Introduction

Although coronary angioplasty with routine implantation of bare-metal stents has been established as the reperfusion strategy of choice for patients with acute myocardial infarction [2, 8, 13, 22, 23, 32, 34, 51], the long-term success of this strategy has been hampered by the frequent development of restenosis [28, 44, 46, 51]. Drug-eluting stents have been shown to reduce the restenosis risk and need for repeat revascularization procedures associated with use of bare-metal stents in various lesion and patient subsets [12, 20]. However, there has been limited evidence on the role of drug-eluting stents in patients with acute myocardial infarction. The benefits and safety issues related to the use of these devices have been evaluated in several randomized and non-randomized studies. While the superior clinical efficacy of drug-eluting stents has not been consistently shown in these studies [10, 25, 30, 40, 42, 49, 50], there have been reports that use of drug-eluting stents during primary percutaneous coronary intervention could be associated with an increased risk of stent thrombosis [6, 16, 35]. After performing a meta-analysis of randomized trials of patients with acute myocardial infarction we found that drug-eluting stent reduce reintervention rate without increasing the risk of adverse outcomes as compared to bare-metal stents [19]. However, the number of patients studied was not large enough to allow a reliable estimation of rare adverse events. In a recent statement, the Food and Drug Administration considered that “off label” use of drug-eluting stents, such as their use in patients with acute myocardial infarction, was associated with an increased risk of adverse outcome compared with “on label” use and asked for additional data to determine optimal treatments for complex patients [47].

Since we performed our meta-analysis of 8 randomized trials including 2,786 patients [19], several other randomized trials comparing drug-eluting stents with bare-metal stents in acute myocardial infarction, including Harmonizing Outcomes with Revascularization and Stents in Acute Myocardial Infarction (HORIZONS-AMI) study, have been completed [4, 11, 18, 21, 45, 48]. Important new data obtained from the evaluation of 5,000 additional patients enrolled in these trials have now become available. Furthermore, the new trials have provided more information on the paclitaxel-eluting stent which, compared to the sirolimus-eluting stent, has been less frequently studied in the early trials of drug-eluting stents in patients with acute myocardial infarction. Therefore, we incorporated the new evidence into an updated meta-analysis to consolidate and extend current knowledge on the safety and efficacy of drug-eluting stents after primary percutaneous coronary intervention.

Methods

Search and selection process

We searched the MEDLINE database, the Cochrane Central Register of Controlled Trials, and Internet-based sources of information on the results of clinical trials in cardiology (http://www.cardiosource.com/clinicaltrials, http://www.theheart.org, http://www.clinicaltrialresults.com, http://www.clinicaltrials.gov, http://www.europcr.com, and http://www.tctmd.com) through January 2010. We used these key words: “acute myocardial infarction”, “primary angioplasty”, “stenting”, “bare-metal stent”, “drug-eluting stent”, “sirolimus-eluting stent”, “paclitaxel-eluting stent”, “randomized trial”. We also identified previous meta-analyses and relevant reviews and editorials from major medical journals published within the last year and assessed these sources for possible information on trials of interest.

Eligible for this meta-analysis was randomized studies comparing drug-eluting stents with bare-metal stents in patients with acute myocardial infarction undergoing primary percutaneous coronary intervention, if information on outcomes of interest were reported or made available by the trial investigators for a mean follow-up period of at least 6 months. No restriction criteria were imposed with regard to the form of study publication.

Data extraction and quality assessment

Studies were searched and reviewed independently by two of the authors (A.D. and K.T.); those meeting the inclusion criteria were selected for further analysis. A total of 15 trials were identified [4, 10, 11, 18, 21, 25, 30, 36, 40, 42, 45, 48–50]. The trial of Pasceri et al. [36] was excluded because it reported only preliminary data of the first 34 patients over a follow-up of 4 months. Finally, 14 trials, shown in Table 1, were included in this meta-analysis [4, 10, 11, 14, 18, 21, 25, 30, 36, 40, 42, 45, 48–50].

Table 1 Main characteristics of the trials
Full size table

Individual patient data were obtained from primary investigators for 9 trials with 3,846 patients (Table 1). Summary data were obtained from publication source for 4 trials. In the remaining trial, we used risk estimates reported in the presentation of the results of this trial [45].

