A MARGINAL REGRESSION MODELING FRAMEWORK FOR EVALUATING MEDICAL DIAGNOSTIC-TESTS

Technological advances continue to develop for early detection of dise ase. Research studies are required to define the statistical propertie s of such screening or diagnostic tests. However, statistical methodol ogy currently used to evaluate diagnostic tests is limited. We propose the use of marginal regression models with robust sandwich variance e stimators to make inference about the sensitivity and specificity of d iagnostic tests. This method is more flexible than standard methods in that it allows comparison of sensitivity between two or more tests ev en if all tests are not carried out on all subjects, it can accommodat e correlated data, and the effect of covariates can be evaluated, This last feature is important since it allows researchers to understand t he effects on sensitivity and specificity of various environmental and patient characteristics. If such factors are under the control of the clinician, it provides the opportunity to modify the diagnostic testi ng program to maximize sensitivity and/or specificity. We show that th e marginal regression modelling methods generalize standard statistica l methods. In particular, when we compare two screening tests and we t est each subject with both screens, the method corresponds to McNemar' s test. We describe data from an ongoing audiology screening study and we analyse a simulated version of the data to illustrate the methodol ogy. We also analyse data from a longitudinal study of PCR as a diagno stic test for cytomegalovirus. (C) 1997 by John Wiley & Sons, Ltd.