The introduction of the first-generation sirolimus-eluting (SES) and paclitaxel-eluting stents (PES) has led to markedly reduced restenosis rates and reduced need for target lesion revascularization, compared with bare-metal stents, in DM patients as well as non-DM patients ((1),(2),(3),(4),5). However, DM remains associated with increased risk of in-stent restenosis, target lesion revascularization, and target vessel revascularization in patients undergoing percutaneous coronary interventions (PCI) (6). The second-generation everolimus-eluting stent (EES) has recently been found to be superior to the first-generation PES for reduction of target lesion revascularization, target vessel revascularization, and stent thrombosis (ST) in 2 large randomized trials; however, these significant improvements in safety and efficacy endpoints were limited to the nondiabetic subgroup of patients, because no differences in treatment effect between these 2 stents were observed in DM patients in both trials ((7),8). These findings were further confirmed by a large patient-level pooled analysis from 4 randomized clinical trials comparing EES with PES (9). Whether these results hold true in larger all-comer populations is unknown. There is a paucity of data on differences in clinical outcomes between EES and SES for treatment of DM patients, because the only data available derive from a relatively small series of patients and therefore are not adequately powered to detect low-frequency endpoints (10). Different issues with regard to the impact of metal alloy, strut thickness, polymer biocompatibility, and especially the effect of eluted active principle in patients with DM still remain unanswered. Therefore, we compared the safety and efficacy of the second-generation EES with the most-studied first-generation drug-eluting stent (DES), represented by the SES and PES in diabetic patients, with the data from the SCAAR registry (Swedish Coronary Angiography and Angioplasty Register) (11).

For the present analysis we studied all PCI-treated DM patients from the SCAAR database. During the period from January 18, 2007, to July 27, 2011, 71,639 PCIs with stent implantations were performed in Sweden, of which 13,830 (19.3%) were in DM patients. A total of 110,610 stents were implanted. Of these, 21,962 (19.8%) stents were used in DM patients, 87,789 in non-DM patients, and 859 in patients without information about DM status. Of the 21,962 stents implanted in DM patients, 11,493 were BMS, and 47 were not classified as BMS or DES. Of the remaining 10,422 DES, 30 were excluded in the analysis due to missing data. Of these 10,422 DES, 2,836 were PES, 1,370 were SES, and 3,929 were EES, whereas the remaining 2,258 stents represent Biolimus-eluting and Zotarolimus-eluting stents (not included in these analyses).

The SCAAR registry has been previously described ((11),12). Briefly, this registry holds data on consecutive patients from all 29 centers that perform coronary angiography and PCI in Sweden. The registry is sponsored by the Swedish Health Authorities and is independent of commercial funding. The technology is developed and administered by the Uppsala Clinical Research Centre. Since 2001, the SCAAR registry has been web-based, with recording of data online through a Web-interface in the catheterization laboratory; data are transferred in an encrypted format to a central server at the Uppsala Clinical Research Centre. All patients undergoing a coronary angiography or a PCI procedure nationwide are included. Since May 2005, all information with respect to restenosis and ST of previously treated patients that return in the catheterization laboratory for subsequent coronary angiography or PCI is entered in the SCAAR registry as well as the indication of such procedures. The web-based system provides each center with immediate and continuous feedback on processes and quality-of-care measures. Monitoring and verification of registry data are periodically performed in all hospitals since 2001 by comparing 50 entered variables in 20 randomly selected interventions/hospital and year with patient hospital records.

The SCAAR registry includes follow-up data for every implanted stent device, permitting device-oriented as well as patient-oriented endpoint analysis. For the current analysis the EES was compared individually with PES and SES in DM patients. Diabetes mellitus was defined either by patient-reported diagnoses on clinical files at baseline or use of antidiabetic medication before procedure. The primary endpoint of the study was a composite safety and efficacy endpoint of all-cause mortality, ST, and restenosis at 1 year.

