Keywords: diagnostic test, decision theory, expected utility, relative utility, net benefit, weighted accuracy, decision curve, test trade-off
In disease interception, a patient is treated before disease is clinically detectable. The patient may either have the disease at a preclinical disease phase or may be healthy but at high risk for developing the disease in the future. The decision of whether or not to intercept the disease with treatment may depend on the result from a diagnostic test for early detection or risk prediction. Merely evaluating the test for diagnostic accuracy would not address the clinical consequences of making treatment decisions based on the test result. Unfortunately, clinical trials in which patients are randomized to either an investigational treatment for disease interception or a standard of care may be prohibitive to conduct due to expense, follow-up time required to obtain patient outcomes, and difficulty in designing such trials to separate test performance from treatment effect. Recently, clinical utility measures have been developed that can be helpful for evaluating the clinical consequences of using a test result for treatment decision making without having to conduct a burdensome clinical trial. Some measures depend on the relative importance r of a false positive to a false negative testing error. Others depend on a patient preference called the risk threshold T, the risk at which a patient is indifferent to being treated or not. An intriguing measure due to Baker is the test trade-off, the minimum number of tests per true positive test result at which a test has positive net benefit at a particular risk threshold T. We discuss some of these measures and illustrate them with specific examples to evaluate tests for treatment decision making in disease interception.