Clinical Outcome of Patients With and Without Diabetes Mellitus After Percutaneous Coronary Intervention With the Resolute Zotarolimus-Eluting Stent

2-Year Results From the Prospectively Pooled Analysis of the International Global RESOLUTE Program

Patient selection

For this analysis, we pooled patient-level data for 5,130 patients from 5 controlled studies, with 1 randomized (21,22) and 4 single-arm (20,23–25) designs of the RESOLUTE Global Clinical Trial Program (Table 1). In concordance with the FDA, the predefined subset of the pooled cohort was based on the inclusion and exclusion criteria of studies in patients with diabetes in the published data at the time the study was designed (Table 2). Because the enrollment criteria of the randomized RESOLUTE All Comers (21,22) and the observational RESOLUTE International (24) studies included more complex patients, the high-risk subsets from these 2 studies were excluded from the primary analysis. Patients with acute myocardial infarction (MI) within 72 h of the index procedure, lesion length >27 mm, >2 lesions/vessel, in-stent restenosis, unprotected left main lesions, bifurcation lesions, total occlusions, bypass grafts, thrombus, left ventricular ejection fraction <30%, or serum creatinine ≥1.6 mg/dl (≥140 μmol/l, according to Table 2) were excluded.

Because the RESOLUTE First-in-Man (20), RESOLUTE US (23), and RESOLUTE Japan studies (25) used the same criteria, all patients from these 3 studies were included in the pre-specified analysis.

Thus, our pre-specified analysis population for the FDA comprised 878 patients with diabetes (Tables 1 to 4).⇓⇓ We also compared these results with the 1,903 patients without diabetes but having the same inclusion/exclusion criteria. In addition to this pre-specified diabetic cohort, we evaluated the safety and efficacy of the R-ZES in all 1,535 patients with diabetes, including the 657 complex patients, compared with all 3,595 (also, including complex) patients without diabetes (Table 1).

Diabetic status

Patients were classified as having diabetes if they were on a regimen of insulin and/or taking oral antidiabetic agents or on a modified diet to control diabetes. Patients were considered insulin-treated if they were taking insulin. Patients were considered noninsulin-treated if they were taking only oral antidiabetic agents or on a modified diabetes diet only or both oral agents and a modified diet. The same criteria were used for all trials in the global RESOLUTE Clinical Program.

All RESOLUTE trials required the same dual antiplatelet therapy. A daily dose of at least 75 mg aspirin was prescribed indefinitely, and 75 mg of clopidogrel was prescribed daily for at least 6 months.

Study management

Each patient consent form was externally verified for all studies. Source files were reviewed for all patients in the RESOLUTE First-In-Man (20), All Comers (21,22), US (23), and Japan (25) studies. For the RESOLUTE International study, although 25% of patient charts were externally monitored, there were no differences in clinical outcomes between fully monitored and partially monitored patients through 1 year (24). Independent Clinical Events Committees (CEC) reviewed all serious adverse events. All studies were conducted according to the Declaration of Helsinki; all study protocols were approved by local ethics committees.

In addition to the CEC event adjudication process specific for each study, a CEC Global Oversight Committee was organized to ensure consistency in clinical data review and to harmonize the interpretation of event definitions across the CECs from the 3 largest trials, the RESOLUTE All Comers, RESOLUTE International, and RESOLUTE US trials. The Global Oversight Committee evaluated the consistency of event adjudication on the basis of interpretation of event guidelines and provided recommendations to the respective primary CEC. The cross-adjudication process included a minimum of 10% of the events from each of the 3 trials (maximum of 100 events/trial) to be cross-adjudicated by an alternate participating CEC. The selection of cross-adjudicated events was based on event type: death, MI, stent thrombosis, target lesion, or vessel revascularization.

Choice and definitions of the clinical outcome parameters

The primary objective for the pre-specified FDA analysis population was to compare the rate of target vessel failure (TVF), defined as a composite of cardiac death, target vessel myocardial infarction (TV-MI), and ischemia-driven target vessel revascularization (TVR) in patients with diabetes at 12 months against a pre-defined performance goal (see the following text).

