Bone Marrow Mononuclear Cell Therapy for Acute Myocardial Infarction: A Perspective from the CCTRN

Introduction

Recent meta-analyses of cell therapy clinical trials have suggested that bone marrow mononuclear cell (BMC) delivery following acute myocardial infarction (AMI) may result in modest improvement in left ventricular LV function.¹ In spite of this, the uniformly null findings emerging from the most current trials, TIME², LateTIME³, and SWISS-AMI⁴ have prompted careful reconsideration of this approach.

Background

By late 2006, multiple preclinical models of AMI suggested that delivery of BMC-derived cells improved LV function following AMI.^{5, 6} While the original study by Orlic suggested trans differentiation as the mechanism of action, this was not confirmed by others.^{7, 8} However the study by Balsam demonstrated functional benefits in spite of failing to provide data supporting trans differentiation. It is thought that BMCs might have pleiotropic and diverse effects in this setting including stimulating angiogenesis and other paracrine effects.⁹ These findings fueled intense interest in assessing effects of autologous BMC delivery on LV function in clinical trials.^10–13 While initial studies showed somewhat mixed results, meta-analyses supported a significant effect of intracoronary delivery of BMCs on left ventricular ejection fraction (LVEF). Three points are important to note. (1) These trials varied widely in design and subject characteristics. (2) None of these meta-analyses used patient level data. (3) The trial responsible for driving the perceived benefit was REPAIR-AMI¹⁰, the largest (n=204) trial reported by 2006, with an almost two-fold difference in LVEF improvement in the active group vs. placebo, and nonrandomly allocated assignment of time of cell delivery. The CCTRN was designed to execute multiple, simultaneous cell therapy protocols in the clinical setting of LV dysfunction.¹⁴ After consideration the investigators decided to test the effect of timing of administration of BMCs while using standard approaches to other features of trial design common to other studies.

Although a pre-specified variable in REPAIR AMI, the timing of cell delivery was not subject to randomization in it or earlier studies. Thus, the TIME and LateTIME trials were initiated to study the impact of timing of BMC delivery following AMI. During the same time period, Swiss investigators, working independently, also decided to focus on timing of BMC administration following AMI (SWISS-AMI).¹⁵ TIME evaluated cell delivery at Day 3 or Day 7 following reperfusion (primary percutaneous coronary intervention (PCI) with stenting), LateTIME evaluated cell delivery 2–3 weeks post-reperfusion, and SWISS-AMI compared the effects of delivery on Day 5 to 7 versus 3–4 weeks post-PCI (Figure 1).

Figure 1.

Placebo adjusted effect size for change in LVEF over time as a function of time from primary PCI in four clinical trials. The upper and lower bar limits reflect the upper and lower confidence interval of the effect size; the bar width is proportional to the sample size.

Internal Consistency of TIME, LateTIME and SWISS-AMI

TIME, LateTIME, and SWISS-AMI were contemporaneous, prospective, randomized, controlled trials. TIME and LateTIME recruited predominantly ST segment elevation myocardial infarctions (STEMIs) and were both placebo controlled and double-blind design; SWISS-AMI recruited exclusively STEMIs but did not have placebo controls and was an open design and using mostly LV angiograms post-PCI to qualify subjects. The primary endpoints focused on global LVEF measured by cardiac magnetic resonance (cMR) imaging, but the timing of this was different (TIME/LateTIME at 6-months and SWISS-AMI at 4 months). Intention-to-treat analyses were conducted in each study. Each was designed to detect moderate- to-large placebo adjusted changes in LVEF. Randomizations were 2:1 (active: placebo) in both TIME and LateTIME and 1:1:1 in SWISS-AMI. Cell processing was by manual Ficoll processing at a central center in SWISS-AMI, whereas the two CCTRN studies utilized on-site automated Ficoll processing (SEPAX, Biosafe, SA). Cell dose and delivery were the same in each of these three studies using the intracoronary stop-flow technique. While differences did exist between the two studies (in SWISS-AMI there was an open design, use of LV angiography immediately post-PCI for qualify, no requirement of primary PCI or stents in eligible patients, and central cell processing requiring > 24 hour delayed delivery of BMC), the similarities suggested that a comparison of their results would be productive.

Overall, the primary results for TIME², LateTIME³, and SWISS-AMI⁴ were each null with no detectable benefit of cell therapy evident when administered at Day 3, Day 7, 2–3 weeks, or 3–4 weeks post PCI. Thus, in spite of prior clinical studies and recent meta-analyses¹⁶ supporting an effect of BMC delivery on echocardiogram-derived LV function post-AMI, these three studies did not detect a significant treatment effect on LV function. Evaluation of the clinical endpoints revealed no safety concerns but the intracoronary administration of BMCs did not improve LV function following AMI irrespective of the timing of administration.

