Weiss , NS . Professor, Department of Epidemiology, School of Public Health, University of Washington; and Member, Fred Hutchinson Cancer Research Center, Seattle, WA

Commentary on: Alan Morrison . Case definition in case-control studies of the efficacy of screening . Am J Epidemiol 1982;115:6-8

Cite as: Weiss NS . Morrison's “Case definition in case-control studies of the efficacy of screening”, revisited . The People's Epidemiology Library (www.epidemiology.ch). Accessed: DATE
© Noel Weiss, 2009.

Author contact details: Noel S. Weiss, University of Washington , Box 357236 , Seattle , WA 98195 , USA . E-mail: nweiss@u.washington.edu

Screening can reduce the risk of death from a given form of cancer by leading to the diagnosis of the cancer (or a pre-malignant lesion) at a stage at which it has an increased likelihood of being cured. A number of studies have documented that a particular mode of cancer screening was the means of diagnosis of early-stage cancers more often than it was of late-stage cancers of the same organ. This suggests that the screening modality is sensitive in identifying early cancer. However, whether such screening could reduce mortality also requires that treatment of the screen-detected lesion be more effective than no treatment (or than treatment administered later in the natural history of that cancer, when it would have been diagnosed in the absence of screening). Yet for obvious reasons, there are essentially no studies among patients with early cancer that compare mortality between those who are given prompt therapy and those given delayed or no therapy.

Randomized trials of screening directly assess the impact of screening on cancer mortality. For example, a lower rate of death from breast cancer in women assigned to a program of mammography and clinical examination of the breast than in women not so assigned argues that the screening program can identify breast cancer relatively early and that early treatment is relatively effective. However, data from randomized trials are not available to address all questions pertaining to cancer screening efficacy.

Alan Morrison's contribution to this area of research (Am J Epidemiol 1982;115:6-8) was to describe the structure of a case-control study that, in the absence of confounding and the other biases that often beset non-randomized studies, could provide the same information as a randomized trial of screening. The focus of his paper was on the definition of a “case” group. Rather than incident cases of the cancer in question, Morrison explained the need to restrict attention to just those cases that screening sought to prevent, i.e. persons who died as a result of the cancer. He recognized that for those forms of screening for which records would not be easily (or at all) available (e.g. breast or skin self-exam), it would be necessary to obtain screening histories directly from persons with cancer whose illness was likely to later prove fatal, and he expanded his recommendation accordingly.

In the nearly three decades since Morrison published this article, a large number of case-control studies of the efficacy of cancer screening have been conducted. Also, considerable thought has been given to other methodologic issues that arise in case-control studies of screening, particularly to control selection and to the ascertainment and characterization of screening history. To a great extent, Morrison's paper served as a stimulus for this further work. He took the first steps and pointed the way, and then the rest of us followed once we saw that there was a path on which to travel.