Introduction

Congenital limb defects (CLDs) are a group of congenital abnormalities characterized by significant hypoplasia or aplasia of one or multiple bones of the extremities, primarily including polydactyly, syndactyly, limb shortening and clubfoot1. As one of the most prevalent congenital limb malformations, CLDs rank as the second most common category of birth defects2. The prevalence and phenotypic manifestations of these abnormalities vary significantly across populations and geographic regions2,3, and the incidence of CLDs among the Chinese population is approximately 3.9 per 1000 births4. These congenital abnormalities not only severely impair children’s motor function, physical and mental development and quality of life but also impose a substantial economic burden on families and society2,3. However, the etiology of CLDs remains highly complex and largely unknown. Studies have shown that several limb defects are associated with genetic mutations or chromosomal abnormalities5,6, but approximately 50% of CLDs lack identifiable causes, suggesting that environmental exposures may play potential critical roles7,8. Animal studies have demonstrated that maternal exposure to teratogens during pregnancy can disrupt limb development and lead to abnormal skeletal differentiation6,9. Nevertheless, the evidence for associations between environmental factors and limb defects in population studies remains controversial, and further exploration is urgently needed to elucidate the underlying mechanisms.

Ambient air pollutants have been identified as risk factors for many adverse pregnancy outcomes10,11,12. In recent years, numerous epidemiological studies have explored the relationships between maternal air pollutant exposure during pregnancy and congenital malformations in offspring but have focused mainly on cardiac defects and oral and facial defects11,13,14,15. Very few studies have been conducted on air pollutant exposure and CLDs, and the conclusions are consistent16,17,18. A multicenter case-control study performed in the United States revealed that maternal CO, SO2 and NO2 exposure during pregnancy was associated with CLDs, whereas no association was found between PM2.5, PM10 and O3 and limb defects16. A case-control study conducted in Changsha, China, which was based on a medical record system, revealed that maternal NO2 and SO2 exposure in the first trimester was significantly linked to overall CLDs and their subtypes in offspring19. Another case-control study conducted in Liaoning revealed that maternal exposure to PM10 and SO2 during both 3 months before conception and the first trimester was associated with an increased risk of polydactyly and syndactyly18,19. A matched case-control study conducted in Taiwan revealed that each 1 ppb increase in SO2 concentration at 9–12 weeks of gestation and during the first trimester increased the risk of limb shortening, and each 1 ppb increase in O3 concentration at 1–4 weeks of gestation increased the risk of limb shortening, while no association between PM10, NO2 and CO exposure during pregnancy and limb defects was found20. Although few studies have explored the associations between air pollutant exposure and limb defects, such as polydactyly, syndactyly, and limb shortening, the available evidence is limited and inconsistent.

Although few studies have investigated the associations between maternal exposure to air pollutants and the risk of CLDs in offspring, there are still some shortcomings that need to be improved. First, most studies are case-control designs with relatively small sample sizes, particularly those lacking large-scale population-based cohort studies16,17. Second, most studies have focused mainly on the effects of single pollutants, neglecting the joint effects of pollutant mixtures. Air pollutant exposure is a combination of multiple pollutants, and a single-pollutant model might not fully reflect the impact of air pollutant exposure on the risk of CLDs19. Third, current studies have focused mainly on maternal exposure to air pollution during 2–8 weeks of pregnancy16,20, which might inadequately capture all critical exposure periods. For example, exposure to pollutants during the preconception or early pregnancy period might exert long-term effects on fetal development21,22. Fourth, there are many subtypes of CLDs, including isolated and nonisolated types and different subcategories, and more detailed investigations targeting different types of CLDs are warranted16. Fifth, the modifying effects of sociodemographic factors, such as maternal age, socioeconomic status, education level, and the residential environment, have not been adequately considered; these factors may influence air pollution exposure and the occurrence of birth defects, thereby interfering with the associations of air pollutant exposure with the risk of CLDs23. In summary, more comprehensive studies on the relationships between maternal air pollutant exposure and CLDs in offspring are urgently needed to address these gaps by expanding sample sizes, considering heterogeneity across regions and susceptible populations, elucidating multipollutant interactions, stratifying CLD subtypes, and extending exposure windows.

