A Prospective, Randomized Clinical Trial of Hemodynamic Support With Impella 2.5 Versus Intra-Aortic Balloon Pump in Patients Undergoing High-Risk Percutaneous Coronary Intervention: The PROTECT II Study

PROTECT II was a prospective, multicenter, randomized trial conducted in 112 sites in the United States, Canada, and Europe (sites and investigators are presented in the online-only Data Supplement). The study was designed to assess whether a high-risk percutaneous revascularization strategy with the support of the Impella 2.5 device would result in better outcome than a revascularization strategy with IABP support. Each site had to demonstrate previous experience with hemodynamic support for nonemergent PCI. Predetermined need for hemodynamic support, assessed by the treating physician, was required to qualify the patient for enrollment Patients were included who were aged ≥18 years and scheduled to undergo a nonemergent PCI on an unprotected left main or last patent coronary vessel with a left ventricular ejection fraction (LVEF) ≤35%. Patients with 3-vessel disease and LVEF ≤30% were also eligible. Major exclusion criteria included recent myocardial infarction (MI) with persistent elevation of cardiac enzymes, left ventricular thrombus, platelet count ≤75 000/mm³, creatinine ≥4 mg/dL (patients already on dialysis were eligible), and severe peripheral vascular disease that precluded passage of the Impella 2.5 catheter or IABP. Complete inclusion and exclusion criteria are presented in the online-only Data Supplement.

The study was conducted in accordance with the Declaration of Helsinki. The study protocol was reviewed and approved by the Ethics Review Committee of each participating center, and all patients provided written informed consent before enrollment. Food and Drug Administration Investigational Device Exemption (IDE G050017) was obtained by the study sponsor (Abiomed, Danvers, MA) in November 2007.

The Impella 2.5 (Abiomed, Danvers, MA) is a 12F axial flow, rotary blood pump mounted on a 9F catheter deployed in a retrograde fashion across the aortic valve. The pump provides nonpulsatile forward blood flow of up to 2.5 L/min at its maximal rotation speed of 51 000 rpm. The device provides direct left ventricular unloading by aspirating blood from the left ventricle and expelling it into the aorta, thus increasing total cardiac output, reducing myocardial oxygen consumption, and decreasing the pulmonary capillary wedge pressure.^8–10 The device became commercially available in the United States during the course of the trial in June 2008 under 510(k) clearance.

After informed consent was obtained, right and left heart catheterization was generally performed by the use of femoral access. Following iliac angiography and vascular access assessment for suitability, operators declared their treatment plan based on the coronary anatomy and myocardium at jeopardy. Randomization to either the Impella 2.5 (n=225) or a commercially available IABP (n=223) was then executed in a 1:1 ratio through an automated interactive voice response system. Randomization was performed through 2 strata: geographical region (n=5) and angioplasty indication (unprotected left main/last remaining vessel versus 3-vessel disease). Crossover from one study arm to the other was not permitted. Operators were asked to aim for the most complete revascularization of the myocardium at jeopardy in a single procedure. The use of intracoronary drug-eluting or bare-metal stents, as well as adjunctive therapies such as rotational atherectomy, embolic protection devices, puncture site closure technique, anticoagulants, and glycoprotein IIb/IIIa receptor antagonists were left to the discretion of the operator. Therapeutic anticoagulation with activated clotting time >250 seconds was required. Revascularization was performed and hemodynamics, including right heart pressure, aortic pressure, and cardiac output, were measured every 15 minutes. Investigators were asked to discontinue hemodynamic support before discharge from the catheterization laboratory if the patient was deemed hemodynamically stable. Following PCI, if patients could be weaned from hemodynamic support, hemostasis was achieved by direct manual pressure when activated clotting time fell to <180 seconds or by direct surgical stitches that were deployed before the procedure. This technique, known as preclose, involves deployment of 2 Perclose devices sequentially (Abbott Vascular, IL). The stitches were deployed at 10 o'clock and 2 o'clock and isolated from the puncture site. At the end of the procedure, stitches were tied with the knot pusher, and, if necessary, manual pressure was applied to achieve complete hemostasis. Only sites experienced in this technique were allowed to use it. Patients were then admitted to the coronary care unit for observation and were discharged when appropriate. Follow-up was scheduled at 30 and 90 days postprocedure. The use of antiplatelet therapy was expected, but was left at the discretion of the investigator. Patients with intolerance to heparin, aspirin, and ADP receptor inhibitors were excluded from the study.

