WHAT IS ALREADY KNOWN ON THIS TOPIC

In previous studies, we showed that early administration of granulocyte colony-stimulating factor (G-CSF) reduces adverse left ventricular (LV) remodelling and function and infarct size in successfully reperfused patients with large ST-elevation myocardial infarction (STEMI). Whether G-CSF has beneficial effects on long-term outcomes remains unexplored.

WHAT THIS STUDY ADDS

This randomised clinical trial was prematurely terminated after enrolling 532 patients and can no longer be considered adequate to test the hypothesis. In post hoc analyses, we found a decrease in the primary composite outcome in G-CSF-treated patients with low bone marrow (BM) cell mobilisation or with severe LV systolic dysfunction at discharge.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICYIntroduction

Despite the widespread application of timely revascularisation through percutaneous coronary intervention (PCI), the incidence of heart failure (HF) in patients with ST-elevation myocardial infarction (STEMI) due to left ventricular (LV) remodelling is still substantial.1 To further reduce myonecrosis extent, several therapeutic approaches have been proposed, the vast majority of which have failed.2 Among such attempts, cell-based repair therapies have been studied over the past 15 years without providing definitive evidence of clinical benefit.3 A key concern has been the lack of well-designed phase III trials. Recently, the results of a large phase III clinical trial (Bone marrow cell therapy in Acute Myocardial Infarction - BAMI) on intracoronary delivery of autologous bone marrow-derived mononuclear cells (BM-MNCs) in patients with STEMI with LV depression after primary PCI were published.4 Unfortunately, BAMI did not qualify to evaluate the efficacy of the administration of BM-MNCs on all-cause mortality because of low recruitment and event rates.

In recent years, we tested the alternative hypothesis that early administration of granulocyte colony-stimulating factor (G-CSF) in patients with large STEMI has a favourable impact on adverse LV remodelling5 and long-term outcomes.6 G-CSF is an endogenous haematopoietic cytokine that mobilises granulocytes, stem and progenitor cells from the bone marrow (BM) into the bloodstream.7 In addition, G-CSF directly activates prosurvival signalling in ischaemic cardiomyocytes.8

Prompted by previous evidence, we designed the nationwide ‘STEM cell mobilization in Acute Myocardial Infarction OUTCOME’ (STEM-AMI OUTCOME) trial to verify whether early therapy with G-CSF in patients with large STEMI may favourably impact long-term outcome by decreasing cardiac mortality and morbidity.9 As surrogate endpoints, we have already provided solid evidence in an adequately powered cardiac magnetic resonance (CMR) substudy that G-CSF is beneficial for LV remodelling and function, infarct size and myocardial strain.10

MethodsStudy design and participants

The STEM-AMI OUTCOME was a multicentre, randomised, controlled, open-label, phase III trial conducted in 44 centres in Italy (ClinicalTrials.gov ID: NCT01969890). The study was conducted by the Declaration of Helsinki. The trial was designed to test the hypothesis that G-CSF on top of standard-of-care (SOC) therapy might favourably affect the long-term outcome of patients who experienced a large STEMI.9 The primary endpoint was the reduction in a composite outcome including (1) all-cause death, (2) recurrence of myocardial infarction (MI) or (3) hospitalisation due to HF. The secondary endpoints were: (1) coronary revascularisation, (2) cardiovascular death, (3) fatal and non-fatal stroke, (4) hospitalisation for any cause, (5) cardiovascular hospitalisation, (6) resuscitation and (7) automated implantable cardioverter-defibrillator therapy. Patients were randomised to G-CSF or SOC in a 1:1 ratio using algorithmically generated intercentre and intracentre balanced blocks and a web-based central allocation system. Patients randomised to treatment received, within 24 hours after reperfusion, 5 μg/kg G-CSF (Filgrastim, Hexal, Holzkirchen, Germany) subcutaneously two times per day for six consecutive days. In line with previous reports,11 12 G-CSF administration was discontinued if the white cell count rose above 50×109/L.

Patients were monitored for 2 years with clinical evaluations, ECG, blood tests and 2D echocardiography. Visits were scheduled after 1 month, 6 months, 1 year and 2 years for all participants discharged alive from the hospital. Clinical and adverse events were recorded.

