The effect of physical activity on sleep disorders in pregnant people: a meta-analysis of randomized controlled trials

Abstract

Background

Sleep disorders are highly prevalent in pregnant people and have been associated with significant complications and morbidities for both pregnant people and their offspring. Despite this, the management of sleep issues during pregnancy remains suboptimal. There is an urgent need to explore novel treatment approaches that are safe, feasible, and widely implementable in daily routines. Given the demonstrated positive effects of physical activity (PA) on sleep in non-pregnant populations, PA interventions are a promising option. This meta-analysis was performed to evaluate the association between PA interventions and sleep disorders in pregnant people.

Methods

A systematic search of seven databases was conducted for English and Chinese articles published from inception to January 2024 using MeSH headings and keywords for ‘sleep disorder’, ‘pregnancy’, ‘physical activity’, and ‘randomized controlled trial’. Two independent researchers selected studies assessing the effects of PA interventions on sleep disorders in pregnant people compared with a control condition. Data extraction was performed independently by two reviewers, and quality was assessed using the Cochrane Risk of Bias V1.0 tool. A random-effects model was applied for the meta-analysis, with results reported as standardised mean difference (SMD) and 95% confidence interval (CI).

Results

Eighteen studies involving 1,541 pregnant people were included, with 14 studies included in the meta-analysis. The results suggested that PA interventions were associated with a reduction in sleep disorders compared with the control condition (SMD = − 1.48, 95% CI = − 2.06 to − 0.90, P < 0.00001; night sleep MD = 0.52, 95% CI = 0.42 to 0.62, P < 0.00001; proportion of night sleep time MD = 5.65, 95% CI = 4.78 to 6.52, P < 0.00001). Subgroup analyses indicated that intervention characteristics (e.g., less than 8 weeks and more than 60 min at a session, individual or group settings, location, and activity type such as water-based exercise, progressive muscle relaxation, and aerobics) and participant characteristics (with or without complications) influenced the overall treatment effect.

Conclusion

This meta-analysis demonstrates that PA interventions positively impact sleep disorders in pregnant people, with effects influenced by participant type, duration, delivery method, and activity form. These findings provide valuable insights for healthcare professionals and hold significant implications for developing comprehensive, evidence-based guidance.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12884-024-07129-z.

Sleep is an essential activity for human beings, and high-quality sleep is crucial for maintaining physical and mental health. Sleep is frequently affected during pregnancy due to hormonal fluctuations, increased nocturia caused by an enlarging uterus, and foetal movements at night [1]. Therefore, sleep disorders are more common during pregnancy and represent heterogeneous conditions, including poor sleep quality, insomnia, and restless legs syndrome. A 2020 meta-analysis estimated that the overall prevalence of insomnia during pregnancy is 38.2% [2] and found that almost half of pregnant people (45.7%) experience poor sleep quality [3]. Furthermore, restless legs syndrome is likely to affect pregnant people approximately three times more frequently than the general population [4]. Multiple systematic reviews and meta-analyses have shown that sleep disorders during pregnancy can lead to a variety of adverse maternal and neonatal outcomes, such as gestational diabetes, gestational hypertension, perinatal psychological disorders, prolonged labour, caesarean section, preterm birth, and stillbirth [5–8]. Timely, safe, and effective intervention for sleep disorders in pregnant people can reduce the occurrence of adverse maternal and infant outcomes, enhancing their quality of life [9].

There are two primary categories for treating sleep disorders: non-pharmacological treatments and pharmaceutical treatments. The most common therapeutic approach involves pharmaceutical treatments, which work by suppressing the central nervous system to reduce anxiety and stress levels, inducing a hypnotic and relaxed state [10]. However, some scholars have suggested that pharmacological treatment may increase the potential risks of preterm birth and low-birth-weight infants, making the use of medication during pregnancy possibly extremely contraindicated [11]. Alternative non-pharmaceutical methods without side effects are more readily accepted [12]. In 2017, Mindel et al. conducted an internet-based survey on sleep patterns and sleep problems among 2,427 pregnant people in the United States and found that 76% of them reported poor sleep quality [13]. Nevertheless, sleep disorders in pregnant people are often overlooked for various reasons, and they are not typically included in clinical trials for sleep treatments. This may lead to underdiagnoses and under-treatment of sleep complaints during pregnancy [14]. Therefore, there is an urgent need to explore novel treatment approaches that can be safely, feasibly, and widely implemented in the daily routines of pregnant people with sleep problems. Physical activity (PA) interventions are potential complementary or alternative treatments for sleep problems in pregnant people. Such interventions have been shown to alleviate sleep disorders in the non-pregnant population [15] and have been endorsed by international guidelines (e.g., the American College of Obstetricians and Gynecologists and the World Health Organization) [16, 17].

