From an initial cohort of 1,341 patients, we enrolled 562 pediatric B-ALL patients who completed WTS at diagnosis under the CCCG-ALL-2015 protocol. As shown in Supplemental Data table S1, the study population had a median age of 4.8 years (range: 3.2–7.9 years) with a male predominance (60.14% male vs 39.86% female). With a median follow-up of 2.67 years (IQR: 1.53–3.80 years), we identified RAS pathway mutations in 224 patients (39.86%), including NRAS (n = 120), KRAS (n = 109), PTPN11 (n = 57), and NF1 (n = 11) variants. Additionally, FLT3 mutations were detected in 82 patients (14.59%).
Co-occurrence analysis between FLT3 and RAS pathway genes
As illustrated in Fig. 1, nearly half of the patients with FLT3 mutations also harbored concurrent alterations in RAS pathway genes. To assess potential associations between FLT3 mutations and RAS pathway alterations, we performed Pearson residual analysis and chi-square tests. Notably, as shown in Fig. 2, co-mutation of FLT3 and NF1 occurred more frequently than expected by chance, as evidenced by a high Pearson residual (3.5). Statistical analysis confirmed significant positive associations between FLT3 and NF1 mutations (Phi coefficient = 0.16, χ² = 11.518, p < 0.0001), and between NRAS and KRAS (Phi coefficient = 0.17, χ² = 15.713, p < 0.0001), indicating non-independence and a strong tendency for co-occurrence. In contrast, no significant correlation was observed between FLT3 mutations and alterations in NRAS, KRAS, or PTPN11 (Supplemental Data Table S2).
Clinical characteristics of the enrolled patients
We analyzed the clinical and cytogenetic characteristics of patients with RAS pathway and FLT3 mutations (Supplemental Data Table S3) and stratified the 562 cases into eight non-overlapping molecular subtypes based on WTS data. Patients harboring RAS pathway alterations exhibited a significantly higher prevalence of hyperdiploidy and DUX4 rearrangement (all p < 0.05), whereas the frequencies of ETV6::RUNX1, BCR::ABL1, and TCF3::PBX1 subtypes were markedly reduced compared to those without RAS pathway mutations (all p < 0.05). Among FLT3-mutated cases, hyperdiploidy was more common (p < 0.001), while TCF3::PBX1 and ETV6::RUNX1 subtypes were underrepresented (p < 0.05). Compared to wild-type patients, those with NRAS/KRAS co-mutations showed no significant differences in clinical characteristics, except for a markedly higher prevalence of hyperdiploidy (p < 0.001) and significantly lower frequency of ETV6-RUNX1 fusion (p = 0.03).
MRD was assessed on days 19 and 46 post-treatment. As shown in Supplemental Data table S3, PTPN11-mutated patients showed a significantly higher rate of MRD ≥ 0.1% than those without PTPN11 mutations (p < 0.001) at day 19. By day 46, two of six cases (33.33%) with concurrent FLT3 and NF1 mutations had MRD ≥ 0.1%, a proportion significantly elevated compared to patients lacking these co-mutations (1.55%, p = 0.005).
While not reaching statistical significance (p = 0.061), we observed that 4 of 6 (66.7%) patients with FLT3/NF1 co-mutations exhibited abnormal karyotypes. In contrast, all 5 patients with NF1 mutations alone showed normal karyotypes (Table 1). Among the FLT3-mutated cases, molecular characterization revealed that 5 patients harbored FLT3-tyrosine kinase domain mutations (TKD), while 1 case presented with a juxtamembrane region insertion mutation-both classified as non-ITD domain alterations. Additionally, all patients with FLT3/NF1 co-mutations were classified as B-other subtype and could not be assigned to any current molecular subtype of B-ALL.
Overall Survival
The overall 3-year OS rate for all enrolled patients was 97.89% ± 0.69%. Supplemental Data Table S4 demonstrates that both initial WBC (white blood cell) count ≥ 100×10⁹/L and Day 46 MRD ≥ 0.01% were significantly associated with inferior OS (p = 0.0072 and p = 0.04, respectively). Kaplan-Meier analysis of RAS pathway mutations and FLT3 mutations individually showed no significant prognostic impact (Supplemental Data Table S4).
Strikingly, patients with concurrent FLT3 and NF1 mutations exhibited markedly worse outcomes, with a 3-year OS of 75.0% ± 21.65% compared to 98.09% ± 0.64% in wild-type patients (p < 0.01; Fig. 1). This survival difference reached statistical significance (p = 0.003), while no other co-occurring mutations showed significant prognostic value (Supplemental Data Table S5).
