Secondary Logo

Journal Logo

Original Research Article

Prenatal and childhood exposure to air pollution and traffic and the risk of liver injury in European children

Garcia, Erikaa,*; Stratakis, Nikosa; Valvi, Damaskinib; Maitre, Léac; Varo, Neread; Aasvang, Gunn Marite; Andrusaityte, Sandraf; Basagana, Xavierc; Casas, Maribelc; de Castro, Montserratc; Fossati, Serenac; Grazuleviciene, Reginaf; Heude, Barbarag; Hoek, Gerardh; Krog, Norun Hjertagere; McEachan, Rosemaryi; Nieuwenhuijsen, Markc; Roumeliotaki, Theanoj; Slama, Rémyk; Urquiza, Josec; Vafeiadi, Marinaj; Vos, Miriam B.l; Wright, Johni; Conti, David V.a; Berhane, Kirosm; Vrijheid, Martinec; McConnell, Roba; Chatzi, Lidaa

Author Information
Environmental Epidemiology: June 2021 - Volume 5 - Issue 3 - p e153
doi: 10.1097/EE9.0000000000000153


What This Study Adds

Nonalcoholic fatty liver disease has become an epidemic in children and is now one of the most common pediatric chronic liver diseases. We used a population-based multicohort study spanning six European countries to assess whether prenatal or childhood exposure to air pollution or traffic relate to biomarkers of liver injury and suspected nonalcoholic fatty liver disease in children. Results indicate no evidence of elevated liver enzyme levels in association with exposure to air pollution and traffic in either prenatal or postnatal periods. Findings from interaction analyses, however, suggest PM10 effect estimates may be higher in children who are overweight or obese.


Nonalcoholic fatty liver disease (NAFLD) has become an epidemic in the pediatric population and is now one of the most common chronic liver diseases in children,1,2 currently affecting 3–13% of the general pediatric population in Western countries.3 In large population studies, the prevalence of elevated serum levels of alanine aminotransferase (ALT), an established biomarker of liver injury and clinical screening tool for pediatric NAFLD, has almost tripled,4 coinciding with the childhood obesity epidemic of the past several decades. Children with NAFLD may suffer future comorbidities, which can have lifelong effects, including low-bone mineral density, type 2 diabetes, an adverse cardiometabolic risk profile, and development of cardiovascular disease.1,5 In the next decade, liver complications due to NAFLD are predicted to become the most frequent indication for liver transplantation.5

Emerging experimental evidence indicates that ambient and near-roadway pollution causes liver injury and may contribute to NAFLD development.6–9 Mice chronically exposed to airborne particulate matter with an aerodynamic diameter <2.5 μm (PM2.5) developed hepatic steatosis, inflammation, and fibrosis.8,9 The epidemiologic literature is scant with only two studies having examined this relation in children. A small prospective study of 74 US children of mean age 14 years recruited from an obesity clinic found an association of childhood nitrogen dioxide (NO2) exposure and residential traffic volume with cytokeratin-18 (CK-18), a biomarker hepatocyte apoptosis.10 The generalizability of these results is unclear since all study participants were overweight or obese. In a cross-sectional study of 150 newborns maternal residential exposure to particulate matter (PM) was associated with higher levels of liver injury biomarkers, including ALT, aspartate aminotransferase (AST), and gamma-glutamyltransferase (GGT) in cord blood.11 Thus, whether prenatal pollution and traffic exposures relate to liver injury later in childhood is uncertain. Given the public health importance of the pediatric NAFLD epidemic, the ubiquity of air pollution and traffic exposure, and the scarcity of human studies, further investigations, including prospective studies, are urgently needed to clarify the effect of pollution and traffic exposure on liver injury and NAFLD risk in children.

We present the first large-scale population-based human study to examine the relationship between early life air pollution exposure and risk of liver injury in childhood. We used data from a well-established population-based multicohort study across six European countries, the Human Early Life Exposome (HELIX) study, to assess whether prenatal or childhood exposure to air pollution or traffic relate to four noninvasive clinical biomarkers of liver injury and suspected NAFLD: ALT, AST, GGT, and CK-18. We examined exposure to outdoor air pollutants, indoor pollutants, and traffic. Finally, we examined the potential for effect heterogeneity by sex and overweight or obesity status, given prior evidence of differential NAFLD prevalence by these factors.3


Study population

The study population was drawn from the HELIX study,12 a collaborative project across six established and ongoing longitudinal population-based birth cohort studies in six European countries. This includes the Born in Bradford (BiB) study in the United Kingdom,13 the Étude des Déterminants pré et postnatals du développement et de la santé de l’Enfant (EDEN) study in France,14 the INfancia y Medio Ambiente (INMA) cohort in Spain,15 the Kaunas cohort (KANC) in Lithuania,16 the Norwegian Mother, Father, and Child Cohort Study (MoBa),17 and the RHEA Mother Child Cohort study in Crete, Greece.18 The full HELIX protocol and database have been described in detail previously.19 Briefly, a subcohort of 1,301 mothers and their singleton children across the six cohorts (approximately 200 children per cohort) was followed up in 2014–2015 for clinical examination, interview with the mothers, and collection of biologic samples. Data collection was standardized across cohorts and performed by trained staff.

