OUP user menu

Prospective randomized trial of direct endomyocardial implantation of bone marrow cells for treatment of severe coronary artery diseases (PROTECT-CAD trial)

Hung-Fat Tse , Sukumaran Thambar , Yok-Lam Kwong , Philip Rowlings , Greg Bellamy , Jane McCrohon , Paul Thomas , Bruce Bastian , John K.F. Chan , Gladys Lo , Chi-Lai Ho , Wing-Sze Chan , Raymond Y. Kwong , Anthony Parker , Thomas H. Hauser , Jenny Chan , Daniel Y.T. Fong , Chu-Pak Lau
DOI: http://dx.doi.org/10.1093/eurheartj/ehm485 2998-3005 First published online: 5 November 2007


Aims Experimental studies have demonstrated that bone marrow (BM) cells can induce angiogenesis in ischaemic myocardium. Recently, several non-randomized pilot studies have also suggested that direct BM cells implantation appears to be feasible and safe in patients with severe coronary artery diseases (CAD).

Methods and results We performed a randomized, blinded, and placebo-controlled trial in 28 CAD patients. After BM harvesting, we assigned patients to receive low dose (1 × 106 cells/0.1 mL, n = 9), high dose (2 × 106 cells/0.1 mL, n = 10) autologous BM cells or control (0.1 mL autologous plasma/injection, n = 9) catheter-based direct endomyocardial injection as guided by electromechanical mapping. Our primary endpoint was the increase in exercise treadmill time and our secondary endpoints were changes in Canadian Cardiovascular Society (CCS) and New York Heart Association (NYHA) class, and myocardial perfusion and left ventricular ejection fraction (LVEF) assessed by single-photon emission computed tomography and magnetic resonance imaging, respectively. A total 422 injections (mean 14.6 ± 0.7 per patient) were successfully performed at 41 targeted ischaemic regions without any acute complication. Baseline exercise treadmill time was 439 ± 182 s in controls and 393 ± 136 s in BM-treated patients, and changed after 6 months to 383 ± 223s and 464 ± 196 s [BM treatment effect +0.43 log seconds (+53%), 95% CI 0.11–0.74, P = 0.014]. Compared with placebo injection, BM implantation was associated with a significant increase in LVEF (BM treatment effect +5.4%, 95% CI 0.4–10.3, P = 0.044) and a lower NYHA class (odds ratio for treatment effect 0.12, 95% CI 0.02–0.73, P = 0.021) after 6 months, but CCS reduced similarly in both groups. We observed no acute or long-term complications, including ventricular arrhythmia, myocardial damage, or development of intramyocardial tumour or calcification associated with BM implantation.

Conclusion Direct endomyocardial implantation of autologous BM cells significantly improved exercise time, LVEF, and NYHA functional class in patients with severe CAD who failed conventional therapy.

  • Angiogenesis
  • Bone marrow
  • Coronary artery disease


Coronary artery disease (CAD) remains a major cause of morbidity and mortality worldwide. Recent advances in medical therapy and coronary revascularization techniques with either percutaneous or surgical procedures have improved the prognosis and survival in patients with CAD. However, with improvement in survival, the number of CAD patients who have exhausted or failed, or are not suitable candidates for coronary revascularization, is increasing.1,2 These patients with severe CAD follow an inexorable downhill course with disabling symptoms and progressive cardiac failure, despite maximal medical therapy.3 In recent years, therapeutic use of bone marrow (BM)-derived or circulating progenitor cells have been investigated as a potential novel therapy in patients with CAD, who are not amenable to conventional treatment.47 Although these studies indicate feasibility and safety of direct endomyocardial injection of autologous BM cells into the chronic ischaemic, non-infarcted myocardium the clinical efficacy of this therapy remains unclear. We performed a randomized controlled trial to investigate the effect of direct endomyocardial injection of autologous BM cells in patients with severe CAD who had failed conventional therapy.