Study outcome

The initial study design was conceived assuming that treatment with G-CSF would lead to a 25% reduction in 2-year primary endpoint events. Based on Italian registries,13 14 we had conservatively estimated a major adverse cardiac event rate of 15% in the control cohort. The sample size calculation was made under the following assumptions: event rate of 15% in the control group, 11.25% in the treatment group; exponential distribution; power, 80%; α, 0.05; two-sided test; accrual time, 3 years; follow-up time, 2 years. A total of 1530 patients (ie, 765 per group) were planned to be enrolled.9

The initial population included patients affected by anterior STEMI with an LV ejection fraction (LVEF) ≤45% after reperfusion and a symptom-to-balloon time ≥2 and ≤12 hours. The study was subsequently amended to increase the enrolment rate, including all patients with STEMI presenting with an LVEF <45%, regardless of symptom-to-balloon time.

Despite these changes, due to a markedly reduced number of patients with STEMI with LV depression after PCI, reduced mortality and morbidity in the target population, and limited funding for study extension, the data and safety monitoring board deemed it futile to continue enrolment. Therefore, the trial was stopped in February 2016 at patient 532. The sponsors approved this decision.

The 2-year follow-up was completed as per protocol. Endpoints were adjudicated by an independent clinical event committee blinded to patient treatment allocation.

Statistical analysis

Descriptive statistics included proportions for categorical variables and mean (SD) or median (IQR) for continuous variables, based on data distribution skewness. There were little missing data (online supplemental table 1) and no assumptions were made.

A modified intention-to-treat analysis was conducted, stratifying randomised patients by treatment arm. Subjects were analysed according to the initially assigned group, excluding those for whom the study stopped before the treatment and whose outcome data were not collected. Statistical differences in proportions between groups were assessed with the χ2 test or Fisher’s exact test, depending on the number of expected cases. The Shapiro-Wilk test was used to assess normality in data distribution. Accordingly, continuous variables were compared using the t-test or Wilcoxon rank-sum test.

We performed Kaplan-Meier analyses to estimate the likelihood of occurrence of outcomes within 2 years after enrolment, using the log-rank test to compare survival curves between groups. Cox proportional hazards models were applied to assess the association between the first occurrence of clinical events at 2 years and treatment. The effect of G-CSF treatment was estimated by the HR using the SOC group as the reference category. In addition, we applied a generalisation of the Cox model (Andersen-Gill model) to assess the impact of treatment on the frequency of recurrent events. HR was reported with a 95% CI.

In post hoc analyses, differences between G-CSF patients stratified according to a cut-off value in white cell count after the last infusion were assessed using the same approaches. We also investigated the relationship between treatment and clinical outcomes by running separate logistic regression models for each echocardiographic parameter (LVEF; indexed LV end-systolic volume (LVESVI); indexed LV end-diastolic volume (LVEDVI)) at baseline and discharge. These models included an interaction term to examine whether the effect of treatment varied according to the parameter values.

All p values were two sided, with p<0.05 considered statistically significant. Analyses were performed with R V.3.5.2 (R Project for Statistical Computing, http://www.R-project.org) and SAS software V.9.4 (SAS Institute).

ResultsPatient characteristics

Centres participating in the study recruited a total of 532 consecutive individuals between November 2013 and February 2016 and, after eligibility assessment, randomly assigned 525 patients to G-CSF or SOC (figure 1). Four patients (three in the G-CSF and one in the control group) were excluded from the study before receiving treatment (due to documented malignancies, logistics or technical problems uncovered after allocation) and no outcome data were collected. Therefore, 521 patients (260 G-CSF and 261 SOC treated) were included in the modified intention-to-treat analysis.

Figure 1

Consolidated Standards of Reporting Trials (CONSORT) diagram for patient flow in the STEM-AMI OUTCOME trial. AMI, acute myocardial infarction; EF, ejection fraction; G-CSF, granulocyte colony-stimulating factor; PCI, percutaneous coronary intervention; TIMI, thrombolysis in myocardial infarction.