The study population comprised 1,102 (85%) mother-child pairs from the HELIX subcohort, following inclusion criteria regarding availability of data on all four liver enzymes, ALT, AST, GGT, and CK-18. All participating families provided written informed consent. Approval for the HELIX project was obtained from the local ethical committees at each site. Additionally, the current study was approved by the University of Southern California Institutional Review Board.

Air pollution and traffic exposure assessment

Here, we provide a brief description of the assessment methods for exposures included in this study. A detailed description of methods used to assess air pollution and traffic is provided in Tamayo-Uria et al.20 The following ambient air pollutants were examined: NO2 and particulate matter with an aerodynamic diameter <2.5 μm (PM2.5) and <10 µm (PM10). These were assessed for the pregnancy period (averages for each trimester and across entire pregnancy) and in the year before childhood study visit based on both home and school address using land-use regression (LUR) models developed in the context of the European Study of Cohorts for Air Pollution Effects (ESCAPE) project21–25 or dispersion models,26 temporally adjusted to measurements made in local background monitoring stations.20 Site-specific ESCAPE LUR models were used for most cohorts.20–25 Exposure assessment for PM2.5 and PM10 in the BiB cohort (United Kingdom) was conducted using the ESCAPE LUR model for London/Oxford, United Kingdom, adjusting for background PM concentration based on air monitoring data from Bradford.20,27 In the EDEN cohort (France), assessment of PM2.5 exposure was conducted based on the European-wide ESCAPE LUR model28 and assessment of NO2 and PM10 (only for pregnancy period) exposure was conducted based on dispersion models.20,26 Routine background ambient air monitoring stations that were active during the whole study period provided daily background concentration data used for temporal adjustment.20 The following markers of traffic were examined: inverse distance to nearest road, traffic load on all roads within a 100 m buffer, and traffic density on nearest road. These were assessed for the full pregnancy period and in the year before childhood study visit based on both home and school address and were calculated from road network maps following the ESCAPE protocol,21,23 applying the land-use regression (LUR) methods and Geographic Information System (GIS) predictor variables used within the ESCAPE project as described in Eeftens et al.23 We examined the following indoor air pollutants: NO2, PM2.5, benzene, and the sum of benzene, toluene, ethylbenzene, and xylene (BTEX). These were assessed for the year before childhood study visit and were estimated using a prediction model that combined measurements in the homes of a subgroup of children with questionnaire data from the subcohort.20 As part of a child panel study nested within the HELIX subcohort (all cohorts except MoBA), indoor NO2, benzene, and TEX (toluene, ethylbenzene, and xylene) were measured in the homes of 157 participants. PM2.5 was measured in INMA, BiB, and EDEN. Panel study participants were followed for one week in two seasons. NO2, benzene, and TEX sampling were conducted over 7 days, and PM2.5 sampling was conducted over 24 hours. The last day of the first week of measurements was the same day as the subcohort examination, which included the main HELIX questionnaire.20

Biomarkers of liver injury

Collection and laboratory analysis of liver enzyme levels have previously been reported in detail.29 Identical predefined standardized protocols across all six cohorts were followed to collect and process blood samples. Briefly, at the end of clinical examinations as part of the subcohort follow-up visit blood samples following a median fasting time of 3.3 hours were collected from children into 4 ml silica plastic tubes. Samples were gently inverted 6–7 times, spun down at 2,500 g for 15 minutes at 4°C, and then frozen at −80°C under optimized and standardized procedures. Concentrations (IU/L) of ALT, AST, and GGT in serum were assessed by Biochemistry Laboratory of the Clínica Universidad de Navarra using homogenous enzymatic colorimetric methods on a Colorimetry Cobas 8000 analyzer according to the manufacturer’s instructions (Roche Diagnostics GmbH Mannheim). CK-18 in serum was measured by ELISA (M30 Apoptosense® ELISA, PEVIVA) according to the manufacturer´s instructions. All coefficients of variation were less than 3%.