The restenosis and ST are performed at device level, whereas mortality is analyzed at patient level. The same definition for restenosis, as defined by the SCAAR steering committee, was used. The SCAAR definition of restenosis is defined as a stenosis assessed by angiographic visual estimation (>50%) or by fractional flow reserve value of <0.80 in a previously stented segment identified by coronary angiography for any clinical indication in any of the 29 centers in Sweden ((11),12). The clinical relevance of restenotic lesions was detected by symptoms, routine noninvasive functional testing (exercise test, nuclear scan) and/or invasive functional evaluation by fractional flow reserve. The clinically driven detected restenosis, the efficacy endpoint in this study, differs from the classical target lesion revascularization endpoints used in other large studies. The target lesion revascularization combines revascularizations performed due to restenosis as well as ST and therefore might introduce biases, especially for high-risk populations treated with first-generation DES, where the ST can generate a consistent percentage of the target lesion revascularization events.

Stent thrombosis was defined in the SCAAR registry as an angiographic occlusion or a nonocclusive thrombus in a previously implanted stent with an acute clinical presentation (13), a definition that highly resembles the American College of Radiology definition of definite ST (14).

All-cause mortality data were obtained from the National Population Death Registry. The merging of the registries was performed by the Epidemiologic Centre of the Swedish National Board of Health and Welfare and approved by the local ethical committee at Uppsala University.

The baseline clinical and angiographic characteristics were compared by means of the chi-square test for categorical variables and analysis of variance for the continuous variables. Stent groups were compared with survival analyses. The Kaplan-Meier estimator was used to compute cumulative hazards and Cox proportional hazards regression to estimate hazard ratios (HRs). Differences between stent groups in baseline characteristics were adjusted with propensity score methods. Two different propensity scores were created, both defined as the conditional probability of having an EES, 1 in populations of stents with only EES or PES, and the other in a population with only EES or SES. The following variables were forced into logistic regression models: previous myocardial infarction, coronary artery bypass grafting, or PCI; sex; hypertension; hyperlipidemia; smoking; prior aspirin, clopidogrel, or warfarin use; anticoagulant use; glycoprotein IIb/IIIa inhibitor use; number of vessel disease; graft; restenosis; chronic occlusion; indication for PCI; complete revascularization; bifurcation lesion; type of lesion; and hospital. These propensity scores were entered in Cox proportional hazards models together with the type of stent drug and the year of PCI procedure to calculate the adjusted HR for ST and restenosis. In a similar way 2 additional propensity scores were created and used in the outcome analysis of all-cause mortality, including the following variables: previous myocardial infarction, coronary artery bypass grafting, or PCI; sex; hypertension; hyperlipidemia; smoking; number of vessel disease; indication for PCI; complete revascularization; bifurcation lesion; and hospital. Discrimination, calibration, and distribution of all 4 different propensity scores were checked. The C-index value and the p value for the Hosmer and Lemeshow test in the EES versus PES Restenosis and ST propensity score were 0.775 and 0.81, respectively; for the EES versus SES Restenosis and ST propensity score: 0.897 and 0.450, respectively; for Everolimus versus Paclitaxel Mortality propensity score: 0.791 and 0.335, respectively; and for Everolimus versus Sirolimus Mortality propensity score: 0.897 and 0.917, respectively.

The restenosis and definite ST endpoint were evaluated in medically treated DM patients as stratified for insulin treatment. Statistical interaction between insulin treatment and stent type was examined by introducing these variables as interaction terms in the Cox proportional hazard models.

All statistical analyses were performed with SPSS (version 19, SPSS, Chicago, Illinois).

A total of 4,751 DM patients were treated with PCI and were included in the present study. These patients received 8,134 DES: 3,928 EES (average 2.23 ± 1.25 stent/patient), 2,836 PES (average 2.03 ± 1.11 stent/patient), 1,370 SES (average 2.34 ± 1.34 stent/patient). Baseline characteristics of the groups are shown in (Table 1). Although baseline clinical and angiographic results were in general well-balanced, some significant differences were detected.

Clinical outcome data are presented in (Table 2).

