Introduction

Allogeneic hematopoietic stem cell transplantation (allo-SCT) is a potentially curative therapy and has become a standard of care for various hematologic diseases [1, 2]. The outcomes of patients undergoing allo-SCT have improved over the last decades [3, 4]. However, graft failure (GF) is still a life-threatening complication of the procedure due to the prevailing use of cord blood (CB) or human leukocyte antigen (HLA) mismatched donors as stem cell sources, and reduced-intensity conditioning [5,6,7,8,9]. Studies have shed light on the mechanism of GF [10, 11], which includes rejection by residual recipient T lymphocytes [12], absence of facilitating donor T [13] or natural killer cells [14], donor-specific antibody reactions [15, 16], low counts of infused donor cells [6, 7], stromal cell damage [17], inflammatory cytokinemia [18, 19], and concomitant severe infections [20]. GF occurs in as many as 4.0–7.3% of allo-SCTs, with a higher incidence in the setting of CB transplantations [5,6,7,8,9].

The survival rate of patients who developed GF is dismal, at only 11–58% [6, 7, 21,22,23,24,25]. This can be attributed to severe infections caused by prolonged neutropenic periods, organ failure due to conditioning [26], acute graft-versus-host disease (GVHD) after salvage transplantation [27], and disease relapse or progression. CB is favored for urgent salvage transplantations due to its availability [22,23,24,25]. During the last decade, haploidentical donors (HIDs) have emerged as alternative donors due to the promising survival outcomes reported [28,29,30]. HIDs are also considered for salvage transplantations because of their availability and the fast engraftment they afford [22, 31,32,33,34,35,36]. A Japanese study showed the superiority of bone marrow (BM) or peripheral blood (PB) to CB, but they included HLA-matched and -mismatched donors, and their BM and PB populations were small due to GF rarity [22]. To date, no large-scale studies have directly compared CB and HID as stem cell sources in the setting of salvage transplantation after allo-SCT, and the optimal donor selection strategy for GF remains unclear.

To address this question, we analyzed transplant outcomes for patients who developed GF after first allo-SCT and underwent salvage transplantation from HID or CB stem cells from data on a nationwide registry of the Japan Society for Hematopoietic Cell Transplantation.

Methods

The Transplant Registry Unified Management Program 2 of the Japanese Data Center for Hematopoietic Cell Transplantation provided the clinical data for this study [37, 38]. We included data from patients over 16 years who were suspected or diagnosed with GF and underwent a second allo-SCT using CB or HID between 2007 and 2016 (n = 744). Because of the registration data-based retrospective study, we could not distinguish primary and secondary GF and therefore included both. We excluded data from patients who underwent a second allo-SCT without antecedent conditioning regimens (n = 21), and from those who developed GF concomitantly with a relapse diagnosis (n = 11). In addition, we excluded records from patients lacking data on clinical outcomes (n = 1, data for survival) and infused cell dose (n = 12). Finally, 699 patients were included. Our study protocol adhered to the principles of the Declaration of Helsinki, and the Institutional Review Board of Tokai University School of Medicine approved this retrospective study.

Overall survival (OS) was defined as the time between transplantation and death or the time of last visit. Progression-free survival (PFS) was defined as the time between transplantation and the first event (relapse or death). Nonrelapse mortality (NRM) was defined as death without relapse or disease progression. Neutrophil recovery required an absolute neutrophil count of at least 0.5 × 109/L for 3 consecutive days. The first day was considered the recovery day; platelet recovery required a platelet count of at least 20 × 109/L without transfusions for 3 consecutive days. We defined HID as a related donor with HLA disparity of at least two serological levels. We classified the conditioning intensity [39] and disease risk index [40] in line with the published criteria. Acute and chronic GVHD was diagnosed and graded at each center according to published criteria [41, 42]. Finally, we defined acute GVHD-related death as that caused primarily or secondarily by GVHD; and infection-related death as that due to GF or infection, but not to acute GVHD.

