The present study is a phase 2, randomised, double-blind, placebo-controlled multicentre trial. The participating centres were the Leiden University Medical Center (LUMC) and the University Medical Center of Utrecht (UMCU). The LUMC has been the coordinating centre that provided trial management and data analysis. The study protocol was in accordance with the declaration of Helsinki and complied with the Guideline for Good Clinical Practice (CMPP/ICH/135/95—17th July 1996). The protocol was approved by the institutional ethical committees of both research centres and the Dutch Central Committee on Research Involving Human Subjects (CCMO). Overall safety examination was performed by an independent data monitoring safety board (DSMB), as well as by independent institutional safety review boards of each clinical centre. The study has been registered at the Dutch trial registry (www.​trialregister.​nl, no. NTR2516).

The study population consisted of patients with coronary artery disease and chronic HF (NYHA class 2, 3 or 4) despite optimised medical therapy, and were recruited by the 2 participating centres. Full inclusion and exclusion criteria are provided in Tab. 1. A 1:1 randomisation was executed by a statistician of the LUMC (Fig. 1). The randomisation was stratified by presence of stress-inducible ischaemia, to assure a similar amount of patients with and without stress-inducible ischaemia in both treatment and placebo group, and by clinical centre, to equally assign patients who underwent cell-based therapy versus placebo to both hospitals.

At baseline, a complete medical history was obtained and laboratory tests including blood count, electrolytes, cardiac markers, hepatic and renal function, and infectious disease serology were performed. Further baseline evaluation consisted of stress-rest Tc-99m tetrofosmin gated single-photon emission tomography (SPECT), fluorodeoxyglucose (FDG) SPECT, metaiodobenzylguanidine (MIGB) imaging, NYHA classification, quality of life assessment using the Dutch translation of the Minnesota Living with Heart Failure Questionnaire (MLHF), and a bicycle exercise test with volume oxygen maximum (VO2 max) measurement. Next, patients were hospitalised for bone marrow aspiration and the intramyocardial injection procedure. To evaluate peri-procedural and short-term safety of bone marrow cell treatment, patients were admitted to the Coronary Care Unit immediately after treatment. Monitoring of vital signs, heart rhythm and biochemical cardiac markers were performed during 2 days after treatment. Before discharge, echocardiography was performed to evaluate the presence of pericardial effusion. A follow-up SPECT was repeated at 3 and 12 months. NYHA class was reassessed on follow-up at 3, 6 and 12 months, and MLHF and bicycle test were repeated at 3 and 6 months. We evaluated long-term safety during outpatient visits at 1.5, 3, 6 and 12 months after treatment to ensure the collection of all adverse events, including all-cause death, cardiovascular death, hospitalisation for worsening HF, rhythm disorders and myocardial infarction. A 24-hour Holter recording was obtained in patients without implantable cardioverter defibrillator (ICD) at 1.5 and 6 months to monitor the occurrence of arrhythmias.

On the day of the injection procedure, ≥80 ml of bone marrow was aspirated from the posterior iliac crest under local anaesthesia. During the procedure patients were under continuous ECG and oxygen saturation registration. Bone marrow was collected in flasks containing Hanks balance salt solution and heparin. Mononuclear bone marrow cells were isolated by Ficoll density gradient centrifugation according to strict GMP conditions following Standard Operating Procedures of the Stem Cell Laboratory at the LUMC (CST-WV-2014, CST-WR 2014, CST0PF-2000.BM/MNC and CST-Do-2000). Bone marrow cells were suspended at a concentration of 40 × 10⁶ cells/ml in a solution of 0.9% sodium chloride, 0.5% human albumin and 11.3 IU/ml heparin [9]. The final cell product contained 100 × 10⁶ cells. The bone marrow cell population was checked for clots, stained for the presence of bacteria and analysed by fluorescence-activated cell sorting for the presence and percentage of CD14-positive, CD34-positive and CD45-positive cells [9].

A blinded syringe with either bone marrow solution or placebo solution, which contained a similar sodium chloride/human albumin solution without bone marrow-derived cells, was delivered at the catheterisation laboratory. Patient’s allocation was only known to the stem cell laboratory assistant.

