep, difficulty falling asleep, and overall sleep problems might be predictors for the development of BD. Conclusion Results from the meta-analysis indicated that BO had greater daytime sleepiness. Qualitative results showed that the offspring of parents with BD have an increased likelihood of experiencing an adverse sleep pattern.


Introduction
Bipolar disorder (BD) is a chronic mental illness that affects about 2% of the adult population, considering all of the spectrums. 1 BD has a high heritability rate, and family history of BD is one of the main risk factors for developing the disorder. 2,3 Offspring of parents with BD (BO) are therefore an identifiable high-risk group that may provide relevant information about the course of the emerging disease.
Given the chronic and debilitating nature of the disorder, many recent studies have been dedicated to identifying prodromes of BD in BO, in order to be able to target the prodromes of BD through strategies of prevention and early treatment. [4][5][6][7] Diagnosing BD in young people has proven challenging, considering that prodromic symptoms are nonspecific. 8,9 Thus, the sleep patterns of individuals who have parents with BD have been considered an important prodrome of the disorder. [10][11][12][13] However, changes in sleep are complex characteristics that can involve many specifics. For example, when sleep is assessed by subjective measures, the instruments present many different domains for sleep assessment, 7,[13][14][15][16] and studies using objective measures are not consistent on how the macrostructure of sleep is affected in BO. 11,17,18 In addition, many prior studies considered symptomatic BO in their analyses, making it difficult to determine whether sleep is a prodrome of development of the disease or a consequence of the already established disorder.
In recent years, two systematic reviews have been conducted investigating the link between onset of sleep problems and subsequent development of BD. 19,20 One of those systematic reviews 19 aimed to describe the current evidence regarding chronotype and circadian rhythm patterns in patients with BD. Forty-two studies were included, involving 3,432 patients with BD. The systematic review concluded that depression was more frequently associated with circadian alterations than euthymia in patients with BD. Mania was also associated with irregular rhythms, although few studies evaluated it. Considering biomarkers, preliminary evidence showed dysregulation of daily levels of melatonin and cortisol in patients with BD. In conclusion, the vast majority of studies showed a disruption of circadian rhythm and an evening preference in patients with BD, independent of mood status. However, the impact on mood status is still unclear.
The other systematic review 20  Moreover, hypersomnia seems to precede BD episodes. Therefore, sleep alterations frequently happen for a long time before the onset of BD and seem to be specifically related to the polarity of the index episode.
However, none of these systematic reviews included a meta-analysis. Additionally, to the best of our knowledge there are no systematic reviews that have included studies that performed independent analyses of sleep among symptomatic and asymptomatic BO and/or including only studies comparing unaffected BO to offspring of controls. Furthermore, previous reviews were not limited to investigating the sleep patterns in BO, but they also included studies that included firstdegree relatives 20 and individuals with clinical risk (individuals in the general population who developed BD) 19 in their analyses. Our systematic review will add to the existing literature, considering that we will analyze sleep patterns in a homogeneous and specific population: BO.
Therefore, the objectives of our systematic review and meta-analysis were to describe the changes in

The Preferred Reporting Items for Systematic
Reviews and Meta-analysis (PRISMA) 21 guidelines were followed for the present review.

Protocol registration
A protocol for this systematic review was registered prospectively in PROSPERO under the ID CRD42020203654 on September 17, 2020.

Search strategy
A literature search was conducted on August 18, 2020, with no publication date or language restrictions using the following databases: PubMed, PsycINFO, and Embase. We searched for a combination of the following search items ("sleep*") AND ("bipolar disorder*" OR "mania" OR "hypomania" OR "bipolar depression") AND ("ultra-high risk" OR "high risk" OR "offspring" OR "first degree relatives"). The search yielded 587 articles: (Pubmed = 168, PsycINFO = 99, and Embase = 320), with 415 remaining after removal of duplicates.
We used the following inclusion criteria to determine whether an article was relevant to our study: (1) the study should present original data; (2) the study should include BO; (3) the study should include offspring of parents without BD as a non-exposed control group; and (4)  of the search strategy are illustrated in Figure 1.

Data extraction
Two researchers (KRA and MDC) conducted the data extraction process. We extracted: authorship, year of publication, country where the study took place, study aims, study design, characteristics of the population, presence of mood symptoms in the sample, assessments, and main results.

Quality assessment
Each manuscript included was independently assessed by two blinded researchers (KRA and MDC) using the Newcastle-Ottawa Quality Assessment Scale (NOQAS). Disagreements were resolved by consensus among all authors.

Statistical analysis
Random effects meta-analyses were performed using RevMan 5.3. This was conducted to assess differences in sleep parameters between BO and CO. To achieve this, the means, sample sizes, and standard deviations reported in the studies were used to compute the standardized mean difference or the mean difference in sleep patterns between offspring of parents with BD and offspring of controls. Significance was set at p < 0.05. Cochrane's Q test was performed to screen for statistical heterogeneity and the Higgins I2 statistic was used to determine the extent of variation between sample estimates with values ranging from 0-100%. If information needed was not reported in the paper, we contacted the authors asking for additional information in order to include the paper in the meta-analysis.

Results
The literature search yielded 587 studies. Of these, 172 were duplicates and 357 studies were excluded because the titles and abstracts were not relevant to the Four of the studies provided enough data to be included in the meta-analysis. 5,11,17,25 There was only enough data to perform the meta-analysis for aim 1. Table 1 shows an overview of the studies included.
The quality assessment of the studies included showed that most of the studies (n = 4) scored 7 out of 9, indicating good quality (Table 1). Although BO's objective sleep seems to be better in this study, their subjective experience was the opposite (this will be reported later).

