Long-term follow-up of haploidentical hematopoietic stem cell transplantation without in vitro T cell depletion for the treatment of leukemia

Nine years of experience at a single center




Many patients who require allogeneic hematopoietic stem cell transplantation (allo-HSCT) lack a human leukocyte antigen (HLA)-matched donor. Recently, a new strategy was developed for HLA-mismatched/haploidentical transplantation from family donors without in vitro T cell depletion (TCD).


Over the past 9 years, 756 patients underwent haploidentical transplantation using a protocol developed by the authors, which combines granulocyte-colony stimulating factor–primed bone marrow (G-BM) and peripheral blood stem cells without in vitro TCD. The long-term outcome with this treatment modality was reported, and a risk-factor analysis was provided.


Of these patients, 752 (99%) achieved sustained, full donor chimerism. The incidence of grades 2 through 4 acute graft-versus-host disease (GVHD) was 43%, and the 2-year cumulative incidence of total chronic GVHD was 53%. The 3-year cumulative incidence of nonrelapse mortality was 18%. The 2-year cumulative incidences of relapse were 15% and 26% in the standard-risk and high-risk groups, respectively. Of the 756 patients, 480 survived throughout the follow-up period of 1154 days (range: 335-3511 days) with the 3-year leukemia-free survival rates of 68% and 49% in the standard-risk and high-risk groups, respectively. Lower leukemia-free survival was associated with high-risk disease status (P = .001), chronic myelogenous leukemia disease type (P = .004), neutrophil engraftment beyond 13 days after transplant (P = .012), and the occurrence of grades 2 through 4 acute GVHD (P = .019).


The results from the authors' 9-year experience showed that G-BM combined with peripheral blood stem cells from haploidentical donors, without in vitro TCD, is a reliable source of stem cells for transplantation by using the protocol developed by the authors. Cancer 2013. © 2012 American Cancer Society.

Haploidentical/human leukocyte antigen (HLA)-mismatched hematopoietic stem cell transplantation (HSCT) has been performed for more than 20 years. Many researchers focused on the techniques of ex vivo and in vitro T cell depletion (TCD) of the grafts,1, 2 but poor posttransplant immune reconstitution and infection-related mortality remain major obstacles to patient survival. To overcome these shortcomings, certain researchers, including our group, have focused on unmanipulated allografts and posttransplant immune suppression.3-12 We produced an intensive in vivo immunosuppressive protocol for haploidentical transplantations without in vitro TCD.5, 6 There are 4 elements to this protocol: granulocyte colony-stimulating factor (GCSF) for all donors, intensive immune suppression, anti-human thymocyte immunoglobulin (ATG), and combination of GCSF–primed peripheral blood stem cells (PBSCs) and bone marrow stem cells (G-BM), Promising results have been achieved using this protocol for the past 9 years.5, 6, 13-15 More patients and a longer follow-up are needed to confirm its efficacy. Therefore, we initiated the current study to update our data and to confirm the reliability of this technique. This study describes the largest cohort of patients with haploidentical HSCTs so far treated uniformly without the in vitro TCD modality. This report will focus on the clinical features and long-term outcome of our treatment modality and provide a risk-factor analysis.


Patient Eligibility

A total of 756 patients with leukemia underwent haploidentical allo-HSCT between May 2002 and December 2010 at the Peking University Institute of Hematology, Beijing, China. Of the 756 patients, 138 of the 620 patients with acute leukemia (AL) were previously reported in 2009,13 and 58 of the 136 patients with chronic myeloid leukemia (CML) were previously reported in 2008.14 All of these previously reported patients were enrolled and further followed in this study. The Institutional Review Board of Peking University, Beijing, China, approved this study, and all of the patients and their donors gave written informed consent. A pretransplantation cytomegalovirus (CMV) serologic analysis showed that 1%, 14%, and 85% of the patients were low risk (recipient [R]−, donor [D]−), intermediate risk (R−, D+), and high risk (R+, D+) for CMV reactivity after HSCT, respectively. The characteristics of the 756 patients and donors are summarized in Table 1.

