Methods
The patient population in the FREEDOM trial has been described in detail previously. In brief, the multicenter, open-label prospective randomized superiority trial evaluated PCI/DES versus CABG in 1,900 diabetic patients in whom revascularization was indicated with stenosis of more than 70% in 2 or more major epicardial vessels, involving at least 2 separate coronary artery territories, and without left main coronary stenosis. Consenting diabetic patients with MVD were randomized on a 1:1 basis to either CABG or multivessel stenting with DES and observed up to 7 years (minimum 2 years; median 4 years). Patients randomized to the PCI/DES arm received any approved DES (sirolimus-eluting and paclitaxel-eluting were the predominant stents) per operator's choice. In this post-hoc analysis, we classified subjects into PCI/DES versus CABG based on the actual treatment received (non–intention-to-treat [ITT]) and analyzed outcomes occurring post-procedure: 1,850 FREEDOM subjects with this definition had an index procedure: 956 PCI and 894 CABG procedures were performed.
Measures
Patients were categorized as ITDM if they indicated baseline use of insulin (either alone or in combination with other oral antidiabetic medication). To systematically assess angiographic complexity, SYNTAX (Synergy Between PCI With TAXUS and Cardiac Surgery) score, including chronic total occlusions, were calculated and evaluated based on angiographic core laboratory interpretations conducted at the Cardiovascular Research Foundation in New York, New York. Overall, chronic total occlusion and bifurcation statistics were based on total number of lesions (N = 11,219), and not on number of subjects. The primary outcome was the composite (earliest occurring) of all-cause death/MI/stroke using all available follow-up data.
Statistics
Categorical variables are described using a count and percentage, and continuous variables by mean and SD, or median and interquartile range. The distributions of categorical variables by treatment group and by ITDM status were compared using a Fisher exact test. The Wilcoxon rank sum test was used to compare the distributions of blood urea nitrogen (BUN), creatinine, euroSCORE, by treatment arm and by ITDM status, whereas the Student t test was used for the remaining continuous variables. Event-free survival rates were estimated using the Kaplan-Meier method, and the time to event for ITDM versus non-ITDM patients was compared using the log-rank test, using all available follow-up data. Two- and 5-year point estimates are presented. For the primary endpoint and secondary endpoints, a Cox proportional hazards regression test of interaction (treatment received by ITDM status) was used to assess whether there was a differential treatment effect by ITDM status using all available follow-up. Multivariable Cox regression of the primary endpoint was utilized to identify whether ITDM status is an independent predictor of outcome. Key results also are presented in the ITT population and contained in the Online Appendix http://content.onlinejacc.org/data/Journals/JAC/930880/61182_mmc1.docx. To construct the multivariable model, we first examined univariate Cox models, and variables that were at least marginally associated with the endpoint (p < 0.20) were included in a model in which stepwise selection was used for predictor selection at each step. Additional candidate variables were included into the multivariable model if there were significant treatment by predictor interactions (p ≤ 0.05). Finally, we also included measures that were significant predictors of ITDM status. Subgroup analyses were performed for ITDM status and treatment group, according to the angiographic SYNTAX score. Our findings are reported as hazard ratios (HRs) with 95% confidence intervals (CIs). A p value <0.05 was considered significant.