The outcomes of interest were the composite of all-cause death or recurrent myocardial infarction (primary safety endpoint), reintervention (primary efficacy endpoint), all-cause death, recurrent myocardial infarction, and stent thrombosis. The definitions of outcomes of interest used in individual studies are shown in Table 2. To assure consistency in the definition of stent thrombosis across different trials we used only figures of definite stent thrombosis from those trials, where stent thrombosis was defined according to the classification of Academic Research Consortium [5].

Table 2 End point definitions in each trial
Full size table

Each trial was evaluated for the adequacy of allocation concealment, performance of the analysis according to the intention-to-treat principle, and blind assessment of the outcomes of interest. We used the criteria recommended by Altman et al. [1] and Jüni et al. [17] to assess the adequacy of allocation concealment. In 3 trials, a modified intention-to-treat principle, i.e. exclusion of patients who did not receive the study stent, was used [11, 42, 50].

Data synthesis and analysis

Hazard ratios (HRs) with 95% confidence intervals (CIs) were computed as summary statistics. The pooled HRs were calculated using the DerSimonian and Laird method for random effects [9]. The individual study HRs were calculated using the Mantel–Cox test for 9 trials with individual patient data (Table 1). For the HORIZONS-AMI trial we used the hazard ratios as recently reported [45]. For the remaining 4 trials, risk estimates were calculated based on the summary data obtained from publication source (Table 1).

We used the Cochran-test to assess heterogeneity across trials. We also calculated the I 2 statistic to measure the consistency between trials with values of 25, 50, and 75% showing respectively, low, moderate, and high heterogeneity [15].

We performed sensitivity analyses by comparing the treatment effects obtained with each trial removed consecutively from the analysis with the overall treatment effects. Meta-regression analysis was used to explore the influence of the recommended duration of clopidogrel treatment and length of follow-up on the main safety endpoint as well as the influence of the recommended duration of clopidogrel treatment on stent thrombosis. We assessed publication bias with respect to the primary endpoint of safety—the composite of death and recurrent myocardial infarction was evaluated using a funnel plot as well as the adjusted rank correlation test [3]. Results were considered statistically significant at two-sided P < 0.05. Statistical analysis was performed by using the Stata software, version 9.2 (Stata Corp, College Station, Tex).

Results

Included in this meta-analysis were 14 trials with 7,781 patients (Fig. 1). Table 1 shows the main characteristics of these trials. Patient populations of individual trials varied from 80 patients to 3,006 patients. The mean age of study participants of individual studies varied from 59 to 64 years. The sirolimus-eluting stent was the single drug-eluting stent evaluated in 6 trials [11, 30, 42, 48–50], the paclitaxel-eluting stent was the single drug-eluting stent evaluated in 5 trials [4, 14, 18, 25, 45], both sirolimus and paclitaxel-eluting stents were evaluated in 3 trials [10, 21, 40]. In one of the trials either the sirolimus-eluting stent, the paclitaxel-eluting stent, or the zotarolimus-eluting stent was compared to the bare-metal control [21]. The recommended duration of thienopyridine therapy after the procedure was 3 [48, 49], 6 [10, 18, 25, 40, 42], 9 [4, 11], and 12 months [14, 21, 30, 50]. Average length of follow-up varied from 7 to 24 months. A clinical endpoint was the primary endpoint in 8 studies [10, 11, 18, 25, 40, 42, 48]. Angiographic restenosis, late loss or stent volume obstruction by intravascular ultrasound were the primary endpoints in 5 trials [4, 14, 21, 30, 50]. One study used a composite of clinical and angiographic outcomes as the primary endpoint [49].

