Improved survival for young acute leukemia patients following a new donor hierarchy for allogeneic hematopoietic stem cell transplantation: A phase III randomized controlled study
Mingming Zhang and Haowen Xiao contributed equally to this manuscript.
Abstract
The aim of our study was to evaluate the most optimal donor for young acute leukemia (AL) patients with multiple donors available for allogeneic hematopoietic stem cell transplantation (allo-HSCT), including HLA-matched sibling donors (MSDs), HLA-matched unrelated donors (URDs), haploidentical parental donors (HPDs), and haploidentical sibling donors (HSDs). From March 2008 to December 2016, 430 AL patients ≤ 35 years of age were included in the discovery, retrospective study. Patients who received transplantation from a MSD or a HSD had better 5-year OS rates compared with patients who received transplantation from a URD or a HPD. A superior graft-versus-leukemia effect was observed for high-risk patients undergoing HSD-HSCT with a lower relapse rate (p = 0.014) and a higher disease-free survival (DFS) rate (p = 0.029) compared with those undergoing MSD-HSCT. Outcomes of high-risk patients receiving an URD or HPD were equivalent. For intermediate/standard-risk patients, either a MSD or HSD may be the front-line donor selection with comparable outcomes. HLA-matched unrelated donors were preferred over HPDs with reduced non-relapse mortality and higher overall survival (OS) and DFS rates. We further conducted an independent prospective randomized study to evaluate the survival advantage with the new donor hierarchy. Two hundred and fifty patients were randomly assigned to follow our new donor hierarchy or the traditional donor hierarchy at a 2:1 ratio. The new donor hierarchy contributed to significantly superior 2-year OS and DFS rates (OS: 76.2% vs. 67.8%, p = 0.046; DFS: 71.8% vs. 64.5%, p = 0.039).
1 INTRODUCTION
The number of patients who receive allogeneic hematopoietic stem cell transplantation (allo-HSCT) has been on the rise with improved transplantation techniques and increased donor availability with the use of alternative graft sources. The traditional donor hierarchy is HLA-matched sibling donors followed by matched unrelated donors, then other donors (haploidentical donors or unrelated cord blood). Recent years have witnessed significant improvement in the outcomes of allo-HSCT from HLA-haploidentical donors (HIDs), which have achieved outcomes equivalent to the outcomes from HLA-matched sibling donors (MSDs) or unrelated donors (URDs).1-4 However, the traditional donor hierarchy has been challenged. There is particularly a concern for the patients who may have access to multiple donor sources, and selecting an optimal donor remains an area of ongoing study.
HLA-haploidentical donors are divided into several types according to family relationships, such as haploidentical parental donors (HPDs), haploidentical sibling donors (HSDs), and haploidentical offspring donors. According to the effect of characteristics of different types of HIDs on transplantation outcomes, a study from Peking University in China that focused on T-replete HID-HSCT using anti-thymocyte globulin (ATG) suggested that use of young, male, and non-inherited maternal antigen mismatched donors results in the best overall survival (OS) rates, so haploidentical offspring donors are the preferred donor type in HID-HSCT.5 The recommendations from the European Society for Blood and Marrow Transplantation (EBMT) on donor selection for T cell replete haploidentical transplants also preferred young male sibling or offspring haploidentical donors6; however, the majority of patients ≤ 35 years of age may lack the preferred donor type of HIDs and the haploidentical offspring may be too young to donate. Patients ≤ 35 years of age have haploidentical parental donors (HPDs), who may be older than URDs, may have haploidentical sibling donors (HSDs) within the same age group as URDs. A relevant question is whether or not outcomes are superior for young patients (≤ 35 years) with an older HPD (usually 2–3 decades older than the patient) or a young URD or HSD.
Furthermore, there are few studies that have compared the outcomes of patients with acute leukemia receiving allo-HSCT from a specific type of HIDs with conventional MSDs or URDs. Recently, Karam et al.7 analyzed 406 older allo-HSCT recipients with a median age of 54 years in haplo-HSCT with post-transplant cyclophosphamide (PT-Cy) platforms and concluded that patients undergoing transplantation with a younger HID (≤ 35 years) had a similar OS rate, lower rates of chronic GVHD (cGVHD), and better cGVHD-free, relapse-free survival compared with a MSD or URD ≥ 35 years of age. Wang et al.8 also suggested that kinship and donor age, rather than HLA disparity, predominantly influence survival in older acute leukemia patients undergoing haplo-HSCT in an ATG and granulocyte colony-stimulating factor (G-CSF)-based protocol; however, donor selection for young acute leukemia (AL) patients undergoing allo-HSCT has not been defined. To address this issue, we performed a two-stage cohort study including a retrospective study followed by an independent prospective randomized controlled study to develop a feasible hierarchy guiding the selection of the best donor in the context of young AL patients (≤ 35 years of age) with multiple donors available, including MSDs, URDs, HPDs, and HSDs.
2 METHODS
2.1 Patient population
From March 2008 to December 2019, all eligible patients were included if they fulfilled the following criteria: (1) patients ≤ 35 years of age. (2) being diagnosed with de novo AL according to WHO criteria (except acute promyelocytic leukemia and mixed phenotype acute leukemia). (3) being prepared to receive allo-HSCT. (4) having two or more donor sources available, lacking suitable haploidentical offspring donors. The exclusion criteria were: (1) AL secondary to myelodysplastic syndrome, chronic myeloid leukemia or myeloproliferative neoplasm. (2) treatment-related AL. (3) significant dysfunctions in vital organs. (4) ECOG performance status ≥ 3.
