Haploidentical vs identical-sibling transplant for AML in remission: a multicenter, prospective study

Acute myeloid leukemia (AML) is a common hematologic malignant disease. HLA-identical sibling donor (ISD) hematopoietic stem cell transplantation (HSCT) offers significantly improved overall survival (OS) for intermediate- or high-risk AML patients compared with chemotherapy alone^1-5 ; thus, it is recommended as a first-line postremission treatment in this clinical situation. Recent progress in haploidentical donor (HID) transplantation provides the benefits of rapid and near-universal donor availability.^6-8  However, whether haploidentical HSCT could serve as a first-line postremission therapy in intermediate- or high-risk AML remains to be determined.

In recent years, our group established an unmanipulated haploidentical blood and marrow transplantation protocol in which anti-thymocyte globulin (ATG) in the conditioning and the powerful posttransplantation graft-versus-host disease (GVHD) prophylaxis probably provided a major contribution to controlling alloreactivity.^7,8  Additionally, our recent single-institute, patient self-selected study showed that haploidentical HSCT was superior to chemotherapy alone for patients with intermediate- or high-risk AML in the first complete remission (CR1).⁹  Undoubtedly, randomized trials are needed to determine if haploidentical transplantation is, in fact, superior to chemotherapy for AML in CR1. However, this may be difficult to determine in this setting because of ethical and practical reasons given the inferior outcomes observed after chemotherapy-based postremission strategies compared with those after allo-HSCT.^1,3-5  Conversely, prior single-center analyses demonstrated similar survival after unmanipulated HID HSCT using ATG vs ISD HSCT for hematologic malignancies.^10,11 

Taken together, we hypothesized that haploidentical HSCT is a valid option as a postremission treatment of AML in CR1. The results from haploidentical HSCT have not formally been compared in prospective studies with those from ISDs to evaluate the value of haploidentical HSCT for the specific, homogenous disease of intermediate- or high-risk AML in CR1. In addition, evaluating a uniform haplo-HSCT protocol in more clinical centers would enable prospective, multicenter trials on the standardized treatment. Therefore, based on the current prospective, multicenter, disease-specific study, we compare the outcomes of consecutive AML patients in CR1 undergoing T-cell–replete haploidentical HSCT uniformly performed at 3 centers with all contemporaneous ISD HSCT.

We conducted a disease-specific, multicenter, nonrandomized trial. Patients were assigned to undergo haploidentical or ISD HSCT according to donor availability. The primary end point was disease-free survival (DFS), OS, and relapse rates. Secondary end points included an evaluation of immunologic reconstitution. Enrollment began in July 2010 and ended after the accrual goal was met, which occurred in November 2013. The analysis includes data collected as of November 30, 2014. The median follow-up for surviving patients is 952 days (range; 365-1612 days) after HSCT. This article focuses on the primary end points. Analyses of immune reconstitution are ongoing.

Eligible patients were aged 15 to 60 years and had intermediate- or high-risk AML in CR1 with no contraindications to HSCT. HLA-ISD was the first choice for allotransplantation. If an HLA-ISD was unavailable, subjects without a suitable closely HLA-matched unrelated donor (>8 of 10 matching HLA-A, B, C, DR, and DQ loci; and >5 of 6 matching HLA-A, B, and DR loci) after 2 cycles of consolidation were eligible for HLA-haplotype transplantation. Consecutive patients receiving HID (n = 231) or ISD (n = 219) HSCT during the study period were enrolled. For this comparative analysis to arrive at comparable patient cohorts that received transplants during the same time interval, we excluded patients who received unrelated donor HSCT (n = 58).

The study was performed in accordance with the Declaration of Helsinki and was approved by the local institutional review board. All donors and recipients gave written informed consent before enrollment. All authors vouch for the accuracy and completeness of the reported data and analyses and for the adherence of the study to the protocol. This study was registered as #ChiCTR-OCH-10000940 at www.chictr.org.