Clinical and acute myocardial infarction (AMI)-related features at baseline are listed in table 1. The two groups were similar in clinical characteristics, cardiovascular risk factors and medical therapy at admission, apart from a higher prevalence of patients with diabetes in the SOC group and chronic obstructive pulmonary disease in the G-CSF group. 429 patients (82.3%) had anterior MI. LVEF at enrolment was comparable in the two groups (38.5±5.2% vs 38.9±4.8% in the G-CSF and SOC groups, respectively). The G-CSF group exhibited a significant trend for a more severe AMI than the SOC group, as indexed by troponin peak levels, the number of ECG leads involved and LVESVI values, as well as non-significant trends for worse Killip score and longer symptom-to-balloon time.

Table 1

Baseline characteristics and myocardial infarction-related parameters of the study population stratified by treatment

Patients in the treatment arm were administered G-CSF within a median of 14.63 hours after PCI, with a median of 9 (6–11) doses. According to the study protocol, treatment was stopped prematurely in 218 patients who reached a white cell count of 50 cells ×109/L, presented with side effects, had technical or personal problems or by physician’s choice (online supplemental table 2). These patients received a median of 7 (5–10) G-CSF administrations. No differences were observed in drug treatment at discharge between the two groups, but patients treated with G-CSF had a significantly higher mean LVEDVI than those treated with SOC (online supplemental table 3).

Safety

No patients withdrew from the study because of G-CSF adverse effects, although four patients discontinued treatment prematurely because of side effects (online supplemental table 2). After 2 years, there were no significant differences between the two groups for any adverse events or serious adverse events (n=65 and n=58 in the G-CSF and SOC groups, respectively; OR 1.17; 95% CI 0.78 to 1.75; p=0.45; figure 2). We observed one patient with a malignant gastrointestinal tumour and two patients with lung cancer, all in the SOC group. Four events of malignant arrhythmia were observed during the follow-up period, with no differences between the two groups.

Figure 2

Safety endpoints in the STEM-AMI OUTCOME trial. The Kaplan-Meier curves illustrate estimated probabilities for (A) any adverse event, (B) serious adverse events, (C) malignancies and (D) malignant arrhythmias up to 2-year follow-up in the two treatment groups. The log-rank test was used to compare survival curves between groups. G-CSF, granulocyte colony-stimulating factor; SOC, standard of care.

Clinical outcomes

Control and G-CSF patients received the SOC pharmacological treatment during the entire follow-up, without appreciable differences between the two groups (data of 475 survivors who completed the 2-year follow-up; online supplemental table 4).

Kaplan-Meier analysis illustrates the temporal occurrence of the composite primary and secondary outcomes within 2 years from enrolment (figure 3A,B, respectively). A comparison of the survival curves did not show significant differences between the two groups for both endpoints.

Figure 3

Clinical composite outcomes in the STEM-AMI OUTCOME trial. Survival analysis showing composite primary and secondary outcomes in the study population stratified by treatment and/or white blood cell mobilisation after G-CSF. The plots show the Kaplan-Meier curves for (A) the primary endpoint and (B) the secondary endpoint over the 2 years of follow-up for the two treatment groups overall, and (C) the primary endpoint and (D) the secondary endpoint in the SOC group versus low and high mobilisers (defined as patients with white cell count ≤50×109/L and >50×109/L after the last infusion, respectively). The log-rank test was used to compare survival curves between groups. G-CSF, granulocyte colony-stimulating factor; SOC, standard of care.

Table 2 presents the primary and secondary outcomes registered after hospital discharge in both patient groups during the 2-year follow-up. Mortality was 2.31% (6 patients) in G-CSF and 2.68% (7 patients) in the control group (HR 0.88; 95% CI 0.29 to 2.60; p=0.81). There were no significant differences in either primary (HR 1.20; 95% CI 0.63 to 2.28; p=0.59) or secondary composite endpoints (HR 1.13; 95% CI 0.79 to 1.62; p=0.50), individual clinical outcomes or major adverse cardiac and cerebrovascular events (MACCE, defined as all-cause death, reinfarction, acute HF, urgent revascularisation and stroke) between the two groups. The number of patients with recurrence of primary outcomes was higher in the SOC group (p=0.015, table 2), but the probability for recurrent events was not significant (online supplemental table 5).

