Author + information
- Received April 16, 2009
- Revision received July 17, 2009
- Accepted September 4, 2009
- Published online November 1, 2009.
- Jason B. Lindsey, MD⁎,
- Steven P. Marso, MD⁎,
- Michael Pencina, PhD†,
- Joshua M. Stolker, MD⁎,
- Kevin F. Kennedy, MS⁎,
- Charanjit Rihal, MD‡,
- Greg Barsness, MD‡,
- Robert N. Piana, MD§,
- Steven L. Goldberg, MD∥,
- Donald E. Cutlip, MD¶,
- Neal S. Kleiman, MD#,
- David J. Cohen, MD, MSc⁎,⁎ (, )
- EVENT Registry Investigators
- ↵⁎Reprint requests and correspondence:
Dr. David J. Cohen, Saint Luke's Mid America Heart Institute, 4401 Wornall Road, Kansas City, Missouri 64111
Objectives Our aim was to examine the prognostic importance of hemorrhagic and ischemic complications after percutaneous coronary intervention (PCI) in unselected patients.
Background In randomized trials of PCI, major bleeding and periprocedural myocardial infarction (pMI) have been associated with increased mortality. Whether similar associations exist among un-selected PCI patients is unknown.
Methods We used data from the EVENT (Evaluation of Drug Eluting Stents and Ischemic Events) registry—a multicenter registry of unselected patients undergoing PCI—to examine the association between both in-hospital bleeding and pMI and 1-year mortality. Cardiac enzyme levels were assessed in all patients, and pMI was defined as a peak creatine kinase-MB value ≥3× the upper limit of normal. Post-PCI bleeding was classified by Thrombolysis In Myocardial Infarction criteria.
Results After excluding patients with elevated pre-PCI creatine kinase-MB values and ST-segment elevation myocardial infarction at presentation (n = 1,626), a total of 5,961 patients were available for evaluation. Rates of post-PCI bleeding and pMI were 3.0% and 7.1%, respectively; 1-year all-cause mortality was 2.8%. After multivariable adjustment, both post-PCI bleeding (adjusted hazard ratio [HR]: 3.83, 95% confidence interval: 2.48 to 5.90, p < 0.001) and pMI (adjusted HR: 1.84, 95% confidence interval: 1.17 to 2.89, p = 0.009) were independently associated with 1-year mortality. Time period-specific analyses demonstrated that the adjusted HR for bleeding was similar for 30-day mortality and mortality between 1 month and 1 year, while the adjusted HR for pMI was greater for 30-day mortality as compared with mortality between 1 month and 1 year.
Conclusions Among unselected PCI patients, both post-PCI bleeding and pMI are independently associated with increased 1-year mortality. Continued efforts to reduce these complications after PCI are warranted.
Both periprocedural myocardial infarction (pMI) and bleeding after percutaneous coronary intervention (PCI) have been associated with increased short- and long-term mortality (1–11). However, most studies evaluating the prognostic significance of these events were based on randomized trial populations (1,4–8,10) or were published before the current drug-eluting stent (DES) era (2,9,11–13). Therefore, a paucity of data exists regarding the prognostic significance of post-PCI bleeding and pMI in unselected patients undergoing PCI in the current era of antithrombotic and antiplatelet therapies. Although administrative data could potentially be used to address these issues, these data often suffer from incomplete or inconsistent ascertainment of clinically relevant but “nonbillable” complications (14). In particular, assessment of pMI is poor in most administrative datasets and even some clinical registries (15).
In addition, most studies evaluating post-PCI bleeding have only reported “major” bleeding (with a variety of definitions) as an independent predictor of mortality (6,7,9), while the impact of less severe bleeding events is less certain (5). Finally, until recently few studies have attempted to examine the independent effects of bleeding and pMI on post-PCI mortality in the same population. Consequently, the relative prognostic significance of these 2 complications in the post-PCI population is unknown. In this study, we used data from the EVENT (Evaluation of Drug-Eluting Stents and Ischemic Events) registry to identify the frequency and severity of both bleeding and ischemic complications of PCI and to examine the relationship between these complications and 1-year mortality among unselected patients.
