Optimal donor for severe aplastic anemia patient requiring allogeneic hematopoietic stem cell transplantation: A large-sample study from China

HLA-haploidentical hematopoietic stem cell transplantation (HSCT) may be an option for severe aplastic anemia (SAA) patients. However, to date, no large-sample studies have been performed to determine which types of SAA patients are suitable for HLA-haploidentical HSCT. We retrospectively studied 189 consecutive patients with SAA who underwent HLA-identical or HLA-haploidentical HSCT at seven transplant centers in China. Propensity score matching (PSM) was applied in this study to reduce the influence of potential confounders. The 5-year overall survival (OS) rate was 72.0% in the HLA-haploidentical group and 76.5% in the HLA-identical group. The median time to achieve engraftment and the incidence of acute GVHD/chronic GVHD were not significantly different between the two groups. In the subgroup analysis, the outcome of patients older than 40 years in the HLA-haploidentical group was significantly poorer than that of patients younger than 40 years in the same group and that of patients older than 40 years in the HLA-identical group. Based on the above results, we suggest that HLA-haploidentical relative HSCT should be considered as a valid alternative option for patients younger than 40 years with SAA for whom no matched sibling donor is available.

Survival. year OS rate was 74.8% in the HLA-haploidentical group versus 78.1% in the HLA-identical group before PSM. No significant difference was observed between the groups (χ 2 = 1.624, p = 0.203) (Fig. 1A). After PSM, the overall survival of the two groups was still comparable; the estimated 5-year OS rate was 72.0% in the HLA-haploidentical group and 76.5% in the HLA-identical group (χ 2 = 2.044, p = 0.153) (Fig. 1B).
In the subgroup analysis, we classified all the propensity score-matched patients into three age levels (<20 years, 20-40 years, and ≥40 years). In the HLA-haploidentical group, the OS rates of the three age levels were different (χ 2 = 13.020, p = 0.001) ( Fig. 2A). The estimated 5-year OS rate was 75.5% among patients <20 years old (n = 32), and the estimated 1-year OS rates among patients aged 20-40 years old (n = 30) and patients aged ≥40 years old (n = 8) were 78.5% and 37.5%, respectively. However, in the HLA-identical group, the OS rates of the three age levels were similar (χ2 = 0.717, p = 0.699) (Fig. 2B). The estimated 1-year OS rates of patients <20 years old and patients 20-40 years old were 85.9% and 62.4%, respectively, and the estimated 4-year OS rate was 83.3% among patients ≥40 years old (n = 12). Next, we compared the two transplant groups at the same age level. At the <40-years level, the 8-year OS rates were 76.9% and 75.6% in the HLA-haploidentical and HLA-identical groups, respectively (χ 2 = 0.443, p = 0.506) (Fig. 2C). However, in the ≥40-years level, the outcome of the HLA-haploidentical group was significantly poorer than the outcome of the HLA-identical group (χ 2 = 5.210, p = 0.022) (Fig. 2D). The estimated 1-year OS rate was 37.5% in the HLA-haploidentical group, whereas the 4-year OS was 83.3% in the HLA-identical group.
To analyze the effects of the conditioning regimen on OS, we divided the patients into three subgroups according to the conditioning regimen protocol (Flu + CY, BU + CY, and CY). In both the HLA-haploidentical and HLA-identical HSCT groups, the OS rates were similar among the three conditioning regimen subgroups. In the HLA-haploidentical group, the 8-year OS rate in the Flu + CY subgroup was 68.5%, the 5 year OS rate in the BU + CY subgroup was 93.8%, and the 1-year OS rate in the CY subgroup was 80.8% (χ 2 = 3.739, p = 0.154) (Fig. 3A). In the HLA-identical group, the 8-year OS rate in the Flu + CY subgroup was 75.4%, the 5 year OS rate in the BU + CY subgroup was 87.5%, and the 2-year OS rate in the CY subgroup was 87.5% (χ 2 = 0.174, p = 0.917) (Fig. 3B).

