Joint associations between maternal exposure to ambient air pollutants during pregnancy and low Apgar score in newborns: a birth cohort study in China Authors: Zhijian Chen, Bingyi Lin, Zhongai Ouyang, Weigui Ni, Long Jiang, Bo Wu, Shaowu Chen, Lijuan Lai, Siqi Chen, Haoqu Zheng, Minting Zhu, Yi Jing, Xi Yu, Jingjie Fan Published: 2026 ABSTRACT BackgroundEpidemiological evidence on the risk of low Apgar score in newborns associated with ambient air pollutants, their joint associations, and sensitive exposure windows is limited. This study aimed to investigate the independent and joint associations between exposure to ambient air pollutants during different pregnancy trimesters and low Apgar score in newborns.MethodsTwenty thousand eight hundred ninety-four pregnant women were included in the prospective birth cohort in Shenzhen. Concentrations of PM2.5, PM10, NO2, O3, SO2, and CO were assessed based on residential addresses of participants. Generalized additive model (GAM) was used to assess the independent associations of air pollutants exposure with low Apgar score, and restricted cubic spline (RCS) model was used to analyze the dose-response trends. Quantile g-computation (Qgcomp) and weighted quantile sum (WQS) regression models were further used to investigate the joint associations and constituent contributions.ResultsSecond-trimester exposure to PM2.5, PM10, and NO2 was associated with an increased risk of low 1-minute Apgar score (PM2.5: OR = 1.64, 95% CI: 1.10, 2.44; PM10: OR = 1.59, 95% CI: 1.01, 2.50; NO2: OR = 1.48, 95% CI: 1.11, 1.98, per interquartile range increase). In the mixed-pollutant models, both Qgcomp and WQS models showed that each one-quartile increase in the air pollutant mixture was significantly associated with low 1-minute Apgar score (Qgcomp: OR = 1.48, 95% CI: 1.06, 2.07; WQS: OR = 1.15, 95% CI: 1.09, 1.21). PM2.5 and NO2 were the primary contributors to the risk (Qgcomp: 48.0%, 34.5%; WQS: 23.6%, 39.4%).ConclusionsSecond-trimester exposure to air pollutants was associated with an increased risk of low 1-minute Apgar score, and PM2.5 and NO2 may be the key relative contributors. These findings support the need for further research and increased public health attention to air pollutant exposure during pregnancy to prevent adverse neonatal health outcomes.Trial registrationNot applicable.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28296-y. Introduction The Apgar score is one of the most widely used clinical tools for assessing the health status of newborns at birth. Initially developed by Dr. Virginia Apgar in 1953, it involves a rapid evaluation of newborns at 1 min after birth [1]. The Apgar score consists of five components: color, heart rate, reflexes, muscle tone, and respiration. This assessment aims to determine the clinical status of the newborn and identify the need for prompt interventions to provide respiratory support [2]. Subsequent studies have shown that the 5-minute Apgar score is a good predictor of neonatal survival [3]. From a clinical perspective, the 1-minute Apgar score reflects the immediate health status of the newborn at birth more directly than the 5-minute Apgar score [2]. Numerous studies have provided evidence that a low Apgar score is a key predictor of neonatal asphyxia and is closely associated with a higher risk of adverse long-term health outcomes, such as neurodevelopmental disorders in childhood [4–6]. Various factors can affect the Apgar score, including pregnancy complications, medication use, and gestational age [7–9]. In recent years, studies have also suggested that environmental factors may significantly influence the Apgar score [10]. According to the Global Burden of Disease (GBD) studies, exposure to ambient air pollution is a major risk factor for disease burden and has become a critical public health challenge that urgently needs to be addressed worldwide [11]. Pregnant women, as a special population, are particularly sensitive to exposure to air pollutants. Studies have demonstrated that exposure to air pollutants is closely related to adverse health conditions in pregnant women, such as gestational diabetes mellitus, gestational hypertension, and liver function impairment [12–14]. The intrauterine environment serves as the direct environment for fetal growth and development. Any disruption to maternal physiological homeostasis may further increase the risks of adverse pregnancy outcomes such as preterm birth and low birth weight. Consequently, this may affect the newborn’s immediate health condition at birth [15–17]. Therefore, a comprehensive investigation of the associations between prenatal air pollution exposure and key newborn health indicators, such as the Apgar score, can provide scientific evidence for clarifying the potential hazards of air pollution to maternal and fetal health. This holds significant public health and clinical implications. Few previous epidemiological studies have assessed the associations between exposure to air pollutants and Apgar score in newborns. These studies have advanced the understanding of how air pollution exposure affects newborn health at birth. However, these studies have mainly focused on the effects of single ambient particulate matter, with limited research on other gaseous air pollutants [18, 19]. Additionally, existing studies have overlooked the actual characteristics of human exposure to a complex mixture of multiple air pollutants in real-world environments. Therefore, it is particularly important to systematically assess the joint health impacts of air pollutant mixture. However, limited research has investigated the joint associations between prenatal air pollutant exposure and low Apgar score in newborns, and the differences in potential sensitive windows during different pregnancy trimesters remain unclear. Shenzhen, a typical immigrant city in China, has a large population of women of childbearing age with relatively high education levels [20, 21], providing an ideal population basis for investigating the effects of environmental factors on maternal and neonatal health. In recent years, with the implementation of a series of air pollution control policies, air quality in Shenzhen has improved [22]. However, the concentration of air pollutants such as NO2 remains higher than the national average [23], and O3 levels are still on the rise [24]. Moreover, the health impacts and economic burden associated with PM2.5 and O3 remain substantial [22]. Therefore, based