Previous meta-analyses have examined the effect of PA, providing limited evidence on the relationship between PA and sleep disorders in pregnant people [18, 19]. In 2020, Yang et al. [19] published a systematic review of seven randomised controlled trials, concluding that exercise has a positive impact on sleep disorders in perinatal people. Another meta-analysis [18] reached the same conclusion. However, there were several limitations to these reviews. Both included small sample sizes, with the largest review analysing only seven studies. In addition, they only reported the overall effects of PA on the sleep of pregnant people and did not explore the frequency, duration, type, or delivery methods of PA. These factors are crucial because they have significant implications for clinical practice. In recent studies, although Cannon et al. [20] and Weng et al. [21] attempted to conduct a systematic quantitative assessment of the relationship between PA and sleep during pregnancy, they were unable to do so because of the small sample sizes.

With this meta-analysis, we aimed to include a wider range of studies to provide updated critical insights into the quantitative association between PA and sleep disorders in pregnant people. Furthermore, we aimed to identify possible participant- and trial-related characteristics, such as session duration and sites, to determine the optimal dose and mode of delivery of PA for managing sleep problems.

Methods

Date sources and search strategy

We searched the Cochrane Database of Systematic Reviews and PubMed Clinical Queries to prevent any duplication of the review topic. We searched English and Chinese electronic scientific literature databases, including PubMed, Embase, The Cochrane Library, Web of Science, CNKI, CBM, and Wanfang, from inception to January 2024. We included insomnia symptoms, poor sleep quality, RLS, S-SDB, and OSA as the sleep disorders for this analysis. The search terms were used to probe title, abstract, and keywords, which included (sleep or sleep disorder or insomnia or sleep quality or sleep-disordered breathing or obstructive sleep apnea or restless legs syndrome or Willis Ekbom disease) AND (pregnancy or pregnant women or gravidity or prenatal or pregnan* or gestation* or antenatal) AND (exercise or physical activity or aerobic* or exercise* or sport or yoga or relaxation or training or walk) AND (randomized controlled trial). In addition to the database search, we also searched reference lists and review articles for additional studies that met the inclusion criteria. The study protocol was registered with the International Prospective Register of Systematic Reviews (PROSPERO), number CRD42024496338.

Inclusion and exclusion criteria

The inclusion criteria for the study were based on the PICOS principle, assessing the following aspects: participants, intervention, comparison, outcomes, and study design, which were used as the selection criteria [22]. According to the design, participants were pregnant people, and the intervention was PA regardless of its types or forms (e.g., aerobic exercise, stretching and relaxation, yoga, or walk and so on) versus non-active intervention (e.g., education or usual care) as a comparison. Any one of the sleep disorders was considered an outcome. Regarding study design, only RCTs were assessed. The following studies were excluded: (1) Other types of paper (e.g., review, case reports, abstract, or clinical trials), (2) numerical data not provided or specified on specific tools, and (3) not accessible.

Study selection and data extraction

Two reviewers completed the study selection and data extraction. They selected studies based on titles and abstracts, and then full texts of the selected studies were retrieved and read in full in an unblinded and independent manner. Disagreements between the two reviewers were resolved by discussion or consultation with a senior advisor.

