Ozone and risk of COPD in the USA: demographic insights from the All Of Us Research Program Authors: Shelton Lo, Gabriel A. Goodney, Hantao Wang, Jungeun Lim, Susan Valerie Czach, Jared A. Fisher, Maryam Hashemian, Véronique L. Roger, Rena R. Jones, Jason Y.Y. Wong Published: 2026 ABSTRACT ExtractWhile smoking is the prominent risk factor for COPD, environmental exposures play a significant aetiological role in both smokers and never-smokers in the USA [1, 2]. Furthermore, COPD rates vary among population subgroups with different exposure profiles [3]. Reported associations between ozone exposure and COPD have been inconsistent [4], and there is limited understanding of the interrelationships between ozone, atmospheric conditions, and the air exposome, which partly explain subgroup differences. To the Editor: While smoking is the prominent risk factor for COPD, environmental exposures play a significant aetiological role in both smokers and never-smokers in the USA [1, 2]. Furthermore, COPD rates vary among population subgroups with different exposure profiles [3]. Reported associations between ozone exposure and COPD have been inconsistent [4], and there is limited understanding of the interrelationships between ozone, atmospheric conditions, and the air exposome, which partly explain subgroup differences. Additionally, whether ozone exposure consistently below the current US National Ambient Air Quality Standard (70 ppb average per 8 h) is associated with increased COPD risk remains unclear, especially among subgroups. We addressed these knowledge gaps using data from the All of Us Research Program (AoU), a prospective mega-cohort of >864 000 US adults [5]. Here, we investigated the complexities surrounding the respiratory effects of ozone and provide greater precision regarding which subgroups might be more susceptible to COPD, which could inform risk stratification and targeted exposure reduction efforts. AoU was previously described [5]. Among 633 547 participants followed from 2017 to 2023, we analysed 376 535 participants who met inclusion criteria and data requirements. Annual average ozone concentrations derived from satellite-based measurements (2000–2016) were linked to residential addresses. Air monitoring data were obtained from the National Aeronautics and Space Administration's Socioeconomic Data and Applications Center and Earth Observing System Data and Information System. We extracted high-resolution, well-validated predictions generated using ensemble models for daily and annual concentrations of PM2.5 (particulate matter with a maximum diameter of 2.5 μm; in μg·m−3), ozone (ppb), and NO2 (ppb) at the zip code level, and then calculated the concentrations of each pollutant for each three-digit postal region [6, 7]. We extracted temperature values from National Oceanic and Atmospheric Administration – National Centers for Environmental Information databases. Environmental exposures were assigned based on residential duration and participants lacking linkage data were excluded from analyses. Incident COPD was identified via linked electronic health records (EHRs) using validated ICD-10 (J44) and SNOMED CT (13645005) codes. We estimated hazard ratios and 95% confidence intervals for incident COPD in relation to ozone concentrations using multivariable Cox regression models. We adjusted for age at enrolment (years), sex (male, female), race (white, black or African American, Asian, other), Hispanic status, income (USD 150 000), educational attainment (high school graduate or less, some college, college graduate, or advanced degree or higher), body mass index (kg·m−2), smoking status (never- or ever-smoker), smoking pack-years and PM2.5 concentrations (µg·m−3). Heterogeneity among subgroups was evaluated using stratified and multiplicative interaction analyses. Joint effects of ozone, NO2 and temperature were assessed with cross-product terms. Cell weighting using the 2018–2022 American Community Survey data was used for sensitivity analyses to account for non-probabilistic sampling. Among 376 535 participants, 7907 incident COPD cases were identified during a mean 4-year follow-up. Ozone concentrations (median 37.75 ppb, range 29.79–51.39 ppb) were similar across subgroups. We found that higher ozone concentrations were associated with increased COPD risk, even at exposure levels consistently below the current US regulatory limit. Overall, we observed a positive, non-monotonic relationship between ozone and COPD risk (figure 1a and b). The association was most apparent at higher ozone concentrations (quartile (Q)3 versus Q1 HR 1.02, 95% CI 0.95–1.11; Q4 versus Q1 HR 1.09, 95% CI 1.01–1.18). Our findings remained robust after adjusting for co-exposures and atmospheric conditions, and correcting for volunteer bias from non-probabilistic sampling. We found evidence that temperature strengthened the association (p-value for interaction=3.53×10−8). By integrating additional environmental data, we provide greater understanding of the air exposome and COPD risk. Associations between ozone concentrations and incident COPD risk in the All of Us Research Program. a) Overall associations between ozone exposure and incident COPD risk. b) Associations between ozone exposure and incident COPD modelled as a continuous exposure (per 3 ppb increase). c) Associations between ozone exposure and incident COPD risk across joint categories of sex and smoking status, with men in the lowest ozone quartile as the reference