Fig. 1
figure 1

Flowchart of included studies. DES drug-eluting stent, BMS bare-metal stent

Full size image

Figure 2 shows the hazard ratio of the primary safety endpoint—the composite of all-cause death or recurrent myocardial infarction—associated with drug-eluting stents versus bare-metal stents. There was no significant difference in the hazard of this endpoint between the two treatment modalities using the random effects model: pooled HR, 0.91 (95% CI 0.75–1.09; P = 0.28). The fixed effects model yielded identical results: pooled HR, 0.91 (95% CI 0.75–1.09; P = 0.28). There was no significant heterogeneity (P = 0.55) and no inconsistency across trials (I 2 = 0%). Sequential exclusion of each individual trial from the analysis of overall mortality yielded HRs that ranged from 0.87 (95% CI 0.72–1.04) to 0.94 (95% CI 0.77–1.13) and were not significantly different from the overall HR (P ≥ 0.74). Meta-regression analysis showed that neither the length of follow-up (P = 0.98) nor the recommended duration of clopidogrel treatment (P = 0.70) had any influence on the overall result.

Fig. 2
figure 2

Hazard ratios of all-cause death or recurrent myocardial infarction associated with drug-eluting stent versus bare-metal stent. DES drug-eluting stent, BMS bare-metal stent, CI confidence interval. The size of the data marker is proportional to the weight of the individual studies

Full size image

Figure 3 shows the hazard ratio of the primary efficacy endpoint—reintervention—associated with drug-eluting stents versus bare-metal stents. There was a large reduction in the hazard of reintervention with drug-eluting stents compared to bare-metal stents using the random effects model: (pooled HR, 0.41 [95% CI 0.32–0.52; P < 0.001]). The fixed effects model yielded similar results: pooled HR, 0.44 (95% CI 0.36–0.52; P < 0.001). There was no significant heterogeneity (P = 0.14) and low-to-moderate inconsistency across trials (I 2 = 29.9%). Sequential exclusion of each individual trial from the analysis of death yielded HRs that ranged from 0.38 (95% CI 0.30–0.48) to 0.43 (95% CI 0.34–0.55) and were not significantly different from the overall HR (P ≥ 0.67). Meta-regression analysis showed that neither the type of drug-eluting stent (P = 0.27) nor the length of follow-up (P = 0.51) had any influence on the overall result. We evaluated separately the outcome of reintervention for trials with angiographic follow-up (pooled HR, 0.36 [95% CI 0.27–0.48]) and without angiographic follow-up (pooled HR, 0.54 [95% CI 0.36–0.8]) and we found no significant difference in the results (P for interaction = 0.12).

Fig. 3
figure 3

Hazard ratios of reintervention associated with drug-eluting stent versus bare-metal stent. DES drug-eluting stent, BMS bare-metal stent, CI confidence interval. The size of the data marker is proportional to the weight of the individual studies

Full size image

Figure 4 shows the hazard ratio of all-cause death associated with drug-eluting stents versus bare-metal stents. There was no significant difference in the hazard of all-cause death between the two treatment modalities using the random effects model: (pooled HR, 0.90 [95% CI 0.71–1.15; P = 0.41]). The fixed effects model identical similar results: HR, 0.90 ([95% CI 0.71–1.15; P = 0.41).There was no significant heterogeneity (P = 0.51) and no inconsistency across trials (I 2 = 0%).

Fig. 4
figure 4

Hazard ratios of all-cause death associated with drug-eluting stent versus bare-metal stent. DES drug-eluting stent, BMS bare-metal stent, CI confidence interval. The size of the data marker is proportional to the weight of the individual studies

Full size image

Figure 5 shows the hazard ratio of recurrent myocardial infarction associated with drug-eluting stents versus bare-metal stents. The hazard of myocardial infarction was lower among patients treated with drug-eluting stents versus bare-metal stents using the random effects model, although the difference did not achieve statistical significance (pooled HR, 0.81 (95% CI 0.63–1.04; P = 0.10]). The fixed effects model yielded identical results: HR, 0.81 (95% CI 0.63–1.04; P = 0.10). There was no significant heterogeneity (P = 0.73) and no inconsistency across trials (I 2 = 0%).

