Overall 5-year survival has long been used as an approximate measure of the proportion of patients cured of a particular cancer under study. For more accurate estimation of the cure rate, we applied the Boag parametric cure model to follow-up data from cancer patients.
 Three datasets were used, comprising 1,410 patients with gastric cancer, 1,754 with colonic cancer and 10,447 with pancreatic cancer; all underwent surgical resections and were followed up. The patients were classified into 14 groups by cancer type and TNM stage and for each group the 5-year survival rate was compared with the cure rate, which was estimated by the maximum likelihood method based on the Boag model. The effects of pathological factors including tumor size, serosal invasion, nodal involvement and venous invasion on the cure rate were evaluated by the Boag model and the Gamel-Boag regression.
 The results showed that, in 11 of the 14 cancer-stage groups, the 5-year survival was greater than the corresponding cure rate. Of the three cancers the pancreatic cancer showed the lowest cure rate: the curable fraction was about 16% with tumor size <= 4 cm. In contrast, patients with colonic cancer were highly curable:75% of them were cured even when their tumor size >4 cm. Only the group with positive venous invasion showed a cure rate less than 75%. This lower curability is explained by the fact that liver metastasis, a common form of relapse of colonic cancer, is preceded by venous invasion. Gastric cancer was intermediate in curability between pancreatic and colonic cancers.
 In conclusion, the estimation of cure rate provides a clear and deep insight into the relationship between patient outcome and prognostic factors, permitting patients and clinicians to make better clinical decisions.