Association of fine particulate matter exposure during pregnancy and stillbirth rates in Pakistan: a cross-sectional study Authors: Waqas Hameed, Bilal Ahmed Usmani, Saqib Ur Rehman, Muhammad Ahmed, Asad Allana, Anmol Minaz, Zafar Ahmed, Zafar Fatmi Published: 2025 ABSTRACT Abstract AbstractObjectivesThis study assessed the association between fine particulate matter (PM2.5) exposure during pregnancy and stillbirth in Pakistan. We hypothesised that higher PM2.5 exposure is linked to increased stillbirth risk.DesignA cross-sectional study using secondary data from the 2017 to 2018 Pakistan Demographic and Health Survey (PDHS), combined with satellite-derived PM2.5 exposure data.SettingThe study covered urban and rural areas across Pakistan, including all four provinces (Punjab, Sindh, Khyber Pakhtunkhwa and Baluchistan) and administrative regions (Gilgit-Baltistan (GB), Islamabad, Federally Administered Tribal Areas and Azad Jammu Kashmir).ParticipantsThe study included 9172 married women aged 15–49 with at least one birth in the past 5 years. Women with incomplete pregnancy outcome data were excluded.Exposure assessmentPM2.5 exposure was estimated using satellite data, matching PDHS clusters with the nearest air quality point via MATLAB. Monthly average exposure was categorised into quartiles.Primary outcomeStillbirth, defined as pregnancy loss at ≥28 weeks gestation.ResultsMultivariable logistic regression was used to assess the association between PM2.5 and stillbirth, adjusting for maternal age, gravidity, wealth index, birth interval, previous adverse pregnancy outcome and region of residence. The stillbirth rate in Pakistan for the most recent pregnancy was 17.0 (14.5–19.9) per 1000 births, with highest rates (28.9) in Baluchistan province. The mean level of PM2.5 exposure in Pakistan was 53.96 (SD 20.42; range 5.9–209.4) µg/m3. PM2.5 exposure was higher for urban (56.43) than rural (51.87) pregnancies, highest in Sindh (78.06) and lowest in GB (13.41) provinces. For every 1 µg/m3 average increase in PM2.5 during the pregnancy period, there was approximately 1% increase in stillbirth.ConclusionsIncreased PM2.5 exposure was strongly associated with stillbirth risk. This underscores the need for targeted public health interventions, such as government regulations, emission controls and clean energy initiatives to protect pregnant women in high-risk areas.Keywords: Community child health, Pregnant Women, PUBLIC HEALTH STRENGTHS AND LIMITATIONS OF THIS STUDY The study uses a large, nationally representative dataset from the Pakistan Demographic and Health Survey (PDHS), increasing generalisability to the Pakistani population. The study uses a large, nationally representative dataset from the Pakistan Demographic and Health Survey (PDHS), increasing generalisability to the Pakistani population. The study accounts for spatial variations in air pollution exposure by linking PDHS clusters to the nearest air quality point using MATLAB. The study accounts for spatial variations in air pollution exposure by linking PDHS clusters to the nearest air quality point using MATLAB. Adjustments for key confounders, such as maternal age, socioeconomic status and urban–rural differences, strengthen the validity of the findings. Adjustments for key confounders, such as maternal age, socioeconomic status and urban–rural differences, strengthen the validity of the findings. Particulate matter (PM2.5) exposure was estimated at the cluster level rather than individual level, which may introduce exposure misclassification. Particulate matter (PM2.5) exposure was estimated at the cluster level rather than individual level, which may introduce exposure misclassification. The study does not account for maternal health conditions (eg, hypertension and diabetes) that could independently contribute to stillbirth risk. The study does not account for maternal health conditions (eg, hypertension and diabetes) that could independently contribute to stillbirth risk. Introduction Stillbirth is an adverse pregnancy outcome, with one stillbirth occurring every 16 s globally.1 Stillbirth frequently has lasting negative impacts on parents’ physical, psychological, social and financial well-being. It affects women and their family dynamics, influences relationships and can impact future pregnancies and parenting.2 Annually, 2 million deaths occur due to stillbirth. Of these, 98% occur in low- and middle-income countries (LMICs).3 Stillbirth is a pregnancy loss that occurs≥28 weeks of gestation and before the birth of the baby.4 In 2015, Pakistan had the highest stillbirth rate in the world (43.1 stillbirths per 1000 total births) compared with a global estimate of 18.4.5 Stillbirths are overlooked as a critical issue in various countries, including Pakistan. Understanding all major risk factors for stillbirth is crucial for planning effective interventions. Air pollution, particularly exposure to fine particulate matter (PM2.5) and other pollutants, has been associated with various health problems, including adverse pregnancy outcomes.6 PM2.5 exposure occurs from a diverse range of manmade and naturally occurring materials, including accumulated heavy metals and