Hospital versus random digit dialing controls in the elderly. Observations from two case-control studies.

Finding an optimal source of controls is a major consideration in the design of case-control studies. While hospital patients can provide a relatively economical and convenient source of controls, hospital controls may have diseases independently associated with the exposures being studied. Results of a recent case-control study of multiple myeloma suggest that the problems of using hospital controls may be exacerbated in studies of diseases affecting the elderly, especially when considering risk factors which might be associated with chronic diseases. Apparently, the authors have encountered a modern analog of "Berkson's bias" in which cancer cases are referred to a tertiary care center for a single disease. On the other hand, noncancer patients who might serve as controls tend to be referred selectively if they have multiple diseases. This article reports some of the difficulties encountered in a case-control study whose mean case age was 63.4 years and suggests that community controls selected by random digit dialing may be preferable to hospital controls for hypothesis-generating case-control studies of diseases in the elderly.