Although uncommon, stent thrombosis (ST) after percutaneous coronary intervention (PCI) is associated with high rates of morbidity and mortality (1- 4). Conceptually, the prognosis following ST may depend on how rapidly reperfusion is restored. One might presume that out-of-hospital ST would have a more dire prognosis than ST occurring in-hospital due to the absence of readily available resuscitation or coronary angiography/PCI capabilities. However, no systematic large series of ST has reported on the potential differences in outcomes when an ST event occurs during a hospitalization versus in an out-of-hospital setting.

We have previously reported that ST occurred in approximately 4% of patients during the 2-year follow-up after primary PCI for ST-segment elevation myocardial infarction (STEMI) in patients enrolled in the HORIZONS-AMI (Harmonizing Outcomes with Revascularization and Stents in Acute Myocardial Infarction) trial (5- 6). In the present analysis, we report the clinical outcomes after in-hospital versus out-of-hospital ST from the HORIZONS-AMI trial, including 3-year follow-up after the original index STEMI and study enrollment.

The HORIZONS-AMI trial has been described in detail previously (5- 6). In brief, 3,602 patients admitted with STEMI presenting within 12 h after symptom onset undergoing a primary PCI management strategy were randomized (1:1) to receive either bivalirudin monotherapy (plus bailout glycoprotein IIb/IIIa inhibitor [GPI]) or unfractionated heparin (UFH) plus a GPI before PCI. Following angiography, stents were implanted in 3,202 patients, including 3,006 who were randomized (3:1) to TAXUS Express2 paclitaxel-eluting stents or otherwise identical bare metal stents (Boston Scientific, Natick, Massachusetts). Clinical follow-up was performed at 30 days, 6 months, 1 year, 2 years, and 3 years. The outcomes of patients in whom ST developed during the hospital period versus out-of-hospital were compared.

Stent thrombosis was defined according to the Academic Research Consortium ST definite or probable criteria (7), and all ST events were adjudicated by an independent clinical events committee blinded to pharmacology and stent assignment after review of original source documentation. The HORIZONS-AMI trial definitions for ischemic and bleeding events were used throughout this study, as previously reported (8).

Categorical variables were compared with Fisher's exact test, and continuous variables with Student's t test. Starting from the time of the first ST event per patient, time-to-event curves were derived using Kaplan-Meier methods. Landmark analyses were utilized to distinguish adverse events occurring in 3 distinct time periods after the occurrence of ST: 0 to 7 days, 7 to 30 days, and 30 days to 1 year. Time-to-event data were compared using log-rank tests and univariate Cox proportional hazards methods.

Logistic regression was used to identify the independent predictors of in-hospital and out-of-hospital ST. Two multivariable time-updated covariate-adjusted Cox proportional hazards models were used to investigate the relative impact of in- versus out-of-hospital ST on mortality. To prevent over-fitting, the first model utilized a limited number of covariates, and the second model used a propensity score covariate. The following covariates were included in the first model: diabetes mellitus, randomization to treatment with UFH plus a GPI or bivalirudin, and current smoking. These covariates were selected because of their previously described clinical relationship to ST (9). Additionally, pre-procedural Thrombolysis In Myocardial Infarction (TIMI) flow grade 0/1 (vs. >1), final post-procedural TIMI flow grade 3 (vs. <3), peak creatine phosphokinase, and protocol-defined non-coronary artery bypass graft surgery–related major bleeding were added to the model because of the well-known relationship with mortality after STEMI. The multivariable model was built by stepwise variable selection with entry and exit criteria set at the p = 0.1 level. The propensity score was estimated from a logistic regression model for occurrence of in- or out-of-hospital ST. To calculate the propensity score, the following variables were entered into the model: baseline TIMI flow grade 1/0, final TIMI flow grade 3, peak creatine phosphokinase, protocol-defined non-coronary artery bypass graft surgery–related major bleeding, pre-randomization heparin, age, current smoking, randomization to bivalirudin versus UFH plus a GPI, diabetes mellitus, and platelet count (Online Appendix). The propensity score was added as the only covariate in addition to in- versus out-of-hospital ST in the second multivariable Cox model. Formal interaction tests were performed to determine whether diabetes or randomization to bivalirudin versus UFH+GPI affected the relative risk of mortality after in-hospital versus out-of-hospital ST.

