Introduction The therapeutic strategy of combining the hypomethylating agents (HMAs) Azacitidine (Aza) and Decitabine (Deci) with the BCL-2 inhibitor Venetoclax (Ven) has significantly improved the clinical outcomes of patients with Acute Myeloid Leukemia (AML), reaching overall response rates of up to 70% in clinical trials (Wei et al, JCO 2021). Still, on an individual basis, clinicians struggle to predict which patients will respond to Ven/HMA treatment. Furthermore, there is little data to guide selection of which HMA to use and the decision is generally based on physician preference and logistics. Functional ex-vivo testing using patient-derived primary cells has emerged as a promising approach to guide clinical decision (Letai et al, Cancer Cell 2022). This is particularly relevant for heme malignancies, given the ease of access to patient-derived cells from peripheral blood and/or bone marrow. We conducted a blinded study using OncoPrecision's proprietary Patient Micro-Avatars (PMAs) to predict clinical responses to Ven/HMA treatment. Methods De-identified diagnostic bone marrow aspirate samples from the leukemia program biobank at Weill Cornell Medicine that received Ven/HMA were provided to OncoPrecision for ex-vivo testing without information about the specific HMA used for treatment nor clinical response. PMAs are based on a triple co-culture platform that, by promoting the long term ex-vivo survival of Patient-Derived Cells (PDCs), allows testing the performance of different treatments. PDCs are co-cultured with two types of heterologous engineered cells: i) bone marrow stroma cells that act as a feed layer and as a control of non-specific toxicity (Tox-Control) and ii) a known outcome AML cell line that serves as an internal control of the system robustness and reproducibility in the response to drugs (System-Control). PMAs also leverage a multi-tagging approach coupled to high-throughput flow cytometry to track the differential response to treatments in subpopulations of pathological and healthy cells from each patient. This technology previously showed robust predictive power in observational studies in AML (García et al, Blood - 2022 ; Andino et al, Blood - 2023). Results PMAs were used to screen 39 AML patients who were treated with either Aza+Ven or Deci+Ven and from whom bone marrow or peripheral blood samples were stored prior to the start of treatment. OncoPrecision blindly tested these samples ex-vivo against a full dose-response curve of Aza+Ven and used their machine learning algorithm to predict clinical outcomes. When the predictions were matched against the clinical outcomes of the 27 patients of the cohort that received Aza+Ven at the clinic, the predictive accuracy (85%) was similar to the one previously reported in OncoPrecision's observational study (80%). Remarkably, however, when Aza+Ven predictions were matched to the clinical outcomes of the 12 patients in the cohort who received Deci+Ven, the predictive accuracy dropped to 50%. Interestingly, such a dramatic drop in the predictive power cannot be explained by a decrease in the negative predictive value (which remained >80%), but rather to a subpar positive predictive value (<30%), thus indicating that Aza may clinically outperform Deci in several of these patients. It is clear that ex-vivo testing using PMAs may have the ability to discern treatment sensitivity between Aza vs. Deci, and that there is potential to utilize this technology to tailor Ven/HMA therapy for individual patients. This is valuable not only to maximize potential disease response in a patient, but to also better prioritize patient quality of life during treatment, considering that Aza more commonly causes adverse effects like nausea and constipation than Deci. Conclusions Ex-vivo testing using PMAs has the potential to identify patients likely to respond to Ven/HMA and to discern treatment sensitivity between Aza vs. Deci. Our findings suggest that Aza and Deci may not always have equivalent efficacy for patients. This variation between Aza and Deci merits further exploration, especially since the therapeutic landscape for AML is likely to further expand with additional HMAs.
Read full abstract