Patient population

This study was a randomized, blinded, placebo-controlled trial performed in Hong Kong and Australia. Eligibility for inclusion was based on a history of stable Canadian Cardiovascular Society (CCS) class III or IV angina refractory to medical therapy, no conventional percutaneous or surgical revascularization option as determined by interventional cardiologists and cardiothoracic surgeons, ability to complete >3 min but <10 min of treadmill exercise using modified Bruce protocol, and one or two coronary territories of viable, ischaemic myocardium as documented by dipyridamole single-photon emission computed tomographic (SPECT) perfusion study. The main exclusion criteria consisted of unprotected left main CAD, decompensated heart failure, and/or severe systolic left ventricular (LV) dysfunction, as reflected by an echocardiographic LV ejection fraction (LVEF) ≤30%, acute coronary syndrome, or stroke within the past 3 months, significant renal, liver, or haematological abnormalities, active systemic infection or malignancy, and inability to perform percutaneous electromechanical mapping in the LV, due to atrial fibrillation, LV thrombus, severe peripheral vascular disease, severe aortic stenosis, or an aortic prosthetic valve. All patients gave informed consent and were willing to comply with the planned follow-up. The protocol was approved by the institutional review boards of the two participating centres. This study is registered with www.hkclinicaltrials.com, number HKCTR-3.

Study protocol

Patients were randomly assigned into low or high BM groups or control group in 1:1:1 ratio using a randomization table. Randomization was constrained, stratified on the study centre, and conducted via a system of sealed and numbered envelopes provided to each investigational centre. Two different dosages of BM mononuclear cells were used: low dose (1 × 106 cells/0.1 mL injection) and high dose (2 × 106 cells/0.1 mL injection), to allow us to explore the potential dose-dependent treatment effect of BM cell treatment. Patients randomized to the control group received autologous plasma injection. After randomization, the study processes were blinded to the patients, study coordinators, and the investigators who were responsible for patients assessment. Baseline evaluation in both groups included functional status [New York Heart Association (NYHA) and CCS angina classification],8 treadmill exercise testing (modified Bruce treadmill protocol),9 dipyridamole SPECT perfusion scanning, and cardiac magnetic resonance imaging (MRI).

On the day of the procedure, BM was harvested from every patient by an experienced haematologist via posterior iliac crest puncture under local anaesthesia. A total of 40 mL of BM blood was aspirated, and an adequate trephine biopsy was performed. Mononuclear cells were isolated by Ficoll density gradient centrifugation as previously described.6 BM cells were washed twice in phosphate-buffered saline, re-suspended in phosphate-buffered saline enriched with 10% autologous plasma to either 1 or 2 × 107 mononuclear cells/mL and returned directly to cardiac catheterization laboratory for use. In the placebo preparation for the control group, BM cells were not included in the final suspension, which consisted merely of phosphate-buffered saline with 10% autologous plasma. BM suspensions were tested by flow cytometry (Elite, Beckman Coulter, Fullerton, CA, USA) with directly conjugated antibodies against CD34 (Beckman Coulter, Krefeld, Germany).

At an average of 3–4 h after BM harvest, we performed non-fluoroscopic LV electromechanical mapping (NOGA System, Biosense-Webster, CA, USA) to identify the foci of ischaemic myocardium as previously described.46 During the procedure, systemic anticoagulation was achieved with intravenous heparin to maintain an activated clotting time of 250–300 s throughout the procedure. The targeted injection regions were selected by matching the area of ischaemic myocardium identified by SPECT. After completion of the LV electromechanical mapping, the mapping catheter was replaced by a modified mapping catheter incorporated with a 27 G needle at the tip (Myostar, Biosense-Webster) that could be used for direct endomyocardial injection. Depending on the size of the region, eight to 12 injections of 0.1 mL of BM cell or placebo preparation were delivered evenly in each ischaemic region.

Clinical outcomes

All patients were observed for 24 h after the procedure in the coronary care unit. Serial myocardial isoenzymes (CPK-MB) and electrocardiograms were measured for the first 24 h. A standard transthoracic echocardiogram was also performed within the first 24 h to exclude pericardial effusion post-operatively.

At 3 and 6 months after discharge, patients had follow-up examinations to assess their clinical and functional status by study coordinator in each centre who was blinded to treatment assignment. At 6 months’ follow-up, modified Bruce treadmill testing, dipyridamole SPECT, and cardiac MRI were repeated in all patients.