harmful organic pollutants.7 8 PM2.5 can enter blood circulation through the lungs due to its small size and be transported throughout the body.9 Pakistan’s air quality, specifically PM2.5 concentrations, has consistently exceeded WHO interim targets (above 35 µg/m3) due to inadequate monitoring and political indifference, posing severe health risks to the population.10 Thus, it is imperative to assess the risk of air pollution exposure on pregnancy outcomes. Several risk factors for stillbirths are known, including advanced maternal age (≥35 years) and adolescent (≤19 years) pregnancies, malnutrition among women, history of previous pregnancy losses and complicated pregnancies.11 Low socioeconomic status of women and contextual health system factors also have an association with high stillbirth.12,14 Furthermore, household pollution from solid fuel for cooking and heating is a potential risk factor for stillbirth in LMICs.15 16 According to the current mechanistic frameworks, exposure to fine particulate matter (PM₂.₅) can trigger a cascade of biological responses, including oxidative stress, systemic and placental inflammation, endothelial dysfunction, altered coagulation and haemodynamic disturbances that collectively impair uteroplacental blood flow and oxygen exchange.17 Oxidative stress induced by the reactive metal and organic components of PM₂.₅ can damage placental DNA and lipids, while inflammatory cytokine release (eg, Interleukin-6 (IL-6) and Tumor Necrosis Factor-alpha (TNF-α)) disrupts trophoblastic function and vascular remodelling. In parallel, PM₂.₅ exposure may enhance maternal blood coagulability and cause endothelial injury, reducing placental perfusion and nutrient delivery to the fetus. These processes lead to placental hypoxia, growth restriction and potential fetal demise.17 18 A few studies have assessed an association between a mother’s pregnancy exposure to PM2.5 and stillbirth.18 19 Zang et al reported an increased risk of stillbirth (relative risk (RR)=1.14, 1.11, 1.15 and 1.14) for every 10 µg/m3 rise in PM2.5 during the overall pregnancy duration and across the first, second and third trimesters, respectively.19 Similarly, Yang et al also reported a 5% higher risk of stillbirth with every 10 µg/m3 increase in PM2.5.18 A case–control study conducted among South Asian women, including Pakistani women, with live and stillbirth from same women as cases and controls, also reported a higher risk of stillbirth with exposure to PM2.5.20 Air pollution is an escalating environmental and public health concern, particularly in LMICs like Pakistan, where rapid industrialisation, urbanisation and inadequate pollution control measures have led to persistently high levels of fine particulate matter (PM2.5). Although extensive research has linked PM2.5 exposure to various adverse health outcomes, including respiratory and cardiovascular diseases, its impact on pregnancy outcomes remains underexplored in the Pakistani context. Stillbirth is a significant yet often overlooked reproductive health issue, with Pakistan recording one of the highest stillbirth rates globally. It is crucial to assess its potential role as a risk factor for stillbirth. While studies from other regions have investigated the association between PM2.5 exposure and stillbirth, the data from South Asia, where air pollution levels rank among the highest in the world, remain limited. This evidence gap is partly due to limited availability of integrated environmental and health datasets, weak air quality monitoring systems and a lack of population-based studies in the region. As a result, South Asia, despite its disproportionate exposure burden, has been under-represented in the global discourse on air pollution and adverse pregnancy outcomes. Given Pakistan’s dual burden of extreme PM2.5 exposure and persistently high stillbirth rates, it presents a critical setting to examine this association. To our knowledge, this is among the first studies in Pakistan to link nationally representative demographic and health survey data with high-resolution satellite-derived PM2.5 estimates for the pregnancy period. Building on this, the current study aimed to investigate the association between exposure to PM2.5 during pregnancy and stillbirth in Pakistan. Findings from this study can contribute to context-specific evidence to inform maternal and environmental health policy in similar high-exposure, low-resource settings. The hypothesis is that exposure to PM2.5 during pregnancy increases the risk of stillbirth. Methods Study design We analysed cross-sectional data from the Pakistan Demographic Health Survey (PDHS 2017–2018), which was retrospectively linked with monthly satellite-derived PM2.5 data from 2012 to 2018. This approach enabled us to assess average PM2.5 exposure during pregnancy and its association with stillbirth. The study provides a snapshot of this relationship using nationally representative health data and environmental exposure estimates aligned with each respondent’s pregnancy period. Patient and public involvement None. Patients and the public were not involved in the design, conduct, reporting or dissemination plans of this research. Study