A matrix was used to tabulate the information extracted from each study. The following information was extracted: basic information of the article (first author, published year, country), age, gestational age, sample size(s), type(s) of the intervention(s), comparator(s), type of the pregnancy as well as outcome measures. Details for the description of the PA interventions include the mode of delivery (e.g., at home or at the hospital; individuality or group), the frequency of the intervention, and its duration. In cases of missing or questionable information, the authors were contacted to gather more information. Any disagreements were sorted out with the help of a third reviewer.

Assessment of study quality

Using the Cochrane Risk of Bias Tool V1.0 to assess the risk of bias in each trial, which is the most commonly used evaluation tool for RCTs [23]. The assessment included: generation of random sequence, allocation concealment, blinding of participants and personnel, blinding of outcome assessors, completeness of outcome data, selective reporting of results, and other biases. The criteria for assessment were: low risk of bias, high risk of bias, and unclear.

Statistical analysis

RevMan software (Review Manager version 5.4.1, The Nordic Cochrane Centre, Copenhagen) was used to synthesize the data and conduct the meta-analysis. As for data synthesis of continuous variables, the results of the individual studies were calculated as mean difference (MD) or standardized MD (SMD), with 95% confidence intervals (CIs). When the pooled trials used different rating scales, the absolute MD divided by the SMD was applied. Cochran’s Q test (chi-square test) and I² statistics were used to evaluate statistical heterogeneity. The statistical significance of the chi-square test for significant heterogeneity was set as P<0.10. The interpretation of heterogeneity based on I² is as follows: <50% (homogeneity), ≥ 50% (heterogeneity). If heterogeneity was observed (I² ≤50%, P≥0.1), we applied a fixed-effects model. If heterogeneity was observed (I²>50%, P < 0.1), further analysis of the sources of heterogeneity should be conducted. After excluding the influence of obvious clinical heterogeneity, a random-effects model can be adopted for Meta-analysis. Obvious clinical heterogeneity will be dealt with by methods such as subgroup analysis or sensitivity analysis, or only descriptive analysis will be conducted. Predefined groups were developed to identify the essential features of intervention through subgroup analyses on the basis of different types of pregnant people, as well as duration, delivery, and form of PA interventions.

Risk of publication bias

Potential publication bias was evaluated using the Egger’s Intercept Test, with P < 0.05 signifying significant bias.

Results

Search outcome

Figure 1 illustrates the process of searching and selection process. A total of 2612 articles were found, and 2095 articles were screened after curating 517 duplicates using the Endnote X9 program. A total of 56 full-text articles were examined, and 38 articles were excluded due to various reasons such as outcomes not mentioned or not relevant, not accessible, clinical trials, etc.; finally, 18 RCTs were selected, of which four did not provide specific results on PSQI after the intervention. We contacted the corresponding authors three times by email, but they did not reply. Thus, only 14 were included for meta-analysis, and the remaining 4 articles would be presented in a systematic review.

Fig. 1.

Flow chart of the identification of eligible studies

Characteristics of included studies

Participants

These articles have almost been published in the last ten years. The sample size ranged from 14 to 270 and included 1541 pregnant people, with 784 in the PA intervention group and 757 in the nonphysical activity control group. 27.8% of the studies were from developed countries, and 72.2% were conducted in developing countries. Of the 18 RCTs that were included in Tables 1, 11 RCTs included healthy pregnant people. The other 7 RCTs included restless legs syndrome symptoms in pregnant people (2), pregnant people with a current or a history of depression and/or anxiety (2), lumbopelvic pain in pregnant people, and pregnant people with sleep disorders or complaints of insomnia and fatigue.

Table 1.

