Author + information
- Received September 1, 2005
- Revision received May 5, 2006
- Accepted May 16, 2006
- Published online October 3, 2006.
- Marc Cohen, MD, FACC⁎,⁎ (, )
- Kenneth W. Mahaffey, MD, FACC†,
- Karen Pieper, MS†,
- Charles V. Pollack Jr, MD‡,
- Elliott M. Antman, MD, FACC§,
- James Hoekstra, MD∥,
- Shaun G. Goodman, MD, FACC¶,
- Anatoly Langer, MD, FACC¶,
- Jacques J. Col, MD#,
- Harvey D. White, MD⁎⁎,
- Robert M. Califf, MD, FACC†,
- James J. Ferguson, MD, FACC††,
- SYNERGY Trial Investigators
- ↵⁎Reprint requests and correspondence:
Dr. Marc Cohen, Cardiac Cath Lab Administration, Newark Beth Israel Medical Center, 201 Lyons Avenue, Newark, New Jersey 07112.
Objectives The purpose of this study was to compare the effect of receiving pretreatment with antithrombin before randomization as well as overall efficacy and safety of enoxaparin versus unfractionated heparin (UFH) in the SYNERGY (Superior Yield of the New Strategy of Enoxaparin, Revascularization, and Glycoprotein IIb/IIIa Inhibitors) trial.
Background The SYNERGY trial results demonstrated noninferiority in outcomes with enoxaparin compared with UFH. Randomized treatment was independent of prerandomization treatment.
Methods Analyses were first performed on the 4 prerandomization subgroups: patients who received no antithrombin therapy and those who were treated with enoxaparin or UFH or both. Then, we focused on the subgroup of patients who received no pretreatment or were pretreated with and randomized to the same drug. Of the 9,978 patients, 2,440 did not receive prerandomization therapy and 6,138 received consistent therapy through randomization. The primary end point was the composite of death and nonfatal myocardial infarction (MI) at 30 days.
Results After adjustment for differences among the subgroups, no significant difference in the association between the 4 pretreatment groups and death or MI remained (p = 0.171). The randomized treatment effect on 30-day death or MI tended to vary with pretreatment (p = 0.055 for interaction test after adjustment). Patients who received consistent therapy with enoxaparin had significantly less death or MI than patients randomized to UFH (adjusted p = 0.041) with a trend toward increased bleeding.
Conclusions Treatment with antithrombin therapy before randomization had potential impact on comparison of study drug effects. After adjustment for differences in baseline characteristics between subgroups, consistent therapy with enoxaparin might be superior to UFH in reducing death or nonfatal MI, with a modest excess in bleeding.
The SYNERGY (Superior Yield of the New Strategy of Enoxaparin, Revascularization, and Glycoprotein IIb/IIIa Inhibitors) trial (1) was conducted to compare subcutaneous enoxaparin and intravenous unfractionated heparin (UFH) in patients with high-risk non–ST-segment elevation acute coronary syndromes (NSTE ACS). The protocol recommended early invasive management and guidelines-based therapy, including aspirin, glycoprotein (GP) IIb/IIIa inhibitor, and clopidogrel (2,3).
In contrast to earlier trials in which enoxaparin proved superior to UFH in NSTE ACS (4–8), the SYNERGY trial demonstrated no significant difference in the incidence of the primary end point of death or myocardial infarction (MI) at 30 days. The SYNERGY patient population was higher risk than prior trial cohorts, a more invasive and earlier management strategy was used, and contemporary evidenced-based secondary prevention was much more common. Whether these factors attenuated any true difference between antithrombin agents is unknown. Another important difference between the SYNERGY trial and prior trials was that a substantial proportion of the SYNERGY patients had been initiated on enoxaparin or UFH by a treating physician before enrollment. Randomization was made independently of pre-enrollment antithrombin therapy (pretreatment). Because of the expectation that some randomized patients would be pretreated, a series of analyses to assess the impact of prior therapy was prespecified in the SYNERGY trial protocol. In this paper, we will examine the results of these analyses.