Covariate assessment

Using directed acyclic graph theory,30 a set of variables considered to be sufficient for confounding adjustment were decided upon a priori. The covariates were cohort, maternal age (years), maternal prepregnancy BMI (kg/m2), maternal education level (low, middle, high), paternal education level (low, middle, high), and maternal active smoking during pregnancy (yes, no). Additionally, data on child’s sex (male, female) and age (years) and BMI (kg/m2) at follow-up visit were assessed. Information on maternal age at birth, maternal prepregnancy BMI, maternal education, paternal education, and smoking status during pregnancy from each study participant was obtained by each cohort during pregnancy or at birth by questionnaire or medical records. Birthdate and newborn sex were obtained at birth. During the follow-up examination, anthropometric data were collected using regularly calibrated instruments. Height was measured with a stadiometer and weight with a digital weight scale, both without shoes and with light clothing. Height and weight measurements were converted to BMI for age-and-sex z-scores using the international WHO reference curves to allow comparison with other studies.31 Overweight and obese children were defined as those above the age-and sex-specific 85th and 95th percentiles, respectively, as recommended by WHO ( Maternal alcohol consumption during pregnancy (yes, no) obtained in each cohort during pregnancy or at birth by questionnaire or medical records is used in a sensitivity analysis.

Statistical analyses

Skewed exposure variables were transformed to improve model fit. The following were natural log transformed: ambient NO2, inverse distance to nearest road, and all indoor pollutants (NO2, PM2.5, benzene, and BTEX). The following were cube root transformed: traffic load on all roads within a 100 m buffer and traffic density on nearest road. Analyses were conducted on transformed variables, but plots are shown with back-transformed values for interpretability. Single imputation of missing data was done using a chained equations method,32 as described previously in detail.20 The proportion of missing data was minimal, ranging from 1.2% for maternal age at birth to 4.9% for paternal education. The associations between liver injury biomarkers with air pollution exposures and markers of traffic were estimated separately based on generalized additive models (GAM) using the R package “mgcv.”33 Smooth functions of exposure were fitted using a penalized regression spline33 to allow for possible nonlinear exposure-response functions. Covariates included in the models were cohort, paternal education level, and maternal age, education level, prepregnancy BMI, and active smoking during pregnancy. Maternal age and prepregnancy BMI were each fitted with a smooth function using penalized regression splines. We also examined associations with categories of exposure based on tertiles using the same modeling approach as described above. We conducted a sensitivity analysis further adjusting for maternal alcohol consumption during pregnancy as a potential confounding variable. We then evaluated possible interaction between exposure and child’s sex (male; female) or overweight/obese status at follow-up assessment (dichotomous: not overweight or obese; overweight or obese). Because GAM models indicated linear exposure-response curves for almost all exposure-outcomes, interactions were assessed using an interaction term in models with exposure included as a continuous variable with a linear term. For exposure-outcomes with significant nonlinearity, interactions were additionally assessed using smooth parameterizations that allowed for quantification of interactions.

All hypotheses were tested assuming a 0.05 significance level and a two-sided alternative hypothesis. P values from GAM models with smooth terms for exposure were adjusted for multiple comparisons for each biomarker using a method of computing P values adjusted for correlated tests (PACT).34 It was not possible to use the PACT adjustment for the P values from the categorical or interaction models, so we instead used the false discovery rate (FDR) procedure to adjust these P values.35 All statistical analyses were performed using R


Characteristics of the study populations

Distribution of child and parental sociodemographic characteristics among 1,102 study participants are shown in Table 1. The median maternal age at participant’s birth was 31 years (interquartile range [IQR]: 27.8–34.1 years) with a median age at follow-up in childhood of 8.2 years (IQR: 6.6–9.1 years). There were slightly more males than females (596 [54.1%] compared with 506 [45.9%], respectively) and most children came from parents of middle to high education. Distribution of liver enzyme concentrations among participants are shown in Table 1. Exposures distribution of ambient (outdoor) air pollutants, markers of traffic, and indoor air pollutants among study participants are shown in Table 2. Median exposure concentrations during pregnancy for ambient NO2, PM10, and PM2.5 were 18.4, 22.5, and 14.9 µg/m3, respectively.

Table 1. - Distribution of sociodemographic characteristics and liver injury biomarkers in the study population
N 1,102
 MoBa, Norway 270 (24.5)
 INMA, Spain 211 (19.1)
 EDEN, France 196 (17.8)
 KANC, Lithuania 166 (15.1)
 RHEA, Greece 165 (15.0)
 BiB, United Kingdom 94 (8.5)
Maternal characteristics
 Age at birth, years, median (IQR) 31.0 (27.8–34.1)
 Prepregnancy BMI, kg/m2, median (IQR) 23.5 (21.1–26.7)
  High 505 (45.8)
  Middle 434 (39.4)
  Low 163 (14.8)
 Active smoking during pregnancy
  No 945 (85.8)
  Yes 157 (14.2)
Paternal characteristics
  High 596 (54.1)
  Middle 382 (34.7)
  Low 124 (11.3)
Child characteristics
 Age at follow-up, years, median (IQR) 8.2 (6.6-9.1)
  Male 596 (54.1)
  Female 506 (45.9)
Liver enzyme concentrations, IU/L, median (IQR)
 ALT 14.4 (11.6-18.1)
 AST 28.8 (25.1-34.2)
 GGT 12.0 (10.4-14.3)
 CK-18 71.2 (60.7-88.9)
N (%), unless otherwise noted.
ALT indicates alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; CK-18, cytokeratin-18; GGT, gamma-glutamyltransferase; IQR, interquartile range.

