ORLANDO – A biomarker algorithm was better than was clinical response at predicting outcomes after 1 week of systemic steroid treatment for graft-versus-host disease, according to findings from a multicenter study.
The findings have implications for early decision making regarding treatment course, Hannah Major-Monfried reported at the combined annual meetings of the Center for International Blood & Marrow Transplant Research and the American Society for Blood and Marrow Transplantation.
“The same GVHD algorithm that stratifies patients at day 7 after transplant, at [GVHD] diagnosis, also stratifies them after 1 week of steroid treatment into two groups with distinct risks for treatment failure, 6-month nonrelapse mortality, and overall survival,” she said.
The biomarker algorithm, which includes measures of ST2 and REG3-alpha, was previously shown to predict day-28 treatment response and 6-month non-relapse mortality (NRM) when applied at day 7 post transplant before the onset of GVHD and at the time of diagnosis, said Ms. Major-Monfried, a third-year medical student at Icahn School of Medicine at Mount Sinai, New York.
For the current analysis, levels of the biomarkers were measured after 1 week of treatment in 378 patients with acute GVHD from 11 centers in the Mount Sinai Acute GVHD International Consortium.
In a test cohort that included 236 of the patients, the measurements were used to generate a new predicted probability, or treatment score, for 6-month NRM, which had a value between 0 and 1.
Of the 236 patients, 93 (39%) were considered to have high posttreatment probability of NRM, and the remaining patients (61%) had low posttreatment probability of NRM, based on their treatment scores.
“High-risk patients were significantly less likely to respond to treatment than low-risk patients,” she said, noting that very similar results were found in a validation cohort of the remaining 142 patients, which had a similar proportion of high- and low-risk patients as did the test cohort.
The overall 6-month NRM for patients treated for GVHD was 27% in the test cohort. When the biomarker algorithm was used to separate the cohort into high- and low-risk groups, the NRM rate was found to be approximately 4 times higher among the high-risk patients than among the low-risk patients.
Overall survival was also significantly worse among high- vs. low-risk patients in both the test and validation cohorts.
“We can conclude that the increased NRM seen in the high-risk groups can explain these large differences in overall survival,” Ms. Major-Monfried said.
Because treatment decisions are often made after 1 week based on early clinical response, she and her colleagues also explored whether treatment response after 1 week could similarly predict NRM.
In the test cohort, early response – which includes complete or partial response in GVHD symptoms after 1 week of steroids – was observed in 48% of patients, while 52% were early nonresponders. NRM occurred in 17% of the early responders, compared with 36% of the nonresponders in the test cohort, and similar results were found in the validation cohort.
“These differences are independent of biomarkers,” she noted. “These are solely based on observed clinical response.”
When the biomarker algorithm was applied, prediction of NRM was more precise.
“We started with the early responders,” she explained. “When we used the biomarker algorithm to stratify these patients into high- and low-risk groups, we found that 28% of early responders were actually high risk, and that they experienced 38% NRM – significantly higher than the 8% observed in low-risk patients. Similar results were found again in the validation cohort.”
When the biomarker algorithm was used to stratify patients who were nonresponders at 7 days into high- and low-risk groups, 50% were found to be low risk, and those patients experienced 17% NRM, significantly lower than the 57% seen in the high-risk patients. Similar results were again seen in the validation cohort.
“Early responders with high posttreatment probability have high NRM, and perhaps should not be tapered despite the improvement of their clinical symptoms, while early nonresponders with low posttreatment probability have lower NRM and may not need treatment escalation,” Ms. Major-Monfried said.
“In data not shown, many of these patients are actually what we could call ‘slow responders’ who ultimately fare well,” she noted.
Ms. Major-Monfried reported having no disclosures.