Introduction

Childhood growth trajectory is an important content of life course epidemiological studies and a potential factor determining the occurrence of chronic diseases in adulthood1,2,3,4. It can reflect children’s nutritional status, social and economic level, family feeding behavior, disease and social environment, and other factors, and can more effectively evaluate the process of children’s individual development3,4. An inappropriate growth trajectory in early childhood is associated with the risk of chronic non-communicable diseases, such as hypertension, diabetes, cardiovascular disease, etc., in adolescence and later adulthood1,2,5. Moreover, the growth trajectory can intuitively reflect the change in children’s growth rate and the deviation of children’s growth in time, which can help clinicians and child health personnel to intervene in children as early as possible6. Although studies on children’s growth trajectories have received wide attention, there are still some challenges, such as the difficulty in obtaining repeated measurements of physical growth indicators such as height (length), weight, and head circumference6,7. Therefore, it is particularly important to carry out researches on children’s growth trajectory to promote the health of children and even later.

The existing researches primarily examine the relationships between children’s growth trajectory and maternal pregnancy complications, sociodemographic characteristics, food intake and other factors8,9,10, while the literature regarding the impact of environmental pollutants exposure on child’s growth trajectory is relatively limited, especially the impact of maternal environmental pollutant exposure during pregnancy on the growth trajectory of offspring11,12,13. There is evidence that air pollutants may increase the risk of intrauterine growth and birth defects in newborns7,14, but only a few studies investigated the longitudinal impacts of prenatal exposure to ambient air pollutants on childhood growth, and most evidence comes from studies using a single growth measurement.

Limited researches have determined the associations of prenatal air pollutants exposure with early childhood growth trajectories, and the results are inconsistent, especially the associations between prenatal fine particulate matter (PM2.5) exposure and childhood growth trajectories11,12,15,16. One prospective cohort study performed in Spain found that maternal PM2.5 exposure during pregnancy reduced the risk of children’s rapid growth compared with those with normal growth trajectories11. Another study in Boston demonstrated that prenatal exposure to PM2.5 during pregnancy was related to childhood weight growth trajectories and found significant sex-specific differences16. Moore’s study in Colorado showed that high PM2.5 exposure in the third trimester was associated with the rapid rate of BMI growth15. However, the study in Project Viva had showed that PM2.5 exposure during pregnancy did not appear to affect childhood growth trajectory12. In China, only one study has focused on the associations between prenatal air pollutants exposure and childhood growth trajectories and found an increased risk of deviation of childhood growth associated with air pollutants exposure during pregnancy, but it failed to investigate the relationships between PM2.5 exposure and childhood growth trajectories13. It can be seen from the above that there is limited evidence on the relationships of prenatal PM2.5 exposure with childhood growth trajectory, and results are inconsistent, and more studies are needed. In addition, there is no report on the associations between prenatal PM2.5 exposure during pregnancy and childhood growth trajectories from birth to 6 years old in China. Challenges in conducting this study include both the identification of childhood growth trajectories and the difficulty of collecting repeated measurements of children’s growth throughout early childhood.

In this study, we aimed to assess whether maternal PM2.5 exposure during the specific trimesters and the entire pregnancy was related to children’s growth trajectories from birth to 6 years old using a birth cohort study design. We further performed stratified analyses by children’s gender, maternal age, residential area, pre-pregnancy BMI, gravidity, and parity to explore their potential modifying effects in relation to the growth trajectory of children and maternal PM2.5 exposure.

Materials and methods

Study population

The study population was identified from a population-based prospective birth cohort of mothers and their children born between January 1, 2011, and December 31, 2013, and described in detail previously13,17. The population-based birth cohort was performed based on the Wuhan Maternal and Child Health Management Information System (WMCHMIS), which was a governmental registration information system for women and child healthcare. All information of regular examinations of mothers from their first clinic visit after pregnancy until delivery and the growth measurements of children from birth to 6 years old were required to be filled in this system. As described previously13, the population was restricted to all singleton births without a congenital defect and those mothers who had resided in Wuhan for at least one year before delivery, and those children born prematurely (less than 37 weeks gestation), with low birth weights (less than 2500 g at birth), with serious illnesses at birth, who had less than five physical measurements, or whose weight or height values were implausible were excluded. This process yielded 62,540 pairs of mothers and children for analysis.