Adverse events related to direct endomyocardial injection procedures and BM cells implantation were carefully evaluated. Specifically, the occurrence of ventricular arrhythmias was detected by 24 h Holter ECG recording, and the development of myocardial tumour formation was assessed by MRI at 6 months’ follow-up. Furthermore, computed tomography (CT) of thorax was performed at 12 months after procedure to assess for development of intramyocardial tumor or calcification.

Exercise treadmill testing

At baseline and at 6 months after study enrolment, all patients underwent standard exercise treadmill testing per the modified Bruce protocol.9 Each exercise treadmill testing was monitored by a study coordinator who was not aware of the patients’ treatment group assignments. Patients were encouraged to perform as much exercise stress as possible, while their symptoms and ECG signs of cardiac ischaemia were continuously assessed. Specific indications for termination of exercise treadmill testing include: patients’ desire to stop the study due to fatigue or symptoms or development of ECG evidence of myocardial ischaemia/injury by ≥1 mm ST deviation.

Dipyridamole single-photon emission computed tomographic perfusion

Dipyridamole stress and resting SPECT myocardial perfusion imaging was performed at baseline and 6 months’ follow-up. Gated rest and stress SPECT imaging of the heart were acquired Helix (General Electric, Milwaukee, WI, USA) or Marconi Prism 3000 (Philips Medical Systems, Bothell, WS, USA) gamma camera systems. SPECT imaging was performed either as a 1 day rest/stress protocol (sestamibi 10 mCi intravenous injection at rest followed by 30 mCi post peak pharmacological-induced stress by dipyridamole infusion) or a 2-day protocol (sestamibi 20 mCi rest and 20 mCi post-stress). Imaging was performed at least 20 min after the rest injection and 30 min post-stress injection. Analysis of SPECT myocardial perfusion imaging was performed by consensus of two experienced investigators (J.A.P. and T.H.H.). Pairs of stress and rest images from baseline and 6 months’ follow-up were interpreted as a group by investigators blinded to the order of the image pairs. Defect severity (0 = normal, 1 = equivocal, 2 = moderate, 3 = severe, and 4 = absent) was scored in each of 20 segments from a computer display of short-axis, vertical long-axis, and horizontal long-axis images. The sum of stress scores (SSS), sum of rest scores, and sum of difference scores (SDS) were calculated as simple sums of the respective scores.

Cardiac magnetic resonance imaging

Cardiac MRI examination was performed at baseline and 6 months’ follow-up using a 1.5 T scanner (Signa CV/i, General Electric, Milwaukee, WI, USA; Magnetom Sonata; Siemens, Erlangen, Germany; Intera, Philips, Eindhoven, The Netherlands) and a four-channel cardiac phased-array coil. Images were acquired during repeated breath-holds with the patient in supine position. Cine images covering the entire LV were obtained using an ECG-gated steady-state free precession pulse sequence, with the following imaging parameters: TR3.8 ms, TE1.5 ms, flip angle 45°, matrix size 192×160, field of view 28–34 cm, slice thickness 8 mm with no gap, views per segment 16, and number of excitation 1.

All image analyses were performed by two investigators who were unaware of treatment allocation (W.S.C. and R.Y.K.), using a commercially available off-line software package (CineTool 4.5.2, General Electric Healthcare). Measurement of LV volume and calculation of stroke volume and EF were based on previously validated techniques.10,11 For each slice location, the end-diastolic and end-systolic cine frames were defined as the ones showing the largest and smallest LV cavity, respectively. The endocardial border was manually traced, and the end-diastolic and end-systolic volumes were the sums of LV cavity sizes across all contiguous slices in the corresponding phase of the cardiac cycle (Simpson's rule). LVEF was calculated as stroke volume divided by end-diastolic volume and multiplied by 100%. Regional wall thickening was calculated by the percentage change of end-systolic thickness over end-diastolic thickness using a 20-segment model.

Statistical analysis

Primary outcome was the change from baseline in total exercise time on a modified Bruce protocol at 6 months’ follow-up. Secondary outcomes were changes in LVEF, NYHA, and CCS angina classification and sum of different scores on SPECT. The primary comparison was performed between the combined BM-treated groups with two different doses of BM cells vs. control group.

