A Study on the Relationship between Ex Vivo Drug Sensitivity and Clinical Outcome of Acute Lymphoblastic Leukemia

Novel immune-cellular target therapies have been actively developed and introduced in the real-world treatment of acute lymphoblastic leukemia (ALL). However, we are still using multi-agent chemotherapy to reduce the initial tumor burden and facing many relapsed or refractory (R/R) patients with poor survival outcomes. Conventional risk factors and molecular cytogenetic profiles substantially influence treatment selection, but the expected response to a specific treatment based on the data from a large group of patients with a similar mutational profile is not always equal to the actual response observed on an individual basis. Previous studies suggest that ex vivo drug sensitivity (DS) analysis enables more accurate prediction of individual responses, allowing personalized treatment of lymphoma and acute myeloid leukemia (AML). In this study, we measured the ex vivo sensitivity of patient-derived cells to as many as 21 different anti-cancer drugs and assessed their prognostic utility. The results confirmed a strong utility of the drug sensitivities in predicting clinical outcomes.

Out of 37 ALL patients who participated, we performed ex vivo DS analysis on 27 patients (9 Ph-negative, 14 Ph-positive, and 4 T-ALL at diagnosis (n=17) and at relapse (n=10)) using bone marrow aspirates. The reasons behind not being able to conduct the DS analysis were insufficient number of cancer cells in the aspirates (90%) and delayed shipping (10%). Except for those cases, we were able to conduct DS analysis using the previous reported media and protocol for AML. Briefly, the isolated blast cells were incubated for 72 hours with various concentrations of up to 21 anti-cancer drugs, and changes in viability were measured using the Alamar Blue assay. For each drug, we calculated the individual patient's drug sensitivity in terms of the three metrics-IC₅₀ (the drug concentration to inhibit cell growth by 50%), area under the curve (AUC, area under the dose vs viability curve), and E_max (the remaining viability at the highest drug concentrations tested). Smaller values of IC_50, AUC, and E_max correspond to high drug sensitivities.

After sample acquisition, all patients received one of the following therapies: Modified hyper-CVAD (cyclophosphamide, vincristine, doxorubicin, dexamethasone, n=6) plus tyrosine kinase inhibitors (n=8) or L-asparaginase (n=2), inotuzumab ozogamicin (n=3), tyrosine kinase inhibitor (TKI) alone (n=3), blinatumomab (n=2), chimeric antigen receptor T-cells (CAR-T, n=1), and others (n=2). Complete response was achieved in 24 patients (89%), and minimal residual disease (MRD) was observed in 16 (67%) among the responders. During the follow-up, with median of 22 months, there were 3 refractory and 2 relapsed patients, 4 succumbed to death in remission, and 3 died due to leukemia progression.

Given the various disease subtypes and treatment modalities, we focused on the patients with available DS analysis results treated with modified hyper-CVAD +/- imatinib/L-asparaginase (n=16) for further analysis. Upon evaluating the predictive utility of the drug sensitivity metrics, we found that the E_max of cyclophosphamide or other alkylating agents (melphalan, busulfan or ifosfamide) was effective (ROC-AUC of 1.00, n=6) in predicting early disease-progression (<3 months) of Ph+ ALL. On the other hand, the E_max of decitabine was effective (ROC-AUC of 1.00, n=5) in predicting early disease progression of Ph- B-ALL. For T-ALL, the E_max of multiple drugs-decitabine, alkylating agents including cyclophosphamide, and mitoxantrone-were effective in predicting early disease progression (ROC-AUC of 1.00, n=2). Furthermore, the unsupervised clustering of the drug AUCs was effective in predicting MRD positivity (ROC-AUC of 0.85, n = 27).

In conclusion, we report that our proprietary ex vivo drug sensitivity platform may have clinical utility in identifying ineffective therapy for ALL patients across multiple subtypes and/or settings. DS analysis results of single drugs were able to prediction clinical outcome of modified hyper-CVAD +/- imatinib/L-asparaginase, which consists of at least four chemotherapeutic agents. E_max, which correlates to the proportion of the drug-resistant cancer cells, was the most useful in discriminating early disease progression, coinciding with our previous observation in AML.

Park:ImpriMed, Inc.: Consultancy, Current holder of stock options in a privately-held company. Lim:ImpriMed, Inc.: Current Employment, Current equity holder in private company. Ham:ImpriMedKorea, Inc.: Current Employment. Lee:ImpriMedKorea, Inc.: Current Employment. Park:ImpriMedKorea, Inc.: Current Employment. Yi:ImpriMedKorea, Inc.: Current Employment. Koo:ImpriMed, Inc.: Current Employment, Current equity holder in private company.