Long-term outcomes by biomarker probabilities in early treatment–resistant patients. Early treatment–resistant patients were subdivided based on biomarker probabilities into low and high groups. (A) Test cohort of patients (n = 122). Twelve-month cumulative incidence of NRM (low 22% vs high 63%, P < .001) and OS (low 68% vs high 34%, P < .001) and proportion of patients resistant to treatment at week 4 (low 33% vs high 82%, P < .001). (B) Validation cohort 1 (n = 80). Twelve-month cumulative incidence of NRM (low 13% vs high 67%, P < .001) and OS (low 76% vs high 26%, P < .001) and proportion of patients resistant to treatment at week 4 (low 45% vs high 71%, P = .03). (C) Validation cohort 2 (n = 68). Twelve-month cumulative incidence of NRM (low 14% vs high 75%, P < .001) and OS (low 78% vs high 14%, P < .001) and proportion of patients resistant to treatment at week 4 (low 29% vs high 68%, P = .004).