Hematopoietic stem cell transplantation for acute lymphoblastic leukemia: why do adolescents and young adults outcomes differ from those of children? A retrospective study on behalf of the Francophone Society of Stem Cell Transplantation and Cellular Therapy (SFGM-TC)

In the acute lymphoblastic leukemia (ALL) landscape, adolescents and young adults (AYA) often present high-risk diseases and increased chemotherapy-related toxicity. Studies analyzing the outcomes of AYA after hematopoietic stem cell transplantation (HSCT) are scarce. Our study aimed to compare the outcomes of children and AYA with ALL after HSCT and to determine the factors influencing potential differences. 891 patients, from the SFGM-TC registry, aged between 1 and 25 years who received HSCT between 2005 and 2012 were included. The outcomes of AYA were compared to the ones of their younger counterparts. Five-year OS and GRFS were lower in AYA: 53.1% versus 64% and 36% versus 47% (p = 0.0012 and p = 0.007, respectively). WhileCIR was similar in both groups, 5 year-treatment related mortality was higher in AYA: 19% versus 13% (p = 0.04). The lower GRFS in AYA was mainly explained by a higher chronic graft versus host disease (cGvHD) incidence: 32% versus 19% (p < 0.001). Use of peripheral blood stem cells and use of anti-thymoglobulin appeared to be the main factors impacting cGvHD occurrence in AYA. AYA have worse outcomes than children after HSCT for ALL because of a greater risk of TRM due to cGvHD. HSCT practices should be questioned in this population.


Introduction
The prognosis of childhood acute lymphoblastic leukemia (ALL) has improved continuously during the last 30 years. The currently reported overall 5-year survival (OS) is about 90% in privileged countries (Curran and Stock 2015). The prognosis of AYA ALL is poorer than the ones of their younger counterparts, as 5-year survival decreases from 85.8% in children (ages between 0 and 14 years) to 62.2% in adolescents (15-19 years) and 52.8% in young adults (20-39 years) (Curran and Stock 2015;Trama et al. 2016). The biology of ALL differs between children and AYA. Indeed high-risk cytogenetic abnormalities are more frequent in this population (Boissel and Baruchel 2018;Boissel et al. 2003;Burke and Douer 2014). Moreover, AYA show higher incidence of acute toxicities after intensive treatments (Boissel and Baruchel 2018;Boissel and Sender 2015;Burke and Douer 2014;Stock et al. 2011). Nevertheless, despite this increased toxicity, the prognosis of AYAs ALL is better following pediatric protocols than after adults protocol as shown in the first French comparative study (Boissel et al. 2003), subsequently confirmed by other studies led by pediatric and adult cooperative groups (Boissel and Baruchel 2018;de Bont et al. 2004;Ibrahim et al. 2014;Pui et al. 2011;Stock et al. 2008).
Either in pediatric or in adult protocols, hematopoietic stem cell transplantation (HSCT) indications have been progressively restricted to patients with early poor response to chemotherapy (e.g., persistence of MRD after consolidation therapy). This has led to a decrease of HSCT indications in AYA, which are still more frequent than in the pediatric patients as in the NOPHO 2008 trial (5.5% in 1-17 years, and 15.8% in 18-45 years) (Boissel and Baruchel 2018;Toft et al. 2018). The outcomes of AYAs after HSCT seemed to be worse than the ones of children in three previous studies published in 2013,2014 and 2019 (Burke et al. 2013;Hangai et al. 2019;Wood et al. 2014). In two of these studies, the lower overall survival in AYA was attributed to an excess of treatment-related mortality (TRM) (Burke et al. 2013;Hangai et al. 2019).
Our study aimed to compare the outcomes of HSCT for ALL between pediatric and AYA patients in a large cohort, with sufficient follow-up, to determine factors influencing OS and TRM.

