Preterm birth (PTB) is defined as delivery before 37 weeks of pregnancy, which has long-term negative impacts on the development of newborns and may lead to stunting. PTB is becoming more and more common in developing and even some developed countries [1], although the underlying causes are still unknown. Premature delivery is a common adverse pregnancy outcome. Because of its serious impact on life and health, it brings a great burden on families and society, leading to perinatal and early neonatal death (especially in developing countries) and immature organ development, which leads to physical disability or mental disorders [2, 3]. PTB is regarded as a public health issue and one of the key risk factors contributing to the Global Burden of Disease. As a result, there is an urgent need to investigate PTB risk factors in order to reduce the occurrence of PTB and thus improve human health.

PTB has been linked to a variety of factors, including social and economic status, race, genetic influences, medical conditions, mental disorders, and environmental exposures [4]. Air pollution is a major component of environmental pollution, and it is considered as a risk factor of many diseases. According to the latest estimate of the World Health Organization, about 90% of the world’s population breathes air with high concentrations of pollutants, and as many as 7 million people die every year due to environmental and indoor air pollution [5]. The impact of air pollution on health has become a global public health problem (6). In addition to causing respiratory [7, 8] and cardiovascular diseases [9], more and more studies have investigated the influence of air pollution on premature delivery, including China [10], South Korea [11], Japan [12] and the United States [13]. Although most published studies have reported that various air pollutants are related to PTB, the nature of the pollutants studied and the related pregnancy outcomes are different. At present, there is a lack of toxicological data to guide pregnant women to choose the most relevant vulnerable exposure window [14]. Previous surveys have used a wide range of exposure windows (i.e., weeks, months and 3 months) to determine the time relationship between air pollution and adverse pregnancy outcomes [15]. Some studies have reported that exposure in the first trimester of pregnancy is associated with an increased risk of PTB [16]. Other studies have shown that exposure in the third trimester has a greater impact [17]. Therefore, the relationship between different pollutants exposure and different adverse pregnancy outcomes is complicated, and different periods of pollutant exposure during pregnancy have different effects on pregnancy outcomes [18]. Some previous studies have discussed the effect of air pollution on PTB [19–21], but they are mainly carried out in developed countries. Although the level of air pollution in Asian countries is usually higher, there has been a lack of research in these countries [10, 22]. Fetuses are particularly sensitive to environmental hazards because cell proliferation, differentiation and organ development accelerate during pregnancy, which may directly damage the fetus or interfere with placental function [23]. Several possible biological approaches have been put forward in previous studies. On the one hand, placental DNA adducts produced by air pollution are more common in mothers exposed to high levels of air pollution [24]. The high expression of fetal DNA adducts may reduce the ability of DNA repair and detoxification enzymes [25]. In previous studies, DNA adducts were associated with shortened pregnancy [26]. On the other hand, its mechanism may be related to hematological factors. The increase of blood viscosity caused by exposure to air pollution and hematological factors may be related to insufficient placental perfusion, which leads to PTB(24). Animal studies also support the finding that exposure to urban air pollution during pregnancy will inhibit fetal growth, which is closely related to the risk of premature delivery [27]. Therefore, this paper plans to collect the previous cohort and case-control studies, effectively increasing the sample size, and further improving the evidence level by including only the studies with high Newcastle-Ottawa Quality Score (NOS) and discussing the effects of air pollution exposure during different pregnancies on PTB.