On the basis of the data from our pilot study, we took a standard deviation of 95 s for the change of total exercise time at 6 months’ follow-up from baseline. In order to have at least 80% power to detect a difference of 115 s (30 percentage points) between the control and the combined BM groups, including two different doses of BM cells with a 5% maximum false positive error rate, we would need a total 27 patients with 2:1 randomization to BM(n = 18) and placebo (n = 9) by a two-tailed t-test. The inclusion of two different doses of BM cells in this study was to explore for the potential of dose–response effects. The analysis had taken account of the three-group randomization and was performed in all randomized subjects according to the intention-to-treat principle. The exercise time for the withdrawn patient was available. For other analysis, the baseline value was used to impute the missing value.

Continuous variables were presented as mean ± standard deviation. Categorical data were presented as frequencies and percentages. Comparison of between groups was performed using Student's t-test. The logarithmic transformation was used for the total exercise time due to the presence of non-constant variance in residuals. For the analysis of total exercise time (in log), MRI, and SPECT variables, we did an analysis of covariance (ANCOVA) analysis to assess the differences between the treatment groups at 6 months, after adjusting for baseline values. Specifically, values at 6 months were taken as the dependent variable and the associated baseline values and a factor for treatment were taken as independent variables. Model adequacy was assessed by examining the standardized residuals. We estimated treatment effects by computing the differences (combined BM vs. control) between the adjusted means and their corresponding 95% CIs. Furthermore, paired t-tests were performed to determine the regional changes in SPECT scores and percentage of wall thickening measured by MRI according to the coronary circulation territories in the target and non-target regions for intramyocardial injection. The relation between the total number of CD34+ cells injected into the ischaemic myocardium and the subsequent total exercise time changes were assessed with Pearson's correlation coefficient. Comparisons of NYHA functional class and CSS angina, taken as ordinal outcomes, between BM-treated patients and controls at 3 and 6 months were made by generalized estimating equations with the cumulative logit link function and adjustment on the baseline values. The resulting treatment effects were expressed as proportional odds. All statistical tests were two-sided and used 0.05 as the nominal level of significance. The statistical analysis was performed in the Statistical Analysis System version 9.


Between January 2002 and September 2005, 38 patients were enrolled and 28 patients were randomized into this study (Figure 1). In this study, 15 and 13 patients were randomized in Hong Kong and Australia, respectively. After randomization, one patient in the control group was withdrawn because of failure to access LV to perform mapping and endomyocardial injection due to severe bilateral peripheral artery disease. The patient was, however, included in the analysis by the intention-to-treat principle. The baseline characteristics of these patients are shown in Table 1. After discharge, all patients were maintained on stable cardiovascular medications throughout the study period (Table 1).

Figure 1

Flow chart of the study design. MNC, mononuclear cell.

View this table:
Table 1

Patients’ characteristics

Control group (n = 9)Bone-marrow group (n = 19)
Age, years68.9 (6.3)65.2 (8.3)
Men7 (88)14 (79)
Diabetes mellitus5 (63)8 (42)
Hyperlipidaemia9 (100)19 (100)
Hypertension6 (75)13 (68)
Current cigarette smoker4 (50)8 (42)
Percutaneous coronary intervention7 (88)9 (47)
Coronary artery bypass surgery5 (63)13 (68)
Medication at baseline
 ACE-inhibitors or ARB4 (50)6 (32)
 Aspirin and/or clopidogrel9 (100)19 (100)
 β-Blockers7 (88)18 (95)
 Calcium channel blockers4 (50)8 (42)
 Nitrates4 (50)8 (42)
 Statins9 (100)19 (100)
Medication at 6 months
 ACE-inhibitors or ARB4 (50)7 (37)
 Aspirin and/or clopidogrel9 (100)19 (100)
 β-Blockers6 (75)17 (89)
 Calcium channel blockers4 (50)7 (37)
 Nitrates4 (50)9 (47)
 Statins9 (100)19 (100)
  • Data are means (SD) or n (%) unless otherwise stated.

  • ACE, angiotensin-converting enzyme; ARB, angiotensin receptor.

With respect to BM cell assessment, all the microbiological cultures were negative, and BM trephine biopsy revealed no abnormality. Flow cytometry analysis revealed no significant differences in the number of CD34+ cells in the BM preparation between control (5.4 ± 5.8 × 106 cells) and BM groups (3.6 ± 3.0 × 106 cells, P = 0.32).

