Author + information
- Received February 17, 2005
- Revision received August 26, 2005
- Accepted September 9, 2005
- Published online February 21, 2006.
- Sunil V. Rao, MD⁎,⁎ (, )
- Kristi O’Grady, MS⁎,
- Karen S. Pieper, MS⁎,
- Christopher B. Granger, MD, FACC⁎,
- L. Kristin Newby, MD, MHS, FACC⁎,
- Kenneth W. Mahaffey, MD, FACC⁎,
- David J. Moliterno, MD, FACC†,
- A. Michael Lincoff, MD, FACC‡,
- Paul W. Armstrong, MD, FACC§,
- Frans Van de Werf, MD, FACC∥,
- Robert M. Califf, MD, FACC⁎ and
- Robert A. Harrington, MD, FACC⁎
- ↵⁎Reprint requests and correspondence:
Dr. Sunil V. Rao, The Duke Clinical Research Institute, P.O. Box 17969, Durham, North Carolina 27715.
Objectives The goal of this study was to determine the association between Thrombolysis In Myocardial Infarction (TIMI) and Global Use of Strategies to Open Occluded Coronary Arteries (GUSTO) bleeding and clinical outcomes.
Background There are limited data on the relative utility of either scale at predicting clinical outcomes in patients with non–ST-segment elevation acute coronary syndromes (ACS).
Methods Pooled data from two randomized trials of patients with ACS (n = 15,454) were analyzed to determine the association between TIMI and GUSTO bleeding and 30-day and 6-month death/myocardial infarction (MI) using Cox proportional hazards modeling that included bleeding as a time-dependent covariate.
Results There was a stepwise increase in the adjusted hazard of 30-day death/MI with worsening GUSTO bleeding (hazard ratio [95% confidence interval], GUSTO mild 1.20 [1.05 to 1.37]; moderate 3.28 [2.88 to 3.73]; severe 5.57 [4.33 to 7.17]), and an increased risk with all three levels of TIMI bleeding (TIMI minimal 1.84 [1.63 to 2.08]; TIMI minor 1.64 [1.31 to 2.04]; major 1.45 [1.23 to 1.70]). When both bleeding scales were included in the same model, the risk with GUSTO bleeding persisted; however, the association between TIMI bleeding and outcome was no longer significant.
Conclusions Both scales identify ACS patients with bleeding complications at risk for adverse outcomes. In a model that included both definitions, the risk with GUSTO bleeding persisted while the risk with TIMI bleeding did not. This suggests that bleeding assessed with clinical criteria is more important than that assessed by laboratory criteria in terms of outcomes. Future clinical trials should consider using a combination of the GUSTO bleeding scale and the need for transfusion to assess bleeding complications.
Contemporary therapy for the treatment of acute ischemic heart disease includes anti-thrombotic medications and the use of early invasive risk stratification (1). While this combination of treatments has improved the outcomes of patients with acute coronary syndromes (ACS), the risk of bleeding remains a significant issue (2,3). Cognizant of this fact, clinical investigators systematically identify bleeding events that occur during the course of randomized clinical trials. Although guidelines exist for classifying bleeding events (4), many clinical trials utilize one of two bleeding classifications: the Thrombolysis In Myocardial Infarction (TIMI) classification, and/or the Global Use of Strategies to Open Occluded Coronary Arteries (GUSTO) classification. The existence of these two systems often leads to disparities in the reported incidence of bleeding events within the same clinical trial. For example, in the Platelet Glycoprotein IIb/IIIa in Unstable Angina: Receptor Suppression Using Integrilin Therapy (PURSUIT) trial, the rate of TIMI major bleeding among patients assigned to the platelet inhibitor eptifibatide was 3.0%, yet the rate of GUSTO severe bleeding in this same group was 1.1%. Moreover, the proportion of patients receiving eptifibatide deemed to have no TIMI bleeding was 84.2%, but the proportion deemed to have no GUSTO bleeding was 68.8%. There are limited data on the relative merits of either scale at predicting clinical outcomes. We analyzed data from two large multicenter randomized clinical trials to compare the association between the different degrees of in-hospital TIMI and GUSTO bleeding and short- and intermediate-term clinical outcomes.