Table 2. - Distribution of exposures to ambient air pollutants, markers of traffic, and indoor air pollutants in the study population
Ambient air pollutants
NO2, µg/m3
 Pregnancy 18.4 (13.4, 26.2)
 Trimester 1 19.0 (13.9, 28.2)
 Trimester 2 17.1 (12.5, 26.5)
 Trimester 3 17.1 (13.0, 26.5)
 Childhood, home 19.7 (11.2, 30.2)
 Childhood, school 18.5 (12.0, 29.9)
PM10, µg/m3
 Pregnancy 22.5 (15.9, 27.7)
 Trimester 1 21.9 (16.1, 28.1)
 Trimester 2 22.0 (15.6, 28.3)
 Trimester 3 20.2 (15.1, 27.5)
 Childhood, home 25.2 (18.8, 31.5)
 Childhood, school 24.9 (18.4, 31.7)
PM2.5, µg/m3
 Pregnancy 14.9 (13.0, 17.0)
 Trimester 1 14.5 (12.2, 17.7)
 Trimester 2 14.1 (11.8, 17.2)
 Trimester 3 13.7 (11.6, 16.4)
 Childhood, home 13.7 (11.5, 14.9)
 Childhood, school 13.8 (11.6, 14.9)
Markers of traffic
Inverse distance, m-1
 Pregnancy 0.051 (0.019, 0.108)
 Childhood, home 0.019 (0.007, 0.053)
 Childhood, school 0.015 (0.008, 0.03)
Traffic load, vehicles/day-m
 Pregnancy 212,135 (0, 1,501,124)
 Childhood, home 229,783 (0, 1,514,536)
 Childhood, school 263,992 (0, 1,579,946)
Traffic density, vehicles/day
 Pregnancy 1,210 (500, 4,302)
 Childhood, home 2,979 (500, 10,898)
 Childhood, school 3,398 (798, 10,000)
Indoor air pollutants
 NO2, µg/m3: childhood, home 1.8 (1.4, 2.4)
 PM2.5, µg/m3: childhood, home 29.1 (16.9, 94.3)
 Benzene, µg/m3: childhood, home 9.5 (7.7, 13.8)
 BTEX, µg/m3: childhood, home 21.0 (14.4, 30.6)
Median (interquartile range).
BTEX indicates sum of benzene, toluene, ethylbenzene, and xylene; NO2, nitrogen dioxide; PM10, particulate matter <10 μm; PM2.5, particulate matter <2.5 μm.

Associations with liver enzymes

Modeled associations between ambient air pollution exposure during different age windows and ALT are shown in Figure 1, and P values unadjusted for multiple comparisons are in eTable 1; No results were statistically significant after adjustment for multiple comparisons. Null associations with ALT were also observed for markers of traffic exposure in utero and in childhood and indoor air pollution exposure during childhood (eFigures 1; and 2;; eTable 1; Modeled associations for AST, GGT, and CK-18 with ambient air pollution, markers of traffic, and indoor air pollution are shown in the supplement (eFigures 2–8;;; eTable 1; No results were statistically significant after adjustment for multiple comparisons. AST was borderline statistically significantly associated with ambient PM10 exposure in trimester 1 (P = 0.055 after adjustment for multiple comparisons) with a positive association observed for PM10 exposures above approximately 30 μg/m3 (eFigure 3; Null associations were similarly observed for models with tertiles of exposure (eTables 2–5; Additional adjustment for maternal alcohol consumption during pregnancy did not markedly change model results (results not shown).

Figure 1.
Figure 1.:
Model results for ambient air pollution and ALT. Associations between ambient air pollution exposure (NO2, PM10, and PM2.5) during different age windows and ALT modeled separately using generalized additive models, adjusting for cohort, maternal age, maternal prepregnancy BMI, maternal education level, paternal education level, and maternal active smoking during pregnancy. Null referent line is shown. No results were statistically significant after adjustment for multiple comparisons. ALT indicates alanine aminotransferase; BMI, body mass index; NO2, nitrogen dioxide; PM10, particulate matter <10 μm; PM2.5, particulate matter <2.5 μm.