on data from a birth cohort in Shenzhen, this study investigated the associations between exposure to air pollutants during different trimesters of pregnancy and low Apgar score in newborns. Drawing on previous environmental epidemiological studies, we focused on six common air pollutants (PM2.5, PM10, NO2, O3, SO2, and CO) [25, 26]. Various mixed-pollutant models were employed to assess their joint associations with low Apgar score and to determine the contribution of each air pollutant to the overall effect. The findings provide scientific evidence to support the development of targeted intervention strategies aimed at reducing the potential adverse effects of air pollution on pregnant women and their offspring. Methods Study design and participants This study was conducted based on a birth cohort established at Shenzhen Maternity and Child Healthcare Hospital, which was designed to investigate the impact of environmental factors on maternal and child health. The baseline characteristics of this birth cohort have been described in previously published articles [27, 28]. Briefly, pregnant women were recruited during their first prenatal visit at the hospital and provided written informed consent after face-to-face explanation of the protocol. All participants completed a questionnaire at enrollment. The procedures were conducted in accordance with the principles of the Declaration of Helsinki. Information regarding ethical approval and informed consent has been provided in the Declarations section. A total of 32,176 pregnant women were recruited from July 2021 to April 2023 in a birth cohort from Shenzhen. After excluding those aged less than 18 years (N = 13), those with multiple pregnancy (N = 1,131), and those who had miscarriage or stillbirth (N = 509), 30,523 participants were initially included in the study. Subsequently, 4,119 pregnant women who were lost to follow-up were excluded (13.49% of the initial included participants). Additionally, participants were excluded if they did not deliver at this hospital (N = 2,262), or had missing data on key variables such as air pollutants and meteorological factors (N = 3,248). Finally, a total of 20,894 pregnant women were included in the analysis (Fig. S1). Table S1 shows that most baseline variables were not statistically different between the included group and the lost-to-follow-up group, and Table S2 shows similar results between the included group and excluded group. Although a few variables showed statistically significant differences due to the large sample size, the absolute differences were small. Overall, baseline characteristics were generally balanced across the groups, suggesting no substantial selection bias. Exposure assessment to environmental variables Daily air pollution data from January 1, 2021, to January 31, 2024, including inhalable particulate matter (PM10), nitrogen dioxide (NO2), sulfur dioxide (SO2), and carbon monoxide (CO), were obtained from the China High Air Pollutants datasets (CHAP, http://www.geodata.cn). Fine particulate matter (PM2.5) and daily maximum 8-hour average ozone (MDA8 O3) were obtained from the Tracking Air Pollution in China (TAP, http://tapdata.org.cn/). Both datasets integrate multi-domain information such as ground-based observations, satellite remote sensing products, and atmospheric reanalysis to generate high-spatial-resolution daily air pollutant concentration data (CHAP: 1 km × 1 km; TAP: 10 km × 10 km). The results of the cross-validation coefficient of determination and root-mean-square error indicated that the CHAP and TAP datasets has good accuracy [29–32]. Based on the daily air pollution data, the average exposure levels of air pollutants were calculated for each trimester. The geographic distribution of study participants’ residences and trimester-specific air pollutant levels are shown in Fig. S2. Daily mean temperature (TM) and daily relative humidity (RH) data were obtained from the ERA5 dataset provided by the European Centre for Medium-Range Weather Forecasts (ECMWF). This is a global climate-atmospheric reanalysis dataset with a spatial resolution of 0.1° × 0.1°, providing high-precision meteorological information resources [33, 34]. Based on the acquired meteorological data, the mean temperature and relative humidity during each trimester were calculated for each pregnant woman. In this study, trimesters of pregnancy are defined as follows: the first trimester is from the conception date to the 13th week of gestation; the second trimester is from the 14th to the 27th week of gestation; and the third trimester is from the 28th week of gestation to the delivery date. Outcome definition Apgar scores are routinely measured at 1, 5, and 10 minutes after birth, with a total score ranging from 0 to 10. Since the study population consisted of Chinese population, a low Apgar score was defined as 7 or below according to Chinese diagnostic criteria and guidelines, and with reference to previous studies conducted in China [19, 35]. In this study, newborns with an Apgar score ranging from 8 to 10 were classified as having a normal score, while those with an Apgar score ranging from 0 to 7 were defined as having a low Apgar score. Based on the clinical significance of the Apgar score, the 1-minute Apgar score reflects the immediate health status of the newborn at birth, while the 5-minute and 10-minute Apgar scores indicate the recovery of the newborn after medical interventions. In this study, we did not analyze the 5-minute and 10-minute Apgar scores, owing to insufficient statistical power resulting from the very low number of cases, as well as the potential confounding effects of medical interventions on these two Apgar scores. Covariates Based on the directed acyclic graph (DAG) shown in Fig. S3 and previous studies [36–39], potential covariates and confounders were selected from the questionnaire (Table S3) and medical records. Variables derived from the questionnaire included maternal demographic characteristics (including maternal age, pre-pregnancy BMI), socioeconomic status (educational level), and lifestyle habits (smoking and alcohol drinking). Educational level was categorized as high school or below, undergraduate, or above undergraduate. Smoking status was categorized as never smoking, former smoking (smoked before pregnancy but not during pregnancy), occasional smoking during pregnancy, or frequent smoking during pregnancy. Alcohol drinking was categorized as no alcohol