The rationale, enrollment criteria, design, and end points of the SYNERGY trial as well as the approach to statistical analysis have been previously described (1,9).
Patients and study design for the SYNERGY trial
In brief, eligible patients had ischemic symptoms for at least 10 min within the 24 h preceding enrollment and at least 2 of the following high-risk characteristics: age ≥60 years, troponin or creatine kinase elevation above the upper limit of normal, or ST-segment changes on the electrocardiogram. Exclusion criteria included known or suspected pregnancy, contraindications to anticoagulation, planned spinal or epidural anesthesia, percutaneous coronary intervention (PCI) or thrombolysis within the 24 h preceding enrollment, and calculated creatinine clearance <30 ml/min. Written informed consent was obtained from all patients.
Eligible patients were randomized in an open-label fashion to enoxaparin (1 mg/kg every 12 h) or UFH (bolus of 60 U/kg [maximum of 5,000 U] and initial infusion of 12 U/kg/h [maximum of 1,000 U/h initially]) to achieve an activated partial thromboplastin time (aPTT) of 1.5 to 2.0 times the institutional upper limit of normal or 50 to 70 s. Patients receiving enoxaparin before random assignment to enoxaparin were given their first study dose 12 h after their last injection of enoxaparin. Patients already receiving enoxaparin who were randomly assigned to UFH started UFH infusion at 12 U/kg/h (without bolus, 1,000 U/h maximum) 8 h after their last enoxaparin dose. If their last enoxaparin dose was between 8 and 12 h before randomization, UFH infusion at 12 U/kg/h (1,000 U/h maximum) was started after a 30-U/kg bolus (5,000 U maximum). Patients pretreated with UFH who were randomly assigned to UFH continued the drug with aPTT checks every 6 h; those patients randomly assigned to enoxaparin had their UFH stopped and were given trial enoxaparin immediately regardless of aPTT.
Study drug was continued until the treating physician judged that no further anticoagulation was required but at least through angiography and revascularization, if performed. Recommendations concerning antithrombin dosing for patients undergoing PCI or coronary artery bypass grafting (CABG), sheath removal, and use of closure devices were provided and are described elsewhere (1,9). For emergency invasive procedures, enoxaparin or UFH was stopped and surgery performed regardless of the timing of the last dose.
All patients received aspirin or clopidogrel daily beginning at the time of enrollment. Glycoprotein IIb/IIIa inhibitor use was encouraged. All other medications were administered at physician discretion.
Subgroups, derived from the 9,978 SYNERGY trial patients, were defined on the basis of antithrombin treatment before randomization as reported by the investigator (none, enoxaparin only, UFH only, or both). To fully assess the impact of pretreatment on the clinical outcomes of the randomized treatment groups, analyses were first performed on the 4 prerandomization subgroups. Subsequent analyses were focused on the subgroup of patients who either received no pretreatment or were pretreated with and randomized to the same drug. This subgroup, the consistent therapy group, was treated in a consistent manner with one antithrombin from initial presentation through randomization. Included within the consistent therapy subgroup were 501 patients who underwent postrandomization crossover to the other antithrombin, directed by the treating physician; these patients were not excluded.
The primary efficacy end point was the combined incidence of all-cause death and nonfatal MI during the 30 days after randomization. Secondary efficacy end points included the combined incidence of death or nonfatal MI at 14 days; the combined incidence of all-cause mortality, nonfatal MI, stroke, or recurrent ischemia requiring revascularization; and individual components of this composite at 14 and 30 days after enrollment. The primary safety end point was the incidence of major bleeding. Bleeding was assessed according to both Thrombolysis In Myocardial Infarction (TIMI) and Global Utilization of Strategies to Open Occluded Arteries (GUSTO) trial criteria as well as by the total number of blood transfusions received (9). A blinded clinical events committee adjudicated all suspected incidents of MI and stroke (1,9).
