[Criteria for the selection and evaluation of diagnostic tests]

When evaluating the result of a diagnostic test with respect to a reference interval, two additional sources of variability should be taken into account, i.e. the variability of the analytical procedure and the biological variability which comprises an intraindividual and an interindividual component. The test assessment chart presented in this paper is a graphical representation allowing for the simultaneous evaluation of all three variables. In particular, the test assessment chart yields information relating to the effectiveness of a test in early detection of a disease. Moreover, the types of errors inherent in a test may be defined, i.e. either errors of the first type or those of the second type, leading to false positive and false negative results, respectively. The test assessment chart also shows whether repeated measurements improve the result, and whether a test basically is uncertain. Diagnostic sensitivity, specifity, and predictive values of a test may be arrived at if a threshold of discrimination is introduced. An example is discussed relating serum creatinine concentration with glomerular filtration rate. This approach reduces quantitative results to binary ones (positive/negative). More advanced methods of estimating the probability of the existence of a disease have come into use more recently. This bases on the likelihood quotients calculated (class-wise) from test results of well defined groups of patients (and healthy individuals). Examples of calculating the risk for gout attack and of EPH gestosis, depending on serum uric acid level, is presented. The application of the methods for diagnostic support today is restricted, so far, to certain special classes fo clinical problems.