Identifying Critical Windows and Joint Effects of Prenatal Air Pollution and Temperature Exposure and Lung Function in Schoolchildren: Findings from a Prospective Birth Cohort Study

Abstract

Background:

Air pollution and extreme temperatures exposure during pregnancy is associated with lung function in schoolchildren.

Research Question:

What are the critical time windows during pregnancy when exposure to air pollution (PM_2.5 and NO₂) and temperature affects lung function in schoolchildren, and do these exposures interact?

Study Design and Methods:

Within the PROGRESS study, daily residential levels of pollutant/temperature exposures during pregnancy were generated from satellite-based models. Lung function was evaluated at ages 8–14 years-old, and was modeled as z-scores adjusted for age, height, and sex. We used distributed lag non-linear models to evaluate overall and sex-specific associations of exposures with lung function outcomes. Interactive effects were evaluated through the relative excess risk due to interaction and the attributable proportion.

Results:

A total of 429 mother-child dyads were included. Prenatal higher PM_2.5 exposure was associated with reduced lung function parameters, including FEV₁ z-score (weeks 1–21, cumulative change: −0.23 [95% CI: −0.39, −0.07]), FVC z-score (weeks 13–19, cumulative change: −0.04 [95% CI: −0.08, −0.00]), FEF_25–75% z-score (weeks 1–20, cumulative change: −0.20 [95% CI: −0.36, −0.04]), and FEV₁/FVC ratio (weeks 6–16, cumulative change: −0.57 [95% CI: −0.11, −0.04]). Similarly, increased NO₂ exposure was associated with reduced FEV₁ z-score (weeks 1–16, cumulative change: −0.16 [95% CI: −0.31, −0.02]), FEF_25–75% z-score (weeks 13–16, −0.02 [95% CI: −0.04, −0.00]), and FEV₁/FVC ratio (weeks 6–15, −0.48 [95% CI: −0.96, −0.01]). In contrast, both warmer (weeks 1–8) and colder temperatures (weeks 9–18) showed positive associations with FVC z-score. Stronger associations were found in females. No interactive effects of air pollution and temperature were found.

Interpretation:

Our findings emphasize detrimental effects of early-life air pollution exposure on long-term respiratory health and suggest potential sex-specific vulnerabilities, informing targeted interventions to protect child health.

The DOHaD hypothesis posits that early-life environmental exposures, particularly during pregnancy, shape long-term health and disease risk ¹. Lung development starts at 3–4 weeks post-conception and continues into early adulthood, involving complex structural and functional maturation processes ^2,3. Environmental perturbations during critical developmental windows may compromise lung growth, with implications for lifelong respiratory health ^3–5. Early lung function deficits can persist, increasing risks of chronic respiratory diseases and cardiorespiratory complications in adulthood ^6,7.

Air pollution is a key environmental factor affecting respiratory health, with longitudinal studies linking early-life exposure to asthma, wheeze, and reduced lung function growth in childhood and adolescence ^8–11. Notably, improved air quality has been associated with better lung function trajectories, highlighting the significance of modifiable environmental factors ^12,13. While previous research primarily focused on air pollution, emerging evidence suggests temperature may modify its effects via physicochemical and physiological pathways ^14,15. High temperatures can increase secondary pollutant formation, enhance bioavailability of particle-bound compounds, and alter respiratory and inflammatory responses, potentially heightening susceptibility to air pollution ¹⁶.

Our previous study examined prenatal air pollution and temperature associations with asthma and wheeze at 4–6 years using caregiver-reported outcomes ¹⁷. While informative, these questionnaire-based measures are prone to reporting bias and may fail to detect subclinical lung function deficits. Lung function measurements provide objective, quantitative assessments capable of identifying impairments preceding clinical manifestations. However, previous studies ^18–20 reported inconsistent findings and predominantly examined trimester-averaged air pollution exposure, despite potentially shorter critical windows requiring precise temporal identification. Moreover, autocorrelation across adjacent exposure periods complicates accurate sensitive window detection ²¹, and evidence regarding air pollution-temperature interactions remains limited.

