Abstract
Background
We aimed to estimate the association of air pollution and term low birth weight (LBW) by maternal education and assess the distributional impact of meeting European Union and World Health Organization air pollution targets and universal tertiary maternal education on LBW cases.
Methods
We analyzed administrative data on 1 409 084 term births in Spain (2010–18) and high spatial resolution estimates of maternal particulate matter with a diameter of 2.5 micrometers or less (PM2.5) and nitrogen dioxide (NO2) exposure. Single-exposure distributed lag non-linear models were used to estimate exposure–response functions (ERF) for LBW (<2500 g) overall and by maternal education. We then applied G-computation to estimate changes in LBW cases under counterfactual scenarios: target levels of PM2.5 (10 and 5 µg/m³) and NO2 (20 and 10 µg/m³), and universal tertiary maternal education.
Results
A 10 µg/m³ increase in PM2.5 and NO2 was associated with 6% [95% confidence interval (CI): 0%–11%] and 1% (95% CI: 0%–3%) higher odds of LBW. Relative to tertiary education, secondary and primary education were associated with increased LBW odds [1.32 (95% CI: 1.30–1.35); 1.76 (95% CI: 1.71–1.82)]. Meeting the 10 and 5 µg/m³ PM2.5 limits would have prevented 662 (95% CI: −4 to 1271) and 2032 (95% CI: −74 to 3900) LBW cases in 2010–18. Meeting the 20 and 10 µg/m³ NO2 limits would have prevented 432 (95% CI: −16 to 848) and 938 (95% CI: −50 to 1860) LBW cases in 2010–18. Universal attainment of tertiary education would have prevented over 9200 LBW cases. The primary maternal education group showed the highest shares of LBW cases, but they also experienced the largest relative reductions under air pollution intervention scenarios.
Conclusions
Expanding access to higher education may yield large reductions in the burden of LBW. Reducing air pollution exposure could also meaningfully reduce LBW burden, particularly among infants from lower educated mothers, and contribute to narrowing related health inequalities.
Keywords: air pollution, low birth weight, NO2, PM2.5, maternal education, health inequality
Key Messages.
We estimated potential reductions in low birth weight from meeting air quality targets and improving maternal education using Spanish birth registry data.
Particulate matter with a diameter of 2.5 micrometers or less (PM2.5) and nitrogen dioxide (NO2) exposure increased the risk of low birth weight, particularly among infants born to mothers with lower education.
Achieving World Health Organization air quality guidelines or universal tertiary education could prevent thousands of low birth weight cases and reduce social inequalities in low birth weight.
Introduction
Low birth weight (LBW) has profound and lasting implications across the lifespan, affecting health and development from infancy through adulthood. While some of its effects, such as poorer school performance and suboptimal growth, diminish over time, others, including mild cognitive, attention, and neuromotor issues, persist into later childhood and adolescence [1, 2]. LBW is also a significant risk factor for chronic conditions in adulthood, including hypertension, type 2 diabetes, cardiovascular diseases, and chronic kidney disease. These long-term risks are partly attributed to the fetus adapting to scarcity, which can become maladaptive decades later, such as underlying insulin resistance that begins at birth [3–6]. Furthermore, LBW increases the likelihood of asthma, cerebral palsy, and osteoporosis, while being associated with earlier menopause in women [7, 8]. Early preventive interventions targeting modifiable environmental exposures and social determinants can provide large cumulative benefits across the life course [1, 2].
Air pollution is a major global public health concern, with serious implications for maternal and infant health. Exposure to pollutants like particulate matter with a diameter of 2.5 micrometers or less (PM2.5) and nitrogen dioxide (NO2) during pregnancy has been linked to LBW, preterm birth (PTB), and small for gestational age (SGA) infants [9–13]. These outcomes increase the risk of long-term health issues and poorer economic attainment [14, 15]. Reducing pollution levels could help reduce these risks. A previous study documented that a 10 µg/m³ reduction in NO2 could prevent 8% of LBW cases [12].
The burden of air pollution, however, is not equally distributed across the population. Lower socioeconomic groups are disproportionately affected by exposure to air pollutants and the consequent adverse outcomes [17]. Moreover, mothers with lower educational attainment are particularly susceptible to the negative effects of air pollution on birth outcomes [18]. This heightened susceptibility reflects the combined influence of chronic stress, poorer nutrition, and reduced healthcare, which may exacerbate placental oxidative stress and inflammatory responses triggered by air pollution exposure [19], highlighting the interplay between environmental and social determinants of health [20].
Our objectives were to (i) quantify the association of long-term air pollution (NO2 and PM2.5) exposure and LBW according to maternal education, (ii) quantify the impact of meeting air pollution targets on cases of LBW and on socioeconomic inequalities in LBW, and (iii) compare the impact of meeting air pollution targets with other potential policy options, specifically increasing maternal education.
Methods
Study population
We used birth registry data provided by the National Statistics Institute, including all births on Spanish territory except the Canary Islands, Ceuta, and Melilla between 1 January 2010 and 31 December 2018. The study population included a total of 1 409 084 term births between 37 and 40 weeks of gestation. The registry included geocoded maternal home addresses with added spatial noise of about 40 m to pseudonymize the data.
Low birth weight
Birth weight was measured in grams. LBW at term was defined as a birth weight <2500 g.