Patients were classified as intermediate- or high-risk based on National Comprehensive Cancer Network cytogenetic abnormalities.¹²  Information on the monosomal karyotype,¹³  Center for International Blood and Marrow Transplant classification,¹⁴  nucleophosmin (NPM), and fms-related tyrosine kinase 3 (FLT3) is listed in Table 1. The protocol required DNA typing of the patient and donor at intermediate resolution for HLA-A, B and at high resolution for DRB1. HID subjects received a graft from a family member sharing 1 HLA haplotype with the recipient but differed to a variable degree for the HLA-A, B, and DR antigens of the unshared HLA haplotype. Granulocyte colony-stimulating factor (G-CSF; 5 µg/kg of body weight per day for 5 days) was used to mobilize the BM and PB. The target mononuclear cell count was >6 × 10⁸/kg of recipient weight. Unmanipulated BM (harvested on day 4 after G-CSF) and PB stem cells (harvested on day 5 after G-CSF) were infused into the recipient on the day of collection. All HID patients received both BM and PB mobilized with G-CSF.

The conditioning therapy for the HID group was as follows: cytarabine (4 g/m² per day) IV on days −10 to −9; busulfan (3.2 mg/kg per day) IV on days −8 to −6; cyclophosphamide (1.8 g/m² per day), IV on days −5 to −4; methyl chloride hexamethylene urea nitrate (Me-CCNU) (250 mg/m² per day), orally once on day −3; and ATG (2.5 mg/kg per day; Sang Stat, Lyon, France) IV on days −5 to −2. Patients in the ISD group received hydroxycarbamide (80 mg/kg) orally on day −10 and a lower dose of cytarabine (2 g/m² per day) on day −9, but otherwise, an identical regimen to the HID patients without ATG was employed. GVHD prophylaxis regimen consisted of cyclosporine A, mycophenolate mofetil, and short-term methotrexate.¹¹  Between May 2012 and November 2013, 25 subjects in Peking University were randomized to receive low-dose corticosteroid prophylaxis (www.clinicaltrials.gov, #NCT01607580).

After completion of the study treatment, BM samples were analyzed at 1, 2, 3, 4.5, 6, 9, and 12 months after transplantation and at 6-month intervals thereafter for the monitoring of minimal residual disease (MRD), defined as previously reported.^15,16  In Peking University, modified donor lymphocyte infusion (DLI) would be given before hematologic relapse as the intervention therapy (preemptive DLI) after 3 months post-HSCT following a trial of immunosuppressant withdrawal. The detailed criteria for preemptive DLI administration included the following: (1) patients scored as MRD+ if they had 2 consecutive positive results using flow cytometry or Wilms’ tumor gene 1 or were both flow cytometry–positive and Wilms’ tumor gene 1– positive in a single sample within 1 year after transplantation; (2) no uncontrolled GVHD or life-threatening infection; and (3) with donor availability and willingness. Patients with GVHD first received GVHD therapy. After GVHD was controlled, MRD testing was repeated, and those patients that remained MRD+ received modified DLI. In total, 23 patients (13 HID, 10 ISD) required this preemptive approach, which occurred at a median of 171 days (range, 109-319 days) after transplantation. The modified DLI regimen was as previously described.¹⁷  When a hematologic relapse was diagnosed after HSCT, posttransplant immune suppression was immediately discontinued. If patients did not develop GVHD within 2 weeks, and if patients agreed to receive targeted therapeutic modified DLI and their donors also agreed to undergo PB stem cell collection again, the patients would receive chemotherapy followed by modified DLI; otherwise, the patients would receive chemotherapy alone.¹⁸  Antileukemia chemotherapy before DLI included aclacinomycin (10 mg/m² per day for 5 days) and Ara-C 100 (mg/m² per day for 5 days); fludarabine (30 mg/m² per day for 5 days), Ara-C (1.0 g every 12 hours for 10 doses), and G-CSF; or harringtonine (2 mg/m² per day for 5 days), aclacinomycin (10 mg/m² per day for 5 days), and Ara-C (100 mg/m² per day for 5 days). In total, 18 patients (11 HID, 7 ISD) were given this targeted therapeutic DLI (see “Results”). The median number of CD3⁺ cells infused in each patient was 0.58 (interquartile range, 0.13–2.03) × 10⁸/kg. Chimerism analyses were done by DNA fingerprinting of short tandem repeat on blood samples.

