This study was performed retrospectively in eight hospitals of South Korea from 2005 to 2015. In total, 352 patients suffering from acute leukemia or myelodysplastic syndrome (MDS) who underwent HSCT from HLA-mismatched unrelated or haploidentical family donors were included in the study. All patients received busulfan-based conditioning regimen, including busulfan plus cyclophosphamide or busulfan plus fludarabine. Patients who had been treated using total body irradiation (TBI)-based conditioning regimens were not included in this study. The pre-transplant disease status was divided into first or second complete remission (CR1 or CR2), refractory or relapsed, and untreated status. The definition of CR followed the recommended criteria [11]. To evaluate the degree of HLA matching, HLA A, B, C, and DR were identified using polymerase chain reaction with sequence specific primer followed by high-resolution typing, and the degree of HLA matching was classified as 1 antigen mismatch, 2 antigen mismatches, 3 antigen mismatches, and 4 antigen mismatches. Stem cell sources, such as bone marrow or peripheral blood were identified and CD34+ cell counts (× 10⁶/kg) were measured to evaluate the amount of stem cells. This study was approved by Institutional Review Board (IRB) of Kosin University Gospel Hospital (approval number 201809011) and the IRBs granted a waiver of informed consent because this study was a retrospective analysis that involved no more than minimal risk for patients.

Busulfan and fludarabine (BuFlu) or busulfan and cyclophosphamide (BuCy) were used as conditioning regimens, including Bu4Cy (busulfan [3.2 mg/kg] for 4 days along with cyclophosphamide [60 mg/kg] for 2 days), Bu-4Flu (busulfan [3.2 mg/kg] for 4 days along with fludarabine [30 mg/kg] for 6 days), Bu3Flu (busulfan [3.2 mg/kg] for 3 days along with fludarabine [30 mg/kg] for 6 days), Bu2Flu (busulfan [3.2 mg/kg] for 2 days along with fludarabine [30 mg/kg] for 6 days), and Bu1Flu (busulfan [3.2 mg/kg] for 1 day along with fludarabine [30 mg/kg] for 6 days). Bu4Cy and Bu4Flu were defined as myeloablative conditioning (MAC), and Bu3Flu, Bu2Flu, and Bu1Flu were defined as RIC. Rabbit ATG (thymoglobulin, Sanofi-Aventis, Cambridge, MA, USA) was administered to all patients at various dosages to prevent GVHD. We classified patients who did not receive ATG as no ATG group (0 mg/kg) and those who received 2.5 to 7.5 mg/kg (less than 8 mg/kg) as low to medium ATG group. Patients who received 9 to 12 mg/kg ATG were classified as high ATG group. ATG was given as an equally divided dose for 2 or 3 days from day –3. All patients received calcineurin inhibitors, including cyclosporine or tacrolimus with or without short-term methotrexate, as immunosuppressants to prevent GVHD.

Survival probabilities were estimated using the Kaplan-Meier method and differences in survival distributions were compared using the log-rank test. Overall survival (OS) rates were calculated from the time of HSCT to the date of death or last follow-up. Event-free survival (EFS) rates were defined from the time of HSCT to the date of relapse, progression, graft failure, donor lymphocyte infusion, or death from any cause. To identify factors affecting OS and EFS, the log rank test and Cox proportional hazard model were used in univariate and multivariate analysis, respectively. ATG dose and other variables having p < 0.05 in the univariate analysis were included for multivariate analysis. We included ATG doses as a variable in multivariate analysis to validate the clinical impact of ATG dose on survival outcomes and GVHD incidence, although there were no significant differences with respect to ATG doses in univariate analysis. Consensus grading criteria was used to assess the degree of aGVHD [12]. The cumulative incidence rate (CIR) of aGVHD was calculated through day 100. A cGVHD was assessed using the traditional grading criteria [13]. For all analyses, the p values were two-sided, and p < 0.05 was considered statistically significant. The statistical analyses were performed using SPSS version 24.0 software (IBM Corp., Armonk, NY, USA).

The CIR of relapse and GVHD were generated using confidence interval (CI) estimates adjusting for competing risks and compared by Gray’s method. For the analysis of CIR of relapse, patients who were alive without relapse were censored.