Air pollution exposure
Air pollution data for NO2 and fine particulate matter (PM2.5) were derived from geographically and temporally weighted regression models developed for Europe from 2000 to 2019 as part of the EXPANSE project and described in detail elsewhere [21, 22]. Briefly, models integrated routine monitoring data with spatiotemporal predictors such as land use, satellite data, and meteorological variables to predict monthly air pollution at a high spatial resolution of 100 m × 100 m (see Extended Methods in Supplementary Material).
We defined the pregnancy period for each birth using the reported date of birth and gestational age, dividing it into consecutive four-week intervals (with a shorter final interval for weeks 37 and 40). Monthly air pollution concentrations were assigned to geocoded maternal home addresses via spatial linkage with pollution grids. In utero exposure for each interval was calculated as a weighted average of monthly concentrations based on the number of gestational days within each month. Supplementary Fig. S1 shows the mapped temporal evolution of in utero air pollution exposure between 2010 and 2018.
Maternal education
Maternal education was measured in 10 categories at the time of birth. The variable was recoded according to the International Standard Classification of Education 2011 (ISCED 2011) scale and subsequently categorized into “primary,” “secondary,” and “tertiary” education (see Supplementary Table S1) [23].
Statistical analyses
The statistical analyses were performed in two steps. First, we estimated single-exposure associations between air pollution and LBW to obtain exposure–response functions (ERF) overall and by individual-level maternal education level, as well as the association between maternal education and LBW. We estimated single-exposure models to avoid multicollinearity and to ensure stable estimation of the ERF used for the G-computation analyses in the next step. Second, we predicted LBW cases under different exposure scenarios consistent with air quality and education targets.
Associations between air pollution and low birth weight
We used single-exposure distributed lag non-linear models (DLNM) with a binomial link function to estimate the ERF between monthly average PM2.5 and NO2 exposure during pregnancy and LBW at term, overall and stratified by maternal education [16]. Models estimated (i) the overall cumulative ERF of PM2.5 or NO2 and LBW and (ii) the association of air pollution and LBW at each lag (i.e. 10 four-week intervals or pregnancy months) to identify windows of susceptibility [9]. The association of maternal education with LBW was estimated from the overall model.
Model selection for the exposure-lag-response functions was guided by prior literature, visual inspection of resulting graphs, and the Bayesian Information Criterion and Akaike Information Criterion (Supplementary Table S2) [9, 24, 25]. The final selected models included linear functions for the exposure–response and the lag-response for both PM2.5 and NO2. Effect modification by maternal education was tested using an air pollution cross-basis × education interaction and comparing it to a model without interaction with a likelihood ratio test.
All models were adjusted for covariates identified by using a directed acyclic graph (DAG) and from previous literature (Supplementary Fig. S2) [9, 26–28]: Birth year to adjust for secular trend, infant’s sex (female/male), singleton (yes/no), maternal age at birth in years, maternal marital status (married/not married), maternal citizenship status (Spanish/not Spanish), parity (nulliparous/multiparous), paternal citizenship status (Spanish/not Spanish), degree of urbanization of maternal residence’s municipality (city/town or suburb/rural) [29], 2011 area-level deprivation index [30], yearly mean normalized difference vegetation index (NDVI) in a 300-m buffer for years 2010, 2015, 2020 (births were assigned the closest value to date of birth). Monthly mean temperature was modeled using a natural cubic spline with 2 df for the exposure–response and a penalized spline for the lag-response consistent with previous literature [30, 31]. We also included a random intercept at the NUTS 1 level (i.e. autonomous communities grouped into regions: North-West, North-East, Madrid, Center, East, South) to account for spatial autocorrelation. The overall model was additionally adjusted for maternal education level (primary/secondary/tertiary). While the DAG informed the minimally sufficient confounder set, we additionally adjusted for established predictors of LBW (infant sex, singleton status, maternal age, and parity) to improve model precision and account for baseline risk differences.
In an exploratory analysis, we ran models for continuous birth weight [see Supplementary Material, Appendix A1, for PM2.5 and birth weight ERF (Supplementary Fig. S7) and lag response (Supplementary Fig. S8), and NO2 and birth weight ERF (Supplementary Fig. S9) and lag response (Supplementary Fig. S10)].
Hypothetical intervention scenarios
We considered hypothetical interventions to retrospectively reduce air pollution concentrations and limit pregnancy air pollution exposure according to the revised EU Ambient Air Quality Directive (AAQD) and 2021 WHO Air Quality Guidelines (AQG) (Table 1) [31, 32]. We also considered a social intervention in which all mothers attained tertiary education (see Supplementary Table S1).
Table 1.
Description of scenarios.
| Scenarios | Description |
|---|---|
| Reference | Observed exposure for PM2.5 and NO2 |
| EU AAQD | Individual PM2.5 and NO2 exposure limited at 10 and 20 µg/m³, respectively |
| WHO AQG | Individual PM2.5 and NO2 exposure limited at 5 and 10 µg/m³, respectively |
| Universal tertiary education | Mothers universally attain tertiary education (i.e. advanced vocational training, or Bachelor’s degree or higher) |
Notes: EU AAQD = European Union Ambient Air Quality Directive, WHO AQG = World Health Organisation Air Quality Guidelines. To analyze the counterfactual scenarios, we capped each infant’s in-utero air pollution exposure at the corresponding limit, i.e. for each lag the observed exposure was artificially set to have a maximum equal to the limit.
