Nikolaos Katsivelos, MD, and John Levine, MD, MS, on How Serial Clinical and Biomarker Monitoring During Treatment Can Stratify Patients With Low-Risk GVHD

Dr. Levine: Acute graft-versus-host disease (GVHD) requires treatment in about 40% of allogeneic bone marrow transplant patients. Despite advances in prophylaxis such as post-transplant cyclophosphamide, GVHD treatment remains high-dose steroids tapered slowly over weeks to months despite a shift towards less severe phenotypes. Although real-world steroid tapering practices are very heterogeneous, cumulative steroid exposure is very high because physicians taper steroids slowly. Even in responders, they do this because they fear loss of GVHD control even though steroids cause many common complications. Thus, there's an unmet need to identify patients who can be safely tapered aggressively. Previous research has shown that clinical and biomarker criteria can classify new-onset GVHD into different risk groups. For example, the Minnesota Risk Stratification system stratifies patients by various combinations of target organ severity into two groups, a high-risk group and a standard-risk group. However, Minnesota standard-risk patients still have more than 20% six-month non-relapse mortality, which is too high to test the aggressive steroid tapers that might reduce steroid complications. The Mount Sinai Acute GVHD International Consortium, or MAGIC, has used serum biomarkers to generate risk scores called MAGIC Algorithm Probabilities or MAPs. MAPs are powerful predictors of response to treatment and non-relapse mortality. We previously published that patients with newly diagnosed GVHD that is both Minnesota standard risk and has a low MAP represent a group of low-risk patients with an 85% response rate to steroids and a non-relapse mortality of 10%. This group includes half of all patients treated for GVHD. We also have published that MAPs measured during treatment predict outcomes better than symptoms alone, even in responding patients. Given the data that clinical and biomarker criteria at the onset of GVHD symptoms and during treatment predict outcomes, we hypothesized that serial monitoring of GVHD biomarkers and clinical symptoms could stratify low-risk patients into clinically meaningful groups. In this study, we selected patients with low-risk GVHD, Minnesota standard risk, and a low MAP at the start of systemic treatment from the MAGIC database and biorepository, and developed new risk strata based on clinical and biomarker responses at days seven and 14 post-treatment. Nik Katsivelos, the postdoc in our MAGIC research team, performed these analyses. Dr. Katsivelos: The study population comprised 450 patients with low-risk GVHD at onset, whose MAPs we retrospectively determined at days seven and 14. The median age was 54, and the majority of patients were treated for skin GVHD, although 23% had lower GI tract involvement. We observed several different trajectories during the first two weeks of treatment. The majority of patients were early responders with low MAPs at day seven, and they maintained both at day 14. A second group of slow responders achieved a clinical response by day 14 and had low MAPs at both time points. Together, early and slow responders with low MAPs at both time points formed the group we called ultra-low risk. They had a very high response rate to steroids of 93% and very low six-month non-relapse mortality of 4%. This ultra-low risk group comprised 69% of the total population and 35% of all patients requiring systemic treatment. Patients who progressed by day 14, did not respond to steroids by day 14, had a transient clinical response at day seven that was lost before day 14, or who developed a high MAP at both time points were classified as non-ultra-low risk. This group was 31% of the original population, and their outcomes were significantly worse compared to the ultra-low risk patients. Only 50% of the patients responded to steroid treatment by day 28, and their non-relapse mortality was 13%. Thus, we were able to identify a group of ultra-low-risk patients with really excellent treatment outcomes from within the low-risk population. When we analyzed the cause of death by serial monitoring, we found that the majority of deaths in patients with ultra-low-risk GVHD were from fatal infections while GVHD was in sustained clinical response, a finding that highlights the downside of intense immunosuppression. In contrast, uncontrolled GVHD was causing the majority of deaths in the non-ultra-low-risk population. We also found that biomarkers stratified risk even in patients grouped by clinical response. Patients with clinical response and low biomarkers—the ultra-low-risk group—had six-month non-relapse mortality of 4% compared to 25% of clinical responders who developed a high MAP on either day seven or 14. Likewise, clinical non-responders who maintained low MAPs through day 14—the majority of non-responders—had non-relapse mortality of 8%, highlighting the ability of biomarkers to predict outcomes even in patients who do not clinically respond. Outcomes were terrible for the few patients who had both a clinical non-response and developed a high MAP, with non-relapse mortality of 50%. Dr. Levine: So to summarize: First, this study found that monitoring both clinical symptoms and biomarker scores from the start through day 14 of treatment generates new risk strata with distinctly different short- and long-term outcomes. Second, biomarker scores during treatment identify patients at both high and low risk irrespective of their clinical response. Third, ultra-low-risk GVHD—a large group that represents about a third of patients treated for GVHD—has exceptional outcomes. These patients may benefit from treatment de-escalation strategies such as shorter courses of lower-dose steroids, a strategy that we are currently testing in a MAGIC clinical trial.