Variables addressed in these studies

Study population

Since compelling work from the REPAIR-AMI trial¹⁷ suggested that AMI patients with the greatest impairment of LVEF appeared to gain the most benefit from BMC therapy, the CCTRN chose to study patients with infarctions resulting in an LVEF of <45% following successful reperfusion by PCI. Given the need to randomize patients in TIME by day 2, local echocardiographic readings were used to screen patients whereas baseline and endpoint values were determined by core laboratory assessment of cMR imaging. In TIME and LateTIME, these qualifying echocardiograms, which were obtained closer in time to reperfusion than the following baseline cMRs, revealed lower LVEF compared with baseline cMR (Figure 2A), resulting in the inclusion of a population with less LV dysfunction than proposed. As a result, a significant part of our patient population in both TIME and LateTIME had less LV dysfunction (as measured by cMR) than anticipated. Reducing the threshold for enrollment to say, LVEF ≤ 40%, or obtaining screening core cMR closer to the time of delivery, are admissible alternatives for future trials, although each comes with greater logistical challenges, financial cost and risks to timely recruitment.

Figure 2.

Figure 2A demonstrates the relationship between the baseline LVEF MRI based assessment and the screening echocardiographic based LVEF. The correlation is substantial and in general the core lab assessment is greater than the echo based assessment.

Figure 2B depicts the relationship between the change in LVEF over time and the baseline ejection fraction in the TIME and LateTIME trials. On average, the greater the baseline LVEF, the smaller the increase in LVEF from baseline to six months.

In SWISS-AMI that randomized subjects to early treatment (5–7 days), late treatment (3–4 weeks), or control, patients were screened by LV angiogram or echocardiography (<45%) the day of or after AMI. The median baseline LVEF was 37% by cMR. Delivery of BMCs demonstrated no benefit in spite of the greater baseline degree of dysfunction. Thus, we believe that it is unlikely that the degree of baseline LV dysfunction was a major reason for the null results.

In the face of these null findings for LVEF, power becomes a critical factor. SWISS-AMI was powered to detect a 3.5 (absolute LVEF unit) placebo adjusted change (over four months) in EF. TIME was powered to detect a 5 unit placebo adjusted change (over six months). Although each of TIME and LateTIME were adequately powered overall, the sample sizes in the LVEF ≤ 40 subgroups were too small and underpowered to detect these same effect sizes.

The planned similarities between TIME and LateTIME permit the opportunity to conduct further evaluation of the combined datasets. An analysis was completed using a dataset containing 81 of the 87 patients from LateTIME, and 112 of the 120 patients from TIME, all of whom had paired cMR LV images at baseline and six months. We observed no overall effect of BMC therapy on the change in LVEF over time (placebo adjusted change in LVEF −1.4 ± 9.5: p=0.967; 95% CI −4.2 to 1.5) in this combined dataset. Furthermore, the placebo corrected changes from baseline to six months in the two studies were not statistically different from each other.

Examination of this combined data set for the effects of age, baseline LVEF, and time from PCI to infusion identified only baseline LVEF as significantly associated with change in LVEF regardless of treatment (b= −0.22 p = 0.001; 95% CI = −0.34 to −0.10) (Figure 2B). This effect remained after adjusting for age and time from PCI to infusion. Neither age nor time from PCI to cell infusion (days) had a significant relationship with the change in LVEF from baseline to six months in either study or the combined dataset as was the case in SWISS-AMI.⁴ These analyses suggest but do not prove that the greatest change in LVEF over the study period occurs in the cohort with the most severe baseline LV dysfunction.

Size of the study population

In view of the results, questions have been raised regarding the size and power of these studies. These three trials were powered to detect placebo-adjusted LVEF increases from 3.5–5%. Great variability and heterogeneity across clinical centers all but precludes identifying small effect sizes. However, we did not anticipate the small effect sizes that we observed. These miniscule effects were not presaged by the literature, which instead reported (e.g., REPAIR-AMI) much larger effects of cell therapy. Presuming these moderate to large effects would be discoverable in our trials, we focused on whether the timing of administration of cells would influence these effects. In addition, these effect sizes were, as the reviewer points out, beyond the ability of clinical centers to measure with requisite precision.

Additionally, smaller study sizes may result in incomplete randomization of baseline variables as was the case in these studies. Additionally, and perhaps unexpectedly, the randomization to different delivery times impacted the study populations. The intended delay between enrollment and delivery resulted in greater numbers of patients withdrawing from the studies from groups that received delivery at later time points. Thus, considerations balancing the costs of larger trials and the inherent uncertainty of smaller trials are critical to the field of cell therapy. Furthermore, in trials of AMI in which subjects need to return to the hospital following discharge, designs need to account for the possibility of withdrawal of subjects once discharged.

Randomization models

In the CCTRN studies, a 2:1 ratio of BMC-treated to placebo was utilized to balance the rigor of the study with the needs to recruit subjects to a trial that required BM harvest and invasive delivery of cells. However, in addition to reducing statistical power, it also generated a placebo group 50 percent smaller than that of an equal randomization model, creating inequalities in the baseline characteristics and increased variability due to small sample sizes in the placebo group. The CCTRN favors designs that include a placebo group as robust as possible for comparisons; in general, the Network endorses equal randomization as was done in SWISS-AMI.