This study enrolled patients with acute leukemia or MDS who underwent HLA-mismatched HSCT. Their baseline characteristics are summarized in Table 1. The median age was 46 years (range, 15 to 75). Two hundred and nine patients were diagnosed with acute myeloid leukemia (AML, 59.4%), 62 patients with acute lymphoblastic leukemia (ALL, 17.6%), five patients had mixed phenotype acute leukemia (MPAL, 1.4%), and 76 patients were diagnosed with MDS (21.6%). Of these 352 patients, 172 patients (48.8%) were confirmed as CR1 status, 29 patients (8.2%) as CR2 or subsequent remission, 55 patients (15.6%) as refractory, 34 (9.7%) as relapse, and 62 patients (17.6%) as untreated before HSCT. Of all recruited patients, 264 patients (75.0%) received RIC while 88 patients (25.0%) received MAC for conditioning. One hundred and seventy-one patients (48.6%) received stem cells from haploidentical donors and 181 patients (51.4%) from HLA-mismatched unrelated donors. HLA 1 locus mismatch was detected in 107 patients (30.4%), 2 mismatches in 53 patients (15.1%), 3 mismatches in 39 patients (11.1%), and 4 mismatches in 153 patients (43.5%). Among the patients who received stem cells from mismatched family donors, 161 patients (94.2%) had HLA 3 or 4 mismatches, and among those who received from unrelated donors, 31 patients (17.1%) had HLA 3 or 4 mismatches. Forty-three patients (12.2%) did not receive ATG during transplantation, 60 patients (17.0%) were treated with 2.5 to 7.5 mg/kg of ATG, 192 patients (54.5%) with 9.0 mg/kg of ATG, and 57 patients (16.2%) with 12.0 mg/kg of ATG.

In univariate analysis, there were significant differences in 5-year OS rates according to disease status (39.1% in CR1, 54.0% in ≥ CR2, 11.5% in relapse/refractory, and 38.8% in untreated status, p < 0.001). In addition, conditioning intensity had significant impact on OS (18.2% in MAC group vs. 36.3% in RIC group, p = 0.026). However, there was no significant difference in 5-year OS according to the other clinical factors: sex (male 30.3% and female 34.8%, p = 0.842), age (< 50 years 37.8% and ≥ 50 years 24.3%, p = 0.096), diagnosis (AML 30.1%, ALL 34.1%, MPAL 53.3%, and MDS 40.1%, p = 0.829), HLA mismatch (1 mismatch 30.8%, 2 mismatches 44.8%, 3 mismatches 23.2%, and 4 mismatches 32.9%, p = 0.326), stem cell source (bone marrow 34.8% and peripheral blood 33.5%, p = 0.238), CD34+ cell dose (< 2.0 × 10⁶/kg 51.4% and ≥ 2.0 × 10⁶/kg 32.1%, p = 0.40), and median donor age (< 30 years 36.0%, 30 to 49 years 32.3%, and ≥ 50 years 20.1%, p = 0.067). There were significant differences in the 5-year EFS based on disease status (38.7% for CR1, 44.5% for ≥ CR2, 9.4% for relapse/refractory, and 33.6% for untreated status, p < 0.001) and median donor age (< 30 years 37.0%, 30 to 49 years 26.3%, and ≥ 50 years 17.8%, p = 0.010). However, there was no significant difference in EFS on the basis of other clinical factors, including sex (male 28.9% and female 31.2%, p = 0.97), patient’s age (< 50 years 34.6% and ≥ 50 years 23.6%, p = 0.364), diagnosis (AML 28.1%, ALL 37.3%, MPAL 30.0%, and MDS 35.2%, p = 0.299), conditioning intensity (MAC 25.3% and RIC 32.6%, p = 0.132), HLA mismatch (1 mismatch 32.5%, 2 mismatches 38.5%, 3 mismatches 17.6%, and 4 mismatches 28.1%, p = 0.174), stem cell source (bone marrow 29.8% and peripheral blood 32.3%, p = 0.168), and CD34+ cell dose (< 2.0 × 10⁶/kg 39.3% and ≥ 2.0 × 10⁶/kg 30.5%, p = 0.886). The 5-year CIR of relapse, relapsed or refractory disease status at the time of HSCT was significantly associated with higher relapse compared to CR or untreated status (83.2% in relapse/refractory vs. 39.3% in CR1, 48.2% in ≥ CR2, and 30.4% in untreated status, p < 0.001). However, there was no significant association between CIR of relapse and other clinical factors, such as sex (male 51.4% and female 48.9%, p = 0.906), age (< 50 years 49.3% and ≥ 50 years 49.4%, p = 0.868), conditioning intensity (MAC 50.4% and RIC 49.3%, p = 0.958), stem cell source (bone marrow 48.6% and peripheral blood 48.6%, p = 0.402), CD34+ cell dose (< 2.0 × 10⁶/kg 50.9% and ≥ 2.0 × 10⁶/kg 49.3%, p = 0.532), and median donor age (< 30 years 46.9%, 30 to 49 years 51.4%, and ≥ 50 years 55.3%, p = 0.614). However, three ATG groups did not show significant differences in their 5-year OS (30.7% in no ATG group, 46.2% in low to medium dose ATG group, and 32.2% in high dose ATG group, p = 0.615) and, 5-year EFS (37.6% in no ATG group, 41.5% in low to medium dose ATG group, and 29.0% in high dose ATG group, respectively; p = 0.289). In addition, there was also no significant difference in 5-year CIR of relapse among the three ATG groups (39.9% in no ATG group, 42.1% in low to medium dose ATG group, and 51.4% in high dose ATG group, respectively; p = 0.289).

