Predictive Value of Dynamic Peri-Transplantation MRD Assessed By MFC Either Alone or in Combination with Other Variables for Outcomes of Patients with T-Cell Acute Lymphoblastic Leukemia

We performed a retrospective analysis to investigate dynamic peri-hematopoietic stem cell transplantation (HSCT) minimal/measurable residual disease (MRD) on outcomes in patients with T-cell acute lymphoblastic leukemia (T-ALL). A total of 271 patients were enrolled and classified into three groups: unchanged negative MRD pre- and post-HSCT group (group A), post-MRD non-increase group (group B), and post-MRD increase group (group C). The patients in group B and group C experienced a higher cumulative incidence of relapse (CIR) (42% vs. 71% vs. 16%, P<0.001) and lower leukemia-free survival (LFS) (46% vs. 21% vs. 70%, P<0.001) and overall survival (OS) (50% vs. 28% vs. 72%, P<0.001) than in group A, but there was no significant difference in non-relapse mortality (NRM) among three groups (14% vs. 12% vs. 8%, P=0.752). Multivariate analysis showed that dynamic peri-HSCT MRD was associated with CIR (HR=2.392, 95% CI, 1.816–3.151, P<0.001), LFS (HR=1.964, 95% CI, 1.546–2.496, P<0.001) and OS (HR=1.731, 95% CI, 1.348–2.222, P<0.001). We also established a risk scoring system based on dynamic peri-HSCT MRD combined with remission status pre-HSCT and onset of chronic graft-versus-host disease (GVHD). This risk scoring system could better distinguish CIR (c=0.730) than that for pre-HSCT MRD (c=0.562), post-HSCT MRD (c=0.616) and pre- and post-MRD dynamics (c=0.648). Our results confirm the outcome predictive value of dynamic peri-HSCT MRD either alone or in combination with other variables for patients with T-ALL.

transplantation outcomes in patients with T-ALL. In addition, we also tried to establish a risk score principally based on the dynamic peri-HSCT MRD combined with other parameters, such as remission status pre-HSCT and onset of chronic GVHD demonstrated by others [12,26,27] and us [2,16,19] , which might provide better relapse risk determination for T-ALL patients.

Study Design
This retrospective study included T-ALL subjects who were enrolled at the Peking University People's Hospital between January 2010 and December 2018. For patients with human leukocyte antigen (HLA)matched sibling donors (MSDs), MSDs were chosen. If cases without MSDs, HLA-matched unrelated donors (MUDs) were chosen. If cases without MSDs and MUDs, then haploidentical donors were chosen [28,29] . All of the included subjects signed an informed consent form. The study protocol was in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of Peking University. All of the cases were treated according to the transplant protocol as previously described [1,17,30] .

Transplant Procedures
Recombinant human granulocyte colonystimulating factor (G-CSF; 5 μg/kg per day for 5 days) were administered to healthy donors for bone marrow stem cells (BMSCs, collected on day 4 after G-CSF) and peripheral blood stem cells (PBSCs, collected on day 5 after G-CSF) mobilization [21,30] . Subjects received BMSCs and/or PBSCs as allografts.

MFC Detection of MRD
Bone marrow aspirate samples were obtained as part of the baseline assessment before SCT, as well as 1, 2, 3, 4.5, 6, 9, and 12 months posttransplantation and at 6-month intervals thereafter according to previous studies [2,16,21,30] . Six-to eight-color MFC was performed for MRD evaluation according to previous studies [2,16,21] . A panel of antibody combinations recognizing cTdT, mCD3, cCD3, CD5, CD7, CD34, CD45, and CD2 or CD99 was used for MRD determination. Any measurable level of MRD was considered positive, otherwise was defined as negative. The definition for quantitative dynamics of pre-MRD and post-MRD included: (1) post-MRD increase after allograft compared with the pre-HSCT baseline; (2) post-MRD non-increase was defined as not meeting the criteria of (1) and (3); (3) unchanged negative MRD pre-and post-HSCT.