Table 1. Characteristics of Patients and Donors
CharacteristicAll Patients (n = 756)
  1. Abbreviation: Ph, Philadelphia chromosome.

Median age, y25 (3-57)
Disease status
  Standard-risk561 (74.2%)
  High-risk195 (25.8%)
Disease type
Acute myeloid leukemia
  First complete remission234
  Second complete remission29
  ≥Third complete remission5
Acute lymphoid leukemia
  Ph-negative first complete remission183
  Ph-negative second complete remission38
  Ph-negative ≥ third complete remission4
  Ph-negative nonremission14
Chronic myeloid leukemia
  First chronic phase77
  Beyond first chronic phase59
Matched human leukocyte antigen locus
Donor-recipient sex
Donor-recipient blood type
  Minor mismatch149
  Major mismatch140
  Minor + major52
Donor-recipient relation
CD34+ count, median (range)2.21 (0.27-55.27) × 106/kg
CD3+ count, median (range)1.53 (0.1-8.29) × 108/kg

Donor and Stem Cell Harvesting

Donor selection, HLA typing, and stem cell harvesting have been described in detail.13

Conditioning Regimen and Graft-Versus-Host Disease Prophylaxis

The conditioning therapy was consisted of cytarabine (4 g/m2/day, on days −10 to −9), busulfan (4 mg/kg/day, orally on days −8 to −6 before January 2008 and 3.2 mg/kg/day, intravenously on days −8 to −6 after January 2008), cyclophosphamide (1.8 g/m2/day, on days −5 to −4), 1-(2-chloroethyl)-3-(4-methylcyclohexyl)-1-nitrosourea (Me-CCNU; 250 mg/m2, once on day −3), and ATG (20 mg/kg/day, porcine; the Bioproduct, Wuhan, China, 14 cases, between 2003 and 2004; or 2.5 mg/kg/day, rabbit; Sang Stat, Lyon, France, all the other 742 cases, on days −5 to −2). Graft-versus-host disease (GVHD) prophylaxis were previously described in detail.13 All transplant recipients received cyclosporine A, mycophenolate mofetil, and short-term methotrexate.

Relapse Prevention, Monitoring, Intervention, and Treatment

Philadelphia chromosome–negative AL patients

For high-risk Philadelphia chromosome (Ph)-negative AL patients (n = 76, Table 1), a prophylactic donor lymphocyte infusion (DLI) was administered to approximately half of the patients who received transplants before July 2009 (28 of 53 patients) according to the physicians' and patients' intention, and after July 2009, nearly all of the patients without active GVHD or early relapse received a prophylactic DLI (21 of 23 patients). We made several modifications to the classic DLI, as previously described in detail.16 Starting in January 2006, the minimal residual disease (MRD) targets were regularly monitored. The bone marrow samples were defined as abnormal if they contained > 0.001% of leukemia-associated immunophenotypic patterns17 or > 0.6% of Wilms' tumor gene 1 (WT1).18 DLI or interleukin-2 was administered according to donor availability to MRD-positive patients. When a hematologic relapse was diagnosed after HSCT, the relapse was treated with chemotherapy followed by a therapeutic DLI.19 For patients with acute myeloid leukemia (AML), aclacinomycin was administered at a dose of 10 mg/day for 5 days and cytarabine was administered at a dose of 150 mg/m2 for 7 days; for patients with acute lymphoid leukemia (ALL), methotrexate was administered at a dose of 1.5 g/m2 for 1 day. The median number of CD3+ cells infused in each patient was 0.58 (0.13-2.03) × 108/kg.

Ph-positive patients

Starting in May 2005, imatinib (Novartis, Basel, Switzerland) treatment was initiated for the eligible Ph-positive ALL patients after transplantation for the prevention of leukemia relapse. The imatinib treatment protocol has been described in detail.20 Any Ph-positive ALL or CML patients who experienced a cytogenetic or hematologic relapse after HSCT received imatinib for therapeutic use.

