Short-Term Consequences of Angiographically-Confirmed Coronary Stent Thrombosis

Objectives To conduct a meta-analysis to quantify the real-world incidence of in-hospital or 30-day death or myocardial infarction (MI), and angiographically-confirmed ST-related treatment costs. Background The short-term clinical and economic consequences of coronary stent thrombosis (ST) are thought to be significant. Methods We searched MEDLINE, Embase and Scopus from January 2000-July 2012 to identify observational/registry studies that evaluated a cohort of ≥25 patients experiencing angiographically-confirmed thrombosis of a drug-eluting or bare-metal stent, required the use of dual-antiplatelet therapy for guideline-recommended durations, and reported incidences of in-hospital or 30-day death or MI and/or ST-related treatment costs. Incidences and costs from each study were pooled using random-effects meta-analysis. Results Twenty-three studies were included. Of the 13 studies reporting in-hospital outcomes, 12 (N=8,832 STs) reported mortality data, with the pooled incidence rate estimated to be 7.9%, 95%CI=5.4%-11.3%, I2=86%. Ten studies (N=1,294 STs) reported 30-day death, with a pooled incidence of 11.6%, 95%CI=8.8%-15.1%, I2=55%. Patients experiencing early ST (within 30-days of implant) had higher in-hospital and 30-day mortality than those experiencing very-late ST (interaction p<0.04 for both). Stent type had no significant effect on in-hospital or 30-day mortality. In the 5 studies (N=542 STs) and 3 studies (N=180 STs) reporting in-hospital and 30-day MI, respectively, the pooled incidence rates were 6.1%, 95%CI=2.1%-16.2%, I2=88% and 9.5%, 95%CI=3.8%-22.0%, I2=65%. One study reported costs associated with ST, estimating the median/patient cost of hospitalization to treat early ST at $11,134 (in 2000US$). Conclusions Regardless of stent type used, the short-term consequences of coronary ST appear significant.


Introduction
The clinical consequences for patients experiencing coronary stent thrombosis (ST) are thought to be dire [1][2][3]. A pooled analysis of six multicenter bare-metal coronary stent clinical trials demonstrated patients experiencing an angiographicallyconfirmed ST had a 30-day incidence of the combined endpoint of death or myocardial infarction (MI) of 64% [3]. However, the incidence of death and MI observed in this analysis may not accurately reflect what would be seen in a real-world population. In addition to the clinical consequences, the occurrence of ST has been associated with a significant economic impact as well, as depicted by retrospective study estimating the median total hospital cost to treat a ST to be upwards of $11,134 per patient (in 2000 U.S. dollars) [4].
As no such systematic review and meta-analysis of realworld data/studies has been published to date, we sought to conduct a meta-analysis to better quantify the real-world incidence of in-hospital or 30-day death or MI and/or ST-related treatment costs derived from observational studies and coronary stent registries.

Methods
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for the reporting of this systematic review and meta-analysis [5].
A systematic literature search using the Medline (January 2000-July 2012), Embase (January 2000-July 2012) and Scopus (January 2000-July 2012) computerized databases was conducted. The search began at the year 2000 (inclusive) to limit the identification of studies using outdated practices (outdated stents and implantation practices, or no/suboptimal use of guideline recommended dual antiplatelet therapy (DAPT)). The complete search strategy for Medline is available in Appendix S1 in File S1. Review of the reference sections of eligible studies as well as review articles was also performed to identify additional relevant studies.
To be included in this systematic review and meta-analysis, studies had to 1) be an observational or registry study (to assess "real-world" outcomes), 2) evaluate a cohort of ≥25 patients experiencing angiographically-confirmed (definite) thrombosis of a drug-eluting stent (DES) or bare-metal stent (BMS) [6], 3) require the use of DAPT according to guideline recommendations of the time, 4) provide data on the incidence of in-hospital or 30-day death or MI (new or re-infarction, not counting the ST-defining ischemia) and/or ST-related treatment costs, and 5) be a full-text publication in the English language. In studies with more than one published report on the same study population, the most recent publication was selected for analysis to avoid double-counting participants, although previous publications were reviewed to supplement for missing/ additional data where applicable. When studies with nonidentical but overlapping populations were identified, the most inclusive publication was selected for analysis. When needed, authors of identified studies were contacted for clarification or additional data. In all situations, two investigators (CGK and MA) determined study eligibility independently, with disagreements resolved by discussion or by a third investigator (CIC).