Therefore, we conducted this large prospective population-based cohort study to investigate whether there are dose-response relationships between maternal air pollutant exposure in early pregnancy and the risk of CLDs, as well as their subtypes, such as polydactyly, syndactyly, limb shortening, and clubfoot, and to examine whether similar associations exist for exposure during preconception. We further analyzed whether individual sociodemographic characteristics affect these associations. These findings could help to further clarify the relationships between maternal air pollutant exposure and the risk of CLDs, and provide a high-quality scientific basis for formulating environmental intervention strategies to prevent birth defects.

Materials and methods

Study population

Participants were recruited from a birth cohort established through the Wuhan Maternal and Child Health Management Information System (WMCHMIS). A detailed description of the registry system for data collection, case identification, and quality control has been described previously24,25. In brief, all pregnant women who received their first prenatal care visit were required to register in this government system to collect and perform quality control on health surveillance data concerning mothers and children, and all subsequent examinations during pregnancy and birth information were recorded by trained hospital staff. The maternal and birth information included maternal sociodemographic characteristics, past medical history, gestational history, antenatal care visits, pregnancy complications, delivery information, birth defects, postnatal check-ups for mothers and infants, etc. The system has established a strict three-level quality control system composed of municipal, county and community data to ensure the accuracy and integrity of the data. In this study, we recruited all pregnant women and births, including live births, stillbirths, or terminations of pregnancy due to prenatal diagnosis of congenital anomalies, from January 1, 2011, to September 30, 2017. A total of 588,581 births from mothers who lived permanently in Wuhan during the study period were included, and mothers with missing residential addresses, gestational ages less than 28 weeks, birth defects other than CLDs and other key perinatal information missing were excluded. The final analysis included 510,550 mother-infant pairs, of which 1,864 births were diagnosed with CLDs. All personal information about mothers and infants in this study was anonymous and did not involve personally identifiable information. Approval for the study protocol was granted by the Institutional Review Board of Wuhan Children’s Hospital (NO. 2024R061E01).

Congenital limb defects (CLDs)

Data on CLDs were obtained from the governmental birth defect surveillance system, which was a submodule system of the WMCHMIS and was established according to the requirements of the China Maternal and Child Health Monitoring Manual and National Hospital Birth Defects Surveillance of China. All birth defect cases from 28 weeks of gestation and 7 days after birth were required to report as the Birth Defects Registration Card to the monitoring system, and all birth defects were diagnosed by trained obstetricians or pediatricians based on clinical observations, physical examinations, and image examination results according to the 10th Revision of the International Classification of Diseases (ICD-10). All CLD cases in this study were identified and categorized into the following four subtypes based on the ICD-10 code: polydactyly (Q69), syndactyly (Q70), limb shortening (Q71, Q72) and clubfoot (Q66.0).

Exposure assessment

The daily concentrations of ambient air pollutants (PM2.5, PM10, SO2, NO2, CO, and O3) used in this study were obtained from the Wuhan Environmental Monitoring Center. The details of data collection were described in our previous studies24; Tan et al.25, and the average 24-hour daily concentrations of these pollutants from 2010 to 2017, which covered the period of maternal pregnancy in this study, were acquired from twenty-one government automatic air quality monitoring stations. Individual exposure concentrations were calculated via the inverse distance weighted method on the basis of residential address during pregnancy22. The inverse distance weighted method is based on the principle that the closer a residential address is to a monitoring station, the higher weight assigned to it. We calculated the longitudes and latitudes of each pregnant woman’s residence registered at the first physical examination during pregnancy, and then calculate the distances from each residence to monitoring stations. Based on these distances, inverse-distance weight coefficients for the stations relative to each residence were assigned, and the daily average air pollutants concentrations at each station were inversely weighted according to these coefficients to estimate individual exposure concentrations. To ensure the accuracy of the exposure assessment, the maximum distance between an individual residential address and a monitoring station was limited to 20 km. The spatial distributions of the study participants and air quality monitoring stations are presented in Fig. 1. Considering that the first trimester of pregnancy is a critical period for fetal limb development, we selected the first, second and third months of pregnancy as exposure windows of interest and estimated monthly average air pollutant concentrations for each pregnant woman during the first trimester of pregnancy. In addition, previous studies have shown that exposure of pregnant women to air pollutants before conception can increase the risk of birth defects19,22; therefore, we extended the exposure period of three months before conception to assess the impact of maternal air pollutant exposure on CLDs.

