PROGNOSTIC SCORES AND MOLECULAR RESPONSE IN CHRONIC MYELOID LEUKAEMIA: COMPARATIVE ANALYSIS AND PREDICTIVE MODELLING IN A REAL-WORLD COHORT

Prognostic stratification in chronic myeloid leukaemia (CML) is essential to guide early therapeutic decisions and monitor response to tyrosine kinase inhibitors (TKIs). Traditional scores such as EUTOS and ELTS, alongside molecular parameters like BCR::ABL1 levels, may offer complementary insights into disease aggressiveness and treatment outcomes. To compare clinical, laboratory, and molecular features according to EUTOS, ELTS, and baseline BCR::ABL1 levels, and to evaluate their predictive power for early molecular response (EMR) at three months. A retrospective analysis was conducted in a CML cohort (n = 176). Patients were stratified by EUTOS (cut-off: 87), ELTS (cut-off: 2.21), and BCR::ABL1 baseline load (cut-off: 100%). Variables were analysed using Mann–Whitney U tests and Spearman's correlations. Predictive modelling for EMR (BCR::ABL1 <10% at month 3) was performed using logistic regression and ROC curve analysis. Patients with EUTOS >87 had significantly lower haemoglobin (8.15 g/dL vs 10.4; p < 0.001), higher basophils (8.0% vs 3.0%; p < 0.001), splenomegaly (14.0 cm vs 5.0 cm; p < 0.001), and higher BCR::ABL1 at baseline and month 3. Time to TKI switch was shorter (20.7 vs 40.1 months; p = 0.029). ELTS >2.21 was associated with higher circulating blasts (19.45% vs 4.5%; p < 0.001), eosinophilia, and spleen size. Patients with BCR::ABL1 >100% showed poorer EMR (34.5% vs 5.1%; p = 0.014), though no other significant clinical or haematological differences were observed. Spearman analysis revealed moderate positive correlations between EUTOS and basophils (ρ = 0.65) and spleen size (ρ = 0.60), and negative correlation with age (ρ = –0.21) and haemoglobin (ρ = –0.31). ELTS was positively correlated with circulating blasts and spleen size, and negatively with age and haemoglobin. BCR0 was positively associated with CD34+ expression and LDH levels. In univariate logistic regression, BCR0 had the strongest predictive power for EMR failure (OR = 1.022; AUC = 0.763), followed by EUTOS (OR = 1.017; AUC = 0.703). ELTS had no predictive value (AUC = 0.471). In multivariate analysis, only BCR0 remained a significant independent predictor of EMR (OR = 1.019; p = 0.038), while EUTOS and ELTS lost significance. This study reinforces the clinical relevance of BCR::ABL1 baseline quantification as an early predictor of molecular response more robust than traditional prognostic scores in both univariate and multivariate contexts. While EUTOS retained some predictive value in univariate analysis, its overlap with molecular markers may limit independent utility. ELTS did not demonstrate significant discriminative capacity for early response. Baseline BCR::ABL1 transcript level is a robust and independent predictor of early molecular response in CML in this cohort. EUTOS may offer additional stratification in select contexts, but ELTS appears limited in its utility for predicting short-term outcomes. Incorporating molecular load at diagnosis could enhance individualised therapeutic strategies in CML management.