Analyses were performed on the 4 pretreatment subgroups and on the patients who received consistent therapy versus those who did not. Continuous factors are presented as medians with 25th and 75th percentiles. Categorical factors are presented as frequencies and percentages. A global test of at least 1 pretreatment group being different from the others was used to derive p values across the baseline factors. A Pearson’s chi-square test was used for categorical factors and a Wilcoxon signed rank test for continuous variables.
Tests were first performed to evaluate differences in the association between pretreatments and outcomes regardless of the randomized therapy. A Pearson’s chi-square test was used for unadjusted comparisons across the key primary and secondary end points.
Multivariable logistic regression models were developed in this population for the end points of GUSTO severe bleed and TIMI major bleed. Covariates considered for these models were randomized treatment, age, weight, heart rate, creatinine clearance, baseline hemoglobin, gender, transient ST-segment elevation, ST-segment depression, diabetes, hypertension, history of CABG, history of PCI, enrolling country, and prerandomization use of a GP IIb/IIIa inhibitor before coronary artery disease and before MI. Both stepwise and backwards variable selection techniques were used to determine the final models. These models were then used to adjust for covariates when evaluating the effect of pretreatment on the 2 key bleed outcomes.
A multivariable model of 30-day death or MI had been previously developed and validated in a similar population (10). This model was used to adjust for confounders when comparing pretreatment strategies. Multivariable models were developed in this population for the end points of GUSTO severe bleed and TIMI major bleed. Covariates considered for these models were randomized treatment, age, weight, heart rate, creatinine clearance, baseline hemoglobin, gender, transient ST-segment elevation, ST-segment depression, diabetes, hypertension, history of CABG, history of PCI, U.S. versus non-U.S., and prerandomization use of a GP IIb/IIIa inhibitor. The interactions of the randomized treatment with these factors were also considered.
Both stepwise and backwards variable selection techniques were used to determine the final models. These models were then used to adjust for covariates when evaluating the effect of pretreatment on the 2 key bleed outcomes.
Analyses were also performed including the randomized treatment. These models included the interactions of pretreatment with randomized therapy to evaluate whether the type of pretreatment was associated with differences in the randomized treatment effect. The critical value for statistical significance was 0.05. All analyses were performed with the SAS statistical program (SAS Institute, Cary, North Carolina).
Derivation of patient subgroups and demographics
Of the 9,978 patients in the SYNERGY trial, 2,440 did not receive any prerandomization antithrombin therapy, 4,294 received enoxaparin, 2,940 received UFH, and 304 received both enoxaparin and UFH before randomization (Fig. 1).The baseline characteristics of the 4 pretreatment subgroups are shown in Table 1.There were statistically significant differences in the baseline demographics between groups of patients defined by presence or absence of prerandomization antithrombin therapy or type of pretreatment if used. However, there was no clear trend in how these groups were selected by physicians for pretreatment relative to known cardiovascular risk factors. For example, patients receiving enoxaparin were more likely to be female and have all 3 inclusion criteria whereas patients pretreated with UFH were more likely to have ST-segment elevation and a history of MI and angina. Patients with no prerandomization therapy were less heavy, more often female, more often white, had less ST-segment elevation and T-wave inversion, and less frequently had all 3 enrolling risk factors. The patients who received UFH only before randomization were similar to enoxaparin-only patients with regard to baseline risk.
There is a significant difference in the association between the 4 pretreatment groups and death or MI (unadjusted p = 0.017). Those who received UFH only before randomization seemed to have more MIs (Table 2).There is no association between the type of pretreatment received and having a TIMI major bleed (p = 0.632) or GUSTO severe bleed (p = 0.923) according to unadjusted analyses (Table 3).
The effect of prerandomization therapy on death and death/MI was not significantly different after adjustment for baseline differences (adjusted p = 0.171). In addition, after adjustment for baseline differences, the lack of association between bleeding outcomes and pretreatment remains (TIMI major bleed p = 0.592, GUSTO severe bleed p = 0.711). Therefore, it seems that the type of pretreatment alone is not associated with increased risk of ischemic or bleeding events.