Interaction with sex and overweight/obese status

Results of interaction models by sex are shown in the supplement (eFigures 9–17; No clear differences by participant’s sex were observed in associations for air pollution or traffic exposure with any liver enzyme biomarker. Results of interaction models by overweight/obese status are shown in Figure 2 for ALT with ambient air pollution and in the supplement for the other liver enzyme biomarkers and exposures (eFigures 18–25; The association between prenatal PM10 exposure and ALT was statistically significantly different by whether participants were overweight or obese at follow-up assessment (P = 0.045 after adjustment for multiple comparisons), with larger positive slopes among those who were overweight or obese (β = 0.120 per 1 µg/m3; SE = 0.065) compared with a slightly negative slope among those who were not overweight or obese (β = –0.047 per 1 µg/m3; SE = 0.052). Overall ambient PM10 exposures associations with ALT, slopes appear larger, although not always statistically significantly different, among participants who were overweight or obese compared with those who were not overweight or obese (Figure 2). No other clear differences by participant overweight/obese status were observed in associations for other air pollution/traffic exposure with any liver enzyme biomarker. For the five exposure-outcomes with significant nonlinearity (eTable 1;, plots of exposure smooth with statistically significant interaction with either sex or overweight/obese status are shown in eFigure 26;, but these are not adjusted for multiple comparisons and should be interpreted with caution.

Figure 2.
Figure 2.:
Model results for ambient air pollution with overweight or obese status interaction and ALT. Associations between ambient air pollution exposure (NO2, PM10, and PM2.5) during different age windows and ALT, stratified by overweight or obese status (orange line represents those who are overweight or obese; purple line represents those who are not overweight or obese). Modeled separately using generalized additive models, adjusting for cohort, maternal age, maternal prepregnancy BMI, maternal education level, paternal education level, and maternal active smoking during pregnancy. *Association for prenatal PM10 exposure was statistically significant after adjustment for multiple comparisons. No other results were statistically significant. ALT indicates alanine aminotransferase; BMI, body mass index; NO2, nitrogen dioxide; PM10, particulate matter <10 μm; PM2.5, particulate matter <2.5 μm.


Using data from a well-characterized multi-cohort study across several European countries, we examined for the first time in a population-based prospective study the associations between prenatal and childhood air pollution and traffic exposure with biomarkers of child liver enzymes. We found no clear association of air pollution and traffic both in the prenatal and postnatal periods with liver enzyme levels in the overall study population. Notably, stratified analysis by child obesity status revealed a stronger association between prenatal PM10 exposure and ALT among children who were overweight or obese compared with children who were not overweight or obese; the same pattern was observed in trimester-specific associations and for childhood PM10 exposures.

There is emerging evidence that environmental factors may play a role in the onset and progression of NAFLD.7,37 Air pollution exposure is linked with oxidative stress, systemic low-grade inflammation, and alterations in insulin/insulin-like growth factor and insulin resistance, which are all etiological factors related to NAFLD.37 Among the few studies in adult populations, associations have been reported for PM2.5 with ALT,38,39 AST,38 and GGT40; PM10 with ALT41,42 and AST42; NO2 with ALT and AST38,41,42; and blood benzene with CK-18.43 Only two studies have examined the effect of air pollution or markers of traffic exposure on liver injury and NAFLD risk in children. In contrast to our findings, a small cross-sectional study on 150 newborns from Sabzevar, Iran, reported higher maternal exposure to PM <1 μm, PM2.5, and PM10 to be associated with increased ALT, AST, and GGT in newborn cord blood.11 Positive associations were also observed with higher street length in a 100 m buffer around the home for ALT, AST, and GGT, and an inverse association for distance to major roads with AST. Pollution levels were a lot higher in this newborn study in comparison with the present analysis; for example, the median (IQR) PM2.5 was 46.8 (40.1–73.3) μg/m3 compared with 14.9 (13.0–17.0) μg/m3, respectively. It is possible air pollution effects may differ by level of pollutant, with larger effects at higher pollutant concentrations and smaller, less detectable effects at lower concentrations, such as those in the present study. A prospective study of 74 children with mean age of 14 years who were overweight or obese from the Yale Pediatric Obesity Clinic followed for two years reported association for CK-18 at follow-up with NO2 and traffic volume at baseline residence.10 An IQR increase in NO2 (1.91 ppb) was associated with 11 U/L higher CK-18 (SE = 5.4), and an IQR increase in residential traffic volume within a 1-km buffer was associated with 15 U/L higher CK-18 (SE = 5.2) per 110,000 vehicle-km. No statistically significant associations of AST or ALT with NO2 or traffic volume were found. This study, however, comprised only overweight and obese children recruited from an obesity clinic, whereas our study sample is population based. Given their health status, the children in the Yale study may have been more susceptible to the effects of pollution. Although we did not observe the same positive association between CK-18 and NO2 and traffic volume, we did find similar stronger associations between PM10 and ALT among children who were overweight or obese. These findings of higher pollution effects in children who are overweight/obese should be more carefully examined in future research.

