Predictive scoring systems for molecular response assessment in pediatric chronic-phase chronic myeloid leukemia.

Abstract

256 Background: For pediatric patients with chronic myeloid leukemia (CML), accurately predicting at the start of imatinib therapy the likelihood of achieving a deep molecular response (DMR; at least MR4) is clinically important and may inform treatment goals and strategies. Methods: We analyzed 98 consecutive children with chronic-phase CML treated with frontline imatinib as a training cohort. Cumulative incidence functions were estimated and Fine–Gray subdistribution hazard models were used for univariable and multivariable analyses to identify covariates associated with MR4. A predictive scoring system was derived from the final multivariable Fine–Gray model. Results: In univariable analyses, older age, higher WBC, lower hemoglobin, lower platelet count, higher peripheral blood basophils, larger spleen size, and a lower 0–3-month BCR::ABL1 log-reduction were significantly associated with a lower cumulative incidence of DMR (all P < 0.05). In the multivariable Fine–Gray model, a lower 0–3-month BCR::ABL1 log-reduction (P < 0.001) and higher baseline WBC (P = 0.002) remained independently associated with a lower likelihood of achieving DMR. We developed a DMR predictive score using the final model coefficients: DMRscore = 0.6903 × (0–3-month BCR::ABL1 log-reduction) –0.3019 ×(WBC/100). Patients were stratified into low-, intermediate-, and high-score groups according to score tertiles: low (DMRscore ≤ −0.1389, n = 32, 33.3%), intermediate (−0.1389 < DMRscore < 0.9594, n = 32, 33.3%), and high (DMRscore ≥0.9594, n = 32, 33.3%). The cumulative incidence of DMR differed significantly across groups (Gray test P < 0.001); at 24 months, the DMR cumulative incidence was 22.6%, 69.3%, and 82.9% in the low-, intermediate-, and high-score groups, respectively. The model showed good discrimination with time-dependent AUROC values of 0.90, 0.84, and 0.80 at 12, 24, and 36 months, respectively, and favorable predictive accuracy (Brier score: 0.1555/0.1983/0.2114 at 12/24/36 months). Conclusions: We developed a MR4 risk score using routine clinical and early molecular kinetics in pediatric chronic-phase CML receiving frontline imatinib, enabling early risk stratification to guide individualized treatment planning.