Discussion
The aim of this study was to compare the outcomes of HLA-haploidentical relative HSCT and HLA-identical sibling HSCT for SAA patients. To the best of our knowledge, this study enrolled the largest number of patients for the comparison of outcomes between HLA-haploidentical relative HSCT and HLA-identical sibling HSCT using similar doses and graft compositions and identical conditioning regimens. The results suggested that, with the exception of a higher incidence of bacterial and fungal infections, HLA-haploidentical relative HSCT was comparable to HLA-identical sibling HSCT in terms of OS, incidence rates of aGVHD and cGVHD, and time to engraftment. In our retrospective study, we found that the OS rates of the HLA-haploidentical and HLA-identical HSCT groups were all greater than 70% after PSM. These results were different from those of Gupta's study in 2010 23 and were indicative of the rapid advances in transplantation technology 24,25 . In the HLA-identical sibling HSCT group, the SAA patients in different age groups exhibited similar survival rates. Thus, in the 2015 edition of the British Guide for SAA 1 , the age limit for hematopoietic stem cell transplantation in SAA patients was expanded to 50 years, and the 40-year-old boundary became increasingly blurred for HLA-identical HSCT 26,27 . HLA-haploidentical relative HSCT was the first alternative transplantation regimen to be considered among other treatment options after failure to respond to IST in the absence of an HLA-identical sibling donor or a suitably matched unrelated donor 1 . Under those circumstances, who are the suitable donors for HLA-haploidentical HSCT for the treatment of SAA? In our study, the prognosis was significantly better among patients younger than 40 years and was equal to the prognosis achieved with HLA-identical sibling HSCT. In contrast, patients older than 40 years had a significantly poorer prognosis, with 5/8 patients older than 40 years in the HLA-haploidentical group dying in the first 6 months after transplantation. The direct causes of death were infection (4/5) and heart failure (1/5), and 4 of these patients had first graft failure. The relatively high rates of graft failure-and transplant-related death were quite different from the rates in elderly patients with hematological malignancies who underwent HLA-haploidentical HSCT 28 . Thus, the 40-year-old boundary remains of great significance in HLA-haploidentical HSCT. Moreover, HLA-haploidentical relative HSCT should be considered as an alternative transplantation option for SAA patients younger than 40 years.
The hematopoietic engraftment rate and time to engraftment of the HLA-haploidentical and HLA-identical groups were comparable to that in previous reports 5,9,[29][30][31] . The cumulative incidence of aGVHD in the HLA-haploidentical group was slightly higher than that in the HLA-identical group, whereas the cumulative incidence of cGVHD was lower in the HLA-haploidentical group than in the HLA-identical group; however, the differences were not statistically significant. Notably, both groups had higher incidence rates of aGVHD and lower incidence rates of cGVHD compared with the rates observed in our previous report 15 and other recently published studies 6,26 ; these differences may be attributed to the different diagnostic criteria or the enrollment of older patients in our study. Among the three different GVHD prophylaxis subgroups, we observed no significant differences in the GVHD incidence rates. There was a lower incidence of grade II to IV aGVHD in the PT-CY subgroup, but the sample size was too small; a study with a larger sample size is required to draw a reliable conclusion. The incidence of bacterial and fungal infections was higher in the HLA-haploidentical group, which was consistent with a previous report 32 . Chang et al. suggested that the early delayed immune reconstitution after HLA-haploidentical HSCT probably led to an increased incidence of infection 33 . In the multivariate analysis,   previous RBC transfusion was revealed to be an adverse factor for SAA patient survival after HSCT, which may be explained by decreased organ function due to iron deposition prior to HSCT. Although iron chelation therapy was administered to these patients before HSCT, the impact of RBC transfusion seemed to be irreversible. This is also the reason why upfront HLA-haploidentical HSCT is the preferred treatment. The GVHD prophylaxis regimens also seemed to have an impact on survival, but the number of patients in the PT-CY subgroup were too small to draw a persuasive conclusion.
In this study, all the SAA patients were given timely intravenous immunoglobulin (IVIG) as a basic infection prevention therapy. Only 23.6% of the patients in the HLA-haploidentical group and 14.7% of the patients in the HLA-identical group developed CMV infections; these rates were significantly lower than those in another SAA HLA-haploidentical HSCT study without the use of timely IVIG 34 . Although our systematic review and meta-analysis showed no advantage from IVIG for the prevention of CMV infection 35    that meta-analysis included only patients with hematologic malignancy, and SAA was not within the scope of the systematic review. The baseline analysis before PSM showed that patients in the HLA-haploidentical HSCT group were significantly younger and that their donors were older than those in the HLA-identical HSCT group. Due to the one-child policy in China over the past 30 years, parent donors accounted for 77.4% (89/115) of the HLA-haploidentical group. The difference in the patient ages was either due to the one-child policy or the preference of the doctors and patients, as the previously prevailing opinion was that younger patients had better outcomes. Because the present study was not a randomized controlled trial, the doctors' and patients' choice of conditioning regimens was based on the knowledge that HLA-haploidentical transplantation had a higher graft rejection rate. Thus, more intensive conditioning regimens and sufficient immunosuppression were required. As a result, the patients in the HLA-haploidentical group were more likely to receive the Bu + CY conditioning regimen than the patients in the HLA-identical group. Similarly, FK506-based GVHD prophylaxis was more often used in the HLA-haploidentical group, as those patients were believed to be at greater risk of severe GVHD. Due to the retrospective nature of this study, we could not control for selection bias and exposure factors (e.g., the diverse conditioning regimens and GVHD prophylaxis regimens). Thus, we performed PSM to reduce selection bias and other potential influences. The results from the cohort after matching were similar to the results obtained before matching. However, there are still limitations in this study. Due to the relatively small sample sizes, subgroup analyses were not performed for the different conditioning regimens that were observed to affect OS or for the different age groups that showed an effect on GVHD and infection incidence. These results should therefore be validated by a well-designed, multicenter prospective study with a larger sample size.