Our analysis revealed an interaction between prenatal PM10 exposure and overweight or obese status in childhood for liver injury biomarkers and especially for ALT. This suggests that prenatal air pollution might make the liver vulnerable to effects of increased weight status. This observation is consistent with the multiple-hits hypothesis for NAFLD pathogenesis,44 whereby prenatal PM10 exposure serves as an initial hit, leaving the liver compromised and sensitive to further insults, such as obesity or being overweight (possibly a proxy for high-fat/proinflammatory diet), acting as an additional hit promoting disease progression. Synergistic interaction between air pollutant exposure and high-fat diet, a precursor to increased weight status, have been reported based on animal experiments. Mice exposed to an average of 15 μg/m3 PM2.5 and fed high-fat chow showed significantly increased lobular inflammation, hepatocyte ballooning, and Mallory bodies compared with either PM2.5 exposure or diet alone.8 It has also been reported that PM2.5 acts synergistically with high-fat diet to promote other metabolic outcomes, such as adiposity, insulin resistance and type 2 diabetes.45,46 It has been posited that environmental exposure—the first hit—may compromise the liver’s protective responses against overnutrition—a subsequent hit—, promoting fatty liver disease from high-fat diets.47 The synergistic interaction between prenatal PM10 exposure and overweight or obese status merits further investigation. If these findings hold, then regulatory effects to improve air quality may potentially reduce the hits to the liver and lower the risk of NAFLD development.

Several toxicological studies have examined the role of air pollution in NAFLD and possible mechanisms, with most focusing on PM. Inhaled PM particles can reach the liver where they activate Kupffer cells and induce an inflammatory response through the activation of several molecular pathways, such as c-Jun N-terminal kinases-activator protein 1, nuclear factor-κB, and Toll-like receptor 4.7 In a mouse study, PM2.5 exposure significantly increased Kupffer secretion of cell interleukin-6,8 a proinflammatory cytokine associated with human NAFLD and those with higher steatohepatitis compared with simple fatty liver.48 PM particles may also affect peroxisome proliferator-activated receptors activity, altering lipid and glucose metabolism in Kupffer cells, hepatocytes, and hepatic stellate cells.7,9 In mice with 6-month PM2.5 exposure, accelerated upregulation of tumor necrosis factor alpha (TNF-α) caused hepatic inflammation and oxidative stress, disrupting the balance of lipid metabolism in the liver.9 However, given the species-specific toxico-kinetics of PM, extrapolation from animals to humans is difficult.49 Although the toxicological literature supports an effect of air pollution on NAFLD development, we did not observe such associations in our longitudinal epidemiologic study. It may be that liver injury biomarkers being examined here in a population of apparently healthy children are not sufficiently sensitive to detect small perturbations in hepatic inflammation and lipid metabolism or that the levels of exposure in our analysis are below some effect threshold.

The main strength of our study is that it is the first large-scale epidemiologic study of the impact of air pollution on child liver injury, using data from six European birth cohorts with prospectively collected data. In addition to using the same protocol for the outcome assessment, these six cohorts have detailed information on air pollution exposure assessed using standardized protocols in two critical developmental age periods, in utero (including estimates of trimester-specific outdoor air pollution exposure) and early childhood. Our study has also a number of limitations. First, we used serum liver enzymes as our measure of liver injury rather than the current diagnostic gold standard of liver biopsy for NAFLD. Large-scale liver biopsies, however, are not feasible in large population studies due to ethical considerations and high costs; any outcome misclassification is not expected to have been differential by exposure level. Second, although we had detailed exposure assessment regarding participants’ residential addresses, we did not have information about total exposure, including maternal occupational exposure during pregnancy or exposure at other locations such as school for indoor pollutants, which might have affected the observed results. Third, we were not able to examine other potentially informative markers of traffic, such as distance to nearest major road or total street length in select buffers since such estimates were not available for the HELIX study. It is important to note that traffic exposure captures not only near-roadway air pollution but also traffic noise and possibly aspects of the built environment, such as green space and opportunity for outdoor physical activity. Finally, because of the large number of exposures being tested and our appropriate adjustment for multiple comparisons, the statistical power was limited. A narrower list of focused exposures could have partly addressed this issue, however, we wanted to take full advantage of the rich air pollution and traffic exposure data available in HELIX. Regardless of statistical significance, researchers will be able to examine the exposure-response curves reported for the four liver enzymes with the pollution and traffic exposures to inform their own work. Furthermore, P values not adjusted for multiple comparisons are presented in the Supplement for the benefit of the reader.

This multicohort study of over 1,100 European children did not find prenatal or childhood air pollution or traffic exposure to be associated with biomarkers of liver injury in children. Findings from interaction analyses suggest PM10 effect estimates may be higher in children who are overweight or obese, consistent with the multiple-hits hypothesis for NAFLD pathogenesis. Although additional research is needed to confirm these findings, this synergistic interaction suggests that reduction of particulate air pollution levels may be a possible intervention to lower the risk of liver injury and NAFLD development in children.