Study characteristics for all 18 included studies

First author, year	Country	Age		Sample size		Gestational age	Type of intervention	Comparator	Type of the pregnancy	Intervention details	Site	Outcome measures
		EG	CG	EG CG
Hyun 2022 [28]	Korea	34.14 ± 3.82	31.71 ± 3.03	7	7	20–24 weeks	Pilates	Standard care	Healthy pregnant people	Twice a week for 50 min per day for 8 weeks; individuality; online	At home	PSQI
Shamekh 2022 [44]	Egypt	25.87 ± 5.29	25.90 ± 5.33	30	30	-	Stretching exercises include four main steps: calf muscle, front thigh, hip flexor stretches, and ankle rotations.	Thermotherapy	Pregnancy with restless legs syndrome symptoms	30-minute stretching exercises per day for one week; individuality	At home	GSQS
Hayase 2018 [52]	Japan	33.8 ± 4.7	32.6 ± 4.9	38	53	20–23 weeks	Yoga	Standard care	Healthy pregnant people	Once a week for approximately 60 min until delivery; group;	At hospital	Sleep log
Shen 2021 [49]	China	33.3 ± 4.19	32.8 ± 3.83	51	47	16–30 weeks	Yoga	Standard care	Healthy pregnant people	Three times per week for 20-min per day for 12 weeks at home; individuality	At home	PSQI
Alomairah 2023 [53]	Saudi Arabia	31.1 ± 4.3	32.0 ± 4.6	87	45	≤ 15 weeks	Exercise training at a gym and swimming pool	Standard care	Healthy pregnant people	1 h moderate intensity 3 times per week in a gym and swimming pool; group; online	At hospital	PSQI、Activity Tracker
Broberg 2022 [54]	Danish	31.9 ± 3.8	31.7 ± 3.9	133	137	17–22 weeks	10-min warm-up, 20 min of endurance training on treadmills, exercise bikes or cross trainers, 25 min of strength training, and 15 min of stretching and relaxation	Standard care	Pregnant people with current or a history of depression and/or anxiety	Twice a week for 60-min per session for 12 weeks; group;	At hospital	PSQI
Yıldırım 2023 [48]	Turkey	30.8 ± 7.0	28.8 ± 5.6	17	17	14–24 weeks	Pilates	Standard care	Pregnancy with lumbopelvic pain	Twice a week for 60-min per session for 12 weeks; group	At hospital	PSQI
Azward 2021 [55]	Indonesia	-	-	30	30	≥ 28 weeks	Yoga	Standard care	Healthy pregnant people	Four times in two weeks; group	At community health center	PSQI
Kim 2022 [45]	Korea	39.71 ± 2.01	38.14 ± 1.39	8	8	24–28 weeks	Pilates	Standard care	Healthy pregnant people	Twice a week for 50 min per day for 8 weeks; individuality; online	At home	PSQI
Field 2013 [50]	America	24.4 ± 4.7	26.0 ± 5.6	37	38	-	Tai chi/yoga	waitlist control group	Clinically depressed pregnant people	20 min session per week for 12 weeks; group	At hospital	VSH Sleep Scale
Akbas 2023 [41]	Turkey	28.03 ± 4.45	26.00 ± 4.07	26	26	26–27 weeks	Progressive muscle relaxation	Standard care	Pregnant people diagnosed with restless legs syndrome symptoms	At least exercise three times a week for 8 weeks; individuality	At home	PSQI
Özkan 2018 [56]	Turkey	27.93 ± 4.56	27.79 ± 3.90	42	42	28–34 weeks	Progressive muscle relaxation	Standard care	Healthy pregnant people	Four weeks and was composed of 5 courses, and each course lasted for 3 h; group	At hospital	PSQI
Blanque 2017 [57]	Spain	32.12 ± 4.43	30.58 ± 4.75	70	70	12–20 weeks	Physical activity in water	Standard care	Healthy pregnant people	Three times for one hour per session for 17 weeks; group	At hospital	PSQI
Golmakani 2015 [43]	Iran	25.30 ± 4.11	25.64 ± 4.47	33	34	29–32 weeks	Progressive muscle relaxation	Standard care	Pregnant people with sleep disorders (PSQI ≥ 5)	Twice a day for 4 weeks; individuality	At home	PSQI
Rashed 2019 [42]	Saudi Arabia	-	-	40	40	29–32 weeks	Progressive muscle relaxation	Standard care	Healthy pregnant people	Twice a day for 4 weeks; individuality	At home	PSQI
BA 2010 [58]	Nigeria	31.8 ± 7.7	31.0 ± 7.1	15	15	-	Aerobics exercise	Standard care	Complaints of insomnia and fatigue	Six weeks; group	At hospital	ISI
Shu 2018 [47]	China	29.29 ± 4.8	29.16 ± 4.46	52	50	10–32 weeks	Mindfulness yoga	Standard care	Healthy pregnant people	Once a week for 60 min per course for 4 weeks and assigned homework require to practice at home once a day for at least 20 min; group	At hospital	SRSS
Mao 2020 [59]	China	27.35 ± 4.37	27.43 ± 4.65	68	68	-	Prenatal aerobics includes three parts: aerobic exercise, strength training, and stretching training	Standard care	Healthy pregnant people	Including three parts: aerobic exercise, strength training, and stretching training; group	At hospital	PSQI

