[Methodology of diagnostic validation studies. Errors in planning and analysis].

Abstract

The evaluation of diagnostic tests play an important role in clinical research. Though detailed standards have been developed for treatment evaluation, widely accepted methodological standards in diagnostic test research are still lacking. Following recent methodological research German biostatisticians usually distinguish between 4 types of diagnostic validation studies (Phase I to IV). The diagnostic value of a test can be described by indexes like sensitivity, specificity and predictive values. Sensitivity and specificity are so-called nosological probabilities whereas predictive values describe the probabilities of the disease given a positive or negative test result, respectively. These posterior probabilities depend on the prevalence of the disease in question. The methodological problems that arise in the planning of diagnostic studies differ from those encountered in treatment evaluation. This contribution presents an annotated checklist of the most important faults and shortcomings in the planning and the analysis of diagnostic validation studies. Avoiding these deficiencies may help that the study results are accepted by biostatisticians, clinicians and cost units.