The Mount Sinai Acute Gvhd International Consortium (MAGIC) Model: An Integrated Clinical and Biomarker Grading System for Acute Graft-Versus-Host Disease (GVHD)

• Published in: Transplantation and cellular therapy 2024-02, Vol.30 (2), p.S52-S53
• Main Authors: Akahoshi, Yu, Spyrou, Nikolaos, Weber, Daniela, Aguayo-Hiraldo, Paibel Ixia, Ayuk, Francis, Bunworasate, Udomsak, Choe, Hannah, Eder, Matthias, Etra, Aaron, Grupp, Stephan A., Hexner, Elizabeth O, Hogan, William J., Kitko, Carrie L., Kraus, Sabrina, Merli, Pietro, Qayed, Muna, Reshef, Ran, Schechter, Tal, Ullrich, Evelyn, Vasova, Ingrid, Wölfl, Matthias, Zeiser, Robert, Baez, Janna, Beheshti, Rahnuma, Gleich, Sigrun, Katsivelos, Nikolaos, Kowalyk, Steven, Morales, George, Young, Ms. Rachel, Chen, Dr. Yi-Bin, Nakamura, Ryotaro, Ferrara, James, Levine, John E
• Format: Article
• Language: English
• ISSN: 2666-6367; 2666-6367
• DOI: 10.1016/j.jtct.2023.12.087

Current acute graft-vs-host disease (GVHD) risk stratification systems use either biomarkers (e.g., Ann Arbor scores) or clinical symptoms (e.g., Minnesota risk), but not both. The Minnesota system (standard or high risk) lacks a low risk stratum needed for treatment de-escalation. This MAGIC study had two goals: to determine whether a risk model based on clinical symptoms alone could define three separate groups and to determine whether inclusion of biomarker risks could further improve the prediction of risk. We analyzed 1863 patients who underwent a first HCT for hematological disorders between 2014 – 2021 who received systemic treatment for acute GVHD. We randomly divided them into training (n=1306) and validation (n=557) cohorts so that there were sufficient numbers of patients with less common clinical presentations (e.g. isolated stage 2 or stage 3 GI GVHD) in the training cohort. We divided patients in the training cohort into 15 groups based on clinical similarity and non-relapse mortality (NRM) and then used a classification and regression tree (CART) model to create 3 Manhattan risk groups that were independently confirmed by unsupervised K-means clustering (Table 1). We next compared the Minnesota and Manhattan risk systems in the validation cohort. Half of Minnesota standard risk patients were now categorized as Manhattan low risk and the three Manhattan strata had distinctly different NRM and overall response rates (ORR) to treatment at day 28 (Figure 1). Manhattan risk had a significantly greater area under the receiver operating characteristic curve (AUC) than Minnesota risk in predicting 6-month NRM (0.69 vs. 0.64, P = 0.009). Each Manhattan risk group (low, intermediate, high) was further stratified by Ann Arbor (AA) biomarker scores (1-3) for risk of NRM (data not shown). We again applied a CART analysis to the training cohort to the 9 combinations of Manhattan risk and AA scores to create a new MAGIC risk model of 3 strata that was also confirmed by K-means clustering (Table 2). This analysis only included patients (78%) with serum samples at the time of GVHD treatment initiation. In the validation cohort, the incidence of NRM increased for each successive MAGIC risk group and the ORR decreased accordingly (Figure 2). The AUC of the MAGIC model was greater than the Manhattan model (0.79 vs. 0.70, P = 0.009). In summary, the symptom-based Manhattan model contains 3 risk groups that accurately predicts treatment-response and NRM. The MAGI