Patients
All patients aged between 0 and 25 years, who received a first HSCT in treatment for ALL between 2005 and 2012, from the Francophone Society of Bone Marrow Transplantation and Cellular Therapy (SFGM-TC) registry were included in this retrospective and multicentric study.
Cells provided by a matched sibling donor (MSD) defined genoidentical setting. A transplant from an HLA-matched unrelated donor (MUD) (9/10 or 10/10) defined a phenoidentical setting. All cord blood sources were pooled regardless of HLA matching, since HLA compatibility was not fully captured in the registry for this stem cell source.

Outcomes
Outcomes were OS (time from HSCT to death), DFS (time from HSCT to relapse or death), TRM, GvHD, cumulative incidence of relapse (CIR) and GvHD and relapse-free survival (GRFS). GRFS was defined as time from HSCT to grade 3-4 acute GVHD (aGvHD) or relapse or chronic GvHD (cGvHD) or death, whichever comes first. Chronic GvHD was described as limited or extensive GvHD as the NIH consensus was not available at the studied period (Martin et al. 2006). Data cutoff was June 2015.

Statistical analysis
Diseases and HSCT procedures were compared between pediatric patients and AYA by Student or Wilcoxon test for quantitative variables and by Chi 2 or Fischer test for qualitative ones.
5-year OS, 5-year DFS and 5-year GRFS were determined by Kaplan-Meier method. Risk factor analyses were done using Cox models. The cumulative incidence of relapse (CIR) was estimated taking into account death as a competing event. The cumulative incidence of cGvHD was estimated taking into account death and relapse as competing events. A leukemia-related death was considered as a competing event for the TRM estimation. Risk factors were analyzed using Fine and Gray models A p value of ≤ 0,05 was considered as significant in all statistical tests.
Statistical analysis was performed using R version 4.0.2

Ethical aspects
Patients or their parents/guardians received information about the SFGM-TC registry before HSCT and provided a signed consent to be included. There was a trend for a higher T-cell ALL incidence in the AYA population (29.8% versus 25.1%, p = 0.1275). High-risk cytogenetic abnormalities were found in 25.9% of children and 29.5% of AYA (p = 0.273).

Patients, disease, and conditioning regimens
HSCT was performed in the first CR for 56.8% of the AYAs, whereas 57.5% of children received HSCT in the second or higher CR (p < 0.001). Note that 7.5% of AYAs and 3.7% of children presented a refractory disease at the time of transplant (p = 0.015).
HSCT procedures mainly included a myeloablative conditioning regimen (MAC) based on total body irradiation (TBI) or chemotherapy. See detailed MAC administered in supplementary data. TBI was used more frequently in AYA than in children (90.1% versus 83.1%, p = 0.003). Performance status at transplant was similar in both groups. Nevertheless, AYA more often received a reduced intensity conditioning (RIC) than children (5.8% versus 2.4%, p = 0.01).
Bone marrow (BM) or CB was often used in children (60.2% and 29.4%, respectively) versus 55.6% and 16.4% in AYA, respectively, whereas peripheral blood stem cells (PBSC) were used for 28% of AYA and 10.3% of children (p < 0.0001). Moreover, when being transplanted in an adult center, PBSC were more commonly used for AYA (30% of AYA's HSCT in adult centers versus 21.2% of AYA's HSCT in pediatric centers, p = 0.051). No difference in the choice of stem cell source depending on the center of transplant was observed for children. BM and PBSC were provided by an MSD in 40.2% of children and 43.4% of AYA and from an MUD in 57.2% and 55.1% of cases, respectively (p = 0.474) (other patients received CB, for which HLA compatibility was not fully captured in the registry). GvHD prophylaxis mainly included ciclosporin (n = 835), methotrexate (n = 461), mycophenolate mofetil (n = 141), and corticosteroids (n = 91). Antithymoglobulins (ATG) were used for 48% of children and 26% of AYA (p < 0.001). When AYA received HSCT in a pediatric center, they more often received ATG as GvHD prophylaxis than in adult centers, 47.5% versus 16.8% (p < 0,001), as well as observed for children: 49.8% of children received ATG in pediatric center versus 31% in adult center (p = 0.02). See patient's characteristics in Table 1.