Definitions and Assessments

Patients with leukemia were categorized as “standard-risk” if they were in their first or second complete remission (CR1 or CR2) of AL without the translocation between the ABL and BCR genes [t(9;22)(q34;q11)], or in the chronic phase of CML. Patients were classified as “high-risk” if they had AL with [t(9;22)(q34;q11)] regardless of disease stage, AL in its third complete remission (CR3) or greater, or in nonremission regardless of cytogenetics, or CML beyond the first chronic phase. Assessments of engraftment, chimerism, GVHD, and relapse were described in detail.13


The cumulative incidences were estimated for engraftment, GVHD, nonrelapse mortality (NRM), and relapse to accommodate the competing risks. The associations between the potential factors and outcomes were evaluated using an add-on package for the R statistical software that allows for the estimation of the semiparametric proportional hazards model for the subdistribution of a competing risk analysis, as proposed in the literature.21 For the continuous variables, the median value was used as the cutoff point. Engraftment and GVHD were studied as time-dependent variables. The probabilities of overall survival and leukemia-free survival (LFS) were estimated by the Kaplan-Meier method. The potential prognostic factors were evaluated in univariate analyses by the log-rank test, with P < .05 considered statistically significant. In the multivariate analysis, the forced factors, including disease status, HLA disparity, and all of the factors found to influence the outcomes in univariate analysis with a P < 0.1, were included into a Cox proportional hazards model using time-dependent variables. The potential interaction between the main effect and all of the significant covariates was tested. No interactions were detected. SAS version 8.2 (SAS Institute, Cary, NC) and S Plus 2000 (Mathsoft, Seattle, Wash) were used for most of the analyses. The endpoint of the last follow-up for all of the surviving patients was December 1, 2011.



A total of 752 patients (99%) achieved sustained myeloid engraftment. Polymerase chain reaction DNA fingerprinting of the short tandem repeats on the recipient BM cells was used to confirm 100% donor chimerism in these patients. The median time to reach an absolute neutrophil count above 0.5 × 109cells/L was 13 days (range: 8-49 days). The cumulative 30-day myeloid engraftment probability was 99.1% (CI,98.9%-99.3%). During the follow-up period, 694 patients (92%) exhibited platelet engraftment, and the median time to reach a platelet count above 20 × 109cells/L was 16 days (range: 5-195 days). The 50-day cumulative platelet engraftment probability was 88% (95% CI = 87%-88%). The median time to achieve myeloid and platelet engraftment was 12 days versus 13 days, and 14 days versus 17 days (P < .001) for patients with lower (≤ median) versus higher (> median) numbers of infused CD34 cells, respectively. The risk factor analysis is shown in Tables 2 and 3.

Table 2. Univariate Analysis of Haploidentical Transplant Outcomes
Risk FactorsMyloid EngraftmentAcute GVHD, Grades 2-4Extensive Chronic GVHDRelapseNonrelapse MortalityLeukemia-Free Survival
  1. Abbreviations: GVHD, graft-versus-host disease; HLA, human leukocyte antigen.

HLA disparity
Donor-patient relation
Donor-patient sex
Donor-patient blood type
 Minor mismatch993923141967
 Major mismatch944429212058
Median CD34
Median CD3
Median MNC
Disease status
Disease type
 Acute myeloid leukemia994221221564
 Acute lymphoid leukemia994325202157
 Chronic myloid leukemia99442591973
Patient age
Transplant year
Table 3. Multivariate Analysis of Haploidentical Transplant Outcomes
OutcomeHazard Ratio (95% Confidence Interval)P
  • a

    Four degrees of freedom test.

  • b

    Based on final model, time to engraftment, aGVHD and extensive cGVHD had significant effect on non-relapse related mortality: relative risk (RR) (95% confidence interval [CI]) for myeloid engraftment of ≤ 13 days versus > 13 days = 0.53 (0.35-0.82; P = .004), for platelet engraftment of ≤ 16 days versus > 16 days = 0.56 (0.35-0.88; P = .012), RR (95% CI) for aGVHD of grades 0 to 1 versus 2 to 4 = 0.51 (0.34-0.75; P = .001), for extensive cGVHD of no versus yes = 0.63 (0.43-0.93; P = .011);

  • c

    Two degrees of freedom test.