Methods for MRD Intervention and Relapse Treatment
Donor lymphocyte infusion (DLI) was performed as described previously by our group [1,17,31] . Other methods for positive MRD intervention and relapse treatment, such as interferon-γ (IFN-γ), were administered according to our previous studies [1,17,31] .

Outcomes
The primary study end point was the cumulative incidence of leukemia relapse. The secondary end points were the cumulative incidence of non-relapse mortality (NRM) and the probabilities of leukemia-free survival (LFS) and overall survival (OS).
The engraftment, infection, NRM, relapse, LFS, and OS were defined according to our previous studies [1,17,31,32] . The definition and grades of acute GVHD were based on the pattern and severity of organ involvement [32] . The chronic GVHD was defined and graded according to the National Institute of Health criteria [33] .

Statistical Analysis
Patient characteristics were compared between the negative MRD and positive MRD groups with the χ 2 statistic for categorical variables and the Mann-Whitney test for continuous variables. Cumulative incidence curves were used in a competing risk setting, with relapse treated as a competing event to calculate NRM probabilities, and with death from any cause as a competing risk for GVHD, engraftment, and relapse. The probabilities of LFS and OS were estimated with the Kaplan-Meier method. The variables in table 1 were included in the univariate analysis. Only variables with P<0.1 were included in a Cox proportional hazards model with time-dependent variables. We calculated C-statistics (c), whereby a c value of 1.0 indicates perfect discrimination, and a c value of 0.5 is equivalent to chance [34] . Unless otherwise specified, P values were based on two-sided hypothesis tests. Alpha was set at 0.05. Most analyses were performed with SPSS software, version 16.0 (Mathsoft, USA).

Effects of Peri-HSCT MRDs on Outcomes
We first investigated the effects of pre-HSCT MRDs or post-pre-HSCT MRDs on transplant outcomes in total cases. The patients in the positive pre-HSCT MRD group experienced higher CIR (45% vs. 22%, P=0.003), lower LFS (44% vs. 65%, P=0.012) and OS (50% vs. 67%, P=0.051) than those in the negative pre-HSCT MRD group, but there was no significant difference in NRM between the positive and negative pre-HSCT MRD groups (11% vs. 13%, P=0.805). Patients with positive post-HSCT MRDs had higher CIR (72% vs. 19%, P<0.001), lower LFS (22% vs. 67%, P<0.001) and OS (32% vs. 69%, P<0.001) than those with negative post-HSCT MRDs, but no significant difference was found in NRM (6% vs. 14%, P=0.193) ( fig. 1 and table 2 We further evaluated the effects of peri-HSCT (Continued from the last page)    The predictive value of peri-HSCT MRD on outcomes was also observed after analysis in the adult and pediatric subgroups (data not shown).

A Risk Score for CIR Prediction
We applied multivariate Cox regression analysis with stepwise forward selection based on the data of total patients. The final model included remission status before transplantation (CR1 scores: 0; ≥CR2 scores: 1), onset of chronic GVHD (with chronic GVHD scores: 0; without chronic GVHD scores: 1) and dynamics of pre-and post-HSCT MRD (pre-MRDneg and post-MRDneg, MRD non-increase, and MRD increase, scores: 0, 1, and 2, respectively). According to the risk score categories, we classified each patient into one of four prognostic risk groups: low-risk (score=0), intermediate-risk (score=1), high-risk (score=2) and very high-risk (score=3).