1. Bush H, Golabi P, Younossi ZM. Pediatric non-alcoholic fatty liver disease. Children (Basel). 2017; 4:E48.
2. Alisi A, Feldstein AE, Villani A, Raponi M, Nobili V. Pediatric nonalcoholic fatty liver disease: a multidisciplinary approach. Nat Rev Gastroenterol Hepatol. 2012; 9:152–161.
3. Anderson EL, Howe LD, Jones HE, Higgins JP, Lawlor DA, Fraser A. The prevalence of non-alcoholic fatty liver disease in children and adolescents: a systematic review and meta-analysis. PLoS One. 2015; 10:e0140908.
4. Welsh JA, Karpen S, Vos MB. Increasing prevalence of nonalcoholic fatty liver disease among United States adolescents, 1988-1994 to 2007-2010. J Pediatr. 2013; 162:496–500.e1.
5. Mantovani A, Scorletti E, Mosca A, Alisi A, Byrne CD, Targher G. Complications, morbidity and mortality of nonalcoholic fatty liver disease. Metabolism. 2020; 111S:154170.
6. Zheng Z, Xu X, Zhang X, et al. Exposure to ambient particulate matter induces a NASH-like phenotype and impairs hepatic glucose metabolism an animal model. J Hepatol. 2013; 58:148–154.
7. Arciello M, Gori M, Maggio R, et al. Environmental pollution: a tangible risk for NAFLD pathogenesis. Int J Mol Sci. 2013; 14:22052–22066.
8. Tan HH, Fiel MI, Sun Q, et al. Kupffer cell activation by ambient air particulate matter exposure may exacerbate non-alcoholic fatty liver disease. J Immunotoxicol. 2009; 6:266–275.
9. Xu MX, Ge CX, Qin YT, et al. Prolonged PM2.5 exposure elevates risk of oxidative stress-driven nonalcoholic fatty liver disease by triggering increase of dyslipidemia. Free Radic Biol Med. 2019; 130:542–556.
10. Hsieh S, Leaderer BP, Feldstein AE, et al. Traffic-related air pollution associations with cytokeratin-18, a marker of hepatocellular apoptosis, in an overweight and obese paediatric population. Pediatr Obes. 2018; 13:342–347.
11. Pejhan A, Agah J, Adli A, et al. Exposure to air pollution during pregnancy and newborn liver function. Chemosphere. 2019; 226:447–453.
12. Vrijheid M, Slama R, Robinson O, et al. The human early-life exposome (HELIX): project rationale and design. Environ Health Perspect. 2014; 122:535–544.
13. Wright J, Small N, Raynor P, et al.; Born in Bradford Scientific Collaborators Group. Cohort profile: the Born in Bradford multi-ethnic family cohort study. Int J Epidemiol. 2013; 42:978–991.
14. Heude B, Forhan A, Slama R, et al.; EDEN mother-child cohort study group. Cohort profile: the EDEN mother-child cohort on the prenatal and early postnatal determinants of child health and development. Int J Epidemiol. 2016; 45:353–363.
15. Guxens M, Ballester F, Espada M, et al.; INMA Project. Cohort profile: the INMA–INfancia y Medio Ambiente–(environment and childhood) project. Int J Epidemiol. 2012; 41:930–940.
16. Grazuleviciene R, Danileviciute A, Nadisauskiene R, Vencloviene J. Maternal smoking, GSTM1 and GSTT1 polymorphism and susceptibility to adverse pregnancy outcomes. Int J Environ Res Public Health. 2009; 6:1282–1297.
17. Magnus P, Birke C, Vejrup K, et al. Cohort profile update: the Norwegian mother and child cohort study (MoBa). Int J Epidemiol. 2016; 45:382–388.
18. Chatzi L, Leventakou V, Vafeiadi M, et al. Cohort profile: the mother-child cohort in Crete, Greece (Rhea Study). Int J Epidemiol. 2017; 46:1392–1393k.
19. Maitre L, de Bont J, Casas M, et al. Human early life exposome (HELIX) study: a European population-based exposome cohort. BMJ Open. 2018; 8:e021311.
20. Tamayo-Uria I, Maitre L, Thomsen C, et al. The early-life exposome: description and patterns in six European countries. Environ Int. 2019; 123:189–200.
21. Beelen R, Hoek G, Vienneau D, et al. Development of NO2 and NOx land use regression models for estimating air pollution exposure in 36 study areas in Europe—the ESCAPE project. Atmos Environ. 2013; 72:10–23.
22. Eeftens M, Tsai M-Y, Ampe C, et al. Spatial variation of PM2.5, PM10, PM2.5 absorbance and PMcoarse concentrations between and within 20 European study areas and the relationship with NO2—results of the ESCAPE project. Atmos Environ. 2012; 62:303–317.
23. Eeftens M, Beelen R, de Hoogh K, et al. Development of Land Use Regression models for PM(2.5), PM(2.5) absorbance, PM(10) and PM(coarse) in 20 European study areas; results of the ESCAPE project. Environ Sci Technol. 2012; 46:11195–11205.
24. Beelen R, Hoek G, Pebesma E, Vienneau D, de Hoogh K, Briggs DJ. Mapping of background air pollution at a fine spatial scale across the European Union. Sci Total Environ. 2009; 407:1852–1867.