drinking, occasional alcohol drinking during pregnancy, or frequent alcohol drinking during pregnancy. In subsequent analyses, occasional and frequent smoking during pregnancy were combined into a “current smoking” category, and occasional and frequent alcohol drinking during pregnancy were combined into an “alcohol drinking” category. Information on gestational age at delivery, gestational hypertension, gestational diabetes mellitus, and infant gender was obtained from medical records. Season of conception was determined based on the date of last menstrual period recorded in medical records, with May to October defined as the warm season and November to April as the cold season [40]. It should be noted that although gestational age, gestational hypertension and gestational diabetes mellitus may lie on the potential causal pathway between air pollutant exposure and Apgar score, we included them as covariates in the analysis. Previous studies have identified these variables as known risk factors for adverse neonatal outcomes, which may exert potential confounding effects [19, 38]. To further assess whether these variables act as potential mediators, we performed mediation analyses. The results showed that none of the three variables had a statistically significant indirect effect (all ACME P‑values > 0.05, Table S4), suggesting that they may not serve as significant mediators in the associations between air pollutant exposure and low Apgar score in our study. Therefore, including them as covariates in the analysis is unlikely to introduce meaningful overadjustment. In addition, these variables were included stepwise in multiple models to evaluate the stability of the effect estimates after adjustment for these variables. Statistical analysis The characteristics of participants are reported as Mean ± SD or number (%). Air pollutant and meteorological factor exposures are described using the median, percentiles (P25, P75), and interquartile range (IQR). Pearson correlation analysis was used to assess the correlations between different air pollutants within the same trimester. Firstly, we used Generalized Additive Model (GAM) with logistic regression to separately estimate the independent associations between ambient air pollutants and low 1-minute Apgar scores in newborns, with the results presented as odds ratios (ORs) and 95% confidence intervals (CIs) for each IQR increase in air pollutants. All pollutants were entered into the model using their original concentration values. Meanwhile, mean temperature and relative humidity were adjusted for using natural cubic spline functions in both unadjusted and adjusted models. Based on previous research experience and considering the root mean square error values of generalized cross validation, the degrees of freedom (df) of the function were set to 3 (df = 3) [41, 42]. Three sequential models were constructed. The Crude Model included only the adjustment for mean temperature and relative humidity using natural cubic splines (df = 3). Model 1 was further adjusted for maternal age, pre-pregnancy BMI, infant gender, conception season, education level, smoking status, and alcohol-drinking status based on the Crude Model. Model 2 was additionally adjusted for gestational age, gestational hypertension, and gestational diabetes mellitus based on Model 1. To assess the robustness of the model after adjustment, we further explored the dose-response relationships between exposure to each air pollutant and low 1-minute Apgar score using a restricted cubic spline (RCS) model with four knots. The reference values (OR = 1) were set at the 5th percentile, and the knots were set at 5th, 35th, 65th and 95th percentiles of the concentrations of each air pollutant. Secondly, to address the issue of multicollinearity among air pollutants, this study further used different mixed-pollutant models to estimate the joint associations of co-exposure to air pollutants with low 1-minute Apgar score, including weighted quantile sum (WQS) regression and quantile g-computation (Qgcomp) [43]. Similarly, we included the six air pollutants directly into the mixed-pollutant models using their original concentration values. WQS regression is an analytical method specifically designed to estimate mixed exposure effects. It can effectively handle the correlations between exposure variables and assess the relative importance of each variable in influencing the outcome through the weights assigned to each exposure variable by the model. This method is applicable when the associations between each component of the mixed exposure and the outcome indicator are linear and in the same direction [44]. Qgcomp is a method further improved based on WQS regression. It enables pollutants to exhibit complex interactions rather than simple linear effects, does not require the assumption of directional homogeneity, and allows for non-linear effects and associations in different directions between exposures and outcomes [45]. In this study, six air pollutants (PM2.5, PM10, NO2, O3, SO2, and CO) were included in the model, and overall OR with 95% CI, as well as the contribution ratio of each pollutant, were estimated. Thirdly, we included Apgar score as a continuous variable in the GAM model and the mixed-pollutant models, adjusted for the aforementioned covariates, to investigate the associations between ambient air pollutant exposure and 1-minute Apgar score from different perspectives. Finally, a stratified analysis was conducted, and interaction terms were used to assess the potential impact of covariates, including maternal age, pre-pregnancy BMI, infant gender, conception season, education level, status of alcohol drinking, gestational hypertension and gestational diabetes mellitus. Sensitivity analysis We conducted a series of sensitivity analyses to test the robustness of the results. First, we adjusted the df of the spline function in the GAM model to 4–6, respectively, to examine the robustness of the model. Second, we constructed two-pollutant models to explore whether the results remained stable after controlling for other air pollutants [46]. Prior to constructing the two-pollutant models, a correlation analysis was performed among the pollutants. To avoid multicollinearity issues, any two pollutants with a high correlation (r > 0.7) were not included together in the two-pollutant models. Third, studies have shown that preterm birth is associated with low Apgar scores [47]. we therefore excluded