Abbreviations EG = Experimental group; CG=; Control group PSQI = Pittsburgh Sleep Quality Index; GSQS = Groningen Sleep Quality Scale; VSH = Verran and Snyder-Halpern; ISI = Insomnia Severity Index; SRSS = Self Rating Scale of Sleep

Intervention

The mean duration of the prescribed PA program was 8 weeks (range, 1–17 weeks). The frequency of the PA sessions ranged from 1 to 7 days per week, with 2 or 3 days per week being the most commonly used frequency (8 studies). The mean duration of the PA sessions was 60 min (range, 20–180 min). Most interventions were fully supervised at the hospital (10 studies). 2 studies shifted the intervention from hospital to home due to the impact of the epidemic, but pregnant people were supervised through online meetings. We identified five distinct types of PA: pilates (n = 3), yoga (n = 5), activity in water (n = 2), progressive muscle relaxation (n = 4), and aerobics exercise (n = 4). See Table 1 for the characteristics of aerobics exercise.17 studies compared to usual care for their region. Only one study used thermotherapy as a control intervention.

Outcome

Nearly all studies used subjective tools as a method of data collection. Pittsburgh Sleep Quality Index (PSQI) was administered in the 13 studies and was the most frequently used sleep quality measure among the eighteen studies. The remaining 5 studies respectively implemented the Insomnia Severity Index(ISI), sleep log, Groningen Sleep Quality Scale (GSQS), Verran and Snyder-Halpern (VSH) Sleep Scale, and Self Rating Scale of Sleep (SRSS). It is noteworthy that GSQS, SRSS, and VSH sleep scale are also tools for measuring sleep quality. However, the sleep log was different. It needed each participant to record every instance of sleep and wake for seven consecutive days, and then researchers used a sleep and wake rhythm analysis program to analyze the night sleep time, day sleep proportion of night sleep time (ratio of night sleep time to total sleep time over1 day). Therefore, it was analyzed separately during the meta-analysis. Only one study used objective measurements (Activity Tracker), but it did not provide data.

Risk of bias

A graphical representation of the risk-of-bias assessment is represented in Fig. 2. Based on the distribution of each item in the bias risk assessment, the methodological quality of all included studies was judged as “moderate.” All studies claimed to be randomized; however, 4 studies did not reveal their content and method of random sequence. 14 studies did not report methods applied to perform adequate allocation. Most studies offered no data material on blinding. 2 studies clearly reported that participants and personnel were blinded. Only 4 studies used intention-to-treat (ITT) for statistical analysis.

Fig. 2.

Assessment of risk of bias of the included RCTs

Effects of PA interventions on sleep disorders

Initially, a meta-analysis was conducted to investigate the effects of all the included PA interventions on sleep disorders. Eventually, 14 articles were included in the meta-analysis. The test for overall effect indicates that participants in the PA group demonstrated an obvious sleep disorders improvement based on the random effects model (SMD=-1.48; 95%CI=-2.06 to -0.90; P < 0.00001; n = 13) with high heterogeneity results between studies (I² = 95%). The outcomes are illustrated in Fig. 3. One study reported participants’ night sleep and proportion of night sleep time by using the sleep log, with the results of continuous variables revealing a significant difference between the yoga group and standard care group (MD = 0.52; 95%CI = 0.42 to 0.62; P < 0.00001; n = 1); (MD = 5.65; 95%CI = 4.78 to 6.52; P < 0.00001; n = 1). The outcomes are illustrated in Fig. 4.

Fig. 3.