Pre-specified secondary objectives included major adverse cardiac events (MACE), defined as a composite of any death, any MI, emergent coronary artery bypass, and ischemia-driven revascularization by any method and target lesion failure (TLF), defined as a composite of cardiac death, TV-MI, and ischemia-driven target lesion revascularization (TLR). Any death was considered cardiac unless a noncardiac cause could be confirmed. Any MI for which a clear ascription to a target vessel could not be determined was counted as a TV-MI. All endpoint definitions have been previously described in detail and are consistent among the 5 studies (20–25,27). Definite and probable stent thrombosis was adjudicated according to the Academic Research Consortium (ARC) criteria (28).

Pre-defined hypothesis and statistical methods

All studies in the RESOLUTE Global Clinical program were prospectively designed with similar inclusion and exclusion criteria, clinical documentation forms, endpoint definitions, adjudication procedures, and statistical methodologies. To evaluate the performance of the R-ZES in patients with diabetes against previously published evidence, the following pre-specified analysis was designed in conjunction with the FDA to support the approval of an expanded indication for treatment of patients with diabetes with the R-ZES in the United States.

All data were analyzed according to the intention-to-treat principle. To develop the pre-specified performance goal, we performed a meta-analysis of studies reporting outcomes of patients with diabetes, treated with sirolimus- (SES) and paclitaxel-eluting stents (PES) (7,8,13,15,29), and used data from the ENDEAVOR pooled studies (30) to define an expected rate of TVF in a diabetic patient population after treatment with various DES. Thus, the weighted meta-analytic rate for the 12-month TVF rate was calculated to be 11.08%. The performance goal was established at 14.5%, which includes a clinically acceptable margin. The hypothesis for this analysis accounted for the differences in the protocols of the individual studies in the published data. Rejection of the null hypothesis at a 1-sided 0.05 level of significance would signify that the upper 95% limit around the TVF rate would be <14.5%. With 80% power, this resulted in a sample size requirement of 604 evaluable patients. This performance goal was developed and agreed in collaboration with the U.S. FDA and the Harvard Clinical Research Institute (Boston, Massachusetts).

For baseline comparisons, the 2-sample t test was used for continuous variables, and data are presented as mean ± SD. The Fisher exact test was used for binary variables. Time to event analyses is reported with the Kaplan-Meier method. A p value of <0.05 was considered statistically significant. The Harvard Clinical Research Institute, an independent clinical research organization, performed all statistical analyses with SAS (version 9.1 or higher; SAS Institute, Inc., Cary, North Carolina).

Primary endpoint at 1 year for FDA approval

The 12-month rate of the primary endpoint of TVF in the 878 R-ZES patients in the pre-specified diabetic cohort was 7.8% (upper 95% confidence interval: 9.51%) and significantly lower than the performance goal of 14.5% (p = 0.001). Therefore, the pre-set performance goal for the FDA was met. Additional safety and efficacy outcomes at 12 months for the pre-specified diabetic cohort were as follows: TLF was 6.6%, all death was 2.8%, the combination of cardiac death and TV-MI was 3.6%, MACE was 7.5%, and ARC definite and probable stent thrombosis was 0.3%.

Clinical outcomes at 2 years

Clinical outcomes at 2 years for the pre-specified patient population are shown in Tables 4 and 5.⇓ The rates of MACE, all death, TVF, and TLF were significantly higher in the pre-specified diabetic cohort than in the nondiabetic cohort (Table 4). These outcomes were also significantly higher in patients with diabetes taking insulin as compared with the noninsulin-treated cohort (Table 5). The incidence of ARC definite and probable stent thrombosis through 2 years was very low, regardless of the presence or severity of diabetes: in the pre-specified cohort, the rates of stent thrombosis were 0.4% for the nondiabetic patients and 0.3% in the diabetic group (p > 0.99) (Table 4). Of the 250 patients taking insulin, there were 2 patients with stent thrombosis, and of the 628 diabetic patients not treated with insulin, only 1 patient had stent thrombosis throughout the 2 years (Table 5). In the entire pre-specified diabetic cohort of 878 patients, not a single patient had a very late stent thrombosis (Tables 4 and 5). The 30-day, 6-month, and 1- and 2-year rate of dual antiplatelet therapy usage was 96.1%, 94.8%, 90.3%, and 59.1%, respectively, for the pre-specified diabetic population.