Fig. 1
Fig. 1
Full size image

Spatial distributions of the study participants and air monitoring stations in Wuhan. The map was generated using ArcGIS (version 10.8, Esri, URL: https://www.esri.com).

Statistical analysis

Frequencies and percentages were utilized to display the categorical sociodemographic characteristics of mothers and infants, with air pollutant concentrations described by means and standard deviations. Multiple imputation methods were used to handle missing data to ensure that bias in risk estimates was minimized. Given that the air pollutant concentration data exhibited right-skewed distributions, Spearman correlation coefficients were evaluated to assess the relationship between each pair of air pollutants. Multivariable logistic regression models were adopted to evaluate the relationships between ambient air pollutant exposure during different months before and after conception and the risk of CLDs in single-pollutant models, controlling for potential confounders: maternal age, occupation, residential area, parity, infant sex, preterm birth and birth type, which were chosen on the basis of their established associations with air pollutants exposure or CLDs in prior literatures (Chio et al., 2019; Tan et al.24,. The odds ratios (ORs) and 95% confidence intervals (CIs) for the risk of CLDs per 10 µg/m3 increase in PM2.5, PM10, SO2, NO2, O3, and per 100 µg/m3 increase in CO. The relationships between air pollutants and various CLD subtypes were examined separately via the same procedure. To assess the robustness of SO2 associations, we conducted two-pollutant models where SO2 was simultaneously adjusted for one additional pollutant at a time (SO2 + PM2.5, SO2 + PM10, etc.). To ensure the robustness of our two-pollutant models, we conducted multicollinearity diagnostics using variance inflation factors (VIF) prior to modeling, and the results indicated that the selected pollutant combinations did not exhibit significant multicollinearity (all VIFs values < 5).

Given the varying susceptibility to pollutants based on maternal age and infant sex19,25[,26, as well as the potential differences in exposure opportunities and concentrations due to maternal occupation, residential area, and season of conception19,25, along with the possibility that preterm birth might be more closely associated with birth defects27, stratified analyses were also conducted to assess the potential effect modification of the following factors: maternal age (≤ 35 vs. >35 years), occupation (professionals vs. liberal professions), residential area (urban vs. rural), infant sex (male vs. female), season of conception [warm season (April- October) vs. cold season (November-March)] and preterm birth (yes vs. no).Warm season (approximately from early April to late October) was classified when the 7-day rolling average temperature was ≥ 15 °C, while cold season (approximately from early November to late March) was classified when it was < 15 °C. Effect modification was assessed through two complementary approaches: (1) multiplicative interaction testing using product terms in logistic regression models, evaluated via likelihood ratio tests; and (2) between-stratum heterogeneity testing using Cochran’s Q test with inverse-variance weighting. Interaction p-values < 0.10 indicated significant effect modification, while Q p-values < 0.10 indicated significant heterogeneity across strata. We also performed sensitivity analyses by limiting CLDs diagnosed without other anomalies and removing stillborn and dead fetuses from the CLD group to evaluate the independent effects of air pollutants on the risk of CLDs. All analyses were implemented via SAS version 9.4 (SAS Statistical Institute, Inc., Cary, NC), with statistical significance set at a two-tailed P < 0.05.

Results

A total of 510,550 pairs of mothers and infants eligible for inclusion during the study period were included in this study, of which 1,864 infants were diagnosed with CLDs, with an incidence of 3.7/1,000. The demographic characteristics of the study subjects are shown in Table 1.

Table 1 Baseline characteristics of the study population.
Full size table

Table 2 provides a description of the distribution of ambient air pollutant concentrations in Wuhan during the study period. The median and range (25th percentile to 75th percentile) were 55.0 (35.0–83.0) µg/m3 for PM2.5, 90.0 (59.0–129.0.0.0) µg/m3 for PM10, 14.0 (8.0–28.0) µg/m3 for SO2, 44.0 (29.0–64.0) µg/m3 for NO2, 973.0 (733.0–1,256.0) µg/m3 for CO, and 85.0 (49.0–128.0.0.0) µg/m3 for O3. The Spearman correlations between the two air pollutants were significant (P < 0.05). All ambient air pollutants were positively correlated with one another, with moderate to high Pearson correlation coefficients ranging from 0.421 to 0.843, except for O3. O3 was negatively correlated with other air pollutants, with correlation coefficients ranging from − 0.036 to −0.261 (Table S1).

