The purpose of this paper was to establish the diagnostic value of the C-reactive protein measurement (CRP) in patients attending their general practitioner (GP) with a new complaint for which the GP considers determination of the erythrocyte sedimentation rate (ESR) to be indicated. During 4 successive months in 1992, 11 GPs in four general practice centres in the Netherlands identified patients indicated for the erythrocyte sedimentation rate. ESR and CRP were determined at the local hospital laboratory. One year later, an independent GP established the follow-up diagnoses. By comparing the test results with the follow-up diagnoses, using logistic regression analysis and Receiver Operating Characteristic curves, sensitivities, specificities, predictive values and odds ratios were established. In 396 patients the prevalence of inflammatory diseases and malignancies ('pathology') was found to be 26% in males and 15% in females. Both ESR and CRP were valuable in discriminating pathology from harmless, often self-limiting diseases. The optimal upper limits of reference values ('cut-off points') for ESR were found to be 31 mm in both males and females. At these cut-off points, the diagnostic gains from positive and negative test results (positive predictive value minus prevalence, and negative predictive value minus 100% minus prevalence) were 45% and 4% for males and 30% and 2% for females. The optimal cut-off point for CRP was found at 15 mg L-1 in males and 34 mg L-1 in females. The diagnostic gain from positive and negative test results were 18% and 9% in males and 25% and 4% in females.(ABSTRACT TRUNCATED AT 250 WORDS)