We compared distributions of patient characteristics using Fisher’s exact test for categorical variables and the Mann–Whitney U test for continuous variables. We estimated OS probabilities using the Kaplan–Meier method, and analyzed differences between groups using the log-rank test. We used a cumulative incidence method to determine the incidences of NRM, engraftment, acute and chronic GVHD, and Gray’s test to analyze differences between the groups. We considered relapse and disease progression as competing risks in the NRM analysis and NRM as competing risk in the relapse analysis. Relapse and NRM were considered to be competing risks in the analysis of engraftment, whereas relapse, NRM, and GF after salvage transplantation were considered in the analysis of acute and chronic GVHD. We performed multivariate analyses with the Cox proportional hazard regression model for OS and PFS or the Fine–Gray proportional hazards model for NRM, relapse, engraftment, and the incidence of acute and chronic GVHD. We included factors demonstrating significance at P values < 0.05 in the univariate analysis into the multivariate analyses, and calculated hazard ratios (HR) with 95% confidence intervals (CIs). We considered the following covariates in the univariate models for each analysis: stem cell source (HID vs. CB), age (<55 vs. ≥55 years), sex (male vs. female), donor source at first SCT (CB vs. BM/PB), disease risk index (DRI) (low/intermediate vs. high/very high), presence of anti-HLA antibody (absence vs. presence), presence of donor-specific anti-HLA antibody (absence vs. presence), conditioning intensity (myeloablative, reduced-intensity, or nonmyeloablative conditionings), alkylating agents use (no vs. yes), total body irradiation (TBI) (no vs. yes), GVHD prophylaxis (calcineurin inhibitors [CNI] plus methotrexate [MTX], CNI plus mycophenolate mofetil [MMF], CNI plus prednisolone [PSL], CNI alone, or another combination), ABO incompatibility (match vs. mismatch), transplantation year (2007–2011 vs. 2012–2016), and days from first SCT to salvage SCT (<42 days vs. ≥42 days). All P values were two-sided, and we performed all statistical analyses using EZR, a graphical user interface for R software (The R Foundation for Statistical Computing, version 2.13.0, Vienna, Austria) [43].

Results

Patient and transplant characteristics

We included data from 129 eligible patients in the HID group, and 570 in the CB group. Table 1 shows the characteristics of patients and transplants. We found significant differences between the groups in terms of donor source during the first SCT, DRI, presence of anti-HLA antibody, and donor-specific anti-HLA antibody, conditioning intensity, use of alkylating agents and TBI, GVHD prophylaxis, use of antithymocyte globulin (ATG), and ABO incompatibility. The median CD34+ cell counts were 25.0 × 105/kg (range, 6.90–97.0) in the HID group and 0.78 × 105/kg (range, 0.08–4.86) in the CB group. In the HID group, ten patients (7.8%) received posttransplant cyclophosphamide (PT-Cy) as GVHD prophylaxis. Other characteristics were similar between the groups. All but one in the HID group received PB, rather than BM, as a stem cell source. No patient received donor cell with ex vivo T-cell depletion in the HID group. In the CB group, all patients received single CB. The median time from first SCT to salvage SCT was 42 days in both groups (ranges: HID, 17–757; CB, 19–2250). The median follow-up period for survivors was 1111 days.

Table 1 Characteristics of patients and transplants.
Full size table

Engraftment

The median days of neutrophil recovery were 13 and 28.5 days, in the HID and CB groups, respectively. The cumulative incidence of neutrophil recovery after 30 days from transplantation was significantly higher in the HID group (79.7%) than the CB group (52.5%; P < 0.001; Fig. 1a). In the multivariate analysis, the HID group was significantly associated with neutrophil recovery (HR, 2.94; 95% CI, 2.11–4.10; P < 0.001; Table 2). Other significantly favorable factors for neutrophil recovery were GVHD prophylaxis other than CNI + MTX and use of TBI and alkylating agents. Among patients who survived longer than 28 days after salvage SCT, 2.9% (3/102) in the HID and 19.5% (86/440) in the CB groups again developed GF (P < 0.001). CB was also a significant risk factor for GF after salvage transplantation. The cumulative incidence of platelet recovery after 60 days was also significantly higher in the HID group (49.6%) than in the CB group (34.1%; P < 0.001; Fig. 1b).

Fig. 1: Engraftment rate after salvage transplantation.
figure 1

Cumulative incidence of neutrophil (a) and platelet recoveries (b) after salvage transplantation according to the stem cell source were shown. Relapse and NRM are competing risks in each analysis.

Full size image
Table 2 Multivariate analysis for neutrophil recovery, OS, and NRM.
Full size table