  • d

    Based on final model, time to myeloid engraftment, aGVHD had significant effect on leukemia-free survival: RR (95% CI) for myeloid engraftment of ≤ 13 days versus > 13 days = 0.70 (0.53-0.92; P = .012), RR (95% CI) for aGVHD of grades 0 to I versus II to IV = 0.74 (0.58-0.95; P = .019).

  • Abbreviations: aGVHD, acute graft-versus-host disease; cGVHD, chronic graft-versus-host disease.

Myeloid engraftment
 HLA disparity 2-3 vs 0-11.02 (0.81-1.28).837
 Disease status standard vs high risk1.29 (1.08-1.53).004
 Other significant risk factors
   CD34 ≤ median vs > median0.77 (0.66-0.90).001
Platelet engraftment
 HLA disparity 2-3 vs 0-10.83 (0.83-1.05).126
 Disease status standard vs high risk1.24 (1.02-1.50).025
 Other significant risk factors
  Mononuclear cell count ≤ median   vs > median0.83 (0.70--0.97).026
  CD34 ≤ median vs > median0.70 (0.59-0.82).001
Acute GVHD grade 2-4
 HLA disparity 2-3 vs 0-11.29 (0.89-1.87).165
 Disease status standard vs high risk0.82 (0.64-1.06).141
 Other significant risk factors
  Transplant year 2002-2005 vs 2006-20101.39 (1.02-1.87).033
  Patient-donor relation .001a
   Father-child0.67 (0.48-0.85).007
  Sibling-sibling0.59 (0.40-0.74).001
   Child-parent0.19 (0.08-0.39).001
   Other0.60 (0.29-1.22).075
Chronic GVHD, extensive
 HLA disparity 2-3 vs 0-11.27 (0.78-2.05.).325
 Disease status standard vs high risk0.70 (0.50-0.99).049
 Other significant risk factors
  Transplant year 2002-2005 vs 2006-20101.54 (1.06-2.21).001
  CD3 ≤ median vs >median0.67 (0.49-0.91).012
Non-relapse related mortalityb
 HLA disparity 2-3 vs 0-12.14 (0.98-4.73).054
 Disease status standard vs high risk0.54 (0.37-0.80).007
 HLA disparity 2-3 vs 0-10.80 (0.50-1.28.).361
 Disease status standard vs high risk0.36 (0.25-0.52).001
 Other significant risk factors
  Diagnosis .001c
   Chronic myeloid leukemia1.0 
   Acute lymphoid leukemia3.02 (1.64-5.54).001
   Acute myeloid leukemia3.40 (1.84-6.52).001
Leukemia-free survivald
 HLA disparity 2-3 vs 0-11.12 (0.75-1.66).567
 Disease status standard vs high risk0.45 (0.35-0.59).001
 Other significant risk factors
  Diagnosis .004c
   Chronic myeloid leukemia1.0 
   Acute lymphoid leukemia1.83 (1.26-2.66).001
   Acute myeloid leukemia1.79 (1.22-2.61).003


At 100 days after transplant, the cumulative incidences of grade 2 through 4 acute GVHD and grade 3 or 4 acute GVHD were 43% (95% CI = 40%-46%) and 14% (95% CI = 11%-17%), respectively. The cumulative incidences of total chronic GVHD and extensive chronic GVHD were 53% (95% CI = 51%-55%) and 23% (95% CI = 21%-26%) at 2 years after transplantation, respectively. According to the US National Institutes of Health consensus criteria, there was 1 case of late-onset acute GVHD and 31 cases of overlap syndrome. Among all of the donor-recipient sex combinations, the female-to-female combination had the highest grade 2 acute GVHD (41%), whereas the female-to-male combination had the highest grade 3 or 4 acute GVHD (19%). The risk factor analysis is shown in Tables 2 and 3. The cumulative incidences of grade 2 through 4 acute GVHD for various donor-recipient pair relationships are shown in Figure 1.

