Functional precision approach in patients with very high risk acute lymphoblastic leukaemia in India: a single-centre cohort study

Abstract

Background: Persistence of measurable residual disease (MRD) and high-risk cytogenetics are established predictors of relapse in childhood acute lymphoblastic leukaemia (ALL).

Methods: Outcomes of children with ALL treated with the ICiCLe-ALL-2014 protocol at a single centre, between August 2013 and May 2023 were analysed. Co-culture ex-vivo drug response profiling (DRP) was performed on diagnostic or relapsed samples. Patients classified as very high risk (VHR) received DRP guided therapeutic modifications. Event free (EFS) and overall (OS) survival were compared across risk categories.

Findings: Among 715 patients, at a median 55 (50-58) months, the 3-year EFS for standard-risk, intermediate-risk, high-risk B, T-ALL and VHR were 71% (64%-78%), 67% (58%-75%), 77% (70%-82%), 81% (71%-88%), and 38% (24%-52%) respectively (p < 0.0001). Persistent MRD at end of consolidation was associated with inferior EFS (40.3%, p ≤ 0.0001). Drug sensitivity scores from DRP performed on 112 samples identified panobinostat (median DSS 23.4), venetoclax (20.7), daunorubicin (17.9), selinexor (12.7) and bortezomib (12.1) as effective in VHR or relapsed ALL. From November 2020, 25 VHR patients received a modified treatment block incorporating venetoclax and bortezomib. At 1.5-year, landmark EFS was 81.8% (58%-93%) with the modified regimen vs 67.7% (49-81) with standard therapy (p = 0.0324). Venetoclax sensitivity correlated with MRD clearance (p = 0.0070).

Interpretation: DRP enabled identification of effective agents for integration into therapy of VHR paediatric ALL. The addition of venetoclax and bortezomib was well tolerated and associated with improved early survival outcomes. These findings support prospective evaluation of DRP-guided treatment regimens in VHR ALL.

Funding: DBT-Wellcome India Alliance, Tata Consultancy Services.

Keywords: Childhood cancers; Drug response profiling; Leukaemia; Precision oncology; Very high risk.

Fig. 1

Patient risk stratification and treatment assignment for children with acute lymphoblastic leukaemia (ALL). Flow diagram illustrating risk stratification for 715 children diagnosed with ALL.

Fig. 2

Ex-vivo drug response profiling reveals heterogeneous sensitivity patterns across clinical subgroups in ALL. (a) Schematic representation of the drug response profiling (DRP) workflow. (b) Hierarchical clustering heatmap of DSS values across 112 patient samples and 12 drugs. (c) Comparison of DSS between group 1 (SR/IR) and group 2–4 (HR/VHR/relapse). Boxplots display DSS distribution for each drug. (d) Comparison of DSS between newly diagnosed and relapsed ALL patients. Boxplots show drug-wise DSS variation across the two time-points. p < 0.05 considered significant. ^∗indicates p < 0.05; ^∗∗indicates p < 0.01, ns indicates not significant differences.

Fig. 3

Ex-vivo drug sensitivity in samples from patients with Relapse/VHR ALL. Boxplots showing Drug Sensitivity Scores (DSS) across 67 samples from patients with relapsed/VHR ALL. Each dot represents an individual sample.

Fig. 4

Modification of delayed intensification (DI) block based on findings of drug response profiling improves response and survival rates in VHR ALL. (a) Schematic of modified DI block. Agents highlighted in purple are the new drugs. (b) Decision-making algorithm for eligibility for modified DI. VHR criteria, general condition, and DRP results were reviewed by a leukaemia tumour board. (c) Longitudinal MRD response in VHR patients treated with modified DI. (d) Landmark analysis comparing event free survival (EFS) between modified DI and standard DI in VHR ALL.

Fig. 5

Correlating ex-vivo drug sensitivity (venetoclax, bortezomib) with clinical response to modified delayed intensification. (a, b) Correlation between Drug Sensitivity Scores (DSS) and end-of-modified DI (Eo-Modified DI) MRD levels. (c, d) Receiver operating characteristic (ROC) curves evaluating DSS as a predictor of MRD response. (e, f) Contingency matrices comparing DSS-based classification with clinical MRD response.