During cell preparation, biplane left ventricular angiography was performed. Based on ventricular size, the mapping catheter curve (D or F) was selected. Via femoral artery access and a retrograde aortic approach, the catheter was inserted into the left ventricle. A non-fluoroscopic electromechanical map of the left ventricle was constructed using the NOGA system (NOGAStar catheter, Biosense-Webster, Waterloo, Belgium) [19]. The electromechanical map was used to guide the injection catheter with a 27-gauge retractable needle (MyoStar catheter Biosense-Webster) to the target region, which is the area of stress-inducible ischaemia on SPECT for the patients with ischaemia [19], and to the peri-infarction zone for patients without ischaemia [20]. Subsequently, 8–12 injections of approximately 0.2–0.3 ml each were delivered.

Gated SPECT imaging was performed with a triple-head SPECT camera system (GCA 9300/HG, Toshiba Corp., Tokyo, Japan). A 2-day stress-rest protocol was used for Tc-99m tetrofosmin SPECT imaging. Since the infusion of adenosine should be continued for 2–3 minutes after injection of the radiopharmaceutical, adenosine was given as a continuous infusion (0.14 mg/kg/min) for 6 minutes with injection of Tc-99m tetrofosmin (500 MBq) at 3 minutes after the start of the adenosine infusion. Stress imaging was performed 45 minutes after injection. On the second day, rest images were obtained 1 hour after injection of Tc-99m tetrofosmin (500 MBq). To obtain left ventricular ejection fraction (LVEF) and left ventricular volumes, rest imaging studies were acquired using ECG gating [9]. We carried out a quantitative assessment of left ventricular end-systolic and end-diastolic volumes and LVEF at rest and stress using the 4DM software (Corridor 4DM, INVIA, Ann Arbor, University of Michigan Medical Center) [21]. To analyse myocardial perfusion, the gated SPECT images were divided in 17 segments and we categorised segmental tracer uptake, at rest and stress imaging, on a 4 point scale: 1 = tracer activity >75%; 2 = tracer activity 50–75%; 3 = tracer activity 25–49%; 4 = tracer activity <25%. When segmental tracer uptake during stress was >75% of maximum tracer uptake, perfusion was considered to be normal [22]. The summed stress score was calculated by summation of the segmental scores at stress and summed rest score was calculated by summation of the segmental scores at rest. The summed differences score was calculated by summation of the differences in stress and rest segmental scores and reflects the extent of stress-inducible ischaemia. A summed difference score of 1 was defined as presence of stress-inducible ischaemia.

FDG imaging was performed to assess myocardial viability [23]. The plasma glucose level was assessed and regulated by an oral dose of 500 mg acipimox and 45 minutes later a low-fat carbohydrate-rich meal [23]. FDG (185 MBq) was injected two hours thereafter, tracer uptake imaging was performed after 45 minutes of rest using the dual-head Toshiba, GCA 7200A/DI ultra high-energy collimator. Myocardial tracer uptake was scored on a 17 segment model and scored on a 4-point scale: 1 = normal tracer uptake (activity >75%); 2 = mildly reduced tracer uptake (activity 50–75%); 3 = moderately reduced tracer uptake (activity 25–49%); 4 = severely reduced tracer uptake (activity <25%). Segments with >75% of maximum tracer uptake were considered as viable (normal) and segments with tracer uptake <75% were considered to contain some extent of scar. An increase in segmental uptake by 1 point was considered an improvement [23].

¹²³I-MIBG imaging was performed to assess myocardial (sympathetic) innervation. Patients were pre-treated with sodium iodide or potassium iodide 1 hour prior to injection to block uptake of free ¹²³Iodine by the thyroid gland. 185 MBq of ¹²³I-MIBG was injected and ¹²³I-MIBG planar imaging was performed in the supine position after 15 minutes and 4 hours. The heart-to-mediastinum ratio was calculated for the late planar images [24].

Exercise capacity was assessed by a bicycle exercise test. Patients performed a symptom-limited bicycle exercise test with a 20W starting load and an increment of 10 W/min, including VO2max measurement. Test endpoints were angina pectoris, physical exhaustion, dyspnoea and significant decrease in systolic blood pressure (>10 mm Hg), as measured every 2 minutes [9].