The influence of air pollutants on premature delivery has become the focus of attention. This Meta-analysis included 18 eligible studies to investigate the correlation between the effects of each 10 μg/m³ increase in the mean concentration of five major ambient air pollutants (including PM_2.5, PM ₁₀, NO₂, SO₂ and O₃) during pregnancy exposure on preterm birth. The results showed that exposure to PM_2.5 throughout pregnancy and O₃ in late pregnancy was positively correlated with PTB. Every 10 μg/m³ increase in the average concentration of PM_2.5 in the whole pregnancy increased the risk of premature delivery by 4%, and every 10 μg/m³ increase in the average concentration of O₃ in the third trimester increased the risk of premature delivery by 1%. Every increase of 10 μg/m³ in the exposure concentration of residual pollutants during different pregnancy has no significant effect on the incidence of premature delivery. As a major component of air pollution mixture, PM_2.5 is more likely to harm human health due to its special characteristics, such as small size, large specific surface area, long floating time in the air, and adsorption capacity [29]. SUN, Chen et al. [29, 30] similarly found that PM_2.5 exposure throughout pregnancy was associated with preterm delivery, and an increase of PM_2.5 in early to mid-late pregnancy 10 μg/m³, the risk of preterm birth occurrence was not significant. Luo et al. [31] found that PM_2.5 exposure throughout pregnancy was not associated with preterm birth. Liu et al. [32] reported results shows that there is a significant correlation between low exposure of PM_2.5 in early pregnancy and premature delivery. This inconsistency may be due to the different number of studies involved due to different increments in pollutant exposure levels, search times, and search scopes, which may be the main reason for the discrepancy in study results. Current epidemiological evidence suggests that women in the first trimester of pregnancy tend to be less active outdoors when reported PM_2.5 concentrations are high, which may overestimate ambient PM_2.5 exposure and lead to bias. However, during periods of relatively low reported PM_2.5 concentrations, they are relatively less aware of self-protection against air pollution, which may lead to a high risk of preterm birth.

Pregnancy is a time when people are most susceptible to the effects of air pollution, although past researches have produced mixed findings. Early gestation (first or third month), according to some studies, is the window of susceptibility, however later gestation is supported by other studies (third gestation, last month, or last week). In the present study, the effect of different gestational pollutant exposures on preterm birth was investigated and only O₃ exposure in late pregnancy was found to promote preterm birth. The results of Ju et al. [33] also showed that O₃ exposure in early and mid-late pregnancy was significantly associated with preterm birth. This is similar to our results, suggesting that the susceptibility window may be related to the specificity of different pollutants and needs to be further explored.

It is well known that the toxicity and health effects of air pollutants may vary by geographic region. Therefore, it is reasonable to assess the effects of different gestational air pollutant exposures on PTB by regional subgroup analysis. Subgroup analyses in different regions showed regional differences in the effects of air pollution on PTB. The results showed that the exposure of PM_2.5 in the northern area and PM₁₀ in southern area during pregnancy was significantly related to the increase of PTB, but other pollutants did not find this correlation. Due to the differences in living habits, climate and pollutant concentrations between the northern and southern regions, it was found that the concentration of PM_2.5 in the northern region was higher than that in the southern region due to coal burning in winter. The quality of the results from long-term cohort studies is obviously better than that of case-control studies. Especially the prospective cohort studies, is considered to be the most reliable results in observational epidemiology. Therefore, we used different research methods to carry out subgroup analysis. The results showed that SO₂ exposure in the second and third trimester of pregnancy was significantly correlated with the decrease of PTB, while in the case-control study, SO₂ exposure in the first trimester of pregnancy was significantly correlated with the increase of PTB. The cohort study found that NO₂ exposure in early pregnancy was significantly correlated with the decrease of PTB, while NO₂ exposure in middle and late pregnancy was not significantly correlated with premature delivery.

We discovered substantial variability among the included studies, which is predictable in view of the limitations of our meta-analysis. In order to quantitatively aggregate individual estimates from studies with substantial heterogeneity, we employed a random effects model. Subgroup analysis was also used to investigate the causes of heterogeneity. The results show that, even though the study design and study area only partially explain the variability, most subgroup analyses still revealed significant heterogeneity. These results imply that due to the small number of related studies, variability of the included studies may be affected by other factors that are not considered in this analysis, such as socio-economic status and chemical composition of pollutants. Therefore, more studies are needed to investigate the causes of heterogeneity in the future. In several studies, personal samplers were used to measure the exposure of pregnant women to air pollutants, and the results were more accurate. Therefore, more research on personal information might be helpful for future analysis.