Fig. 5
figure 5

Hazard ratios of recurrent myocardial infarction associated with drug-eluting stent versus bare-metal stent. DES drug-eluting stent, BMS bare-metal stent, CI confidence interval. The size of the data marker is proportional to the weight of the individual studies

Full size image

Figure 6 shows the hazard ratio of stent thrombosis associated with drug-eluting stents versus bare-metal stents. The hazard of stent thrombosis was comparable between patients treated with drug-eluting stents compared to patients treated with bare-metal stents using the random effects model: (pooled HR, 0.84 [95% CI 0.61–1.17; P = 0.30]). The fixed effects model yielded identical results: pooled HR, 0.84 (95% CI 0.61–1.17; P = 0.30). We found no significant heterogeneity (P = 0.70) and no inconsistency across trials (I 2 = 0%). Meta-regression analysis showed that the recommended duration of clopidogrel treatment (P = 0.80) had no influence on stent thrombosis.

Fig. 6
figure 6

Hazard ratios of stent thrombosis associated with drug-eluting stent versus bare-metal stent. DES drug-eluting stent, BMS bare-metal stent, CI confidence interval. The size of the data marker is proportional to the weight of the individual studies

Full size image

We found no evidence of publication bias with respect to the composite of death or recurrent myocardial infarction using the Begg funnel plot (Fig. 7) and rank correlation test (P = 0.75).

Fig. 7
figure 7

Funnel plot (with pseudo 95% confidence intervals) to detect possible publication bias

Full size image

Finally, we compared sirolimus-eluting stents with paclitaxel-eluting stents regarding the different outcomes evaluated in this meta-analysis. The pooled hazard ratio for the outcome of death or recurrent myocardial infarction was 0.71 (95% CI 0.54–0.93) for trials evaluating sirolimus-eluting stents and 1.02 (95% CI 0.81–1.28) for trials evaluating paclitaxel-eluting stents; we found a significant difference in the results (P for interaction = 0.05). The pooled hazard ratio for the outcome of reintervention was 0.30 (95% CI 0.22–0.41) for trials evaluating sirolimus-eluting stents and 0.56 (95% CI 0.45–0.70) for trials evaluating paclitaxel-eluting stents; the difference in the results (P for interaction = 0.001). In addition, we analyzed separately the outcomes of death, recurrent myocardial, and stent thrombosis for trials evaluating sirolimus-eluting stents and trials evaluating paclitaxel-eluting stents and we found no significant difference in the results (P for interaction = 0.41, 0.23, and 0.68, respectively).

Discussion

This updated meta-analysis has several strengths. First, inclusion of HORIZONS-AMI and several other trials that were recently completed more than doubled the number of patients who were evaluated in the prior meta-analyses [7, 19, 37]. Availability of a large amount of data regarding the outcome of 7,781 patients (4,727 treated with drug-eluting stents) enrolled in 14 randomized trials enabled us to perform a more reliable meta-analytic evaluation of rare adverse events and benefits related with the use of drug-eluting stents as compared with bare-metal stents in acute myocardial infarction. Second, the new trials included a large number of patients treated with paclitaxel-eluting stents. Earlier studies mainly focused on the sirolimus-eluting stent and evidence on the outcome of patients treated with the paclitaxel-eluting stent was limited. Inclusion of more than 2,500 additional patients treated with paclitaxel-eluting stents in this meta-analysis allows drawing reliable conclusions regarding the performance of both sirolimus-eluting and paclitaxel-eluting stents in patients undergoing percutaneous coronary intervention for acute myocardial infarction.

Coronary implantation of drug-eluting stents in patients with acute myocardial infarction remains an “off label” indication [47]. Although some of the studies in this field have questioned the clinical efficacy of drug-eluting stents versus bare-metal stents [25], the main concern related to their use in patients with acute myocardial infarction has been an increased risk of adverse outcomes [39]. Several studies have reported that, compared to bare-metal stents, sirolimus-eluting and paclitaxel-eluting stents provoke delayed healing, impaired endothelialization, vascular dysfunction, hypersensitivity reaction, and persistent fibrin deposition which produce the conditions that have been related to the development of stent thrombosis [27]. Recently, findings from an autopsy study suggested an increased risk of thrombotic complications in patients treated with drug-eluting stents for acute myocardial infarction and this was explained with the delayed vessel healing at the culprit site in patients with acute myocardial infarction as compared to patients with stable angina [31].

