The Effects of Sensitivity and Specificity of Case Selection on Validity, Sample Size, Precision, and Power in Hospital-Based Case-Control Studies

The consequences of imperfect sensitivity and specificity in disease diagnosis in epidemiologic studies have conventionally been assessed by models of misclassification which assume a fixed number of study participants. This assumption is not usually applicable to case-control studies in which disease diagnosis is part of the case selection process and sensitivity and specificity will, for a given time period and source of cases, affect the size of the case group. In this paper, a mathematical model that incorporates this is developed in the framework of a hospital-based case-control study. The separate and combined effects of imperfect sensitivity and specificity of case diagnosis on validity, sample size, precision, and power are assessed. The authors conclude that if several diagnostic procedures are available, specificity of case diagnosis should usually take precedence over sensitivity for the sake of validity. Although increasing specificity and sacrificing sensitivity may compromise precision to some extent, the latter can often be fully compensated for by an increased control:case ratio. Imperfect specificity also compromises power despite increased sample size. Since clinical diagnoses tend to focus on high sensitivity and sacrifice some specificity, their uncritical adoption for case recruitment in case-control studies may compromise their validity.