The methods and population of the EVENT registry have been described previously (16). Briefly, the EVENT registry was a collaborative effort to assess the contemporary practice of stenting by performing a prospective evaluation of un-selected patients undergoing attempted implantation of an approved intracoronary stent at more than 50 PCI centers in the U.S. Although enrollment in the registry was limited to pre-specified recruitment “waves,” specific efforts were made to enroll patients consecutively during each enrollment period (e.g., on pre-determined days of the week) to minimize selection bias.
The patient population for our study was based on all patients who underwent PCI in the first 3 waves of the EVENT registry (n = 7,587), which were enrolled between July 2004 and January 2006. Patients were excluded if the primary indication for PCI was treatment of ST-segment elevation myocardial infarction (STEMI) or if baseline creatine kinase (CK)-MB levels were elevated (because of difficulty in ascertaining pMI in such individuals) (Fig. 1). In addition, we excluded those subjects who were missing ≥1 candidate variable for the proposed multivariable models.
Data collection and end points
Data regarding patient characteristics, presentation, treatment, and outcomes were collected prospectively on standardized case report forms and submitted to the data coordinating center. CK and CK-MB levels were assessed at baseline (within 1 h before the procedure) and every 8 h for a minimum of 2 samples after the procedure and assayed using each site's clinical laboratory and reference values. If a myocardial infarction (MI) was clinically suspected at a later point, additional biomarkers were obtained as indicated.
Patients were contacted by telephone at 6 and 12 months after the index PCI. Events ascertained at follow-up included death, stent thrombosis, revascularization, and MI. All clinical outcomes were adjudicated by 2 cardiologists blinded to treatment factors and subsequent outcomes. The study protocol was approved by ethical review committees at all participating institutions, and all patients provided written informed consent before participation.
The primary end point of the present analysis was 1-year all-cause mortality. pMI was defined as elevation of CK-MB ≥3× the upper limit of normal (ULN) (as determined by the local reference laboratory). Bleeding was defined using the Thrombolysis In Myocardial Infarction (TIMI) criteria (17). A bleeding event was defined as major if it was intracranial or if clinically overt signs of hemorrhage were associated with a drop in hemoglobin of >5.0 g/dl. Minor bleeding was defined as clinically overt hemorrhage (including that visualized on imaging) associated with a fall in hemoglobin concentration of 3.0 to 5.0 g/dl.
Continuous variables are described as mean ± SD and were compared using unpaired t tests. Categorical variables are described as counts and percentages and were compared using the chi-square test or the Fisher exact test. One-year mortality was estimated by the Kaplan-Meier method, and univariate associations were assessed by the log-rank test for both post-PCI bleeding and pMI.
To examine the independent associations between post-PCI bleeding, pMI, and 1-year mortality, multivariable-adjusted hazard ratios (HRs) and their 95% confidence intervals (CIs) were calculated using the Cox proportional hazards model. A sequentially saturated model was used with adjustments for demographic variables (age, sex, body mass index); clinical variables (diabetes mellitus, hypertension, hypercholesterolemia, congestive heart failure, peripheral arterial disease, estimated glomerular filtration rate, prior MI, prior coronary artery bypass grafting, acute coronary syndrome [ACS] at presentation); and angiographic and procedural variables (multivessel PCI, PCI of saphenous vein graft, PCI of proximal left anterior descending coronary artery) in a hierarchical fashion. To reduce the likelihood of model overfitting, we restricted the number of covariates to ∼1 per 10 outcome events (18).
