Efficacy of cognitive behavioral therapy for insomnia in the perinatal period: a meta-analysis of randomized controlled trials

Abstract

The meta-analysis aims to explore the effect of cognitive behavioral therapy for insomnia (CBT-I) in the perinatal period. Randomized controlled trials (RCTs) assessed the effects of CBT-I in perinatal women with insomnia, published in English, were eligible. Electronic searches were performed using PubMed, Embase (Elsevier), PsycINFO (Ebsco), and Web of Science (Clarivate Analytics). Insomnia Severity Index (ISI) as the primary outcome was used to estimate the pooled effects and durable efficacy of CBT-I. The secondary outcome measures were Edinburgh Postnatal Depression Scale (EPDS) and Pittsburgh Sleep Quality Index (PSQI). Of 46 studies reviewed, seven studies met the inclusion criteria. The meta-analysis indicated significant improvement in insomnia as measured with the ISI (standardized mean difference (SMD) =  − 0.62, 95% confidence intervals (CI) − 0.77, − 0.47, I² = 28%). At the follow-up time point, the meta-analysis indicated the durable efficacy of CBT-I (SMD = − 0.47, 95% CI − 0.90, − 0.03, I² = 73%). Definite improvement of CBT-I on EPDS (SMD = -0.31, 95% CI − 0.55, − 0.06, I² = 33%) and PSQI (SMD = − 0.82, 95% CI − 1.27, − 0.38, I² = 68%) score change post-intervention were found. In sub-analyses, CBT-I had similar effect sizes, independent of possible modifiers (study population, comparison group, delivery format, etc.). This meta-analysis demonstrates that CBT-I is effective in alleviating insomnia, depression, and sleep quality among perinatal women. It is equally important to find that CBT-I has a durable efficacy on insomnia in the perinatal period. However, it is necessary to include larger samples and conduct rigorous RCTs to further explore this issue.

Supplementary Information

The online version contains supplementary material available at 10.1007/s41105-023-00502-z.

Introduction

Insomnia in perinatal women (the period from conception to 12 months post birth) [1] is greatly prevalent relative to the general population of women of childbearing age (40–60% versus 11%) [2–4]. According to the 3-P model of insomnia [5], there are several contributing factors to insomnia in the perinatal population. Specifically, personality characteristics and the female sex are usually considered predisposing factors [6, 7]. As for precipitating factors, probably due to hormonal effects [8], they tend to engage in nocturnal rumination (i.e., repetitive negative thinking at night) compared with non-pregnant women [9]. Moreover, pregnancy and childbirth as inevitable events may contribute to worry around bedtime, which probably was considered a perpetuating factor among perinatal women. Prior research indicates insomnia can adversely affect a person’s development and overall health [10]. A perinatal woman on a weaker side is more vulnerable. Meanwhile, perinatal women with insomnia are at high risk for numerous perinatal complications involving pre-term birth [11], gestational hypertension [12], and maternal psychiatric illness, particularly depression, and anxiety [13]. All of these can contribute to an adverse effect on the health of the mother and fetus. Therefore, effective therapy is imperative.

Research on interventions to relieve insomnia symptoms in the perinatal period is limited. Compared with pharmacotherapy and acupuncture, cognitive behavioral therapy for insomnia (CBT-I) has been identified by pregnant women as their treatment of choice [14]. CBT-I consists of two core components, sleep restriction therapy (SRT) and stimulation control therapy (SCT), and two adjunctive, cognitive therapy(CT) and sleep hygiene(SH) [15]. As for the effect of CBT-I, the American College of Physicians recommended that CBT-I is the recommended first-line treatment for chronic insomnia [16], and has been confirmed as an effective therapy among non-perinatal adults, which shows average treatment response rates between 70 and 80% [17–19]. In terms of perinatal women, a review suggested that CBT-I played a role in relieving insomnia symptoms during pregnancy and the early postpartum period to some extent [20]. However, only three studies focused on the efficacy of CBT-I in the perinatal period were mentioned and there were no concrete data to support the result. The efficacy of CBT-I post-intervention was also evaluated in a meta-analysis [21] among only pregnant women.

Furthermore, there are a small number of studies focused on the effect of CBT-I on insomnia among perinatal women. Also, up to now, no meta-analyses have been developed examining the persistent efficacy of CBT-I for the treatment of perinatal insomnia. Therefore, the current study aims to analyze the available research to include recent trials to determine a more precise estimate of the efficacy of CBT-I for perinatal insomnia using a meta-analysis of randomized controlled trials (RCTs). Moreover, we conducted an additional meta-analysis to examine the durability effect sizes of CBT-I.