Cytogenetic abnormalities of AML were classified as favorable-risk, intermediate-risk or adverse-risk according to 2017 European LeukemiaNet (ELN) recommendations for diagnosis and management of AML in adults.9 Taking into account that there is no necessity for AML patients with favorable genetic abnormalities to receive allo-HSCT in the first complete remission (CR1). If AML patients with favorable genetic abnormalities experienced refractory to chemotherapy or relapse were further classified as intermediate-risk AML. The ALL patients were classified as standard- or high-risk according to previously published criteria.10, 11 High risk also was defined by failure to achieve CR after two cycles of induction chemotherapy, and/or patients in CR2 or beyond.12
2.2 Study design
We conducted a two-stage cohort study. First, a discovery, retrospective study was performed to compare outcomes of young AL patients undergoing allo-HSCTs from different types of donor at our center from March 2008 to December 2016. Patients were assigned to undergo MSD-HSCTs, or URD-HSCTs, or HID-HSCTs followed the traditional donor hierarchy. An MSD remains the preferred donor option. Patients without an MSD would consider a HLA-suitably-matched URD (≥ 8 of 10 matching HLA-A, HLA-B, HLA-C, HLA-DRB1, and HLA-DQB1 allele loci and ≥ 5 of 6 matching HLA-A, HLA-B, and HLA-DRB1 antigen loci) as the alternative donor. If a suitably matched URD was not available or lacked timely access to a suitably matched URD because of the patient's disease status or clinical circumstances, patients were permitted to receive HSCT from HIDs.
Then we developed a new donor hierarchy based on the results from the discovery, retrospective dataset. From January 2017 to December 2019, we further performed an independent prospective randomized controlled study to evaluate the survival advantage with the new donor hierarchy.
2.3 Sample size calculation in the prospective study
The estimation of sample size was performed using Power Analysis and Sample Size 11.0 software (NCSS, Kaysville, UT). The 5-year OS rates after transplantation were approximately 60%–70% for MSD-HSCTs, URD-HSCTs, or HID-HSCTs according to published reports from our center and other centers.1, 4 This present study proposed to improve the 5-year OS rate from 65% to 75% by following our new donor hierarchy. The settings included one-sided α = 0.05, β = 0.2, and a sample size of 231 patients, of which 154 patients were assigned to the experimental cohort and 77 patients to the control cohort at a 2:1 ratio, was determined to achieve the statistical power (80%) needed to detect significant differences between two cohorts after calculation. In consideration of sample lost to follow-up and other unexpected factors, an additional 5% sample size was added; the final sample size was approximately 243 patients.
All patients gave their signed informed consent. The study was approved by the Ethics Committee of The First Affiliated Hospital at Zhejiang University. The study was registered as ChiCTR2000029535 at www.chictr.org.cn.
2.4 Randomization and masking
Patients were screened for eligibility by study site staff before randomization (30–60 days pre-transplantation). Eligible patients were randomly allocated (2:1) to the experimental cohort (following the new donor hierarchy) or the control cohort (following the traditional donor hierarchy) according to the randomization principle after signing the informed consent form. This trial used a simple randomization method using an allocation table generated by SAS software (SAS Institute, Cary, NC), and a random allocation envelope was printed, which was generated independently of the study by an independent statistician. Furthermore, study staff who did the data collection and assessments of outcomes were masked to treatment allocations.
2.5 Transplantation procedures
The details of transplant procedures have been described previously.1 All patients received myeloablative conditioning involving busulfan and cyclophosphamide without total body irradiation. Our previously reported approach of T-cell-replete haploidentical HSCT with low-dose anti-T-lymphocyte globulin1 was used in all patients receiving HID-HSCTs. All patients received the same graft-versus-host disease (GVHD) prophylaxis, consisting of cyclosporin A (CSA), methotrexate (MTX), and low-dose mycophenolate mofetil (MMF). Rabbit antithymocyte globulin (ATG, Thymoglobulin, Genzyme, Cambridge, MA) was also administered to patients receiving URD-HSCTs (4.5–6 mg/kg total dose). Grafts were peripheral blood stem cells mobilized with granulocyte colony-stimulating factor without ex vivo T-cell depletion.
2.6 Study endpoints, definitions, and statistical analysis
The primary end point was OS (time from transplantation until death from any cause). The secondary end points were non-relapse mortality ([NRM] time from transplantation until death from any cause, with relapse as a competing event); relapse incidence (defined as bone marrow blasts >5%, reappearance of blasts in peripheral blood or development of extramedullary leukemia, with NRM as a competing event); disease-free survival ([DFS] time from transplantation until relapse or death from any other cause); and GVHD and GVHD-free, relapse-free survival ([GRFS] time from transplantation until relapse, death, severe acute GVHD [grades 3–4], or moderate/severe chronic GVHD). Acute GVHD was diagnosed and graded using established criteria.13 Chronic GVHD was classified as mild, moderate, or severe by the National Institutes of Health consensus criteria.14 Death without GVHD was considered a competing event for all GVHD assessments.
Cumulative incidence using the competing risk method was used to estimate the incidence of GVHD, relapse, and NRM. A Kaplan–Meier curve was used to estimate the probability of OS and DFS, and comparisons between cohorts were made using the log-rank method. The χ2 and Fisher exact tests were used to compare categorical variables among cohorts. The Kruskal-Wallis test was used to compare continuous variables among cohorts.