To further account for differences in baseline characteristics between subjects with and without post-PCI bleeding or pMI, we performed secondary analyses in which we used propensity score matching of patients with post-PCI bleeding and pMI where 1 patient with an event (post-PCI bleeding or pMI) was matched with up to 4 without an event. The propensity scores for these analyses were calculated as predicted probabilities of each complication (post-PCI bleeding and pMI) based on logistic regression models that included the covariates listed in the preceding text. Matching required the propensity scores for patients with and without events to be within 0.01 (i.e., 1%). For the propensity-matched analyses, we used a conditional proportional hazard model, stratified by quintile of the propensity score.
Finally, to evaluate the prognostic impact of post-PCI bleeding and MI on short- (30 days) and longer- (between 30 days and 1-year) term mortality, we performed time period-specific analyses using 2 time-dependent exposure variables with a change point of 30 days. We also tested for the interaction between post-PCI bleeding and pMI with the change point time variable to determine if there was a difference in their prognostic associations with short versus longer-term mortality.
Probability values of <0.05 were considered to indicate statistical significance. All statistical analyses were performed using SAS version 9.1 (SAS Institute, Cary, North Carolina).
Patient population and in-hospital outcomes
Baseline characteristics of the study population and index revascularization procedures are described in Table 1. The patient population was broadly representative of the general U.S. population undergoing nonemergent PCI. The mean age was 65 ± 11 years, 68% were men, 35% had diabetes, and 23% were current smokers. The indication for PCI was an ACS (without STEMI) in 34% and stable coronary artery disease in 60%. DES were used in 92%, and use of either glycoprotein IIb/IIIa inhibitors or direct thrombin inhibitors (mainly bivalirudin) was common. Among the 5,961 patients, the overall post-PCI bleeding rate was 3.0% (0.7% TIMI major and 2.3% TIMI minor), and the overall pMI rate was 7.1% (3.5% with CK-MB values 3 to 5× ULN, 2.1% with CK-MB values >5 to 10× ULN, and 1.5% with CK-MB values >10× ULN). The 1-year all-cause mortality rate was 2.8% (1.6% cardiac).
Patients with post-PCI bleeding were older, more likely to be women, more likely to have diabetes, hypertension, heart failure, peripheral arterial disease, to have renal insufficiency (or be receiving chronic hemodialysis), more likely to have undergone previous coronary artery bypass grafting, and more likely to have ACS at presentation. Among angiographic and procedural characteristics, those with post-PCI bleeding were more likely to have undergone PCI of the left main coronary artery or a saphenous vein graft, multivessel PCI, and PCI of type C lesions. Use of heparin alone and heparin + glycoprotein IIb/IIIa antagonists was slightly higher among those with post-PCI bleeding, and use of direct thrombin inhibitors was lower (Table 1).
Patients with pMI were less likely to have hypertension or diabetes, and more likely to have renal insufficiency and ACS at presentation. There were no differences among the groups with regard to age or sex. Among angiographic characteristics, those with pMI were more likely to have undergone PCI of the left anterior descending coronary artery or a saphenous vein graft, PCI of a type C lesion, multivessel PCI, and treatment of a bifurcation lesion. With regard to procedural characteristics, individuals with a pMI had a greater number of stents implanted and longer total stent lengths (Table 2).
Association between bleeding and mortality
Compared with those without post-PCI bleeding, patients with any post-PCI bleeding (combined TIMI major and minor) had increased 1-year mortality (15.6% vs. 2.4%, p < 0.001) (Fig. 2A). When post-PCI bleeding events were stratified according to the TIMI criteria, individuals who experienced TIMI major bleeding had the highest 1-year mortality (21.9%), while patients with TIMI minor bleeding had intermediate 1-year mortality (13.7%) (Fig. 2B). There were no significant differences in baseline demographic, clinical, or angiographic characteristics among those with TIMI minor bleeding versus TIMI major bleeding (data not shown).