Standardization of cell processing

To limit the potential sources of variability related to cell product, the CCTRN adopted distributed and automated cell processing (Sepax, Biosafe) at each of the five regional centers as opposed to centralized processing This decision was based on several lines of reasoning. First, use of open Ficoll systems for BMC isolation was becoming less standard in the U.S. with the advent of cell mobilization and isolation for clinical hematopoietic stem cell transplantation. Second, use of a closed-automated system enabled local preparation and standardization. Despite of the striking similarity to the negative results of SWISS-AMI, which used manual Ficoll processing, concerns have been raised about the functionality of the BMC product used in TIME and LateTIME. These included the potential effects of erythrocytes and heparin (ref 31) in the study product.¹⁸ We have presented data to suggest that these concerns are not warranted.^{19, 20} Notably, in the FOCUS-CCTRN trial, with Sepax processed cells, there was an observed increase in LVEF in the BMC group vs. placebo at six months in patients with chronic LV dysfunction.²¹

Additional studies performed by the Network compared the effects of human BM separated by Sepax or by manual Ficoll preparation in two distinct immunodeficient murine models, hind limb ischemia and myocardial infarction. Results of these studies indicate that Sepax and manual Ficoll-isolated cells resulted in similar effects in these complementary models (Supplemental Figure). While a number of differences exist that distinguish SWISS-AMI from TIME and LateTIME taken together, the data suggest that the negative results of TIME, LateTIME, and SWISS-AMI are more likely to be due to the inherent nature of BMCs than the means by which they are isolated or stored.

Timing of cell harvest and delivery

The timing of harvest and delivery in SWISS-AMI affected the content of the BM product. CD34 cell content was marginally higher at three to four weeks than at five-seven days following MI (%CD34: 1.31% late; 1.02% early, P=0.01) but this was not seen when comparing cells from TIME and LateTIME. These studies do not support a major impact of timing of cell harvest on the BMC product.

End point selection

Each study used cMR as the most rigorous method to assess LVEF while TIME and LateTIME used a co-primary endpoint of regional LV function. CCTRN followed the concept used in BOOST, employing wall motion in the infarct zone and border zone.¹² Precision for the assessment of regional LV function was substantially better than for global LV function, suggesting that the sample size for a two armed clinical trial is substantially smaller when designed around regional function rather than global LV function.

Future directions for cell therapy after AMI

In 2007, cell therapy clinical trials of BMCs in AMI were considered highly innovative with a growing safety profile and hopes for effects. In spite of the early positive studies including BOOST, which utilized cMR as the primary endpoint, a meta-analysis of studies using LVEF by cMR as the primary endpoint did not show a statistically significant effect of un fractionated BMCs on LVEF.¹⁶ Why, in aggregate, studies using echo endpoints demonstrate differences in LV function assessment while those using cMR do not, remains an unsettled question.

Early preclinical and clinical trial findings suggested BMC delivery could improve LV function. This, coupled with the strong public demand for new interventional strategies propelled this collection of clinical investigations. While the impact of BMCs on survival following AMI remains to be determined, there may be a future for incorporation of some aspects of bone marrow-derived cells with selected or enriched populations. However the promise of a major impact of BMCs on LV function appears unfounded. A potential effect on mortality will be examined in the prospective BAMI trial, a Phase 3 trial in Europe aimed to test the hypothesis that BMCs improve two year survival following AMI.²² The CCTRN eagerly awaits the results of BAMI as they will effectively make all of the ongoing discussion surrounding the effects of BMC following AMI on LVEF moot, as the study is purely designed to test the effects on mortality.

Meanwhile, concepts of “off-the shelf” cell delivery following AMI with an allogeneic cell (e.g., allogeneic MSCs²³) delivered at multiple doses and timing (even as early as reperfusion) are promising. Another alternative in which cells are delivered in the post-acute period following the initial phases of recovery and remodeling which would permit their intramyocardial effect to develop in the presence of stable LV function. This research pathway is lit by a collection of post MI LV dysfunction trials including FOCUS, POSEIDON, TAC-HFT, SCIPIO, CADUCEUS, and C-CURE^{21, 24–28} Although BMC use in AMI patients has dimmed (at least temporarily), the future of cell therapy for LV dysfunction resulting from AMI remains bright.

Concise summary.

To understand the role of bone marrow mononuclear cells (BMC) in the treatment of acute myocardial infarction (AMI), this overview offers a retrospective examination of strengths and limitations of three contemporaneous trials with attention to critical design features and provides an analysis of the combined dataset and implications for future directions in cell therapy for AMI.

Non-Standard Abbreviations and Acronyms

AMI

acute myocardial infarction

BMC

bone marrow mononuclear cells

CCTRN

Cardiovascular Cell Therapy Research Network

LV

left ventricular

LVEF

left ventricular ejection fraction

MR

magnetic resonance

MSC

mesenchymal stem cells

NHLBI

National Heart, Lung, and Blood Institute

PCI

percutaneous coronary intervention

SDF

stromal derived factor

STEMI

ST elevation myocardial infarction