Natural experiments and large databases in respiratory and cardiovascular disease

A number of scientific questions cannot be tested in a laboratory, clinic or clinical trial setting. In many cases, observational data can be used to test such hypotheses. This article illustrates how epidemiology can contribute and shows the different ways of using observational data through three approaches: 1) prospective cohort study design; 2) time series analysis; and 3) a nested case–control design in pharmacoepidemiology. In a prospective cohort study design, three cohorts were merged to study lung function decline, testing the importance of different trajectories of lung function decline for developing chronic obstructive pulmonary disease (COPD). Using these three well-described cohorts it was documented that maximally attained lung function in early adulthood is as important as excess decline in forced expiratory volume in 1 s for the development of COPD. Time series analysis is used to examine exposures and disease over time. In a recent review of cardiovascular disease some interesting associations, and not least lack of associations, were presented. Assessing effects of drugs in database studies is challenging. In a nested case–control design in a large cohort study, statins were found to reduce the risk of COPD exacerbations. These findings will be discussed. Observational data from large databases, as well as carefully collected data in cohort studies, can be used to test hypotheses that may not be addressed in a traditional experimental setting.