As in the pre-specified cohort (Tables 4 and 5), in the whole population, including the complex patients, the rates of stent thrombosis were not significantly different between patients with and without diabetes (1.2% vs. 0.8%) as well as between insulin-treated and noninsulin-taking patients with diabetes (1.8% vs. 0.9%) (Table 6). Dual antiplatelet therapy usage in the overall diabetic and nondiabetic groups at 1 year was 89.9% and 89.2% and at 2 years was 52.8% and 42.5%, respectively.

Figure 1 shows the cumulative incidence of TLF and its components for patients with diabetes according to insulin treatment and for patients without diabetes at 1- and 2-year follow-up for the pre-specified analysis population. Time-to-event analyses show that the incidence of TLF (Fig. 1A) and its components of cardiac death and TV-MI (Fig. 1B) and TLR (Fig. 1C) in patients with noninsulin-treated diabetes are comparable to that of patients without diabetes. The clinical events are significantly higher in the insulin-treated patients than in the noninsulin-treated diabetic and nondiabetic patients. The analysis of all R-ZES patients, including the nondiabetic and diabetic complex patients, shows the same pattern with significantly higher event rates in the insulin-treated patients but no difference between the noninsulin-treated diabetic and the nondiabetic patients (Fig. 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Cumulative Incidence of TLF and its Components for Patients in Pre-Specified Analysis

Cumulative incidence of target lesion failure (TLF) (A) and its components cardiac death or target vessel myocardial infarction (TVMI) (B) and ischemia-driven target lesion revascularization (TLR) (C) by diabetic status for patients in the pre-specified analysis population. DM = diabetes mellitus; ITDM = insulin-treated diabetes mellitus.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Cumulative Incidence of TLF and its Components for All Patients

Cumulative incidence of TLF (A) and its components, cardiac death or TVMI (B), and ischemia-driven TLR (C) by diabetic status for all patients from the RESOLUTE Clinical Trial Program, including the complex patients. Abbreviations as in Figure 1.

The concept of a pre-specified prospective cohort analysis relative to a historical control

In general, randomized trials are more rigorous than observational studies, yet they are often limited in power to address many individual clinical questions, due to the sample size of individual populations. Therefore, an acceptable compromise of a prospectively planned analysis compared with a performance goal was derived from historical data. This acceptability of a performance goal study on the basis of historical data was used for U.S. approval of the first-generation Taxus Liberté PES (Boston Scientific, Natick, Massachusetts) (31). The TAXUS ATLAS trial (TAXUS ATLAS: A Multi-center, Single-arm Study of the TAXUS Liberté-SR Stent for the Treatment of Patients With de Novo Coronary Artery Lesions)—a prospective, single-arm study in de novo coronary lesions—compared treatment with the Taxus Liberté PES versus a matched historical control group of patients from the TAXUS-IV and -V trials treated with the Taxus Express PES (31). The Taxus Liberté PES met its performance goal as compared with the Taxus Express PES for the primary clinical endpoint of TVR at 9 months. Subsequently, the Promus Element everolimus-eluting stent (EES) (Boston Scientific) gained U.S. approval by meeting a performance goal for the primary surrogate endpoint of late loss at 9 months (32).