The EES was associated with significantly lower rates of adverse events compared with SES (SES vs. EES: HR: 1.99; 95% confidence interval [CI]: 1.19 to 3.08). A similar trend was observed also when compared with PES (HR: 1.33; 95% CI: 0.93 to 1.91) ((Figure 20_gr1),Table 2). No significant differences were observed in the insulin-dependent DM, as in the noninsulin-dependent DM, but a trend toward better outcomes was observed in these subgroups when comparing with SES as well as PES.

The EES was associated with significantly lower rates of all-cause mortality, compared with SES or PES after adjustment for propensity scores (SES vs. EES: HR: 2.02, 95% CI: 1.03 to 3.98; PES vs. EES: HR: 1.69; 95% CI: 1.06 to 2.72) (Figure 20_gr2). A trend toward lower mortality rates with EES was maintained irrespective of insulin treatment but reached significance only for the EES and SES comparison in the insulin-treated group, after stratifying for insulin treatment (Table 2).

No differences were seen in restenosis rates when comparing EES with SES or PES in all DM patients as well as after stratifying for insulin treatment (Figure 20_gr3). A trend toward better outcomes with EES was observed when compared with SES in noninsulin-dependent DM patients (SES vs. EES: HR: 1.76, 95% CI: 0.82 to 3.77, p = 0.14), whereas no differences were observed in insulin-treated DM patients (SES vs. EES: HR: 0.95, 95% CI: 0.48 to 1.88, p = 0.88), p for interaction = 0.001. In the EES versus PES comparison, no differences were observed between groups; however, a small trend for better outcomes with EES was observed in the insulin-treated DM patient group (PES vs. EES: HR: 1.21, 95% CI: 0.74 to 1.98, p = 0.44), whereas the opposite trend was observed in noninsulin-treated DM patient group (PES vs. EES: HR: 0.81, 95% CI: 0.41 to 1.59, p = 0.54), p for interaction <0.001.

Rates of early definite ST were significantly higher for both PES and SES as compared with EES (PES vs. EES: HR: 3.78, 95% CI: 1.31 to 10.39; SES vs. EES: HR: 4.45, 95% CI: 1.24 to 15.91) (Figure 20_gr4). At 1 year EES was associated with numerically lower risk of definite ST compared with PES and SES but reached statistical significance only when compared with SES (Table 2). A trend toward better outcomes with EES as compared with PES was observed in the insulin-dependent DM subgroup (PES vs. EES: HR: 2.37, 95% CI: 0.97 to 5.81, p = 0.06), whereas the opposite was observed in the noninsulin-dependent DM patients (PES vs. EES: HR: 0.69, 95% CI: 0.13 to 3.67, p = 0.67), p for interaction <0.001. At 1 year, better outcomes were observed with EES as compared with SES in both insulin-dependent DM (SES vs. EES: HR: 2.67, 95% CI: 0.76 to 9.44, p = 0.13) and noninsulin-dependent DM subgroups (SES vs. EES: HR: 3.41, 95% CI: 0.70 to 16.57, p = 0.13), p for interaction = 0.31.

The major findings of the present analysis, the largest study to date comparing the outcomes of first- and second-generation DES in DM patients, are: 1) in all-comer DM patients, use of EES was associated with lower rates of composite endpoint of restenosis, definite ST, and all-cause mortality compared with first-generation stents—with significant lower rates when compared with SES, and an important trend also when compared with PES; 2) treatment with EES was associated with a significantly lower all-cause mortality when compared with SES and PES; 3) EES was associated with lower rates of 1-year definite ST when compared with SES and a trend for lower ST when compared with PES; 4) no significant differences in restenosis rates were observed when comparing EES with PES and SES; and 5) there is a significant interaction for restenosis outcomes in EES and PES as well as EES and SES when stratified for noninsulin-treated DM versus insulin-treated DM patients.

Because DM is becoming the leading cause of cardiovascular mortality in the western world, and taking into account the epidemic increase of this condition in countries with fast-developing economies in Asia and South America, the findings of this study—the first in the row to show improved clinical outcomes with newer-generation drug-eluting stent devices when compared with their predecessors—in this high-risk category of patients gain particular importance.