setting, population and data This study was conducted using data from Pakistan, a South Asian country with a population of over 255 million. Pakistan is divided into four provinces (Punjab, Sindh, Khyber Pakhtunkhwa (KPK) and Baluchistan) and additional administrative areas, including Gilgit-Baltistan (GB), Islamabad Capital Territory (ICT), Federally Administered Tribal Areas (FATA) and Azad Jammu Kashmir (AJK). The nationally representative PDHS 2017–2018 survey data were used. This was the fourth round of survey conducted as a part of the global Demographic and Health Survey (DHS) programme under the National Institute of Population Studies and funded by the US Agency for International Development.21 Standard DHS employed trained fieldwork staff who interviewed members of eligible households, including married women of reproductive age (MWRA, 15–49 years), ensuring high-quality control throughout the process. The survey used a two-stage stratified cluster sampling strategy to select a nationally representative sample of 16 240 households. The total population of Pakistan has been divided into primary sampling units (PSUs), which are clusters of 250–300 households. Using a population proportional to size sampling scheme, 580 PSUs were selected in the first stage. In the second stage, 28 households were randomly selected from each PSU. Among 16 240 households who gave consent for the survey, we restricted our analysis to MWRA who gave birth to at least one child in the recent 5 years preceding the survey (n=9712). The data are also publicly available. Outcome variable The PDHS survey used four types of questionnaires: household, women, men and community-level questionnaire. We used the women’s questionnaire to extract our relevant variables for the study. Using the variables from individual record for women, a composite variable for pregnancy outcomes was generated, which included live birth, stillbirth, miscarriage and abortion. Subsequently, for our outcome variable of stillbirth, a separate binary variable was created as stillbirth (present or absent). The calendar data were transformed to extract the information for most recent stillbirth using pregnancy loss, corresponding time and gestational length, total number of births and date of childbirth. Stillbirth was classified as having any pregnancy loss from 28 weeks of gestation until delivery. For women with multiple pregnancies, only the most recent to the survey date was selected to reduce recall bias. In PDHS 2017–2018, the geographical locations of the surveyed clusters were also recorded using global positioning system (GPS) devices, which are publicly available (DHS website). Exposure variable: PM2.5 data The PM2.5 dataset used in our study provides high-resolution global coverage at 1 km × 1 km. The data are derived from satellite aerosol optical depth measurements (MODIS (Moderate Resolution Imaging Spectroradiometer), MISR (Multi-angle Imaging SpectroRadiometer) and SeaWiFS) combined with the GEOS-Chem chemical transport model and subsequently calibrated against ground-based PM2.5 observations using geographically weighted regression (GWR). This approach ensures both spatial completeness and accuracy. Satellite retrievals provide near-global coverage, including regions without dense ground monitoring networks, while the chemical transport model fills gaps where satellite data are unavailable due to cloud cover or other observational limitations. The GWR calibration against ground stations further corrects for regional biases and enhances representativeness. As a result, the dataset offers continuous spatial coverage with minimal missing values, capturing both urban and rural areas worldwide. While minor uncertainties remain particularly in regions with sparse ground observations, these are mitigated by the combination of multiple satellite instruments, chemical transport modelling and statistical calibration. Overall, the dataset provides a comprehensive and reliable basis for high-resolution exposure assessment in our study. PM₂.₅ exposure from 2012 to 2018 was retrospectively matched to each respondent’s pregnancy period. We used each respondent’s reported date of childbirth and gestational age (in months) from the PDHS to retrospectively estimate the start and end months of their pregnancy and assigned exposure to PM2.5 using respective monthly averages. Linking the PM2.5 data with the PDHS data—exposure assignment PDHS 2017–2018 provides cluster-level geographical location (latitude and longitude) using GPS devices, allowing us to link the satellite-derived PM2.5 data to PDHS clusters. The 1 km × 1 km spatial resolution of the PM2.5 data was used to approximate ambient PM2.5 exposure values for each PDHS cluster. We did an approximation analysis to locate the nearest available air quality data point from available PM2.5 data adjacent to the respective clusters using MATLAB software. In the first step, the array, representing PM2.5 values within the 1-km grid cells surrounding each cluster, was analysed by calculating the maximum and minimum values to understand the range of exposures in the vicinity. Second, we calculated the absolute difference between