preterm infants and reanalyzed the main findings in the term infant population to verify the robustness of the associations between ambient air pollutant exposure and Apgar scores. Fourth, according to the commonly used international definition [1, 47], a low Apgar score is defined as < 7 (i.e., 0–6 as low and 7–10 as normal). To assess whether our results were sensitive to the choice of cut-off, we redefined the outcome based on this definition and repeated the main analyses. Statistical analyses in this study were performed using R software (Version 4.5.1), with a two-tailed P value < 0.05 considered statistically significant. Results Characteristics of study participants and distributions of environmental variables Table 1 presents the basic characteristics of the 20,894 participants. Approximately half of the participants (50.46%) were aged between 30 and 35 years. The mean gestational age at delivery was 38.67 ± 1.46 weeks, and the mean pre-pregnancy BMI was 21.04 ± 2.77 kg/m2. Over 70% of the participants had an undergraduate education. The vast majority of pregnant women had no history of smoking or alcohol consumption. A total of 864 pregnant women (4.14%) were diagnosed with gestational hypertension, while 4,035 pregnant women (19.31%) were diagnosed with gestational diabetes. The gender ratio of infants was approximately 1:1, and pregnancies were also evenly distributed between cold and warm seasons. Among the 20,894 participants included in the analysis, a total of 230 participants (1.10%) delivered newborns with a low 1-minute Apgar score. Table 1General characteristics of total participants (N = 20,894)CharacteristicsMean ± SD or N (%)Maternal age (years)31.45 ± 4.11 < 30 years6924 (33.14) 30–35 years10,543 (50.46) > 35 years3427 (16.40)Pre-pregnancy BMI (kg/m2)21.04 ± 2.77 < 18.53437 (16.45) 18.5–2414,587 (69.81) > 242870 (13.74)Gestational age (weeks)38.67 ± 1.46 ≥ 3719,718 (94.37) < 371176 (5.63)Education level High school or below3196 (15.30) Undergraduate15,031 (71.94) Above undergraduate2667 (12.76)Smoking Never20,432 (97.79) Current69 (0.33) Former393 (1.88)Alcohol drinking No19,026 (91.06) Yes1868 (8.94)Conception season Cold9350 (44.75) Warm11,544 (55.25)Gestational hypertension No20,030 (95.86) Yes864 (4.14)Gestational diabetes mellitus No16,859 (80.69) Yes4035 (19.31)Infant gender Male11,169 (53.46) Female9725 (46.54)1-minute Apgar score 0–7230 (1.10) 8–1020,664 (98.90)5-minute Apgar score 0–736 (0.17) 8–1020,858 (99.83)SD standard deviation, BMI body mass index Table 1General characteristics of total participants (N = 20,894)CharacteristicsMean ± SD or N (%)Maternal age (years)31.45 ± 4.11 < 30 years6924 (33.14) 30–35 years10,543 (50.46) > 35 years3427 (16.40)Pre-pregnancy BMI (kg/m2)21.04 ± 2.77 < 18.53437 (16.45) 18.5–2414,587 (69.81) > 242870 (13.74)Gestational age (weeks)38.67 ± 1.46 ≥ 3719,718 (94.37) < 371176 (5.63)Education level High school or below3196 (15.30) Undergraduate15,031 (71.94) Above undergraduate2667 (12.76)Smoking Never20,432 (97.79) Current69 (0.33) Former393 (1.88)Alcohol drinking No19,026 (91.06) Yes1868 (8.94)Conception season Cold9350 (44.75) Warm11,544 (55.25)Gestational hypertension No20,030 (95.86) Yes864 (4.14)Gestational diabetes mellitus No16,859 (80.69) Yes4035 (19.31)Infant gender Male11,169 (53.46) Female9725 (46.54)1-minute Apgar score 0–7230 (1.10) 8–1020,664 (98.90)5-minute Apgar score 0–736 (0.17) 8–1020,858 (99.83)SD standard deviation, BMI body mass index General characteristics of total participants (N = 20,894) SD standard deviation, BMI body mass index Table 2 shows the distributions of environmental variables. Most air pollutants were not normally distributed, and their concentrations are reported as median (25th-75th percentile). In the first trimester, the concentrations of PM2.5, PM10, NO2, O3, SO2, and CO were 21.02 (16.20-25.14) µg/m3, 36.92 (29.62–43.26) µg/m3, 26.87 (23.35–30.86) µg/m3, 92.20 (85.49–99.11) µg/m3, 7.15 (6.51–7.81) µg/m3, and 0.70 (0.66–0.74) mg/m3, respectively. In the second trimester, the corresponding values were 21.22 (16.09–25.04) µg/m3, 37.64 (30.28–43.60) µg/m3, 26.31 (22.35–30.22) µg/m3, 90.46 (85.11–97.40) µg/m3, 6.95 (6.38–7.63) µg/m3, and 0.71 (0.66–0.73) mg/m3. In the third trimester, most pollutant concentrations declined: PM2.5 was 18.76 (13.71–22.79) µg/m3, PM10 was 33.98 (26.24–40.68) µg/m3, NO2 was 23.94 (21.33–27.50) µg/m3, O3 was 88.96 (81.55–98.01) µg/m3, SO2 was 6.77 (6.23–7.39) µg/m3, and CO was 0.69 (0.65–0.72) mg/m3. The IQR for each pollutant equals the difference between the 75th and 25th percentiles. Mean temperature, relative humidity and a comparison of air pollutant exposure concentrations across the three trimesters are all presented in Table 2. There were significant differences in the exposure concentrations of PM2.5, PM10, NO2, O3, SO2, and CO among the three trimesters. Positive correlations were observed between the concentrations of various air pollutants. Across all trimesters, the correlation between O3 and SO2 was consistently the highest, while the correlation between PM2.5 and NO2 was consistently the lowest (Fig. S4). Table 2Descriptive statistics of air pollutant concentrations and meteorological variablesExposure factorsFirst trimesterSecond trimesterThird trimesterP-valueMedian (25th−75th)IQRMedian (25th−75th)IQRMedian (25th−75th)IQRPM2.5 (µg/m3)21.02 (16.20–25.14)8.9421.22 (16.09–25.04)8.9518.76 (13.71–22.79)9.08< 0.001PM10 (µg/m3)36.92 (29.62–43.26)13.6437.64 (30.28–43.60)13.3233.98 (26.24–40.68)14.44< 0.001NO2 (µg/m3)26.87 (23.35–30.86)7.5126.31 (22.35–30.22)7.8723.94 (21.33–27.50)6.17< 0.001O3 (µg/m3)92.20 (85.49–99.11)13.6290.46 (85.11–97.40)12.2988.96 (81.55–98.01)16.46< 0.001SO2 (µg/m3)7.15 (6.51–7.81)1.306.95 (6.38–7.63)1.256.77 (6.23–7.39)1.16< 0.001CO (mg/m3)0.70 (0.66–0.74)0.080.71 (0.66–0.73)0.070.69 (0.65–0.72)0.07< 0.001Mean temperature (°C)21.80 (17.95–26.26)8.3121.28 (17.71–25.78)8.0724.40 (19.19–27.13)7.94< 0.001Relative humidity (%)76.89 (72.09–83.43)11.3477.02 (72.07–83.34)11.2780.59 (74.71–85.13)10.42< 0.00125th-75th: the 25th and 75th percentiles (first and third quartiles)IQR interquartile range, PM2.5particulate matter with an aerodynamic diameter ≤ 2.5 μm, PM10particulate matter with an aerodynamic diameter ≤ 10 μm, NO2nitrogen dioxide, O3ozone, SO2sulfur dioxide, COcarbon monoxide Table 2Descriptive statistics of air pollutant concentrations and meteorological variablesExposure factorsFirst trimesterSecond trimesterThird trimesterP-valueMedian (25th−75th)IQRMedian (25th−75th)IQRMedian (25th−75th)IQRPM2.5 (µg/m3)21.02 (16.20–25.14)8.9421.22 (16.09–25.04)8.9518.76 (13.71–22.79)9.08< 0.001PM10 (µg/m3)36.92 (29.62–43.26)13.6437.64 (30.28–43.60)13.3233.98 (26.24–40.68)14.44< 0.001NO2 (µg/m3)26.87 (23.35–30.86)7.5126.31 (22.35–30.22)7.8723.94 (21.33–27.50)6.17< 0.001O3 (µg/m3)92.20 (85.49–99.11)13.6290.46 (85.11–97.40)12.2988.96 (81.55–98.01)16.46< 0.001SO2 (µg/m3)7.15 (6.51–7.81)1.306.95 (6.38–7.63)1.256.77 (6.23–7.39)1.16< 0.001CO (mg/m3)0.70 (0.66–0.74)0.080.71 (0.66–0.73)0.070.69 (0.65–0.72)0.07< 0.001Mean temperature (°C)21.80 (17.95–26.26)8.3121.28 (17.71–25.78)8.0724.40 (19.19–27.13)7.94< 0.001Relative humidity (%)76.89 (72.09–83.43)11.3477.02 (72.07–83.34)11.2780.59 (74.71–85.13)10.42< 0.00125th-75th: the 25th and 75th percentiles (first and third quartiles)IQR interquartile range, PM2.5particulate matter with an aerodynamic diameter ≤ 2.5 μm, PM10particulate matter with an aerodynamic diameter ≤ 10 μm, NO2nitrogen dioxide, O3ozone, SO2sulfur dioxide, COcarbon monoxide Descriptive statistics of air pollutant concentrations and meteorological variables 25th-75th: the 25th and 75th percentiles (first and third quartiles) IQR interquartile range, PM2.5particulate matter with an aerodynamic diameter ≤ 2.5 μm, PM10particulate matter with an aerodynamic diameter ≤ 10 μm, NO2nitrogen dioxide, O3ozone, SO2sulfur dioxide, COcarbon monoxide Associations between ambient air pollutants and low 1-minute Apgar score Table 3 presents the OR and 95% CI for the associations between each IQR increase in air pollutant concentrations and low 1-minute Apgar score. Single-pollutant model analyses revealed that second-trimester exposure to PM2.5, PM10, and NO2 was associated with increased risks of low 1-min Apgar score (PM2.5: OR = 1.64, 95% CI: 1.10, 2.44; PM10: OR = 1.59, 95% CI: 1.01, 2.50; NO2: OR = 1.48, 95% CI: 1.11, 1.98), while exposures during the first and third trimesters showed no statistically significant effects. Further RCS model curves demonstrated a positive dose-response trend between second-trimester exposures to PM2.5, PM10 and NO2 and low 1-minute Apgar score, which was characterized by a linear relationship (P for non-linear association > 0.05) (Fig. 1). Table 3Associations between exposure to ambient air pollutants and low 1-minute Apgar score in single-pollutant modelAir pollutants/ModelsFirst trimesterSecond trimesterThird trimesterOR (95%CI)P-valueOR (95%CI)P-valueOR (95%CI)P-valuePM2.5 Crude model a0.98 (0.64, 1.51)0.9421.57 (1.06, 2.34)0.026*1.22 (0.77, 1.92)0.399 Model 1 b0.99 (0.64, 1.53)0.9651.62 (1.09, 2.42)0.018*1.33 (0.85, 2.08)0.210 Model 2 c1.01 (0.65, 1.55)0.9811.64 (1.10, 2.44)0.016*1.04 (0.68, 1.59)0.859PM10 Crude model1.64 (0.94, 2.87)0.0841.53 (0.98, 2.38)0.0611.09 (0.65, 1.85)0.746 Model 11.58 (0.90, 2.75)0.1091.55 (0.99, 2.43)0.0561.14 (0.65, 1.97)0.654 Model 21.55 (0.89, 2.71)0.1191.59 (1.01, 2.50)0.045*0.87 (0.52, 1.45)0.591NO2 Crude model1.12 (0.86, 1.46)0.4051.46 (1.10, 1.95)0.010*1.25 (0.92, 1.70)0.150 Model 11.12 (0.84, 1.49)0.4341.46 (1.10, 1.95)0.010*1.24 (0.91, 1.68)0.180 Model 21.14 (0.86, 1.52)0.3681.48 (1.11, 1.98)0.007**1.14 (0.86, 1.52)0.376O3 Crude model1.03 (0.80, 1.32)0.8471.05 (0.88, 1.27)0.5900.80 (0.64, 1.00)0.046* Model 10.99 (0.78, 1.25)0.9151.04 (0.86, 1.26)0.6680.81 (0.65, 1.01)0.062 Model 21.01 (0.79, 1.29)0.9321.05 (0.87, 1.27)0.6150.83 (0.68, 1.01)0.067SO2 Crude model1.14 (0.90, 1.44)0.2681.22 (0.97, 1.53)0.0861.03 (0.83, 1.28)0.797 Model 11.15 (0.90, 1.47)0.2641.23 (0.97, 1.54)0.0831.02 (0.82, 1.27)0.871 Model 21.16 (0.91, 1.49)0.2331.24 (0.98, 1.56)0.0690.97 (0.78, 1.19)0.742CO Crude model1.30 (0.97, 1.74)0.0821.18 (0.90, 1.54)0.2371.14 (0.91, 1.43)0.245 Model 11.27 (0.95, 1.71)0.1131.17 (0.89, 1.53)0.2511.12 (0.89, 1.40)0.341 Model 21.25 (0.93, 1.68)0.1421.16 (0.89, 1.52)0.2741.00 (0.79, 1.26)0.996OR and 95%CI were calculated for per IQR increment for each air pollutant. a Crude Model: adjusted only for mean temperature and relative humidity using natural cubic splines (df = 3); b Model 1: further adjusted for maternal age, pre-pregnancy BMI, infant gender, conception season, education level, smoking status, and alcohol-drinking status based on Crude Model; c Model 2: further adjusted for gestational age, gestational hypertension, and gestational diabetes mellitus based on Model 1OR odds ratio, CI confidence interval* P < 0.05, ** P < 0.01 Table 3Associations between exposure to ambient air pollutants and low 1-minute Apgar score in single-pollutant modelAir pollutants/ModelsFirst trimesterSecond trimesterThird trimesterOR (95%CI)P-valueOR (95%CI)P-valueOR (95%CI)P-valuePM2.5 Crude model a0.98 (0.64, 1.51)0.9421.57 (1.06, 2.34)0.026*1.22 (0.77, 1.92)0.399 Model 1 b0.99 (0.64, 1.53)0.9651.62 (1.09, 2.42)0.018*1.33 (0.85, 2.08)0.210 Model 2 c1.01 (0.65, 1.55)0.9811.64 (1.10, 2.44)0.016*1.04 (0.68, 1.59)0.859PM10 Crude model1.64 (0.94, 2.87)0.0841.53 (0.98, 2.38)0.0611.09 (0.65, 1.85)0.746 Model 11.58 (0.90, 2.75)0.1091.55 (0.99, 2.43)0.0561.14 (0.65, 1.97)0.654 Model 21.55 (0.89, 2.71)0.1191.59 (1.01, 2.50)0.045*0.87 (0.52, 1.45)0.591NO2 Crude model1.12 (0.86, 1.46)0.4051.46 (1.10, 1.95)0.010*1.25 (0.92, 1.70)0.150 Model 11.12 (0.84, 1.49)0.4341.46 (1.10, 1.95)0.010*1.24 (0.91, 1.68)0.180 Model 21.14 (0.86, 1.52)0.3681.48 (1.11, 1.98)0.007**1.14 (0.86, 1.52)0.376O3 Crude model1.03 (0.80, 1.32)0.8471.05 (0.88, 1.27)0.5900.80 (0.64, 1.00)0.046* Model 10.99 (0.78, 1.25)0.9151.04 (0.86, 1.26)0.6680.81 (0.65, 1.01)0.062 Model 21.01 (0.79, 1.29)0.9321.05 (0.87, 1.27)0.6150.83 (0.68, 1.01)0.067SO2 Crude model1.14 (0.90, 1.44)0.2681.22 (0.97, 1.53)0.0861.03 (0.83, 1.28)0.797 Model 11.15 (0.90, 1.47)0.2641.23 (0.97, 1.54)0.0831.02 (0.82, 1.27)0.871 Model 21.16 (0.91, 1.49)0.2331.24 (0.98, 1.56)0.0690.97 (0.78, 1.19)0.742CO Crude model1.30 (0.97, 1.74)0.0821.18 (0.90, 1.54)0.2371.14 (0.91, 1.43)0.245 Model 11.27 (0.95, 1.71)0.1131.17 (0.89, 1.53)0.2511.12 (0.89, 1.40)0.341 Model 21.25 (0.93, 1.68)0.1421.16 (0.89, 1.52)0.2741.00 (0.79, 1.26)0.996OR and 95%CI were calculated for per IQR increment for each air pollutant. a Crude Model: adjusted only for mean temperature and relative humidity using natural cubic splines (df = 3); b Model 1: further adjusted for maternal age, pre-pregnancy BMI, infant gender, conception season, education level, smoking status, and alcohol-drinking status based on Crude Model; c Model 2: further adjusted for gestational age, gestational hypertension, and