Approval to conduct the study was obtained from the Institutional Review Board of Wuhan Children’s Hospital (No. 2019R019-E03). This study was performed in accordance with Declaration of Helsinki, all personal information was anonymous and no identifiable information at the individual level was in the data set, therefore, the need for informed consent was waived by the Institutional Review Board of Wuhan Children’s Hospital, Tongji Medical College, Huazhong University of Science & Technology.

Growth trajectories

The outcomes of the study were growth trajectories of children’s BMI. The BMI was calculated as weight in kilograms divided by height (length) in meters squared (kg/m2). Body weight, recumbent length (generally before 36 months), and standing height (generally after 36 months) were collected from the WMCHMIS. These measurements were usually performed and recorded by a pediatrician or pediatric nurse at a primary care center to monitor the child’s growth at 1, 3, 6 and 12 months, every 6 months until 3 years, and once a year from 3 to 6 years according to the National Basic Public Health Service protocol.

Exposure and assessment

Air pollutants data were collected from Wuhan Environmental Monitoring Center during the study period. As described previously, all automatic monitoring stations continuously collected the concentrations of atmospheric pollutants, and were set up according to the national environmental air quality monitoring standards, and are well distributed in various administrative regions of Wuhan17,18. Daily mean concentrations of PM2.5, particulate matter of less than 10 µm in diameter (PM10), sulfur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO), and ozone (O3) were obtained from each station. All initial monitoring concentrations of air pollutants were audited and processed by staffs from the Wuhan Environmental Monitoring Center. Since the Ministry of Environmental Protection of China required PM2.5 to be monitored and published as an air quality indicator until the end of 2012, we only obtained daily concentrations of PM2.5 in this study from the monitoring stations that had conducted pilot monitoring of PM2.5 concentrations. Ambient air pollutants exposure for individuals were estimated using the inverse distance weighted (IDW) method, which inversely weighted air pollutants concentrations based on the distance of the subject’s residence from nearby monitoring stations19. We geocoded each pregnant woman’s residential address according to the corresponding residential parcel at the first physical examination and the addresses of monitoring stations, then calculated the distances from the residential addresses to each monitoring station, and calculated the weight coefficients to different monitoring stations according to the distances from the residential addresses to each monitoring stations, and inversely weighted the daily average concentrations of each monitoring station according to the weight coefficient. Those pregnant women who did not live within 10 km of a monitoring station were excluded from the study, making a final analysis sample of 47,625 pairs of mothers and children.

To explore the effects of prenatal exposure to air pollutants on childhood growth trajectories at each of the different trimesters of pregnancy, the average concentrations of PM2.5 and other ambient air pollutants in every trimester as well as the entire pregnancy were calculated. Referring to the classifications in previous similar studies18,20, the trimester 1 was defined by the first day of gestation until 13+6 weeks, and the trimester 2 was the period from 14 weeks of gestation to 27+6 weeks of gestation, the trimester 3 was from 28 weeks of gestation to the day of delivery, and the entire pregnancy was from the first day of gestation to the day of delivery.

Other covariates

Individual covariates of known or hypothesized factors that might be associated with child growth were collected from the WMCHMIS, including maternal age, maternal educational status, residential areas, pre-pregnancy BMI, pregnancy-induced hypertension syndrome, gestational diabetes, gravidity, parity, birth type, children’s gender, and birth weight.