Multivariable Cox regression models were applied to analyze HSCT outcomes. A stepwise forward model was built for each outcome by selecting adjusted factors using a threshold of 0.1 for entry and 0.05 for retention. All variables and categorization as listed in Table 1 were considered in the stepwise model for all outcomes, with p < 0.05 considered significant for the covariates. Adjusted plots were created on the basis of the stratified Cox model15 for OS and DFS, and a subdistribution hazards model16 was created for cumulative incidences of NRM and relapse.
| Characteristics | MSD HSCT (n = 78) | URD HSCT (n = 136) | HPD HSCT (n = 134) | HSD HSCT (n = 82) | p |
|---|---|---|---|---|---|
| Patient age (median, range), years | 27 (12–35) | 22 (10–35) | 21 (9–35) | 26.5 (16–35) | <0.001 |
| Donor age (median, range), years | 27 (14–42) | 30 (20–50) | 45 (32–61) | 27.5 (15–45) | <0.001 |
| Underlying diseases, n(%) | 0.199 | ||||
| AML | 42 (53.8) | 56 (41.2) | 53 (39.6) | 34 (41.5) | |
| ALL | 36 (46.2) | 80 (58.8) | 81 (60.4) | 48 (58.5) | |
| Donor-patient gender, n (%) | <0.001 | ||||
| Female to male | 22 (28.2) | 14 (10.3) | 44 (32.8) | 25 (30.5) | |
| Others | 56 (71.8) | 122 (89.7) | 90 (67.2) | 57 (69.5) | |
| Donor-patient blood type relation, n (%) | <0.001 | ||||
| Identical | 50 (64.1) | 48 (35.3) | 78 (58.2) | 43 (52.4) | |
| Incompatibility | 28 (35.9) | 88 (64.7) | 56 (41.8) | 39 (47.6) | |
| Donor-patient HLA match, n (%) | NA | ||||
| 10/10 (HLA-A, B, C, DRB1, DQB1) | 63 (80.8) | 65 (47.8) | |||
| 9/10 | 37 (27.2) | 1 (1.2) | |||
| HLA-A mismatched | 6 (4.4) | ||||
| HLA-B mismatched | 4 (2.9) | ||||
| HLA-C mismatched | 12 (8.8) | ||||
| HLA-DRB1 mismatched | 5 (3.7) | ||||
| HLA-DQB1 mismatched | 10 (7.4) | ||||
| 8/10 | 10 (7.4) | 4 (3.0) | 4 (4.9) | ||
| HLA-A and –C mismatched | 2 (1.5) | ||||
| HLA-A and –DQB1 mismatched | 1 (0.7) | ||||
| HLA-B and –C mismatched | 1 (0.7) | ||||
| HLA-B and –DQB1 mismatched | 1 (0.7) | ||||
| HLA-C and –DQB1 mismatched | 3 (2.2) | ||||
| HLA-DRB1 and -DQB1 mismatched | 2 (1.5) | ||||
| 7/10 | 22 (16.4) | 7 (8.5) | |||
| 6/10 | 27 (20.1) | 16 (19.5) | |||
| 5/10 | 66 (49.3) | 44 (53.7) | |||
| 6/6 (HLA-A, B, DRB1) | 15 (19.2) | 24 (17.6) | |||
| 5/6 | 1 (0.7) | 1 (1.2) | |||
| 4/6 | 4 (3.0) | 1 (1.2) | |||
| 3/6 | 10 (7.5) | 8 (9.8) | |||
| Months from diagnosis to HCT (median, range) | 7 (4–96) | 8 (4–108) | 8 (4–120) | 8 (4–130) | 0.515 |
| Disease status at transplant, n (%) | 0.448 | ||||
| CR1 | 53 (67.9) | 102 (75.0) | 83 (61.9) | 57 (69.5) | |
| ≥CR2 | 19 (24.4) | 25 (18.4) | 40 (29.9) | 20 (24.4) | |
| Advanced stage | 6 (7.7) | 9 (6.6) | 11 (8.2) | 5 (6.1) | |
| Risk stratification, n (%) | 0.607 | ||||
| Intermediate /standard risk | 31 (39.7) | 53 (39.0) | 43 (32.1) | 30 (36.6) | |
| High risk | 47 (60.3) | 83 (61.0) | 91 (67.9) | 52 (63.4) | |
| Median CD34+ count, ×106/kg (range) | 6.26 (0.72–22.71) | 4.85 (0.34–29.64) | 4.92 (1.28–28.7) | 5.2 (1.06–20.63) | 0.065 |
To adjust for multiple testing of the donor variable for several outcomes, p < 0.008 (0.05/6) for the donor variable was considered statistically significant for direct pairwise comparisons between multiple donor types. All p values are two-sided. Data were analyzed by using SAS software version 9.4 (SAS Institute, Cary, NC).