For the first time, for the purpose of obtaining this specific indication with the U.S. FDA, a pre-specified performance goal was prospectively established to determine the safety and effectiveness of a DES for patients with diabetes. Inclusion and exclusion criteria (Table 2) for the pre-specified analysis cohort were aligned across the 5 RESOLUTE trials to support appropriate comparisons with patients in the studies used to derive the performance goal. Therefore, all 5 RESOLUTE trials used consistent endpoint definitions, and a Global Oversight Committee cross-adjudicated a subset of events, to ensure consistency in clinical endpoint review. Patient-level data were then pooled from the 5,130 patients from the RESOLUTE Global Clinical Program (Table 1). The rate of TVF in the pre-specified analysis cohort for the R-ZES–treated patients with diabetes at 12 months was 7.8%, considerably less than the predefined DES performance goal of 14.5% (p = 0.001). Thus, the primary endpoint of 1-year TVF met the performance goal, and the R-ZES achieved the first FDA approval of the specific expanded indication for PCI in patients with diabetes. The R-ZES is the only stent in the United States approved for use in patients with diabetes.

As shown in Figure 1, the 2-year clinical outcomes with the R-ZES in patients with diabetes not taking insulin overlap the outcome in patients without diabetes, even though the patients with diabetes not taking insulin had significantly more cardiovascular risk factors (Table 3). The same observation was also made when including the more-complex patients (Fig. 2). Therefore, the presence of noninsulin-requiring diabetes does not seem to affect clinical outcomes after treatment with the R-ZES.

Superiority of DES versus BMS in patients with diabetes

The DIABETES (Randomized comparison of sirolimus-eluting stent versus standard stent for percutaneous coronary revascularization in diabetic patients: the diabetes and sirolimus-eluting stent) trial (7) was the first prospective randomized trial comparing a DES versus BMS in patients with diabetes as an inclusion criterion (33). In 160 patients, the primary surrogate endpoint of in-segment late lumen loss at 9 months was significantly reduced with the Cypher SES (Cordis Corporation, Miami Lakes, Florida) compared with a BMS (0.06 ± 0.4 mm vs. 0.47 ± 0.5 mm). In the SCORPIUS trial (A German Multicenter, Randomized, Controlled, Open-Label Study of the Cypher Sirolimus-Eluting Stent in the Treatment of Diabetic Patients With De Novo Native Coronary Artery Lesions) (8), the same primary surrogate endpoint for the same DES was reached after 8 months in 200 patients with a significant reduction from 0.75 ± 0.59 mm to 0.17 ± 0.45 mm. In the DESSERT (An Italian Multicenter, Randomized, Single Blind Study of the Sirolimus Eluting Stent in the Treatment Of Diabetic Patients With De Novo Coronary Artery Lesions) study (9), the primary surrogate endpoint of in-stent late lumen loss was significantly reduced from 0.96 ± 0.61 mm to 0.14 ± 0.33 mm after treatment with the same DES after 8 months in 150 patients treated with oral anti diabetic medication and/or insulin.

Drug-eluting stents are preferred over BMS for PCI in patients with diabetes according to European and U.S. guidelines (34–36). However, although patients with diabetes are an everyday challenge in clinical practice and are regularly treated with DES, there is no randomized trial of DES versus BMS with a primary clinical endpoint based on an adequately calculated sample size for the clinical outcome in patients with diabetes as an inclusion criterion.

Are there differences between first-generation DES in patients with diabetes?

The ISAR-DIABETES trial (Intracoronary Stenting and Antithrombotic Regimen: Diabetes Mellitus) (37), was the first randomized study directly comparing 2 DES in patients with diabetes. The primary endpoint was the surrogate parameter of in-segment late luminal loss after a mean follow-up of 196 days. In 205 patients, the primary angiographic endpoint was significantly higher for the Taxus PES (0.67 ± 0.62 mm) than for the Cypher SES (0.43 ± 0.45 mm). Although this study was underpowered for the clinical outcome, the reported TLR rate was 12.0% in the PES group and 6.4% in the SES group with a wide confidence interval of 0.82 to 4.27 (p = 0.13). The discussion of the comparison of the 2 first-generation DES in patients with diabetes has been controversial over the last years with pros and cons for either PES or SES (38–40). With the fading use of these 2 first-generation DES, however, this discussion today is of less practical importance.

Are newer generation DES better for patients with diabetes?

