In the framework of chemical risk assessment, dose-response assessment first quantifies the relationship between dose and critical health effects, and it then extrapolates the biologically/ statistically significant dose (i.e., point of departure [POD]) to a reference dose for regulatory use. Typically, the point of departure for chemical carcinogens is the benchmark dose (BMD) or its statistical lower bound (i.e., BMDL). From a conservative perspective, the US EPA uses linear extrapolation as a default approach to derive a cancer slope factor, which may be employed to estimate cancer risk at a given dose. If sufficient evidence of the mode of action (MOA) can support a conclusion that the dose-response relationship is nonlinear in the low-dose region (i.e., to confirm that the agent is nonmutagenic), an oral reference dose (RfD) or an inhalation reference concentration (RfC) can be calculated instead. However, calculated RfD/RfC lacks the utility to quantify the risk of adverse effects in support of a probabilistic risk assessment. Therefore, this dissertation aims to develop a new framework of dose-response assessment for chemical carcinogens with sufficient data to characterize their MOAs. There are three significant steps in the framework: (1) key quantifiable event identification and dose-response data extraction; (2) dose-response modeling and critical dose derivation for each endpoint; and (3) pathway dose-response relationship evaluation and POD calculation. Three chemical carcinogens were analyzed using the proposed framework, including TCDD and PCB 126, with the MOA of AhR, and DEHP with PPARα. The results for the examples demonstrate excellent stability and flexibility of the proposed method. By quantitatively integrating human variability, the pathway POD can be used to derive a risk-specific reference dose. Rather than using a specific health effect to derive a POD in the current dose-response assessment framework, the new approach not only considers the mechanism of tumor formation but also probabilistically quantifies the dose level that may induce a defined critical key event. Linear and nonlinear methods of low-dose extrapolation can be unified before inducing to a human reference dose. The new framework of dose-response assessment will provide for more meaningful, scientifically based cancer risk assessment.