G-computation was implemented in the DLNM models that generated the exposure–response relationships to predict the probability of LBW under counterfactual exposure levels [33–35]. First, we obtained the predicted log-odds of LBW for each infant in the “reference” scenario (i.e. under observed exposure, Table 1) among all infants and stratified by maternal education (see Extended Methods in Supplementary Material).
Second, to evaluate the air quality interventions, we capped each infant’s in utero exposure at the corresponding EU AAQD and WHO AQG limits, setting observed lag-specific exposures to the guideline maximum where exceeded. With this altered exposure matrix, we recomputed the cross-basis, which was then used to predict the probability of LBW for each infant. The sum of these probabilities yielded the predicted LBW cases under the improved air quality scenarios for all infants and each stratum of maternal education. To account for parameter uncertainty, we applied a parametric simulation approach to calculate confidence intervals (CIs) (see Extended Methods in Supplementary Material).
Third, we evaluated a social intervention, changing all mothers’ education level to “tertiary.” We predicted the absolute number and change and percent change of LBW cases under each scenario overall and stratified by maternal education level.
Analyses were conducted using R version 4.4.2, with models fitted using the dlnm and mgcv packages [36–38].
Results
Table 2 presents descriptive statistics overall and by maternal education. Among 1 409 084 term births, 11.2%, 38.7%, and 50.1% were to mothers with primary, secondary, and tertiary education, respectively. Mean pregnancy PM2.5 was similar across groups [11.41 (SD: 2.24), 11.44 (SD: 2.27), and 11.60 (SD: 2.30) µg/m³ for primary, secondary, and tertiary, respectively). Mean NO2 exposure was comparable for primary and secondary [21.31 (SD: 9.16) and 21.88 (SD: 9.68) µg/m³] but higher for tertiary education [24.06 (SD: 10.57) µg/m³] (Supplementary Fig. S3). Exposures remained stable throughout pregnancy (Supplementary Fig. S4). LBW prevalence decreased with increasing maternal education (5.3%, 4.9%, and 4.3% for primary, secondary, and tertiary, respectively).
Table 2.
Characteristics of the study population by maternal educational level.
| Overall N = 1 409 084 | Primary N = 158 037 | Secondary N = 545 463 | Tertiary N = 705 584 | |
|---|---|---|---|---|
| Birth weight (mean in g, SD) | 3234.80 (449.50) | 3239.44 (475.10) | 3231.12 (457.07) | 3236.60 (437.52) |
| Low birth weight (N, %) | 65 027 (4.6%) | 8334 (5.3%) | 26 678 (4.9%) | 30 015 (4.3%) |
| PM2.5 exposure (µg/m³ average across pregnancy, SD) | 11.52 (2.28) | 11.41 (2.24) | 11.44 (2.27) | 11.60 (2.30) |
| NO2 exposure (µg/m³ average across pregnancy, SD) | 22.91 (10.15) | 21.31 (9.16) | 21.88 (9.68) | 24.06 (10.57) |
| Year of birth, N (%) | ||||
| 2010 | 57 158 (4.1%) | 5618 (3.6%) | 22 954 (4.2%) | 28 586 (4.1%) |
| 2011 | 212 989 (15%) | 20 275 (13%) | 83 142 (15%) | 109 572 (16%) |
| 2012 | 210 761 (15%) | 21 676 (14%) | 81 771 (15%) | 107 314 (15%) |
| 2013 | 195 181 (14%) | 20 006 (13%) | 73 825 (14%) | 101 350 (14%) |
| 2014 | 194 525 (14%) | 20 563 (13%) | 73 328 (13%) | 100 634 (14%) |
| 2015 | 183 442 (13%) | 20 616 (13%) | 70 151 (13%) | 92 675 (13%) |
| 2016 | 134 625 (9.6%) | 17 518 (11%) | 52 789 (9.7%) | 64 318 (9.1%) |
| 2017 | 134 911 (9.6%) | 18 929 (12%) | 53 372 (9.8%) | 62 610 (8.9%) |
| 2018 | 85 492 (6.1%) | 12 836 (8.1%) | 34 131 (6.3%) | 38 525 (5.5%) |
| Sex, N (%) | ||||
| Male | 723 138 (51%) | 81 515 (52%) | 279 518 (51%) | 362 105 (51%) |
| Female | 685 946 (49%) | 76 522 (48%) | 265 945 (49%) | 343 479 (49%) |
| Singleton, N (%) | 1 368 448 (97%) | 155 322 (98%) | 531 464 (97%) | 681 662 (97%) |
| Maternal age at birth, mean (SD) | 32.51 (5.18) | 27.88 (6.28) | 31.66 (5.20) | 34.20 (3.95) |
| Mother is married, N (%) | 860 100 (61%) | 66 839 (42%) | 301 431 (55%) | 491 830 (70%) |
| Maternal citizenship status, N (%) | ||||
| Spanish | 1 231 003 (87%) | 104 640 (66%) | 458 502 (84%) | 667 861 (95%) |
| Not Spanish | 178 081 (13%) | 53 397 (34%) | 86 961 (16%) | 37 723 (5.3%) |
| Parity, N (%) | ||||
| Multiparous | 711 713 (51%) | 93 155 (59%) | 281 756 (52%) | 336 802 (48%) |
| Nulliparous | 697 371 (49%) | 64 882 (41%) | 263 707 (48%) | 368 782 (52%) |
| Paternal citizenship status, N (%) | ||||