Although the performance of EES and PES in DM patients has been extensively studied in subgroup analysis from large randomized trials as well as patient-level pooled databases, evidence from dedicated randomized trials is still lacking, and there is a paucity of data comparing EES with SES in DM patients ((8),(9),(15),16). However, all available studies comparing EES with SES or PES have been inadequately powered to provide answers on endpoints, such as all-cause mortality and ST, as well as provide insights on treatment effect differences on DM patients when further stratified according to insulin treatment. Because dedicated randomized and adequately powered trials to study such low-frequency events might represent significant organizational or financial challenges, data from large and well-organized national databases can be used to provide insights that can further guide our daily clinical practice and future research, as is the case here.

The observed lower rates of primary endpoint with EES reflects mainly in the lower rates of safety endpoints (ST and mortality), because no significant differences were observed in clinically meaningful restenosis rates, the efficacy endpoint of this study. Strikingly, there was a significant difference in all-cause mortality between EES and SES and PES. Such a finding should mainly be attributed to the magnitude of this analysis, which—with its almost 5,000 patients—represents the largest study that has analyzed the widely used first-generation SES and PES in comparison with a second-generation DES in a DM patient population. As shown in Figures (Figure 20_gr2) and (Figure 20_gr4) and (Table 2), the observed lower mortality rates with EES as compared with SES or PES in DM patients closely parallels the rates of definite ST. In both analyses the differences arise in the early phase, with curves that remain separated in the late phase; and therefore the observed lower mortality rates with EES might be attributed to lower rates in 1-year definite ST. Indeed, previous studies have shown that ST is strongly associated with higher rates of mortality or myocardial infarction (17). Furthermore, these results are in line with the findings of Sarno et al. (18), in an all-comer (DM patients included) population from the SCAAR database, where significant lower rates of definite ST and mortality were observed with newer-generation stents compared with first-generation and bare-metal stents.

The second-generation EES has emerged as a very promising stent with low rates of ST, as concluded from several large prospective randomized trials with moderate-risk or all-comer patients, but this is the first time that such a finding was observed specifically in a large cohort of DM patients ((7),8). Also the pattern of ST reduction with EES observed in this study strongly resembles findings previously reported from the aforementioned trials and pooled database analysis (19). Indeed, the difference in ST was already observed in the early phase and was maintained during the 1-year follow-up. The pathophysiological mechanism(s) underlying the marked lower rates of ST after EES implantation, although speculative, might relate to specific design features of this stent. The combination of thin, fracture-resistant struts; low dose of everolimus; and thrombo-resistant, non-inflammatory proprieties of the fluorinated polymer might contribute to the lower rates of early ST with EES ((20),21). By contrast, as observed in preclinical animal models, the low dose of everolimus elution might prompt a more rapid and complete stent re-endothelialization (21), whereas everolimus, mammalian target of rapamycin inhibition-dependent, selective clearage of macrophages might reduce inflammation (22). The reduction in inflammation, which is even more severe in DM patients than non-DM patients, and the faster and more complete re-endothelialization might in turn relate to lower rates of 1-year ST (23). Such advantages offered from EES as compared with PES in non-DM patients, although in a lesser magnitude, can be observed also in DM patients, as shown from the present analysis (24). Interestingly, significant lower rates of definite ST with EES as compared with PES were also observed in the insulin-treated DM patients, whereas an opposite trend was observed in noninsulin-treated DM patients. Although such findings can be due to chance, it is also known that insulin-treated DM represents the most severe form of this condition, which often has a longer duration and therefore can be expected to be associated with a more severe form of coronary disease. This could be the reason why the rates of ST with PES in the insulin-treated DM patients (as shown in Table 2) were higher, as similarly observed in other high-risk populations (25). However, it cannot be excluded that different treatment effects of EES as compared with PES might be present in DM patients when stratified for insulin treatment; therefore more evidence is needed to clarify this issue.