Materials and methods

The protocol of this study has been previously registered at the International Prospective Register of Systematic Reviews (PROSPERO) (registration number: CRD42022374201). However, in the process of conducting later analysis, we added two subgroup analyses for “pregnancy” vs. “postpartum” of the study population and “digital” vs. “face-to-face” of delivery format.

Search strategies

We developed and searched relevant studies in PubMed, Embase (Elsevier), PsycINFO (Ebsco), and Web of Science (Clarivate Analytics). Search strategies were customized to each database and included controlled words and free text synonyms combining the terms of insomnia, perinatal period, and cognitive behavioral therapy, including individual intervention elements (e.g., sleep hygiene, mindfulness interventions, etc.). The full search strategies for these databases are presented in Supplementary Material S1.

Study selection

We independently reviewed all identified abstracts for qualification. We only selected RCTs. All RCT research evaluating CBT-I among perinatal women was included in a full-text review. The inclusion criteria were as follows: (1) study design followed a randomized controlled trial; (2) perinatal women were the primary targeted population; (3) CBT-I was used alone as the intervention method, not connected with other active treatment; (4) study outcomes were presented via insomnia severity or sleep quality measured with any validated tool; (5) studies of perinatal women with insomnia. The excluded criteria included non-perinatal period, no RCT, no CBT-I intervention, no sleep measures of interest, lack of Mean or standard deviations (SD) of Insomnia Severity Index (ISI) assessment outcomes, and not full texts. Authors were got in touch by email if their researches were eligible design but inadequate information. Eligible studies were recognized after full-text reading by the first author individually. Any arguments were reviewed and solved by consensus with the corresponding author.

Data extraction

All accessible data from contained studies were extracted for further analysis. Data extraction was carried out independently by two authors. Disagreements were discussed and harmonized with an expert author. The following data were extracted: first author, publication year, sample features and study design (sample size, age, perinatal period, insomnia duration, diagnostic criteria of insomnia, comparison, control condition), CBT-I intervention (composition, number, frequency, and length of session), method (individual or group), delivery format (e.g., traditional in-person or by remote intervention), outcome measures, points of time for assessments, and the number of discontinued treatment. If an eligible study had missing data, we contacted relevant authors by email to obtain necessary data for effective size calculation, or else, missing information was calculated and transformed from given data following the contact guidance from the Cochrane Handbook for Systematic Reviews of Interventions [22].

Outcome measures

The primary outcome variable of interest was the scores from ISI, one of the most frequently used self-reported sleep surveys. The ISI has been one of the most widely utilized questionnaires for screening insomnia and is a sensitivity screening and therapy outcome measure. It is a brief measure with only seven items, easily administered, developed based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition [23–25].

We found that included studies usually contained the assessment of depression and sleep quality, and almost half of the included studies had the results from the assessment of the Edinburgh Postnatal Depression Scale (EPDS) and Pittsburgh Sleep Quality Index (PSQI). We conducted a meta-analysis as the secondary outcome.

Quality assessment

Two authors individually evaluated the risk of bias for all eligible RCT studies according to the Cochrane risk of bias tool [22] with six components. Each component was assessed as "low-risk (score of 0)," "high-risk (score of 2)," or "uncertain risk (score of 1)". For each study, a score below 4 indicates a low risk and high quality; a score of 5–9 indicates moderate risk and medium quality; a score of 10–16 indicates a high risk and low quality. And two authors conducted a discussion or consulted a third author to resolve the disagreement.

Statistical analysis

Data analyses were conducted using Review Manager 5.4.1. All measures for pre- and post-intervention were extracted for analysis. Likewise, data from follow-ups were also used to develop the statistical analysis. Data input involved common formats from reported results including mean and SD in each group. If SD were not reported, they were calculated from CIs and the mean. Standardized mean difference (SMD) conveyed the size of the intervention effect in each study relative to the variability observed. For the common measure of effect across research, we reported SMD. Forest plots were produced to visually evaluate the SMD and corresponding 95% confidence intervals (CI) across studies. The heterogeneity test I² statistic explained that heterogeneity generates the percentage of variability in effect estimates. If the included studies had significant heterogeneity (I² > 50%), a random-effects model was used. Further sensitivity analysis was performed to assess the impact of each study on the pooled results. Publication bias was assessed via a visual funnel plot. In the meanwhile, we conducted a meta-analysis of the discontinuation rate between the CBT-I group and the control group using Stata 17.0.