Table 2 Distribution of air pollutants during the study period.
Full size table

Table 3 shows the adjusted ORs and 95% CIs of the relationships between air pollutant exposure before conception and during the first three months after conception and between CLDs and their subtypes. We found that maternal SO2 exposure was significantly related to a greater risk of overall CLDs during the first (OR = 1.033, 95% CI: 1.005–1.062), second (OR = 1.041, 95% CI: 1.013–1.070), and third months of conception (OR = 1.043, 95% CI: 1.015–1.072) months, whereas no statistically significant effect was found between exposure to PM2.5, PM10, NO2, CO or O3 and the risk of overall CLDs. In addition, air pollutant exposure at 3 months before conception did not correlate with a higher risk of overall CLDs. In the subgroup analysis, the odds of polydactyly and limb shortening remained elevated for every 10 µg/m3 increase in SO2 concentration during the first three months of conception, whereas no similar associations were observed in the other subgroups. We additionally observed no significant correlation between exposure to air pollutants and an increased risk of syndactyly and clubfoot, except that exposure to O3 in the third month of conception increased the risk of syndactyly. After either PM2.5, PM10, NO2, CO, or O3 was added to the two-pollutant model, the associations between maternal SO2 exposure and increased risk of CLDs remained unchanged, and the estimated effect of SO2 increased (Table S2). The robustness of the associations between air pollutant exposure during different gestational stages and CHD risk was validated by performing sensitivity analyses. The results indicated that after infants with CLD accompanied by other congenital abnormalities were excluded, no significant changes were observed in the relationships between air pollutant exposure and CHD risk. Specifically, maternal SO2 exposure during the second and third months was significantly associated with overall CLD risk. Table S3 presents the ORs and 95% CIs for CLD risk related to air pollutant exposure.

Table 3 Odds ratios (ORs) and 95% confidence intervals (95% CIs) for risk of CLDs and their subtypes attributable to maternal air pollutants exposure during different gestation periods.
Full size table

Table 4 presents the modifying effects of maternal age, occupation, residential area, season of conception, infant sex and preterm birth on the associations of maternal SO2 exposure with the risk of CLDs. We observed that both multiplicative interaction testing and heterogeneity analysis provided consistent evidence that maternal occupation significantly modifies the association between SO₂ exposure in the first (P for interaction = 0.026, P for Q test = 0.001), second (P for interaction = 0.006, P for Q test < 0.001), and third months (P for interaction = 0.023, P for Q test < 0.001) and the risk of CLDs, with stronger effects observed among mothers who were in professional occupations. Although the multiplicative interaction did not reach statistical significance, we observed significant stratified heterogeneity between maternal age and season of conception in the first three months. We observed statistically significant multiplicative interactions in the stratified analyses for residential area (3 months before conception: P for interaction = 0.023, 3rd month after conception: P for interaction = 0.061) and preterm birth (3 months before conception: P for interaction = 0.044), but the heterogeneity between strata did not reach conventional significance levels (P for Q test>0.05). Although we observed that male infants presented a relatively greater risk of CLD associated with SO2 exposure, neither multiplicative interaction testing nor heterogeneity analysis provided evidence of effect modification by infant sex.

Table 4 Specific ORs (95% CIs) for risk of CLDs associated with SO2 exposure during different gestation stratified by individual characteristics.
Full size table

Discussion

This study was the first to systematically investigate the associations of air pollutant exposure before and after conception with the risk of CLDs and its sociodemographic modifying effects based on a large prospective population cohort. We observed that exposure to SO2 during the first three months after conception was significantly associated with an increased risk of overall CLDs, whereas preconception exposure was not significantly associated. The associations between SO2 exposure and the risk of CLDs in the two-pollutant model remained relatively unchanged. In subgroup analyses, we observed that SO2 exposure in the first three months after conception still increased the risk of polydactyly and limb shortening but not syndactyly or clubfoot. We also observed that sociodemographic factors significantly modified the SO2-CLD association.