3 RESULTS
3.1 Outcomes of allo-HSCT following the traditional donor hierarchy in the discovery, retrospective study
The retrospective analysis involved 430 consecutive younger patients with de novo acute leukemia who received a transplant from an MSD (n = 78), HLA-matched URD (n = 136), or an HLA-haploidentical parental donor (HPD, n = 134), or an HLA-haploidentical sibling donor (HSD, n = 82). In the URD-HSCT cohort, 65 and 24 patients received a 10/10 or a 6/6 HLA-matching donor, respectively. Thirty-seven and 10 patients received a 9/10 or a 8/10 HLA-matching donor, respectively. Data from 189 consecutive younger patients were previously reported and further followed in this study.1 The MSD-HSCT, URD-HSCT, HPD-HSCT, and HSD-HSCT cohorts were well-matched and there were no significant differences with respect to underlying disease, disease status at the time of transplantation, disease risk stratification, and time from diagnosis-to- transplantation. The median donor age for HPD-HSCTs was 45 years (range, 32–61 years), which was significantly older than the MSD-HSCTs (median age, 27 years; age range, 14–42 years), URD-HSCTs (median age, 30 years; age range, 20–50 years), and HSD-HSCTs (median age, 27.5 years; age range, 15–45 years) (p < 0.001). Based on the donor-patient gender relationship and blood type match, there were fewer matching female donors with a male patient, but more donor-patient ABO blood group incompatibility in the URD-HSCT cohort compared with the MSD-HSCT, HSD-HSCT, and HPD-HSCT cohorts (p < 0.001) (Table 1).
The end point of the last follow-up for all the survivors was 30 June 2019. The median follow-up of all the survivors was 68.2 months (range, 30.3–143.4 months).
3.2 Overall survival, disease-free survival and GVHD-free, relapse-free survival
The 5-year OS rates were 74.7% (95% CI, 65.3–85.4), 62.3% (95% CI, 54.6–71), 57.4% (95% CI, 49.3–66.9), and 80.9% (95% CI, 72.6–90.2) for MSD-HSCT, URD-HSCT, HPD-HSCT, and HSD-HSCT cohorts, respectively. The 5-year DFS rates were 68.5% (95% CI, 58.7–79.9), 59.3% (95% CI, 51.6–68.2), 55.2% (95% CI, 47.1–64.8), and 80.5% (95% CI, 72.4–89.5) for MSD-HSCT, URD-HSCT, HPD-HSCT, and HSD-HSCT cohorts, respectively (Table S1).
In multivariable analysis, patients receiving HSCT from a URD (for OS: HR, 2.58; 95% CI, 1.42–4.68; p = 0.002; for DFS: HR, 2.41; 95% CI, 1.38–4.23; p = 0.002) or a HPD (for OS: HR, 2.83; 95% CI, 1.56–5.16; p = 0.001; for DFS: HR, 2.63; 95% CI, 1.49–4.64; p = 0.001) were associated with significantly inferior OS and DFS as shown in Figure S1A,B and detailed in Table 2. Furthermore, patients with lymphoblastic malignances have inferior OS and DFS (for OS: HR, 1.62; 95% CI, 1.14–2.3; p = 0.008; for DFS: HR, 1.57; 95% CI, 1.13–2.2; p = 0.008). In the direct pairwise comparisons in the same multiple testing model, we found no difference in OS and DFS between the MSD- and HSD-HSCT cohorts (for OS: p = 0.364; for DFS: p = 0.126). There was also no difference in OS and DFS between the URD- and HPD-HSCT cohorts (for OS: p = 0.657; for DFS: p = 0.672).
| Variablea | HR (95% CI) | p | |
|---|---|---|---|
| Grades II–IV aGVHD | |||
| Patient ageb | 0.98 (0.95–1.01) | 0.16 | |
| Donor-patient gender | Female to male | 1.49 (1.01–2.19) | 0.046 |
| Others | |||
| Donor-patient bloodtype relation | Incompatibility | 1.29 (0.91–1.81) | 0.15 |
| Identical | |||
| Donor type | HSD | 0.39 (0.2–0.74) | 0.004 |
| HPD | 2.3 (1.24–4.25) | 0.008 | |
| URD | 2.68 (1.44–5) | 0.002 | |
| MSD | 1 | ||
| Relapse | |||
| Patient ageb | 1.0 (0.97–1.03) | 0.98 | |
| Underlying disease | ALL | 1.69 (1.12–2.57) | 0.013 |
| AML | |||
| Donor-patient gender | Female to male | 0.65 (0.4–1.05) | 0.075 |
| Others | |||
| Disease status at HSCT | Advanced stage | 4.03 (2.29–7.09) | <0.001 |
| ≥CR2 | 1.9 (1.21–2.98) | 0.006 | |
| CR1 | 1 | ||
| Donor-patient bloodtype relation | Incompatibility | 1.02 (0.7–1.48) | 0.94 |
| Identical | |||
| NRM | |||
| Patient ageb | 1.01 (0.95–1.07) | 0.82 | |
| Donor-patient gender | Female to male | 2.79 (1.46–5.36) | 0.002 |
| Others | |||
| Donor-patient blood type relation | Incompatibility | 0.8 (0.47–1.38) | 0.42 |
| Identical | |||
| Donor type | HSD | 0.7 (0.12–4.04) | 0.69 |
| HPD | 7.28 (1.64–32.3) | 0.008 | |
| URD | 8.65 (2.18–34.3) | 0.002 | |
| MSD | 1 | ||
| DFS | |||
| Patient ageb | 1.01 (0.98–1.04) | 0.451 | |
| Donor-patient gender | Female to male | 1.03 (0.71–1.5) | 0.863 |
| Others | |||
| Donor-patient blood type relationship | Incompatibility Identical |
0.93 (0.67–1.29) | 0.652 |
| Underlying disease | ALL | 1.57 (1.13–2.2) | 0.008 |
| AML | |||
| Disease status at HSCT | Advanced stage | 2.82 (1.73–4.6) | <0.001 |
| ≥CR2 | 1.63 (1.13–2.35) | 0.009 | |
| CR1 | 1 | ||
| Donor type | MSD | 1.63 (0.87–3.05) | 0.126 |
| URD | 2.41 (1.38–4.23) | 0.002 | |
| HPD | 2.63 (1.49–4.64) | 0.001 | |
| HSD | 1 | ||
| OS | |||
| Patient ageb | 1.02 (0.99–1.05) | 0.312 | |
| Donor-patient gender | Female to male | 1.0 (0.68–1.49) | 0.992 |
| Others | |||
| Donor-patient blood type relationship | Incompatibility | 0.85 (0.61–1.2) | 0.366 |
| Identical | |||
| Underlying disease | ALL | 1.62 (1.14–2.3) | 0.008 |
| AML | |||
| Disease status at HSCT | Advanced stage | 2.56 (1.51–4.34) | 0.001 |
| ≥ CR2 | 1.67 (1.14–2.45) | 0.009 | |
| CR1 | 1 | ||
| Donor type | MSD | 1.37 (0.69–2.71) | 0.364 |
| URD | 2.58 (1.42–4.68) | 0.002 | |
| HPD | 2.83 (1.56–5.16) | 0.001 | |
| HSD | 1 | ||
- a The donor types and clinical factors included in the multivariate analysis were stated in the table.
