We explored the type I error rate (false positive rate) associated with exposure-response (ER) analyses for two compounds in a fixed-dose combination product through simulations. In the simulations, at least one compound was assumed to be inactive, whereas the active compound followed E(max) model at different concentration ranges. The simulated data were independently evaluated by pre-specified univariate or multivariate linear, log-linear models, and mixed linear log-linear models. The type I error rate was evaluated by comparing the total number of falsely identified significant slope estimates with the total number of models with successful convergence. We demonstrated that ER analyses results based on data from fixed-dose combination products at various dose levels should be interpreted with caution. A univariate analysis, even though is appropriate to guide dose selection, is inadequate to identify the active compound. Multivariate analyses can be applied to determine the active compound only when the underlying ER relationship for each compound (especially for the active compound) has been adequately defined or approximated. The false positive rate in determining a significant ER relationship is elevated, when the underlying ER relationship (especially for the active compound) is erroneously or inadequately defined. Without the assurance of the correct structural models, the identified significant ER relationship does not necessarily indicate that the compound associated with the significant slope estimate is pharmacologically active.