Two investigators (CGK and CIC) performed all data extraction. Data collected for each study included author and publication year, number of STs identified, timing of data collection (prospective or retrospective), inclusion/exclusion criteria, stent type(s) evaluated (BMS and/or DES), country where study was conducted, time frame for patient/case inclusion, definition of ST [6], timing of ST (acute, sub-acute, early, late, very late) [6], use of DAPT at time of ST, funding source and role, and the incidence of in-hospital or 30-day death or MI and ST-related treatment costs.
The validity of included studies for assessing the incidence of in-hospital or 30-day death or MI and thrombosis-related treatment costs was assessed using the assessment tool included in Appendix S2 in File S1. The tool assessed the following attributes : (1) whether post-ST outcomes and costs were the primary endpoints of the identified study, 2) whether the inclusion/exclusion criteria for the study were clearly described, 3) adequacy of sampling by assessing whether consecutive and unselected patients/ST cases were evaluated, 4) use of prospective data collection, 5) whether the methods of data collection for ST and outcomes data, completeness of follow-up and reasons for loss-to-follow-up were clearly described, 6) the use of a standard/acceptable definition of ST (Appendix S3 in File S1) and MI [6], 7) independent adjudication of ST and MI diagnoses, 8) whether studies' discussion and conclusion sections were consistent with their results, 9) discussion of previous studies evaluating post-ST outcomes and study limitations were provided, and 10) study funding and the role of the funder in the research was described. We did not attempt to give a summary validity rating to each individual study in this systematic review. Study validity was conducted for informational purposes only, and was not used as an exclusion criterion or to perform any type of statistical adjustment.
Numerators (n = number of patients reporting the outcome of in-hospital or 30-day mortality or MI) and denominators (N = total number of STs evaluated) were extracted from each study in order to compute incidences with accompanying 95% confidence intervals (CIs). Proportions were then pooled using DerSimonian-Laird weights (a random-effects model). Between-study heterogeneity was assessed using the I 2 statistic with a threshold of 50% used to define an important degree of heterogeneity. To assess for the potential for publication bias, we reviewed Egger's weighted regression statistic p values (with p <0.05 suggesting a higher likelihood of publication bias). As only one study reporting ST-related treatment costs was identified, no meta-analysis was performed for this endpoint.
We conducted various subgroup analyses to examine the effect of ST timing (early, late and very late) and stent type (BMS and DES only analyses) on the meta-analysis' results. We considered P <0.05 statistically significant for all analyses. All analysis was conducted using Comprehensive Meta-Analysis version 2 (Biostat, Englewood, NJ).
Included studies were published between 2000 and 2012 (patient inclusion between 1993 and 2010) and followed as few as 23 and as many as 7,315 ST cases. The patients experiencing angiographically-confirmed ST came from various geographic regions, 30% in the United States as well as France, Hungry, Israel, Italy, Japan, Netherlands, South Korea, Spain and Switzerland. Eleven studies included a mix of patients with both BMS and DES, 10 studies included a single stent type and two did not report stent type. While inclusion criteria required all patients to have angiographically-confirmed ST, a majority of included studies (87%) required patients meet the Academic Research Consortium's (ARC's) criteria for "definite" ST [acute myocardial ischemia (ECG major ST abnormality or any biomarker elevation) and angiographic or autopsy evidence of stent occlusion or thrombus] or some close variation (two of the three studies not meeting the ARC definite definition were conducted and published prior to the definitions creation) [6]. Most studies included cases of angiographically-confirmed ST regardless of their timing (early, late or very late) and thus had a mixture of such cases; however, 10 studies reported in-hospital or 30-day data for those experiencing early ST, 6 studies reported data for those experiencing late ST, and 4 for very late ST. Three studies stated they were funded by an industry partner, two and one study, respectively, were funded by an academic institution and a foundation, two studies reported being unfunded, and the remainder did not provide information of funding.