Patient population and treatments
Clinical data from the multicenter international PURSUIT trial (n = 10,798) and the Platelet IIb/IIIa Antagonism for the Reduction of Acute Coronary Syndrome Events in a Global Organization Network (PARAGON) B trial (n = 5,225) were pooled. The details of the trials have been published elsewhere (5,6). Briefly, patients with acute non–ST-segment elevation ACS enrolled in the PURSUIT trial were randomly assigned to receive the intravenous glycoprotein IIb/IIIa inhibitor eptifibatide or placebo. Patients with non–ST-segment elevation ACS enrolled in the PARAGON B trial were randomly assigned to receive the intravenous glycoprotein IIb/IIIa inhibitor lamifiban or placebo. For the current study, the analysis was limited to the patients from the two trials that had complete data on end points.
Concomitant treatment with aspirin (dose ranges between 80 and 325 mg daily) and antithrombin agents was recommended by protocol in both trials. The use of other medications and procedures was at the discretion of the treating physicians.
Definitions and end points
The TIMI bleeding classification is a laboratory-based scale (7) while the GUSTO bleeding classification is a clinically based scale (8) (Table 1).The TIMI definition of bleeding uses four categories: major, minor, minimal, and none. The GUSTO bleeding definition also uses four categories: severe or life-threatening, moderate, mild, and none. The PURSUIT investigators used both definitions to classify bleeding events. The PARAGON investigators defined bleeding complications as major or life-threatening, and intermediate. Major or life-threatening bleeding was defined as any intracranial hemorrhage or bleeding leading to hemodynamic compromise requiring intervention. Intermediate bleeding was defined as bleeding requiring transfusion or a decrease in hemoglobin 5 g/dl or more (or decrease in hematocrit ≥15% when hemoglobin was unavailable). For the purpose of this analysis, the TIMI and GUSTO classifications were reconstructed from the detailed clinical data (Appendix).
Data on the date, time, severity, and location (including “unidentifiable”) of each in-hospital bleeding event were collected prospectively. Detailed clinical data, including treatments (such as blood transfusion), baseline, and nadir hemoglobin or hematocrit values after each bleeding event, and hemodynamic status during each bleeding event, were collected on all patients in both trials. For patients who experienced more than one bleeding episode, only the most severe bleeding episode was considered.
The primary end point of our study was the occurrence of death or recurrent myocardial infarction (MI) at 30 days. The secondary end point was the occurrence of death or recurrent MI at six months. Myocardial infarction was defined according to the protocol of each trial (5,6). All death and MI events for each trial were adjudicated by an independent blinded events committee.
Patients were grouped according to the presence or absence of a bleeding event. Patients who experienced a bleeding event were further classified based on bleeding severity according to the TIMI and GUSTO scales. Categorical variables are expressed as percentages, and continuous variables are expressed as medians and interquartile ranges. Baseline characteristics were compared using chi-square tests for categorical variables and the non-parametric Kruskal-Wallis test for continuous variables. Baseline differences with p values <0.01 were considered significant.
We compared unadjusted rates of the primary and secondary end points among patients with no bleeding and those within the various categories of TIMI and GUSTO bleed severity. To determine the association between bleeding severity as defined by the two scales and the primary and secondary outcomes, we constructed separate models for each bleeding definition. Because bleeding is a post-randomization event that can change over time, and can influence and be influenced by treatments (e.g., procedures) and adverse events (e.g., MI), both models incorporated bleeding as a time-dependent covariate in a Cox regression. This technique minimizes confounding by considering only those bleeding events that transpired before the occurrence of the end points (9). Variables entered into both models were based on a comprehensive set of baseline variables from a validated model of outcome among patients with non–ST-segment elevation ACS (10) (c-index = 0.81) and included patient characteristics, presenting signs and symptoms, and treatments, including blood transfusion.
