| First author [ref.]/ location | Year of study/ population/ age at observation | Study design/ follow-up | Markers of exposure to air pollution | Definition of COPD outcome | Effect estimate (95% CI) | Study limitation |
| Finkelstein [24]/Hamilton, Canada | 1985 and 1999/5228 patients referred to pulmonary laboratory | Cohort/9 years | Pollutants: none Traffic proximity: distance to nearest major road | ICD-9 codes for COPD mortality; measures of FEV1 and FVC | Rate of advancement period for living close to a major road: 3.4 (0.8–6.0) | No direct pollution measurements, only GIS data was used to assign air pollution exposure |
| Lepeule [25]/USA | Baseline: 1974–1977 Last follow-up: 2009/8096 subjects from the Harvard Six Cities study | Cohort/20 years (death from 1979–2009) | Pollutants: local centrally measured levels of PM10 and PM2.5 Traffic proximity: none | ICD-10 codes for COPD mortality; all patients underwent spirometry, measures of FEV1 and FVC | Adjusted RR for 10 μg·m−3 increase in PM2.5: never-smoker, 0.85 (0.36–2.02); former smoker, 1.64 (0.92–2.93); current smoker, 1.10 (0.74–1.62) | Only adjustment for baseline factors, PM2.5 was not measured in same location throughout the study period |
| Naess [26]/Oslo, Norway | 1992–1998/143 842 subjects from the general population of Oslo/51–70 and 71–90 years | Cohort/14 years | Pollutants: dispersion model of NO2, PM10 and PM2.5 to calculate individual daily average exposure estimates Traffic proximity: None | ICD-9 codes 490–496 (COPD) and ICD-10 codes I00–I119 for COPD mortality Limitation: no lung function measures | Age group 51–70 years: HR for highest quartile >42 μg·m−3 NO2: 1.21 (1.05–1.39); PM10: 1.29 (1.12–1.48); PM2.5: 1.27 (1.11–1.47) | Study based on registry data only, no information about confounders such as smoking |
| Pope [27]/USA | 1979–1983 and 1999–2000, 16-year follow-up/500 000 subjects from the general population | Cohort | Pollutants: background measures of PM2.5 Traffic proximity: none | ICD-10 codes I00–I119 for COPD mortality Limitation: no lung function measures | RR for increase of 10 μg·m−3 PM2.5: 0.84 (0.77–0.93) | No lung function, coding from death records only |
| Yorifuji [16]/Shizuoka, Japan | 1999–2006/14 001 subjects from the general population | Cohort/7 years | Pollutants: NO2 modelled (LUR models) Traffic proximity: none Pollutants: daily monitoring measures to create yearly averages of PM10 Traffic proximity: none | ICD-10 codes for COPD mortality Mortality (case fatality) Limitation: no lung function measures | Adjusted HR for 10 μg·m−3 increase in NO2: 1.11 (0.78–1.56); PM10 in older subjects: 1.14 (1.12–1.16); younger subjects: 1.11 (1.08–1.13) | No additional information about confounding available, study Population was defined by COPD admission, but no information on smoking or other confounders available |
| Zanobetti [15]/USA | 1985–1999/1 039 287 hospital discharges of patients hospitalised with COPD (ICD-9 code 491;492;494–496 used to define COPD)/≥65 years | Cohort/4 years | Pollutants: daily monitoring measures to create yearly averages of PM10 Traffic proximity: none | Mortality (case fatality) Limitation: no lung function measures | HR for 10 μg·m−3 increase in PM10 in older subjects, 1.14 (1.12–1.16); younger subjects, 1.11 (1.08–1.13) | Population was defined by COPD admission, but no information on smoking or other confounders available |
ICD: International Classification of Diseases; PMx: particular matter with an aerodynamic cut-off diameter of ≤x μm; LUR: land use regression; FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; RR: relative risk; HR: hazard ratio; GIS: geographic information system.