the minimum and maximum values, providing a measure of spatial variability in PM2.5 levels around each cluster, essential for identifying the potential exposure range. Finally, we extracted the value of PM2.5, which was nearest to the cluster (PSU). These values are approximated to three decimal places for precision. We calculated the average PM2.5 exposure (the mean of monthly average) for the entire duration of pregnancy based on the variable length of pregnancy. To extract data based on pregnancy duration from available monthly PM2.5 data, we used the following steps in MATLAB software22: tracked the months as the first and last month of each pregnancy; tracked the months as the first and last month of each pregnancy; extracted data for each respective pregnancy according to its length; extracted data for each respective pregnancy according to its length; calculated average of monthly PM2.5 exposure. calculated average of monthly PM2.5 exposure. Covariates The covariates considered in our analysis included age of mother at the time of survey (20–34 years as reference, compared with ≤19 and ≥35 years age groups), age of mother at first birth (20–34 years as reference, compared with ≤19 and ≥35 years age groups), gravidity (categorised as 2–4 pregnancies as reference, primigravida (first pregnancy) and multigravida which included five or more pregnancies), birth by skilled attendant (previous antenatal care by skilled attendant), preceding birth interval (categorised as <24 months versus ≥24 months as reference), previous pregnancy outcome (no adverse previous pregnancy outcome as reference versus ≥1 previous adverse pregnancy outcomes), wealth index (richest as reference), place of residence (urban vs rural), administrative location of residence (for contextual differences of health systems), cigarette/pipe smoking status and use of biomass fuel for cooking. Statistical analysis Data were analysed using Stata V.17.0 (codes are provided as online supplemental file 1).23 The average rates (weighted proportions) of stillbirth across Pakistan, its provinces and urban and rural regions, with 95% CIs, were calculated. To ensure nationally representative estimates, we applied sample weights from the DHS dataset, adjusting for the complex survey design, including clustering and stratification. This standard approach aligns with DHS recommendations, accounting for differential selection probabilities across the sample. Participants with missing data for key variables, such as PM2.5 exposure or pregnancy outcomes, were excluded from the analysis, which was missing at random. We determine mean PM2.5 exposures among MWRA in Pakistan (aggregate and stratified into urban and rural areas) and separately for seven administrative areas, including four provinces (Punjab, Sindh, KPK and Baluchistan) and other areas (including GB, AJK and FATA). A logistic regression analysis to quantify the association between PM2.5 and stillbirth was conducted. In the univariate unadjusted model, only PM2.5 was used with all covariates. The multicollinearity among covariates was assessed and interactions between all biologically plausible variables were also assessed. We used the average monthly PM2.5 exposure as a continuous variable. For univariate logistic regression analyses, covariates with a p value<0.25 were considered for inclusion. Variables that remained significant at p<0.05 and/or were supported by biological plausibility were retained in the final multivariable logistic regression model assessing the association between PM2.5 exposure and stillbirth. For the final multivariate model, we considered the following variables: wealth index, the current age of mothers, gravidity, preceding birth interval, previous adverse pregnancy outcome and administrative location of residence. Results The total MWRA included in this analysis who gave at least one birth during the last 5 years preceding the survey was 9172. More than two-third (71.0%) were between 20–34 years of age and few were younger (3.8%) and a quarter of them were older (25.2%) than this age range. About 46% of women were living in defined urban locations. About 21% of women were living in Punjab, followed by 18% in Sindh, 24% in KPK/FATA and 12.1% in Baluchistan. The ICT, GB and AJK were the other regions where 25% of the women were residing at the time of the survey. Around half of the women had no education, while barely over 15% had higher education. About 16% were primigravida and 84% were multigravida. Approximately, 54% of these women also had exposure to household air pollution due to cooking with biomass fuel (table 1). Table 1Characteristics and exposures of MWRA, Pakistan (n=9172)Key variablesWeighted percentages (n)Current age of mother (yrs) ≤193.8 (351) 20–3471.0 (6511) ≥3525.2 (2310)Age of mother at first birth (yrs) ≤1937.6 (3339) 20–3461.7 (5482) ≥350.7 (68)Place of residence of women Urban45.9 (4207) Rural54.1 (4965)Province/regions Punjab21.3 (1956) Sindh17.7 (1620) KPK/FATA24.0 (2206) Baluchistan12.1 (1106) Other regions*24.9 (2284)Women’s highest level of education No education49.8 (4568) Primary13.7 (1257) Secondary21.3 (1951) Post-secondary or higher15.2 (1396)Birth by skilled attendant No16.6 (1377) Yes83.4 (6903)Preceding birth interval ≥24 months55.4 (5084) <24 months23.1 (2116) No previous birth16.1 (1480) Missing5.4 (492)Gravida Primi16.1 (1480) 2–449.0 (4491) ≥534.9 (3201)Previous pregnancy outcome Never pregnant before16.1 (1480) All previous live births56.3 (5163) ≥1 Previous adverse pregnancy outcome27.5 (2526) Missing0.03 (3)Cigarette/pipe smoking No94.6 (8676) Yes5.4 (496)Biomass fuel use for cooking No42.8 (3927) Yes54.0 (4950) Other†3.2 (295)Wealth quintile Poorest21.3 (1956) Poorer22.3 (2046) Middle19.8 (1819) Richer18.1 (1660) Richest18.5 (1691)*GB/ICT/AJK.