| Spanish | 1 232 859 (87%) | 104 784 (66%) | 460 973 (85%) | 667 102 (95%) |
| Not Spanish | 176 225 (13%) | 53 253 (34%) | 84 490 (15%) | 38 482 (5.5%) |
| Area-level deprivation index (standardized mean, SD) | −0.22 (0.99) | 0.40 (1.01) | −0.01 (0.93) | −0.51 (0.91) |
| Degree of urbanization of maternal residence municipality, N (%) | ||||
| City | 761 354 (54%) | 76 331 (48%) | 268 303 (49%) | 416 720 (59%) |
| Town/suburb | 492 865 (35%) | 63 755 (40%) | 209 144 (38%) | 219 966 (31%) |
| Rural | 154 865 (11%) | 17 951 (11%) | 68 016 (12%) | 68 898 (9.8%) |
| NDVI in 300 m around maternal residence (yearly mean, SD) | 0.27 (0.12) | 0.25 (0.10) | 0.27 (0.12) | 0.28 (0.12) |
| Temperature (average °C across pregnancy, SD) | 16.41 (2.60) | 16.82 (2.57) | 16.52 (2.60) | 16.22 (2.60) |
| Season at birth, N (%) | ||||
| Spring | 343 091 (24%) | 36 507 (23%) | 130 268 (24%) | 176 316 (25%) |
| Summer | 351 629 (25%) | 40 369 (26%) | 137 474 (25%) | 173 786 (25%) |
| Autumn | 377 785 (27%) | 43 066 (27%) | 147 628 (27%) | 187 091 (27%) |
| Winter | 336 579 (24%) | 38 095 (24%) | 130 093 (24%) | 168 391 (24%) |
| NUTS 1 region, N (%) | ||||
| ES1 (North-West) | 98 333 (7.0%) | 6081 (3.8%) | 37 545 (6.9%) | 54 707 (7.8%) |
| ES2 (North-East) | 170 924 (12%) | 16 059 (10%) | 55 074 (10%) | 99 791 (14%) |
| ES3 (Madrid) | 180 178 (13%) | 13 573 (8.6%) | 62 196 (11%) | 104 409 (15%) |
| ES4 (Center) | 158 643 (11%) | 18 400 (12%) | 65 542 (12%) | 74 701 (11%) |
| ES5 (East) | 464 411 (33%) | 51 535 (33%) | 180 115 (33%) | 232 761 (33%) |
| ES6 (South) | 336 595 (24%) | 52 389 (33%) | 144 991 (27%) | 139 215 (20%) |
Notes: SD = standard deviation, NDVI = yearly mean normalized difference vegetation index
Associations between air pollution, maternal education, and low birth weight
Main results are presented as exposure–response curves showing the cumulative association of PM2.5 or NO2 exposure (with the reference being 0 µg/m³ exposure) separately and the overall odds ratios (ORs) for LBW across the entire lag period [16, 24]. Results for the “windows of susceptibility” can be found in the Supplementary Material (see lag–response relationships).
PM2.5 and low birth weight
We found a positive linear association between PM2.5 and LBW in the total sample, with the CI including the null (OR: 1.06; 95% CI: 1.00, 1.11 per 10 µg/m³ increase in exposure) (Fig. 1A). The association differed by maternal education (likelihood ratio test P-value = .02); CI for primary, but not for other education groups, excluded the null (Fig. 1B).
Figure 1.
Odds ratios for low birth weight per PM2.5 exposure, cumulative over the entire pregnancy, compared to 0 µg/m³, for all infants (A) and stratified by maternal education (B).
NO2 and low birth weight
A 10 µg/m³ increase in NO2 was associated with a 1% (95% CI: 0%–3%) increased odds of LBW (Fig. 2A), with no difference according to maternal education (likelihood ratio test P-value = .98) (Fig. 2B).
Figure 2.
Odds ratios for low birth weight per NO2 exposure, cumulative over the entire pregnancy, compared to 0 µg/m³, for all infants (A) and stratified by maternal education (B).
Maternal education and low birth weight
Compared with infants of mothers with tertiary education, infants of mothers with secondary and primary education had 32% (95% CI: 30%–35%) and 76% (95% CI: 71%–82%) higher odds of LBW, respectively.
Predicted low birth weight cases under hypothetical intervention scenarios
Between 2010 and 2018, we predicted 65 036 LBW cases (95% CI: 64 584–65 507) under the reference PM2.5 scenario (Fig. 3A) and 65 045 (95% CI: 64 576–65 556) under the reference NO2 scenario (Fig. 3B).
Figure 3.
Predicted low birth weight cases (with 95% confidence intervals) under PM2.5 (A) and NO2 (B) scenarios among all infants, 2010–18.
Under PM2.5 counterfactuals, limiting exposure to 10 and 5 µg/m³ was associated with reductions of 662 (95% CI: −4 to 1271) and 2032 (95% CI: −74 to 3900) LBW cases (Fig. 4A), corresponding to a decrease in LBW cases from 4.62% (95% CI: 4.58%–4.65%) to 4.47% (95% CI: 4.34%–4.62%) (Fig. 5A). However, CI for these reductions included the null.
Figure 4.
Predicted reduction of low birth weight cases (with 95% confidence intervals) under PM2.5 and NO2 scenarios compared with the reference scenario (observed exposure), overall and by maternal education, 2010–18. EU AAQD limits are 10 and 20 µg/m³ and WHO AQG limits are 5 and 10 µg/m³ for PM2.5 and NO2, respectively.
Figure 5.