Results

Included studies

The original literature search yielded 1430 studies, of which 120 were duplicate studies (Fig. 1). In the process of screening, 1268 researchers were excluded based on criteria for reviewing the titles and abstracts, then 42 records were detected and they were assessed for eligibility based on the full texts. Ultimately, 7 studies were left in the meta-analysis.

Fig. 1.

Flow diagram of study selection

Seven types of research published between 2017 and 2022 were developed in this article. The details of the included studies are provided in Table 1, involving the study population, perinatal period (baseline), diagnosis of insomnia, sample size, the mean age of CBT-I and CTRL (control therapy) groups, CBT-I components, outcome measures, and assessment time point. The 7 RCTs [4, 26–31] included 708 samples, and there were 360 samples in CBT-I groups and 348 samples in CTRL groups. The studied subjects included patients in the perinatal period, and most of them were prenatal. There was one research conducted with postpartum women.

Table 1.

Characteristics of included RCTs of CBT-I

Study	Study population	The perinatal period (baseline)	Insomnia defined by	Numbers of CBT-I group	Numbers of comparison group	Mean age (years)	Treatment components	Outcome measures	Measures and follow up
Cain [28]	Nulliparas and multiparous	13–20 weeks’ gestation	Self-reported	27	26	CBTI, 28.5; CTRL, 29.8	SH, SC, RT	ISI, PSQI	Pre-intervention, post-intervention, follow up 6–8 weeks later
Manber [29]	Nulliparas and multiparous	18–32 weeks’ gestation	DSM-5, DSM-IV-TR	89	90	CBTI, 33.4; CTRL, 32.6	SH, SR, SC, CT	ISI, EPDS	Pre-intervention, post-intervention
Bei [4]	Nulliparas	Mean (27.59 weeks)	DSISD, DSM-5	81	82	CBTI, 33.13; CTRL, 33.39	SC, SH	ISI	Pre-intervention, post-intervention, follow up 6 months,18 months later
Kalmbach [27]	Nulliparas and multiparous	25–30 weeks ‘gestation	ISI score ≥ 10	46	45	CBTI, 28.91; CTRL, 29.16	SR, SC, RT, SH	ISI, PSQI, EPDS, PSAS-C	Pre-intervention, post-intervention
Verma [26]	Nulliparas	4–12 months postpartum	ISI score > 7	39	39	CBTI, 32.85; CTRL, 31.42	SE, SC, SH, RT	ISI	Pre-intervention, post-intervention, follow up 1 month later
Forsell [30]	Nulliparas and multiparous	Mean (17.1 weeks)	None	33	39	CBTI, 31.2; CTRL, 30.8	SH, SR	ISI, EPDS, GAD-7	Pre-intervention, post-intervention
Felder [31]	Nulliparas and multiparous	28 weeks’ gestation	DSM-5	105	103	CBTI, 33.9; CTRL, 33.2	SE, SH, SC, SR, CT, RT	ISI, EPDS, GAD-7, PSQI	Pre, post, follow up 18 weeks later

CBT-I cognitive behavioral therapy for insomnia, ISI insomnia severity index, DSM-IV-TR diagnostic and statistical manual of mental disorders, fourth edition, text revision, DSM-5 DSM 5 criterion, DSISD duke structured interview for sleep disorders, CTRL control therapy, SE sleep education, SH sleep hygiene, RT relaxation training, SC stimulus control, SR sleep restriction, CT cognitive therapy, post post-intervention, pre pre-intervention, mid in the middle of intervention period, PSQI Pittsburgh sleep quality index, EPDS Edinburgh postnatal depression scale, PSAS-C pre-sleep arousal scale's cognitive factor, GAD-7 general anxiety disorder 7-item scale

CBT-I components and sessions

A majority of CBT-I interventions were the synthesis of at least three main sections. All eligible studies involved SH; 85.7% of studies involved SCT; few studies involved SRT (42.9%), CT (14.3%), and relaxation training (RT) (42.9%). Some studies delivered CBT-I interventions via sessions or the synthesis of sessions and emails (71.4%) and others conducted by emails or messages independently (28.5%). In general, almost all studies could conduct sessions more than five times and more than 50 min.