This study aims to address these gaps by using distributed lag nonlinear models (DLNMs) to identify week-specific critical windows while accounting for autocorrelation, and formally assessing air pollution-temperature interactions. Given sex differences in respiratory maturation and airway structure ²², we additionally examine sex-specific effects. Using lung function data from preadolescents and adolescents aged 8–14 in the PROGRESS (Programming Research in Obesity, Growth, Environment, and Social Stressors) birth cohort, this study expands beyond caregiver-reported symptoms to provide comprehensive analysis of early-life environmental determinants of respiratory health in a developing country context.

Study Design and Methods

Study population

PROGRESS, an ongoing prospective birth cohort study, was established in Mexico City in 2007. Pregnant women receiving prenatal care through the Mexican Social Security System (Instituto Mexicano del Seguro Social) were recruited from July 2007 to February 2011 ²³. Eligibility required gestational age <20 weeks, age ≥18 years, completed primary education, intent to reside in Mexico City for the following three years, telephone access, and no history of heart/kidney disease, daily alcohol use, or anti-epileptic/steroid medication use.

The study protocol received approval from the ethics committees of both the National Institute of Public Health in Mexico, the National Institute of Perinatology in Mexico City, and the Icahn School of Medicine at Mount Sinai. Participants provided written consent during their research visits, with children offering assent at age seven. Of 948 mothers with live births, 429 mother-child dyads with complete lung function data, prenatal exposure information (PM_2.5, NO₂, and temperature), and covariate data were included in the final analysis. The detailed flow chart of participant selection is shown in e-Figure 1. Included and excluded subsamples showed no significant differences in baseline characteristics (e-Table 1).

Exposure assessment

Prenatal PM_2.5 and NO₂ exposures were estimated using validated hybrid spatiotemporal resolved models ^24,25. PM_2.5 models combined extreme gradient boosting and inverse-distance weighting, integrating aerosol optical depth, meteorology, and land-use data for daily residential predictions. NO₂ exposure used an ensemble approach ²⁴ that synthesized satellite and ground-based variables with geospatial factors via geostatistical and machine learning methods (XGBoost, RF). Daily temperature was modeled using land-use regression, calibrating satellite surface temperatures with monitoring station data ²⁶. Exposures were calculated at 1-km resolution for each participant’s residence, accounting for address changes. Daily prenatal exposures were averaged weekly to identify critical windows, with additional calculations for trimester-specific periods.

Outcome measurement

Lung function was measured by trained professionals per standardized protocols. Height (stadiometer, 0.1 cm precision) and weight (electronic scale, 0.1 kg precision) were recorded. Lung function was assessed using a portable MedGraphics^™ spirometer with heated screen pneumotachograph, calibrated before each session. All procedures followed ATS guidelines for acceptability and reproducibility ²⁷. Testing occurred at participants’ homes or the study clinic, with participants rescheduled if experiencing recent respiratory symptoms. To ensure accuracy, participants withheld medications before testing: short-acting bronchodilators and anticholinergics (4 hours), long-acting beta-agonists (12 hours), and extended-release theophylline preparations (24 hours). Each participant performed 3–8 maneuvers, recording FEV₁, FVC, FEV₁/FVC ratio, and FEF_25–75%. A pediatric pulmonologist reviewed all tests for acceptability and reproducibility. FEV₁, FVC, and FEF_25–75% raw values were regressed on age, sex, and height; residuals were converted to z-scores using the standard deviation of residuals. Z-scores were used instead of raw values or percent predicted values to provide population-specific standardization that accounts for developmental factors and enables valid comparison across age groups within our study cohort. The FEV₁/FVC ratio was calculated using the raw values.

Covariates

Potential confounding variables were selected a prior based on existing literature ^8,28 a DAG (e-Figure 2). Analyses were adjusted for maternal age, pre-pregnancy BMI (used pre-gestational weight estimated via linear mixed-effects models²⁹), parity, educational attainment at enrollment, ETS exposure, child sex, and seasonality. Seasonality was adjusted using sine and cosine functions of time of year for DLNMs ³⁰ and season of conception for average exposure models ³¹. Sensitivity analyses for average exposure models were additionally adjusted for SES (derived from a six-level index developed by the Asociación Mexicana de Agencias de Investigación de Mercados y Opinión Pública ³², consolidated from six to three levels) and maternal asthma history. The details of covariates coding were shown in e-Appendix 1.