Table 2

Events occurred during the follow-up in the study population stratified by treatment

Clinical outcomes and response to G-CSF

In a post hoc analysis, we sought to determine whether the clinical outcome might depend on the response to the drug, that is, the extent of cell mobilisation. We stratified patients according to blood cell count after the last infusion into high mobilisers (white cell count >50×109/L) and low mobilisers (white cell count ≤50×109/L). MI severity or damage/inflammatory indices at baseline were not different between the two groups, except for a 9% higher white blood cell number in the high mobilisers (online supplemental table 6).

Kaplan-Meier curves showed a lower trend in the cumulative incidence of primary and secondary clinical events in low mobilisers than in high-mobilisation patients (figure 3C,D, respectively). Specifically, we observed a lower trend in the composite primary outcome in patients with low mobilisation (HR 2.86; 95% CI 0.96 to 8.56; p=0.06) with a fourfold lower total number of patients with events (p=0.042; table 3). Secondary outcomes also appeared cumulatively lower in low mobilisers than in high mobilisers (28 vs 69, respectively). The analysis of recurrent events suggested that the probability of composite secondary outcomes was significantly higher in high mobilisers (HR 1.62; 95% CI 1.03 to 2.54; p=0.037; online supplemental table 5).

Table 3

Events during the follow-up in the study population stratified by WBC mobilisation peak after G-CSF infusion

Clinical outcomes and echocardiographic parameters

In another post hoc analysis, we then investigated whether cardiac function at baseline or at discharge, as measured by echocardiographic parameters, influenced treatment response in terms of clinical outcomes. To do so, we used logistic regression models with an interaction term between the treatment effect and the echocardiographic variable. The results (table 4) show that there is no effect of baseline cardiac function parameters. In contrast, the interaction effect between G-CSF and LVESVI at discharge is associated with a decrease in both the primary composite outcome (β±SE, −0.08±0.04; p=0.034) and MACCE (β±SE, −0.07±0.03; p=0.024). Treatment effects in relation to each echocardiographic parameter and estimated OR are illustrated in online supplemental figures 1 and 2.

Table 4

Logistic regression models for assessing treatment effects on composite primary outcome and MACCE according to left ventricular parameters at baseline and discharge

We found no correlation between BM cell mobilisation and cardiac volumes at discharge (not shown).

Discussion

Since it was discontinued due to low recruitment and event rates, the STEM-AMI OUTCOME trial is not suitable to evaluate the efficacy of early administration of G-CSF on all-cause mortality and heart-related morbidity in successfully reperfused patients with large STEMI. Consequently, the data presented in this paper are descriptive. However, to the best of our knowledge, this is the largest study investigating the use of G-CSF in STEMI and the largest clinical trial that falls under the broad definition of cardiovascular repair medicine, even compared with the BAMI study.4 Therefore, this study provided insights into the use of a growth factor approach in the treatment of AMI, as well as valuable information on conducting a large nationwide phase III study of STEMI with a high likelihood of adverse LV remodelling.

Similar to the BAMI trial, we observed very low mortality and morbidity rates at follow-up (the occurrence of the primary endpoint was 7.1% at 2 years), confirming that in the scenario of primary PCI the SOC available in a developed country for reperfused patients with STEMI and LV dysfunction is highly effective and associated with progressive improvement in survival.15 As a consequence, the assumptions we considered based on existing registries when we designed the trial 9 years ago (event rate of 15% in controls and reduction of the primary endpoint by 25% after 2 years) are no longer applicable. Nevertheless, we believe that insights arising from the follow-up data are worth being reported.

We found an interaction between LVESVI at discharge and treatment associated with a decreased risk of both primary outcomes and MACCE in the G-CSF group. These results suggest that the greater the level of myocardial damage, the greater the likelihood of G-CSF benefit. These clinical findings parallel our previous data in the STEM-AMI trial5 6 and the STEM-AMI OUTCOME CMR substudy,10 in which the beneficial effects of G-CSF on adverse remodelling appeared to be most evident in the presence of dysfunctional myocardium at late reperfusion times. A similar observation was reported in the CMR imaging substudy of the REPAIR-AMI trial on intracoronary administration of autologous BM progenitor cells in STEMI16: a reduced LVEF at baseline was associated with significant improvement at follow-up. These findings suggest that G-CSF therapy for STEMI could be reserved for patients with systolic dysfunction and a higher probability of adverse remodelling.5 10