Statistical analysis

The Group-based trajectory model (GBTM) was applied to categorize participants into growth trajectories with statistically distinct changes over time from birth to 6 years, using PROC TRAJ in SAS version 9.2. As previously elaborated10,13,21, the GBTM model assumes heterogeneity in the population, i.e., there are several potential subgroups with different trajectories or patterns in the population, each subgroup has a specific intercept and slope, while the model can categorize individuals into subgroups with different intercepts and slopes based on their growth measurements over time, thereby better reflecting the heterogeneity of the data. The model by categorizing the variability of children’s growth could provide a better understanding of how child growth changes over time and provide a way to understand the causal factors behind different growth trajectories. To determine the optimal number of categories for children’s BMI trajectories, we made multiple attempts and determined the best-fitting model based on the following factors: (a) the model with the highest Bayesian information criterion value, with a minimum of 3-point change in absolute Bayesian information criterion between the models required to accept the improved model; (b) the shape of the most minimal trajectory with statistically significant polynomial coefficients; (c) whether additional trajectories revealed important features in the data; and (d) interpretability10,13.

Categorical sociodemographic characteristics of mothers and children were presented as frequencies and percentages, and the concentrations of air pollutants were summarized as means and standard deviations (SD). The Chi-square tests were conducted to compare the difference in sociodemographic characteristics of participants among different growth trajectory groups. We estimated the associations of prenatal exposure to PM2.5 in different stages of pregnancy with child’s growth trajectories by both single-pollutant and two-pollutant models with multivariate multinomial logistic regression, with controlling for maternal age, educational status, pre-pregnancy BMI, residential area, gravidity, parity, birth type, and children’s gender. The associations were presented with Odds Ratios (ORs) and 95% confidence intervals (95% CI) for per 10 μg/m3 increase in the concentrations of PM2.5.

To investigate the potential effect modifications of specific groups on the relationships between prenatal exposure to PM2.5 and child’s growth trajectories, we conducted stratified analyses to explore the potential effects of children’s gender (male or female), maternal age (≤ 35 years or > 35 years), residential area (urban or rural), pre-pregnancy BMI (< 24.9 kg/m2 or ≥ 25 kg/m2), gravidity (first pregnancy or not) and parity (first delivery or not), based on prior studies suggesting the possibility of these factors concerning prenatal air pollutants exposure or childhood growth10,13. Considering the air pollutants concentrations varying in different birth seasons and years, the sensitivity analyses were performed to determine the potential effect modifications of PM2.5 exposure and child’s growth trajectories. All data analyses were carried out using SAS version 9.2 (SAS Statistical Institute, Inc. Cary, NC).

Results

Population characteristics

This cohort population covered 47,625 mother–child pairs with PM2.5 exposure data available and these children were classified into three trajectory groups (Fig. 1): slow growth (n = 13,671, 28.7%), normal growth (n = 29,736, 62.4%), rapid growth (n = 4218, 8.9%). Table 1 summarizes the characteristics of the study population. In total, 87.8% of mothers’ maternal age was 22–35 years old, more than half (55.2%) had a college or above degree and 79.0% were living in urban areas. 36.8% of mothers had a previous pregnancy and 18.2% had a history of delivery. Compared to the children with slow and normal growth trajectories, children with rapid growth trajectory were more inclined to be born from mothers with pre-pregnancy BMI ≥ 25 kg/m2. The main birth type of children in our study was cesarean Sect. (70.9%), and the proportion of cesarean section of mothers in children with rapid growth trajectory was significantly highest among three growth trajectory groups (P < 0.05). A higher proportion of females was found in the rapid growth group than in other growth groups (P < 0.05). Children with rapid growth trajectory had higher birth weight than children with other growth trajectories (P < 0.05). There was no statistically significant difference in gestational diabetes and pregnancy-induced hypertension syndrome among mothers of the three growth trajectory groups (P > 0.05).

Figure 1
figure 1

Body mass index (BMI) trajectory groups. The blue line represent the rapid growth trajectory group, the orange line represent the normal growth trajectory group and the yellow line represent the slow growth trajectory group.

