Day 7 mount sinai acute GVHD international consortium (MAGIC) algorithm probability predicts GVHD and non-GVHD-related non-relapse mortality in a contemporary cohort of allogeneic transplant patients

Abstract The Mount Sinai Acute GVHD International Consortium (MAGIC) developed and validated the MAGIC algorithm probability (MAP), which is a biomarker-based risk model using serum levels of ST2 and REG3α as predictive biomarkers of acute graft-versus-host disease (aGVHD)-related, non-relapse mortality (NRM), as early as 7 days after allogeneic hematopoietic stem cell transplantation (HCT) (Hartwell et al. JCI Insight 2017). This study included patients transplanted between 2005-2015, with a limited number receiving post-transplant cyclophosphamide (PTCy), and no specific analyses for older adults, racial/ethnic minorities or non-malignant diseases. We evaluated the impact of Day 7 MAP on HCT outcomes in a contemporary cohort of patients at our center between Jan 2021-Aug 2024. The primary endpoint was 1-year NRM. Secondary endpoints included overall survival (OS), progression-free survival (PFS), cumulative incidence of relapse (CIR), aGVHD, and GVHD-related mortality. We retrospectively reviewed a consecutive cohort of 721 patients who underwent commercially available MAP testing on Day 7 post-HCT (median: 9, range 5-14). The median age at HCT was 56 years (range 2-80), with 58.1% male; and 32.5% aged ≥ 65 years. The cohort reflected the demographics of our geographic catchment area: 40.8% Hispanic, 40.5% non-Hispanic White, or other (18.7%). Indications for HCT included AML (33.3%), ALL (22.5%), MDS/MPN (23.8%), lymphoma (6.7%), non-malignant disorders (8.3%), or other diagnoses (5.4%). HCT was from matched related (27.9%), matched/mismatched unrelated (56%), and haploidentical (16.1%) donor after myeloablative (35.2%) or non-myeloablative/reduced intensity conditioning (64.8%). GVHD prophylaxis was tacrolimus/sirolimus-base (59.1%), CNI/methotrexate-based (8.3%), PTCy-based (32%), and others (0.6%). We stratified 164 patients as high risk (HR) and 557 as low risk (LR) by MAP. The median follow-up for surviving patients (n=539) was 18.0 months. HR patients had significantly higher 1-year NRM vs LR patients: 29.8% (95% CI, 22.9-37.0%) vs 9.5% (95% CI, 7.1%-12.1%), p<0.01. Day 7 MAP was also associated with 1-year OS: 63.8% (HR) vs 84.0% (LR), p<0.01. MAP risk category was not associated with 1-year CIR (HR: 12.6% vs LR: 12.8%, p=0.36), grade 2-4 aGVHD (HR: 37.2% vs LR: 34.9%, p=0.35), or grade 3-4 aGVHD (HR: 12.8% vs LR: 7.9%, p=0.26). In multivariate analysis, MAP remained independently associated with NRM (hazard ratio 3.4; 95% CI 2.4-5.0; p<0.01) after adjusting for age (<65 vs ≥65; HR=1.6; p=0.048), HCT-CI (<3 vs ≥3; HR=1.5; p=0.047) and conditioning regimen (MAC vs RIC, HR=1.2, p=0.55). Cause-specific mortality analysis showed that HR patients had higher GVHD-related mortality (10.1% vs 4.3%, p<0.01) as well as non-GVHD-related NRM (19.7% vs 5.3%, p<0.01), primarily from infection or organ dysfunction. We evaluated the impact of Day 7 MAP on NRM within specific subgroups of interest. High MAP was associated with significantly higher incidence of NRM in adults aged ≥ 65 years (HR [n=71] 34.0% vs LR [n=163] 14.6%, p<0.01), those who received PTCy (n=230, HR [n=51] 39.5% vs LR [n=179] 9.8%, p<0.01), and racial/ethnic minority populations – specifically Hispanic patients (n=294, HR [n=62] 24.6% vs LR [n=232] 9.3%, p<0.01). MAP also stratified NRM risk, (HR [n=102] 33.0% vs. LR [n=325] 9.6%, p<0.01) in non-Hispanic patients (n=427). In an exploratory analysis, ST2 and REG3α levels demonstrated an inverse correlation in the HR group (Pearson R= -0.3317, p<0.0001, R2=0.1100), but no significant correlation was observed in the LR group (R=-0.0506, p=0.1167, R2= 0.0026). When the biomarker levels were evaluated individually in tertiles, there was a trend towards greater ST2 levels (high/medium tertiles) associated with non-GVHD NRM risk compared with GVHD-NRM (p=0.09), and not observed with REG3α.In this contemporary and diverse cohort of HCT recipients, the Day 7 MAP score was strongly predictive of NRM, including both GVHD-related and non-GVHD related causes. MAP effectively stratified risk across key subgroups, including PTCy recipients, older adults and racial/ethnic minorities. Day 7 MAP did not predict the incidence of grade 2-4 or 3-4 aGVHD, consistent with its role as a prognostic, not a diagnostic, tool. Our data support the continued development of GVHD-specific or non-specific preemptive strategies for HR patients, potentially guided by further analysis on the individual biomarkers and their expression patterns.