The effects of PA interventions on sleep disorders

Fig. 4.

The effects of PA interventions on night sleep and proportion of night sleep time

The effects of the intervention on sleep disorders of different types of pregnant people are shown in Fig. 5. The test for subgroup differences was not statistically significant, with low heterogeneity (P = 0.57, I² = 0%). Meta-analysis showed that both healthy pregnant people and pregnant people with complications improved their sleep compared to the control group (SMD = -1.34; 95%CI = -2.06 to -0.63; P = 0.0002; n = 7); (SMD = -1.73; 95%CI = -2.82 to -0.63; P = 0.002; n = 6).

Fig. 5.

The effects of the intervention on sleep disorders of different types of pregnant people

The effects of the intervention time on sleep disorders are presented in Figs. 6 and 7. Meta-analysis showed a large effect in sleep disorders in the ≤ 8 weeks (SMD = -2.30; 95%CI = -3.15 to -1.44; P < 0.00001; n = 8) and one session ≥ 60 min (SMD = -0.86; 95%CI = -1.54 to -0.19; P = 0.01; n = 5). Interventions with>8 weeks (SMD=-0.35; 95%CI=-0.78 to 0.08; P = 0.11; n = 5) and one session < 60 min (SMD=-0.42; 95%CI=-1.06 to 0.23; P = 0.2; n = 3) had no effect for sleep disorders.

Fig. 6.

The effects of the intervention time on sleep disorders

Fig. 7.

The effects of the intervention time on sleep disorders

The effects of the delivery on sleep disorders are presented in Fig. 8. The test for subgroup differences was not statistically significant, with low heterogeneity (P = 0.28; I² = 14.8%). Meta-analysis showed a large effect for sleep disorders by home-based individual interventions (SMD = -2.02; 95%CI = -3.62 to -0.42; P = 0.01; n = 5). Similarly, interventions with group and hospital-based were also found to have statistically significant results (SMD = -1.09; 95%CI = -1.64 to -0.54; P = 0.0001; n = 8).

Fig. 8.

The effects of the sites of PA interventions on sleep disorders

The effects of the mode of PA interventions on sleep disorders are illustrated in Fig. 9. Overall, the test for subgroup differences suggests that there is a statistically significant subgroup effect based on a random model with high heterogeneity (P < 0.00001; I² = 87.5%). Meta-analysis suggested no effect for sleep disorders in the Pilates and yoga subgroups (SMD = -0.68; 95%CI = -1.94 to 0.57; P = 0.29; n = 2); (SMD = -0.37; 95%CI = -0.88 to 0.13; P = 0.15; n = 3). The estimates showed significant effect for sleep disorders in the activity in water, progressive muscle relaxation and aerobics exercise subgroups (SMD = -1.08; 95%CI = -1.44 to -0.73; P < 0.00001; n = 1); (SMD = -2.74; 95%CI = -3.41 to -2.07; P < 0.00001; n = 4); (SMD = -1.58; 95%CI = -2.78 to -0.38; P = 0.01; n = 3).

Fig. 9.

The effects of the intervention on sleep disorders of different forms of intervention

Sensitivity analysis

By sequentially excluding each included study, it was observed that the key outcomes, such as the pooled effect size (SMD) and heterogeneity index (I²), did not exhibit significant fluctuations. This demonstrates that the results of the meta-analysis are relatively robust and less influenced by individual studies.

Publication bias

Egger’s test was used to assess publication bias (P = 0.005 < 0.05). The results suggested the possibility of publication bias for all studies included in the meta-analysis.

Narrative syntheses

Among the 18 selected trials, 4 studies could not be meta-analyzed (Hyun et al., 2022; Shamekh et al., 2022; Alomairah et al.,2023; Azward et al., 2021) because their results were presented in different ways and could not be pooled. All these studies utilized the PSQI to assess the effectiveness of PA on sleep. Based on our primary findings in the meta-analysis, four studies reported the effectiveness of PA in improving sleep quality among pregnant people and showed statistically significant results. Alomairah et al. also complemented records of sleep time with Activity Tracker. However, they found that objective sleep decreases as pregnancy progresses and did not differ significantly between groups.