Primary endpoint was defined as the difference in change of LVEF, determined by SPECT, between baseline and follow-up. A prespecified composite endpoint consisted of LVEF, summed stress score (SPECT), NYHA class, MLHF score, exercise capacity, VO2 max, 6‑minute walking distance, viability (FDG SPECT) and heart-to-mediastinum ratio (MIBG). Improvement per patient was quantified on a scale ranging from −9 (deterioration on every test) to +9 (improvement on every test), baseline compared to 3‑month follow-up. A subanalysis was performed to compare the effect of cell-based therapy in patients with and without stress-inducible ischaemia.

Serious adverse events (SAEs) and suspected unexpected serious adverse reactions (SUSARs) were reported to the CCMO, as well as to the DSMB. The DSMB evaluated whether adverse events in the patients included in the study were related to any of the diagnostic or therapeutic procedures used in the protocol. Furthermore, an annual safety report was submitted to the CCMO and DSMB. This provided an overview of all SUSARs and SAEs, accompanied by a brief report, highlighting the main points of concern.

During 12-month follow-up, 2 cell-treated patients and 3 placebo-treated patients died. One of the cell-treated patients died 2.5 months after the injections. The other deceased cell-treated patient requested euthanasia at 7.5 months after the procedure because of incurable suffering from chronic pain that was not related to his cardiac disease. Of the placebo-treated patients, 1 patient died from complication of acute myocardial ischaemia 7 months after the procedure, 1 patient died following an out-of-hospital cardiac arrest 6.5 months after the procedure, and the other patient died from cardiorespiratory insufficiency due to terminal HF 3.5 months after the procedure.

In 2 cell-treated and 2 placebo-treated patients a serious adverse event resulted in medical intervention and exclusion of further follow-up. Both cell-treated patients underwent cardiac resynchronisation therapy (CRT). In 1 of these patients, at 6‑month follow-up after a high-grade atrioventricular block and pacemaker dependency, a CRT was indicated and in the other patient at 10-month follow-up, a CRT was indicated because of progressive intraventricular conduction delay. One placebo-treated patient had a non-ST-elevation myocardial infarction 9 months after the procedure and underwent coronary artery bypass grafting. The other placebo-treated patient was admitted at 7‑month follow-up because of decompensated HF and underwent minimal invasive mitral valve repair.

Other SAEs included a monomorphic ventricular tachycardia in a cell-treated patient at 1.5 months, which was stabilised with amiodarone, and a near collapse due to a non-sustained ventricular tachycardia in a placebo-treated patient at 6.5 months. All SAEs are summarised in supplementary Tab. 1.

Mean procedural time was 112 ± 40 minutes in the bone marrow cell group and 110 ± 44 minutes in the placebo group (P = 0.87). In both groups, 19 patients received intramyocardial injections. Patients in the bone marrow cell group received 9.5 ± 0.8 injections and patients in the placebo group received 9.1 ± 1.3 injections (P = 0.29). All patients from the cell group received 100X10⁶ cells, with a CD34+ fraction of 1.6% ± 0.6%.

At baseline all 39 patients underwent SPECT. At 3 months paired SPECT studies were available in 18 patients from both treatment groups due to previously described deaths (N = 2) and unsuccessful injection procedure (N = 1). At 12-month follow-up, SPECT scans of 14 cell-treated patients and 13 placebo-treated patients were available (missing scans due to previously described reasons, 1 scan unavailable due to logistics reasons).

In the cell-treated group no improvement in LVEF at rest, assessed by SPECT, was detected (P = 0.61 at 3‑month and P = 0.81 at 12-month follow-up). Also in the placebo group, at 3‑month follow-up LVEF at rest did not change (P = 0.62). In the placebo group at 12-month follow-up, there was a significant increase in LVEF (P = 0.01). However, we did not detect any significant differences (LVEF change in percentage points from baseline) between both groups at 3‑month and 12-month follow-up (respectively, P = 0.47 and P = 0.08). We also did not observe any differences between groups in LVEF at stress, left ventricular end-systolic volume and left ventricular end-diastolic volume at 3‑month and 12-month follow-up, as shown in Tab. 3 and Fig. 2.

Summed stress score did not significantly change in the cell group (P = 0.20 at 3 months and P = 0.86 at 12 months follow-up). In the placebo group at 3‑month follow-up, there was a small improvement in summed stress score (P = 0.03). However this improvement did not sustain up to 12 months (P = 0.21) and was not significantly different compared with the cell group (P = 0.48 at 3‑month and P = 0.45 at 12-month follow-up).