25. Cyrys J, Eeftens M, Heinrich J, et al. Variation of NO2 and NOx concentrations between and within 36 European study areas: results from the ESCAPE study. Atmos Environ. 2012; 62:374–390.
26. Rahmalia A, Giorgis-Allemand L, Lepeule J, et al.; EDEN Mother-Child Cohort Study group. Pregnancy exposure to atmospheric pollutants and placental weight: an approach relying on a dispersion model. Environ Int. 2012; 48:47–55.
27. Schembari A, de Hoogh K, Pedersen M, et al. Ambient air pollution and newborn size and adiposity at birth: differences by maternal ethnicity (the Born in Bradford Study Cohort). Environ Health Perspect. 2015; 123:1208–1215.
28. Wang M, Beelen R, Bellander T, et al. Performance of multi-city land use regression models for nitrogen dioxide and fine particles. Environ Health Perspect. 2014; 122:843–849.
29. Stratakis N, V Conti D, Jin R, et al. Prenatal Exposure to Perfluoroalkyl Substances Associated With Increased Susceptibility to Liver Injury in Children. Hepatology. 2020; 72:1758-1770.
30. Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology. 1999; 10:37–48.
31. de Onis M, Onyango AW, Borghi E, Siyam A, Nishida C, Siekmann J. Development of a WHO growth reference for school-aged children and adolescents. Bull World Health Organ. 2007; 85:660–667.
32. White IR, Royston P, Wood AM. Multiple imputation using chained equations: Issues and guidance for practice. Stat Med. 2011; 30:377–399.
33. Wood SN. Generalized Additive Models: An Introduction with R. CRC Press; 2017.
34. Conneely KN, Boehnke M. So many correlated tests, so little time! Rapid adjustment of P values for multiple correlated tests. Am J Hum Genet. 2007; 81:1158–1168.
35. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Series B Methodol. 1995; 57:289–300.
36. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2018.
37. Kelishadi R, Poursafa P. Obesity and air pollution: global risk factors for pediatric non-alcoholic fatty liver disease. Hepat Mon. 2011; 11:794–802.
38. Kim KN, Lee H, Kim JH, Jung K, Lim YH, Hong YC. Physical activity- and alcohol-dependent association between air pollution exposure and elevated liver enzyme levels: an elderly panel study. J Prev Med Public Health. 2015; 48:151–169.
39. Pan WC, Wu CD, Chen MJ, et al. Fine particle pollution, alanine transaminase, and liver cancer: a Taiwanese prospective cohort study (REVEAL-HBV). J Natl Cancer Inst. 2016; 108. doi: 10.1093/jnci/djv341.
40. Markevych I, Wolf K, Hampel R, et al. Air pollution and liver enzymes. Epidemiology. 2013; 24:934–935.
41. Dey T, Gogoi K, Unni B, et al. Role of environmental pollutants in liver physiology: special references to peoples living in the oil drilling sites of Assam. PLoS One. 2015; 10:e0123370.
42. Kim HJ, Min JY, Seo YS, Min KB. Association of ambient air pollution with increased liver enzymes in Korean adults. Int J Environ Res Public Health. 2019; 16:1213.
43. Werder EJ, Beier JI, Sandler DP, et al. Blood BTEXS and heavy metal levels are associated with liver injury and systemic inflammation in Gulf states residents. Food Chem Toxicol. 2020; 139:111242.
44. Buzzetti E, Pinzani M, Tsochatzis EA. The multiple-hit pathogenesis of non-alcoholic fatty liver disease (NAFLD). Metabolism. 2016; 65:1038–1048.
45. Sun Q, Yue P, Deiuliis JA, et al. Ambient air pollution exaggerates adipose inflammation and insulin resistance in a mouse model of diet-induced obesity. Circulation. 2009; 119:538–546.
46. Xu X, Yavar Z, Verdin M, et al. Effect of early particulate air pollution exposure on obesity in mice: role of p47phox. Arterioscler Thromb Vasc Biol. 2010; 30:2518–2527.
47. Wahlang B, Jin J, Beier JI, et al. Mechanisms of Environmental Contributions to Fatty Liver Disease. Curr Environ Health Rep. 2019; 6:80–94.
48. Wieckowska A, Papouchado BG, Li Z, Lopez R, Zein NN, Feldstein AE. Increased hepatic and circulating interleukin-6 levels in human nonalcoholic steatohepatitis. Am J Gastroenterol. 2008; 103:1372–1379.
49. Schwarze PE, Ovrevik J, Låg M, et al. Particulate matter properties and health effects: consistency of epidemiological and toxicological studies. Hum Exp Toxicol. 2006; 25:559–579.

Supplemental Digital Content

Copyright © 2021 The Authors. Published by Wolters Kluwer Health, Inc. on behalf of The Environmental Epidemiology. All rights reserved.