Angiosarcomas have traditionally been considered high-grade lesions for which histologic features and grading have played no role in prognostication and, consequently, they have been excluded from the American Joint Committee on Cancer staging system. We have, therefore, analyzed 69 cutaneous angiosarcomas seen in consultation and not associated with lymphedema or prior radiation therapy to determine if a combination of histologic and clinical parameters could be used to differentiate indolent from aggressive tumors. The clinical features analyzed included patient age, location, size, depth, and focality of the lesion; the histologic features studied included pattern of growth (vasoformative vs. solid), nuclear grade (high vs. low), necrosis (present/absent), cell type (epithelioid or spindled), extent of inflammatory infiltrate (minimal vs. marked), and mitotic rate. Lesions occurred on the head/neck (49), extremities (15), and trunk (5) of adults (21 to 94 y) and predominated in males (41 males; 28 females). Tumors ranged in size from 0.3 to 15 cm (average 3.1 cm) and involved the papillary (n=9), reticular (n=16), or deep dermis/ subcutis (n=30). They could be predominantly vasoformative (n=28) to solid (greater than 50% solid, n=41). Most lesions were of high (n=65) as opposed to low (n=4) nuclear grade, were mitotically active (0 to 99/10 high power fields), and occasionally displayed necrosis (n=14) and epithelioid features (n=21). Inflammatory infiltrates were minimal in most cases. Follow-up information was obtained for all patients. Recurrences developed in 18 patients (26%) and metastasis in 15 (22%) to the following sites: lung (6), lymph node (7), liver (2), bone (2), and parotid gland (1). Forty-seven patients died (30 of disease) and 22 were alive at last known follow-up (range, 16 to 158 mo; mean, 65; median, 36). Five-year disease-specific survival was 48%. By univariate analysis only older age, anatomic site, necrosis, and epithelioid features correlated with increased mortality. Tumors were stratified into low (n=41) or high (n=28) risk groups based on necrosis and/or epithelioid features. By multivariable analysis, high-risk group (hazard ratio 4.07, P=0.0004) and age >70 (hazard ratio 2.79, P=0.012) were associated with increased mortality, and tumor depth (P=0.048) correlated with the risk of local recurrence. The high-risk group had a significantly worse prognosis than the low-risk group with 3-year survival of 24% and 77%, respectively. No patients with high-risk features survived 5 years. In conclusion, we report that a combination of clinical and histologic features allows stratification of angiosarcoma patients into 2 risk groups that are strongly associated with marked differences in clinical course. These features seem to diminish in importance with increased tumor size and are probably most useful in tumors less than 5 cm in maximum dimension.