The primary end point was the composite rate of intra- and postprocedural major adverse events (MAEs) at discharge or 30-day follow-up, whichever was longer. A follow-up of the composite primary end point was performed at 90 days. The composite primary end point components included all-cause death, Q-wave or non–Q-wave MI, stroke, or transient ischemic attack, any repeat revascularization procedure (PCI or coronary artery bypass grafting), need for a cardiac or a vascular operation (including a vascular operation for limb ischemia), acute renal insufficiency, severe intraprocedural hypotension requiring therapy, cardiopulmonary resuscitation or ventricular tachycardia requiring cardioversion, aortic insufficiency, and angiographic failure of PCI. Non–Q-wave MI was defined as creatinine kinase-MB isoenzyme ≥3 times the upper limit of the normal range within 72 hours of the procedure or ≥2 times upper limit of the normal range beyond 72 hours postprocedure. Cardiac troponin values with the same thresholds were used if creatinine kinase-MB isoenzyme was not available. The development of new, pathological Q waves in ≥2 continuous leads was required to diagnose a Q-wave MI. Definitions of all primary end-point components are listed in the online-only Data Supplement.

Data collection, management, and monitoring, events adjudication, and statistical analysis were conducted by an independent academic clinical research organization (Harvard Clinical Research Institute, Boston, MA). Independent academic central core laboratories analyzed all angiographic (Beth Israel Deaconess, Boston, MA) and echocardiographic (Duke Clinical Research Institute, Durham, NC) data to determine angiographic success and device safety (including aortic insufficiency), respectively. An independent Clinical Events Committee adjudicated all MAEs and study end points blinded to treatment group assignment, and an independent data safety monitoring board (DSMB) monitored the safety trends on a monthly basis and provided oversight of data at 25% and 50% of planned study enrollment. In addition, the DSMB was provided with 1 formal preplanned interim analysis that included the first 50% of patients (n=327). The interim analysis aimed to evaluate the feasibility of the study with a futility guideline, and possible sample size adjustment, as well, based on the conditional power at the interim mark.¹¹ The conditional power is the probability of observing a statistically significant treatment effect at the end of a trial, conditional on the data observed at interim and under specific assumptions on the true treatment trends for the remaining 50% patients to be enrolled. A conditional power <40% at interim was the cutoff at which the DSMB could recommend stopping the study for futility. Coordination between the core laboratories, Clinical Events Committee, and DSMB was conducted by Harvard Clinical Research Institute. The study design and conduct were controlled by an executive committee chaired by the study Principal Investigator (W.W.O.).

The study was powered assuming a 30-day MAE rate of 30% in the IABP arm and a relative Impella treatment effect of 33%, resulting in a 30-day MAE rate of 20% in the Impella 2.5 arm, based on the prevalence of events in previous observational IABP studies^4,5 and the PROTECT I feasibility trial.¹² Accounting for 10% missing data, a total of 654 patients (327 per arm) were necessary to detect this treatment effect difference between Impella 2.5 and IABP at 80% power and a 2-sided α-error of 5%.

The primary end point analysis is reported for all consented randomly assigned patients undergoing high-risk PCI in the study on the intent-to-treat (ITT) principle regardless of the protocol compliance and duration of follow-up. The prespecified per protocol (PP) population includes all consented randomly assigned patients who met the protocol eligibility criteria. The treatment comparison on the primary end point (30-day MAE) and on 90-day MAE were performed by using the χ² test. As an additional supportive analysis, Kaplan–Meier estimates of the cumulative incidence of MAE through 30 and 90 days were performed, and a log-rank test was used to compare the curves between the 2 study arms at these time points.

Secondary end points included in-hospital efficacy and safety end points consisting of efficacy of hemodynamic support assessed by maximal decrease of cardiac power output from baseline, creatinine clearance change from baseline 24 hours post-PCI, device failure assessed as Impella flow <1 L/min for >5 minutes at the performance level 5 or higher (out of 9) and rate of in-hospital MAEs.

In general, the remaining data are expressed as mean±SD, median (range), or proportion as appropriate. Univariate parametric analysis was performed by using a 2-tailed unpaired t test or a nonparametric Mann–Whitney test for continuous outcomes. χ² test or Fisher exact tests were used as appropriate for nominal data. A 2-way analysis of variance with repeated measures was performed to assess the difference in hemodynamics between the 2 study arms at different time points.

