BACKGROUND: In occupational epidemiology, it is common to use cross-sectional studies comparing exposed with matched non-exposed populations. When several such past studies are available within a relatively short time frame, it may be possible to use their control groups to reduce the number of controls in a new study and thus reduce its cost. METHODS: We compare existing non-exposed populations within a standard random-effects model, to explore the natural between population variability and to compute the power increase when including these populations. An example is given from respiratory epidemiology, with the forced expiratory volume in 1 s as the outcome, using 13 previously-collected control populations obtained in recent studies. RESULTS: We show that the reduction in the number of controls needed for a new study is equal to the ratio of the between subjects and the between group variance, observed in controls in previous studies. In the present example, the reduction in the number of controls needed for a new study was found to be 18. An optimal allocation of resources is obtained in this framework. CONCLUSIONS: The use of past available control groups provides a significant reduction in cost only if they are very homogeneous, compared to the between-subject variance.
BACKGROUND: In occupational epidemiology, it is common to use cross-sectional studies comparing exposed with matched non-exposed populations. When several such past studies are available within a relatively short time frame, it may be possible to use their control groups to reduce the number of controls in a new study and thus reduce its cost. METHODS: We compare existing non-exposed populations within a standard random-effects model, to explore the natural between population variability and to compute the power increase when including these populations. An example is given from respiratory epidemiology, with the forced expiratory volume in 1 s as the outcome, using 13 previously-collected control populations obtained in recent studies. RESULTS: We show that the reduction in the number of controls needed for a new study is equal to the ratio of the between subjects and the between group variance, observed in controls in previous studies. In the present example, the reduction in the number of controls needed for a new study was found to be 18. An optimal allocation of resources is obtained in this framework. CONCLUSIONS: The use of past available control groups provides a significant reduction in cost only if they are very homogeneous, compared to the between-subject variance.