Poorer results of HSCT in adolescents and young adults
2-year OS and 5-year OS were significantly lower in the AYA group (60% versus 71% and 53% versus 64%, respectively, p = 0.0012) with multivariate analysis hazard ratio (HR) of 1.40 [1-1.95], (p = 0.05), see Fig. 1. In multivariate analysis, an altered performance status, a refractory disease at transplant, or use of PBSC or CB was associated with lower OS. In subgroup analysis the 5-year OS difference remained significant in patients who received HSCT in CR2 or higher (42% in AYA, versus 59% in children, p = 0.0003).
Interestingly, analysis of different subgroups of age at transplant, excluding patients with refractory diseases at transplant, showed a gradual significant decrease of OS (see Fig. 2 and Table 2).

ALL subtype and relapse do not explain the worst outcomes of HSCT in AYAs
The 2-year and 5-year DFS were lower in AYA than in younger patients (54.5% versus 62.3% and 49.2% versus 59.5%, respectively, p = 0.0071), but CIR was similar in both groups (32% and 27% at 5 years, p = 0.19) (Fig. 1). There was no significant difference of DFS probability or CIR in each age subgroup (Table 2). In multivariate analysis, a refractory disease at transplant, a low performance status and a transplant from CB were associated with a lower DFS probability. In patients treated for a T-ALL, 5-year DFS and CIR were not significantly different between AYA and children (see in Supplementary Data).
In multivariate analysis, having received a TBI-based conditioning regimen was associated with a higher DFS probability (HR 0.7 [0.49-0.01], p = 0.06). When comparing outcomes of children and AYA who received a TBIbased conditioning, CIR was similar in both groups (25% versus 31%, p = 0.081). Moreover, even for TBI-conditioned patients, 5-year OS, 5-year DFS and 5-year GRFS were still lower in older patients; see supplementary data.

Chronic GvHD strongly impacts AYA outcomes after HSCT for ALL
aGvHD cumulative incidences were similar in both groups (61% in AYAs and 59% in children, p = 0.62). On the contrary, cGvHD occurred more frequently in AYA than in children (32% versus 19%, p < 0.001), among which 44% Percentages of TBI based conditionings are given in italic **HLA matching: CB recipients were excluded from this analysis, since HLA compatibility was not fully captured in the registry for this stem cell source AYA adolescents and young adults, HSCT hematopoietic stem cell transplantation, ND not determined, CR complete remission, TBI total body irradiation, RIC reduced intensity conditioning regimen, PBSC peripheral blood stem cells, CB cord blood; quantitative variables are given as mean ± standard deviation (SD) a As evaluated at the time of HSCT and 31% were extensive, respectively (Fig. 1). Having received ATG (HR 0.63 [0.43-0.92], p = 0.016) or presenting a good performance status was associated with a lower risk of cGvHD in multivariate analysis, while having received PBSC was associated with a higher risk of cGvHD (HR: 1.46 [0.99-2.14], p = 0.05) (Fig. 3).
In patients who received bone marrow stem cells after an MAC, cGvHD incidence was still significantly higher in AYA (30% versus 17%, p < 0.001).

