Metabolism studies are one of the important steps in pharmaceutical research. LC-MS combined with metabolomics data-processing approaches have been developed for rapid screening of drug metabolites. Mass defect filter (MDF) is one of the LC/MS-based metabolomics data processing approaches and has been applied to screen drug metabolites. Although MDF can remove most interference ions from an incubation sample, the true positive rate of the retaining ions is relatively low (approximately 10%). To improve the efficacy of MDF, we developed a two-stage data-processing approach by combining MDF and stable isotope tracing (SIT) for metabolite identification. Pioglitazone (PIO), which is an antidiabetic drug used to treat type 2 diabetes mellitus, was taken as an example drug. Our results demonstrated that this new approach could substantially increase the validated rate from about 10% to 74%. Most of these validated metabolite signals (13/14) could be verified as PIO structure-related metabolites. In addition, we applied this approach to identify uncommon metabolite signals (a mass change beyond the window of 50 Da around its parent drug, MDF1). SIT could remove most interference ions (approximately 98%) identified by MDF1, and four out of five validated metabolite signals could be verified as PIO structure-related metabolites. Interestingly, a lot of the verified metabolites (10/17) were novel PIO metabolites. Among these novel metabolites, nine were thiazolidinedione ring-opening signals that might be related to the toxicity of PIO. Our developed approach could significantly improve the efficacy in drug metabolite identification compared with that of MDF.