No single test is perfect and without false-negative and/or false-positive results. Consequently, the clinician is perpetually confronted with incertitude about the true disease state of the patient. In oncology, these diagnostic errors may have harmful consequences for the patient. It is, therefore, imperative that the clinician knows how often these errors occur, which implies a quantitative evaluation of a test. With this knowledge, the test result must subsequently be interpretated within the clinical framework. Bayes' theorem provides a simple and useful mathematical model for the integration of measures of test performance and clinical data. Traditionally, sensitivity and specificity are used to describe test performance. However, this approach requires that the conclusion of the test is dichotomised into 'normal' and 'abnormal'. Few tests have a natural binary outcome. A test parameter that is applicable to all types of test outcome scales and, at the same time, provides the opportunity to determine the gain in diagnostic information by applying Bayes' theorem, is therefore mandatory. The likelihood ratio meets these conditions. The application of this concept for both the evaluation and the interpretation of various types of tests used in cancer patients is demonstrated.