Comparative survival analysis among molecular subtypes revealed distinct prognostic patterns (Fig. 1). Patients with FLT3/NF1 co-mutations demonstrated comparable OS to KMT2Ar subtype (p = 0.30), but exhibited significantly worse outcomes compared to non-KMT2Ar subtypes (p = 0.0022). Univariate analysis identified five clinical variables significantly associated with reduced OS (Supplemental Data Table S6). Subsequent multivariate Cox proportional hazards regression analysis established three independent predictors of shorter OS: FLT3/NF1 co-mutations (HR = 18.66, 95% CI = 2.20-158.11, p = 0.007), WBC ≥ 100×10⁹/L at diagnosis (HR = 6.83, 95% CI = 1.70-27.48, p = 0.007), and MEF2D mutated subtype (HR = 21.72, 95% CI = 2.57–183.50, p = 0.005) (Fig. 2).
Event-free Survival
The 3-year EFS for all enrolled patients was 89.86% ± 1.53% (Supplemental Data Fig. 1). Survival analysis identified several significant prognostic factors: male gender (p = 0.0014), WBC ≥ 100×10⁹/L (p < 0.0001), and MRD ≥ 0.01% at both day 19 and day 46 (all p < 0.05) were all associated with inferior EFS (Supplemental Data Table S5). Molecular characterization revealed that FLT3/KRAS co-mutations and ETV6::RUNX1 fusion correlated with favorable outcomes, while KMT2Ar predicted poor prognosis (all p < 0.05, Supplemental Data Table S5).
As shown in Fig. 1, comparative analysis demonstrated significantly worse 2 year-EFS in patients with FLT3/NF1 co-mutations versus those without (75.0% vs 93.30% at 2 years; p = 0.0061). When compared to other B-ALL subtypes, FLT3/NF1 co-mutated cases showed inferior outcomes. Notably, the prognosis of FLT3/NF1 co-mutated patients was comparable to KMT2Ar cases (p = 0.86) but significantly worse than non-KMT2Ar patients (p = 0.002).
We subsequently conducted univariate and multivariate Cox regression analyses to evaluate prognostic factors (Supplemental Data Table S6, Fig. 2). Among seven variables initially considered for multivariate modeling, five demonstrated independent prognostic significance. The final multivariate model identified FLT3/NF1 co-mutations (HR = 4.99, 95% CI = 1.17–21.30, p = 0.03) and KMT2Ar (HR = 3.92, 95% CI = 1.47–10.47, p = 0.006) as significant predictors of reduced EFS, along with WBC ≥ 100×10⁹/L and male gender (Fig. 2). However, apart from NF1 and KRAS, other genes in the RAS pathway, when co-occurring as mutated genes with FLT3, were not significantly associated with either EFS or OS probabilities using Kaplan-Meier methods based on log rank test (Supplemental Data Table S5).
Univariate and multivariable analysis of risk factors on MRD
Table 1 displays the distribution of MRD positivity rates at different threshold levels on Day 19 and Day 46. As shown in Fig. 2, multivariable analyses revealed that ZNF384 (OR 4.64, 95% CI 1.29–16.72), Ph-like (OR 7.28, 95% CI 1.84–28.72), KMT2Ar (OR 4.98, 95% CI 1.7-14.61) subtypes were independent risk factors for Day 19 MRD ≥ 1%. Notably, ETV6::RUNX1 subtype showed consistent protection across all MRD thresholds (≥ 1%, ≥ 0.1%, and ≥ 0.01%), in contrast to DUX4 which remained a risk factor at all levels (all p < 0.05). Additionally, hemoglobin levels (OR 0.99, 95% CI 0.98-1.00), PTPN11 mutations (OR 2.67, 95% CI 1.41–5.04), and Ph-like subtype (OR 9.91, 95% CI 1.98–49.48) were independently associated with Day 19 MRD ≥ 0.1%, while ZNF384 subtype (OR 5.98, 95% CI 1.27–28.13) was specifically associated with Day 19 MRD ≥ 0.01% (Fig. 2).
For Day 46 MRD analysis, limited case numbers restricted evaluation to the ≥ 0.01% threshold, where male sex (OR 2.51, 95% CI 1.06–5.95), FLT3/NF1 co-mutations (OR 7.08, 95% CI 1.22–41.09), and Ph-like subtype (OR 5.56, 95% CI 1.05–29.53) showed significant independent associations with MRD persistence in multivariable analysis (Fig. 2).