Discussion
Our study confirmed the lower prognosis after HSCT for ALL of AYA compared to children. Decrease of OS after HSCT started after 10 years of age. Moreover, excess of mortality in the AYA group was linked to a higher TRM due to excess of cGvHD. Indeed, GRFS probability was significantly lower in AYA than in children, while CIR and aGvHD incidences were similar in both groups and cGvHD incidence was higher in AYA than in children. The main factors involved in TRM and cGvHD excess in AYA were the use of PBSC and the absence of use of ATG. These results pointed to the importance of transplantation practices in AYA.
Our study included patients who received a first HSCT for ALL in the pre-FORUM study era. We showed a significant decrease of 5-year OS in AYA compared to younger patients consistent with previous studies (Burke et al. 2013;Wood et al. 2014). Of note, outcomes of HSCT in our pediatric group (5-year OS of 63.9% and 5-year DFS of 59.5%) were consistent with previous published studies. The BFM study showed a 4-year DFS of 67% and 71% in children who received HSCT in the same era of procedure (Peters et al. 2015). The international FORUM study found higher 2-year OS and DFS than in our study (91% (IC95 86-95%) and 86% (IC95 79-90%), respectively) in children transplanted after a TBI-based conditioning between 2013 and 2019 for ALL (Peters et al. 2020). These differences could be related to improvement of supportive therapy notably through improved early detection of infection as well as prophylactic and preemptive antimicrobial drug (Singh and McGuirk 2016).
Leukemia relapse does not contribute to the excess of mortality in AYA after HSCT. AYA often present highrisk ALL (Iacobucci and Mullighan 2017;Soulier et al. 2003) and have a worse prognosis than younger children even after pediatric-inspired first-line therapeutic protocols (Toft et al. 2018). Nevertheless, as in our study, following HSCT, relapse rates are similar in both children and AYA in several studies (Burke et al. 2013;Goldstone et al. 2008;Wood et al. 2014). In our study, despite a higher incidence of high-risk cytogenetic abnormalities in AYA (not statistically significant), CIR was similar in both groups overall (27% in children, and 32% in AYA, p = 0.19) and when 1 3 stratifying according to immunophenotype. Knowing the impact of TBI on relapse incidence (Peters et al. 2020), we compared AYA and children receiving TBI-based conditioning regimens. Again, relapse rates were similar (25% in children versus 31% in AYA, p = 0.08).
Treatment-related toxicity was the main cause of higher mortality in AYA in our study. Of note, 2-year and 5-year TRM incidences in our pediatric group (12% and 13%, respectively) were consistent with previous studies (Peters et al. 2015). In our study, AYA had a higher 2-year and 5-year TRM than children (16% versus 12% and 19% versus 13%, p = 0.04). In other studies, higher treatment-related toxicity in AYA was previously observed, either following chemotherapy first-line treatment, or HSCT (Boissel and  Rank and Schmiegelow, 2020;Toft et al. 2018;Wood et al. 2014). The Minneapolis comparative study showed a higher TRM at 1 year after HSCT in AYA than in children (28% versus 14%, respectively, p = 0.04). Maybe due to the small numbers, except use of matched CB (RR = 0.31 [0.13-0.78], p = 0.01), no factor significantly impact TRM in their multivariate analysis (Burke et al. 2013). In the Japanese study, AYA presented excess of TRM (11%, 17% and 19% in patients aged between 1-9 years, 10-19 years and 20-29 years, respectively, p < 0.001). In this study, infections-related deaths tended to be more frequent in older AYA. No multivariate analysis on TRM was performed (Hangai et al. 2019). In the HSCT for myeloid acute leukemia setting, a recent retrospective study also showed a lower 2-year OS and a higher TRM in AYA than in children (61.1% versus 71.4%, p = 0.0009, and 10.6% versus 7%; p < 0.0001, respectively) (Pochon et al. 2021).
In our study, 5-year GRFS probability was lower in AYA than in younger patients (36% versus 47%, p = 0.0078), while aGvHD incidence (61% and 59%, p = 0.6) and CIR (32% versus 27%, p = 0.19) was similar in both groups. Thus, post-HSCT morbidity-mortality in AYA was mainly impacted by cGvHD occurrence, which was higher in AYA (32% versus 19% in children, p < 0.001). This could be consistent with the TRM estimation curve, which showed a gap between AYA and children, starting from 3 months after HSCT. Of note, the incidence of cGvHD observed in our pediatric group was consistent with previous pediatric studies (Peters et al. 2015;Zheng et al. 2015). A higher cGvHD incidence in AYA was previously described in other studies. In the Minneapolis comparative study, cGvHD incidence was higher in AYA (15% versus 7% in pediatric patients, p = 0.06) (Burke et al. 2013). In the BFM trial, patients upon than 12 years old, receiving a genoidentical stem cell transplant, had higher incidence of extensive cGvHD than younger patients (Peters et al. 2015).
