Background: Mass spectrometry metabolomics-based data-processing approaches have been developed for drug metabolite profiling. However, existing approaches cannot be used to comprehensively identify drug metabolites with high efficacy. Methods: Herein, we propose a two-stage data-processing approach for effective and comprehensive drug metabolite identification. The approach combines dose-response experiments with stable isotope tracing (SIT). Rosiglitazone (ROS), commonly used to treat type 2 diabetes, was employed as a model drug. Results: In the first stage of data processing, 1,071 features exhibited a dose-response relationship among 22,597 features investigated. In the second stage, these 1,071 features were screened for isotope pairs, and 200 features with isotope pairs were identified. In time-course experiments, a large proportion of the identified features (69.5%: 137 out of 200 features) were confirmed to be possible ROS metabolites. We compared the validated features identified using our approach with those identified using a previously reported approach [the mass defect filter (MDF) combined with SIT] and discovered that most of the validated features (37 out of 42) identified using the MDF-SIT combination were also successfully identified using our approach. Of the 143 validated features identified by both approaches, 74 had a proposed structure of an ROS-structure-related metabolite; the other 34 features that contained a specific fragment of ROS metabolites were considered possible ROS metabolites. Interestingly, numerous ROS-structure-related metabolites were identified in this study, most of which were novel. Conclusion: The results reveal that the proposed approach can effectively and comprehensively identify ROS metabolites.