Before risk adjustment, post-PCI bleeding of any severity was associated with increased risk for 1-year mortality (HR: 6.83, 95% CI: 4.55 to 10.24; p < 0.001). After multivariable adjustment, the impact of bleeding was attenuated substantially; nonetheless, post-PCI bleeding of any severity remained independently associated with mortality (adjusted HR: 3.83, 95% CI: 2.48 to 5.90; p < 0.001) (Fig. 3). The association remained significant in propensity-matched analyses as well (HR: 3.60, 95% CI: 2.16 to 6.00; p < 0.001). When bleeding severity was stratified by TIMI criteria, both TIMI major bleeding (adjusted HR: 6.85, 95% CI: 3.37 to 13.89; p < 0.001) and TIMI minor bleeding (adjusted HR: 3.12, 95% CI: 1.87 to 5.21; p < 0.001) remained independently associated with increased 1-year mortality.
Association between pMI and mortality
Patients who experienced a pMI had 1-year mortality that was approximately twice that of patients without pMI (5.2% vs. 2.7%; p = 0.002) (Fig. 4). Before risk adjustment, pMI was associated with an increased risk for 1-year mortality (HR: 2.00, 95% CI: 1.27 to 3.14; p < 0.001). After multivariable adjustment, the HR remained largely unchanged, and pMI remained independently associated with increased mortality (adjusted HR: 1.84, 95% CI: 1.17 to 2.89; p = 0.009) (Fig. 5). Similar findings were noted in propensity-matched analyses as well (adjusted HR: 1.82, 95% CI: 1.09 to 3.06; p = 0.02).
Short- versus longer-term associations between complications and mortality
Time period-specific analyses demonstrated that the association between post-PCI bleeding and mortality was qualitatively similar for short-term (<30-day) (adjusted HR: 3.92, 95% CI: 1.46 to 10.52; p < 0.001) and longer-term (≥30-day to 1-year) mortality (adjusted HR: 3.79, 95% CI: 2.35 to 6.10; p < 0.001). In contrast, the association between pMI and mortality was significant and stronger for short-term (<30-day) mortality (adjusted HR: 5.89, 95% CI: 2.63 to 13.12; p < 0.001) but was not significant for longer-term mortality (≥30-day to 1-year) (adjusted HR: 1.25, 95% CI: 0.70 to 2.21; p = 0.46). There was a significant interaction between time and pMI indicating a difference in the prognostic impact of these events on short- versus longer-term mortality (p = 0.003).
In this study, we used data from 5,961 unselected PCI patients treated between 2004 and 2006 to assess the frequency of both pMI and bleeding complications in contemporary practice and the association between these events and subsequent mortality. Despite numerous advances in PCI technique, devices, and adjunct pharmacology over the past 2 decades (19), we found that with careful surveillance (including routine monitoring of post-procedure cardiac enzymes) both of these events remain relatively common. pMI was observed in 7.1% of patients, while bleeding complications were noted in 3.0%. The rate of pMI in our unselected population is virtually identical to the 8% rate recently reported among patients who underwent routine assessment of post-procedure cardiac enzymes in the ACC–NCDR (American College of Cardiology–National Cardiovascular Data Registry) (15) and serves to highlight the importance of such surveillance for both monitoring and improving the quality of PCI procedures.
In addition to their frequency, we found that both bleeding (of any severity) and pMI were independently associated with 1-year, all-cause mortality among our “real-world” PCI population. In time-dependent analyses, however, the patterns of association differed somewhat between these 2 complications. For post-PCI bleeding, the adjusted HR for 30-day mortality was qualitatively similar to that for mortality between 1 month and 1 year. In contrast, for pMI, the association with mortality was strongest during the first 30 days and not significant for events beyond this time point.
Several previous studies have demonstrated an independent association between bleeding and subsequent mortality after PCI (5–7,11). For example, Ndrepepa et al. (5) reported pooled data from the ISAR (Intracoronary Stenting and Antithrombotic Regimen) trials that demonstrated that bleeding of any severity after PCI was independently associated with 1-year mortality. Of note, the rates of bleeding after PCI (4%) were similar to those reported in the current study (3%) as was the point estimate for the relative risk of mortality. In addition, results from the ACUITY (Acute Catheterization and Urgent Intervention Triage Strategy) trial (6), which demonstrated that major bleeding was associated with increased mortality, also support our findings. Similar findings have been noted in ACS patients as well, independent of the performance of PCI (6,8,10). Despite the consistency of these findings, however, important questions remain regarding their generalizability as well as the causal nature of the association (20).
