Statistical Bliss: A Novel Framework For Statistical Assessment Of Drug Synergy.

Abstract Combination therapy is a common feature of cancer treatment, aiming to maximize efficacy while minimizing toxicity and opportunities for acquired resistance. It is critical to assess which combinations achieve best effects. This requires rigorous, statistically sound methods of characterizing the relationships between drugs as synergistic, independent, or antagonistic. Current frameworks, including Bliss independence and Loewe additivity as implemented by the Chou-Talalay method, are not statistical and assess effects at discrete points rather than across the entirety of a dose-response curve, limiting their application. We aimed to create a comprehensive statistical framework applicable to dose-response curves of all forms and any number of drugs in combination. We developed a generalized definition of independent action to construct theorized dose-response curves of drugs in combination informed by empirically fitted dose-response curves for individual drugs. We derive uncertainty estimates for theorized curves, which are compared to empirical measurements of fixed- or variable-ratio drug combinations at all measured concentrations. Separately accounting for variability in technical and experimental replicates allows robust statistics at each experimental condition as well as globally across the range of conditions. As an example of its utility, we applied this model to a system of known synthetic lethality. We exposed BRCA-mutant (HCC1937) and wild-type (MDA-MB-231) triple-negative breast cancer cell lines to PARP inhibitor (PARPi) olaparib and RAD51 inhibitors (RAD51i) IBR2 or B02 singly and in variable-ratio combinations and evaluated cell viability at 96 hours post-treatment by MTT assay. We hypothesized that combination of RAD51i, producing a defect in homologous recombination, with olaparib in BRCA-WT cells would phenocopy the synthetic lethal effect of PARPi in BRCA-Mut cells and demonstrate synergy, whereas there would be no synergy in BRCA-Mut cells. HCC1937 cells demonstrated antagonism between olaparib and B02 (p < 0.001) or IBR2 (p < 0.001). MDA-MB-231 cells demonstrated significant synergy of olaparib and B02 (p < 0.001) or IBR2 (p < 0.001), consistent with our hypothesis. However, the interaction of olaparib and B02 was dose-dependent with synergy overall but antagonism at high concentrations. This shows that (1) measurements at single concentrations may be misleading and (2) our new method may guide optimal drug concentration selection. We also assessed a “sham” combination of B02 with itself in MDA-MB-231 cells and predictably detected neither synergy nor antagonism (p = 0.39). Not only do these data provide evidence for expanded use of PARPi alongside RAD51i in BRCA-WT cancers, they validate this statistical approach that offers a robust calculation of synergy or antagonism with broad applicability to any study of combination therapies, especially for high-throughput screening. Citation Format: Richard E. Grewelle, Kalin L. Wilson, Dana M. Brantley-Sieders. Statistical Bliss: A novel framework for statistical assessment of drug synergy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 262.