Comparisons of patient characteristics between the 2 groups were performed using the Mann-Whitney U test for continuous variable and χ² test for categorical data. Relapse, nonrelapse mortality (NRM), engraftment, and GVHD were estimated as cumulative incidences, taking into account competing risks. NRM was defined as death from any cause in the first 28 days post-HSCT or death without evidence of disease recurrence beyond day 28. Relapse was defined as recurrence of BM blasts >5%, reappearance of blasts in the blood, or development of extramedullary disease infiltrates at any site. Assessments of engraftment and chimerism were previously described in detail.¹¹  Chronic GVHD was classified as limited or extensive and was also classified as mild, moderate, or severe by the National Institutes of Health consensus criteria. GVHD was evaluated and graded by a single practitioner within the program.

Survival functions were estimated using the Kaplan-Meier method and compared by the log-rank test. The starting point for comparing outcomes between the cohorts was the point when complete remission was declared. Univariate probabilities of survival and DFS were calculated using a left-truncated version of the Kaplan-Meier estimator with 95% confidence intervals (CIs). Cumulative incidence of relapse (CIR) and NRM were calculated using a left-truncated version of cumulative incidence function to accommodate competing risks. To adjust for the differences in baseline characteristics, left-truncated versions of the Cox proportional hazards regression models were used to evaluate the relative risk of subjects receiving either allotransplant by forcing the main interest variable (HID vs ISD, using ISD as the reference group) into the models. Backward elimination with a criterion of P < .10 for retention was used to select a final model. The following variables were analyzed: patient age (>50 years or not),¹⁹  sex (female or male), WBC count at diagnosis (>50 × 10⁹/L or not), cytogenetic risk group (intermediate or high), and courses to achieving CR1 (>2 courses or not). The assumption of proportional hazards for each factor in the Cox model was tested. The test indicated that the proportionality assumptions hold. End-point events for DFS included morphologic relapse and NRM in CR1. Patients with MRD were not considered relapsed for DFS determination. Since conventional measures of leukemia-free survival (LFS) fail to take into account the capacity of DLI to produce durable molecular remissions in patients who relapse, thus underestimating the ability of allografting to produce sustained DFS even in patients who have relapsed, the current LFS probabilities were also estimated.²⁰  The SAS V. 9.3 (SAS Institute Inc., Cary, NC) and SPSS 13.0 software packages (SPSS Inc., Chicago, IL) were used for data analyses.

As shown in Table 1, the groups were well matched, except that patients undergoing HID-HSCT were significantly younger and there was a little delay from diagnosis to transplantation. Because of the 1-child policy in China over the past 30 years, parent donors accounted for 50% of the HID group. As a result, the HID cohort was younger. The phenomenon of 1 month longer to transplant in the HID vs the ISD group may be explained by the possible hesitation of patients, which is associated with unawareness arising from a lack of a doctor’s referral during chemotherapy in small institutes or insufficient confidence in HID transplantation. Patients undergoing HSCT using HIDs were mismatched at a median of 3 of 6 HLA-A, B, and DRB1 loci by molecular typing (range, 1-3).

The 3-year probabilities of DFS were 74% (95% CI, 67-81) and 78% (95% CI, 72-85; P = .34) among HID and ISD patients (Figure 1A). The 3-year probabilities of current LFS were 76% (95% CI, 64-87) and 80% (95% CI, 70-91) among HID and ISD patients. The 3-year probabilities of OS were 79% (95% CI, 73-85) and 82% (95% CI, 76-88; P = .36) among HID and ISD patients (Figure 1B). The 3-year cumulative incidences of NRM were 13% (95% CI, 7-19) and 8% (95% CI, 4-12; P = .13; Figure 1C). Subgroup analysis with available FLT-ITD results at diagnosis showed that the 3-year DFS rates were 79% (95% CI, 71-88) and 60% (95% CI, 42-78) among patients without FLT-ITD and patients with FLT-ITD (P = .02).