Study validity assessment as it pertains to answering this review's aims are depicted in Tables 3 and 4. The most common validity concerns noted for included studies were: 1) not planning to follow post-ST outcomes as a primary objective of the study (studies were designed for a different purpose), 2) failure to include consecutive and unselected populations, 3) not collecting data in a prospective fashion, 4) not independently adjudicating ST and MIs, and 5) not adequately describing funding or the role of the funder in the research.
Only one study reported costs associated with angiographically-confirmed ST, estimating the median/mean± standard deviation per patient cost of hospitalization to treat early ST at $11,134/$17,400±17,758 (in 2000US$). The majority of these costs were associated with the cardiac catheterization laboratory ($5,496±2,833) followed by time spent in the intensive care unit (mean length-of-stay=1.8±2.3 days, associated costs= $3,692 ± 5,229).

Discussion
To our knowledge, this is the first meta-analysis with the aim of better quantifying the real-world incidence of in-hospital or 30-day death or MI, as well as, angiographically-confirmed STrelated treatment costs to be conducted. Our meta-analyses suggested angiographically-confirmed ST is associated with significant consequences; with nearly 8 out of every 100 patients experiencing an ST dying and 6 of 100 suffering a new or recurrent MI (not counting the ST-defining event) in-hospital. The rate of these same major adverse cardiovascular events increased to greater than 11% and 9%, respectively by 30days. While we identified only a single study [4] describing costs associated with treating angiographically-confirmed ST, available data suggested treating ST may be associated with a significant economic burden to the healthcare system. There are multiple definitions of ST that are utilized in studies evaluating the incidence or outcomes of ST [6]. The "definite" definition requires angiographic or autopsy-confirmation, and therefore maximizes specificity (minimizes false positive diagnoses), but at the cost of underestimating the incidence (or missing cases) of ST. The supplementary use of more clinical definitions of ST (i.e., "probable" and "possible") can add sensitivity (misses fewer cases of ST); however, their utility is highly dependent on the quality of data available to adjudicate ST events [6]. For a number of reasons, our meta-analysis only included studies reporting outcomes associated with "definite" or angiographically-confirmed ST. Since the large majority of studies did not report clinical ST along with definite/ angiographically-confirmed ST, we wanted to avoid pooling studies using disparate definitions to minimize statistical heterogeneity in our analyses. Moreover, since our metaanalysis was focusing on real-world observational studies, many of which were retrospective and failed to independently adjudicate outcomes, we had concerns about a higher risk of misclassification bias in these studies. It must be stressed, however, our restriction to angiographically-confirmed ST (to improve internal validity) means that patients not surviving to angiography were not commonly evaluated in included studies, and our analysis likely underestimates the true negative consequences of ST.