In summary, this clinical study suggests that HLA-haploidentical HSCT achieves outcomes comparable to those of HLA-identical sibling HSCT for SAA patients younger than 40 years. HLA-haploidentical relative HSCT should be considered as a valid alternative option for patients younger than 40 years with SAA for whom no matched sibling donor is available. The SAA diagnosis was based on bone marrow cytomorphology, bone marrow biopsy, and karyotyping according to the Education Program of the American Society of Hematology 36 . All of the patients met the following criteria: (1) voluntary participation in HSCT, including the provision of written informed consent from the patients or their guardians, and (2) the absence of uncontrolled infections and severe liver, renal, lung and heart diseases. Patients with ferritin levels greater than 1000 ng/mL were treated with deferoxamine until the ferritin concentration was less than 1000 ng/mL before undergoing HSCT. This study was approved by the ethics committees of the Second Affiliated Hospital of Third Military Medical University and was conducted in accordance with the Declaration of Helsinki.   Donors and graft sources. All  GVHD prophylaxis and treatment. All of the patients, except those in the PT-CY group, received cyclosporine A (CsA) or tacrolimus (FK506) plus mycophenolate mofetil (MMF) and short-term methotrexate (MTX) for acute GVHD (aGVHD) prophylaxis. HLA-haploidentical HSCT patients from the PT-CY group also received CsA plus MMF, whereas the HLA-identical HSCT patients used only CY (the detailed treatments are described in Fig. 7). aGVHD was defined according to the Fred Hutchinson Cancer Research Center criteria and was treated with 1-2 mg/kg of methylprednisolone per day. Second-line IST, such as CD25 monoclonal antibody (MoAb) (daclizumab; Roche, Basel, Switzerland) or MTX, was given for steroid-refractory aGVHD. Chronic GVDH (cGVHD) was diagnosed according to the 2014 NIH criteria 37 and was treated with 1-2 mg/kg per day of prednisolone equivalents and a full-dose of CsA or FK506.
Supportive care. All of the patients lived in a laminar air-flow room from day −8 to hematopoietic reconstitution and received prophylactic antibiotics when their absolute neutrophil count (ANC) was <0.5 × 10 9 cells/L. Norfloxacin, trimethoprim sulfamethoxazole and ganciclovir were routinely administered according to a previously reported method 38 , and micafungin (50 mg/day), itraconazole (p.o. 200 mg/day) or voriconazole (p.o. 100 mg/day) was administered from day −1 to prevent invasive fungal infections (IFIs). Intravenous immunoglobulin (IVIG 0.4 g/kg once weekly before +100 days and 0.4 g/kg once monthly after +100 days) was given to increase passive immunity. CMV DNA was monitored weekly by PCR. CMV-positive patients were treated with either ganciclovir or foscarnet. Irradiated and filtered red blood cell and platelet transfusions were given to maintain a hemoglobin (Hb) level >80 g/L and a platelet count >20 × 10 9 /L. All of the patients received recombinant human granulocyte colony stimulating factor (G-CSF) from day 01 after hematopoietic stem cell infusion until hematopoietic reconstitution.
Propensity score matching. Patient allocation in this study was based on the HLA-identical or HLA-haploidentical group assignments rather than by random assignment; therefore, the baseline levels of some clinical characteristics were imbalanced between the two groups (see Table 1). To reduce the influence of potential confounders, propensity score matching (PSM) was applied in this study 39,40 . The propensity score that indicated the HLA status for each patient was calculated based on a multivariate logistic regression model. In this model, the dependent variable was a dichotomous variable [the HLA status (identical or haploidentical)], and the previously mentioned imbalanced variables between the two groups were used as covariates (patient and donor age, conditioning regimen, and GVHD prophylaxis). Patients in the HLA-identical group were matched to those in the HLA-haploidentical group using 1:1 nearest neighbor matching with a caliper width of 0.2.

Statistical analysis.
The primary endpoint of the study was overall survival (OS). The secondary endpoints included the incidences of GVHD and infection and the time to engraftment. The time at which PBSCs and BMNCs were infused was the starting point for comparing the outcomes between cohorts. The date of the last follow-up for all surviving patients was January 1, 2015. The median follow-up time was 20.35 months (range: 1-102 months).
Categorical data were summarized as counts and percentages, whereas continuous data were expressed as medians and ranges. Comparisons of the baseline and transplant-related characteristics between the 2 groups were performed before and after PSM using the χ2 test for categorical variables and the non-parametric Mann-Whitney test for continuous variables. Survival functions and the cumulative incidence of aGVHD or cGVHD were estimated by the Kaplan-Meier method and compared using log-rank tests. Multivariate Cox stepwise regression was used to determine the relationship between survival and risk factors; P < 0.05 was considered to indicate statistical significance.
A statistical software package (IBM SPSS Statistics 22, IBM Corp., Armonk, NY, USA) was used for the analyses.