Another important finding of this study is that, despite its large size, it could not detect any significant differences in restenosis rates between EES and PES as well as between EES and SES. These results corroborate previous findings from other studies comparing EES with PES or SES in DM patients ((9),(10),26). No difference in restenosis rates between EES and PES or between EES and SES could be detected, even when the analysis was performed in DM patients stratified for insulin treatment; however, a significant interaction was detected for EES as compared with PES or SES. Indeed, these findings primarily reflect the severity and complexity of the coronary disease in DM patients. These patients are known to have increased oxidative stress and profound endothelial dysfunction, both factors that are believed to strongly impact intracellular signal transduction. As a result, the cell growth and migration inhibition (anti-restenotic effect) exerted from both paclitaxel or rapamycin analogs is altered in these patients, resulting in a similarly affected efficacy for both rapamycin-analogs or PES ((27),28). The imbalance in efficacy outcomes with EES as compared with PES and SES, as indicated from the respective observed interaction when stratified for insulin-dependent DM and noninsulin-dependent DM, might also rely on the impact of the insulin resistance on the mechanism of action of the active principle eluted from these stents. Although the mechanism of action of paclitaxel might be less-influenced from the presence of DM, the rapamycin (and other rapamycin derivatives) inhibition of mammalian target of rapamycin might be affected by the insulin resistance, because it is regulated from glycosylation-dependent enzymes (29). However, because the intracellular signaling pathways in these patients are very complex and not entirely understood, this interpretation should be only regarded as a hypothesis that needs further investigation. In the case of EES versus SES comparison in insulin-dependent DM and noninsulin-dependent DM patients, although both stents do elute a rapamycin-analog, the polymer and release kinetics of the SES and EES differ from each other in that sirolimus is released at a higher dose and for a longer period than everolimus (30). The SES specific release kinetics might offer some advantages in protection from restenosis in DM patients, the magnitude of which seems to parallel the DM severity. However, although these release kinetics of SES might in a way counterbalance the design-related disadvantages—such as thicker stent struts and polymer layer—and in terms of efficacy lead to similar outcomes as EES, it might render this device more vulnerable to ST, as indeed observed.

In light of these findings, the newer, recently introduced, modern-design rapamycin-analogs or PES with durable bio-absorbable or polymer-free platforms might theoretically represent valid alternatives with regard to both safety and efficacy in percutaneous treatment of DM patients and therefore need to be studied in large and dedicated randomized trials.

This analysis represents a retrospective study; therefore, due to lack of randomization, differences might arise on baseline characteristics. However, adjustments for baseline differences were implemented with propensity score analysis for each individual stent comparison incorporating 22 baseline variables. Despite the efforts to carefully adjust for baseline differences arising from the nonrandomized nature of this study, by means of propensity score adjustment—described in detail in Methods—several baseline differences might have not been completely adjusted (31). Other unmeasured variables might have influenced the study observations. In this study the analysis of device performance was performed on the device level and therefore can differ slightly from endpoints used in other trials. Indeed, the clinically driven restenosis detection—the efficacy endpoint in this study—differs from the classical target vessel revascularization/target lesion revascularization endpoint; however, in our opinion it offers a more accurate evaluation of device efficacy than the classical endpoints, as explained in Methods. Another limitation of this study is that the outcomes of myocardial infarction in each cohort were not studied in this analysis, because myocardial infarction events during follow-up are not routinely entered in the SCAAR database. Furthermore, the results on ST are reported only as definite ST, because the rates of probable ST could not be analyzed, due to the previously mentioned limitation. Finally, this analysis incorporates only patients included in the SCAAR registry; these results represent only the Swedish population and the standard of care. However, Sweden has adopted the European Society of Cardiology guidelines, and therefore we do believe that they can be extrapolated to the rest of the western world.

This study shows that differences in clinical outcomes after PCI with DES in DM patients do exist, although to a lesser extent when compared with non-DM patients, and that treatment with EES is associated with improved clinical outcomes, as compared with first-generation PES and SES, in all-comer DM patients. This finding reflects mainly the lower rates of ST and mortality with EES, which in turn might reflect the improvements in stent design of this second-generation device. These findings should prompt further dedicated and properly powered clinical trials to assess the performance of newer-design rapamycin-analogs or PES in DM patients.