To our knowledge, the only attempt to systematically assess outcomes following ST is the 2001 paper pooling 6 multicentered BMS randomized trials by Cutlip and colleagues [3]. That meta-analysis followed a total of 45 patients diagnosed with angiographic ST (0.7% of the pooled population), and reported a lower incidence of 30-day death compared to our meta-analysis (6.7% vs. 11.6%). One explanation for this discrepancy between Cutlip's and our meta-analysis may be their use of randomized trial patient data which is likely not generalizable to the more expansive and sicker populations commonly treated with stents in real-world settings. Cutlip's use of only BMS trials cannot be ruled out as an explanation for differences in 30-day morality estimates, since recent analyses suggest DES use in the modern era may result in better outcomes then BMS use [29,30]. In contrast to the 30-day death outcome, our meta-analysis found a lower incidence of 30-day MI (9.5%) then the aforementioned analysis (15.6% Qwave, 44.4% non-Q-wave). This finding may be a result of detection bias (patients were more aptly followed for MI in the randomized trial setting than in observational studies) or it may be a result of differences in what timing of MI was counted. More specifically, ST studies appear to vary in how they include post-ST MI, with some studies including the initial STdefining ischemic event and some excluding it; counting only new or reinfarction in their in-hospital and 30-day MI outcomes. Because most observational studies of ST include only patients who present with acute coronary syndrome, counting the STdefining ischemic event results in much higher estimates of MI incidence, often as high as 85%-95% by 30-days [31]. Of note, while most of the studies we identified in our systematic review exclude the ST-defining ischemic event from their MI endpoint, not all did. Thus, we had to exclude a small number of studies [23,31] from our MI meta-analysis because they either stated     they included the ST-defining MI or they did not make it clear they did not (albeit efforts to contact authors for clarification were made, with some success). Upon subgroup analysis, our meta-analysis identified a statistical interaction between the incidences of in-hospital and 30-day death and the timing of angiographically-confirmed ST. In both analyses, patients experiencing early ST (within 30-   days) appeared to have higher incidences of death than those experiencing very late ST (>1 year). It has been posited that these differences in fatality rates could be explained, at least partially, by ST timing-dependent differences in patient characteristics or variation in the pathophysiologic mechanisms of early and later occurring ST [7,18,32]. Our finding of no difference in early mortality among patients experiencing a thrombosed bare-metal or DES is consistent with the findings of prior studies [9,20,25]. Of note, we did not try to run subgroup analyses on the in-hospital and 30-day MI outcomes due to the small number of studies identified for these endpoints (N=5 and 3). Hopefully future larger studies or metaanalyses can address whether interactions exists between ST timing or stent type and post-ST MI outcomes.
The Agency for Healthcare Research and Quality's (AHRQ's) Healthcare Utilization Project (HCUP) estimated approximately 640,000 Americans were hospitalized to undergo coronary stent implantation in 2009 [1]. Therefore, despite the relatively low incidence of ST, the absolute number of patients experiencing ST is still quite large and likely to continue to grow as the use of coronary stents increase. If the median cost of treating a ST calculated by Reynolds and colleagues is inflated to a 2012US$ value ($17,686), and we assume a conservative ST rate of 1% [2], estimates of direct hospital costs alone of treating ST likely exceed $113 million per year. It is important to note, the study by Reynolds and colleagues [4] was a small study (N=23 subacute STs) including only patients receiving BMSs. Moreover, it did not estimate costs of ST due to delayed complications, outpatient healthcare utilization, and lost work There are some limitations to our meta-analysis that merit further discussion. First, despite our best efforts to reduce heterogeneity between pooled studies through inclusion/ exclusion criteria, high degrees of statistical heterogeneity in our base-case analyses were still present (I 2 >55% for all). Subgroup analysis suggested that some of this heterogeneity was explained by pooling studies including different ST-timing, but not the pooling of studies including different stent types. Due to the relative paucity of identified studies for most endpoints, we were not able to investigate other sources of heterogeneity on our results. A second limitation of our metaanalysis stems from the fact that the evaluation of post-ST outcomes was not the primary objective of many of the included studies. This likely explained some of the studies' validity deficiencies noted, such as the failure to independently adjudicate ST and major adverse cardiovascular event outcomes. Next, since the majority of included studies in this meta-analyses evaluated ACS patients; our pooled incidence rates are likely most generalizable to this population. Finally, due to inconsistent or incomplete reporting of the time from angiographically-confirmed ST to the occurrence of death and/or MI, a number of studies had to be excluded from our meta-analysis. This latter limitation may in part explain the higher likelihood of publication bias noted some of our analyses.

Conclusions
Regardless of stent type used, the short-term consequences of coronary stent thrombosis (ST) appear significant. While stent type does not seem to affect the incidence of post-ST outcomes, an earlier occurrence of ST may be associated with higher mortality.