Statistical analysis

Maternal and child characteristics were summarized using descriptive statistics, with continuous variables reported as mean (SD) for normal distributions or median (IQR) for skewed and categorical variables as count and percentage (n, %).

Associations between prenatal PM_2.5, NO₂, and ambient temperature exposure and childhood lung function were assessed using DLNMs for infants born at ≥37 weeks gestation ³³. Models used linear exposure-response functions for air pollutants and natural cubic splines for temperature, with lag-response relationships modeled using natural cubic splines. Degrees of freedom (2–6) were selected by minimizing AIC (e-Table 2). Models adjusted for maternal factors, ETS, child sex, and seasonality. Effect estimates were reported per standardized increments (5 μg/m³ for PM_2.5, 5 ppb for NO₂) and for warmer (19°C, 95th percentile) and colder temperatures (11°C, 5th percentile) relative to the median (15°C), with 95% CIs to identify critical exposure windows. Cumulative effects were calculated by summing weekly coefficients over critical exposure windows (95% CIs excluding zero). In consideration of the sexually dimorphic differences in respiratory development, we then examined effect modification by applying DLNMs on the data stratified by child sex.

Multivariable linear regression models were used to compare with DLNMs and assessing associations of exposures during entire pregnancy and individual trimesters with lung function, overall and by sex. Trimester exposures were simultaneously incorporated into a single model to minimize bias from separate adjustments ²¹. Changes in lung function were reported per 5 μg/m³ increase in PM_2.5, 5 ppb increase in NO₂, and 1°C increase in temperature. Sex-specific effects were examined through exposure-sex interactions, with P<0.1 considered evidence of differential associations by sex. Results were presented stratified by sex regardless of interaction significance.

We evaluated potential additive interactions between air pollution and temperature on odds of low lung function, defined as z-score less than −1.64 for FEV₁, FVC, or FEF_25–75%, or FEV₁/FVC ratio less than 80% ³⁴. Exposures were dichotomized at median values. We calculated OR for high air pollution only (OR₁₀), high temperature only (OR₀₁), and co-exposure to both (OR₁₁), with low exposures as reference. Interaction was assessed using RERI and AP, representing additional risk and proportion of total effect attributed to interaction, respectively ³⁵. Bootstrap percentile method with 1000 replicates generated 95% CIs for both measures. Interactions were considered statistically significant if CIs excluded 0. Detailed methodology for interaction assessment, including formulas and interpretation criteria, is provided in e-Appendix 2.

Sensitivity analyses included: (1) additional adjustment for SES and maternal asthma history; (2) multi-pollutant models simultaneously adjusting for PM_2.5, NO₂, and temperature within the same exposure window; and (3) further adjustment for postnatal exposures using eight-year averages of PM_2.5, NO₂, and temperature.

All statistical analyses were performed using R software (version 4.3.3) and the R package “dlnm” (version 2.4.7).

Results

Characteristics of study participants

The baseline characteristics are summarized in Table 1. Mothers averaged 27.5 years (SD: 5.52) at enrollment with pre-pregnancy BMI of 26.51 kg/m² (SD: 4.19). Educational attainment included: less than high school (40.3%), completed high school (37.1%), and post-high school (22.6%). Most mothers (70.2%) reported no ETS exposure, were primiparous (61.3%), and only 0.5% had asthma history. Nearly half of conceptions (46.2%) occurred during the rainy season, with most participants from lower (53.1%) or medium (35.7%) socioeconomic backgrounds.

Table 1.

PROGRESS participant characteristics (N = 429).