Of 3,202 patients in whom stents were implanted, definite/probable ST within 3 years occurred in 156 patients (4.9%). In-hospital ST occurred in 54 patients (34.6%), and out-of-hospital ST occurred in 102 patients (65.4%). The majority of ST events in both groups were definite ST events (46 [85.2%] in-hospital vs. 90 [88.2%] out-of-hospital; p = 0.78). Five patients had 2 ST events each, all of which were definite ST. Two of these patients first had an in-hospital ST, followed 28 and 47 days later by an out-of hospital event; both of these patients were included in the in-hospital ST group. Three patients had 2 out-of-hospital ST events.

The median time to first ST was 1 day (interquartile range: 0 to 5 days] in the in-hospital group and 360 days (interquartile range: 100 to 604 days) in the out-of-hospital group. Mean length of hospital stay was 8.67 ± 6.63 days versus 5.65 ± 5.86 days in the in-hospital versus out-of-hospital groups (p < 0.01).

Baseline clinical and laboratory characteristics in the 2 groups are shown in (Table 1). Patients with in-hospital compared with out-of-hospital ST were less frequently diabetic and current smokers, but more frequently had thrombocytopenia. No significant differences were present in other oral medical therapies prescribed before or during the index hospitalization (data not shown), except for beta-blocker use, which was less frequently used during the index hospitalization in patients with in-hospital compared with out-of-hospital ST (83.3% vs. 96.1%, respectively; p = 0.01). Angiographic and procedural characteristics during the index STEMI treatment are summarized in (Table 2). In comparison with patients with out-of-hospital ST, those with in-hospital ST more frequently had pre-PCI coronary TIMI flow grade 0/1 and post-PCI coronary TIMI flow grade <3 were more likely to have been randomized to bivalirudin monotherapy, and were less likely to have received pre-randomization heparin. By multivariable analysis, baseline TIMI flow grade 0/1, nonuse of pre-randomization heparin, and randomization to bivalirudin were independent predictors of in-hospital ST, whereas insulin-treated diabetes, prior PCI, cigarette smoking, an elevated platelet count, and randomization to heparin plus a GPI were independent predictors of out-of-hospital ST (Table 3).

As shown in (Table 4), the 1-year rates of mortality and major bleeding were higher in patients with in-hospital compared with out-of-hospital definite or probable ST, although the rates of reinfarction were lower. Most of the differences in clinical outcomes (death, myocardial infarction, and bleeding) appeared within 1 week of the ST episode (Figure 1). The results were similar when only definite ST events were analyzed (Figure 2); patients with in-hospital compared with out-of-hospital definite ST had significantly higher 1-year mortality (15.2% vs. 2.3%; p < 0.01) and major bleeding (23.5% vs. 6.7%; p < 0.01), with lower rates of MI (65.5% vs. 82.2%; p = 0.03).

Subgroup analysis indicated that within the in-hospital ST group, mortality after acute ST was 7.1% and after subacute ST was 50% (p = 0.0004). Within the out-of-hospital ST group, mortality was 27.6% after subacute ST, 3.3% after late ST, and 8.0% after very late ST (p = 0.02). Detailed data on baseline clinical and angiographic characteristics as well as clinical outcome after acute, subacute, late, or very late ST are included in the Online Appendix.

By multivariable analysis, in-hospital (compared with out-of-hospital) definite/probable ST was a powerful independent predictor of mortality (Table 5). Diabetes was also associated with increased mortality after an ST event, whereas randomization to bivalirudin (as opposed to UFH+GPI) was associated with improved survival after an ST event ((Table 4), Figure 4). Multivariate models for in- versus out-of-hospital definite ST demonstrated directionally similar results to those for definite/probable ST (Table 5).

With 156 ST events, the present study is 1 of the largest to analyze the relationship between ST and subsequent mortality. We observed that in patients with STEMI undergoing primary PCI with stent implantation, more than one-third of all ST events during the 3-year follow-up period occurred during the hospital phase, shortly after the index procedure. One-year mortality after ST was significantly higher in patients with in-hospital compared with out-of-hospital ST, despite more rapid access to care and similar rates of target lesion revascularization. The occurrence of in-hospital compared with out-of-hospital definite or probable ST was an independent predictor of greater mortality after multivariate adjustment, both with and without the use of a propensity score. In-hospital compared with out-of-hospital ST was also associated with greater rates of major bleeding, but fewer myocardial infarctions.