Univariate analysis treated the patient age in different ways: (1) as a continuous variable; and (2) in addition, a dichotomous variable of 50 years old was used as reported by Yanada et al¹⁹  for matched donor transplant. Univariate analysis showed that patient age did not affect DFS as a continuous variable (P = .39), whereas patient age over 50 years old significantly affected DFS (P = .03). This is likely because of the significantly higher proportion of WBC count at diagnosis as well as higher proportion of cytogenetically high-risk patients within the older group as compared with that in the younger group (45% vs 26%, P = .01; 30% vs 17%, P = .03, respectively). Of note, however, the differences in DFS rates between age groups differed according to donor source. DFS was not significantly different for patients over 50 years or not (62% vs 75%, P = .37) in the HID cohort, whereas DFS was significantly lower among those over 50 years than those under 50 years (61% vs 81%, P = .02) in the ISD group. The previously mentioned different proportion of disease characteristics and the more prominent differences within the ISD group are likely to have contributed to differences in DFS between the 2 age groups. In addition, the effect of patient age on DFS disappeared in multivariate analysis (P = .11).

A multivariate analysis failed to show significant differences in DFS, OS, or NRM rates between the 2 cohorts; cytogenetic risk group and WBC count at diagnosis were independent risk factors associated with DFS and OS, and patient age, sex, or courses required to achieving CR1 did not significantly affect mortality (Table 2).

The 3-year CIRs were 15% (95% CI, 9-21) and 15% (95% CI, 9-21; P = .98) among HID and ISD patients (Figure 1D). Multivariate analysis failed to show significant differences in CIR between the 2 cohorts; cytogenetic risk group and WBC count at diagnosis were independent risk factors affecting CIR (Table 2).

At Peking University, full donor chimerism was detected in 22 of the 23 (96%) patients at the time of preemptive DLI. Among the 23 patients (13 HID, 10 ISD) receiving modified DLI as intervention for positive MRD, 7 (4 of 13 HID and 3 of 10 ISD) eventually relapsed at a median of 194 days (range, 85-780 days) after DLI, and 15 patients were alive without relapse with a median of 940 days (range, 214-1349 days) after DLI; the remaining 1 patient died of infection.

Until the last follow-up, 26 and 27 patients experienced relapse after HID- or ISD-HSCT; of the 26 HID patients, 11 received DLI plus chemotherapy, 10 received chemotherapy alone, and 5 received no further therapy. In the DLI plus chemotherapy group, 6 patients achieved the second complete remission (CR2) and are still alive with a median of 323 days after relapse, whereas 3 died of relapse (76, 209, and 397 days after DLI) and the other 2 died of infection. The patients who received chemotherapy alone or no therapy died at a median of 60 days after relapse. Among ISD patients, 27 experienced relapse; of these, 7 received DLI plus chemotherapy, 8 received chemotherapy alone, and 12 received no further therapy. In the DLI plus chemotherapy group, 4 patients achieved CR2 and are still alive with a median of 844 days after relapse, whereas 3 died of relapse (30, 146, and 185 days after DLI). Among the 8 patients who received chemotherapy alone, 2 patients achieved CR2 and are still alive (94 and 194 days after relapse) and 6 died at a median of 36 days after relapse. Although the 12 patients receiving no further therapy died at a median of 30 days after relapse. In total, 10 of the 18 patients who underwent therapeutic DLI and 2 of the other 35 patients who did not undergo DLI achieved CR2 and are still alive (P < .001). Full donor chimerism was detected in 14 of the 18 (80%) patients at the time of targeted therapeutic DLI. The 1-year CIR after preemptive DLI or targeted therapeutic DLI was 22% and 33% (P = .42), respectively.