†No food being cooked in the household/missing.AJK, Azad Jammu Kashmir; FATA, Federally Administered Tribal Areas; GB, Gilgit-Baltistan; ICT, Islamabad Capital Territory; KPK, Khyber Pakhtunkhwa; MWRA, married women of reproductive age. Characteristics and exposures of MWRA, Pakistan (n=9172) GB/ICT/AJK. No food being cooked in the household/missing. AJK, Azad Jammu Kashmir; FATA, Federally Administered Tribal Areas; GB, Gilgit-Baltistan; ICT, Islamabad Capital Territory; KPK, Khyber Pakhtunkhwa; MWRA, married women of reproductive age. The average monthly exposure to PM2.5 during pregnancy (over the pregnancy period) was 53.96 (±20.42) µg/m3, with a broad range (5.9–209.4 µg/m3), based on cluster-level data. There was a small difference in exposure to PM2.5 between urban and rural clusters (56.43 vs 51.87 µg/m3); however, the highest exposure in urban locations rose up to 209 µg/m3 versus 162 µg/m3 in rural areas. The average exposure to PM2.5 was highest in clusters in Sindh province (78.06±10.30 µg/m3) and lowest in GB (13.41±3.51 µg/m3) (table 2 and figures 1 and 2). Table 2Average monthly PM2.5 exposure and stillbirth rates (weighted estimates) in various strata of PakistanRegionPM2.5 (µg/m3)RangePM2.5 (µg/m3)Mean (SD)PM2.5 (µg/m3)MedianStillbirth rate (95% CI) /1000 (Weighted estimates) (n=9172)Pakistan (n=8979)5.9–209.453.96 (20.42)53.0517.0 (14.5 to 19.9)Urban (n=4115)6.2–209.456.43 (19.04)55.6515.7 (12.2 to 19.9)Rural (n=4864)5.9–162.251.87 (21.30)50.3018.1 (14.6 to 22.2)Punjab (n=1877)29.2–209.460.76 (17.19)57.7815.8 (10.8 to 22.4)Sindh (n=1599)48.7–134.978.06 (10.30)76.117.3 (11.5 to 24.9)KPK (n=1496)12.5–96.351.31 (14.51)50.9113.2 (8.1 to 20.3)Baluchistan (n=1096)27.2–120.151.81 (12.63)48.8628.9 (19.9 to 40.6)GB (n=644)5.92–34.513.41 (3.51)12.859.1 (3.4 to 19.8)ICT (n=612)35.01–81.756.51 (7.14)55.5717.2 (8.6 to 30.5)AJK (n=963)13.9–62.738.76 (8.57)37.6325.3 (16.4 to 37.2)FATA (n=692)21.9–72.445.61 (9.18)45.764.3 (0.8 to 12.6)AJK, Azad Jammu Kashmir; FATA, Federally Administered Tribal Areas; GB, Gilgit-Baltistan; ICT, Islamabad Capital Territory; KPK, Khyber Pakhtunkhwa; PM2.5, fine particulate matter. Average monthly PM2.5 exposure and stillbirth rates (weighted estimates) in various strata of Pakistan AJK, Azad Jammu Kashmir; FATA, Federally Administered Tribal Areas; GB, Gilgit-Baltistan; ICT, Islamabad Capital Territory; KPK, Khyber Pakhtunkhwa; PM2.5, fine particulate matter. Boxplot of average exposure to PM2.5 (µg/m3)among MWRA during pregnancy. MWRA, married women of reproductive age; PM2.5, particulate matter. In spatial distribution, the lowest exposure to PM2.5 (based on quartile) was observed in GB clusters. The second lowest exposure (second quartile) was observed in most parts of AJK and some parts of KPK and upper parts of Baluchistan. The third quartile of exposure was observed randomly in the same cities and town cluster areas of KPK, Baluchistan and northern part of Punjab province. The highest exposure was noted in major cities, including Lahore, Karachi, Peshawar, ICT and Quetta. Also, most of the clusters in Sindh province and a major part of Punjab province had the highest level of exposure (fourth quartile) (figure 2). PM2.5 exposure during pregnancy for MWRA based on quartiles. Each dot represents a PDHS cluster. MWRA, married women of reproductive age; PDHS, Pakistan Demographic and Health Survey; PM2.5, particulate matter. The overall stillbirth rate during the last pregnancy in Pakistan was 17 per 1000 births, higher in rural (18.1) than in urban areas (15.7). The province of Baluchistan (28.9) and areas of AJK (25.3) had higher stillbirth rates comparatively (table 2). In spatial distribution, the lowest stillbirth rates were noted around ICT (first quartile), while slightly higher stillbirth rates (second quartile) were noted in most of KPK, Baluchistan and northern areas. Most of Punjab and Sindh province had the third highest (third quartile) stillbirth rates in the country, whereas several parts of KPK and upper Punjab had the highest stillbirth rates (figure 3). Stillbirth rates based on quartiles for MWRA. Each dot represents PDHS cluster. MWRA, married women of reproductive age; PDHS, Pakistan Demographic and Health Survey. In the adjusted model, the risk for stillbirth was strongly associated with every 1 µg/m3 average increase in exposure to PM2.5 during pregnancy (adjusted odds ratio(aOR) 1.013, 95% CI 1.002 to 1.025) (table 3). The adjusted results found that compared with the richest wealth quintile, richer (aOR 1.4, 95% CI 0.8 to 2.6), middle (aOR 1.6, 95% CI 0.9 to 