Our study adds to the available data by extending these previous observations—which were derived largely from clinical trial populations—to a large, unselected population undergoing PCI using contemporary devices and pharmacologic regimens including DES in ∼90% and pre-PCI clopidogrel loading in ∼40% of patients. Moreover, our study is one of the first to demonstrate a graded relationship between the severity of bleeding and long-term mortality in a PCI population—findings that were first described by Rao et al. (10) in an analysis of clinical trial participants with ACS. Although other explanations remain possible (e.g., unmeasured confounding), the finding of a “dose-response curve” in the bleeding-mortality association provides further evidence of a potential causal relationship (21).
The prognostic importance of pMI after PCI has been reported in prior studies as well (1–5,12,13,22). Many of these studies were summarized in a recent meta-analysis in which Ioannidis et al. (3) found that any increase in CK-MB after PCI, especially a >3-fold CK-MB elevation, was associated with a significant increase in the subsequent risk of death. Similar to the literature on post-PCI bleeding, however, the vast majority of these studies pre-date the current era of DES and prolonged dual antiplatelet era. Thus, our study serves to re-emphasize the importance of ischemic complications as a key prognostic factor after PCI in contemporary practice.
In addition to demonstrating its importance as a prognostic factor after PCI, our study provides several novel insights into the relative magnitude and timing of the prognostic impact of bleeding and ischemic complications after PCI. In particular, we found that as a predictive factor, the association between bleeding (with an overall incidence of ∼3% in our population) and 1-year mortality was about twice as strong as for pMI (with an incidence of ∼6%). These findings suggest that the population-attributable risks are similar for these 2 complications and that a similarly effective effort to further reduce these complications is likely to yield comparable benefits in terms of 1-year mortality.
These extrapolations are based on the assumption that the observed associations with mortality are of a causal nature—an assumption that remains controversial. For pMI, our observation that the association is strongest in the first 30 days after PCI and not significant after 30 days is reasonably consistent with a causal chain. For bleeding, however, the fact that the association is similar in both the first month and in the subsequent 11 months may reflect either unmeasured confounding or a mechanism of mortality that relates to latent effects of bleeding on inflammation or the immune response (20,23).
The results of the current study provide further evidence that both bleeding and pMI after PCI continue to be common, and future efforts should be aimed at a reduction of these events. This is a difficult proposition because use of more potent antiplatelet and antithrombotic drugs tends to reduce ischemic complications after PCI but typically results in increased bleeding rates (24–27). Recent data do suggest that some antithrombotic drugs may be preferable to others because of reduced major bleeding rates after PCI and that, in certain settings, it may even be possible to reduce bleeding without increasing ischemic events (28,29). In addition, the use of radial artery access, as opposed to femoral artery access, appears to reduce the rate of access site bleeding after PCI, suggesting a further therapeutic option for improving both short- and long-term outcomes in PCI patients (30,31). Other strategies to potentially reduce post-PCI bleeding events include use of smaller vascular access sheaths (31), fluoroscopy-guided arterial puncture (32), judicious use of vascular access closure devices (33), and avoidance of excessively high activated clotting times during PCI (31).
First, despite our best efforts to adjust for potential confounders by using multivariable models and propensity matching, given the observational nature of the present study, it is likely that unmeasured confounders could have contributed to the observed differences in mortality among subjects with post-PCI bleeding and pMI. Second, it is conceivable that the rates of pMI and post-PCI bleeding were under-reported. However, in the case of pMI, the fact that CK-MB levels were assessed both before and after PCI in 97% of study subjects makes this possibility relatively unlikely. However, the EVENT registry did not mandate post-PCI assessment of hemoglobin in all patients. Thus, it is possible that some clinically silent bleeding events went undetected. Moreover, post-PCI bleeding was only tracked during the index hospitalization.