All patients achieved donor-cell engraftment. The median time to achieve neutrophil engraftment was 2 days shorter (P = .004; Figure 1E) after HID-HSCT, whereas that of platelet engraftment was 3 days shorter after ISD-HSCT (P < .001; Figure 1F).

The cumulative incidences of grades 2 to 4 acute GVHD at 100 days were 36% (95% CI, 30-42) and 13% (95% CI, 9-17) after HID- or ISD-HSCT, respectively (P < .001; Figure 1G); the cumulative incidences for grades 3 to 4 acute GVHD at 100 days were 10% (95% CI, 6-14) and 3% (95% CI, 1-5), respectively (P = .004). The cumulative incidences of chronic GVHD at 1 year were 42% (95% CI, 36-48) vs 15% (95% CI, 10-20) for HID vs ISD patients (P < .001; Figure 1H); the cumulative rates of severe chronic GVHD at 1 year were 12% (95% CI, 8-16) and 2% (95% CI, 0-4), respectively (P < .001).

To our knowledge, our study represents the first formal comparison of HID transplantation with allo-HSCT using ISDs in a disease-specific population of patients with intermediate- or high-risk AML in CR1. Although the study was nonrandomized, its strengths include the prospective nature, uniform treatment modality at the multiple centers, contemporaneous population of patients studied, and the number of patients analyzed. Furthermore, the comparison provides a particular opportunity for exploring the up-to-date undefined role of HID HSCT as postremission therapy for intermediate- or high-risk AML patients in CR1. Our study is the first to demonstrate that HID-HSCT is comparable to ISD-HSCT in the rate of DFS as postremission treatment of this portion of patients and the rates of OS and relapse were also similar in the 2 groups.

In our analysis, a 1.5-year NRM rate of 10% was achieved in the HID patients, which is comparable with the low 1-year NRM of 7% observed in studies of T-cell–replete HID HSCT using posttransplantation cyclophosphamide in nonmyeloablative settings.²¹  Our analysis demonstrates that low rates of NRM can be achieved among AML patients in CR1 undergoing both HID and ISD transplantation performed in multi-institutions. Nonetheless, both groups are relatively young, and this may have contributed to reduce the severity of NRM. It should also be noted that NRM was substantially higher with haploidentical transplants compared with those with matched related donors, although it did not reach statistical significance. Regarding GVHD, similar to our recently updated results,²²  the rates of GVHD observed after HID transplantation in our population were also comparable to those described in 53 HID transplantations using posttransplantation cyclophosphamide reported by Bashey et al.²³  However, the rates of GVHD observed after ISD transplantation in our population were much lower than those after nonmyeloablative ISD transplant described in Bashey’s report.²³  Recently, a multicenter phase 2 study from the Chinese Bone Marrow Transplant Cooperative Group showed that a combination of cyclosporine A, methotrexate, and mycophenolate mofetil for GVHD prophylaxis significantly decreased the risk of acute GVHD compared with historical controls in ISD transplantation.^24,25  This finding is confirmed in our ISD cohort. Thus, the higher incidence of GVHD in the HID group is partly because of the relatively low incidence in the ISD cohort. Furthermore, the GVHD rates in HID group were quite acceptable as discussed previously and the rates of GVHD-related death were similar between the 2 groups (data not shown) because of effective therapy. The CIR of 15% at 3 years in the HID group was similar to that of 12% reported in our previous study⁹  and to patients who underwent ISD-HSCT in the present study but lower than those reported in other studies.^26-28  Posttransplantation monitoring of MRD and modified DLI intervention guided by MRD level might have further decreased the relapse rate.^16,17 

