RESUMEN
This work is aimed at identifying factors associated with primary jejunum-ileal tumors malignancy, defining a prediction model with sensitivity, specificity and accuracy to distinguish malign from benign neoplasms. These tumors are rare, have highly unspecific presentation and, frequently, are diagnosed late. We reviewed the charts of 42 patients with primary jejunum-ileal tumors treated in the Department of General Surgery of Rio de Janeiro State University Hospital, Rio de Janeiro, RJ, Brazil, from 1969 to 1998. We performed bivariate analyses, based on chi 2 test, searching associations between tumors malignancy and demographic and clinical variables. Then logistic regression was employed to consider the independent effect of variables previously identified on malignancy risk. The malign tumors included 11 adenocarcinomas, 7 leiomyosarcomas, 5 carcinoids and 4 lymphomas; the benign tumors included 10 leiomyomas, 2 hamartomas, and single cases of adenoma, multiple neurilemoma and choristoma. The bivariate analyses indicated the association between malignancy and palpable abdominal mass (P = 0.003), period from signs and symptoms onset to diagnosis (P = 0.016), anemia (P = 0.020), anorexia (P = 0.003), abdominal pain (P = 0.031), weight loss (P = 0.001), nausea and vomit (P = 0.094), and intestinal obstruction (P = 0.066); no association with patients demographic characteristics were found. In the final logistic regression model, weight loss, anemia and intestinal obstruction were statistically associated with the dependent variable of interest. Based only on three variables--weight loss, anemia and intestinal obstruction-the model defined was able to predict primary jejunum-ileal tumors malignancy with sensitivity of 85.2 per cent, specificity of 80.0 per cent, and accuracy of 83.3 per cent