| Objective: To evaluate the risk factors for recurrence of gastrointestinal stromal tumor(GIST)after radical resection.To develop a recurrence-free survival prediction model applicable to the Chinese population,and assess its accuracy by comparing with the modified National Institute of Health(NIH)consensus criteria(Joensuu version).Methods: We retrospectively collected the demography and clinical data of 5285 GIST patients from 23 hospitals in Shandong Province from March 2001 to November 2014,and followed the patients who met the inclusion criteria.The risk of GIST recurrence and metastasis was stratified using the modified NIH consensus criteria.The recurrence-free survival(RFS),disease-specific survival(DSS)and overall survival(OS)were calculated by Kaplan-Meier method and compared by Log-rank method.The Cox’s proportional hazards regression model was adopted to identify the prognostic factors for GIST after radical resection.We also develop prognostic index(PI)model and 1-year,3-year,and 5-year RFS prediction models.The receiver operating characteristics curve analysis(ROC)was used to assess the accuracy of RFS prediction model by comparing with the modified NIH consensus criteria.All statistical analyses were performed using SPSS 24.0 statistical software package.Results: Among the collected 5285 GIST patients,4216 patients accorded with the inclusive criteria and were followed up.The complete follow-up data of 3363 patients were obtained with follow-up rate of 79.8%.1-year,3-year and 5-year RFS were 94.6%(95% confidence interval [CI] 93.8-95.4),85.9%(84.7-87.1)and 78.8%(77.0-80.6),respectively.1-year,3-year and 5-year DSS were 97.6%(97.0-98.2),90.7%(89.7-91.7)and 88.9%(87.7-90.1),respectively.1-year,3-year and 5-year OS were 97.1%(96.5-97.7),89.1%(87.9-90.2)and 84.9%(83.3-84.5),respectively.After being stratified by the modified NIH consensus criteria,no statistically significant difference was observed for the RFS(χ2 =0.001,P=0.982),DSS(χ2 =0.03,P=0.868)and OS(χ2 =0.83,P=0.361)between very low-risk patients and low-risk patients,and OS between very low-risk patients and intermediate-risk patients(χ2 =3.35,P=0.067).There were significant differences among other groups(χ2 =15.71-258.28,P=0.000-0.003).A multivariable analysis by Cox’s proportional hazards regression model indicated that the independent variables related to prognosis were sex(HR=1.310,1.052-1.632,P=0.016),tumor location(HR=1.419,1.144-1.760,P =0.001),tumor size(HR=1.100,1.081-1.120,P<0.001),mitosis count(6-10/50 HPFs vs.≤5/50HPFs: HR=4.231,3.261-5.489,P<0.001;>10/50 HPFs vs.≤5/50HPFs: HR=7.585,5.678-10.134,P<0.001)and rupture(HR=3.522,2.573-4.822,P<0.001).Based on the calculated regression coefficient,Prognostic index(PI)= 0.000(if female)+ 0.270(if male)+ 0.000(if gastric GIST)+ 0.350(if non-gastric GIST)+ 0.000(if no tumor rupture)+ 1.259(if tumor rupture)+ 0.000(tumor mitotic count less than 6 per 50 HPFs)+ 1.442(tumor mitotic count between 6 and 10 per 50 HPFs)+ 2.026(tumor mitotic count more than 10 per 50 HPFs)+ 0.096 × tumor size(cm).The 1-year,3-year and 5-year RFS prediction model were S(12,X)= 0.9926exp(PI),S(36,X)= 0.9739exp(PI)and S(60,X)= 0.9471exp(PI),respectively.In the accuracy assessing,the area under the curve(AUC)of ROC of RFS prediction model was 0.873(95%CI 0.854-0.893,P<0.001),whereas the AUC of the modified NIH consensus criteria was 0.825(0.804-0.846,P<0.001).Conclusions: The variables of sex,tumor location,tumor size,mitosis count,and rupture were independent prognostic factors.The developed 1-year,3-year,and 5-year of RFS prediction model based on PI can be applied to predict the risk of recurrence/metastasis of individual patient. |
PDF Full Text Request