American journal of epidemiology
-
Although clinic-based cohorts are most representative of the "real world," they are susceptible to loss to follow-up. Strategies for managing the impact of loss to follow-up are therefore needed to maximize the value of studies conducted in these cohorts. The authors evaluated adult patients starting antiretroviral therapy at an HIV/AIDS clinic in Uganda, where 29% of patients were lost to follow-up after 2 years (January 1, 2004-September 30, 2007). ⋯ DAGs encoding knowledge about the data-generating process, including the fact that death is a cause of being classified as lost to follow-up in this setting, revealed "collider" bias in the unweighted and IPCW approaches. In a clinic-based cohort in Africa, unweighted and IPCW approaches-which rely on the "missing at random" assumption-yielded biased estimates. A sampling-based approach can in general strengthen epidemiologic analyses conducted in many clinic-based cohorts, including those examining other diseases.
-
In metropolitan areas, road traffic is a major contributor to ambient air pollution and the dominant source of community noise. The authors investigated the independent and joint influences of community noise and traffic-related air pollution on risk of coronary heart disease (CHD) mortality in a population-based cohort study with a 5-year exposure period (January 1994-December 1998) and a 4-year follow-up period (January 1999-December 2002). Individuals who were 45-85 years of age and resided in metropolitan Vancouver, Canada, during the exposure period and did not have known CHD at baseline were included (n = 445,868). ⋯ After adjustment for potential confounders, including traffic-related air pollutants or noise, elevations in noise and black carbon equal to the interquartile ranges were associated with 6% (95% confidence interval: 1, 11) and 4% (95% confidence interval: 1, 8) increases, respectively, in CHD mortality. Subjects in the highest noise decile had a 22% (95% confidence interval: 4, 43) increase in CHD mortality compared with persons in the lowest decile. These findings suggest that there are independent effects of traffic-related noise and air pollution on CHD mortality.