- b Patient age was treated as continuous variable.
We further performed a risk stratification-directed subgroup analysis to explore the effect of donor sources on the outcomes of allo-HSCT. After controlling for intermediate/standard-risk AL (n = 96) and high-risk AL patients (n = 99), a superior graft-versus-leukemia (GVL) effect was demonstrated for high-risk AL patients undergoing HSD-HSCTs with lower relapse rates (HR, 0.37; 95% CI, 0.16–0.82; p = 0.014) and higher DFS rates (HR, 0.76; 95% CI, 0.59–0.97; p = 0.029) compared with patients undergoing MSD-HSCTs. According to intermediate/standard-risk AL patients, better OS and DFS rates were achieved in patients receiving HSCT from a URD compared with patients from an HPD (for OS: HR, 0.41; 95% CI, 0.2–0.85; p = 0.017; for DFS: HR, 0.43; 95% CI, 0.21–0.86; p = 0.018) because of lower NRM rate (p = 0.028) (Table S2).
The 5-year GRFSs were 57% (95% CI, 46.9–69.4), 44.6% (95% CI, 36.9–53.8), 39.6% (95% CI, 32–49.1), and 50% (95% CI, 40.3–62.1) for the MSD, URD, HPD and HSD cohorts, respectively (Table S1). The patients in the MSD cohort had a superior 5-year GRFS rate compared with patients in the HPD cohort (p = 0.008) and exhibited a trend of a superior 5-year GRFS rate compared with patients in the URD cohort (p = 0.065). The patients in the HSD cohort also achieved a trend of a superior 5-year GRFS rate than patients in the HPD cohort (p = 0.087).
3.3 Component end points of aGVHD, cGVHD, NRM and relapse
In multivariable analysis, we found that alternative donors had an adverse impact on the risk of grades II-IV aGVHD (URDs: HR, 2.68; 95% CI, 1.44–5; p = 0.002; HPD: HR, 2.3; 95% CI, 1.24–4.25; p = 0.008; HSDs: HR, 0.39; 95% CI, 0.2–0.74; p = 0.004). Transplantation from a female donor to a male patient was also found to be a significant risk factor for grades II-IV aGVHD (HR, 1.49; 95%CI, 1.01–2.19; p = 0.046) (Table 2). The incidences of grades II–IV aGVHD were similar among the URD-HSCT, HPD-HSCT and HSD-HSCT cohorts (36%, 34.3%, 36.6%, respectively) (Table S1).
The cumulative incidences of cGVHD at 5 years were 31.3%, 39.4%, 42.3%, and 47.6% for patients receiving transplantations from MSDs, URDs, HPDs, or HSDs, respectively. The cumulative incidences of moderate-to-severe cGVHD at 5 years were 15.5%, 22.3%, 20.2%, and 30.5% for MSD-HSCT, URD-HSCT, HPD-HSCT, and HSD-HSCT cohorts, respectively. No significant difference was found in the risk of cGVHD among different donor sources (Table S1).
The NRM rates were significantly influenced by donor type. The 5-year NRM rates were 3.3%, 15.6%, 17.5%, and 3.7% for the MSD-HSCT, URD-HSCT, HPD-HSCT, and HSD-HSCT cohorts, respectively (Table S1). In multivariable analysis, the patients in the MSD-HSCT or HSD-HSCT cohorts achieved lower 5-year NRM rates compared with those in the URD-HSCT or HPD-HSCT cohorts (URDs: HR, 8.65; 95% CI, 2.18–34.3; p = 0.002; HPD: HR, 7.28; 95% CI, 1.64–32.3; p = 0.008) (Table 2, Figure S1C).
In multivariable analysis, lymphoblastic malignances (HR, 1.69; 95% CI, 1.12–2.57; p = 0.013), and disease status at transplantation (≥CR2: HR, 1.9; 95% CI, 1.21–2.98; p = 0.006; advanced stage: HR, 4.03; 95% CI, 2.29–7.09; p < 0.001) contributed to a significantly higher risk of relapse for all patients (Table 2). In the direct pairwise comparison, we found no significant difference in 5-year relapse rates among the four donor sources (MSDs, 28.2%; URDs, 25.1%; HPDs, 27.3%; and HSDs, 15.9%; Figure S1D).
