We undertook this study to determine the clinical, biologic, immunologic, and therapeutic factors associated with the prognoses of polyarteritis nodosa (PAN) and Churg-Strauss syndrome (CSS). Three hundred forty-two patients (260 with PAN, 82 with CSS) followed from 1980 to 1993 were included in a prospective study on prognostic factors. Two hundred eighty-eight of these patients were included in the prospective studies on PAN and CSS. Items to be considered for analysis were collected at the time of diagnosis, during the acute phase of the disease. A survival curve was plotted for each clinical and biologic symptom observed in PAN or CSS. Each treatment arm of the prospective therapeutic trials was also tested: 1) prednisone (CS) + oral cyclophosphamide (CYC) + plasma exchanges (PE) versus CS E, 2) CS + PE versus CS, 3) CS + oral CY versus CS + pulse CY, 4) CS + pulse CY + PE versus CS + pulse CY in severe PAN and CSS, and 5) PE + antiviral agents after short-term CS in hepatitis B virus-related PAN. Of the parameters thus evaluated, the following had significant prognostic value and were responsible for higher mortality: proteinuria > 1 g/d (p < 0.0001; relative risk [RR] 3.6), renal insufficiency with serum creatinine > 1.58 mg/DL (p < 0.02; RR 1.86), GI tract involvement (p < 0.008. RR 2.83 for surgery). Cardiomyopathy and CNS involvement were associated with a RR of mortality of 2.18 and 1.76, respectively; these were not statistically significant. Similar survival rates were obtained with the prospectively tested therapies. The five-factors score (FFS) we established considered the prognostic factors creatinemia, proteinuria, cardiomyopathy, GI tract involvement, and CNS signs. Multivariate analysis showed that proteinuria (due to vascular or glomerular disease) and GI tract involvement were independent prognostic factors. When FFS = 0 (none of the 5 prognostic factors present), mortality at 5 years was 11.9%; when FFS = 1 (1 of the 5 factors present), mortality was 25.9% (p < 0.005); when FFS > 2 (3 or more of the 5 factors present), mortality was 45.95% (p < 0.0001 between 0 and 2, p < 0.05 between 1 and 2). We conclude that an initial assessment of PAN or CSS severity enables outcome and mortality to be predicted. The FFS is a good predictor of death and can be used to help the clinician choose the most adequate treatment. Renal and GI signs are the most serious prognostic factors.