Abstract In theory, the development of combination therapies in Oncology holds the promise of improved short-term response and better disease control. In practice, the added toxicity burden of a second antineoplastic agent can often limit the potential of combination therapies. Many antineoplastic agents, targeted or not, result in severe toxicities in clinical practice, and the paradigm for Oncology drug development still involves dosing to the Maximum Tolerated Dose (MTD). Thus, proactive identification and management of combination toxicity should help combination therapies achieve their full clinical potential. In this work, we devise a mathematical framework for assessing the toxicity potential for a pair of drugs with a single overlapping toxicity, using neutropenia- a very common adverse event for antineoplastics- as a motivating example. First, we introduce a mathematical framework for modeling combination toxicities, based on a combination index that is a measure of the interaction between two agents (analogous to antitumor activity, the overlapping toxicities for a pair of agents can be synergistic, additive or antagonistic). Next, we demonstrate the application of isobolograms for toxicity, which are a set of lines connecting all equally toxic combinations of the two drugs. When the MTD of a pair of combination therapies is determined by an overlapping toxicity, it can thus be expressed in terms of a minimal toxicity model that uses three components- the concentration/toxicity relationship for each individual therapy and the combination index of the toxicities. We show that the magnitude of the combination index remains unchanged whether the toxicity readout is continuous (such as Absolute Neutrophil Count) or categorical (Grade 1,2, 3, or 4 neutropenia). This result enables us to use the combination index calculated from a lower-grade toxicity (e.g. Grade 1 or Grade 2 neutropenia) to anticipate the MTD. We validate this approach using simulated datasets generated from a previously published model of neutropenia. This validation is further extended using experimental datasets based on preclinical toxicities for combinations of targeted agents. Finally, we demonstrate the extension of this approach to the prediction of the Dose Limiting Toxicity for a combination. This method relies on developing several toxicity models in parallel for each potential dose-limiting toxicity, and calculating the combination index from early-stage toxicities for each of them. The combination index for each toxicity is then used to predict the MTD that would result from it, and the toxicity that results in the lowest MTD is then the putative Dose Limiting Toxicity. Taken together, the approaches described here can be used to derive critical information directly from clinical data, and enable the design of rational escalation schemes in Phase I trials for combinations that are at once faster and safer. Citation Format: Ekta Kadakia, Christopher J. Zopf, Mayankbhai Patel, Dean Bottino, Greg Hather, Wen Chyi Shyu, Arijit Chakravarty. Anticipating the maximum tolerated dose for combinations based on early toxicity signals. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 3746. doi:10.1158/1538-7445.AM2014-3746