Prognostic value of blood biomarkers in steroid-refractory or steroid-dependent acute graft-versus-host disease: a REACH2 analysis

• Low levels of most aGVHDs (before/after treatment) and immune cell markers (before treatment) positively affect the probability of response.

• Models (baseline and day 14) can help identify patient characteristics and biomarkers for improved response to ruxolitinib vs BAT.

Systemic steroids are the standard first-line treatment for acute graft-versus-host disease (aGVHD), but ∼50% of patients become steroid-refractory or dependent (SR/D). Ruxolitinib is the only Food and Drug Administration– and European Medicines Agency–approved therapy for patients with SR/D aGVHD. In the phase 3 REACH2 trial (NCT02913261), ruxolitinib demonstrated superior efficacy in SR/D aGVHD, with a significantly higher overall response rate (ORR) on day 28, durable ORR on day 56, and longer median overall survival compared with the best available therapy (BAT). Identifying biomarkers and clinical characteristics associated with increased probability of response can guide treatment decisions. In this exploratory analysis of the REACH2 study (first biomarker study), we developed baseline (pretreatment) and day 14 models to identify patient characteristics and biomarkers (12 aGVHD-associated cytokines/chemokines, 6 immune cell types, and 3 inflammatory proteins) before and during treatment, which affected the probability of response at day 28. Treatment with ruxolitinib, conditioning, skin involvement, and age were strongly associated with an increased likelihood of response in the ≥1 model. Lower levels of most aGVHD and immune cell markers at baseline were associated with an increased probability of response. In the day 14 model, levels of aGVHD markers at day 14, rather than changes from baseline, affected the probability of response. For both models, the bias-corrected area under the receiver operating characteristic values (baseline, 0.73; day 14, 0.80) indicated a high level of correspondence between the fitted and actual outcomes. Our results suggest potential prognostic value of selected biomarkers and patient characteristics.