In the present analysis, despite being distanced from immediate clinical care, out-of-hospital ST was associated with lower subsequent 1-year mortality than in-hospital ST. Importantly, the results were similar when only angiographically confirmed definite ST was considered, which eliminates any possible confounding by the inclusion of unexplained deaths within probable ST as defined by the Academic Research Consortium criteria. Thus, early ST after primary PCI in STEMI may portend a particular poor prognosis, presumably due to the incremental loss of myocardium, than in patients with ST after an elective stent procedure. These findings are consistent with a report by Lasala et al., who studied 184 patients with ST after paclitaxel-eluting stent implantation (10). Mortality was higher after acute (29.4%) or subacute ST (41.7%) compared with late (12%) or very late ST (13%). In contrast, the RESTART (Registry of Stent Thrombosis for Review and Reevaluation) investigators reported similar rates of 1-year mortality in patients with acute or subacute ST (22.4%) and late ST (23.5%), although mortality was less frequent in patients with very late ST (10.5%) (2).

Major bleeding has been shown to be a powerful independent predictor of subsequent mortality (11- 12), and the higher rates of major bleeding in patients with in-hospital compared with out-of-hospital ST may have contributed to the greater mortality of early ST. A recent subanalysis from the HORIZONS-AMI study reported an unadjusted association between in-hospital major bleeding after STEMI PCI and ST at 3-year follow-up (13). Indeed, occurrence of major bleeding characterizes an overall high-risk patient group. This reflects a population-based observation and should not be construed as a statement implying a causal relationship between bleeding and subsequent ST. In fact, multivariate Cox proportional hazards analysis in the present study did not find in-hospital stent thrombosis to be related to a preceding bleeding event. The question of whether bleeding causes ST, by triggering the coagulation cascade and activating platelets, or whether treatment of ST with femoral reinstrumentation, and antithrombotic and antiplatelet therapy, leads to bleeding is difficult to answer. Alternatively, it is not unthinkable that subclinical bleeding may have led to platelet activation, precipitating ST, with subsequent escalation of bleeding as a result of the treatment for ST. Further study is required in order to answer this question with confidence.

Conversely, it may be supposed that the lower rate of reinfarction in patients with in-hospital ST would be protective from death (11- 12). Differences in the underlying pathophysiological mechanism may also be partly responsible for the differences of in-and out-of-hospital ST on mortality. During the early phase, technical and procedural factors are important predictors of ST (1,14- 15), whereas delayed neointimal coverage and ongoing vessel inflammation are associated with late ST (16).

In the present study survival after ST was also worse in patients with diabetes, whereas use of bivalirudin (as opposed to heparin plus a GPI) was associated with lower mortality after ST. The favorable effect of bivalirudin was principally present in patients with in-hospital ST, whereas the effect of diabetes did not have any interaction with the previous variables. Thus, despite the higher rate of acute ST with bivalirudin compared with heparin plus a GPI in STEMI patients undergoing primary PCI (7), the impact of ST on mortality may be less severe when it occurs early after bivalirudin. While the reason for this observation is unknown (and the play of chance cannot be excluded), it may be due to the reduced rate of major bleeding with bivalirudin compared to heparin plus GPI (especially because bleeding complications were increased in patients with in-hospital compared to out-of-hospital ST).

This study was a retrospective analysis from a randomized trial and as such should be considered exploratory. We did not have any information on antiplatelet hyporesponsiveness or related genotypes. Intravascular ultrasound imaging was scarce, not enabling evaluation of the role of angiographically “silent” events such as stent malapposition, underexpansion, or edge dissection. The fact that the original stent implantation was performed during STEMI may have contributed not only to the high rates of early and late stent thrombosis, and may also have relatively worsened the outcome of early versus late ST. Further studies are required to determine whether an early in-hospital ST event may be less catastrophic in patients undergoing elective stent implantation. (Figure 3).

For supplemental tables, please see the online version of this article.