Sleep patterns in BO assessed using subjective measures: evidence from cross-sectional studies (aim 1)
Several different instruments are used to assess sleep subjectively (for more information see Table 1).
The subjective results showed that sleep disorders, 5 increased sleep, and insomnia 11 are characteristics that differ between BO and CO. A study including 25 BO and 22 CO reported more sleep disturbances in BO, with the clearest disturbances observed in the affected children of parents with BD. 5 Post-hoc tests indicated significant differences between affected and non-affected BO (p < 0.001) and between affected BO and unaffected CO (p < 0.001), with greater disturbances among affected BO in ).
• To evaluate aspects of mood dysregulation in their children, parents also completed the PGBI-10M to assess positive mood and energy dysregulation in the past six months 33 and the CALS-P to assess severity of mood lability. 34 • Elevated scores on these mood dysregulation measures (PGBI-10M, CALS-P) have been linked to a BD diagnosis in youth. 34,35 Modified version of the self-report PSQI -sleep patterns and quality in the prior week In BO, sleep duration was negatively correlated with PGBI-10M and CALS-P (p = 0.049). In CO, sleep duration did not correlate with CALS-P and there was not a sufficient range of PGBI-10M scores in this group to perform a correlation analysis.

8/9
Continued on next page    Finally, another study including a sample of 43 BO and 42 controls found that the BO had higher GBI scores (depression p < 0.001; 10-item mania p < 0.001; sleep p = 0.02) than the CO.

measures: evidence from cohort studies (aim 2)
Several different instruments were used to assess sleep subjectively (Table 1) Finally, there was a cohort study that failed to observe the impact of sleep on the development of BD. 16 This study included 335 offspring of parents with BD and 227 offspring of healthy control parents. The main findings showed that in middle adolescence (ages [14][15][16], youth in the "poor sleep group" were significantly more likely to have a parent with BD than those in the "variable sleep group" (p < 0.01), while at ages 16-18 youth in the "poor sleep group" were significantly more likely to have a parent with BD than those in the "good sleep group" (p = 0.02).The poor sleep group had more than four times the odds of developing BD as those in the good sleep group (OR = 4.25), however, the differences were not statistically significant. This study failed to demonstrate statistical significance but suggests clinical significance.

Meta-analysis of studies comparing sleep patterns between BO and CO using subjective measures
Association between daytime sleepiness and BO 5,11 We found that the standardized mean difference  Association between sleep latency and BO 5,26 We found that the standardized mean difference between groups was 0.02 (95%CI -0.39, 0.43; p = 0.93), indicating no significant difference in sleep latency between BO and CO ( Figure 2).
Association between sleep duration and BO 5,25 We found that the standardized mean difference between groups was 0.23 (95%CI -0.18, 0.64; p = 0.27), indicating no significant difference in sleep duration between BO and CO ( Figure 2). 5,25 We found that the standardized mean difference between groups was 0.28 (95%CI -0.13, 0.70; p = 0.17), indicating no significant difference in sleep efficiency between BO and CO ( Figure 2). 5,25 We found that the standardized mean difference between groups was 0.08 (95%CI -0.70, 0.93; p =

0.86), indicating no significant difference in sleep global
index between BO and CO ( Figure 2).
Association between sleep self-assessment and BO 11,17 We found that the standardized mean difference 0.11), indicating no significant difference in sleep selfassessment between BO and CO ( Figure 2).

Meta-analysis of studies comparing sleep patterns between BO and CO using objective measures
Association between sleep duration and BO 5,11,17 We found that the standardized mean difference between groups was -0.03 (95%CI -0.36, 0.29; p = 0.83), indicating no significant difference in sleep duration between BO and CO ( Figure 3).
Association between sleep efficiency and BO 5,11,17 We found that the standardized mean difference between groups was -0.20 (95%CI -0.49, 0.09; p = 0.17), indicating no significant difference in sleep efficiency between BO and CO ( Figure 3).
Association between sleep fragmentation and BO 5,11,17 We found that the standardized mean difference between groups was -0.21 (95%CI -0.80, 0.39; p = 0.50), indicating no significant difference in sleep fragmentation between BO and CO ( Figure 3).
Association between sleep latency and BO 5,11,17 We found that the standardized mean difference  Although BD may present a progressive course, 38 prodromal symptoms signaling the onset of the disease are still nonspecific. 8,9 The high degree of heritability of the disease 2 and the results found in our study demonstrate the importance of early diagnosis of BD and emphasize that sleep management may be a relevant strategy to achieve better prognosis in this high-risk population, in which the chances of developing the disease are increased by 9 times. 39 Changes in sleep and circadian functioning are essential characteristics of the pathophysiology of BD 40 and interventions involving sleep hygiene should be strongly encouraged as they are malleable problems and intervention may help to delay the progression of the disorder.
Our study selection was restricted to studies including non-affected BO. However, we did not find studies that completely excluded symptomatic BO. We therefore decided to include studies that at least analyzed the symptomatic and asymptomatic BO separately Addressing these research questions was a challenge for us and we believe that it will continue to present a great challenge for researchers in the area, considering the difficulty of access to this population when still asymptomatic, at an early age. However, this is a population that can provide many answers in understanding the development of BD. Thus, future studies should invest in prioritizing the homogeneity of their samples as much as possible with respect to age and to their assessment, prioritizing unaffected children, in the absence of any mood disorder. Furthermore, studies should investigate effective sleep-based interventions for this population.