Characteristics	Total sample	Males (n = 232)	Females (n = 197)
Maternal age (years), mean (SD)	27.5 (5.52)	27.7 (5.37)	27.2 (5.70)
Maternal education, n (%)
< High school	173 (40.3)	101 (43.5)	72 (36.5)
High school	159 (37.1)	78 (33.6)	81 (41.1)
> High school	97 (22.6)	53 (22.8)	44 (22.3)
Maternal ever asthma, n (%)
Yes	2 (0.5)	1 (0.4)	1 (0.5)
No	427 (99.5)	231 (99.6)	196 (99.5)
Season of conception
Dry cold	147 (34.3)	82 (35.3)	65 (33.0)
Dry warm	84 (19.6)	46 (19.8)	38 (19.3)
Rainy	198 (46.2)	104 (44.8)	94 (47.7)
SES
Lower	228 (53.1)	123 (53.0)	105 (53.3)
Medium	153 (35.7)	84 (36.2)	69 (35.0)
Higher	48 (11.2)	25 (10.8)	23 (11.7)
ETS, n (%)
Yes	128 (29.8)	73 (31.3)	55 (27.9)
No	301 (70.2)	159 (68.5)	142 (72.1)
Maternal pre-pregnancy BMI, mean (SD)	26.51 (4.19)	26.76 (3.96)	26.22 (4.43)
Parity, n (%)
Primiparous	263 (61.3)	149 (64.2)	114 (57.9)
Multiparous	166 (38.7)	83 (35.8)	83 (42.1)
Age at spirometry (years), mean (SD)	10.7 (1.87)	10.8 (1.85)	10.5 (1.89)
Height at spirometry (cm), mean (SD)	144.6 (12.82)	146.0 (13.55)	143.0 (11.72)
Lung function parameters, mean (SD)
FEV1 z-score a	−0.03 (0.99)	−0.02 (1.03)	−0.04 (0.94)
FVC z-score a	−0.02 (0.98)	−0.02 (0.99)	−0.03 (0.97)
FEF25–75% z-score a	−0.02 (0.99)	−0.01 (1.03)	−0.04 (0.95)
FEV1/FVC	86.79 (6.33)	85.74 (6.12)	88.02 (6.36)

Abbreviations: SD, standard deviation; SES, socioeconomic status, ETS, environmental tobacco smoke; BMI, body mass index. FVC, forced vital capacity; FEV₁, forced expiratory volume in 1 s; FEF_{25–75 %}, forced expiratory flow between 25 % and 75 %.

^aAdjusted for age, sex, and height.

The study included 429 children (232 males [54.1%], 197 females [45.9%]) who averaged 10.7 years (SD: 1.87) with mean height of 144.6 cm (SD: 12.82) at spirometry visit. Mean z-scores were: FEV₁ −0.03 (SD: 0.99), FVC −0.02 (SD: 0.98), FEF_25–75% −0.02 (SD: 0.99), and FEV₁/FVC ratio 86.79 (SD: 6.33). Average exposures for different exposure periods were 22.4 to 22.9 μg/m³ for PM_2.5, 32.7 to 33.3 ppb for NO₂, and 14.9 to 15.1°C for temperature (Table 2).

Table 2.

Distribution of ambient air pollution and temperature.

Exposure and window	Mean	SD	Min.	Q1	Median	Q3	Max.
PM2.5 (μg/m3)
Whole pregnancy	22.7	3	16.4	20	22.9	25	30.3
First trimester	22.9	4.8	12.6	18.9	21.8	26.9	33.5
Second trimester	22.4	5.1	11.5	18	21.3	26.8	32.8
Third trimester	22.9	5.9	11.9	17.7	21.8	28.3	34.4
NO2 (ppb)
Whole pregnancy	33	5.6	17.8	28.8	31.9	36.9	47.6
First trimester	33.2	7.8	16.6	27.3	31.3	37.7	61.9
Second trimester	32.7	7.1	19.9	27	31.6	36.9	56.5
Third trimester	33.3	7.7	16.4	26.9	32.8	37.8	63.7
Temperature (°C)
Whole pregnancy	15	1.4	11.3	14.1	14.9	16	18.1
First trimester	15.1	2	9.1	13.8	15.2	16.5	19.3
Second trimester	15.1	2.1	10.1	13.6	15	16.7	19.2
Third trimester	14.9	2.2	9.3	13.3	14.8	16.6	20.6

Abbreviations: SD, standard deviation; NO₂, nitrogen dioxide; PM_2.5, fine particulate matter.