OS, PFS, NRM, and relapse rate

Both groups demonstrated similar OS (33.1% in the HID group vs. 34.6% in the CB group at 1 year; P = 0.40; Fig. 2a). After adjustments for the other covariates, the HID group tended to give better OS, but the difference did not reach a statistical significance (HR, 0.83, 95% CI, 0.64–1.07, P = 0.16). The significant risk factors for OS included old age, poor DRI, GVHD prophylaxis with other combination, nonuse of TBI, and myeloablative conditioning. GVHD prophylaxis with CNI + MMF was a significant favorable factor for OS, whereas time from first SCT to salvage SCT (<42 vs. ≥42 days) did not affect OS. Regarding PFS, there was no significant difference between the groups (28.3% in the HID group vs. 30.3% in the CB group at 1 year; P = 0.38; Fig. 2b). In multivariate analysis, stem cell source was not a significant risk factor for PFS (Supplementary Table 1). Both groups had similar NRMs (45.1% in the HID group vs. 49.8% in the CB group at 1 year; P = 0.48; Fig. 2c). However, after adjustments for the other covariates, the HID group was associated with a lower NRM than the CB group (HR, 0.71, 95% CI, 0.52–0.98, P = 0.038). GVHD prophylaxis with CNI + MMF and use of TBI were also associated with lower NRMs. On the other hand, old age and GVHD prophylaxis with other combination were significant risk factors for NRM. In the HID group, the use of ATG or PT-Cy did not result in a significant decrease in NRM after 180 days (44.6% in patients administered ATG (n = 87), 30.0% in those administered PT-Cy (n = 10), and 36.1% in those administered neither ATG nor PT-Cy (n = 32); P = 0.52 by univariate Gray’s test). In the CB group, the use of ATG was associated with higher NRM in the univariate analysis (63.5% in patients who received ATG vs. 48.2% in those who did not receive ATG at 1 year; P = 0.029); however, the difference was not significant in the multivariate analysis (HR, 1.16; 95% CI, 0.84–1.61; P = 0.37). Regarding relapses, the cumulative incidence in the HID group was significantly higher than that in the CB group (26.6 vs. 19.9% at 1 year; P = 0.046), although the difference was not significant in the multivariate analysis (Supplementary Table 1).

Fig. 2
figure 2

Overall survival (a), progression-free survival (b), nonrelapse mortality (c), and relapse rate (d) after salvage transplantations according to the stem cell source.

Full size image

GVHD and infections

Given that the faster engraftment did not translate into significantly better OS in the HID group, we focused on acute GVHD development. The cumulative incidence of acute GVHD after 100 days was significantly higher in the HID group than in the CB group, regardless of the severity (grade II–IV, 34.1% in the HID group vs. 16.4% in the CB group, P < 0.001; and grade III–IV, 19.4% in the HID group vs. 4.0% in the CB group, P < 0.001; Fig. 3a, b). After adjustments for other covariates, the HID group was significantly associated with a higher incidence of grade II–IV (HR, 2.76; 95% CI, 1.81–4.21; P < 0.001) and III–IV acute GVHD (HR, 4.46; 95% CI, 2.22–8.97; P < 0.001). In comparison with CNI + MTX, GVHD prophylaxis with CNI + MMF and CNI alone were associated with development of grade II–IV acute GVHD, whereas the latter was also associated with grade III–IV acute GVHD (Table 3). Figure 4 shows the cumulative incidence of acute GVHD-related or infection-related deaths. The HID group demonstrated a lower incidence of infection-related deaths than the CB group (18.7% in the HID group vs. 33.0% in the CB group at 1 year, P = 0.003; Fig. 4a). Conversely, the cumulative incidence of acute GVHD-related deaths was significantly higher in the HID group than in the CB group (12.4% in the HID group vs. 1.6% in the CB group at 1 year, P < 0.001; Fig. 4b). The list of primary causes of deaths is shown in Supplementary Table 2.

Fig. 3
figure 3

Cumulative incidences of grade II–IV (a) and III–IV (b) acute graft-versus-host disease, and all-grade (c) and extensive chronic graft-versus-host disease (d) after salvage transplantations according to the stem cell source. Relapse, NRM, and graft failure after salvage transplantation are competing risks in each analysis.

Full size image
Table 3 Multivariate analysis for the development of acute GVHD.
Full size table
Fig. 4
figure 4

Cumulative incidence of infection-related (a) and acute graft-versus-host disease-related deaths after salvage transplantations according to the stem cell source.

Full size image

Patients with donor-specific anti-HLA antibody

As shown in Table 1, 21 (16.3%) in the HID and 110 (19.3%) in the CB groups had anti-HLA antibody, of which seven (5.4%) and nine (1.6%) were donor specific. For patients with donor-specific anti-HLA antibody, the cumulative incidence of neutrophil recovery at 30 days after salvage SCT was higher in the HID group than in the CB group, although the difference was not significant (71.4 vs. 22.2%; P = 0.18). The 1-year OS was 28.6% in the HID and 0% in the CB groups (P = 0.203). Figure 5 shows the forest plot for OS stratified by patient characteristics, including the presence of donor-specific antibodies.

Fig. 5
figure 5

Forest plot for overall survival stratified by patient characteristics.

