The present systematic review and meta-analysis of 10 observational studies demonstrated that short sleep duration, as well as poor sleep quality were associated with a significantly increased risk of preterm birth. This association highlights the vital significance of pregnant women to reduce the risk of premature birth.
For sleep duration, the results of subgroup analysis stratified by trimester suggested that slight differences existed between studies conducted in the first/second trimester vs. the third trimester, which might be attributed to the involvement of different proportions of short sleepers in first/second vs. third trimester. For instance, Micheli et al.  conducted a cohort study (n = 1091), in which 23% of pregnant women reported a sleep duration of ≤5h in the third trimester. However, Reutrakul et al.  reported that about 56% of pregnant women experienced sleep deprivation in the second trimester, but this study had a relatively small cohort (n = 116). Meanwhile, Li et al.  enrolled participants with a similar proportion of short sleepers in all trimesters and the results were similar with the main findings. Two studies  reported that about 50% of pregnant women experienced sleep deprivation, but the sleep durations differed (< 7 and < 8 h/night, respectively). Furthermore, subgroup analysis stratified by geographical location roughly showed that a short sleep duration may be a risk factor for preterm birth outside of America, but not in America, which might have resulted from the relatively small number of preterm birth cases (n = 419) in the American studies. Preterm birth rates vary among countries, even different regions in the same country. Also, income and education differences may also affect sleep duration and preterm birth rates [37, 38].
The findings of subgroup analysis for sleep quality suggest that geographical location might be a source of heterogeneity. Although the possibility of a chance finding was ruled out by the limited sample size, Warland et al.  speculated that African Americans may exhibit heightened sensitivity to the adverse physiological sequelae of poor sleep quality. Two other studies indicated that pregnant women with clinically disturbed sleep (PSQI > 5) accounted for a similar proportion (about 60%), despite the study being conducted in various regions within the USA [31, 36]. While a study conducted in Japan enrolled pregnant women with a lower proportion of poor sleepers at gestational weeks 16, 24, and 32 (27%, 34%, 37%, and 41% of the samples, respectively) . Whereas a study conducted in the USA reported that pregnant women at gestational week 14–16 accounted for 36.4% (n = 48) of the sample . Moreover, the results of subgroup analysis by trimester showed that the relationship was significant just in the first and second trimesters, but not the third. Therefore, the trimester and geographical location strongly influenced the association between sleep duration and quality, which can guide pregnant women in different regions to improve sleep practice in order to reduce the risk of preterm birth.
The findings of the current study raise questions about the potential mechanisms underlying the increased risk of preterm birth due to sleep disorders. Sleep deprivation partially accounts for the proinflammatory cytokine response [39-41], immune changes , and greater susceptibility to infections . It is well established that inflammation and infection are highly significant risk factors for preterm birth [44, 45]. Additionally, a short sleep duration and poor sleep quality may result from stress and as a physiological stressor per se, causing stress “overload” and activation of the stress system, may lead to prematurity through impairment of the hypothalamus-pituitary-adrenal axis and activation of the proinflammatory system . On the other hand, physiological and hormonal changes also affect sleep practices. Higher levels of estrogen and progesterone during pregnancy contribute to poor sleep quality and also influence the secretion of other hormones, such as cortisol and melatonin, which can increase arousal [47, 48]. Lastly, because disturbed sleep may disrupt normal remodeling of the maternal blood vessels and increased sympathetic activity, placental blood flow was decreased [49, 50], which may be a mechanism underlying preterm birth.
The strengths of the present meta-analysis lie in its quantitative analysis of the association between sleep duration and quality and the risk of prematurity with large numbers of participants (n = 5693) and instances of preterm birth for sleep duration (n=1248) and sleep quality (n=156). Large sample size provided strong power for analyses conducted and conclusions showed. Although Warland et al.  reviewed the literature to access the aforementioned association, there was heterogeneity; therefore, subgroup analyses were limited. However, to further explore the presence of heterogeneity, as well as to better elucidate the relationship between sleep duration and quality and the risk of premature birth, numerous subgroup and sensitivity analyses were conducted. Furthermore, quality assessment showed that all of the included studies were at a low risk of bias.
Findings from the present meta-analysis should be interpreted in light of several limitations. First, the present meta-analysis was prone to inherent recall and selection bias due to the inclusion of original observational studies. Although case-control studies are more susceptible to bias than cohort studies, the results were robust after exclusion of the only case-control study from the analyses. Furthermore, the PSQI is an important clinical and research tool to gauge sleep quality . However, the PSQI includes sleep duration in the scoring, and therefore, short sleep duration was included as an outcome of "sleep quality". Consistently, the pooled effect sizes for poor sleep quality (RR=1.54) was similar with that for shorter duration (RR=1.32) found by the meta-analysis. Moreover, taking the variation of study population enrolled into consideration, geographical location likely contributed to the heterogeneity of effect estimates. Furthermore, since all included studies measured sleep quality and sleep duration using questionnaires, however, self-reported sleep quality and duration are not always perfectly aligned with objective sleep quality and duration. Third, because the pooled effect estimates were mostly derived from observational studies, susceptibility to confounding factors remained a concern. Some common chronic diseases have been correlated with prematurity, as mediators between short sleep duration and preterm birth, such as diabetes [31, 52], hypertension [53, 54], and obesity . Of note, self-reported sleep disturbances are predictive of the incidence of major depression and strongly precede a series of symptoms of depression [56, 57]. The association between depression syndrome and risk of preterm delivery has been observed . Thus, early intervention poor sleep quality or short sleep duration, which may be indicators of early depression, can reduce the risk of preterm birth. However, the observational studies included in this meta-analysis were restricted by the lack of controls for these potentially-relevant confounders. Hence, further studies are warranted with better designs to take these confounders or mediators into account or fully adjust for these confounders and better rule out the potential effects of residual confounding. Fourth, the results slightly differed among trimesters and the type of preterm birth, thus these findings might be attributed to the limited sample size. Additionally, as the comparison of sleep duration differed considerably among the included studies, dose–response analysis was not conducted. Notably, several included studies suggested a potential U-shaped association between sleep duration and preterm birth. Additionally, one included study suggested a potential non-linear (U-shaped) association between sleep duration and preterm birth . However, since a limited number of the included studies met the criteria of linear/non-linear dose-response analysis, we failed to carry out such analysis in the present meta-analysis. Also, seven studies were excluded due to no valid data for analyzing the association between sleep quantity and quality and preterm birth [10, 24-29]. Of note, although we calculated the power of the main analysis which suggested that this study had over 80% statistical power to identify sleep duration/quality for preterm birth with minimum OR of 1.20 (risk factor for sleep duration) and 1.5 (risk factor for sleep quality), limited sample sizes in several subgroup analyses still may restrict the interpretation of our findings. Therefore, on the basis of these limitations, priority should be given to large, adequately powered, cohort studies using standard definitions of maternal sleep duration and quality with effective data analysis. Furthermore, more comparison groups in the primary studies are needed to evaluate the possible non-linear aforementioned association.