Selection of controls for family studies. Biases and implications.

The use of hospital-based controls as a means of comparison for family studies has raised a number of questions concerning the degree to which the control group represents the general population. Of primary interest is whether the control group should be completely free of disease-associated symptoms. To test the effect of the selection of psychiatric symptom-free controls, we examined the control group from our long-term follow-up and family study of major psychoses. Morbidity risks for psychiatric disorder among relatives of surgical controls were estimated, both for an unscreened proband group of 361 patients and for a screened group of 318 from which the relatives of controls with significant "psychiatric history" had been eliminated. The largest difference occurred with affective disorders, which were more frequent among the relatives of unscreened controls than among the relatives of the screened controls. Our results suggest that gathering data on both screened and unscreened control groups will yield more generalizable results than either alone.