Among the 9,978 patients randomized in the SYNERGY trial, 6,138 had consistent therapy through the time of randomization. Investigator-defined consistent therapy (6,138) was based on the investigator indicating on the case report form (CRF) that the patient had received a pretreatment in the emergency department of the same drug as the randomized therapy or had received no pretreatment at all. Among these 6,138 patients, 2,740 (45%) were randomized to UFH and 3,398 (55%) to enoxaparin. Patients receiving consistent therapy through randomization did not demonstrate an improvement in the efficacy end points versus patients who did not receive consistent therapy (Table 4).
Comparison of enoxaparin versus UFH among pretreatment subgroups
The impact of receiving pretreatment on the randomized treatment effect is depicted in Figure 2and Figure 3.The overall test for a statistical interaction between pretreatment and randomized therapy is borderline significant in the unadjusted analyses (p = 0.0498) and in the adjusted analysis (p = 0.055). In the patients with no pretreatment, the odds of death or MI to 30 days were less in the patients randomized to enoxaparin than in those randomized to UFH. There was little randomized treatment effect in the patients who received either UFH only or enoxaparin only, prerandomization.
The TIMI bleed for pretreatment subgroups is presented in Figure 3. There is no evidence that receiving a particular pretreatment was associated with significantly changing the randomized treatment effect (p = 0.270 unadjusted, p = 0.331 adjusted). All 4 pretreatments are associated with a trend toward greater likelihood of having a TIMI bleed for patients who were randomized to enoxaparin versus those randomized to UFH. As with TIMI bleed, there is no statistical evidence that receiving pretreatment was associated with a significant change in the randomized treatment effect for GUSTO severe bleeds (p = 0.267 unadjusted, 0.227 adjusted) (Fig. 4).
An apparent difference in the randomized treatment effect for those with consistent therapy versus those without was observed (Fig. 5)(p = 0.0004 unadjusted, 0.018 adjusted). Those who received consistent therapy with enoxaparin experienced fewer deaths or MIs than those who received consistent therapy with UFH. The incidence of death or MI was 13.3% and 15.9% in enoxaparin-treated and UFH-treated patients, respectively (hazard ratio 0.82; 95% confidence interval 0.72 to 0.94; p = 0.004, adjusted p = 0.041) (Table 5).This treatment difference was seen at 48 h and persisted through 14 and 30 days. Secondary end points at 30 days for the consistent therapy group are shown in Table 5.
The randomized treatment effect on TIMI bleeding did not vary between those with and without consistent therapy (p = 0.989 unadjusted, 0.583 adjusted for interaction test) (Fig. 6).In both randomized treatment groups, most TIMI major bleeding occurred in the setting of CABG. There was no statistical evidence that the randomized treatment effect varied between those with consistent therapy and those without for GUSTO severe bleeds. The interaction tests were not significant (p = 0.265 unadjusted, 0.114 adjusted). It seemed that the interaction test showed that treatment effects on bleeding were not statistically different between those with consistent therapy and those without. There was increased GUSTO severe bleeding with consistent therapy with enoxaparin versus UFH (2.9% vs. 2.1%, p = 0.0465) (Fig. 7).In the patients without consistent therapy these rates were 2.4% and 2.3%, respectively (p = 0.894).
The primary objective of the SYNERGY trial was to test whether enoxaparin was a more effective anticoagulant than UFH in high-risk NSTE ACS patients managed primarily with an early invasive strategy. For the most part, however, the SYNERGY trial became a comparison of these 2 agents in patients who had antithrombin therapy before randomization. In the primary analysis of the SYNERGY trial (n = 9,978), enoxaparin was found to be noninferior to UFH in reducing death or nonfatal MI but with an excess in bleeding.