Percent of low birth-weight cases (with 95% confidence intervals) under PM2.5 (A) and NO2 scenarios (B), overall and by maternal education, 2010–18. EU AAQD limits are 10 and 20 µg/m³ and WHO AQG limits are 5 and 10 µg/m³ for PM2.5 and NO2, respectively.
Under NO2 counterfactuals, limits of 20 and 10 µg/m³ were associated with reductions of 432 (95% CI: −16 to 848) and 938 (95% CI: −50 to 1860) LBW cases (Fig. 4B), corresponding to a decrease from 4.62% (95% CI: 4.58%–4.65%) to 4.55% (95% CI: 4.47%–4.63%) (Fig. 5B). CI for these reductions included the null.
In contrast, the social intervention scenario (all mothers attaining tertiary education) was associated with reductions of 9246 (95% CI: 8777–9737) LBW cases in the PM2.5 model (Fig. 4A) and 9275 (95% CI: 8743–9726) in the NO2 model (Fig. 4B), corresponding to a decrease in LBW cases from 4.62% (95% CI: 4.58%–4.65%) to 3.96% (95% CI: 3.91%–4.00%) (Fig. 5A and B).
Stratified analyses (Fig. 6A and B) indicated the largest relative reductions among infants of mothers with primary education, whereas estimates for secondary and tertiary education groups were relatively smaller and included the null. For PM2.5, estimated reductions among the primary educated group were 225 cases (95% CI: 7–435) at 10 µg/m³ and 759 cases (95% CI: 81–1387) at 5 µg/m³ (Fig. 4A), corresponding to decreases of 0.14 and 0.48 percentage points, respectively (Fig. 5A). For NO2, estimated reductions were 148 cases (95% CI: 23–286) at 20 µg/m³ and 376 cases (95% CI: 67–702) at 10 µg/m³ (Fig. 4B), corresponding to decreases of 0.09 and 0.23 percentage points (Fig. 5B).
Figure 6.
Predicted low birth weight cases (with 95% confidence intervals) under PM2.5 (A) and NO2 (B) scenarios by maternal education, 2010–18. EU AAQD limits are 10 and 20 µg/m³ and WHO AQG limits are 5 and 10 µg/m³ for PM2.5 and NO2, respectively.
Discussion
Our analysis led to several key findings. First, we observed positive linear, but imprecise, exposure–response relationships between PM2.5 and NO2 and LBW. Infants born to mothers with the lowest educational attainment experienced the highest odds of LBW due to PM2.5 exposure compared with other educational groups. Differences in the odds of LBW due to NO2 exposure were less conclusive. Second, our findings provide evidence that reductions in PM2.5 and NO2 exposure during pregnancy could lead to significant declines in LBW cases, with the greatest benefits observed among mothers with lower educational attainment. Third, increasing maternal education led to larger reductions in LBW cases compared to achieving air pollution targets, although reaching these targets could improve social inequalities in LBW cases.
Our findings related to the first objective add to the mixed results observed in previous epidemiological studies reporting on the relationship between air pollution and LBW. Existing evidence linking PM2.5 exposure to LBW is inconsistent [9, 10, 28]; our study observed an association by a small margin due to large CIs, consistent with the results of a recent study in a similar population reporting an OR of 0.99 (95% CI: 0.96–1.01) per 5 µg/m³ exposure increment [25]. Similarly, we observed no relationship between NO2 and LBW due to large CIs, which contrasts with two meta-analyses reporting small but positive associations: one found an OR of 1.05 (95% CI: 1.00–1.09) per 20 ppb increase in NO2 [39], and another reported 1.03 (95% CI: 1.01–1.05) per 10 ppb increase [40]. However, these represent larger exposure increments in terms of mass concentration than the exposure reductions modeled in our intervention scenarios.
Few studies have examined air pollution effects on birth outcomes by individual-level socioeconomic position. In Spain, PM10 exposure was associated with higher LBW risk among infants of mothers with lower education [25]. Similarly, a study in São Paulo reported stronger associations between PM10 and NO2 with LBW among less educated mothers, aligning with our findings of consistently greater impacts among infants born to mothers with lower educational attainment [41]. Other indicators of social disadvantage—including household income, race, and area-level deprivation—have also been shown to amplify pollution-related risks, highlighting greater vulnerability in socioeconomically disadvantaged populations [41–44]. Consistent with this evidence, we observed stronger adverse effects of PM2.5 on LBW among mothers with lower educational attainment despite comparable exposure levels. This heightened vulnerability may reflect compounding stressors, including nutritional deficiencies and poorer prenatal care access [41, 45, 46], alongside poorer maternal health and chronic stress, which may limit the capacity to buffer pollution-related insults [47].
In relation to the second objective, Arroyo et al. reported 8% of LBW cases (9385) in Spain (2001–09) could have been prevented with a 10-µg/m³ reduction in NO2 [12]. In comparison, achieving the WHO 10-µg/m³ guideline in our study period (2010–18) could have prevented approximately 938 cases (1.4%). This difference is likely due to the higher OR reported by Arroyo et al.: 1.09 (95% CI: 1.06–1.12), which was based on aggregate rather than individual-level data [12]. Additionally, the uncertainty intervals in our study included the null value. Our study extends existing literature by estimating PM2.5 and NO2 reductions by maternal education, suggesting larger benefits among infants of lower-educated mothers and potential reductions in social inequalities in LBW through improved air quality.