The first model incorporated TIMI bleed severity as a time-dependent covariate and used “no TIMI bleeding” as the reference. The second model incorporated GUSTO bleed severity as a time-dependent covariate and used “no GUSTO bleeding” as the reference. Due to the potential influence of blood transfusion on outcomes (11), we repeated the analysis after adjusting for transfusion by creating a dichotomous time-dependent variable that coded transfusion of at least 1 U of blood. We also repeated the analysis by constructing a model that included both bleeding scales as time-dependent covariates. Additional models were generated to evaluate the effect of each bleed scale in the presence or absence of the other scale. The models were adjusted for baseline covariates and included the scale of interest (e.g., the GUSTO scale) plus an indicator for any bleed according to the other scale (e.g., any type of TIMI bleed) versus no bleed according to the other scale (e.g., no TIMI bleed of any sort). Another set of models included the interaction of these two variables. Because coronary artery bypass surgery (CABG) can influence the severity of bleeding as well as outcomes, we repeated the analysis again for the entire patient cohort with censoring at the time of CABG. Finally, we used interaction terms to explore the effect of patient age, gender, weight, and renal function on TIMI and GUSTO bleeding levels in predicting short- and intermediate-term death or MI. All analyses were performed using SAS Version 8.2 (SAS Institute, Cary, North Carolina).
Ethics of protocol
The institutional review boards of all participating institutions reviewed and approved the protocols of the PURSUIT and PARAGON B trials. All enrolled patients gave written informed consent.
Timing of events and baseline characteristics
A total of 15,454 patients from the two trials had complete data on end points. Among patients with a bleeding event, the median time from randomization to the most severe TIMI bleeding event was 1.0 days (25th, 75th percentile: 1.0, 2.0). The median time to the most severe GUSTO bleeding event was 2.0 days (1.0, 6.0). The median times to PCI and CABG procedures were 3.8 days (1.2, 9.8) and 9.9 days (4.1, 30.3), respectively. There were 1,151 patients who met criteria for GUSTO bleeding that did not meet criteria for TIMI bleeding, and 765 patients who met criteria for TIMI bleeding that did not meet criteria for GUSTO bleeding. There were 3,758 patients who met criteria for both TIMI and GUSTO bleeding.
With regard to TIMI bleeding, 12.7% of patients experienced a TIMI minimal bleed, 8.5% of patients experienced a TIMI minor bleed, and 8.2% experienced a TIMI major bleed. Table 1shows the baseline characteristics of the patients who developed TIMI bleeding by the severity of the bleeding event. There were significant differences in baseline characteristics across the TIMI bleeding categories so that patients with a TIMI minor bleed were older, more often of non-white race and female, more often had cardiac risk factors, and were sicker at presentation compared with those having either a TIMI minimal or major bleed. In addition, a higher proportion of patients with TIMI minor bleeding had undergone in-hospital PCI compared with those with TIMI minimal or major bleeding. In contrast, a higher proportion of patients who experienced a TIMI major bleed had prior hyperlipidemia, prior MI, prior congestive heart failure, and prior stroke. A higher proportion of patients with TIMI minimal bleeding had undergone cardiac catheterization and CABG compared with those who experienced TIMI minor or major bleeding. Approximately one-third of the patients with TIMI minimal, minor, and major bleeding underwent blood transfusion during the hospitalization. Significantly more patients with TIMI bleeding who underwent CABG were transfused compared with patients with TIMI bleeding who did not undergo CABG (18% vs. 4.6%, p < 0.001).
The proportion of patients with GUSTO mild, moderate, and severe bleeding was 19.2%, 11.4%, and 1.2%, respectively. Table 2shows the baseline characteristics of patients by worsening GUSTO bleed severity. As the GUSTO bleeding severity worsened, there was a gradient of increasing age, increasing proportion of patients of non-white race, and an increasing proportion of patients with hypertension, diabetes mellitus, prior MI, prior stroke, prior PCI, chronic renal insufficiency, and patients presenting with higher Killip class and higher heart rate. Patients who experienced a GUSTO moderate bleed had undergone cardiac catheterization and/or CABG during hospitalization more often than those experiencing mild or severe bleed, while patients who experienced a GUSTO mild bleed had undergone PCI more often than those who experienced a moderate or severe bleed. Very few patients with GUSTO mild bleeding underwent blood transfusion while a large proportion of patients with GUSTO moderate and severe bleeding received blood transfusion. Significantly more patients with GUSTO bleeding who underwent CABG received transfusion compared with patients with GUSTO who did not undergo CABG (53.5% vs. 10.0%, p < 0.001).
