Stata version 11

Continuous variables were compared by using two-sample t tests or the Wilcoxon rank-sum test if nonnormally distributed. Categorical variables were compared by using the Pearson χ² test or the Fisher exact for small cell counts less than six. Survival curves were visualized by using Kaplan-Meier curves and were compared by using log-rank tests. Annual event rates were derived by dividing the number of events by person-years. Predictors of MACE, MI, and death were assessed by using Cox regression modeling, after verifying the assumption of proportional hazards with Schoenfeld residuals. Men and women were matched for risk factors (age, hypertension, high cholesterol, diabetes, smoking, family history, chest pain symptom presence, and log SIS, where the SIS was log-transformed toward normality and then back-transformed into original units for purposes of reporting; matching was done by using the 1:1 Mahalanobis nearest-neighbor algorithm within a caliper of 0.01) and evaluated by using standardized differences (19 ,20 (link)). A post hoc analysis with our observed MACE event rates show 80% power to detect a minimum HR of 1.5 with 11 462 patients and 1.7 in men or women subgroups. All data were analyzed by using Stata version 11.2 software (www.statacorp.com). A P value less than .05 was considered to indicate a significant difference.

PFS and OS were estimated by the Kaplan-Meier method.²⁹ PFS was calculated from both date of metastasectomy and start date of TKI to date of documented radiographic disease recurrence. OS was calculated from both date of metastasectomy and date of TKI initiation to date of death or last follow-up. Regional recurrence (RR) was defined as those occurring within the pelvis, peritoneum (excluding the liver), or retroperitoneum and distant metastasis (DM) was defined as those occurring outside these regions (ie. lung, liver, abdominal wall). The log-rank test was used to determine statistical differences in OS and PFS. Univariate and multivariate regression analysis was performed using a Cox proportional hazards model. Variables that were statistically significant on univariate analysis were included in the corresponding multivariate analyses. A P value of less than 0.05 was considered statistically significant, and statistical calculations and data analysis were performed using Stata version 11.2 software (College Station, Texas).