Findings for the third objective indicated greater reductions in LBW cases under improved maternal education than under air pollution targets. Previous research, including a study across 12 European countries, has reported elevated risks of PTB and SGA among mothers with lower education [48]. Education may influence pregnancy health through pathways such as nutrition, smoking, genital tract infections, prenatal care access and quality, environmental exposures, stress, and other psychosocial factors [49, 50].
A key strength of this study is its application of G-computation, which allows for causal inference under hypothetical intervention scenarios while accounting for complex exposure–response relationships and confounders. Additionally, our use of high-resolution air pollution estimates provides robust exposure assessments at the individual level. Nonetheless, several limitations should be considered. Residual confounding by unmeasured factors (e.g. maternal nutrition, stress, genetic predisposition) cannot be excluded. Exposure misclassification may have occurred because air pollution was assigned at residential addresses at delivery and monthly exposure data were aligned to weekly gestational windows; this likely resulted in non-differential error and attenuation of associations. We assumed stable exposure–response relationships across time and regions, did not adjust simultaneously for co-pollutants to avoid multicollinearity, and specified distributed lag models that are sensitive to smoothing choices. These decisions may introduce modeling uncertainty and limit inference on independent pollutant effects. Finally, the counterfactual scenarios represent hypothetical exposure shifts and do not specify the policy mechanisms or costs required to achieve air quality improvements or universal tertiary education. Thus, the estimates quantify potential health gains under defined exposure changes but do not assess feasibility or implementation pathways.
While our scenarios are not policy-specific, they illustrate potential reductions in LBW from meeting air quality targets; implications for socioeconomic inequalities; and how environmental interventions compare with other policy options like expanding access to higher education. Given LBW’s long-lasting health consequences across the life course, including hypertension, type 2 diabetes, cardiovascular diseases, and chronic kidney disease, minimizing in utero air pollution exposure is crucial [7, 8]. Our findings support calls for aligning EU air quality directives with WHO guidelines to maximize health benefits [51]. Additionally, integrating air pollution reduction efforts with social policies aimed at improving maternal education and healthcare access may yield synergistic benefits, further reducing disparities in birth outcomes. Future research should explore the cost-effectiveness of these interventions and their long-term health and economic impacts.
In conclusion, this study highlights the benefits of air pollution reduction and increased maternal education for low birth weight in Spain, particularly among vulnerable populations. Continued efforts to improve air quality and improve access to higher education could play an important role in reducing the burden of low birth weight infants and concomitant health complications and improving health equity.
Ethics approval
This study used secondary, de-identified data and therefore did not require ethics approval in accordance with applicable regulations.
Supplementary Material
Contributor Information
Stefan Sieber, ISGlobal, Barcelona, Spain; Universitat Pompeu Fabra (UPF), Barcelona, Spain.
Joan Ballester, ISGlobal, Barcelona, Spain.
Mark Nieuwenhuijsen, ISGlobal, Barcelona, Spain; Universitat Pompeu Fabra (UPF), Barcelona, Spain; CIBER Epidemiología y Salud Pública (CIBERESP), Madrid, Spain.
Andrea Pozzer, Max Planck Institute for Chemistry, Mainz, Germany; The Cyprus Institute, Nicosia, Cyprus.
Cathryn Tonne, ISGlobal, Barcelona, Spain; Universitat Pompeu Fabra (UPF), Barcelona, Spain; CIBER Epidemiología y Salud Pública (CIBERESP), Madrid, Spain; Centro de Investigação em Saúde de Manhiça (CISM), Maputo, Mozambique.
Author contributions
S.S. and C.T. designed the study. C.T. provided supervision. S.S. conducted statistical analyses. S.S., J.B., M.N., A.P., and C.T. interpreted the results and drafted the manuscript. J.B. acquired the administrative birth registry data. All authors provided critical revision of the manuscript.
Supplementary material
Supplementary material is available at IJE online.
Conflicts of interest
None declared.
Funding
This research was supported by grants from the European Union’s Horizon Europe research and innovation program under grant agreement No. 101057131, Climate Action To Advance HeaLthY Societies in Europe (CATALYSE). Data was acquired through funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement Nos. 865564 (European Research Council Consolidator Grant EARLY-ADAPT, https://www.early-adapt.eu/), 101069213 (European Research Council Proof-of-Concept HHS-EWS, https://forecaster.health/), and 101123382 (European Research Council Proof-of-Concept FORECAST-AIR). ISGlobal authors acknowledge support from the grant CEX2023-0001290-S funded by MCIN/AEI/10.13039/501100011033, and support from the Generalitat de Catalunya through the CERCA Program. S.S. has received funding from the Swiss National Science Foundation (grant number 217791). J.B. acknowledges funding from the Ministry of Science and Innovation (MCIU) under grant agreement No RYC2018-025446-I (programme Ramón y Cajal).
Data availability
Air pollution concentration data were provided by the EXPANSE project and are available upon request. Administrative birth data is owned by the Spanish National Statistics Institute and was provided by permission. Data will be shared on request to the corresponding author with permission of the Spanish National Statistics Institute.
Use of artificial intelligence (AI) tools
ChatGPT was used to proofread the manuscript for grammar and clarity. All scientific aspects of the study, including the design, data analysis, interpretation of results, and image generation, were conducted by the authors without the use of AI.