gestational diabetes mellitus based on Model 1OR odds ratio, CI confidence interval* P < 0.05, ** P < 0.01 Associations between exposure to ambient air pollutants and low 1-minute Apgar score in single-pollutant model OR and 95%CI were calculated for per IQR increment for each air pollutant. a Crude Model: adjusted only for mean temperature and relative humidity using natural cubic splines (df = 3); b Model 1: further adjusted for maternal age, pre-pregnancy BMI, infant gender, conception season, education level, smoking status, and alcohol-drinking status based on Crude Model; c Model 2: further adjusted for gestational age, gestational hypertension, and gestational diabetes mellitus based on Model 1 OR odds ratio, CI confidence interval * P < 0.05, ** P < 0.01 Fig. 1Restricted cubic spline (RCS) for the associations between exposure to ambient air pollutants and low 1-minute apgar score.The lines represent odds ratio (OR, solid lines) and 95% confidence interval (CI, long dashed lines). Model adjusted for mean temperature, relative humidity, maternal age, gestational age, pre-pregnancy BMI, infant gender, conception season, education level, smoking status, alcohol-drinking status, gestational hypertension, gestational diabetes mellitus. The reference values (OR = 1) were set at the 5th percentiles, and the knots were set at 5th, 35th, 65th and 95th percentiles of the concentrations of each air pollutant. The histograms represent the distribution of concentrations of each air pollutant Restricted cubic spline (RCS) for the associations between exposure to ambient air pollutants and low 1-minute apgar score.The lines represent odds ratio (OR, solid lines) and 95% confidence interval (CI, long dashed lines). Model adjusted for mean temperature, relative humidity, maternal age, gestational age, pre-pregnancy BMI, infant gender, conception season, education level, smoking status, alcohol-drinking status, gestational hypertension, gestational diabetes mellitus. The reference values (OR = 1) were set at the 5th percentiles, and the knots were set at 5th, 35th, 65th and 95th percentiles of the concentrations of each air pollutant. The histograms represent the distribution of concentrations of each air pollutant Joint associations between co-exposure to air pollutants and low 1-minute Apgar score Figure 2 presents the results of mixed-pollutant models. In the Qgcomp model, each one-quartile increase in the air pollutant mixture during the second trimester was associated with an increased risk of low 1-minute Apgar score (OR = 1.48, 95% CI: 1.06, 2.07). Among the pollutants, PM2.5 and NO2 had the highest weights in this positive association, accounting for 48.0% and 34.5%, respectively. A similar positive association was observed in the WQS model (second trimester: OR = 1.15, 95% CI: 1.09, 1.21). Consistently, PM2.5 and NO2 contributed the highest weights to this positive association, at 23.6% and 39.4%, respectively. However, results for the third trimester were not presented because both the positive and negative weighted quantile sum were statistically significant (positive: 1.28, 95% CI: 1.12, 1.48; negative: 0.83, 95% CI: 0.71, 0.96). The WQS model is designed to assess the effect of mixed exposures under the assumption that all exposure effects are in a consistent direction. When both positive and negative weighted quantile sum are statistically significant, standard WQS regression may fail to accurately assign weights to individual pollutants. Fig. 2Joint association between exposure to air pollutant mixture and low 1-minute Apgar score. Odds ratio (OR) and 95% confidence interval (CI) were calculated for per one-quartile increment for the air pollutant mixture. Model adjusted for mean temperature, relative humidity, maternal age, gestational age, pre-pregnancy BMI, infant gender, conception season, education level, smoking status, alcohol-drinking status, gestational hypertension, gestational diabetes mellitus. Results for the third trimester in WQS model are not presented because both the positive and negative weighted quantile sum were statistically significant (positive: 1.28, 95% CI: 1.12, 1.48; negative: 0.83, 95% CI: 0.71, 0.96) Joint association between exposure to air pollutant mixture and low 1-minute Apgar score. Odds ratio (OR) and 95% confidence interval (CI) were calculated for per one-quartile increment for the air pollutant mixture. Model adjusted for mean temperature, relative humidity, maternal age, gestational age, pre-pregnancy BMI, infant gender, conception season, education level, smoking status, alcohol-drinking status, gestational hypertension, gestational diabetes mellitus. Results for the third trimester in WQS model are not presented because both the positive and negative weighted quantile sum were statistically significant (positive: 1.28, 95% CI: 1.12, 1.48; negative: 0.83, 95% CI: 0.71, 0.96) Associations between ambient air pollutants and 1-minute Apgar score (continuous variable) When the 1-minute Apgar score was treated as a continuous variable, the results indicated that each IQR increase in NO2 exposure during the first trimester and each IQR increase in multiple air pollutants (PM2.5, PM10, NO2, SO2) during the second trimester were significantly associated with decreased 1-minute Apgar score (Table S5 and Table S6). Results from the mixed-pollutant models indicated an association between exposure to the air pollutant mixture in the second trimester and reduced 1-minute Apgar score (Table S7 and Table S8). These findings further confirm the negative associations between exposure to air pollutants and Apgar score. Stratified analysis We selected PM2.5 and NO2 for stratified analysis, as these two air pollutants had the highest weights in the joint associations between co-exposure to air pollutants during the second trimester and low 1-minute Apgar score. We found a borderline interaction between conception season and NO2 only when exploring the association between NO2 exposure and low 1-minute Apgar score (P for Interaction = 0.066). Further subgroup analysis revealed a positive correlation between NO2 exposure and low 1-minute Apgar score among pregnant women who conceived in the warm season (Fig. 3). Fig. 3Stratified analysis of the associations between exposure to ambient air pollutants and low 1-minute Apgar score. The stratification factors include maternal age, pre-pregnancy BMI, infant gender, conception season, education level, status of alcohol drinking, gestational