Tables 2 and 3⇓show the unadjusted rates of 30-day and 6-month death or MI for patients who experienced various levels of TIMI and GUSTO bleeding. After multivariable adjustment, there was a stepwise increase in the adjusted hazard of the primary and secondary end points as GUSTO bleeding severity worsened (Figs. 1Aand 1B). There was also an increased adjusted risk with each level of TIMI bleeding, but the point estimates of the hazard ratios for each level of TIMI bleeding were similar to each other. After adjustment for blood transfusion, the stepwise increase in the risk of short- and intermediate-term outcomes with worsening GUSTO bleeding persisted; however, the association between TIMI bleeding severity and six-month adverse outcomes was no longer statistically significant (Figs. 2Aand 2B). When both bleeding scales were included in the same model, the stepwise increase in the risk for 30-day and 6-month death or MI with worsening GUSTO bleeding was evident, but there was no significant increase in risk with any level of TIMI bleeding (Figs. 3Aand 3B). The analysis was repeated with censoring at the time of CABG, and the results were nearly identical (data not shown). There were no significant interactions between patient age, gender, weight, or renal function and TIMI bleeding. There was a significant interaction between age and GUSTO moderate bleeding such that the increased risk of death or MI with GUSTO moderate bleeding was more pronounced among younger patients. There was a significant interaction between GUSTO and TIMI bleeding (p < 0.0001). The risk for the 30-day and 6-month end points for GUSTO bleeding without TIMI bleeding was higher than the risk with GUSTO bleeding with any degree of TIMI bleeding.
Our study has several important findings. First, we have shown that both the GUSTO and TIMI bleeding scales identify patients at increased risk for adverse clinical events. This risk was manifest primarily within 30 days after the bleeding event. Second, we found that each scale identifies patients with bleeding events that are missed by the other scale. Third, after adjustment for blood transfusion, the stepwise increase in risk with worsening GUSTO bleeding persisted while the risk with TIMI bleeding was no longer statistically significant. Finally, when both bleeding definitions were included in the same model only GUSTO bleeding was associated with a worse prognosis. When exploring interaction terms between the two bleeding scales, we found that TIMI bleeding did not affect the risk among patients who also met criteria for GUSTO bleeding. This suggests that in terms of adverse outcomes, assessment of bleeding using clinical criteria is more important than using laboratory criteria.
Our study has important implications for both clinical care and clinical research. Both bleeding definitions were developed to classify bleeding complications of thrombolytic therapy (7,8), and have not been previously validated in the setting of non–ST-segment elevation ACS. Other investigators have explored the incidence and predictors of bleeding complications in this population (2,12,13), but the bleeding definitions used varied across studies, and outcomes were not specifically examined. Our results add to these previous studies by demonstrating that that bleeding complications, regardless of severity, are associated with worse clinical outcomes among patients with acute ischemic heart disease in whom thrombolytic therapy is not used. While this increased risk persisted for all levels of GUSTO bleeding even after adjusting for blood transfusion, it did not persist for TIMI bleeding after accounting for transfusion. Because one difference between the TIMI and GUSTO scales is whether transfusion is necessary, our results suggest that patients who experience TIMI bleeding who do not require transfusion (i.e., asymptomatic decreases in hemoglobin) are not at risk for adverse outcomes. Therefore, liberal use of blood transfusion in these patients may expose them to the risks of transfusion (11,14) without any clear benefit. These data also provide support for clinical practice guidelines that discourage the use of blood transfusions in patients who are clinically stable regardless of their nadir hemoglobin or hematocrit value (15).
Although our study is the first to provide evidence for the association between bleeding as measured by both the GUSTO and TIMI bleeding classifications and clinical outcomes in the setting of non–ST-segment elevation ACS, these definitions have been used either alone or in combination in cardiovascular clinical trials over the last two decades (5,16–24). This has resulted in wide variation in the rates of bleeding complications that have been reported with various antithrombotic agents. For example, in the Superior Yield of the New Strategy of Enoxaparin, Revascularization and Glycoprotein IIb/IIIa Inhibitors (SYNERGY) trial of enoxaparin versus unfractionated heparin, the rate of TIMI major bleeding was 9.7% among patients assigned to enoxaparin, but the rate of GUSTO severe bleeding was only 2.7% (24). In the context of this disparity, our study suggests that the guidelines for measuring bleeding among patients with ACS (4) should be amended to include a more optimal metric—a combination of the clinically-based GUSTO scale and need for blood transfusion. Bleeding complications with limited prognostic value as determined by the laboratory-based TIMI scale may continue to have a role in guiding dosing decisions for new anticoagulants that are in development.
