Accurate risk stratification of gastrointestinal stromal tumors (GISTs) has become increasingly important owing to emerging adjuvant systemic treatments. All GISTs have been considered to have some malignant potential, but this hypothesis is now seriously challenged by studies indicating that microscopic gastric GISTs that are common in the general population probably have little or no malignant potential. The National Institutes of Health (NIH) consensus classification system, based on tumor size and mitotic count, is commonly used to assess patient prognosis after surgical resection. Large retrospective cohort studies from several countries now uniformly indicate that the NIH classification carries substantial prognostic value. In particular, patients with high-risk GIST (approximately 44% of all) have substantially poorer outcome than those with intermediate-risk (24%) or low/very low-risk GIST (32%), whose survival is not markedly inferior to that of the general population in some studies. Gastric GISTs (approximately 58% of all GISTs) have a lower risk of recurrence than nongastric tumors of the same size and mitotic count, and tumor rupture confers clearly increased risk. These 2 important risk stratification factors are not considered in the NIH classification. Patients with certain nongastric tumors (2.1-5 cm and > 5 mitoses per 50 high-power fields or 5.1-10 cm and < or = 5 per 50 high-power fields) and those with tumor rupture are proposed to be included in the NIH high-risk category. High-risk patients defined by the proposed modified system have more than 15% to 20% risk of disease recurrence. The proposed system, if validated, may be useful in identifying which patients might potentially benefit from adjuvant therapy.