A Biomarker-Integrated Day 7 Response Reliably Predicts Outcomes in Minnesota Standard Risk Acute GVHD

Acute graft-versus-host disease (GVHD) response is conventionally assessed 28 days after treatment initiation, although treatment escalation often occurs earlier. In Minnesota standard risk acute GVHD, day 7 clinical response alone provides limited prognostic discrimination because many patients without early improvement subsequently respond without additional therapy. We hypothesized that incorporating serum GVHD biomarkers into a day 7 response assessment would improve early prognostic stratification. We studied 790 patients with Minnesota standard risk GVHD treated with systemic therapy. A MAGIC observational cohort was randomly divided into training (n=294) and validation (n=308) cohorts, and an independent interventional trial cohort (n=188) served as external validation. A classification and regression tree integrated day 7 clinical response with Ann Arbor biomarker scores to predict 6-month non-relapse mortality (NRM). The integrated model reclassified approximately 20% of patients, identifying clinical non-responders with low-risk biomarkers who had favorable outcomes and clinical responders with high-risk biomarkers who remained at increased risk. In the training cohort, the integrated model improved prediction of 6-month NRM compared with day 7 clinical response alone (AUC, 0.76 vs 0.63; P=.002) and performed comparably to conventional day 28 clinical response (AUC, 0.70). These findings were reproduced in both validation cohorts. Among day 7 clinical non-responders, biomarker status identified biologically favorable slow responders and also stratified outcomes among patients who received second-line therapy. A biomarker-integrated day 7 response assessment improves early prognostic stratification in Minnesota standard risk acute GVHD and may provide an earlier, more objective endpoint for clinical decision-making and future therapeutic trials.