Full size image
Table 1 Characteristics of the study population.
Full size table

Air pollutants concentrations

Table 2 summarizes air pollutants concentrations throughout the entire pregnancy for all participants. The average concentrations and their ranges (min–max) were as follows: PM2.5—83.3 μg/m3 (32.8–179.0 μg/m3), PM10—111.5 μg/m3 (42.7–257.1 μg/m3), SO2—34.0 μg/m3 (5.7–72.3 μg/m3), NO2—57.6 μg/m3 (17.3–96.5 μg/m3), CO—1122.1 μg/m3 (498.0–2873.8 μg/m3), and O3—78.1 μg/m3 (19.9–146.7 μg/m3). Table S1–S2 displays air pollutants concentrations during different birth seasons and years, which varied significantly among different birth seasons and years. The correlations between PM2.5 and other pollutants were generally positive, except for O3. The correlation coefficient between PM2.5 and PM10 was 0.513, PM2.5 and SO2 was 0.268, PM2.5 and NO2 was 0.379, PM2.5 and CO was 0.420, and PM2.5 and O3 was − 0.223.

Table 2 Summary of air pollutants concentrations during pregnancy of all participants.
Full size table

Associations between PM2.5 exposure and child’s growth trajectories

Table 3 presents the ORs and 95% CIs for the child’s growth trajectories with per 10 μg/m3 increase in PM2.5 concentration during the specific trimesters and the entire pregnancy in single-pollutant models. Increased prenatal exposure to PM2.5 was related to increased odds of slow growth trajectory, with an adjusted OR of 1.015 (95% CI 1.009–1.021) for trimester 1, 1.040 (95% CI 1.027–1.052) for trimester 2, and 1.016 (95% CI 1.010–1.023) for the entire pregnancy, whereas increased prenatal exposure to PM2.5 was relevant to reduced odds of rapid growth trajectory, with an adjusted OR of 0.975 (95% CI 0.966–0.985) for trimester1, 0.921 (95% CI 0.903–0.939) for trimester 2, 0.980 (95% CI 0.970–0.991) for trimester 3, and 0.982 (95% CI 0.972–0.992) for the entire pregnancy. Similar associations were also observed in the two-pollutant models (Table 4). For instance, prenatal PM2.5 exposure during the entire pregnancy remained positively related to slow growth trajectory after adjusting for SO2, and O3, respectively, and remained negatively associated with rapid growth trajectory after adjusting for SO2, NO2, and O3.

Table 3 The crude and adjusted ORs and 95% CIs for child’s growth trajectories with each 10 μg/m3 increase in PM2.5 concentration during the specific trimesters and the entire pregnancy among study population.
Full size table
Table 4 The adjusted ORs and 95% CIs for child’s growth trajectories with each 10 μg/m3 increase in PM2.5 concentration during the specific trimesters and the entire pregnancy in two-pollutant models.
Full size table

Subgroup analysis

Table 5 exhibites the effect modifications of stratified analyses by children’s gender, maternal age, residential area, pre-pregnancy BMI, gravidity and parity, and the relationships of prenatal exposure to PM2.5 with child’s growth trajectories varied by maternal age, residential area, and pre-pregnancy BMI. Maternal age-stratified analyses showed that the estimated effect of prenatal exposure to PM2.5 on the child’s growth trajectories was higher among mothers over 35 years old than those who were under 35 years old. Children whose mothers live in urban areas were at greater risk of the slow growth trajectory, and children whose mothers living in rural areas had a relatively lower risks of slow growth trajectory. Compared with mothers with a maternal pre-pregnancy BMI < 24.9 kg/m2, prenatal exposure to PM2.5 had a greater effect on slow growth trajectory in children with maternal pre-pregnancy BMI ≥ 25 kg/m2. Moreover, we found that stratified analyses by children’s gender, gravidity, and parity did not significantly alter effect estimates for the relationships between prenatal exposure to PM2.5 and child’s growth trajectories. Table S3 presents that the correlations of prenatal PM2.5 exposure with child’s growth trajectories were not materially varied by birth seasons and birth years.