Pretreatment with antithrombin therapy or switching from one antithrombin to another at the time of randomization could well attenuate differences in the efficacy and safety of antithrombin agents. Because it was the physician’s decision to use a specific therapy before randomization, these patients might have been different from patients not started on therapy before randomization and those patients treated with enoxaparin or UFH might have been different from one another. Also, patients ultimately randomized to UFH might have derived meaningful benefit from pretreatment with enoxaparin or vice versa. Finally, switching from one antithrombin to another might affect efficacy and safety outcomes by creating a gap in therapy (under-anticoagulation) or result in “stacking” of antithrombin effect (over-anticoagulation). Although recommendations regarding the switch from one antithrombin to another at the time of randomization were provided in the protocol, there are no validated algorithms for switching of therapies.
After adjustment there was a trend for the randomized treatment effect to be different within the 4 pretreatment groups (p = 0.055). This was significant for consistent therapy versus no consistent therapy (p = 0.018). The consistent therapy group had a significantly lower rate of death/MI (p = 0.041). The trend was opposite but not statistically significant within the no-consistent-therapy group. Thus it seems that within the consistent therapy subgroup enoxaparin was associated with a statistically significant clinical benefit on death or MI for enoxaparin compared with UFH. There are trends such that receiving consistent therapy was associated with an increased risk of bleed in the enoxaparin arm compared with the UFH arm. Thus, in the subset of patients who did not change from one antithrombin pretreatment to the other at the time of randomization, enoxaparin seemed to have a significant efficacy advantage over UFH at the cost of moderate increases in bleeding.
Although these subgroup analyses are potentially influenced by multiple confounders and biases despite statistical modeling, these results are consistent with those of earlier randomized trials that suggested superiority of enoxaparin to UFH in patient populations with markedly less pretreatment and switching of antithrombin therapy and time of randomization. Specifically, in the ESSENCE (Efficacy and Safety of Subcutaneous Enoxaparin in Non–Q-Wave Coronary Events) trial (4) in which there were no pretreated patients or in TIMI-11B (5) and the INTERACT (Integrilin and Enoxaparin Randomized Assessment of Acute Coronary Syndrome Treatment) (8) trials where only 35% and 20%, respectively, of patients had prior antithrombin therapy, randomization to enoxaparin was associated with a significant reduction in recurrent ischemic events as measured by the triple composite end point (death, MI, refractory angina) and a very strong trend favoring enoxaparin in terms of the incidence of death and MI. In the ACUTE II (Antithrombotic Combination Using Tirofiban and Enoxaparin II) (11) and A-to-Z (Aggrastat to Zocor) (12) randomized trials, in which the majority of patients had received an antithrombin before enrollment and randomization, enoxaparin was found to be noninferior to UFH. From this perspective, there seems to be consistency across trials regarding the efficacy of enoxaparin relative to UFH. Switching therapy to or from enoxaparin might well have attenuated the efficacy effect and biased the overall SYNERGY study results toward the null hypothesis.
Differences exist between the patients receiving no prerandomization therapy and the patients treated with antithrombins before enrollment. Attempts to account for selection bias in pretreatment were made through adjustment for confounders. However, it is possible that there are factors not collected in the study that also influenced the decision to pretreat. The trial was powered to evaluate treatment differences in the total group. Therefore, there is inadequate power to detect treatment differences in these individual subgroups, especially the sample of 2,440 patients not confounded by any prior antithrombin pretreatment.
Clinical and research implications
The overall SYNERGY trial results showed noninferiority of enoxaparin relative to UFH. One perspective might be that the SYNERGY trial results were diminished by the confounding variable of “pretreatment.” However, the purposeful inclusion of pretreated patients generated data suggesting that there was no reason to deliberately switch patients already receiving one antithrombin to another, because those patients with consistent therapy with enoxaparin had the best ischemic outcomes and similar bleeding excess as the overall trial population after adjustment for potential confounders.
Therefore, health care professionals who are “first responders” to a patient with NSTE ACS not started on antithrombin have these data to support initiating therapy with enoxaparin as first-line therapy. The results of the systematic overview analysis of the 6 major prospective randomized trials examining enoxaparin versus unfractionated heparin also support this approach (13). Petersen et al. (13) observed a treatment benefit of enoxaparin across the trials in the cohort of patients that did not receive antithrombin before randomization.