3.0), poorer (aOR 2.1, 95% CI 1.2 to 3.9) and poorest wealth quintile (aOR 2.6, 95% CI 1.4 to 4.6) had a higher risk of stillbirth. Those women who were aged 35 years or more had a current pregnancy with two times higher risk of stillbirth. Primigravida (aOR 3.1, 95% CI 1.9 to 4.8) had a higher risk of stillbirth compared with those mothers who had 2–4 pregnancies. Women with one or more previous adverse pregnancy outcomes had 1.8 times higher risk (aOR 1.8, 95% CI 1.2 to 2.7) for stillbirth compared with those women who had no adverse previous pregnancy outcome. Administrative location of residence of women (for contextual differences of health systems) revealed that Baluchistan (aOR 2.5, 95% CI 1.4 to 4.2) and other regions (GB/ICT/AJK) (aOR 2.4, 95% CI 1.4 to 4.1) had higher risk of stillbirth compared with KPK/FATA (taken as reference as lowest risk). No interaction was noted in the final model. Other covariates, including age of mothers at first birth, birth by skilled attendant, urban/rural location of residence, cigarette/pipe smoking status and use of biomass fuel for cooking, were not found significant. Table 3Univariate and multivariable logistic regression (adjusted for clustering) for the association between average monthly PM2.5 exposure and stillbirth rates in PakistanVariableOR (95% CI)Adj. OR (95% CI) (n=8893)Particulate matter<2.5 µg/m3 (average exposure during pregnancy)1.004 (0.996 to 1.014)1.013 (1.002 to 1.025)Wealth index RichestReferenceReference Richer1.3 (0.7 to 2.4)1.4 (0.8 to 2.6) Middle1.5 (0.8 to 2.7)1.6 (0.9 to 3.0) Poorer1.9 (1.1 to 3.2)2.1 (1.2 to 3.9) Poorest2.2 (1.3 to 3.9)2.6 (1.4 to 4.6)Current age of mother (yrs) 20–34ReferenceReference ≤191.2 (0.5 to 2.8)0.7 (0.3 to 1.6) ≥351.8 (1.3 to 2.6)2.0 (1.4 to 2.8)Age of mother at first birth (yrs) 20–34Reference ≤191.0 (0.6 to 1.4)– ≥351.2 (0.2 to 8.6)–Biomass use Non-biomass userReference– Biomass user1.4 (1.0 to 1.9)–Cigarette/Pipe smoking NoReference– Yes1.7 (1.0 to 3.0)–Birth by skilled attendant NoReference– Yes0.9 (0.5 to 1.8)–Preceding birth interval ≥24 monthsReference– ≤23 months1.0 (0.7 to 1.6)–Gravida 2–4 pregnanciesReferenceReference Primi2.2 (1.5 to 3.4)3.1 (1.9 to 4.8) ≥5 pregnancies1.8 (1.2 to 2.6)1.0 (0.7 to 1.5)Previous pregnancy outcomes No adverse pregnancy outcomeReferenceReference ≥1 adverse pregnancy outcomes1.9 (1.3 to 2.7)1.8 (1.3 to 2.6)Urban/Rural UrbanReference– Rural1.2 (0.8 to 1.6)Administrative region KPK/FATAReferenceReference Punjab1.5 (0.9 to 2.6)1.5 (0.9 to 2.6) Sindh1.7 (1.0 to 2.9)1.2 (0.6 to 2.3) Baluchistan2.8 (1.6 to 4.9)2.5 (1.4 to 4.3) Others1.8 (1.1 to 3.0)2.4 (1.3 to 4.2)FATA, Federally Administered Tribal Areas; KPK, Khyber Pakhtunkhwa; PM2.5, fine particulate matter. Univariate and multivariable logistic regression (adjusted for clustering) for the association between average monthly PM2.5 exposure and stillbirth rates in Pakistan FATA, Federally Administered Tribal Areas; KPK, Khyber Pakhtunkhwa; PM2.5, fine particulate matter. Discussion This study found a strong association of PM2.5 with stillbirth. The risk of stillbirth increases as the average exposure to PM2.5 increases. 1 µg/m3 average increase in PM2.5 leads to approximately 1% increase in stillbirth. Our study also found a higher risk of stillbirth among women belonging to the poor section of society (compared with richest wealth quintile), giving birth at 35 years or more, primigravida, women with adverse previous pregnancy outcomes and women living in specific administrative regions (health systems contextual factors). Age of the mother at first birth, birth by skilled attendant, smoking status and biomass fuel use were not found significant in our study. This is the first study looking at the association of PM2.5 and stillbirth in Pakistan. A previous case–control study (live birth and stillbirth from same mothers as controls and case, respectively) from South Asia (including Pakistan) estimated a 7% increase in stillbirth per 10 µg/m3.20 We believe that it is similar to our study estimates. In the previous study, PM2.5 did not vary considerably for the women at the same geographic location, as reported by authors (cases mean exposure: 53·61±23.78 µg/m3 vs control mean exposure: 52·55±24·45 µg/m3). Our study used an innovative technique to assign the exposure to PM2.5 at cluster level for women, which improved the exposure estimation considerably (described in methods). Thus, exposure differential varied considerably in our study (25th percentile 40.7 vs 75th percentile 68.4 µg/m3). Several studies in the USA,24,28 Europe,29 Asia1830,33 and Africa34 have explored the association between air pollution and pregnancy outcome. However, these were low air pollution exposure settings; therefore, comparison with these was not possible. Yang et al reported an aOR for stillbirth of 1·60 (95% CI 1·34 to 1·91) per 10 µg/m³ increment in PM2·518; these results equate with our study findings. Another study by Zhang et al found a non-linear effect (risk ratio 1·08–1·51) on missed abortion in the first trimester in Beijing, China30 and Ebisu and colleagues estimated the risk ratio of stillbirth to be 1·08 (95% CI 1·01 to 1·16) in California, USA.24 Although