The research team, led by investigators including Cornelia Santoso, Zhebin Yu and Olena Gruzieva, set out to address a gap in the literature. Environmental exposures such as air pollution, transportation noise and residential greenness are increasingly recognised as determinants of mental, cardiovascular and metabolic health, with stress-related mechanisms thought to play a key role. Yet most previous studies of environmental exposures and cortisol have focused on children, adolescents, pregnant women or specific subgroups, and only a few have examined healthy adults. Prior findings have also been inconsistent: some studies reported elevated salivary cortisol among people living near airports, while a systematic review suggested a link between road traffic noise and increased urinary or salivary cortisol, but more recent studies have not found significant associations. Evidence on air pollution has been similarly mixed, with one study linking high ozone exposure to increased serum morning cortisol and PM2.5 and NO2 to lower levels, while others found no association at all.

The study drew on the BAMSE birth cohort, which enrolled 4,089 children born in Stockholm County between 1994 and 1996. Participants were followed repeatedly through infancy, childhood, adolescence and young adulthood with questionnaires and clinical examinations. At the 24-year follow-up, 3,064 participants, 75 percent of the baseline cohort, answered a questionnaire and 2,270 attended a clinical examination where saliva sampling kits were handed out. From the 1,588 participants who provided morning or evening saliva samples, the final analytical sample comprised 1,148 individuals after exclusions for missing exposure and covariate data; 1,140 contributed to morning cortisol analyses and 1,143 to evening analyses. All participants provided written informed consent, and ethical approvals were obtained from the Regional Ethics Review Board at Karolinska Institutet.

Exposure assessment was unusually detailed for a study of this kind. Annual average exposures during the year preceding the clinical assessment were estimated at each participant’s geocoded residential address, accounting for residential mobility by calculating time-weighted averages across all reported addresses. Air pollution levels of PM2.5, PM10, black carbon and nitrogen oxides were modelled using Gaussian dispersion modelling within the Airviro system, drawing on an emission inventory covering road and ferry traffic, industrial zones and residential areas, along with meteorological inputs and a street canyon effect for addresses in the most polluted sections of Stockholm. Road traffic noise was estimated with validated Nordic-based approaches incorporating 3D terrain data, road networks, traffic volumes and speed limits, expressed as free-field Lden at the most exposed building façades. Residential greenness was measured as the mean Normalized Difference Vegetation Index, or NDVI, within 100-metre, 300-metre and 500-metre circular buffers around each home, based on summer satellite images.

Saliva samples were collected at home: a morning sample shortly after waking, within 15 minutes, and an evening sample at bedtime, both taken before any dental cleaning using sterile Salivette swabs. Samples were mailed to the laboratory, centrifuged, stored at minus 80 degrees Celsius and analysed with a radioimmunoassay kit, with all samples from each participant processed in the same batch. Because cortisol levels were not normally distributed, the researchers applied a natural logarithmic transformation and used linear regression models adjusted for age, sex, education, tobacco use, second-hand tobacco exposure, body mass index and season of sampling, estimating associations per interquartile range increase in each exposure. Exploratory post hoc subgroup analyses examined modification by sex, season, tobacco use, BMI category, history of depression, asthma, rhinitis and eczema, with Benjamini-Hochberg false discovery rate correction applied to interaction tests for sex and season.

The headline result was negative. No significant overall associations were observed between any of the air pollutants, road traffic noise, or NDVI and morning or evening salivary cortisol levels. The findings were robust across a battery of sensitivity analyses: results remained unchanged after excluding participants who used inhaled or oral steroids, those whose evening cortisol exceeded their morning level, and those with unreliable cortisol values, and additional adjustment for area-level mean income or physical activity had only a minor impact. A multi-exposure model mutually adjusting for PM2.5, road traffic noise and lack of greenness produced the same picture, and a quantile g-computation model assessing the joint effect of the full mixture of PM2.5, PM10, NOx, black carbon, road traffic noise and lack of greenness found no clear associations with either morning or evening cortisol.