hypertension and gestational diabetes mellitus. The symbols show the Odds Ratio (OR) and the horizontal bars show the 95% confidence interval Stratified analysis of the associations between exposure to ambient air pollutants and low 1-minute Apgar score. The stratification factors include maternal age, pre-pregnancy BMI, infant gender, conception season, education level, status of alcohol drinking, gestational hypertension and gestational diabetes mellitus. The symbols show the Odds Ratio (OR) and the horizontal bars show the 95% confidence interval Sensitivity analysis We adjusted the df from 3 to 6, and no significant changes were observed in the results of the main models (Fig. S5 and Fig. S6). Pairs of air pollutants with a correlation coefficient (r) ≤ 0.7 were included in the two-pollutant model, which generally showed results similar to those of the main models (Fig. S4 and Table S9). However, we found that the results of the two-pollutant model for PM10 during the second trimester were not statistically significant. This might be attributed to the inherently weak positive association between second-trimester PM10 exposure and low Apgar scores (P = 0.045) in the main model, which became statistically non-significant (P > 0.05) after adjustment for another pollutant. After excluding preterm infants, the analysis results in the term infant population also showed consistency with those of the main models (Tables S10-S13). In addition, we redefined low Apgar score according to the commonly used international definition (< 7) and repeated the main analyses, and the results remained stable (Table S14). Discussion In this large-sample birth cohort study, we used single-pollutant models and mixed-pollutant models to assess the independent and joint associations between exposure to air pollutants during each trimester of pregnancy and low Apgar score in newborns. Based on previous literature [25, 26], we selected six air pollutants (PM2.5, PM10, NO2, O3, SO2, and CO) that are routinely monitored in China’s national air quality standards and are common components of multi-pollutant mixtures in environmental epidemiology studies. Results from the single-pollutant models indicated that exposure to air pollutants during the second trimester, including PM2.5, PM10, NO2, and SO2, was associated with a reduction in the 1-minute Apgar score. Second-trimester exposure to PM2.5, PM10, and NO2 showed linear correlations and increasing trends with the risk of low 1-minute Apgar score. Additionally, as a study contributing to the very limited evidence on the joint associations between prenatal air pollutant exposure and Apgar score, our analyses using different mixed-pollutant models consistently revealed a positive association between second-trimester co-exposure to air pollutants and an elevated risk of low 1-minute Apgar score, with PM2.5 and NO2 identified as the major contributing factors in the models. This study suggests that the second trimester may be a sensitive window for the effects of air pollutants on Apgar score in newborns, and PM2.5 and NO2 may be the key influential pollutants. Previous limited epidemiological studies have assessed the associations between air pollutants and Apgar score, with most focusing on ambient particulate matter. A multicenter cohort study in Australia reported a positive association between wildfire-specific PM2.5 exposure during the last four weeks of pregnancy and the risk of low 5-minute Apgar score [18]. A study in the Netherlands focusing on ultrafine particles did not observe an association with low 5-minute Apgar score [48]. These differing results may reflect variations in particle size, exposure levels, population characteristics, and the timing of resuscitation interventions, which can influence the 5-minute Apgar score. The 1-minute Apgar score is an indicator less affected by clinical interventions, so it is potentially more reflective of direct prenatal effects [49, 50]. A retrospective cohort study in Guangzhou, China, involving 5,396 mother-infant pairs found that exposure to soil dust components of PM2.5 during the second and third trimesters was associated with an increased risk of low 1-minute Apgar score [19]. This provided important evidence linking particulate matter exposure during pregnancy to low 1-minute Apgar score. Another small-sample study focusing on pregnant women with preterm birth and specific genetic variants reported a significant association between nitrogen oxides (NOx) exposure and a decreased 1-minute Apgar score. This study provided preliminary evidence for the association in a specific population, but its generalizability to the general pregnant population requires further validation [51]. To build on previous findings, we conducted a large prospective cohort in a general pregnant population. Consistent with the above studies, our study also observed a positive association between second-trimester exposure to ambient particulate matter (PM2.5, PM10) and the risk of low 1-minute Apgar score, and additionally identified a positive association for second-trimester NO2 exposure. However, given differences in study design, population characteristics, pollutant types, and exposure concentrations, the results of these studies may not be directly comparable, and further research is needed to integrate these findings. Ambient air pollutants consist of multiple components, and people are typically exposed to complex mixtures of these pollutants. However, previous studies have mostly evaluated the health effects of air pollutants using single-pollutant models. Due to the strong correlations among different air pollutants, this approach may lead to residual confounding among pollutant-specific effects, thereby hindering accurate estimation of the independent effect of any single pollutant [52]. In recent years, mixed-pollutant models have been applied to evaluate the overall health effects of air pollution [53–55]. In this study, Qgcomp and WQS models were used to investigate the joint associations between exposure to air pollutants and low Apgar score. The results consistently showed that second-trimester co-exposure to air pollutants was associated with an elevated risk of low 1-minute Apgar score. Interestingly, we observed that PM10 was associated with an elevated risk of low 1-minute Apgar score in the single-pollutant model, but this association was not significant in mixed-pollutant models. This may be because PM10 is highly