differences in outcome indicators, levels and constituents of PM2.5 exposure, populations and epidemiological designs make direct comparisons across studies challenging, the estimated associations between PM2.5 and stillbirth reported in these studies are generally in line with our findings.35 Our study has identified factors for stillbirth, which were also identified by previous studies. Poor wealth index, women ≥35 years, primigravida, ≥1 adverse previous pregnancy outcome and specific administrative regions were significantly associated with high stillbirth rates. A study conducted in six LMICs by Patel et al reported an increased risk (OR 1.3) of stillbirth among low socioeconomic status women.36 Another study in East Africa indicated that women aged 35 years and older had a significantly higher stillbirth rate.37 A study in South India reported a high incidence (two times) of stillbirths among primigravida compared with multigravida.38 Al Khalaf et al reported that women who had previous stillbirths had a higher risk of subsequent stillbirth.39 Our study found that specific administrative regions had higher stillbirth rates. This reflects the contextual factors that influence the health outcome, including health systems strength and weaknesses. Baluchistan is one of the remote and low-developed provinces of Pakistan; therefore, it had the highest risk for stillbirth.21 Associations (OR) of stillbirth with PM2.5 increased after adjustment for other risk factors, indicating negative confounding. Our analysis revealed that geographic locations and wealth index were particularly playing a negative confounding role for this association. While higher levels of PM2.5 were anticipated in more affluent areas, stillbirths are typically associated with lower socioeconomic statuses. Aligned with this expectation, women experiencing higher stillbirth rates were predominantly from poorer and remote communities. This relationship touches on the concept of carbon intensity of well-being, which explores how environmental impacts intersect with socioeconomic status to influence well-being across various populations. Here, socioeconomic inequalities appear to play a significant role in both environmental exposures and health outcomes, as seen in this PM2.5–stillbirth association.40 Although our dataset lacked individual-level data on key maternal comorbidities, such as pregestational and gestational diabetes and gestational hypertension, few large population-based studies have examined these factors and still observed significant associations between gestational PM₂.₅ exposure and stillbirth. For instance, Zang et al conducted a large prospective birth cohort study in the coastal area of China, adjusting for a wide range of maternal and environmental covariates, including chronic diseases, such as hypertension and diabetes. They reported significantly elevated stillbirth risks across all trimesters, with adjusted relative risks of 1.14 (95% CI 1.06 to 1.22), 1.11 (95% CI 1.06 to 1.16), 1.15 (95% CI 1.09 to 1.21) and 1.14 (95% CI 1.08 to 1.20) for PM₂.₅ exposure during the first, second, third and entire pregnancy periods, respectively.19 This study strengthens the biological plausibility of our findings by demonstrating that even after controlling for major maternal comorbidities, such as hypertension and diabetes, fine particulate exposure remains an independent and statistically significant predictor of stillbirth. The average monthly exposure to PM2.5 during pregnancy (over the pregnancy period) was 53.96 (±20.42) µg/m3 in this population, which is much higher than the current guidelines of 5 µg/m3 and the interim target of <35 µg/m3.41 Our reference PM2.5 values (first quartile cut-off: 40.8 µg/m3) approximate the interim target values. PM2.5 levels were consistently elevated across all four Pakistani provinces, regardless of urban or rural location. Since some rural districts have developed urban centres with higher air pollution levels, there was no significant difference in PM2.5 between urban and rural areas. Nonetheless, there was a high variation between urban and rural exposures. Our study showed a clear dose–response relationship, showing increased risk of stillbirth for every 1 µg/m3 increase in PM2.5. Most previous studies on PM2.5 and stillbirth have been conducted in high-income countries with substantially lower PM2.5 concentrations and lower stillbirth rates compared with Pakistan.42 Consequently, a direct comparison of dose–response relationships from these studies was not feasible. This study has notable strengths that enhance the robustness and relevance of the findings. First, the use of nationally representative data from the PDHS 2017–2018 provides a comprehensive overview of the population, increasing the generalisability of the findings within Pakistan. The large sample size of 9172 MWRA strengthens the statistical power of the study, allowing for more reliable estimates. The study also employed high-resolution satellite-based PM2.5 data (1 km × 1 km), enabling detailed assessment of ambient air pollution exposure across diverse geographical regions in Pakistan, assigning PM2.5 exposure to participants for the duration from conception to pregnancy outcome. We employed an innovative technique to use satellite data to align it with health data while making a database linking PM2.5 exposure to health data. The high-resolution database allowed for more accurate approximation of fine particulate matter exposure. Using PM₂.