The exploratory subgroup analyses told a more textured story. Among males, higher morning cortisol was associated with PM10, with a beta of 0.058 per interquartile range increase, NOx, with a beta of 0.070, and road traffic noise, with a beta of 0.120, while greenness showed inverse associations, reaching a beta of minus 0.165 for NDVI within 500 metres. Similar trends appeared for evening cortisol. In females the directions were reversed: NDVI was positively associated with morning cortisol, particularly within 100 metres, and the sex-by-exposure interaction terms for NDVI carried FDR-adjusted p-values ranging from 4.3 times ten to the minus five to 0.001. The exposure mixture was associated with increased morning cortisol in males, with a beta of 0.084 per quintile increase across all exposures, driven mainly by NOx, lack of greenness, PM10 and road traffic noise, while the effect in females was close to null. Stratified by season, NDVI was significantly positively associated with higher evening cortisol during summer, particularly within 100 metres, but no seasonal pattern emerged for air pollutants or noise, and after FDR correction none of the season interaction terms remained significant.

The authors themselves urge caution in reading these subgroup signals. They note that statistically significant associations within individual subgroups should not necessarily be interpreted as evidence of significant differences between subgroups, and that for evening cortisol and the season-stratified analyses the confidence intervals generally overlapped between groups. Evidence for sex-specific differences was strongest for greenness and morning cortisol, where estimates differed in direction and the confidence intervals for males and females did not overlap. Given the large number of analyses performed, chance findings cannot be excluded, and the researchers describe the subgroup results as exploratory and hypothesis-generating, requiring replication in independent studies. They also observed that reported annoyance to road traffic noise increased with higher modelled noise levels, with 19.7 percent of participants moderately to extremely annoyed at 55 dB Lden or above compared with 5.5 percent below 45 dB, but the cross-sectional design and the absence of a formal mediation analysis preclude conclusions about whether annoyance mediates any physiological effect.

Several limitations frame the interpretation. Exposure assessment relied on residential addresses and did not capture workplaces, universities, commuting or other locations, and annual average exposure indicators may not reflect the short-term exposures immediately preceding saliva sampling. Only two saliva samples from a single day were available, so the measures reflect momentary HPA-axis activity rather than a true diurnal rhythm, and the authors suggest that future studies should incorporate repeated sampling and longer-term biomarkers such as hair cortisol. Data on psychosocial stressors such as home environment and working conditions were lacking, so residual confounding cannot be excluded, and the cross-sectional design limits causal inference. Selection bias is also possible: compared with the full 24-year follow-up, the analysed sample had a lower prevalence of tobacco use, obesity and depression and more favourable sleep characteristics, which may have reduced variability and contributed to the predominantly null findings, limiting generalisability to young adults with poorer health profiles.

For now, the study’s main contribution is methodological and cautious rather than alarmist. By combining individually modelled exposures to multiple air pollutants, road traffic noise and greenness with morning and evening cortisol measurements in a population-based cohort of young adults, and by examining both single exposures and their joint mixture effects, the researchers provide one of the most comprehensive assessments to date of how the residential environment relates to a physiological stress marker in this age group. Their conclusion is measured: no significant overall associations between air pollution, road traffic noise, greenness and salivary cortisol, with possible sex-specific differences and seasonal patterns that warrant further investigation but should be interpreted cautiously given their exploratory nature.

Subject of Research: Associations of air pollution, road traffic noise, and residential greenness with salivary cortisol in young adults

Article Title: Air pollution, road traffic noise, and greenness in relation to salivary cortisol in young adults

Article References: Santoso, C., Yu, Z., Bahram, F., Pyko, A., Ögren, M., Eneroth, K., Lõhmus, M., Merritt, A.-S., Kull, I., Melén, E., Bergström, A., Pershagen, G., Eriksson, C., & Gruzieva, O. (2026). Air pollution, road traffic noise, and greenness in relation to salivary cortisol in young adults. Journal of Exposure Science & Environmental Epidemiology. https://doi.org/10.1038/s41370-026-00995-z

Image Credits: AI Generated

DOI: 10.1038/s41370-026-00995-z

Keywords: air pollution, road traffic noise, greenness, salivary cortisol, stress biomarker, BAMSE birth cohort, young adults, sex differences, NDVI, HPA axis, PM10, NOx