group. d) Associations between ozone exposure and incident COPD risk across joint categories of sex and race, with men in the lowest ozone quartile as the reference group. e) Associations between ozone exposure and incident COPD risk across joint categories of sex and age group, with men in the lowest ozone quartile as the reference group. All models are based on Cox proportional hazards analyses adjusted for age, sex, race, Hispanic status, body mass index, smoking status, smoking pack-years, annual income, education level, and annual PM2.5 (particulate matter with a maximum diameter of 2.5 μm) levels. Ozone concentration quartiles: Q1 29.79 to ≤35.41 ppb; Q2 >35.41 to ≤37.75 ppb); Q3 >37.75 to ≤40.41 ppb); Q4 >40.41 to 51.39 ppb. The associations between ozone exposure and incident COPD varied across subgroups (figure 1c–e) when using males in the lowest ozone quartile as the reference. Stronger ozone–COPD associations were observed for male never-smokers, with men in the highest ozone quartiles having the highest COPD risks (Q4 HR 1.30, 95% CI 1.05–1.60). However, trends for never-smoking females were unclear. Among ever-smokers, males in the highest ozone quartiles had moderately increased risks (Q3 HR 1.13, 95% CI 1.01–1.26; Q4 HR 1.17, 95% CI 1.05–1.31). However, associations among ever-smoking women were less apparent (Q3 HR 1.10, 95% CI 0.99–1.23; Q4 HR 1.08, 95% CI 0.97–1.22). When stratified by race and sex, black or African American participants had more pronounced ozone–COPD associations than white participants, particularly at higher ozone quartiles, with the strongest association observed among black or African American men (Q4 HR 1.49, 95% CI 1.21–1.84). Age-specific analyses indicated stronger associations among adults ≤65 years, most notably among individuals aged 20–44 years in both men (Q4 HR 1.75, 95% CI 1.22–2.50) and women (Q4 HR 1.58, 95% CI 1.12–2.25). The most striking ozone–COPD association was observed in male never-smokers. Environmental and occupational epidemiology studies have shown that the respiratory effect of smoking is so strong that it can mask the contribution of environmental exposures [8]. Therefore, analyses of never-smokers allow for a clearer examination of the ozone–COPD relationship [9, 10]. Building on recent COPD research on non-smoking risk factors, we provide important new insights into the respiratory effects of ozone that support the need for greater focus on COPD risk in never-smokers [11]. Specifically, there is a notable research gap on COPD among male never-smokers, highlighting the need for comprehensive understanding of the disease across sexes [12]. This study has notable strengths. AoU is a large, demographically varied cohort with extensive EHR and questionnaire data. Targeted recruitment of participants across demographic and geographic strata enabled effective subgroup analyses. The integration of advanced satellite-based air monitoring and atmospheric data improved the accuracy of exposure assessments over time, enhancing the reliability of our findings. However, our study had some limitations. First, residual and unmeasured confounding may be present; therefore, the non-monotonic exposure–response relationship could reflect random variation or residual confounding rather than a true biological effect. However, we performed rigorous adjustments for potential confounders based on prior studies and conducted comprehensive sub-analyses to assess the influence of residual confounding from smoking, co-exposures and others. Second, the use of three-digit zip codes may have introduced exposure misclassification by masking the spatial heterogeneity of ozone. However, ozone displays lower spatial heterogeneity than PM2.5 and NO2 [13–15]. Third, ozone exposure may not have been measured within the aetiological window for COPD development. However, the cohort's residential stability and our use of long-term averaged pollutant estimates likely captured aetiological windows relevant to COPD onset. Fourth, ozone exposure was estimated based on residential location at three-digit zip codes and historical satellite data; this may not truly reflect individual-level exposures or conditions during follow-up because there may be individual-level variations in exposure due to commuting, indoor environments, or occupational settings. Furthermore, the 4-year follow-up period may limit the ability to detect long-term effects of chronic ozone exposure on COPD risk, and the reliance on EHR for COPD diagnosis may lead to under-ascertainment of cases. Finally, the non-probabilistic, volunteer-based cohort may limit generalisability, and multiple subgroup and interaction analyses may raise the possibility of chance findings. In summary, we found a robust relationship between residential ozone concentrations and COPD risk. The observed associations were more prominent among certain demographic subgroups, especially male never-smokers. We observed significant respiratory effects even at ozone concentrations below current US regulatory limits, suggesting that current thresholds may require re-evaluation. This has critical implications for informing targeted air quality interventions aimed at reducing exposure and disease risks. However, caution is warranted when interpreting these findings. We cannot discount the possibility of statistical power issues for some demographic groups. Future studies in well-powered, population-based cohorts are needed to replicate these findings among granular populations.