[The use of laboratory data].

Medical laboratory data are gained by scientific methods. The method-inherent uncertainties always produce an "inexactitude" of varying degree. The physician who uses these data as a support for his decisions should therefore be able to estimate the reliability of the laboratory data; this can be assessed by statistical methods. The most fundamental characteristics of all analytical methods are analytical inaccuracy and analytical imprecision. The limit of detection and the limits of the measuring interval are functions of these figures. A number of illustrations demonstrate the interdependence of analytical performance and of medical significance of laboratory data. Laboratory data, on the other hand, are--to put it simply--used to distinguish diseased from NON-diseased individuals. The "diagnostic validity" of a certain test for a certain medical problem is determined by comparing appropriate collectives. The selection of these collectives and the choice of the cut-off value determine diagnostic sensitivity and specificity respectively. From these figures and prevalence (or "pre-test probability"), the predictive values of a positive (or negative) result can be calculated. The significance of the position of the cut-off value and of prevalence is demonstrated by a number of examples. If the results are not given in a binary form (positive/negative) but in different grades (slightly, moderately, strongly elevated/depressed), and if the results of several tests must be judged in a multivariate manner, the number of theoretically possible patterns can become extremely large. Various approaches designed to overcome these problems are discussed.