A total of 422 percutaneous catheter-based endomyocardial injections to 41 ischaemic regions (inferior = 16, lateral = 9, posterior = 3, and anterior = 13) as guided by the electromechanical mapping were performed in 27 patients. On average, 14.6 ± 0.7 (range 9–22) injections per patient were performed. There was no significant difference in the number of injections given between controls (12.9 ± 4.4) and BM-treated patients (15.3 ± 2.7, mean difference +2.6, 95% CI −0.55 to 5.82, P = 0.12). The mean procedural time and fluoroscopic time was 152 ± 72 min (range 70–370 min) and 33 ± 11 min (range 15–65 min), respectively. There were no acute procedural-related complications, including stroke, transient ischaemic attack, ECG changes, sustained ventricular or atrial arrhythmias, elevation of CPK-MB (more than two times), nor echocardiographic evidence of pericardial effusion within the first 24 h after the procedure. Post-procedure, all patients were discharged from the hospital after 24 h of observation.

Compared with controls, the total exercise time at 6 months’ follow-up was significantly greater in BM-treated patients (+0.43 in log seconds, 95% CI 0.11–0.74; P = 0.014) (Table 2). This represents a total exercise time increased by 53% more in BM treated patients (Figure 2). In the controls, one patient developed dramatically decline in exercise time during follow-up. Further analysis after omitting this patient revealed that the treatment effect of BM cell injection on exercise time remains statistically significant (+0.32 in log seconds, 95% CI 0.03–0.62; P = 0.043). Besides, there was one patient in the control group who developed myocardial infarction 3 months after the procedure. Exclusion of this patient from the analysis showed again a positive effect of BM cell injection on exercise time (+0.47 in log seconds, 95% CI = 0.15–0.80; P = 0.009). Moreover, removal of both patients in the control group resulted in essentially the same conclusion (+0.36 in log seconds, 95% CI = 0.05–0.67; P = 0.032). Furthermore, there was no difference of BM cell effects on exercise time between the two centres after adjusting for baseline values (P = 0.082). In addition, there was no difference of BM cell effects on exercise time between the two centres after adjusting for baseline values (P = 0.082).

View this table:
Table 2

Treadmill exercise parameters, SPECT scores, and left ventricular volume and global left ventricular ejection fraction as determined by cardiac magnetic resonance imaging at baseline and 6 months’ follow-up

Baseline6 monthsChangeBM treatment effect at 6 months (estimatea)P-value
ControlsBM groupControlsBM groupControlsBM group
Treadmill exercise testn = 9n = 19
 Exercise time, log (s)5.99 (0.48)5.91 (0.39)5.74 (0.74)6.05 (0.46)−0.25 (0.46)0.14 (0.33)0.43 (0.11, 0.74)0.014
 Exercise workload, METs4.2 (1.3)4.1 (1.0)4.2 (1.8)5.1 (1.9)0.03 (1.2)1.0 (1.6)0.9 (−0.3, 2.2)0.14
SPECTn = 8n = 19
 Sum of stress scores17.3 (11.0)13.2 (4.8)19.6 (13.5)12.7 (7.0)2.4 (6.9)−0.5 (5.0)−2.7 (−7.6, 2.1)0.28
 Sum of difference scores5.9 (5.4)7.2 (5.6)6.6 (5.6)5.3 (6.5)0.8 (4.9)−1.8 (3.6)−2.4 (−5.8, 1.0)0.17
Cardiac MRIn = 8n = 19
 Global LVEF, %45.7 (8.3)51.9 (8.5)45.3 (8.2)55.6 (8.8)−0.4 (7.5)3.7 (5.1)5.4 (0.4, 10.4)0.044
 LVEDV, mL158.3 (27.3)161.4 (41.3)160.3 (22.8)150.3 (29.3)2.0 (22.1)−11.2 (26.8)−11.7 (−27.4, 4.2)0.16
 LVESV, mL83.9 (25.1)75.9 (33.5)80.8 (18.6)67.0 (21.5)−3.1 (14.4)−8.8 (18.4)−9.3 (−18.4, −0.2)0.058
  • Data are mean (SD) unless otherwise stated.

  • There were no differences between BM group (n = 19) and control group (n = 9) at baseline.