The efficacy of CBT-I

Results from 7 eligible studies using ISI as a primary outcome measure showed pooled effect sizes for CBT-I at post-intervention (SMD = − 0.62, 95% CI − 0.77, − 0.47, I² = 28%) (Fig. 2).

Fig. 2.

Forest plot for studies with ISI as outcome measures. SD standard deviation, CI confidence intervals

Of all the eligible studies, four were evaluated at follow-up after CBT-I, and the results of the meta-analysis showed that CBT-I had a durable effect on perinatal insomnia (SMD = − 0.47, 95% CI − 0.9, − 0.03, I² = 73%) (Fig. 3). However, substantial pieces of evidence of heterogeneity were observed from the results, and the source of heterogeneity could be determined through sensitivity analysis. Verma's study [26] has a great impact on heterogeneity. After the exclusion of this study, the results showed that the heterogeneity of the remaining three studies was reduced (I² = 55%).

Fig. 3.

Forest plot of studies containing follow-up data after the CBT-I. SD standard deviation, CI confidence intervals

Three included studies evaluated the EPDS after the CBT-I, and the results showed that the effect size of CBT-I was SMD = − 0.31 (95% CI − 0.55, − 0.06, I² = 33%) (Fig. 4); three studies evaluated PSQI, and the effect size of CBT-I was SMD = − 0.82 (95% CI − 1.27, − 0.38, I² = 68%) (Fig. 5).

Fig. 4.

Forest plot for studies with EPDS as outcome measures. EPDS Edinburgh postnatal depression scale, SD standard deviation, CI confidence intervals

Fig. 5.

Forest plot for studies with PSQI as outcome measures. PSQI Pittsburgh sleep quality index, SD standard deviation, CI confidence intervals

For the discontinuation rate of treatment, there was no significant difference between the CBT-I group and the control group, and the forest plot was provided in Supplementary Material S2.

Subgroup analysis

Study population

Mixed population (nulliparas and multiparas) vs. nulliparas

Of all of the seven studies, five studies involved nulliparas and multiparas and two studies only included nulliparas. There was no significant difference between multiparas and nulliparas for effect size (p = 0.95). In the studies with nulliparas and multiparas, effect size favored CBT-I at post-intervention (SMD = − 0.67, 95% CI − 0.85, − 0.48, I² = 0%), while a non-significant effect size in the studies with nulliparas was SMD = − 0.64 (95% CI − 1.37, 0.08, I² = 83%).

Pregnancy vs. postpartum

Among the seven eligible studies, six studies involved participants among pregnancy and one study included postpartum. There was no significant difference between pregnancy and postpartum for effect size (p = 0.09). In the studies with pregnancy, the effect size favored CBT-I at post-intervention (SMD = − 0.58, 95% CI − 0.74, − 0.42, I² = 7%), while the effect size in the studies with postpartum was SMD − 1.04 (95% CI − 1.54, − 0.54).

Comparison groups

Among the seven eligible studies, two studies only conducted treatment as usual for the control group (i.e., inactive control), while five studies developed an active control (e.g., receiving health information, healthy diet, digital sleep education control, etc.). The inactive control showed the effect size of CBT-I was SMD = − 0.77 (95% CI − 1.36, − 0.18, I² = 53%), and the active control was SMD = − 0.59 (95% CI − 0.78, − 0.40, I² = 22%) There was no significant difference between inactive group and active group (p = 0.56).

Delivery format

Of all of the seven studies, six studies provided intervention digitally and only one study in a face-to-face way. There was no significant difference between the way of digital and the face-to-face for effect size (p = 0.96). In the studies with digital way, the effect size favored CBT-I at post-intervention (SMD = − 0.62, 95% CI − 0.79, − 0.45, I² = 40%), while the effect size in the studies with the face-to-face way was SMD = − 0.63 (95% CI − 0.98, − 0.28).

The results of the subgroup analysis were provided in Supplementary Material S3.

Study quality and risk of bias

According to the Cochrane Collaboration's risk-of-bias tool, 15% of eligible studies were moderate risk and others were low risk. The risk of bias are shown in Fig. 6. Publication bias analysis was performed for post-intervention of CBT-I, follow-up after CBT-I, EPDS, and PSQI. The results are presented in funnel plot form in Supplementary Material S4. By visual examination, we found that the funnel plot of the post-intervention of CBT-I was symmetrical, suggesting that publication bias may not exist, but the funnel plots of follow-up after CBT-I, EPSD, and PSQI were asymmetrical, suggesting that publication bias exists to some extent.