References
- 1. Matsushima M, Shimizutani S, Yamada H. Life course consequences of low birth weight: Evidence from Japan. J Jpn Int Econ 2018;50:37–47. [Google Scholar]
- 2. Hack M, Klein NK, Taylor HG, Long-term developmental outcomes of low birth weight infants. Future Child. 1995; 5:176–196. [PubMed] [Google Scholar]
- 3. Tian M, Reichetzeder C, Li J, Hocher B. Low birth weight, a risk factor for diseases in later life, is a surrogate of insulin resistance at birth. J Hypertens 2019;37:2123–34. [DOI] [PubMed] [Google Scholar]
- 4. Alexander BT, Henry Dasinger J, Intapad S. Effect of low birth weight on women’s health. Clin Ther 2014;36:1913–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Alexander BT, Dasinger JH, Intapad S. Fetal programming and cardiovascular pathology. Compr Physiol 2015;5:997–1025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Gluckman PD, Hanson MA. The consequences of being born small – an adaptive perspective. Horm Res 2006;65:5–14. [DOI] [PubMed] [Google Scholar]
- 7. Horikoshi M, Beaumont RN, Day FR et al. Genome-wide associations for birth weight and correlations with adult disease. Nature. 2016;538:248–252. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Grillo MA, Mariani G, Ferraris JR. Prematurity and low birth weight in neonates as a risk factor for obesity, hypertension, and chronic kidney disease in pediatric and adult age. Front Med (Lausanne) 2022;8:769734. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Bravo MA, Zephyr D, Fiffer MR, Miranda ML. Weekly prenatal PM2.5 and NO2 exposures in preterm, early term, and full term infants: decrements in birth weight and critical windows of susceptibility. Environ Res 2024;240:117509. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Simoncic V, Enaux C, Deguen S, Kihal-Talantikite W. Adverse birth outcomes related to NO2 and PM exposure: European systematic review and meta-analysis. Int J Environ Res Public Health 2020;17:8116. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Canto MV, Guxens M, García-Altés A et al. Air pollution and birth outcomes: health impact and economic value assessment in Spain. Int J Environ Res Public Health 2023;20:2290. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Arroyo V, Díaz J, Salvador P, Linares C. Impact of air pollution on low birth weight in Spain: an approach to a National Level Study. Environ Res 2019;171:69–79. [DOI] [PubMed] [Google Scholar]
- 13. Arroyo V, Linares C, Díaz J. Premature births in Spain: measuring the impact of air pollution using time series analyses. Sci Total Environ 2019;660:105–14. [DOI] [PubMed] [Google Scholar]
- 14. Lawn JE, Ohuma EO, Bradley E, et al. ; Subnational Vulnerable Newborn Measurement Group. Small babies, big risks: global estimates of prevalence and mortality for vulnerable newborns to accelerate change and improve counting. Lancet 2023;401:1707–19. [DOI] [PubMed] [Google Scholar]
- 15. Ballester F, Estarlich M, Iñiguez C et al. Air pollution exposure during pregnancy and reduced birth size: a prospective birth cohort study in Valencia, Spain. Environ Health 2010;9:6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Gasparrini A. Modeling exposure–lag–response associations with distributed lag non-linear models. Stat Med 2014;33:881–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Fairburn J, Schüle SA, Dreger S, Karla Hilz L, Bolte G. Social inequalities in exposure to ambient air pollution: a systematic review in the WHO European region. Int J Environ Res Public Health 2019;16:3127. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Heo S, Fong KC, Bell ML. Risk of particulate matter on birth outcomes in relation to maternal socio-economic factors: a systematic review. Environ Res Lett 2019;14:123004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Westergaard N, Gehring U, Slama R, Pedersen M. Ambient air pollution and low birth weight - are some women more vulnerable than others? Environ Int 2017;104:146–54. [DOI] [PubMed] [Google Scholar]
- 20. Torche F, Nobles J. Early-life exposures and social stratification. Ann Rev 2024;50:407–30. [Google Scholar]
- 21. Shen Y, Hoogh KD, Schmitz O et al. Europe-wide air pollution modeling from 2000 to 2019 using geographically weighted regression. Environ Int 2022;168:107485. [DOI] [PubMed] [Google Scholar]
- 22. Shen Y, Hoogh KD, Schmitz O et al. Monthly average air pollution models using geographically weighted regression in Europe from 2000 to 2019. Sci Total Environ 2024;918:170550. [DOI] [PubMed] [Google Scholar]
- 23. UNESCO Institute for Statistics. International Standard Classification of Education: ISCED 2011. Montreal, Quebec: UNESCO Institute for Statistics, 2012. [Google Scholar]
- 24. Yitshak-Sade M, Kloog I, Schwartz JD, Novack V, Erez O, Just AC. The effect of prenatal temperature and PM2.5 exposure on birthweight: weekly windows of exposure throughout the pregnancy. Environ Int 2021;155:106588. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Guxens M, Botella N, Stafoggia M et al. Particulate matter exposure during pregnancy and birth outcomes: exposure windows of susceptibility and socioeconomic inequalities. Eur J Epidemiol 2025;40:1105–21. 10.1007/s10654-025-01274-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Ni W, Xing Y, Li G et al. Windows of sensitivity for risk of adverse birth outcomes related to gestational PM2.5 exposure: evidence from a natural experiment. Environ Pollut 2024;347:123759. [DOI] [PubMed] [Google Scholar]
- 27. Hao H, Yoo SR, Strickland MJ et al. Effects of air pollution on adverse birth outcomes and pregnancy complications in the U.S. state of Kansas (2000–2015). Sci Rep 2023;13:21476. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Nyadanu SD, Dunne J, Tessema GA et al. Prenatal exposure to ambient air pollution and adverse birth outcomes: an umbrella review of 36 systematic reviews and meta-analyses. Environ Pollut 2022;306:119465. [DOI] [PubMed] [Google Scholar]
- 29. Dijkstra L, Poelman H. A harmonised Definition of Cities and Rural Areas: The New Degree of Urbanisation. European Commission DG REGIO; 2014. Report No.: WP 02/2014. https://ec.europa.eu/regional_policy/en/information/publications/working-papers/2014/a-harmonised-definition-of-cities-and-rural-areas-the-new-degree-of-urbanisation (24 February 2026, date last accessed).