Mechanistically, G-CSF has a direct protective action on cardiomyocytes8 activating prosurvival pathways mediated by early sensitisation of G-CSF receptors in cardiomyocytes.17 This may explain why the early mobilisation scheme we adopted is crucial in counteracting delayed LV remodelling,5 10 as previously shown in a preclinical porcine model of MI.18 In addition, data supporting mobilisation-dependent effects have been progressively reported, whereby both adaptive and immune compartments are involved in infarct healing.19–21 Inflammatory/immune responses to acute ischaemic injury involve both culprit and remote myocardium.22 23 BM mobilisation exerted by G-CSF may boost a cascade of physiological events in the myocardial healing process that may explain the long-term beneficial effects of G-CSF on adverse remodelling we observed in patients with extensive STEMI.10

The results of the STEM-AMI OUTCOME study can serve as a proof of concept of growth factor-mediated regulation of BM mobilisation after STEMI. Our data suggest that there may be differences between low mobilisers and high mobilisers in primary outcome rates at 2-year follow-up. This observation is hard to frame within the concept that leucocytosis after MI is a negative prognostic marker.24 25 More recent insights have clarified that the early inflammatory phase after AMI, with the recruitment of neutrophils and monocytes, is followed on day 4 by phenotypic changes of immune cells supporting the myocardial healing process.26 Since the peak mobilisation of CD34+ cells by G-CSF falls within this transition,27 we speculate that G-CSF potentiates the physiological healing functions of the granulation phase mediated by the haematopoietic reparative response,28 including angiogenesis and reduction of myocardial fibrosis.29 These results pave the way for further investigation on tailored G-CSF administration regimens in STEMI through dose-finding studies and/or predictors of the degree of BM cell mobilisation to identify in advance who might benefit most from treatment.

Overall, our results complement those of the RIGENERA study,30 which reported after a 10-year follow-up in 32 patients with severe AMI that G-CSF treatment was safe and associated with reduced adverse LV remodelling and increased quality of life. Although inconclusive regarding the primary endpoint, the STEM-AMI OUTCOME trial definitively confirmed the safety of this approach with the largest sample size reported to date over a 2-year follow-up and provided additional information on which high-risk patients with adequately mobilised BM may undergo G-CSF-mediated amplification of the physiological immune healing process after STEMI.

Study limitations

As reported above, the trial was discontinued due to low recruitment and event rates, and as such it can no longer qualify as a hypothesis testing trial, but as an observational study. This precluded adequately powered group comparisons to assess the efficacy of early G-CSF administration in patients with large STEMI after successful reperfusion. Moreover, potential selection bias introduced by analysing post-baseline variables limits the generalisability of our findings. Finally, the trial had an open-label design: intensive care unit staff and patients could not be effectively blinded to the cytokine-induced increase in white cell count during treatment with G-CSF. However, although a blinded design would have been preferable, we expected that the open-label design would not have a substantial effect on the study results, given the nature of the primary clinical endpoints. To prevent bias in data analysis, study variables were collected and evaluated blindly by independent data managers and biostatisticians.

Conclusions

Like the BAMI study,4 the STEM-AMI OUTCOME trial is not adequate to evaluate the efficacy of early G-CSF administration on all-cause mortality and cardiac morbidity in successfully reperfused patients with STEMI. However, the results of this study provide conclusive information on the long-term safety of G-CSF in this setting and add novel insights into the correlation between treatment efficacy, patient profile and haematopoietic organ response to mobilisation.

Data availability statement

Data are available upon reasonable request. Requests for deidentified participant data should be made to FA or GP. Requests for further analyses to support ancillary publications will be submitted to the steering committee of the study for review and approval.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

This study involves human participants and was approved by the Ethics Committee of the ‘Azienda Ospedaliera della Provincia di Lecco’, as the centre promoting the trial (registration number 39/2013), on 8 May 2013. Subsequently, the Italian regulating authority (Italian Medicines Agency) and all local ethics committees in turn approved the protocol. Participants gave informed consent to participate in the study before taking part.

Acknowledgments

We thank the invaluable support of (the late) Attilio Maseri and Maurizio C Capogrossi. Without their vision and encouragement, this study would not have been possible. Furthermore, the authors express their sincere appreciation to all members of the STEM-AMI OUTCOME Trial Investigators team for their time and commitment in providing and caring for the study patients.