Discussion

In this meta-analysis, data from 14 randomised controlled trials were synthesised to examine the effectiveness of PA interventions on sleep problems among pregnant people. The results showed that PA interventions produced greater improvements in sleep problems compared with control conditions. We further explored the optimal target population, duration, delivery methods, and forms of intervention for PA during pregnancy. This study updates and expands upon previous reviews of PA interventions for sleep disorders in pregnant people.

This meta-analysis demonstrated that PA interventions had a positive effect on the sleep of pregnant people, a finding supported by previous studies. You et al. [24] explored the relationships between sedentary behaviour, exercise, and sleep disorders in 22,599 participants from the National Health and Nutrition Examination Survey. The results showed that while sedentary behaviour was a risk factor for sleep disorders, PA effectively reduced them. The mechanisms by which PA improves sleep during pregnancy are not yet fully understood, but inflammation appears to play an important role. Sleep disorders are associated with significantly elevated inflammatory factors [25, 26], which PA suppresses. During PA, muscle fibres trigger the production of IL-6, which stimulates the circulation of other anti-inflammatory cytokines, inhibiting the expression of pro-inflammatory cytokines (mainly TNF-α, IL-1B, interferon-γ, and leptin) in endothelial cells [27]. For pregnant people, the specific mechanism by which PA improves sleep may also involve the alleviation of bodily pain, thereby enhancing sleep quality. A study by Hyun et al. [28] showed that engaging in Pilates significantly relieved low back pain and related insomnia during pregnancy. Additionally, a literature review indicated that PA is a safe strategy that can help prevent pregnancy-related disorders such as gestational diabetes, lumbopelvic pain, anxiety, and prenatal depression [29]. Given these benefits, healthcare professionals should encourage pregnant people to actively participate in various forms of PA while ensuring their safety.

More than 60 min per session and interventions lasting less than 8 weeks induced greater benefits for sleep disorders compared with other durations. These findings align with previous meta-analyses examining the association between PA and sleep quality [30–32], suggesting that the amount of PA may not be strictly correlated with sleep. A recent national cross-sectional study revealed a non-linear relationship between PA and sleep disorders, indicating that after a certain level of PA is reached, further increases do not result in additional sleep improvements [33]. Adherence likely plays a crucial role in intervention duration and effectiveness, as maintaining participants’ attention throughout a trial can be challenging. Researchers may need to prioritise strategies to effectively sustain participant compliance as a critical factor influencing treatment outcomes. While this meta-analysis suggests that PA interventions lasting less than 8 weeks were effective for improving sleep in pregnant people, other factors may also affect these results. High-quality research is needed to determine the optimal dosage of PA for pregnant people, while also considering economic factors such as cost-effectiveness and time efficiency [34].

Our moderator analyses showed that both healthy pregnant people and those with complications benefitted equally from PA. Similar results were reported in a scoping review on PA for various populations, including healthy older adults and patients with cancer or mental illness-related conditions [35]. This finding is encouraging because previous studies have demonstrated that pregnant people with comorbidities tend to experience more sleep problems than their healthy counterparts [36, 37]. Notably, the comorbidities discussed in this article include anxiety, depression, and pain, which have been shown to interact with sleep such that improvements in one often lead to improvements in the other [38, 39]. However, whether PA similarly benefits sleep in individuals with other comorbidities, such as diabetes and hypertension, remains to be studied.

We found no difference in the effects of intervention studies between individuals at home and groups at the hospital, aligning with findings in the related literature [40]. Individual PA interventions at home also improved sleep, possibly because of the use of various supervision methods. For example, Akbas et al. [41] reminded pregnant people via SMS, Rashed et al. [42] and Golmakani et al. [43] used exercise checklists, Shamekh et al. [44] contacted participants daily to ensure they performed the interventions, and Hyun [28] and Kim [45] utilised a real-time video chat app to deliver lessons. The personal relational pressure created by supervision may have facilitated adherence to treatment. Our results suggest that the location of the activity is not the primary factor influencing the effectiveness of PA; rather, the presence or absence of supervision is crucial. Current guidelines related to PA for pregnant people do not specify which delivery format should be used [17, 46]. This study provides the latest evidence that can inform future guidelines: PA is equally effective for managing sleep problems in pregnant people whether conducted individually at home or in groups at the hospital. When hospital-based treatment is unavailable or prohibitively expensive, home-based interventions can be considered a viable alternative.

