Application value of nomogram and risk factors of postoperative recurrence of gastrointestinal stromal tumors

Abstract

Objective To investigate the risk factors of postoperative recurrence of gastrointestinal stromal tumors (GISTs) and verify the clinical value of nomogram of the Memorial Sloan Kettering Cancer Center (MSKCC), and draw a new nomogram model and a calibration curve. Methods The retrospective case-control study was conducted. The clinical data of 90 patients with GISTs who were admitted to the Affiliated Cancer Hospital of Shanxi Medical University between June 2000 and January 2009 were collected. All the patients underwent radical resection of GISTs and didn′t take oral imatinib mesylate of postoperative adjunctive therapy. The clinicopathological data of patients were used for prognostic analysis. Postoperative GISTs of patients were predicted by nomogram of the MSKCC. Area under ROC curves was calculated through the comparison between probability of new samples from Bootstrop and real probability of recurrence of patients. A new nomogram model and calibration curve were built using the R language. Observation indicators: (1) real recurrence situations of GISTs; (2) risk factors analysis affecting the postoperative recurrence of GISTs; (3) a nomogram of MSKCC predicting the postoperative theoretical recurrence situations of GISTs in each patient; (4) establishment of the new nomogram model and calibration curve. The follow-up using telephone interview, outpatient examination and visit was performed to detect the postoperative recurrence of GISTs up to February 2016. The survival curve was drawn by the Kaplan-Meier method. The univariate analysis and multivariate analysis were done respectively using the chi-sqaure test and COX regression model. A risk model of postoperative recurrence of GISTs was built. The new nomogram model and calibration curve of postoperative recurrence-free survival rate were drawn according to the regression coefficients and other data of selected variables in the COX regression model. Results (1) Real recurrence situations of GISTs: all the 90 patients were followed up for 1-82 months, with a median time of 25 months. The 2-, 5-year recurrence-free survival rates were respectively 52.2%(47/90)and 11.1%(10/90). (2) Risk factors analysis affecting the postoperative recurrence of GISTs: results of univariate analysis showed that age, CD117, tumor location, tumor diameter and mitotic figures were risk factors affecting the postoperative recurrence of GISTs (χ2=9.276, 19.911, 31.721, 32.973, 6.482, P<0.05). Results of multivariate analysis showed that non-gastrointestinal tumor, tumor diameter≥10 cm and mitotic figures≥5/50 high power field (HPF) were independent risk factors affecting the postoperative recurrence of GISTs [RR=1.166, 5.905, 1.012, 95% confidence interval (CI): 1.004-1.355, 2.076-16.796, 1.008-1.017, P<0.05]. (3) A nomogram of MSKCC predicting the postoperative theoretical recurrence situations of GISTs in each patient: the area under the curve (AUC) of 2-, 5-year recurrence-free survival rates in a nomogram of MSKCC was respectively 0.727 (95%CI: 0.662-0.791) and 0.724 (95%CI: 0.658-0.791). (4) Establishment of the new nomogram model and calibration curve: the new nomogram model and calibration curve were build using R language. The alignment pattern was close to a straight line with slope 1, and a nomogram can accurately predict the postoperative recurrence risk of GISTs. The AUC in the nomogram model was 0.724 (95%CI: 0.629-0.783), and a modified and standard AUC of the National Institutes of Health (NIH) was 0.434 (95%CI: 0.347-0.521). Conclusion The non-gastrointestinal tumor, tumor diameter≥10 cm and mitotic figures≥5/50 HPF are independent risk factors affecting the postoperative recurrence of GISTs, and the nomogram can assess the postoperative recurrence risk of GISTs. Key words: Gastrointestinal stromal tumors; Nomogram; Recurrence; Risk assessment