₅ exposure as a continuous variable allows for a more precise assessment of how incremental increases in air pollution influence stillbirth risk. This approach captures the dose–response relationship across the entire exposure range, providing critical evidence to guide public health interventions and policymaking. Despite these strengths, the study has several limitations that warrant consideration. The use of satellite-based PM2.5 data, although beneficial for broad spatial coverage, may not accurately capture local variations in air quality, leading to potential exposure misclassification. The spatial resolution, while high, may not reflect microenvironmental conditions experienced by individuals. Because of the time lag between the survey conducted and the occurrence of pregnancy outcomes, there is an inherent recall concern. The data did not observe the migration history during pregnancy and it is assumed that women remained at the same location for the entire duration of pregnancy. However, during the limited period of pregnancy duration, we expect that the pregnant woman will remain at the location where she was interviewed. This may be considered a limitation. Additionally, the reliance on self-reported data from the PDHS introduces the potential for reporting bias, particularly for pregnancy outcomes, such as stillbirth. Residual confounding from unmeasured factors may still influence the observed associations. Although individual-level data on maternal comorbidities, such as hypertension and diabetes, were not available in this study, both are known risk factors for stillbirth and could have contributed to the observed associations. However, available evidence suggests that particulate exposure exerts an independent effect beyond these conditions. Other unmeasured variables, including occupational exposures, genetic susceptibility and coexposure to other environmental pollutants, may also have introduced residual confounding and should be considered when interpreting the findings. Also, jittering in the DHS data is one of our limitations—the GPS coordinates have had random errors introduced to protect participant’s privacy, which means matching to PM2.5 data will have introduced some errors and, thus, affected our findings. Another limitation of this study is the possible spatial autocorrelation in PM2.5 exposure, as nearby participants may have similar exposure levels. We did not account for this through spatial regression models. A further limitation of this study is the use of data from the 2017–2018 PDHS, which may not fully reflect current patterns of air pollution exposure or stillbirth rates. However, given the persistently high levels of PM2.5 in Pakistan and limited improvements in maternal health indicators, the associations observed are likely to remain relevant for current clinical and public health contexts. Conclusions Fine particulate matter (PM 2.5) is strongly associated with stillbirth and there is a dose–response relationship; higher exposure to PM2.5 leads to an increase in stillbirths. Both air quality and stillbirth are very high in LMICs, which should motivate the maternal health and environmental experts to work closely to reduce this risk among the population. Effective control requires government regulations, emission controls and individual actions to promote clean energy. Supplementary material Acknowledgements We acknowledge Sohail Lakhani for his assistance with the initial data analysis. We also appreciate Muhammad Yousuf for his support in developing GIS maps for this study. Additionally, we thank Inara Tharani for proofreading the manuscript and ensuring consistency in formatting. Footnotes Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Prepublication history and additional supplemental material for this paper are available online. To view these files, please visit the journal online (https://doi.org/10.1136/bmjopen-2025-100972). Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Data availability free text: Data and the analysis codes (for STATA software) are available in a public, open access repository in Open Science Framework. The STATA code is provided in the supplementary file. It can be accessed using following: https://osf.io/uq24r/ Map disclaimer: The depiction of boundaries on this map does not imply the expression of any opinion whatsoever on the part of BMJ (or any member of its group) concerning the legal status of any country, territory, jurisdiction or area or of its authorities. This map is provided without any warranty of any kind, either express or implied. Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting or dissemination plans of this research. Data availability statement Data are available in a public, open access repository. References Supplementary Materials Data Availability Statement Data are available in a public, open access repository.