The ESSENCE-DIABETES trial (Randomized Comparison of Everolimus-Eluting Stent Versus Sirolimus-Eluting Stent Implantation for De Novo Coronary Artery DisEase in Patients With DIABETES Mellitus) (10) is the first randomized study directly comparing a newer generation DES (Xience V EES, Abbott Vascular) with a first-generation DES (Cypher Select and Select Plus SES, Cordis Corporation) in 300 patients with diabetes. The EES was shown to be noninferior to the SES for the primary surrogate endpoint of in-segment late loss at 8 months (0.23 ± 0.27 mm vs. 0.37 ± 0.52 mm, p_{noninferiority} <0.001). Although underpowered for clinical outcomes, death, MI, and ischemia-driven TLR were not statistically different between these 2 DES. A recently published post hoc subgroup analysis from 4 pooled randomized trials with 27.6% diabetic patients compared the Xience V EES with a first-generation DES, Taxus PES (Express and Liberté, Boston Scientific) (41). In these 1,869 patients with diabetes, there were no differences in clinical outcomes after 2 years between the first- and newer generation DES.

A comparison of our pooled R-ZES 2-year outcomes with 2-year outcomes from a recently published pooled EES analysis (41) requires caution, because our pre-specified cohort of patients with diabetes excluded the complex cases (Table 2). By contrast, 3 of the 4 studies in the pooled EES data cohort also excluded complex patients with diabetes, and only 1 study enrolled complex patients with diabetes (17.2% of all pooled patients). With this caution in mind, the corresponding 2-year data for the 878 R-ZES (Table 4) and the 1,188 EES (41) patients with diabetes are as follows: ischemia-driven TLR, 4.8% versus 5.5%; ischemia-driven TVR, 7.9% versus 8.3%; and definite or probable stent thrombosis, 0.3% versus 1.6%. Including even the complex patients in our analysis, the 2-year rates for definite or probable stent thrombosis are 1.2% for the 1,335 diabetic patients treated with the R-ZES (Table 6) versus 1.6% for the 1,188 patients treated with the EES (41).

The problem of insulin-treated patients

Diabetes, particularly in patients taking insulin, has consistently been shown to be an independent predictor of adverse outcomes after stent implantation (42). The results in patients with diabetes treated with the R-ZES parallels results of other DES in that the rates for patients taking insulin have progressively worse clinical outcomes than the rates for patients not taking insulin (Figs. 1 and 2). Thus, the overall results in studies of patients with diabetes strongly depend on the percentage of patients with diabetes taking insulin (43). The proportion of patients taking insulin in the studies used to construct the performance goal ranged from 25.2% to 32.3% (7,14,30,44–46), consistent with the 28.5% from our analysis cohort.

In other randomized studies with diabetes as an inclusion criterion, the proportion of insulin-treated patients was 43% in the SCORPIUS (8), 25% in the DESSERT (9), and 15% in the ESSENCE-DIABETES (10) trials. In the post hoc pooled analysis of the EES and PES (41), 26.4% of patients taking insulin were included, which is comparable to 28.5% in our pre-specified cohort. Although these studies were not used to develop our performance goal, because these data were only recently published, the clinical results from the pooled Xience and Taxus studies are consistent with the rates of the studies used to construct our performance goal.

In accordance with our results for the R-ZES (Figs. 1 and 2), the EES also showed, as compared with patients with diabetes not taking insulin, considerably worse 2-year outcomes for insulin-treated patients with diabetes with significantly higher ischemia-driven TLR (10.8% vs. 3.7%) and higher MACE rates (14.8% vs. 8.4%) (41). Although both of the newer generation DES (i.e., the Resolute ZES and the Xience V) have improved outcomes compared with first-generation DES, there is still an opportunity to improve the treatment of CAD in patients taking insulin, and coronary artery bypass surgery should be strongly considered as an alternative revascularization strategy in these patients—especially with multivessel disease—as recommended by European and U.S. guidelines (35,36). The results of the FREEDOM (Future Revascularization Evaluation in Patients With Diabetes Mellitus: Optimal Management of Multivessel Disease) trial, an international study designed to define the optimal revascularization strategy for patients with diabetes and multivessel coronary disease, has provided more data with regard to this unresolved topic (47).