DISCUSSION
In consistent with previous studies either in total ALL patients [16,20] or B-ALL subgroup patients [36] , we, in the present study, found that dynamic peri-HSCT MRD in patients with T-ALL could be better for discrimination of relapse risk than that of pre-HSCT MRD or pos-HSCT MRD. In addition, we showed that a risk score principally based on dynamic peri-transplantation MRD could further achieve better relapse stratification than dynamic peri-HSCT MRD alone (c-index=0.730 vs. 0.648). Overall, our results add new evidence for the application of MRD, suggesting the usefulness of dynamic peri-HSCT MRD for stratification of T-ALL patients with high risk recurrence.
In a recent study including 477 B-ALL patients who underwent allo-HSCT, Cao et al [19] demonstrated that post-HSCT MRD, but not pre-HSCT MRD was associated with higher CIR and shorter survival after multivariate analysis. In contrast to the result by Cao et al [19] , we found that both pre-and post-MRD could be used for discriminating patients into different relapse risk groups for T-ALL patients, although post-HSCT MRD was better than pre-HSCT MRD in predicting leukemia relapse (table 2 and fig. 1). The abovementioned differences might be related to the higher CIR and shorter survival in patients with T-ALL who underwent allografts than in B-ALL patients who received allo-HSCT [2] . However, the results of the present study are consistent with those of previous study including both T-ALL and B-ALL cases receiving allograft, reported by Bader et al [12] . They reported a higher c-index for the post-HSCT MRD than that of pre-HSCT MRD (c-index=0.649 vs. 0.612). Considering MRD was detected at different time points, several studies focused on the dynamic MRD change in predicting CIR for patients with acute leukemia [12,16,20,25,37,38] . For 279 patients with AML receiving allografts, pre-HSCT MRD and post-HSCT MRD were determined by MFC. Zhou et al [38] found that patients with increased MRD levels around the time of transplantation experienced higher CIR and shorter survival than those with decreased MRD levels or those with negative pre-HSCT MRD and negative post-HSCT MRD. We have also confirmed the results reported by Zhou et al [38] either in total AML patients or in pediatric and young adult ALL cases. Here, we confirmed the superiority of dynamic peri-HSCT MRD compared to pre-HSCT MRD and post-HSCT MRD in T-ALL patients receiving allo-HSCT (table 2 and fig. 1). The results of our previous study [16] and other studies [20,38] suggest that dynamic peri-HSCT MRD should be routinely used for CIR discrimination in allograft settings for patients with acute leukemia.
In a large cohort of study including 616 pediatric and young adult ALL patients receiving allografts, Bader et al [12] showed that these patients could be classified into three different relapse risk groups according to a prognostic risk score established based on remission status before transplantation, conditioning regimen and pre-HSCT MRD. In the present study, except for dynamic MRD peri-transplantation, remission status pre-HSCT and onset of chronic GVHD were also independently correlated with CIR and survival. We found, in subgroup patients with T-ALL who underwent allograft, a risk score principally based on dynamic peri-HSCT MRD as well as remission status and chronic GVHD could further classify patients into four subgroups with different CIR and survival (table 2 and fig. 1). Therefore, the results in the present study and previous studies shown by others [12,39] and us [36] suggest that risk scores based on MRD and other variables could be better in predicting transplant outcomes for ALL patients, especially those of T-ALL with a CIR more than 50% (table 2 and fig. 1).
There are limitations of our study. First, this is a retrospective, single center study. Second, the present study only enrolled T-ALL patients who underwent haploidentical HSCT and MSDT. Third, we did not perform subgroup analysis of cases who received MSDT due to the small number of patients in subgroup. Therefore, a prospective, multicenter study with training and validation sets is needed to further confirm whether our findings are suitable for T-ALL cases who either received haploidentical HSCT (including haploidentical HSCT based on immune tolerance induced by post-transplantation cyclophosphamide), MSDT, MUD transplantation [5] or umbilical cord blood transplantation [40] .
In summary, our results suggest a superiority of dynamic peri-HSCT MRD to single time point, including pre-HSCT MRD and post-HSCT MRD, in relapse risk stratification for patients with T-ALL. We further suggest that T-ALL patients who would experience the worst outcome could be discriminated by a risk score principally based on dynamic peri-HSCT MRD, for these cases, novel strategies are needed to improve the transplant outcomes.

Acknowledgement
We would like to thank Miss Xu Wu for her assistance in minimal residual disease detection.

Open Access
This article is licensed under a Creative Commons Attribution 4.0 International License https://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/.