Figure 1.

Cumulative incidence of acute graft-versus-host disease (aGVHD) for various donor-recipient pair relationships is shown after haploidentical hematopoietic stem cell transplantation (P = .001).

CMV Infection

The 100-day cumulative incidence of CMV viremia was 64% (95% CI = 63%-65%). The 100-day cumulative incidence of CMV-associated pneumonia was 4% (95% CI = 3%-4%). The incidence of CMV antigenemia was not associated with the age of the patients, the extent of HLA disparity, or the stage of disease before transplantation. The proportion of patients who became antigenemia-positive during the follow-up was higher among CMV-seropositive patients (D+/R+, 56%; D−/R+, 59%) than CMV-seronegative patients (D+/R−, 42%; D−/R−, 12%).


As of December 1, 2011, a total of 140 patients had recurrent disease after a median of 183 days (range: 10-1700 days), reaching a cumulative incidence of relapse of 18% (95% CI = 15%-21%) at 2 years. Among the high-risk Ph-negative AL patients, 14 of the 49 patients (29%) who underwent a prophylactic DLI and 17 of the 27 patients (63%) without a prophylactic DLI had relapsed after HSCT. Starting in 2006, among the standard-risk Ph-negative AL patients who were MRD-positive, 6 of the 29 patients (21%) who underwent an interventional DLI and 12 of the 24 patients (50%) without an interventional DLI and 62 of the 387 MRD-negative patients (16%) had relapsed after HSCT. Among the Ph-positive ALL patients, 5 of the 43 patients (11%) treated with prophylactic imatinib and 5 of the 17 patients (29%) without prophylactic imatinib treatment had relapsed after HSCT. A chimerism analysis showed full donor chimerism (with 0.5% recipient DNA as the cutoff) before and after prophylactic and interventional DLI.

After relapse, 29 of the 60 patients who underwent a therapeutic DLI achieved CR; 13 of the 29 patients died from other causes when in remission, and 16 patients remained alive and free of leukemia until the last follow-up. Twenty-three Ph-positive patients, comprising 10 Ph-positive ALL and 13 CML patients, received tyrosine kinase inhibitors for the treatment of a cytogenetic or hematologic relapse. Mixed donor chimerism was detected in the BM samples of all of the relapsed patients. The median percentage of recipient DNA in these samples was 29% (range, 6%-97%). Twenty-nine patients achieved full donor chimerism after a therapeutic DLI.

The risk factor analysis is shown in Tables 2 and 3. The cumulative incidences of relapse characterized by disease status are shown in Figure 2.

Figure 2.

Cumulative incidence of relapse characterized by disease status is shown after haploidentical hematopoietic stem cell transplantation (P = .011).

Survival and Long-Term Follow-Up

As of December 1, 2011, 480 patients were alive, with a median time of 1154 days (range, 335-3511 days) after transplantation without disease recurrence. The probabilities of overall survival and LFS at 3 years were 67% (95% CI = 64%-71%) and 63% (95% CI = 60%-67%), respectively. A total of 136 patients died from causes other than relapse (Table 4). The most frequent cause of death was infection, specifically, pneumonia. The NRM rate at 3 years was 18% (95% CI = 15%-21%).

Table 4. Causes of Death
CauseNo. of Patients (n = 240)
Graft-versus-host disease16
Organ failure3

The risk factor analysis is shown in Tables 2 and 3. Particular donor-recipient pair relationships appeared to influence the NRM, although this trend did not reach statistical significance. However, the influence was statistically significant when we compared paternal donors with maternal donors only, with NRM rates of 13% and 24% (P = .001) and LFS rates of 68% and 56% (P = .009, data not shown), respectively. LFS with respect to disease type is shown in Figure 3.

Figure 3.

Probability of leukemia-free survival (LFS) is shown with respect to disease type after haploidentical hematopoietic stem cell transplantation (HSCT) (P = .049).


