|| Checking for direct PDF access through Ovid
Gastrointestinal stromal tumor (GIST) natural history per se has not been extensively investigated yet, with most data being drawn from large studies with a relevant referral bias. Hence, the estimation of prognosis still remains a critical issue. We retrospectively evaluated 929 GISTs resected between 1980 and 2000 in 35 Italian institutions. A total of 526 patients were found to be suitable for refining risk assessment through the development of a survival nomogram. Median follow-up was 126 months. On testing for potential prognostic parameters, age, tumor site, size, and mitotic index proved to be predictors of OS on both univariable and multivariable Cox model analyses, whereas necrosis and cytonuclear atypia were significant on univariable analysis only. The discriminative ability of the model, including the parameters selected after a backward procedure (C=0.72), improved compared with the National Institutes of Health 2002 (C=0.64) and the National Comprehensive Cancer Network 2007 (C=0.63). On the basis of these data we developed a prognostic nomogram for survival that considers site, size, and mitotic index as continuous variables, providing estimates stratified for patients aged ≤65 and >65 years. This nomogram is a tool based on survival. It overcomes problems that result from artificial categorization of continuous variables. We believe that in the future this should also be attempted by nomograms based on the risk of relapse.