Treatment comparisons on the primary end point were also performed within prespecified subgroup by using the χ² test. Relative risks, calculated as the raw Impella 2.5 event rates divided by the raw IABP event rates, and their 2-sided 95% confidence intervals are presented within each subgroup. These prespecified analyses were planned to account for the randomization scheme used in the study, the potential learning curve effect, because no roll-in subject phase was included, and the unblinded access of the investigators to the study device. The prespecified subgroup analyses included (1) use of adjunctive atherectomy: patients treated with or without atherectomy; (2) coronary anatomy: unprotected left main /last patent conduit versus patients with 3-vessel disease; (3) morbidity risk: Society of Thoracic Surgery (STS) risk of <10 versus ≥10; and (4) learning curve effect: the first Impella 2.5 and IABP patient at each site versus all other treated patients.

All probability values were 2-tailed and considered significant when the probability was <0.05.

The statistical analyses for this report were performed by the Harvard Clinical Research Institute using a SAS version 9.2 (SAS Institute Inc, NC). The authors had full access to the data and take responsibility for its integrity. The corresponding author and study chair (W.W.O.) prepared the first and all subsequent drafts of this report, which were then shared with the coauthors for comments. All authors have read and agreed to the manuscript as written.

In both study arms, more lesions were attempted than originally anticipated. The number of attempted lesions and deployed stents were similar between the 2 groups, although the proportion of patients treated with ≥3 stents was slightly higher in the Impella 2.5 arm. There were significant differences between the 2 study arms with respect to the use of adjunctive therapies (Table 2). In the Impella 2.5 arm, glycoprotein IIb/IIIa receptor antagonists were used less frequently, whereas the use of rotational atherectomy tended to be more frequent. The use of rotational atherectomy was also more vigorous in the Impella arm, with more runs and longer durations. Rotational atherectomy was also more frequently performed in unprotected left main lesions in the Impella 2.5 arm. Finally, the volume of contrast used was significantly greater in the Impella 2.5 arm. Patients randomly assigned to IABP had longer duration of support in comparison with those in the Impella 2.5 arm.

At 69% of the planned enrollment, the primary end point of 30-day MAE occurred in 35.1% of patients in the Impella 2.5 arm in comparison with to 40.1% in the IABP arm (P=0.277; Table 3). At 90 days, patients supported with Impella 2.5 showed a trend toward lower MAE rate in comparison with those supported with IABP (40.6% versus 49.3%, P=0.066; Table 3).

There was no significant difference in the occurrence of in-hospital death, stroke, MI, or the composite of death/stroke/MI between Impella 2.5 and IABP. After hospital discharge, fewer irreversible MAEs of death/stroke/MI (7.1% versus 12.8%, P=0.047) and of death/stroke/MI/repeat revascularization events (9.8% versus 18.3%, P=0.01) occurred in the Impella 2.5 arm in comparison with the IABP arm. At 90 days, there were also fewer repeat revascularization events with the Impella 2.5 in comparison with IABP in both hierarchical (3.6% versus 7.8%, P=0.056) and nonhierarchical (6.3% versus 11.9%, P=0.039) analyses.

At 69% of the planned enrollment, 30-day MAE occurred in 34.3% of Impella 2.5 patients in comparison with 42.2% of IABP patients (P=0.092). In comparison with the IABP arm, the 90-day MAE rate was significantly lower in the Impella 2.5 arm (40.0% versus 51.0%, P=0.023) yielding a relative risk reduction of 22% (Table 4). Although there was no difference in in-hospital death, stroke, MI, or the composite of death/stroke/MI between Impella 2.5 and IABP, fewer irreversible MAEs of death/stroke/MI (7.0% versus 12.9%, P=0.042) and of death/stroke/MI/repeat revascularization (9.8% versus 18.6, P=0.009) occurred after hospital discharge in the Impella 2.5 arm in comparison with the IABP arm. At 90 days, there were also fewer repeat revascularization events in the Impella 2.5 arm in comparison with the IABP arm in both hierarchical (3.7% versus 8.1%, P=0.055) and nonhierarchical (6.0% versus 12.4%, P=0.024) analyses.