3.4 Application of the new donor hierarchy to young patients in a prospective phase 3 clinical trial
3.4.1 Development of a new donor hierarchy
Based on results from our discovery, retrospective dataset, we developed a new donor hierarchy for young patients with AL who had multiple donor sources available for allo-HSCT (Figure 1A). In brief, for high-risk AL patients, an HSD is considered to be the most preferred donor option followed by an MSD. Based on URDs versus HPDs, the outcomes of high-risk AL patients undergoing URD-HSCTs were equivalent to patients undergoing HPD-HSCTs. While for intermediate /standard-risk AL patients, an MSD or HSD may be the first-line donor selection because of comparable transplantation outcomes from an MSD or HSD. Based on URDs versus HPDs, URDs were preferred over HPDs because of reduced NRM and better survival rates.
3.4.2 Patient demographics and baseline characteristics
Between January 2017 and December 2019, 264 young patients with AL undergoing their first allo-HSCT were screened at enrolment, of which 250 patients including 170 high-risk AL and 80 intermediate/standard-risk AL patients were finally enrolled and randomly assigned to the experimental cohort and control cohort at a 2:1 ratio, respectively. A CONSORT flow diagram starting with subjects eligible a transplant is shown in Figure S2. The experimental cohort including 113 high-risk AL patients and 53 intermediate/standard-risk AL patients used our new donor selection option, while 57 high-risk AL patients and 27 intermediate/standard- risk AL patients in the control cohort followed the traditional donor hierarchy. The experimental and control cohorts were well balanced with regard to potential prognostic factors, for example, cytogenetic and genetic risk category, disease status at transplantation (in CR1 vs. outside CR1), the donor-recipient gender relationship and blood type match, and the level of HLA matching of donor-patient in the URD-HSCT cohort (Table 3).
| Characteristics | Experimental cohort (n = 166) | Control cohort (n = 84) | p |
|---|---|---|---|
| Patient age (median, range), years | 29 (11–35) | 27 (10–35) | 0.264 |
| Donor age (median, range), years | 30.5 (11–55) | 36 (18–55) | <0.001 |
| Underlying diseases, n (%) | 0.284 | ||
| ALL | 79 (47.6) | 46 (54.8) | |
| AML | 87 (52.4) | 38 (45.2) | |
| Genetics, n (%) | NA | ||
| AML | 87 | 38 | |
| Mutated FLT3-ITD, wild type NPM1 | 10 (11.5) | 5 (13.2) | |
| Mutated FLT3-ITD and NPM1 | 2 (2.3) | 2 (5.3) | |
| Mutated NPM1 without FLT3-ITD | 3 (3.4) | 0 | |
| Mutated RUNX1 and/or ASXL1 | 4 (4.6) | 0 | |
| Mutated TP53 | 1 (1.1) | 2 (5.3) | |
| Biallelic mutated CEBPA | 4 (4.6) | 1 (2.6) | |
| KMT2A rearranged except t(9;11) | 4 (4.6) | 2 (5.3) | |
| CBFB-MYH11 | 2 (2.3) | 2 (5.3) | |
| RUNX1-RUNX1T1 | 9 (10.3) | 6 (15.8) | |
| -7 | 1 (1.1) | 0 | |
| Complex karyotype | 2 (2.3) | 1 (2.6) | |
| ALL | 79 | 46 | |
| BCR-ABL1 | 27 (34.2) | 6 (13) | |
| KMT2A rearranged | 3 (3.8) | 2 (4.3) | |
| Complex karyotype | 4 (5.1) | 1 (2.2) | |
| Hypodiploidy | 0 | 1 (2.2) | |
| Donor-patient gender, n (%) | 0.43 | ||
| Female to male | 41 (24.7) | 17 (20.2) | |
| Others | 125 (75.3) | 67 (79.8) | |
| Donor-patient blood type relation, n (%) | 0.656 | ||
| Identical | 84 (50.6) | 40 (47.6) | |
| Incompatibility | 82 (49.4) | 44 (52.4) | |
| Months from diagnosis to HCT (median, range) | 6.9 (4–49.1) | 7.2 (3.4–33.2) | 0.775 |
| Disease status at transplant, n (%) | 0.899 | ||
| CR1 | 126 (75.9) | 62 (73.8) | |
| ≥CR2 | 24 (14.5) | 14 (16.7) | |
| Advanced stage | 16 (9.6) | 8 (9.5) | |
| Risk stratification, n (%) | 0.973 | ||
| Intermediate/standard risk | 53 (31.9) | 27 (32.1) | |
| High risk | 113 (68.1) | 57 (67.9) | |
| Median CD34+ count, ×106/kg (range) | 5.26 (1.79–16.7) | 5.73 (1.4–18.84) | 0.816 |
| Donor type | <0.001 | ||
| MSD | 38 (22.9) | 16 (19) | |
| URD | 9 (5.4) | 29 (34.5) | |
| HPD | 28 (16.9) | 27 (32.1) | |
| HSD | 91 (54.8) | 12 (14.3) | |
| Donor-patient HLA match in URD, n (%) | 9 URDs | 29 URDs | 0.727 |
| 10/10 matched | 5 (55.6) | 18 (62.1) | |
| Mismatched | 4 (44.4) | 11 (37.9) | |
| 9/10 matched | |||
| HLA-A mismatched | 1 (11.1) | 0 | |
| HLA-B mismatched | 0 | 2 (6.9) | |
| HLA-C mismatched | 0 | 8 (27.6) | |
| HLA-DRB1 mismatched | 1 (11.1) | 0 | |
| HLA-DQB1 mismatched | 1 (11.1) | 1 (3.4) | |
| 8/10 matched | |||
| HLA-C and HLA-DQB1 mismatched | 1 (11.1) | 0 |
The end point of the last follow-up evaluation for all survivors was June 30, 2020. The median follow-up of all the survivors was 24.8 months (range, 6.1–41.8 months).