Third, our assessment of mortality was restricted to 1-year events. Thus, the prognostic impact of periprocedural complications on longer-term mortality cannot be assessed. Other investigators have reported that the prognostic importance of pMI extends beyond 1 year (1,2,13), but to our knowledge no long-term data (>1 year) have been reported for post-PCI bleeding and mortality. Fourth, we excluded STEMI patients from the current analysis; consequently, our results apply only to patients undergoing either nonemergent PCI in either the elective or unstable angina/NSTEMI setting. Of note, however, recent unpublished data from the HORIZONS-AMI (Harmonizing Outcomes with Revascularization and Stents in Acute Myocardial Infarction) study have shown that major bleeding after primary PCI is strongly associated with an increased risk for mortality at 30 days (34). Finally, there was heterogeneous use of antithrombin therapy in the EVENT registry; however, adjustment for the type of antithrombin therapy used did not appreciably alter the point estimates for mortality. Furthermore, the observed differences in utilization rates of antithrombin therapies are likely due to a variety of unmeasured factors; consequently, no conclusions should be drawn regarding the superiority of 1 antithrombin treatment regimen with regard to post-PCI bleeding or pMI.
In this large, unselected population of patients undergoing PCI with DES and contemporary antithrombotic therapy, both pMI and bleeding remain frequent complications that are independently associated with increased long-term mortality. Although the causal nature of these associations remains speculative, these findings suggest that continued efforts to reduce these complications are warranted.
Dr. Marso has received research grant support from The Medicines Company. Dr. Cohen has received research grant support from The Medicines Company, Schering-Plough, Eli Lilly, and Daichi Sankyo, and has been a consultant to Schering-Plough and Eli Lilly.
- Abbreviations and Acronyms
- acute coronary syndrome
- confidence interval
- creatine kinase
- drug-eluting stent(s)
- hazard ratio
- myocardial infarction
- percutaneous coronary intervention
- periprocedural myocardial infarction
- ST-segment elevation myocardial infarction
- upper limit of normal
- Received April 16, 2009.
- Revision received July 17, 2009.
- Accepted September 4, 2009.
- American College of Cardiology Foundation
- Topol E.J.,
- Ferguson J.J.,
- Weisman H.F.,
- et al.
- Ioannidis J.P.,
- Karvouni E.,
- Katritsis D.G.
- Akkerhuis K.M.,
- Alexander J.H.,
- Tardiff B.E.,
- et al.
- Ndrepepa G.,
- Berger P.B.,
- Mehilli J.,
- et al.
- Manoukian S.V.,
- Feit F.,
- Mehran R.,
- et al.
- Eikelboom J.W.,
- Mehta S.R.,
- Anand S.S.,
- Xie C.,
- Fox K.A.,
- Yusuf S.
- Abdelmeguid A.E.,
- Topol E.J.,
- Whitlow P.L.,
- Sapp S.K.,
- Ellis S.G.
- Wang T.Y.,
- Peterson E.D.,
- Dai D.,
- et al.
- Venkitachalam L.K.K.,
- Selzer F.,
- Willensky R.L.,
- et al.
- Rao S.V.,
- Eikelboom J.A.,
- Granger C.B.,
- Harrington R.A.,
- Califf R.M.,
- Bassand J.P.
- Bhatt D.L.,
- Topol E.J.
- Doyle B.
- (2002) Collaborative meta-analysis of randomised trials of antiplatelet therapy for prevention of death, myocardial infarction, and stroke in high risk patients. BMJ 324:71–86.
- Rao S.,
- Ou F.,
- Wang T.,
- et al.
- Doyle B.J.,
- Ting H.H.,
- Bell M.R.,
- et al.
- Dauerman H.L.,
- Applegate R.J.,
- Cohen D.J.