  • BM, bone-marrow; LVEDV, left ventricular end-diastolic volume; LVESV, left ventricular end-systolic volume.

  • aTreatment effects expressed as differences in least-squares means (ANCOVA model) with 95% CI.

Among those BM-treated patients, the mean total number of mononuclear cells injected in the low-dose group (1.67 ± 0.34 × 107 cells) was significantly lower than the high-dose group (4.20 ± 2.80 × 107 cells, P = 0.011). However, there were no significant differences in the mean number of injection (16.7 ± 3.5 vs. 13.8 ± 5.0, P = 0.15) and the mean total number of CD34+ cell injected (0.79 ± 0.66 × 106 vs. 1.38 ± 1.70 × 106, P = 0.38) between patients treated with low- and high-dose BM mononuclear cells injection. The change in total exercise time from baseline to 6 months’ follow-up did not differ between patients treated with low-dose (+89.0 ± 137.9 s) and high-dose (+51.4 ± 163.9s) BM injections (P = 0.59). Furthermore, the improvement in total exercise time in BM-treated patients after 6 months was not correlated with the total number of mononuclear cells (r = −0.27, P = 0.30) nor CD34+ BM cells (r = −0.11, P = 0.66) injected into the ischaemic myocardium.

Figure 2

Treatment effect of bone marrow cells implantation on total exercise time on treadmill test.

The increase in exercise workload during treadmill test as measured by metabolic equivalents from baseline to 6 months in BM treated patients was only slightly higher than controls (Table 2).

Paired cardiac MRI was available in 8 controls and 18 BM treated patients (Figure 3). One control patient withdrew from the study due to the procedural failure refused a repeated MRI examination and one BM treated patient could not undergo MRI examination due to the presence of an implanted pacemaker. Compared with controls, LVEF increased (P = 0.044) and LV end-systolic volume decreased (P = 0.058) in the BM treated patients at 6-months (Table 2). Furthermore, post hoc analysis showed that the percentage of regional wall thickening over the target regions was only significantly increased in BM treated patients at 6 months compared with baseline (+5.33%, 95% CI 2.72–7.94; P = 0.0032) (Figure 4A). Although the changes in LV end-diastolic volume from baseline to 6 months did not differ significantly between the two groups, these tended to decrease in the BM-treated patients (Table 2).

Figure 3

Treatment effect of bone marrow cells implantation on left ventricular ejection fraction as determined by cardiac magnetic resonance imaging.

Figure 4

(A) cardiac magnetic resonance imaging wall thickening and regional changes in (B) single-photon emission computed tomography sum stress score (upper panel) and sum difference score (lower panel) over target and non-target region in control and bone marrow-treated patients at baseline and 6 months’ follow-up.

Paired SPECT myocardial perfusion was available in eight controls and 19 BM-treated patients. One control patient withdrew from the study due to the procedural failure refused a repeated SPECT examination. With respect to SPECT, the extent of stress-induced perfusion defects as determined by SDS was significantly decreased in BM-treated patients at 6 months when compared with baseline (P = 0.04) (Table 2). Furthermore, post hoc analysis showed that the SDS over the target regions was only significantly decreased in BM-treated patients at 6 months compared with baseline (−0.49, 95% CI −0.97 to 0.02; P = 0.0047) (Figure 4B). Although both changes of SSS and SDS from baseline to 6 months did not differ significantly between the two groups, they tended to decrease more in the BM-treated patients (Table 2).

At 6 months, NYHA functional class was significantly lower in BM-treated patients than in controls [odds ratio (OR) for treatment effect 0.12, 95% CI 0.02–0.73, P = 0.021]. However, there was no significant treatment-related difference on CCS angina class between BM-treated patients and controls (OR for treatment effect 0.43, 95% CI 0.09–2.07, P = 0.291) (Table 3).