correlated with other air pollutants, suggesting that in the single-pollutant model, the association between PM10 and low 1-minute Apgar score might have been confounded by other pollutants. This study’s findings emphasize the advantages of the mixed-pollutant model in assessing the health effects of air pollution. This model not only accounts for the mutual interactions among pollutants, but also enables the evaluation of the contribution of each pollutant to the effect. As observed in our study, PM2.5 and NO2 are the important contributing components to low 1-minute Apgar score associated with air pollutant exposure during the second trimester. The potential biological mechanisms underlying the effects of ambient air pollutant exposure on neonatal health outcomes remain unclear. Several studies have suggested that ambient PM2.5 may cross the placental barrier and exert neurodevelopmental toxicity, potentially interfering with neuronal proliferation, migration, and synaptogenesis in the developing fetal brain, leading to abnormal brain development and thereby increasing the risk of neonatal asphyxia and low Apgar score [56–58]. Meanwhile, as major intrauterine toxicants, air pollutants can induce excessive production of reactive oxygen species (ROS), disrupting the balance between oxidative and antioxidant systems and triggering oxidative stress responses. This oxidative stress may not only directly damage cellular components such as lipids, proteins, and DNA, but also promote the expression of inflammatory factors, leading to a systemic inflammatory response that could impair the development of multiple fetal organ systems [59–61]. Additionally, recent evidence has indicated that air pollutant exposure may compromise the integrity of the placental barrier, resulting in elevated fetal cortisol levels and fetal growth restriction, ultimately leading to various adverse neonatal outcomes [62]. Notably, our findings suggest that the second trimester may be a sensitive window for air pollutant exposure. We attempt to explain this by integrating the above mechanisms with the physiological characteristics of the mother and fetus during the second trimester. First, the second trimester is a critical phase for fetal organogenesis and rapid brain growth, including progressive neuronal proliferation, migration, and synaptogenesis [63]. Highly proliferative and metabolically active cells are generally more susceptible to oxidative and inflammatory insults, which may amplify the neurodevelopmental toxicity of air pollutants. Second, by the second trimester, the placenta becomes fully functional and well-perfused [64], potentially facilitating the transfer of inhaled pollutants or their bioactive metabolites into the fetal circulation. In addition, maternal physiological changes during the second trimester, such as increased cardiac output, increased blood volume, and elevated metabolic demands [65], may enhance the absorption and tissue distribution of air pollutants, leading to higher internal exposure doses and ultimately affecting neonatal health. However, these mechanistic hypotheses require further investigation. In the stratified analysis, we observed a borderline interaction between second-trimester NO2 exposure and conception season, although the interaction did not reach statistical significance (P for interaction = 0.066). Future studies with larger sample sizes and a wider range of exposure concentrations are needed to confirm this finding. This study has several strengths, including a birth cohort design, large sample size, and consideration of more potential covariates. However, several limitations should be acknowledged. First, it is a single-center study, which may limit the generalizability of our findings. To mitigate the limitations of a single-center study, we recruited the participants from a well-known tertiary A-Level maternity and child healthcare hospital in China, which provides comprehensive and high-quality medical and healthcare services for pregnant women. Its annual delivery volume has consistently ranked among the highest in Guangdong Province in recent years. Second, the number of positive cases was limited, and the study did not include more participants with abnormal Apgar score, which may have resulted in the failure to detect certain associations, although a series of sensitivity analyses were performed to verify the stability of the results. Third, we did not collect data on the medication use of participants during pregnancy. Medication use may be an important potential confounder, as previous studies have reported that the use of psychotropic medications such as alprazolam during pregnancy is associated with an increased risk of low Apgar score [66]. The absence of this information may have led to residual confounding and could thus have biased our effect estimates. Fourth, air pollution exposure of participants was assessed based on their residential addresses, without considering factors such as individual mobility, commuting, occupational exposure, and indoor pollution, which may lead to exposure misclassification bias. This limitation is of particular concern during pregnancy, as pregnant women may experience dynamic changes in mobility patterns, reduced outdoor activity time, and alterations in occupational exposure throughout gestation, all of which can affect their actual personal exposure levels. Conclusion This study found that exposure to ambient air pollutants during the second trimester may be associated with an increased risk of low 1-minute Apgar score. In the mixed-pollutant models, the air pollutant mixture was associated with an increased risk of low Apgar score, with PM2.5 and NO2 possibly being the major contributors in the models. Given the growing societal concern regarding the impact of air pollution on maternal and infant health, future multicenter studies with larger sample sizes, a wider range of exposure concentrations, and more refined covariates are needed to clarify the associations between maternal air pollutant exposure during pregnancy and fetal development and neonatal health. Our findings support the need for further research and increased public health attention to air pollution exposure during pregnancy, which may help inform the development of intervention strategies to reduce air pollution-related adverse neonatal health outcomes and their long-term risks. Supplementary Information Supplementary Material 1. Supplementary Material 1.