Clinical researchers should also consider these data in the design and conduct of future clinical trials. Controlling for prerandomization therapy, particularly when the agents being evaluated are commercially available, might be important to enhance the detection of efficacy and safety differences between agents (14). In addition, algorithms for switching from one agent to another—a practice that is common in both clinical trials and clinical practice—need to be validated with appropriate outcomes trials.
Pretreatment with antithrombin therapy in the SYNERGY trial might have affected the treatment differences between enoxaparin and UFH. After adjustment for differences in patients with and without pretreatment and with and without consistent therapy, it seems that patients treated only with enoxaparin have better clinical outcomes with a modest excess in bleeding. The common practice of pretreatment and switching of antithrombin therapy in clinical trials and clinical practice needs further evaluation.
The authors would like to thank Louise Traylor, PhD, for her statistical analyses and Erin Allingham for her help in preparing the manuscript.
The SYNERGY trial was funded by Aventis Pharmaceuticals, part of the sanofi-aventis group. Please note the following author disclosures: Dr. Cohen: research grants, speakers’ bureau (Aventis, Sanofi); Dr. Mahaffey: research grants, consult/other, speakers’ bureau (Aventis); Dr. Traylor: employment (Aventis); Dr. Pollock, Jr.: speaker’s bureau, research support (Aventis); Dr. Antman: research grants (Aventis); Dr. Hoekstra: speakers’ bureau (sanofi-aventis); Dr. Goodman: research grants, consult/other, speakers’ bureau (Aventis); Dr. Langer: research grants, speakers’ bureau, consulting fees (Aventis Canada); Dr. Col: consult/other (Aventis); Dr. White: research grants (Aventis); Dr. Califf: research grants, consult (Aventis, Sanofi-Synthelabo, Bristol-Myers Squibb/Sanofi Pharmaceuticals Partnership); Dr. Ferguson: research grants (Aventis, Sanofi-Synthelabo), consult/other (Aventis), speakers’ bureau (Aventis, Sanofi-Synthelabo).
- Abbreviations and Acronyms
- activated partial thromboplastin time
- coronary artery bypass graft
- Global Utilization of Strategies to Open Occluded Arteries trial
- myocardial infarction
- NSTE ACS
- non–ST-segment elevation acute coronary syndromes
- percutaneous coronary intervention
- Superior Yield of the New Strategy of Enoxaparin, Revascularization, and Glycoprotein IIb/IIIa Inhibitors trial
- Thrombolysis In Myocardial Infarction
- unfractionated heparin
- Received September 1, 2005.
- Revision received May 5, 2006.
- Accepted May 16, 2006.
- American College of Cardiology Foundation
- Braunwald E.,
- Antman E.M.,
- Beasely J.W.,
- et al.
- Bertrand M.E.,
- Simoons M.L.,
- Fox K.A.,
- et al.
- Antman E.M.,
- McCabe C.H.,
- Gurfinkel E.P.,
- et al.
- Antman E.M.,
- Cohen M.,
- Radley D.,
- et al.
- Goodman S.G.,
- Cohen M.,
- Bigonzi F.,
- et al.
- Goodman S.G.,
- Fitchett D.,
- Armstrong P.W.,
- Tan M.,
- Langer A.,
- Integrilin and Enoxaparin Randomized Assessment of Acute Coronary Syndrome Treatment (INTERACT) Trial Investigators
- Boersma E.,
- Pieper K.S.,
- Steyerberg E.W.,
- et al.
- Cohen M.,
- Theroux P.,
- Borzak S.,
- et al.,
- ACUTE II Investigators
- Blazing M.A.,
- de Lemos J.A.,
- White H.D.,
- et al.,
- A-to-Z Investigators
- Diener H.C.,
- Bogousslavsky J.,
- Brass L.M.,
- et al.,
- MATCH Investigators