After a mean follow-up of 19 ± 9months, no patient had sudden cardiac death or was lost to follow-up. During the 6-month study period, none of them received further coronary revascularization. One patient in the control group with baseline inferior and lateral wall myocardial ischaemia developed inferior wall acute myocardial infarction at 3 months after the procedure. Another patient in the control group died of acute myocardial infarction at 31 months after the procedure. One patient in the BM group was diagnosed to have carcinoma of the urinary bladder during follow-up. No patients developed symptomatic cardiac arrhythmias and 24 h Holter recording at 6 months did not reveal any sustained or non-sustained ventricular tachycardia. In seven controls and 19 BM-treated patients, cardiac MRI at 6 months did not demonstrate any tumour formation. Again, in six controls and 18 BM-treated patients who reached the 12 months’ follow-up after the initial procedure, CT thorax did not show any tumour or intramyocardial calcification.


The findings of this randomized controlled trial indicated that direct endomyocardial injection of autologous BM cells in patients with severe CAD who failed conventional therapy resulted in significant but only modest improvement in total exercise time, NYHA functional class, and LVEF compared with placebo. Nevertheless, the absence of any acute procedural complications or long-term sequalae, including ventricular arrhythmia, myocardial damage, or development of intramyocardial tumour or calcification, was encouraging in terms of the safety and feasibility of the current novel catheter-based cellular transplantation procedure.

In this study, significant BM treatment effects were observed on exercise time and LVEF after 6 months. At 3 and 6 months, although both groups reported an improvement in symptoms with respect to NYHA functional class and CCS angina class when compared with baseline, these changes tended to disappear in the controls after 6 months. As a result, significant BM treatment effect was observed on NYHA function class at 6 months.

View this table:
Table 3

Number of subjects in each class of the New York Heart Association functional class and Canadian Cardiovascular Society angina class at baseline and 6 months’ follow-up

Control group (n= 9)Bone marrow group (n= 19)
Baseline6 monthsBaseline6 months
NYHA class
CSS class

Recently, BM cell therapy has been investigated as a means to facilitate myocardial regeneration and/or angiogenesis.12 Previous experimental studies13 have suggested that BM-derived cells may have the capacity to regenerate the myocardium after myocardial infarction. However, the notion that BM-derived stem cells can trans-differentiate into cardiomyocytes has been challenged by recent studies.14,15 With the use of genetic rather than immunofluoresence techniques, these studies have shown that only a very low level of transplantation of BM cells trans-differentiation into cardiomyocytes. However, functional studies after BM cells implantation did demonstrate protection against ventricular remodelling and preservation of LV function.15 These functional benefits in the absence of myocardial regeneration after BM cell implantation is likely attributed to an increased angiogenesis.16,17 There may be multiple mechanisms for the enhancement of myocardial angiogenesis by BM cells. Prior studies have demonstrated that human BM cells contain endothelial progenitor cells with expression of CD34 surface antigen, which can be used to induce neovascularization after myocardial infarction.18 Furthermore, BM cells are a natural source of multiple growth factors involved in neovascularization, including vascular endothelial growth factor.19

We did not observe any significant difference in the improvement in the total exercise time between the low and high-dose BM-treated patients. Although the mean total numbers of mononuclear cells injected was significantly higher in the high-dose group, the mean total number of CD34+ cells injected was not different between the two groups. This may account for the lack of difference in the change of total exercise time between them. Recent experimental studies have shown a dose-dependent effect of CD34+ cells implantation on neovascularization after myocardial infarction.20 However, we also show no direct relationships between the improvement of exercise time with neither the total numbers of mononuclear cells nor the total number of CD34+ cells implanted into the ischaemic myocardium. These findings are consistent with the results of recent studies21,22 which also failed to demonstrate a relationship between the number of injected cells and improvement in clinical outcomes. Therefore, in addition to the numbers of cell implanted, other factors including the functional impairment of BM cells associated with ageing and diabetes, cell survival after transplantation, and optimal composition of types of BM cells may account for individual variation in clinical responses.23 Given the large variability in retention, survival and number of injections, a larger difference (5–10-fold) in the number of BM cells injected might be needed to show an incremental benefit of the treatment.