In our multivariate analysis, two factors were associated with a higher risk of cGvHD: use of PBSC ], p = 0.05) and absence of ATG use (HR of use of ATG 0.63 [0.43-0.92], p = 0.016).
Impact of PBSC is not surprising, since it is now well known that using PBSC increases cGvHD incidence (Campregher et al. 2015;Eapen et al. 2004). In our study, AYA more frequently received PBSC (28% of cases) than children (10.3% of cases) (p < 0.001), particularly when they were transplanted in adult centers (30.8% of cases, versus 21.2% of cases in pediatric centers, p = 0.05). This difference of stem cell source is not described in the Minneapolis study, but the small number of patient may explain this discrepancy with our data (Burke et al. 2013). Moreover, preference of PBSC in adult HSCT programs was previously described (Mehta et al. 2018) and could be supported by the fact that stem cell source (PBSC versus BM) does not impact OS or TRM of adult patients (Giebel et al. 2017). Of interest is the fact that our subgroup analysis in patients who received bone marrow as stem cell source, after an MAC, showed no difference of TRM or CIR between AYA and children, but a remaining higher incidence of cGvHD in AYA (30% versus 17% in children, p < 0.001).
In our study, AYA received ATG less frequently than children (26% versus 48.1%, p < 0.001), and they also less frequently received ATG when treated in adult centers (16.8%) than in pediatric centres (47.5%) (p < 0.001). We of course checked that there was no difference of donor type between AYA and children who received a BM and a PBSC transplant. Interestingly, an Italian retrospective study showed no difference of TRM or GvHD incidence between adolescents and children transplanted for ALL in second CR. In this study, the use of PBSC was rare (6% in children and 10% in AYA) without significant difference, and ATG prophylaxis was used with the same frequency in both groups (Dini et al. 2011). Moreover, in the AML setting, use of PBSC and ATG prophylaxis both impact higher cGvHD incidence observed in AYA (Pochon et al. 2021).
Obviously, our study has some limitations due to its retrospective nature. These limitations include HSCT indications and detailed disease biology, disease status at transplant (particularly MRD levels), details concerning stem cell sources and HLA compatibility for MUD and CB, criteria for the choice of GvHD prophylaxis, and particularly ATG indications, timing and doses. In addition, more details about GvHD and cGvHD grading could also be useful. Moreover, our study could not capture data regarding GvHD-oriented treatment adherence of AYA. Indeed, poor adherence to GvHD treatment has been described in AYA which could obviously lead to increase incidence of cGvHD (McGrady et al. 2014;Mehta et al. 2018;Pulewka et al. 2020). A prospective study comprising controlled and homogeneous HSCT indications, conditionings, and GvHD prophylaxis would not be easy to construct, but should confirm our results.

Conclusion
AYA or patients age more than 10 years, compared to children age less than 10 years have worse outcome after HSCT for ALL. Excess number of deaths in this specific population is mainly due to cGvHD. Choice of stem cells source and use of ATG in those patients should be discussed.

Acknowledgements
The authors are particularly thankful to the SFGM-TC who provided data and to all centers from the SFGM-TC who included patients and kindly agreed to participate in this study.
Author contributions AG and FR conceived and designed the study. AG, FR and AB wrote the manuscript. AG collected and assembled data. LP performed all the statistical analysis. All other co-authors included patients and critically reviewed the manuscript. All co-author consented to publish this manuscript.

Fig. 3
Multivariate analysis of overall survival (OS) (A), cumulative incidence of relapse (CIR) (B) treatment-related mortality (TRM) (C) and chronic graft versus host disease incidence (cGvHD) (D). Results are presented with hazard ratio (HR) and 95% confidence interval (IC95). RIC reduced Intensity conditioning, AYA adolescents and young adults, TBI total body irradiation, PBSC peripheral blood stem cells, CB cord blood, HR high risk Funding No funding was secured for this manuscript.