We found that studies on underwater exercise, progressive muscle relaxation, and aerobic exercise improved sleep more effectively than studies on yoga and Pilates. These findings were surprising because previous clinical studies have demonstrated that yoga and Pilates have significant effects on sleep [28, 47]. A potential explanation could be the relatively small sample sizes of individual studies; when these were combined in a meta-analysis, the originally significant effects may have become statistically insignificant due to the larger overall sample size. Although our subgroup analyses indicated that Pilates and yoga were not statistically significant in improving maternal sleep, clinical studies [45, 48–50] have still shown that pregnant people in intervention groups experienced improved sleep compared with control groups. After practicing yoga and Pilates, there is typically an increase in vagal tone, a decrease in sympathetic discharge and postural heart rate response, and lower catecholamine levels, which contribute to relaxation and reduced responsiveness to external stimuli in pregnant people [51]. Despite evidence that PA benefits both pregnant people and foetuses without increasing adverse outcomes, many pregnant people remain hesitant to engage in PA because of concerns about associated risks. It will require more time and effort for healthcare professionals to disseminate knowledge and challenge traditional misconceptions about exercise during pregnancy. While our study aimed to explore the relationship between any form of PA and sleep in pregnant people, we were unable to include studies on unstructured exercise such as walking or room cleaning. Future research could focus on the impact of unstructured exercise on maternal sleep, and a network meta-analysis could be conducted to identify the most effective type of PA for improving sleep in pregnant people. Additionally, pregnant people’s preferences should be taken into account in future studies.

This study has several limitations. First, although all studies claimed to be randomised, only a minority reported details on allocation concealment and the blinding of participants and assessors. Relevant information often had to be inferred from the texts, which may have introduced a certain degree of selection bias. Future research should prioritise more rigorous reporting of study designs and characteristics. Second, the synthesis estimates may be biased by the exclusion of studies published in languages other than Chinese and English. Nevertheless, we searched PubMed, Embase, Web of Science, and Cochrane—comprehensive databases for organising literature—to capture as many relevant reports as possible. Third, variations in the duration, intensity, and delivery of PA interventions, along with significant discrepancies in the number of studies for certain interventions, contributed to heterogeneity among the studies, potentially limiting the generalisability of our findings. Finally, some subgroup analyses included few studies, and all relied on subjective self-reports rather than objective measures of sleep, which may have affected the robustness of the results.

Clinical and public health implications

Our review highlighted the significant impact of PA on sleep among pregnant people, with important implications for clinical practice. It demonstrated that PA is a promising approach for improving maternal health, suggesting that healthcare professionals should encourage pregnant people to engage in various forms of PA, provided there are no contraindications. Additionally, we identified maternal and trial-related characteristics to offer more specific guidance for healthcare professionals. This review represents a crucial step toward developing evidence-based clinical guidelines that integrate PA as a beneficial component of prenatal care services. Future trials should evaluate safety outcomes and the cost of implementing exercise interventions to further promote exercise among perinatal people.

Conclusions

The results of this meta-analysis suggest that PA interventions can effectively alleviate sleep problems in pregnant people, with effects influenced by participant type, duration, delivery method, and activity form. These findings have important implications for providing more detailed guidance to healthcare professionals. However, because of the limited number of subgroups and the high heterogeneity among studies, further high-quality research is needed to validate these conclusions and determine the optimal dose of PA in the future.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

PA

Physical activity

ITT

Intention-to-treat

Author contributions

LDM wrote the main manuscript text, LKQ did the discussion part based on clinical practice, XL and HM did the main searching and data processing part and LXP made decisions over different opinions during research progress.

Funding

The authors declare no funding support.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

IRB approval

The manuscript is a meta-analysis, so there is no IRB Approval.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.