Critical windows of exposure on lung function

Associations between weekly air pollution exposure, temperature, and lung function parameters are presented in Figures 1–2. For prenatal PM_2.5 exposure, critical windows were identified at gestational weeks 1–21 for FEV₁ (Figure 1A), weeks 13–19 for FVC (Figure 1B), weeks 1–20 for FEF_25–75% (Figure 1C), and weeks 6–16 for FEV₁/FVC ratio (Figure 1D). PM_2.5 exposure during these windows was associated with reduced lung function, with estimated decrements of −0.23 (95% CI: −0.39, −0.07) for FEV₁ z-score, −0.04 (95% CI: −0.08, −0.00) for FVC z-score, −0.20 (95% CI: −0.36, −0.04) for FEF_25–75% z-score, and −0.57 (95% CI: −1.11, −0.04) for FEV₁/FVC ratio. For prenatal NO₂ exposure, critical windows were observed at weeks 1–16 for FEV₁ (Figure 1E), weeks 13–16 for FEF_25–75% (Figure 1G), and weeks 6–15 for FEV₁/FVC ratio (Figure 1H). These NO₂ exposures were associated with lung function decrements of −0.16 (95% CI: −0.31, −0.02) for FEV₁ z-score, −0.02 (95% CI: −0.04, −0.00) for FEF_25–75% z-score, and −0.48 (95% CI: −0.96, −0.01) for FEV₁/FVC ratio. For temperature exposure (Figure 2), warmer temperatures during gestational weeks 1–8 were associated with an increase in FVC z-score (0.57; 95% CI: 0.03, 1.11; Figure 2B), while colder temperatures during weeks 9–18 were similarly associated with an elevated FVC z-score (0.51; 95% CI: 0.04, 0.97; Figure 2F).

Figure 1.

Effect estimates of lung function parameters in association with weekly-specific PM_2.5 and NO₂ exposure during 1–37 weeks of pregnancy. Critical windows are highlighted.

Figure 2.

Effect estimates of lung function parameters in association with weekly-specific warmer and colder temperatures exposure during 1–37 weeks of pregnancy. Critical windows are highlighted.

Sex-stratified analysis (Figures 3–6) revealed differential effects. In males, PM_2.5 exposure was associated with alterations in FEV₁, FVC, and FEF_25–75% z-scores. In females, PM_2.5 exposure was linked to changes in FEV₁ and FEV₁/FVC ratio, with the latter showing more extensive critical exposure periods. NO₂ exposure was associated with reduced FEV₁ z-score and FEV₁/FVC ratio exclusively in females. Temperature exposure showed no significant associations with all assessed lung function parameters in either sex.

Figure 3.

Sex-stratified effect estimates of lung function parameters in association with weekly-specific PM_2.5 exposure during 1–37 weeks of pregnancy.

Figure 6.

Sex-stratified effect estimates of lung function parameters in association with weekly-specific colder temperature exposure during 1–37 weeks of pregnancy.

Associations of averaged exposure and lung function

Prenatal exposure to PM_2.5 across the entire pregnancy and during the first trimester was associated with reduced FEV₁ z-score (β = −0.22, 95% CI: −0.39, −0.05; β = −0.17, 95% CI: −0.27, −0.06). Similarly, PM_2.5 exposure during the first trimester and throughout pregnancy was linked to decreased FVC z-score (β = −0.14, 95% CI: −0.25, −0.04) and FEF_25–75% z-score (β = −0.17, 95% CI: −0.35, −0.00) respectively. In contrast, no associations were observed between NO₂ or temperature exposures and lung function parameters (e-Figure 3).

Significant interactions were identified between air pollutant exposures and child sex. PM_2.5 exposure during the second trimester showed a significant interaction with sex for FVC z-score (P = 0.03). Similarly, NO₂ exposure during second (P = 0.03) and third (P = 0.02) trimesters exhibited significant sex interactions. Sex-stratified analyses revealed that whole-pregnancy PM_2.5 exposure was associated with reduced FEV₁ (β = −0.24, 95% CI: −0.48, −0.01) and FEF_25–75% (β = −0.25, 95% CI: −0.49, −0.01) z-scores in females, whereas first-trimester PM_2.5 exposure was linked to decreased FEV₁ (β = −0.20, 95% CI: −0.36, −0.04) and FVC (β = −0.22, 95% CI: −0.37, −0.06) z-scores in males (e-Figure 4). First-trimester NO₂ exposure was associated with reduced FVC z-score (β = −0.12, 95% CI: −0.22, −0.01) exclusively in males (e-Figure 5). No sex-specific effects were observed for temperature exposure (e-Figure 6).