Survival should be the primary end point for assessing the efficacy of a treatment modality for leukemia. Patient survival in this study was similar to that of our own previous study,⁹  but higher than that observed in other previous reports^26-28  because of a more favorable relapse rate and a much lower NRM. Several factors likely contributed to our superior transplantation outcomes. First, the patients in other studies have often included AML patients in CR1 undergoing HSCT with very high-risk features such as poor cytogenetic, high WBC count at diagnosis, delayed CR,^26-28  while patients in our present study with these high-risk features accounted for 56% of all study population. Second, as mentioned previously, younger age in our present study than in other studies (median age of 28 vs 33 to 37 in other haploidentical report^26-28  or 50 years old in ISD transplant), intensive GVHD prophylaxis, and effective infection control may have attributed to the much lower NRM rate (13% vs 32% to 36% at 3 years^26-28 ). Third, quantification of MRD after allo-HSCT is reportedly predictive of relapse, and individualized intervention can decrease relapse and improve survival in high-risk AML patients after transplantation. All the more, although the effects of HLA-matched allo-HSCT on patients with AML in CR1 are well established,^1-5  the application of haploidentical HSCT in AML needs to be broadened.^29,30  Our observations support the rationale for considering haploidentical HSCT as a front-line postremission treatment option for adults with AML without favorable cytogenetics who lack a matched donor. However, one should be particularly cautious in interpreting the results, given the age difference between the 2 groups. Although OS is similar, a matched sibling transplant should still be considered the standard of care, compared with haploidentical transplants since it has lower morbidity (GVHD) and NRM.

As a limitation to our current study, the patients were not randomized to receive HID- or ISD-HSCT for ethical and practical reasons. Although biologically selected postremission treatment with HID-HSCT or ISD-HSCT based on the availability of a matched donor might have minimized this limitation to some extent, there were still imbalanced features between the 2 groups (the HID cohort was younger and had longer time to transplant). In addition, insufficient confidence in HID transplant resulted in less HID transplant (maybe more patients with very high-risk features were tended to be “selected” into HID group as a result) than the number could be, although consecutive patients entering either transplant group were analyzed. We made adjustments in the multivariate analysis that produced no effects of age on DFS. This finding is consistent with those from our own previous studies and previous studies from other groups,^31-33  but it contradicts some reports.^19,34,35  We did left-truncation analysis to adjust for the time-to-transplant bias. Despite our effort on these adjustments, the possible influence of the imbalanced factors on comparable outcomes between the 2 study cohorts cannot be eliminated completely. Another limitation is that insufficient data on FLT3, comorbidities and NK cell alloreactivity limit our ability to incorporate these important variables into analysis, although subgroup results on FLT3 were inconsistent with those reported by Brunet et al.³⁶  Controversial effects of natural killer cell alloreactivity between the Europe and America study and our study as reviewed in detail by our group.^37,38 

In summary, this comparison suggests that outcomes after transplantation using HIDs with ATG are comparable to those after ISDs transplantation for AML patients in remission. HID transplantation as postremission therapy should be recommended as a valid alternative option for patients with intermediate- or high-risk AML in CR1 for whom no matched donor is available.

The authors thank Prof Bob Löwenberg, Prof Nancy Berliner, and Prof Jacob Rowe for kindly reviewing this paper; Dr Zhuang Tao (Beijing Medical Development Clinical Research Institute) for the careful statistical review; and the principal investigators and the skilled teams at each of the participating sites.

This work was partly supported by the Collaborative Innovation Center of Hematology China, the Key Program of National Natural Science Foundation of China (grant 81230013), the National Natural Science Foundation of China (grant 81400143), and the Scientific Research Foundation for Capital Medicine Development (grant 2011-4022-08).

Contribution: Y.W. and Q.-F.L. wrote the manuscript; X.-J.H. and D.-P.W. designed the protocol; all authors contributed patients, provided clinical and laboratory data, and revised and corrected the manuscript; Y.W. and X.-J.H. performed the analysis; and X.-J.H., Q.-F.L., and D.-P.W. approved and recommended the protocol within each institute.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Xiao-Jun Huang, Peking University People’s Hospital, Peking University Institute of Hematology, Beijing 100044, China; e-mail: xjhrm@medmail.com.cn; and De-Pei Wu, The First Affiliated Hospital of Soochow University, Suzhou, Jiangsu 215006, China; e-mail: wudepei@medmail.com.cn.