This updated study confirmed and extended our previous finding that the transplantation of G-PBSCs and G-BM from HLA-mismatched/haploidentical related donors without in vitro TCD by our regimen resulted in a high rate of engraftment and similar incidences of GVHD, relapse, NRM, and survival.5, 6, 13-15, 22 Our protocol is reliable for leukemia patients.

The current results showed that a low number of CD34 cells in allografts and a high-risk disease status were independently associated with slower engraftment, particularly platelet recovery, which confirms the results of Chang et al.23 Therefore, to ensure rapid engraftment, especially in high-risk patients, higher CD34 levels are preferred. There was no significant association between the extent of HLA disparity and the time to engraftment following haploidentical transplantation in our protocol.

In this report, a maternal donor appeared to be unfavorable for GVHD, NRM, and LFS compared with a paternal donor. It seems that the female-to-female combination has the highest rate of acute GVHD, but it does not translate into any higher NRM. However, the female-to-male combination results in a higher NRM than other donor-recipient sex combinations. These results may be explained by the increased GVHD rate observed for the female-to-female combination, which only included grade 2 and responded better to steroids, whereas the female-to-male combination has the highest rate of grade 3 or 4 acute GVHD and extensive chronic GVHD. Our results were not in accordance with previous studies, which showed that the donor-specific suppression of T cell responses against noninherited maternal antigen (NIMA) is associated with lower risks of GVHD and NRM after transplantation from mother to child, compared with father to child.11, 12, 24 Matsuoka et al.25 revealed the involvement of CD4+ CD25+ regulatory T cells in the tolerogenic NIMA effects. The use of ATG potently depletes T lymphocytes in vivo in the long term; thus, it may partially conceal the histocompatibility barriers due to NIMA/noninherited paternal antigen (NIPA) mismatch in the haploidentical transplantation setting. The immune mechanism of the current finding is not clear. Our results were in accordance with the finding of Kim et al.,26 which indicated that a female donor was associated with a higher incidence of severe acute GVHD and extensive chronic GVHD. Furthermore, among 481 donor-patient pairs who underwent haploidentical HSCT in our institute, a female donor was found to be one of the significant factors associated with higher NRM and lower LFS.27 Therefore, our current results might be explained by the sex difference of maternal and paternal donors rather than the NIMA/NIPA effect. Further studies are necessary to better explore the effect of NIMA on haploidentical HSCT.

The relapse rate among high-risk patients in this report was lower than that described in our previous report, which suggested a relapse rate of 39% among 86 patients who received a transplant before 2005.5 The relapse rate for high-risk leukemia was reported to be 44% to 63%.7, 28 The increased application of prophylactic DLI in recent years may contribute to this difference,29 which is by no means certain because there was no proper control group, and because of improved supportive care and a changing referral pattern over time (ie, an increasing acceptance of haploidentical transplantation in the oncology community and among patients may result in earlier referrals such that the high-risk patients of recent years may not be the exact same as they were in earlier years). However, 21% of the patients who received a DLI for MRD and 50% of the patients without interventional DLI had relapsed after HSCT, suggesting that DLI was at least an effective method of preventing relapse in patients with standard-risk acute leukemia.

More recently, our group showed that for every major HSCT endpoint, including relapse, NRM, and survival, partially matched-related HSCT and unrelated HSCT are not significantly different.22 In addition, a superior graft-versus-leukemia effect was achieved with the transplantation of haploidentical donor grafts compared with HLA-identical sibling donor grafts for high-risk acute leukemia.30 This study provides better donor choice at experienced transplant centers, particularly under certain specialized circumstances, and an opportunity for patients to benefit from HSCT when an HLA-matched donor is not available.


This work was supported (in part) by National Natural Science Foundation of China (grant 30971292), National High-tech R&D Program of China (863 Program), Leading Program of Clinical Faculty accredited by the Ministry of Health of China, National Scientific Major Program-major new drug formulation (grant 2008zx09312-026), and Beijing Key Laboratory of Hematopoietic Stem Cell Transplantation.


The authors made no disclosure.