Fig. 6.

Risk of bias of the included studies

Discussion

This current study aimed to synthesize the research on the efficacy of CBT-I in the perinatal period using a meta-analytic approach. Our meta-analysis included 7 RCTs involving 708 perinatal women with insomnia. Overall, the results further demonstrated that CBT-I significantly improved insomnia symptoms compared to treatment as usual or active control. This result was compatible with the study seen in meta-analyses of CBT-I in adults, a wider population [32]. Perinatal women learned to identify the factors that contributed to their insomnia through CBT-I. At the same time, the American College of Physicians had shown that CBT-I was an effective therapy for insomnia and it should be offered to patients as the first-line therapy [16]. Recently a meta-analysis [21] demonstrated that the efficacy of CBT-I for insomnia looked only at pregnant women, whereas our study focused on perinatal women (pregnancy and postpartum). And the included studies we selected were mostly different compared with their meta-analysis. Furthermore, In addition, we used ISI as a uniform outcome measure for all included studies, and their meta-analyses were varied and highly heterogeneous. Last but not least, we found that CBT-I has a durable efficacy on insomnia among perinatal women.

One of the unique features of this study was that we conducted an additional meta-analysis of the follow-up data to show that CBT-I has a durable effect on perinatal insomnia, which also suggested that CBT-I is somewhat more likely to be incorporated into routine perinatal care. However, the heterogeneity was high, possibly due to the inconsistent follow-up evaluation time (1–6 months) after the CBT-I intervention.

Some researchers mentioned that depression, anxiety, and insomnia were comorbid psychiatric conditions in the complexity of mental health disorders [33]. However, there was significant evidence in the non-perinatal adult literature demonstrating that insomnia was a prospective predictor of new-onset psychiatric disease [13]. Women in the perinatal period may tend to suffer from depression while they have insomnia [34]. Thus, the second finding of this study was that there was an overall significant reduction in EPDS scores.

In the present meta-analysis, sub-analyses revealed CBT-I remained similarly efficacious regardless of potential modifiers (study population, comparison design, delivery format). However, to some extent, the analysis results directed that CBT-I was more likely effective in the condition of the mixed population (nulliparas and multiparas) and inactive control. Future studies using adaptive designs can test the optimal sequence and dosage of CBT-I for perinatal women with insomnia. Overall, the acceptability and feasibility of implementing CBT-I for women in the perinatal period in routine care will be a critical future area of research.

Our meta-analysis had centered on a particular population with unique needs, which enhanced our confidence in these findings yet limited our ability to generalize to other populations. We used ISI as a uniform outcome measure for all included studies. Furthermore, it was interesting to note that our eligible studies had slight heterogeneity and no publication bias, which further confirmed the accuracy of our research.

Limitation

In the meantime, we also acknowledged several methodological limitations of the included studies. First, because of the lack of studies in the field, only 7 RCTs were included. Second, due to the limited amount of literature, we could not separate the analysis of different stages of perinatal women. Thirdly, the results of the PSQI meta-analysis showed that I² > 50%, since the value was included in three studies, we did not conduct further sensitivity analysis. Fourthly, we did not find evidence for the use of PSG (polysomnography) to make an exclusionary diagnosis for sleep disorders other than insomnia, thus there was the possibility that patients with sleep disorders other than insomnia were included. Future studies are encouraged to explore CBT-I efficacy via more RCTs. Despite these limitations, the results of this meta-analysis still may guide future research in the field, and assist to inform optimal service delivery for women with perinatal insomnia in the future.

Conclusions

The meta-analysis demonstrates that CBT-I is effective in alleviating insomnia, and depression and improving sleep quality among perinatal women. It is equally important to find that CBT-I has a durable effect on insomnia in the perinatal period. This intervention has the potential to be implemented into routine perinatal care. However, the efficacy and safety of CBT-I in the treatment of perinatal insomnia still need to be verified by further studies with rigorous design and large samples.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

This study was funded by Qingdao Social Science Planning Research Project (Grant/Award Numbers: QDSKL2201075).

Declarations

Conflict of interest

All authors indicate that they have no conflicts of interest.

Ethical approval

No formal consent is required for this type of research.