There were no changes in summed rest score in the cell group (P = 0.89 at 3‑month and P = 0.87 at 12-month follow-up) or in the placebo group (P = 0.25 at 3‑month and P = 0.56 at 12-month follow-up). This was comparable between both groups (P = 0.33 at 3‑month and P = 0.59 at 12-month follow-up).

There was a small but significant improvement in summed difference score in the cell group at 3‑month follow-up (P = 0.05), which did not sustain up to 12-month follow-up (P = 0.48). In the placebo group there was also an improvement at 3 months, which sustained up to 12-month follow-up (respectively, P < 0.01 and P = 0.03). Importantly, no significant differences between both groups were detected (respectively, P = 0.25 and P = 0.24). Data are shown in Tab. 3.

Baseline F18-FDG SPECT images were available for all cell-treated and 17 placebo-treated-patients and for 16 of both cell-treated and placebo-treated patients at 3‑month follow-up. In the cell group at 3‑month follow-up, there was a small but significant improvement in viability score (−1.2 [95% CI −2.4–−0.1]; P = 0.05), which was not seen after placebo treatment (−0.2 [95% CI −1.4–1.0]; P = 0.72). However, no significant treatment effect was observed (group difference −1.0 [95% CI −2.6–0.6]; P = 0.22).

All patients underwent baseline 123-I-MIBG SPECT imaging and at 3‑month follow-up, 18 scans of both groups were available for analysis. At 3‑month follow-up, late heart-to-mediastinum ratio remained unchanged in both cell group and placebo group (estimated mean difference −0.01 [95% CI −0.1–0.1]; P = 0.77 and −0.01 [95% CI −0.1–0.1]; P = 0.75 respectively, group difference 0.001 [95% CI −0.1–0.1]; P = 0.99).

Clinical status, assessed by NYHA score, did not change in both treatment groups on follow-up at 3, 6 and 12 months (group difference P = 0.73, P = 0.13 and P = 0.52, respectively).

Improvement in quality of life MLHF score after 3 months in the cell group was not significant (estimated mean difference −5.7 [95% CI −14.5–3.0]; P = 0.19). However, at 6 months there was a trend to improvement (estimated mean difference −7.2 [95% CI −14.4–0.01]; P = 0.05). In the placebo group, improvement in quality of life score was not significant (estimated mean difference at 3 months −5.8 [95% CI −15.6–4.0]; P = 0.24 and at 6 months −2.2 [95% CI −11.1–6.7]; P = 0.61). Nevertheless, between-group differences were not significant (P = 0.99 at 3‑month and P = 0.38 at 6‑month follow-up). Data are presented in supplementary Tab. 2 and Fig. 3.

There were no significant changes in exercise capacity assessed by bicycle test, after cell or placebo treatment. Both groups were comparable at 3 and 6 months (P = 0.51 and P = 0.36, respectively). Similarly, we did not find any significant changes in VO2max during the bicycle test (group differences at 3 months P = 0.82 and at 6 months P = 0.42). Data are presented in supplementary Tab. 2.

Prespecified subgroup analysis examined the efficacy of cell-based therapy in patients with stress-inducible ischaemia. At baseline, in the cell group 8 patient had ischaemia (mean summed difference score 3.1 ± 2.4) and in the placebo group 13 patients had ischaemia (mean summed difference score 4.2 ± 3.0) Therefore, 21 patients were included in the analysis. Baseline LVEF was 29.6 ± 11.5% in the cell group and 29.7 ± 11.5 in the placebo group. Estimated mean difference in the cell group was −0.3% (95% CI −5.0–4.4%); P = 0.91 at 3‑month follow-up and −0.5% (95% CI −2.9–1.8%); P = 0.64 at 12-month follow-up. In the placebo group estimated mean differences were 1.5% (95% CI −2.1–5.1%); P = 0.39 at 3‑month follow-up and 1.7% (95% CI −0.5–3.9%); P = 0.12 at 12-month follow-up. Between-group difference at 3 months was −1.8% (95% CI −7.7–4.1%); P = 0.54 at 3‑month follow-up and −2.2% (95% CI −5.4–0.9%); P = 0.15 at 12-month follow-up. Also, no significant changes were observed between ischaemic cell-treated and ischaemic placebo-treated patients in secondary endpoints (data not shown).