Serial Clinical and Biomarker Monitoring during Graft-Versus-Host Disease Treatment Identifies Distinct Risk Strata Including an Ultra-Low Risk Group

Background: The standard treatment for acute graft-vs-host disease (GVHD), a common complication following allogeneic hematopoietic cell transplant, remains prolonged courses of high dose corticosteroids. Previous attempts to decrease corticosteroid exposure during GVHD therapy failed because physicians lack the tools necessary to safely reduce and shorten therapy and fear loss of GVHD control in responding patients. Prior studies have shown that a serum biomarker risk score, the MAGIC algorithm probability (MAP), provided prognostic value within groups with similar clinical severity and that patients with GVHD that is Minnesota standard risk by clinical symptoms and who have a low MAP at the start of corticosteroid treatment represent a low risk group with good outcomes.

Objective: This study tested the hypothesis that serial monitoring of GVHD symptoms and the MAP score in patients with low risk GVHD could provide further risk stratification, and would identify a subset with exceptionally low rates of failure with standard treatment, which we term ultra-low risk (ULR) GVHD who might benefit from reduced corticosteroid treatment.

Study design: Weekly monitoring of clinical symptoms and MAPs from initiation to day 14 of treatment was used to further divide 450 patients with low risk GVHD into groups with different outcomes, such as overall response rates at day 28 and non-relapse mortality at six-months.

Results: 310/450 low risk patients (69%) who achieved clinical response by day 14 and had low MAPs at days 7 and 14 constituted an ultra-low risk (ULR) group. that experienced a significantly higher overall response rate at day 28 (93% vs 50%, p<0.001) that was sustained to day 56 (84% vs 45%, p<0.001) and significantly lower six-month NRM (4% vs 13%, p<0.001) compared to the non-ULR patients. Patients who achieved clinical response by day 14 but who developed a high MAP during monitoring (n=20) experienced six-fold higher six-month NRM than the ULR group (25% vs 4%, p<0.001). Among 120 patients who did not achieve a clinical response by day 14, the overwhelming majority (n=112) who maintained low MAPs at both days 7 and 14 of treatment experienced six-fold lower NRM at six months compared to patients with a high MAP at either time point (8% vs 50%, p<0.001). The majority of deaths within the ULR group were due to infections in patients with complete and sustained control of GVHD symptoms while the majority of deaths in the non-ULR group were due to poorly controlled GVHD.

Conclusions: Serial monitoring during treatment can identify a large subset of patients by day 14 who achieve excellent GVHD control but remain at risk for treatment complications with standard treatment and who might be suitable candidates for testing abbreviated corticosteroid courses.