Additive interaction between exposures and lung function

No significant additive interactions were observed between prenatal temperature and air pollutants exposure for any lung function parameters across whole-pregnancy or trimester-specific exposure windows (e-Tables 3–6). For PM_2.5-temperature interactions, RERI point estimates ranged from −6.22 to 5.65, while NO₂-temperature interactions showed RERI values from −5.72 to 1.21, with all 95% CIs including zero. The AP point estimates showed similar patterns: −2.58 to 0.87 for PM_2.5-temperature and −12.98 to 0.79 for NO₂-temperature interactions, with CIs consistently encompassing zero, confirming the absence of significant synergistic effects.

Sensitivity analysis

In sensitivity analyses, adjusting for additional covariates or multi-exposure models produced results largely consistent with primary analyses (e-Tables 7–10). Further adjustment of prenatal exposure models for postnatal exposure showed first-trimester NO₂ exposure linked to reduced FEV₁ z-score, and whole-pregnancy NO₂ and PM_2.5 exposures, plus first-trimester NO₂, associated with decreased FVC z-score (e-Figure 7). Previously identified critical windows for PM_2.5 with FEV₁/FVC ratio, and NO₂ with FEF_25–75% z-score and FEV₁/FVC ratio, were no longer observed (e-Figure 8), while other pollutant–lung function critical windows remained largely unchanged. New critical windows were identified for warmer temperature exposure and FEV₁ z-score (e-Figure 9).

Discussion

Our findings identified early to mid-gestation as critical exposure windows, where prenatal air pollution was linked to reduced lung function, while both warmer and colder temperatures were associated with improved lung function. Sex-stratified analyses revealed more pronounced effects in females, though we observed no combined effects from air pollution and temperature exposure.

This study contributes to the growing evidence on critical exposure windows for prenatal environmental factors affecting lung function. Early-gestation NO₂ exposure was associated with reduced FEV₁ but not FVC, consistent with a previous review ³⁶ indicating that NO₂ effects are more pronounced on airway properties than lung volumes. Epidemiological findings remain inconsistent. Morales et al. (2014) ¹⁸ found only second-trimester NO₂ exposure associated with lower FEV₁ in preschoolers, while our trimester-averaged analyses showed no associations. He et al. (2019) ¹⁹ reported in utero NO₂ exposure linked to reduced lung function at ~17.5 years, consistent with our findings. In contrast, Stapleton et al. (2022) ²⁰ found no significant associations between prenatal NO₂ and lung function at ages 4–11.

Most previous studies calculated exposure by averaging measurements across trimesters or full gestation, which may not accurately capture true vulnerability windows and may introduce seasonality bias ²¹. Our study used continuous exposure assessment with flexible DLNM models, enabling concurrent evaluation of multiple short exposure intervals throughout gestation. A recent study ³⁷ using distributed lag models found PM_2.5 exposure during pregnancy and early childhood linked to reduced lung function, highlighting late pregnancy as potentially critical. Our results suggest adverse effects beginning in early to mid-gestation, spanning crucial developmental stages including embryonic organogenesis and the canalicular stage when capillaries, terminal saccules, and alveolar epithelium develop ³⁸.

Fewer studies have investigated prenatal temperature exposure and lung function in schoolchildren. Guibert et al. (2023) ³⁹ found long-term exposure to both cold and heat from mid-pregnancy to one month postpartum associated with decreased lung volumes in neonates, with stronger effects in females. Contrary to studies linking warmer temperatures to lower lung function in children with asthma ⁴⁰ or the general population ⁴¹, we observed prenatal exposure to both warmer and colder temperatures associated with increased lung function, possibly attributed to Mexico City’s comfortable temperature range.

Previous studies suggested potential interactions between elevated temperatures and air pollution in reducing lung function ^42,43, though findings have been inconsistent ⁴¹. We observed no evidence that high temperatures amplified the negative associations of air pollutants with lung function, consistent with Rice et al. (2019) ⁴¹. This absence may be attributed to Mexico City’s moderate climate (9.1–20.6°C), which may not be sufficiently extreme to induce synergistic effects with air pollutants.

Our results demonstrated sex differences in associations, with females appearing more sensitive to air pollution, in alignment with several studies ^44,45, though findings remain inconsistent across the literature ⁴⁶. These inconsistencies may reflect differences in exposure assessment methods, age groups studied, analytical approaches, and population characteristics including genetic and socioeconomic factors.