Few studies have determined the correlations between air pollutant exposure before and after conception and the risk of CLDs, and the conclusions are inconsistent. Consistent with our main findings, a matched case-control study involving 1,687 cases of CLDs and 16,870 birth controls in Taiwan, China, reported that SO2 exposure during the first three months of pregnancy increased the risk of CLDs, and no significant associations were reported between PM10, NO2, CO, and O3 exposure and the risk of CLDs20. Another case-control study including 972 cases of CLDs and 9,720 controls in Changsha, China, reported that maternal SO2 and NO2 exposure in the first trimester of pregnancy was positively correlated with overall CLDs and their subtypes17. A multicenter case-control study conducted in the United States also revealed a significant association between SO2 exposure during weeks 2–8 of gestation and increased risk of CLDs but reported a null association for CLDs per 10 µg/m3 increase in PM2.5, PM10, and O3 (Chio et al., 2019). Cheng et al. conducted an observational study on air pollutants in the first trimester of pregnancy and birth defects and reported a significant association only between NO2 exposure and CLDs, but no negative effects on CLDs were found for other pollutants21. Similar to the findings of a study conducted in Liaoning19, we also reported that SO2 exposure during early pregnancy was significantly positively associated with the risk of specific CLD subtypes, such as polydactyly and limb shortening, but not with syndactyly or clubfoot. This finding was slightly inconsistent with studies conducted in Changsha and Liaoning, which reported strong links between SO2 exposure and the risk of syndactyly and clubfoot17,19. This discrepancy might stem from heterogeneity in the underlying etiology of these subtypes or from variations in sample sizes across studies. Our study was a prospective cohort study based on a large sample population and had a sufficient sample size for further subtype analyses, but some studies had limited sample sizes or availability for specific subtypes, and further subdivision may have led to heterogeneity in the results. Future studies with adequately powered sample sizes are warranted to explore the subdivisions of CLDs, which would facilitate a more accurate description of the relationship between air pollution and CLDs. The existing conclusions on the associations between SO2 exposure and the risk of CLDs are still inconsistent, and the inconsistencies in these studies may stem from differences in air pollutant concentration levels, exposure assessments, and population susceptibility to air pollution exposure. The pathophysiological mechanisms underlying the relationship between air pollution and congenital limb defects are not fully understood. Previous studies have shown that exposure to air pollutants in early pregnancy may enhance the interaction of genetic and environmental factors with congenital abnormalities, and air pollutants may also affect epigenetic factors associated with congenital abnormalities by altering the sequence of DNA molecules21,28,29. More studies are needed to further confirm this link and its potential pathogenic mechanisms in the future.

Our results also indicated that the relationships between maternal SO2 exposure and the risk of CLDs in offspring were modified by individual characteristics. We observed that maternal occupation significantly modifies the association between SO₂ exposure and the risk of CLDs, with stronger effects observed among mothers who were in professional occupations, suggesting that greater attention should be paid to professional groups in maternal health care. We observed significant between-stratum heterogeneity among maternal age groups despite non-significant multiplicative interaction. This inconsistency suggests that while the per-unit exposure effects migth be similar across age groups, the absolute risks were significantly higher for younger mothers, possibly due to higher susceptibility or additive-scale effect modification among younger mother, and older pregnant women are more cautious and more willing to take precautions to prevent environmental pollution and other risk factors25. We observed that preterm infants had a significantly greater risk of CLDs than term infants did. Animal studies have shown that SO2 can cause chromosome abnormalities and SCE in mammalian cells, that preterm animals have high blood-brain barrier permeability, and that SO2 is more likely to enter developing tissues, directly destroying limb morphogenesis signaling pathways (such as FGF/Wnt), and increasing the risk of limb defects20,30. Although the seasonal multiplicative interaction was not statistically significant, we observed significant heterogeneity between groups, suggesting that the unit exposure effects may be similar across seasons, but the absolute risk was significantly higher in warmer seasons, which was consistent with previous studies17,31. This phenomenon might stem from the fact that warm season serve not only as a surrogate for temperature but also as a comprehensive indicator encompassing air pollutants, biological responses to air pollutants, and lifestyle. For example, during warm season, people tend to engage in more outdoor activities and are exposed to greater concentrations of sulfur dioxide emissions31.