3.4.3 Efficacy of the new donor hierarchy
GVHD
The incidences of grades II–IV aGVHD were similar between the experimental cohort (24.7%, 95% CI, 18.4–31.5) and the control cohort (26.2%, 95% CI, 17.3–36) (p = 0.874). We also found no difference in the 2-year cumulative incidence of cGVHD between patients followed the new donor selection option (36.3%, 95% CI, 28.4–44.3) and those followed the traditional donor hierarchy (32.4%, 95% CI, 21.6–43.7) (p = 0.44). The 2-year cumulative incidence of moderate to severe cGVHD was also comparable (17.8% vs. 15.2%, p = 0.568).
Transplant outcomes
The estimated probability of 2-year OS rate for the patients in the experimental cohort (76.2%, 95% CI, 69.1–84) was significantly superior to the control cohort (67.8%, 95% CI, 57.7–79.7) (p = 0.046, Figure 1B). The 2-year DFS rate of the patients in the experimental cohort was 71.8% (95% CI, 64.4–80.1), which was also significantly superior to the control cohort (64.5%, 95% CI, 54.6–76.1) (p = 0.039) (Figure 1C). A lower NRM rate of patients in the experimental cohort may contribute to the superior survival compared with the control cohort. At 2 years, the cumulative incidence of NRM was 6.3% (95% CI, 2.8–11.6) for those assigned to the experimental cohort and 12.9% (95% CI, 6.5–21.5) for those allocated to control (p = 0.036, Figure 1D), while the relapse rate was not significantly different (22% vs. 22.7%; p = 0.398, Figure 1E). No significant difference was noted in the 2-year GRFS rates between the experimental cohort (52.6%, 95% CI, 44.8–61.8) and the control cohort (48.8%, 95% CI, 38.5–61.8) (p = 0.251).
4 DISCUSSION
Donor selection is always one of the most important decisions for success of allo-HSCT. The traditional donor hierarchy based on the status of donor-patient HLA matching has been well-established. Recent advances in GVHD prophylaxis and evolving practice in HID-HSCTs have resulted in a narrowing gap in outcomes between alternative HLA mismatched donors and conventional HLA-matched donors. Recently, a retrospective analysis of 1205 adult CR1 AML patients reported to the Center for International Blood and Marrow Transplant Research (CIBMTR) database indicated that a lower rate of cGVHD after PT-Cy–based haplo-HSCT versus MSD using calcineurin inhibitor–based GVHD prophylaxis, but similar survival, relapse, and NRM, suggesting that HIDs is a viable alternative to MSD in these patients.17 To date, published retrospective studies involving a comparison of outcomes of allo-HSCT using HIDs with allo-HSCT using conventional MSDs or URDs in patients with hematologic malignancies all suggested that unmanipulated haploidentical family donor transplants using ATG or PT-Cy are an additional option for patients lacking a MSD.1-4, 18, 19 Further prospective randomized clinical trials also confirmed that HID-HSCTs achieve outcomes similar to MSD-HSCTs.20, 21 Even according to patients lacking an HLA-matched related or unrelated donor, HIDs may be the best choice of alternative donors. Recently, Mehta RS et al. compared composite GRFS and cGVHD-free relapse-free survival (CRFS) in 2198 patients reported to the CIBMTR database, who underwent umbilical cord blood-mismatched, HID-mismatched, HLA-mismatched (7/8) bone marrow or peripheral blood URD-HSCTs, and these data support the HIDs had the best GRFS, CRFS, and OS of all groups.22 The detrimental effect of HLA mismatch has been reduced or eliminated. In the absence of a negative effect of increasing HLA mismatch on outcomes, other non-HLA donor characteristics, including patient-donor age, patient-donor gender, and probability of disease recurrence, have gained more attention, which may change the algorithm for donor selection.