Prenatal air pollution may affect lung development through multiple mechanisms ⁴ inducing physiological alterations in pregnant women (oxygen deprivation, oxidative stress, inflammation); direct translocation through the placenta; causing DNA damage and epigenetic alterations; or through growth-regulating factors. Temperature effects may occur because pregnancy compromises thermoregulation ⁴⁷, and animal studies ^48,49 show heat/cold stress can impair placental development and reduce uterine blood flow. The observed sex-specific vulnerability may reflect several biological mechanisms ⁵⁰, including accelerated lung maturation in female fetuses with earlier surfactant production, sex hormones influences on lung development and inflammatory responses, and emerging evidence of sex-specific epigenetic responses to air pollution exposure.

This study has notable strengths, including well-established prospective birth cohort with comprehensive documentation, highly resolved exposure measures using DLNMs for precise identification of crucial exposure periods, investigation of joint effects between air pollution and temperature, and objective lung function measurements that minimizing reporting and recall biases.

Several limitations warrant consideration. First, our exclusion of gestational age and birth weight as potential mediators may limit clinical interpretability and introduce residual confounding, while reducing comparability with studies adjusting for birth characteristics. Second, ETS exposure was only assessed during the second and third trimesters, potentially missing relevant first trimester household smoking that could confound associations during early critical windows, though smoking prevalence was very low in our population. Third, exposure misclassification is possible since our measures only capture air pollution and temperature exposures around maternal residences, excluding other sources like maternal workplaces. This non-differential misclassification would likely bias our effect estimates toward the null, suggesting our findings may represent conservative estimates of the true associations. Fourth, we lacked data on other environmental risk factors, such as noise and greenspaces, which have been associated with child’s lung function and thus act as potential residual confounders. Finally, the limited number of mothers with asthma history prevented investigation of its potential modifying effect.

Interpretation

In conclusion, early to mid-gestation PM_2.5 and NO₂ exposure was associated with reduced lung function in schoolchildren, while prenatal exposure to warmer and colder temperatures was associated with improved lung function. We found that these effects were more pronounced in females, with no significant additive interactions between air pollutants and temperature. These findings underscore the need for pollution mitigation and urban planning to protect children’s respiratory health. Longitudinal studies tracking lung function into adulthood are vital to elucidate the lifelong impacts of early-life environmental exposures.

Figure 4.

Sex-stratified effect estimates of lung function parameters in association with weekly-specific NO₂ exposure during 1–37 weeks of pregnancy.

Figure 5.

Sex-stratified effect estimates of lung function parameters in association with weekly-specific warmer temperature exposure during 1–37 weeks of pregnancy.

Take-home Points.

Study question:

What are the critical prenatal periods when air pollution and temperature exposures affect lung function in Mexican schoolchildren, and do these exposures exhibit interactions?

Results:

In a birth cohort of 429 mother-child pairs, early to mid-pregnancy exposure to PM_2.5 and NO₂ was associated with reduced lung function in children aged 8–14 years, particularly in females. Temperature exposure was positively associated with FVC, with no observed interaction with air pollution effects.

Interpretation:

This study identifies specific prenatal windows of susceptibility to environmental exposures that impact children’s lung function, providing evidence to guide timing of interventions to protect respiratory health.

Abbreviations:

AIC

Akaike information criterion

AP

attributable proportion due to interaction

ATS

American Thoracic Society

BMI

body mass index

CI

confidence interval

DAG

directed acyclic graph

DLNM

distributed lag nonlinear model

DOHaD

developmental origins of health and disease

ETS

environmental tobacco smoke

FEF_25–75%

forced expiratory flow at 25–75% of the vital capacity

FEV₁

forced expiratory volume in 1 second

FVC

forced vital capacity

IQR

interquartile range

NO₂

nitrogen dioxide

OR

odds ratio

PM_2.5

fine particulate matter

RERI

relative excess risk due to interaction

RF

Random Forest

SD

standard deviation

SES

socioeconomic status

XGBoost

eXtreme Gradient Boosting

Data sharing statement:

For data protection reasons, the datasets analyzed during the current study cannot be made available publicly. The datasets are available to interested researchers from the corresponding author on reasonable request, provided the release is consistent with the consent given by the study participants. Ethical approval may be obtained for the release and a data transfer agreement must be accepted.