Table 5 Children’s gender, maternal age, residential area, pre-pregnancy BMI, gravidity and parity specific ORs (95%CIs) for child’s growth trajectories associated with each 10 μg/m3 increase in PM2.5 in the study population.
Full size table

Discussion

This prospective birth cohort study highlights the long-lasting adverse impact of prenatal PM2.5 exposure on child growth trajectory from birth to age 6. In comparison to children with normal growth trajectory, prenatal PM2.5 exposure could increase the risk of the slow growth trajectory, while reducing the risk of the rapid growth trajectory. Maternal age, pre-pregnancy BMI, and previous delivery experience might modify these associations. To our knowledge, this study is one of the few studies to determine the long-term effects of prenatal exposure to PM2.5 on dynamic child growth using repeated measurements of weight and height in the first 6 years of life. The findings provide new evidence that maternal exposure to PM2.5 during pregnancy has long-term negative impact on growth trajectories in early childhood.

The literatures regarding the relationships of prenatal exposure to PM2.5 during pregnancy and child growth trajectories are sparse and inconsistent11,12,22,23. In general agreement with our results, Fossati et al. investigated the associations of PM2.5 and NO2 exposure in early pregnancy with the growth trajectory of children aged 0–4 years including 1724 pairs of mother and child and showed that maternal exposure to high level of PM2.5 in early pregnancy was related to reduce the risk of rapid growth trajectories in children compared with children with normal growth trajectories, but did not observe the significant relation between exposure to PM2.5 and slow growth in children11. Inconsistent with the above findings, a small birth cohort study in Boston (n = 239) found a positive association between exposure to PM2.5 in early pregnancy and increased BMI in boys22. A longitudinal study in Catalonia showed exposure to PM2.5 was correlated to increased BMI growth23. In the Healthy Start study, Moore and colleagues found that high PM2.5 exposure in the third trimester was related to rapid BMI growth from birth to 3 years of age15. Moreover, Fleisch and colleagues followed 1649 children from birth to mid-childhood and did not find a persistent effect of prenatal PM2.5 exposure on BMI growth trajectories12. From the above studies, the correlation of PM2.5 exposure during pregnancy with early childhood growth trajectory did not reach a consensus. Differences in exposure assessment methods and study populations may explain the partial differences among different studies. For example, we used the inverse distance weighted method to assess maternal PM2.5 exposure concentrations, whereas Fleisch et al.12 and Chiu et al.22 used the satellite spatiotemporal models to assess the mothers’ exposure concentrations of air pollutants according to their residential addresses, which may have bias in the accuracy of exposure assessments. Further studies are warranted to identify the exact associations of prenatal exposure to PM2.5 with longitudinal childhood growth.

The potential effect modifications of children’s gender, maternal age, residential area, pre-pregnancy BMI, gravidity, and parity on the relationships of prenatal PM2.5 exposure with child’s growth trajectories were examined. We found there were significant differences in maternal age, pre-pregnancy BMI, and parity on the observed associations of prenatal PM2.5 exposure on child's growth trajectories, suggesting that these demographic characteristics of mothers can alter children’s susceptibility to exposure to air pollutants. This finding had important public health implications, especially considering the special air pollution protection afforded to these pregnant women during pregnancy. Though the underlying cause of these findings remains unclear13,20, mothers with older, overweight, or obese may be more susceptible to air pollution due to changes in their physical functions, and those who had previous deliveries may neglect to protect themselves and are exposed to more outdoor air pollution20. In addition, we did not find there were significant differences in children’s gender on the observed associations, but Rosofsky’s study16 and our previous study showed higher estimated risk of prenatal PM2.5 exposure on child's growth trajectories in girls, Chiu’s study22 founded that increasing prenatal exposure to particulate pollutants increased boys’ BMI z-scores but not girls’. In conclusion, the potential effect modifications of gender on the observed associations need to be determined in future studies.

