ABSTRACT Objective To determine whether early dexamethasone response provides complementary prognostic information to minimal residual disease (MRD) and whether integrating both markers refines risk stratification in pediatric acute lymphoblastic leukemia (ALL). Methods We retrospectively analyzed pediatric ALL patients treated at a single center according to the CCCG‐ALL‐2015 protocol. Day 5 steroid response was classified as dexamethasone good response (DGR) or dexamethasone poor response (DPR). Bone marrow MRD was measured on Days 19 and 46 of induction. A four‐quadrant classification was constructed by combining Day 5 steroid response with Day 46 MRD status (DGR/MRD−, DGR/MRD+, DPR/MRD−, DPR/MRD+) to evaluate event‐free survival (EFS) and cumulative incidence of relapse (CIR). Results Among 312 patients, 220 were DGR and 92 were DPR. MRD negativity rates on Days 19 and 46 were significantly lower in DPR. Among 304 patients with paired MRD measurements, 181, 102, 18, and 3 patients were classified as D19−/D46−, D19+/D46−, D19+/D46+, and D19−/D46+, respectively, with progressively worse 5‐year EFS and increasing 5‐year CIR. In the four‐quadrant analysis, EFS differed significantly across groups, and compared with DGR/MRD−, the other quadrants exhibited higher event risk. In multivariable Cox models, Day 46 MRD remained independently associated with inferior EFS (HR = 2.429, 95% CI: 1.286–4.590; p = 0.006). When the four‐quadrant classification was entered into the multivariable model (reference: DGR/MRD−), DGR/MRD+ (HR = 4.319, 95% CI 1.792–10.414; p = 0.001), DPR/MRD− (HR = 2.207, 95% CI: 1.253–3.888; p = 0.006), and DPR/MRD+ (HR = 3.718, 95% CI: 1.608–8.598; p = 0.002) were each associated with increased event risk. Conclusions MRD kinetics identify patients with delayed clearance or persistent positivity at higher relapse risk. Integrating early steroid response with MRD using a four‐quadrant framework may complement MRD‐based assessment and further refine risk stratification by highlighting clinically meaningful discordant subgroups.
Ou et al. (Mon,) studied this question.