There are several published studies involving donor selection to compare the effect of HLA mismatch and other non-HLA factors influencing alloreactivity involving donor-patient age and the donor-patient gender relationship. A retrospective study from the EBMT involving 1066 male adult patients with de novo AML undergoing a first allo-HSCT in CR1 using a female MSD or a T-replete haploidentical male or female donor, suggested that in men with intermediate-risk AML, allo-HSCT from a sister MSD or a haploidentical donor produces similar GRFS. In men who have high-risk AML, a haploidentical donor may be a better choice.23 To date, there were four published studies involving donor options for old patients receiving allo-HSCT. Robinson et al.24 compared 2143 donor-recipient pairs (n = 218 haploidentical sibling; n = 218 offspring; n = 1707 HLA-matched sibling) with acute leukemia and concluded that in patients 18–54 years of age, there was comparable survival after transplant from an HLA-matched sibling and a haploidentical sibling. A superior outcome was reported in patients 55–76 years of age when using a HLA-matched sibling instead of offspring. In contrast, Wang et al.8 suggested that offspring may be the preferred donor compared with MRDs for acute leukemia patients ≥ 50 years of age based on the lower 3-year TRM and relapse incidence and higher OS and LFS. Perales MA et al. studied whether survival after haploidentical transplantation is comparable to that after matched URD transplantation for 822 patients aged 50–75 years with AML in CR1 or CR2. One hundred and ninety two patients received grafts from haploidentical donors with the media age of 37 years (sibling 25%; offspring 75%) and 631 patients from matched URDs with the media age of 27 years (range 18–40 years). Their data support the view that matched URD transplant with donors younger than 40 years is to be preferred.25 There is only one published study involving donor selection for old patients among URDs, MRDs, and HIDs, in which the outcomes of 406 allo-HSCT recipients with a median age of 54 years after MRD-HSCT with a donor age ≥ 35 years, URD-HSCT with a donor age ≥ 35 years, and HID-HSCT with a donor age ≤ 35 years were compared. The results showed that recipients of HID-HSCT from a young donor ≤ 35 years of age had a similar OS rate, lower rates of cGVHD, and better GRFSs compared with patients undergoing transplantation with a MRD or URD donor ≥ 35 years of age.7 Hence, for old patients receiving allo-HSCT, the donor age may be more important to transplantation outcomes than HLA disparity, which need to be confirmed in more homogeneous cohort of old patients.
Our study so far is the first prospective study to develop a donor selection strategy for the array of all possible donor options for young patients receiving allo-HSCT. The results of this study go beyond the current donor recommendation in allo-HSCT and provide a new feasible approach for donor selection in clinical practice. The results of our trial suggest three important aspects of donor selection for young acute leukemia patients. First, in young allo-HSCT recipients, the impact of donor age also appears to be more relevant to transplantation outcomes than HLA disparity. Patients who received transplantation from an MSD or HSD had superior 5-year OS and DFS rates compared with patients receiving an HPD because of the lower incidence of NRM. Based on patients with intermediate/ standard-risk leukemia, URD-HSCTs also produced better OS and DFS rates than HPD-HSCTs, which may be attributed to the effect of donor age. Because the majority of URDs in donor registries worldwide are <40 years of age, the donors may be younger than HPDs. Advanced donor age has been shown to correlate with unfavorable transplant outcomes in URD-HSCTs26, 27 and HID-HSCTs.28-30 Although the pathophysiology of an unfavorable effect of older donors has yet to be elucidated, there were several factors that may be related including gradually reduced self-renewal and regenerative potential of HSCs,31 clonal hematopoiesis of indeterminate potential,32, 33 age-associated changes in the immune system.34, 35
The second important aspect of our donor selection is development of a relapse- risk stratification–directed donor selection algorithm. Data from our retrospective analysis showed that patients receiving HSCT from an MSD or HSD can achieve equivalent 5-year OS and DFS; however, after controlling for high-risk patients, a superior GVL effect was observed in high-risk patients undergoing HSD-HSCT with lower relapse and higher DFS rates compared to patients undergoing MSD-HSCTs. The priority use of T-cell-replete HID-HSCT with ATG or PT-Cy in high-risk patients has been reported by our group and other groups1, 23, 36; however, according to our study no advantage of HID-HSCTs was observed in intermediate/standard-risk patients for HSDs or HPDs.
Finally, this was a homogenous population of patients who received a consistent allo-HSCT protocol assembled for analysis of donor selection by a retrospective analysis to discovery combined a prospective trial to validation study. The superiority of our new donor selection strategy was strengthened by being validated in an independent prospective data, showing that 2-year OS and DFS rates of the patients following the new donor selection strategy were significantly better than those following the traditional donor selection.
However, it is also worth noting that the follow-up after the new donor selection strategy to our case is short and needs longer follow-up and a larger number and more homogeneous of patient population to confirm the efficiency of the new donor selection strategy. We also admitted that we could not rule out the possibility that if URD and HID patients had received the same GVHD prophylaxis strategy, they could have yielded better outcomes. Future clinical studies are also warranted to determine the possibility of further improvement of URD and HID results with regular doses of ATG or high-dose cyclophosphamide post-transplant.
ACKNOWLEDGMENTS
This work was supported in part by grants from the National Natural Science Foundation of China (81730008, 81800178, and 81870136), and the Key Research and Development Program of Zhejiang Province (2019C03016).
CONFLICT OF INTEREST
The authors have no conflicting financial interests.
AUTHOR CONTRIBUTIONS
Conception and design: Mingming Zhang, Haowen Xiao, Yi Luo, He Huang. Collection, analysis and interpretation of the data: Mingming Zhang, Haowen Xiao, Jimin Shi, Yi Luo, He Huang, Yamin Tan, Yanmin Zhao, Jian Yu, Xiaoyu Lai, Yongxian Hu, Weiyan Zheng. Drafting the article: Haowen Xiao, Mingming Zhang, He Huang. Provision of study materials or patients: Mingming Zhang, Yi Luo, Jimin Shi, Yamin Tan, Yanmin Zhao, Jian Yu, Xiaoyu Lai, Yongxian Hu, Weiyan Zheng, He Huang. Obtaining of funding: Mingming Zhang, Haowen Xiao, He Huang.
Open Research
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.