The biological mechanisms of prenatal PM2.5 exposure with early childhood growth in offspring remain unsolved, but several hypotheses have been proposed, including inflammation responses and oxidative stress24,25,26,27. Several studies suggested that maternal exposure to particulate matter may result in maternal, placental, or fetal inflammation, which is linked to impaired placenta function and decreased weight later in life28,29, our finding also confirmed this hypothesis that maternal PM2.5 exposure during pregnancy could undermine children’s growth. Several animal studies have also confirmed the inflammatory effects of PM26,30, suggesting that low-grade systemic maternal inflammation may cause disruptions in regulation of maternal appetites and metabolism, which could have adversely affect the growth of offspring31. Moreover, PM2.5-induced neuroinflammation can disrupt maternal satiety signals, making offspring more prone to growth disorders24. Additionally, PM2.5 exposure may regulate the expression of white and brown adipose genes, which were responsible for energy storage and dissipation, and change the metabolic characteristics of fetal adipose tissue, suggesting a programming effect that may result in the risk for growth later in life32,33. Finally, there is evidence that PM2.5 components have the ability to induce endothelial dysfunction34, and that they are much more strongly associated with oxidation-inflammation responses than other particulate pollutants35,36, which may lead to more severe abnormal inflammation -related growth. Although the above studies have initially explored the potential mechanism of prenatal PM2.5 exposure on offspring growth, the exact mechanism needs to be clarified in the future.

Our study has several notable strengths. First, it was a large-sample study that included 47,625 children with multiple repeated measurements of height and weight from birth to age 6, which permitted us to determine longitudinal differences in childhood growth trajectories by levels of prenatal PM2.5 exposure. Second, we focused on growth trajectories in early childhood, rather than BMI measurements obtained at a certain time point, which could allow us to understand changes in velocity and slope during specific growth time periods, and provide an opportunity to identify the susceptible phenotypes and perform potential interventions. Third, all physical measurements were obtained from the government registration system, which required the pediatricians to upload measurements of children in time. These data are likely more reliable than self-reported growth measurements and limit the potential recall bias. Lastly, our findings also provide new evidence for the long-term adverse impact of prenatal PM2.5 exposure on early childhood growth. Pregnant women have been considered more vulnerable to the adverse effects of air pollution, so effective policies and stricter enforcement for air quality control are needed to improve maternal and child health.

Several limitations should also be noted in this study. First, we did not take children’s nutritional intake into account due to the unavailability of this variable in the registration system, and we also lacked information on other important factors, physical activity, disease, and living environment, etc., which have been considered as important predictors of childhood growth4,37. Second, we did not consider the impact of postnatal environmental pollutants exposure, which may be jointly associated with to prenatal PM2.5 exposure and children’s growth, so there may be residual confounding in our analysis. Therefore, this critical confounding should be fully considered in future studies on the associations of prenatal exposure to PM2.5 during pregnancy with childhood growth trajectories. Third, exposure misclassification may be unavoidable, since we only assessed ambient PM2.5 exposure to the mother and did not assess the pollution of the mother’s living or working environment where they might spend more time, this may lead to exposure assessment errors and ultimately weaken the true relationships between maternal PM2.5 exposure during pregnancy and the outcome of this study. Forth, although we found that the associations of per 10 μg/m3 increase in PM2.5 exposure on child growth trajectories were significant, but the effect value was really small, the true associations needs to be confirmed by more studies. Further studies are warranted to overcome the shortcomings of this study, to clarify the long-term impact of PM2.5 exposure during pregnancy on child growth, and to better understand the biological mechanisms of their links and interactions.

Conclusions

This cohort study demonstrated that prenatal PM2.5 exposure could raise the risk of the slow growth trajectory but reduce the risk of the rapid growth trajectory, and suggested that prenatal exposure to PM2.5 during pregnancy has a negative long-term effect on child growth from birth to age of 6. We also revealed that maternal age, pre-pregnancy BMI, and previous delivery experience might modify these associations. Additional studies across a variety of different study populations and geographic regions are needed to validate our findings, and it would also be meaningful to consider follow-up into adolescence and adulthood, and to explore the impact of specific growth trajectory phenotypes in childhood on the chronic disease in adults in future studies.