Recent non-randomized studies47 have suggested that direct BM cell implantation into the ischaemic myocardium improved symptoms and exercise capacity and increased myocardial perfusion and function in patients with refractory CAD. However, the impact of placebo effect of treatment in these studies cannot be determined. In this study, all patients of the BM and control groups underwent identical procedures and follow-up. Indeed, quite striking placebo effects on NYHA and CCS class were observed in controls, but there were no significant improvement on objective functional parameters. In contrast, total exercise time was significantly improved in BM-treated patients. This beneficial effect over and beyond the effect noted in controls suggests a potential therapeutic effect of BM cells on chronic ischaemic myocardium and highlights the need for randomized, placebo-controlled group to prove the beneficial effects. Furthermore, serial cardiac MRI also demonstrated increases in global LVEF and regional wall thickening over the target regions in BM-treated patients, suggesting recovery of myocardial function in the ischaemic myocardium. These findings are consistent with the change in myocardial perfusion, as reflected by a modest reduction in the stress-induced perfusion defects on SPECT in BM-treated patients. However, this trial was not powered to examine the treatment effect in these secondary endpoints. Therefore, it remains unclear that the modest functional beneficial effects observed after BM cells therapy is mediated via enhancement of angiogenesis as suggested in experimental studies.16,17

Recent clinical trials have demonstrated the safety and feasibility of intracoronary route of delivery of BM cells after successful coronary revascularization after myocardial infarction.2428 In our patients with severe CAD, intracoronary injection might not have provided optimal cell transfer, due to the presence of total occlusion and diffuse distal diseases, and the lack of potent local signals for the homing of BM cells that might occur during acute myocardial infarction. Therefore, we performed direct endomyocardial injection of BM cells as guided by electromechanical mapping. When compared with intracoronary delivery, direct endomyocardial injection provides a higher efficacy of cellular delivery, but a more uneven distribution of BM cells.29 However, we have not observed any acute or long-term adverse effects related to direct endomyocardial implantation of BM cells in chronic ischaemic myocardium. In contrast, the requirement of a specialized system and the high cost of the procedure may limit the use of technique. Therefore, future development of a more simple and inexpensive technique for catheter-based direct endomyocardial injection will increase the feasibility of this treatment strategy.

This study had several limitations. First, although no statistical significant differences in baseline characteristics between the BM-treated group and placebo group were observed, there were some imbalances in mean exercise time, SPECT score, and LVEF between them due to a small sample size. These might lead to bias in the results. However, we have made the best effort possible to maintain the balance between groups by following proper randomization procedure. Moreover, adjustment of all possible baseline imbalances resulted in essentially the same conclusion of the effect of BM in improvement exercise time. Second, patients were not trained to perform exercise treadmill test before the study. As a result, the impact of the training effects on exercise time cannot be determined. Third, this trial did not have sufficient power and sample size to allow us to determine the optimal cell type or dose of BM cell implantation and to identify any potential clinical predictors for clinical improvements. Furthermore, no functional tests such as colony-forming unit and migration capacity and/or CXCR-4 receptor expression were measured to correlate the degree of clinical improvement after injection. Forth, our current results also cannot determine whether the observed beneficial effects of BM treatment can persist beyond the 6-month period. Therefore, future studies are required to address the optimal number and phenotype of BM cells for therapeutic angiogenesis, as well as the long-term clinical impact of this novel therapeutic approach.

In summary, this randomized, placebo-controlled trial has shown beneficial effects of direct endomyocardial injection autologous BM cells on symptoms, functional capacity, and LVEF in patients with chronic myocardial ischaemia who failed conventional medical and revascularization procedures. The documented safety and clinical benefits of this novel procedure for this patient group prompt future studies to investigate the therapeutic benefit on survival and cardiovascular outcomes of this therapy.


This study was partially supported by the Sun Chieh Yeh Heart Foundation Fund, S.K. Yee Medical Foundation Grant (Project No. 203217), and The Research Grants Council of Hong Kong(HKU 7357/02M) from Hong Kong.


This study was an academia-initiated study. There was no external industry sponsor involved in study design, data collection, data analysis, data interpretation, or writing of the report.

Conflict of interest statement. H.F.T. and S.T. received consultant fee from Biosense-Webster, CA, USA. All other authors declare that they have no conflict of interest.


  • The authors contributed equally to the article as first authors.

  • The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the open access version of this article for non-commercial purposes provided that the original authorship is properly and fully attributed; the Journal, Learned Society and Oxford University Press are attributed as the original place of publication with correct citation details given; if an article is subsequently reproduced or disseminated not in its entirety but only in part or as a derivative work this must be clearly indicated. For commercial re-use, please contact journals.permissions@oxfordjournals.org


View Abstract