Oncogenetic Risk Classification Based on NOTCH1/FBXW7/RAS/PTEN Mutation Profiles Improves Outcome Prediction in Pediatric T-Cell Acute Lymphoblastic Leukemia, Treated According the Fralle 2000 T Guidelines

Background: Risk stratification in childhood T-cell acute lymphoblastic leukemia (T-ALL) is crucial to drive treatment decisions. Since patients with induction failure or relapse are often refractory to further treatment, identifying high risk patients up-front will allow improved treatment. While minimal residual disease (MRD) is the strongest prognosis risk factor used after complete remission (CR), NOTCH1/FBXW7 (N/F) and RAS/PTEN (R/P) mutation profiles at diagnosis have recently been identified to predict outcome in adult T-ALL. Objective: to test whether an oncogenetic classifier using N/F and R/P mutations could improve the detection of children with T-ALL at risk of relapse. Methods: 405 patients with T-ALL aged from 1 to 14 years were treated according to FRALLE T guidelines (FRALLE Study group) between 2000 and 2010. Among them, 220 patients, for whom biological material at diagnosis was available, were tested retrospectively for N/F and R/P mutations. These study cohort patients were representative of overall FRALLE 2000 T-ALLs. CR was achieved in 213 patients. MRD (IgH-TCR markers) tested at CR (day 35) was available for 191 patients. MRD was <10-4 for 114 patients (60%) and ≥10-4 for 77 patients. Patients with N/F mutation and R/P germline (GL) were defined as oncogenetic low risk (LoR), while N/F GL and R/P GL or mutation and N/F mutation and R/P mutation were defined as high risk (HiR). Results: 111 patients were classified as LoR and 109 as HiR. Five-year-CIR and DFS were respectively 35.5% (95% CI, 26.7-44.3) and 59% (95%CI, 50.2-69.6) for HiR versus 13% (95% CI, 6.8-19.2) and 86.8% (80.5-93.5) for the LoR group (Figures A and B). HiR patients were significantly associated with MRD ≥ 10-4 (p=0.0004) and higher risk of relapse (p=0.00002). Among patients with MRD ≥ 10-4, HiR feature worsened the risk of relapse: 5-year-CIR and DFS were respectively 42.8% (95% CI, 28.9-56.7) and 71.1% (95%CI, 56.0-90.2) in HiR versus 28.9% (95% CI, 11.7-46.1) and 50.9% (95%CI, 38.4-67.6) in the LoR group. Among patients with MRD <10-4, 5-year-CIR and DFS were respectively 28.9 % (95% CI, 15.0-42.8) and 71.0% (95%CI, 58.4-86.3) in HiR group versus 4.4% (95% CI, 0-9.2) and 95.5% (95%CI, 90.7-1.00) in LoR group (Figures C and D). As such, the classifier allowed identification of 63% of very low risk patients amongst the MRD<10-4 population. Prognostic values of new oncogenetic risk factors were then analyzed with conventional factors. By univariate analysis, factors identified to predict relapse were male gender (p=0.036), WBC count ≥ 200 G/L (p=0.023), chemoresistance at day 21 (p=0.007), MRD ≥10-4 (p=0.0006) and oncogenetic HiR (p<0.0001). A multivariable cox model including these variables selected the classifier together with WBC count, day 21 chemo-sensitivity and MRD. Based on a stepwise selection procedure, the three most discriminating variables were classifier, WBC count and MRD. The cause specific Hazard Ratio (HR) was 3.22 (95% CI, 1.64-6.28) for oncogenetic HiR versus LoR (p=0.0006), 2.30 (95% CI, 1.26-4.20) for MRD≥10-4 versus MRD<10-4(p=0.0070) and 1.85 (95% CI, 1.01-3.37) for WBC≥200G/L versus <200 G/L (p=0.0456). Based on these three parameters, 8 subsets of patients were defined according to the estimated 5-year CIR. The 58 patients (30%) associating WBC count < 200G/L, classifier LoR and MRD<10-4 were at very low risk of relapse, with a 5-y-CIR of 1.7%. Patients harboring at least one of: WBC count ≥200G/L, classifier HiR or MRD>10-4, demonstrated an increasing CIR, up to 45.8% if all three were associated. Conclusion: in childhood T-ALL, oncogenetic classification using N/F and R/P mutation profiles is an independent predictor of relapse. When combined with MRD and WBC count ≥200 G/L, it significantly improved relapse prediction, particularly amongst the 60% of T-ALLs with MRD <10-4 at day 35. Appropriate integrating these 3 factors, will help optimize treatment. Figure Figure. Disclosures No relevant conflicts of interest to declare.