There have been several observational studies that have linked drug-eluting stent implantation in acute myocardial infarction with an increased risk of stent thrombosis. Park et al. [35] found that primary stenting with implantation of sirolimus-eluting or paclitaxel-eluting stents in patients with acute myocardial infarction was a major predictor for acute and subacute stent thrombosis. Sianos et al. [41] reported that a larger thrombus burden was associated with a higher risk of stent thrombosis. However, registry studies of patients with acute ST-segment elevation myocardial infarction have not shown an increased risk of stent thrombosis or thrombosis-related events with drug-eluting stents as compared with bare-metal stents [26, 33, 38].

Our meta-analysis shows that implantation of drug-eluting stents in patients with acute myocardial infarction is as safe as implantation of bare-metal stents. The pooled hazard ratios for the composite of all-cause death or recurrent myocardial infarction, all-cause death, recurrent myocardial infarction, and stent thrombosis showed a similar hazard for these adverse outcomes among patients treated with drug-eluting and bare-metal stents. Meta-regression analysis did not identify any influence of the type of drug-eluting stent used, either sirolimus-eluting or paclitaxel-eluting stent, on the outcome of mortality. Thus, the use of both drug-eluting stents is safe. It has been suggested that drug-eluting stents increase the risk of late adverse outcome [39]. Mean length of follow-up in the studies included in our meta-analysis varied from 7 to 24 months. Meta-regression analysis that was used to explore the influence of duration of follow-up on all-cause death did not identify any impact of follow-up duration on the primary safety outcome. On the other hand, the 3-year data reported from Kukreja et al. [24] showed not only a similar risk of adverse events among patients with acute myocardial infarction treated with drug-eluting stents as compared to patients treated with bare-metal stents, but even a trend toward a better outcome with one of the drug-eluting stents. Furthermore, the 2-year data from Massachusetts registry showed a decreased risk of mortality with drug-eluting stents [29]. However, the results from these registry studies should be interpreted with caution.

We found that use of drug-eluting stents was associated with a 59% reduction in the hazard of reintervention. This figure clearly shows the significant superiority of drug-eluting stents over bare-metal stents in patients with acute myocardial infarction. The overall result was not influenced by the length of follow-up. Similarly, the type of drug-eluting stent used had no influence on the outcome of reintervention, which shows the efficacy of both drug-eluting stents, the sirolimus-eluting, and paclitaxel-eluting stent. The finding of this meta-analysis regarding the superior efficacy of drug-eluting stents as compared to bare-metal stents in patients with acute myocardial infarction is in the same line with other studies and meta-analyses that have shown the efficacy of drug-eluting stents in other settings of complex patients and lesions [12, 43].

In some of the studies included in this meta-analysis drug-eluting stents were not shown superior to bare-metal stents. This was observed more frequently in those studies which enrolled smaller number of patients or used the paclitaxel-eluting stent as a comparator. Although it has been suggested that routine angiographic follow-up might increase reintervention rates thereby influencing study outcomes, we did not find any difference in this outcome between studies with or without angiographic follow-up.

This meta-analysis has several limitations. Its results cannot be extended to drug-eluting stents other than the sirolimus-eluting and paclitaxel-eluting stents. The maximum length of follow-up reported in studies included in our meta-analysis was 24 months. Considering the concern related to the potential for increased risk of late thrombotic events in patients treated with drug-eluting stents, it is necessary to perform studies with longer follow-up to provide more reassurance on the safety of using drug-eluting stents in patients with acute myocardial infarction. Furthermore, despite the increased statistical power of this meta-analysis, the failure to detect a clinically relevant difference between drug-eluting stents and bare-metal stents regarding the occurrence of rare adverse events is still possible. About half of the patients evaluated in this meta-analysis were contributed by the HORIZONS-AMI trial. No patient-level data were available for this trial and this should be considered as a limitation of this meta-analysis, despite the fact that the removal of this trial from the general analysis did not influence the overall result. Finally, patients with some high-risk characteristics, such as those with cardiogenic shock, have been excluded from the trials that were evaluated in this meta-analysis. Therefore, our conclusion cannot be extrapolated to the whole population of patients with acute myocardial infarction.

In conclusion, the results of this meta-analysis show that during medium-term follow-up the use of drug-eluting stents in patients with acute myocardial infarction is safe and results in a marked reduction in the need of reintervention.