- 30. Duque I, Domínguez-Berjón MF, Cebrecos A et al. Índice de privación en España por sección censal en 2011. Gac Sanit. Sociedad Española de Salud Pública y Administración Sanitaria (SESPAS) 2021;35:113–22. [DOI] [PubMed] [Google Scholar]
- 31. European Union. Directive (EU) 2024/2881. 2024/2881 Nov 20, 2024. https://eur-lex.europa.eu/eli/dir/2024/2881/oj/eng (21 January 2025, date last accessed).
- 32. WHO. WHO Global Air Quality Guidelines: Particulate Matter (PM2. 5 and PM10), Ozone, Nitrogen Dioxide, Sulfur Dioxide and Carbon Monoxide. 1st ed. Geneva: World Health Organization; 2021. [PubMed] [Google Scholar]
- 33. Ahern J, Hubbard A, Galea S. Estimating the effects of potential public health interventions on population disease burden: a step-by-step illustration of causal inference methods. Am J Epidemiol 2009;169:1140–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Snowden JM, Rose S, Mortimer KM. Implementation of G-computation on a simulated data set: demonstration of a causal inference technique. Am J Epidemiol 2011;173:731–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Welch BM, Keil AP, Buckley JP et al. Racial and ethnic disparities in phthalate exposure and preterm birth: a pooled study of sixteen U.S. cohorts. Environ Health Perspect 2023;131:127015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Wood SN. Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. J R Stat Soc Ser B Stat Methodol 2011;73:3–36. [Google Scholar]
- 37. Gasparrini A. Distributed lag linear and non-linear models in R: The Package dlnm. J Stat Soft 2011;43:1–20. [PMC free article] [PubMed] [Google Scholar]
- 38. R Core Team. R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing, 2024. https://www.R-project.org/ (27 January 2025, date last accessed).
- 39. Stieb DM, Chen L, Eshoul M, Judek S. Ambient air pollution, birth weight and preterm birth: a systematic review and meta-analysis. Environ Res 2012;117:100–11. [DOI] [PubMed] [Google Scholar]
- 40. Li C, Yang M, Zhu Z et al. Maternal exposure to air pollution and the risk of low birth weight: a meta-analysis of cohort studies. Environ Res 2020;190:109970. [DOI] [PubMed] [Google Scholar]
- 41. Nascimento F, Almeida M, Gouveia N. Individual and contextual socioeconomic status as effect modifier in the air pollution-birth outcome association. Sci Total Environ 2022;803:149790. [DOI] [PubMed] [Google Scholar]
- 42. Gray SC, Edwards SE, Schultz BD, Miranda ML. Assessing the impact of race, social factors and air pollution on birth outcomes: a population-based study. Environ Health 2014;13:4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Shmool JLC, Bobb JF, Ito K et al. Area-level socioeconomic deprivation, nitrogen dioxide exposure, and term birth weight in New York City. Environ Res 2015;142:624–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Quraishi SM, Hazlehurst MF, Loftus CT et al. Association of prenatal exposure to ambient air pollution with adverse birth outcomes and effect modification by socioeconomic factors. Environ Res 2022;212:113571. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. O’Neill MS, Bell ML, Ranjit N et al. Air Pollution and Mortality in Latin America: the role of education. Epidemiology (Fairfax) 2008;Nov19:810. [DOI] [PubMed] [Google Scholar]
- 46. Šrám RJ, Binková B, Dejmek J, Bobak M. Ambient air pollution and pregnancy outcomes: a review of the literature. Environ Health Perspect 2005;113:375–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Padula AM, Monk C, Brennan PA et al. A review of maternal prenatal exposures to environmental chemicals and psychosocial stressors—implications for research on perinatal outcomes in the ECHO program. J Perinatol. 2020;40:10–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48. Ruiz M, Goldblatt P, Morrison J et al. Mother’s education and the risk of preterm and small for gestational age birth: a DRIVERS meta-analysis of 12 European cohorts. J Epidemiol Community Health 2015;69:826–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49. Panico L, Tô M, Thévenon O. What is the link between mothers’ level of education and low birthweight? Popul Soc. Ined Éditions 2015;523:1–4. [Google Scholar]
- 50. Kramer MS, Séguin L, Lydon J, Goulet L. Socio-economic disparities in pregnancy outcome: why do the poor fare so poorly? Paediatr Perinat Epidemiol 2000;14:194–210. [DOI] [PubMed] [Google Scholar]
- 51. EU should align with WHO air quality guidelines. Lancet Public Health 2023;8:e473. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Air pollution concentration data were provided by the EXPANSE project and are available upon request. Administrative birth data is owned by the Spanish National Statistics Institute and was provided by permission. Data will be shared on request to the corresponding author with permission of the Spanish National Statistics Institute.