Full size image

Discussion

The utility of haploidentical salvage transplantations for GF has been reported only in studies with small patient populations [22, 34,35,36]; no large-scale studies have directly compared the outcomes of salvage transplantation from HID and CB in adult patients. In this study, we demonstrated the faster engraftment and the lower NRM after salvage transplantation for the HID group patients compared with the CB group patients. However, these HID advantages did not translate into significantly better OS. Fuji et al. reported HID’s superiority using nationwide registry data [22]; but their BM/PB donors included HLA-matched or HLA-partially mismatched donors, and the number of patients in the BM/PB group was small. These differences may contribute to the discrepancy in the outcomes between our studies. Another explanation for the similar survival outcomes of HID and CB groups is that, in our study, the relapse rate after salvage transplantation was significantly higher in the HID group than in the CB group. In fact, the poor DRI was observed in the HID group. The difference between the groups could hamper the definitive conclusion. Similar survival outcomes for HID and CB transplants have also been reported in a pediatric cohort in Japan [44]. Although the results should be interpreted with caution, owing to the rapid availability of CB, which does not need a waiting period for the HLA test results, both CB and HID can be reasonable choices in the GF setting.

The stem cell source had significant effects on the cause of death in our study. Acute GVHD-related mortality was higher after HID transplants than after CB transplants, while infection-related mortality was lower after HID transplants than it was after CB transplants. Although the faster neutrophil recovery in the HID group could prevent infection-related deaths to a certain extent, severe GVHD development was a major problem after haploidentical salvage transplantation. The advantages in the HID group were partly negated by acute GVHD development, which also may have contributed to the similar OS rates after both transplantations. A higher incidence of acute GVHD after haploidentical salvage transplantation was also reported in other studies [22, 33]. PT-Cy was introduced in the last decade as an effective GVHD prophylaxis during haploidentical transplantations with fewer episodes of acute GVHD and lower NRM than ATG-based prophylaxis [45]. Unfortunately, our study included only ten patients transplanted with PT-Cy in the HID group, and our study lacked statistical power for analyzing the effects of PT-Cy in the HID group. Future studies should compare GVHD incidences after haploidentical salvage transplantation with PT-Cy to those with other GVHD prophylaxes.

Donor-specific anti-HLA antibody is a risk factor for GF. In our study, 16 patients underwent salvage SCT from the antibody-specific donor. For patients with donor-specific anti-HLA antibody, HID was numerically associated with neutrophil engraftment, although the difference was not significant. It was probably attributed to the small patient sample. However, no patients with donor-specific antibody in the CB group survived at 1 year after salvage SCT, which shows the dismal outcome after salvage SCT with donor-specific antibody. As shown in Fig. 5, no specific population had significantly benefited from each donor source in terms of OS. In the setting of salvage SCT with donor-specific antibody, it might be reasonable to choose a HID donor to ensure the engraftment, although a larger study is required.

Not only the stem cell source, but several transplant characteristics affected the transplant outcomes in our study. In terms of conditioning regimen and intensity, using alkylating agents and TBI were associated with neutrophil recovery, and the latter was also associated with better survival and lower NRM. Also, we identified myeloablative conditioning as an unfavorable factor for survival. Other studies have attributed a similar prognostic relevance to alkylating agents and TBI [22,23,24, 44]; thus, reduced-intensity conditioning using alkylating agents and TBI may become a first choice for salvage transplantations. Regarding GVHD prophylaxis, CNI + MTX was associated with poor engraftment, while it led to fewer episodes of acute GVHD than other prophylaxis methods. On the other hand, CNI + MMF offered better engraftment than CNI + MTX, probably due to the lower myelotoxic effect of MMF over that of MTX [46]. Although the advantage of CNI + MMF was partly counterbalanced by the higher incidence of acute GVHD, the OS and NRM were significantly better in patients who received CNI + MMF than in those who received CNI + MTX. In general, CNI + MMF seems to provide favorable GVHD prophylaxis for salvage transplantations [24], but the optimal GVHD prophylaxis should probably consider the stem cell source, the conditioning regimen, and the use of ATG or PT-Cy. Therefore, the impact of GVHD prophylaxis on salvage transplantation should be discussed under more specific settings.

We are aware of the limitations in our study. First, this was a registry-based retrospective study, and the patients’ characteristics differed significantly between the groups. In fact, patients in the HID group exhibited poor DRI and were more likely to experience disease relapses after salvage transplantation, which would have affected the survival rate. In addition, patients included in this study were heterogeneous in terms of cause of GF (primary or secondary), underlying disease, conditioning regimen, and GVHD prophylaxis. Second, some data on patient characteristics were lacking, especially that on anti-HLA antibody that might have led to some biases. Third, as described above, few patients received PT-Cy, which prevented us from analyzing the effects of different GVHD prophylaxes for haploidentical transplantations. However, to our knowledge, our study included the largest number of patients with GF after allo-SCT until now, and it should provide useful information about donor selection in the GF setting.

In conclusion, we demonstrated a lower NRM after haploidentical transplantations than after CB transplantations, although that did not lead to better survival. The engraftment advantage of haploidentical transplantations was partly counterbalanced by a higher incidence of acute GVHD. A study focusing on the optimal GVHD prophylaxis in the setting of haploidentical transplantations is needed to improve outcomes in these patients.