In aggregate, patient cardiac function and functional status improved significantly after revascularization. At 90-day study exit follow-up, there was an average 22% relative increase in LVEF from baseline (27%±9 versus 33%±11, P<0.001) and a 58% improvement in New York Heart Association functional class III/IV (62% versus 26%, P<0.001). The improvement in LVEF and New York Heart Association was similar between the 2 study groups. However, and as depicted in the Kaplan–Meier curves for both ITT and PP populations (Figure 2A and 2B), patients treated with Impella 2.5 experienced fewer MAEs over the course of the study in comparison with those treated with IABP. Most of the differences in patient MAE outcomes occurred out of the hospital. These new events were overt and mainly driven by death or rehospitalizations for MI, stroke, and repeat revascularization in a patient population that was event free at discharge.

Complete hemodynamic monitoring was obtained in 279 patients (138 for IABP and 141 for Impella 2.5). Cardiac power output,¹³ defined as cardiac output×mean arterial pressure×0.0022, was calculated to account for systemic blood flow and maintenance of physiologically appropriate blood pressure during the procedural ischemic times. As determined by the maximal drop in cardiac power output from baseline, Impella 2.5 provided better hemodynamic support than IABP during these high-risk procedures (−0.04±0.24 versus −0.14±0.27 W, P=0.001). Change in creatinine clearance was similar between Impella and IABP patients 24 hours after PCI in comparison with baseline (4.64±15.06 versus 4.66±13.55, P=0.988), despite the higher volume of contrast media received by Impella patients. There were no Impella device failures, and no difference was observed in in-hospital composite MAE or any of its components between Impella and IABP arms for ITT (32.4% versus 30.9%, P=0.733) or PP populations (31.9% versus 32.7%, P=0.867).

Prespecified subgroup analyses on 30-day and 90-day MAE are depicted in Figure 3A and 3B (ITT) and Figure 4A and 4B (PP), respectively. Of particular interest, in the ITT population not treated with atherectomy (88% of the entire population, n=396), the Impella 2.5 patients had better outcomes in comparison with those who received an IABP, with a significant 25% relative risk reduction in the MAE incidence at 90 days (36.5% versus 48.7%, P=0.014). The 30-day MAE rate was also lower in the Impella 2.5 arm in comparison with the IABP arm (30.6% versus 39.6%, P=0.060). These findings were magnified for the PP nonatherectomy group with a 30% relative reduction in MAE at 30 days (29.3% versus 41.9%, P=0.011) and 90 days (35.5% versus 50.5%, P=0.003) in the Impella 2.5 arm in comparison with the IABP arm, respectively. Patients with STS scores <10 had better 90-day outcomes with Impella 2.5 than with IABP (37.4% versus 48.6%, P=0.030), whereas there was no difference between the 2 groups for patients with STS scores ≥10. Of note, 27% of the Impella patients with STS ≥10 had atherectomy, ≈3 times the rate of atherectomy use of the rest of the population, overlapping with the first subgroup analysis. Patients in the Impella 2.5 arm had strong trends on 30-day MAE and significantly fewer MAEs at 90 days when the first IABP and Impella 2.5 patients enrolled at each site (roll-in subjects) were excluded from the analysis to account for the learning curve (P=0.029 in ITT and P=0.016 in PP populations, respectively).

Because of the DSMB determination of futility, this trial was terminated on the assumption from the first 50% (327) of patients enrolled. Ultimately only 69% (452) of the planned enrollment occurred. Because the trial did not enroll the number of patients it was powered for, definitive statements concerning the primary end point are speculative. Unfortunately, unbeknownst to the DSMB, a significant learning curve occurred in this trial with marked improvement in safety for Impella-supported patients who were treated in the last half of the trial (Figure 5).

Because of the different radiographic appearance, operators could not be blind to treatment assignment. Knowledge of the presence of Impella support unfortunately led to a greater and more aggressive use of rotational atherectomy in this subgroup. These differences confounded the analysis because of the markedly higher rate of cardiac isoenzyme elevations.

Because the difference in 30-day MAE did not reach statistical significance for the entire study, the analysis of 90-day events remains exploratory. It must be emphasized, however, that 90-day follow-up with end-point analysis was prospectively planned and provides a picture of longer-term safety. The increase in overt, clinical events between 30 and 90 days is an important clinical observation.