Our study has some limitations. First, our study was not designed to evaluate the definitions of bleeding per se. Rather, we sought to delineate the association between bleeding defined by two of the most commonly used bleeding scales and clinical outcomes. Second, our results could be the result of ascertainment bias. The identification of TIMI bleeding events could be limited if hemoglobin or hematocrit values are not obtained. This could lead to an underestimation of the number of TIMI bleeds. Third, our study was a post-hoc analysis of pooled clinical trial data that was collected prospectively. Therefore, despite our robust statistical methods, there could be residual confounding. Fourth, we reclassified bleeding events from the PARAGON B trial into the relevant GUSTO and TIMI categories on the basis of the detailed clinical data that was available. Nevertheless, because PARAGON B used a definition of bleeding that was slightly modified from either the GUSTO or TIMI classification, this may limit our ability to draw firm conclusions on the relationship between both definitions and clinical outcomes. Fifth, despite the strength of the associations we found, we cannot definitively state causality between bleeding and death or MI due to the retrospective nature of our study. Finally, our study population was comprised of patients enrolled in clinical trials and may not be fully representative of patients seen in clinical practice (25). Because the study population did not include higher-risk patients, such as those who are older and have more comorbidities, the risks may be underestimated.
In conclusion, the results of our study indicate that both the GUSTO and TIMI scales identify patients with bleeding complications who are at risk for short- and intermediate-term death or MI. However, after adjusting for transfusion and when both scales were included in the same model, the increased risk with worsening GUSTO bleeding persisted while the risk with TIMI bleeding did not. These results suggest that asymptomatic decreases in hemoglobin and/or hematocrit that do not require transfusion may not be associated with a worse prognosis. Therefore, the issue of whether asymptomatic decreases in hemoglobin or hematocrit require aggressive transfusion deserves further scrutiny. This study also suggests that a combination of the GUSTO scale and need for transfusion should be used to assess bleeding complications in subsequent clinical trials of patients with ACS.
For the reconstruction of TIMI and GUSTO bleeding events from the PARAGON B dataset, please see the online version of this article.
A Comparison of the Clinical Impact of Bleeding Measured by Two Different Classifications Among Patients with Acute Coronary Syndromes
This analysis was supported by the Duke Clinical Research Institute, Durham, North Carolina.
- Abbreviations and Acronyms
- acute coronary syndrome
- coronary artery bypass grafting
- Global Use of Strategies to Open Occluded Coronary Arteries
- myocardial infarction
- Platelet IIb/IIIa Antagonism for the Reduction of Acute Coronary Syndrome Events in a Global Organization Network trial
- Platelet Glycoprotein IIb/IIIa in Unstable Angina: Receptor Suppression Using Integrilin Therapy trial
- Thrombolysis In Myocardial Infarction
- Received February 17, 2005.
- Revision received August 26, 2005.
- Accepted September 9, 2005.
- American College of Cardiology Foundation
- Moscucci M.,
- Fox K.A.,
- Cannon C.P.,
- et al.
- Cannon C.P.,
- Battler A.,
- Brindis R.G.,
- et al.
- The PARAGON B Investigators
- Chesebro J.H.,
- Knatterud G.,
- Roberts R.,
- et al.
- Boersma E.,
- Pieper K.S.,
- Steyerberg E.W.,
- et al.,
- The PURSUIT Investigators
- Berkowitz S.D.,
- Stinnett S.,
- Cohen M.,
- Fromell G.J.,
- Bigonzi F.
- Platelet Receptor Inhibition in Ischemic Syndrome Management in Patients Limited by Unstable Signs and Symptoms (PRISM-PLUS) Study Investigators
- The RESTORE Investigators