Although this study did not find evidence that infant sex had no significant effect modification, we also observed that maternal SO2 exposure in early pregnancy was correlated with a greater risk of limb defects in male offspring, which might be related to the expression of sex chromosome-related genes and their interaction with hormonal effects during early fetal development19,32. An animal study indicated that maternal air pollutant exposure could inhibit gene expression in male mice, resulting in increased structural pathological damage to the offspring of these mice, and this effect was more pronounced in male mice32. Moreover, we also found that maternal residential area could modify the associations between prenatal SO2 exposure and the risk of CLD in offspring, suggesting that maternal demographic characteristics might alter the sensitivity of both mothers and offspring to air pollutants. Therefore, the above conclusions are highly important for pregnant women to take personalized protective measures against air pollutants during pregnancy according to their individual characteristics.

This study has several strengths. First, this study was the first to systematically investigate the associations between maternal exposure to air pollutants before and after conception and the risk of CLDs in offspring on the basis of a large prospective cohort population, providing evidence that high-level SO2 exposure in early pregnancy increases the risk of CLDs and their subtypes. Second, the study adopted a large-scale population-based design and utilized a substantial sample size collected from Wuhan’s Maternal and Child Health Information System, which could effectively reduce selection bias, increase statistical accuracy, and enable us to further examine associations across specific CLD subtypes. More importantly, this allowed us to validate the reliability of our findings through sensitivity analyses and subgroup analyses. These findings could help clinicians provide targeted personalized interventions on the basis of the specific characteristics of pregnant women.

This study also has several limitations that need to be addressed. First, we utilized air pollution data from government monitoring stations rather than actual pollutant concentrations at individual residential locations. Although we partially corrected this hard bias by incorporating the spatial gradient of pollutants into the individual exposure assessment system using the inverse distance weighting method, this statistical model-based exposure assessment method could not accurately reflect the true exposure levels of individuals, potentially leading to exposure misclassification33. Second, while growing evidence has confirmed that indoor air pollution is harmful to individual health34,35, our study lacked data on participants’ indoor pollution concentrations and daily activity patterns, making the assessment of individual air pollutant exposure less comprehensive. Third, although we focused on the critical period for germ cell formation and fetal limb development (i.e., three months before and after conception), this timeframe might be insufficient to capture all critical exposure periods. Previous studies have suggested that air pollutant exposure has lagged effects36,37, indicating that exposure during mid-to-late pregnancy may also have an effect on limb defect risk. Fourth, since birth defects data was obtained from the government registration systems, we were unable to obtain data on CLDs in offspring born before 28 weeks of gestation, and limb defects might be missed at birth or within 7 days of birth, which may weaken the extrapolability of the results. Fifth, although we conducted multiple subgroup analyses to identify vulnerable populations, it must be acknowledged that this might increase the risk of Type I error. The interaction findings should be regarded as hypothesis-generating results and require validation in independent cohorts. Future studies could adopt a Bayesian framework for interaction testing to more accurately quantify the strength of evidence. Sixth, although our results suggested that maternal exposure to SO2 during the second and third months of pregnancy had a relatively strong effect on the occurrence of CLDs in offspring, which was consistent with previous studies confirming that the most sensitive period for the embryo was between gestational weeks 6 and 1038,39. However, due to our study design and the lack of precise temporal exposure data, we could not definitively distinguish whether the observed association reflects the effect of a specific critical exposure window or the cumulative effect of multiple exposures across multiple gestational periods. This clear distinction was crucial for implementing potential interventions to prevent the occurrence of limb defects. In future studies, more accurate exposure measurement methods and statistical models are employed to overcome this issue. Finally, owing to data limitations, we could not consider potential confounding factors, such as maternal nutritional status, folic acid intake, genetic predispositions, and residential mobility during pregnancy, which might combine with air pollution to interfere with fetal development, thus weaken the results of our study. In summary, more prospective cohort studies are encouraged in the future to adopt a more complete study design to address the above limitations.

Conclusion

This study provides evidence that increased exposure to SO2 during the first three months after conception increases the risk of CLDs and their subtypes in offspring and that these associations might be modified by individual characteristics. These results highlight the importance of avoiding SO2 exposure in early pregnancy and the effect modification of individual factors. Future studies should further explore the biological mechanisms underlying the relationship between SO2 exposure and the risk of CLDs to develop more effective preventive measures.