Abstract BACKGROUND The benefits of combination therapy are often attributed to synergy, that is, drug interactions resulting in an anti-cancer effect that is more than the sum of its parts. Accordingly, the rationale for designing new drug combinations is often based on synergy measured in preclinical models. However, preclinical metrics of drug interaction are not applicable to clinical trial data, and there has been no established quantitative method to assess synergy versus additivity in clinical settings. We recently showed that because of extensive patient-to-patient heterogeneity in single drug responsiveness, increasing the chance of a good response to at least one drug was a quantitatively sufficient explanation for the clinical efficacy of many approved combination therapies. Some combinations surpass this ‘highest single-agent’ model, which could be due to either drug additivity or synergy. Here we propose and test a model of drug additivity for Progression-Free Survival (PFS) data from clinical trials, to identify if any approved combinations are clinically synergistic, as compared to additive. METHODS We used PFS from trials as the clinical measure of drug efficacy. We defined ‘clinical drug additivity’ as the sum of PFS times observed from each drug in a combination. Inter-patient heterogeneity in drug responses was simulated by sampling from the joint distribution of drugs’ PFS distributions. For each combination, the clinically observed PFS distribution was compared to the additivity model, or the highest single-agent model (Palmer & Sorger, 2017). Synergy is exhibited if an observed PFS distribution is significantly superior to PFS expected from additivity (by Cox Proportional Hazards). To search for drug synergy in clinical data, we analyzed approved combination therapies where synergy was most likely to explain efficacy, which are combinations where part of the regimen is not approved as monotherapy in the same disease. We analyzed 11 approved combination therapies for advanced cancers in the breast, ovary, pancreas, colon, cervix, and lymphatic system. RESULTS None of the 11 approved combinations analyzed were significantly superior to the model of clinical additivity. For five combinations, the additivity model made the most accurate predictions of clinical efficacy (mean R2=0.97 for additivity, versus R2=0.83 for highest single-agent), and the other six combinations were most accurately described by the highest single-agent model (mean R2=0.97 for highest single-agent, versus R2=0.91 for additivity). CONCLUSIONS Approved combination therapies are rarely ‘more than the sum of their parts’ in quantitative terms. A straightforward definition of clinical drug additivity accurately matched trial results for combination therapies where synergy was expected. This suggests that single-agent efficacy by each drug is usually required for the clinical success of combination therapy. Citation Format: Haeun Hwangbo, Adam C. Palmer. Defining and evaluating drug additivity in clinical trials of combination cancer therapy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2739.
Read full abstract