In univariate analysis, three ATG groups did not significantly affect the CIR of grade 2 to 4 aGVHD. However, the low to medium ATG group tended to show lower incidence rate of grade 2 to 4 aGVHD than those of other two groups (29.6% in low to medium ATG vs. 46.3% in no ATG group and 34.5% in high ATG group, p = 0.083) although the difference was not statistically significant. Age and conditioning intensity significantly affected grade 2 to 4 aGVHD incidence; age (< 50 years 38.6% and ≥ 50 years 29.9%, p = 0.018) and conditioning intensity (MAC 48.4% and RIC 30.8%, p = 0.003). However, there was no significant association between grade 2 to 4 aGVHD incidence and other clinical factors, including sex (35.2% in male and 34.8% in female, p = 0.907), disease status (31.6% in CR1, 21.4% in ≥ CR2, 47.4% in untreated, and 37.9% in refractory/relapse status, p = 0.05), stem cell source (42.3% for bone marrow and 32.6% for peripheral blood, p = 0.211), CD34+ cell dose (35.7 % in < 2.0 × 10⁶/kg and 35.2% in ≥ 2.0 × 10⁶/kg, p = 0.76), and median donor age (33.7% in < 30 years, 33.6% in 30 to 49 years, and 46.8% in ≥ 50 years, p = 0.316). However, the CIR of extensive chronic GVHD (ecGVHD) significantly varied among ATG groups (62.1% in no ATG, 27.1% in low to medium ATG, and 36.1% in high ATG group, p = 0.012). With respect to the stem cell source, patients receiving peripheral blood stem cells showed significantly higher ecGVHD rate compared to those receiving bone marrow stem cell (44.2% vs. 22.1%, respectively; p < 0.001). However, there was no significant association between ecGVHD incidence and other factors, such as age (42.1% in < 50 years and 34.8% in ≥ 50 years, p = 0.42), disease status (35.5% in CR1, 37.7% in ≥ CR2, 37.9% in untreated, and 69.0% in refractory/relapse status, p = 0.58), conditioning intensity (70.5% in MAC and 35.5% in RIC, p = 0.557), HLA mismatch (54.2% in 1 mismatch, 23.8% in 2 mismatches, 40.7% in 3 mismatches, and 31.9% in 4 mismatches, p = 0.469), CD34+ cell dose (54.1% in < 2.0 × 10⁶/kg, and 38.7% in ≥ 2.0 × 10⁶/kg, p = 0.071), and median donor age (34.3% in < 30 years, 45.4% in 30 to 49 years, and 28.3% in ≥ 50 years, p = 0.852).

The factors affecting the 5-year OS and 5-year EFS in multivariate analysis are shown in Table 2. The low to medium ATG group showed significantly longer survival compared with other ATG group (hazard ratio [HR], 0.528; 95% CI, 0.311 to 0.897; p = 0.018) (Fig. 1A). Moreover, the 5-year OS of patients conditioned with RIC and with disease status of refractory/relapse was lower than that of patients conditioned with MAC and with disease status of CR1, respectively ([HR, 2.051; 95% CI, 1.404 to 2.996; p < 0.001] and [HR, 2.903; 95% CI, 2.101 to 4.009; p < 0.001]). The 5-year EFS of patients with disease status of refractory/relapse was significantly lower than that of patients with disease status of CR1 (HR, 2.803; 95% CI, 2.036 to 3.858; p < 0.001). But there was no significant association between the 5-year EFS with ATG dose (HR, 0.792; 95% CI, 0.496 to 1.263; p = 0.327) (Fig. 1B). The 5-year CIR of relapse for low to medium ATG group was higher than those of other ATG groups (HR, 0.580; 95% CI, 0.300 to 1.122; p = 0.105) but the 5-year CIR of relapse of patients with disease status of CR1 was lower compared to patients with other disease statuses (Table 2). The results of multivariate analysis for CIR of GVHD are described in Table 3. The CIR of grade 2 to 4 aGVHD was significantly lower in high ATG group than that in other ATG groups (HR, 0.533; 95% CI, 0.287 to 0.992; p = 0.047) (Fig. 2A). However, the CIR of grade 2 to 4 aGVHD was significantly higher for patients of older age (50 years or more) and conditioned with RIC than that of patients less than 50 years old and conditioned with MAC, respectively ([HR, 1.519; 95% CI, 1.012 to 2.278; p = 0.044] and [HR, 2.140; 95% CI, 1.299 to 3.524; p = 0.003]). In addition, the CIR of ecGVHD was significantly lower in low to medium ATG group compared to that in no AGT group (HR, 0.447; 95% CI, 0.224 to 0.889; p = 0.022) (Fig. 2B).