Untargeted NMR-based metabolomics strategies are increasingly applied for metabolite screening in a wide variety of medical conditions. However, the identification of metabolites remains a great challenge in NMR metabolomics research. Here, we used matrix-assisted diffusion-ordered spectroscopy (DOSY) to identify new biomarkers in body fluids. DOSY-based metabolomics can exploit the unique strengths of NMR and, thanks to improvements in its resolution, can be successfully used to address a significantly wider range of biological issues. In this study, a metabolite model consisting of ten common metabolites was employed to optimize the matrix-assisted DOSY experimental parameters. Polyvinylpyrrolidone (PVP), a high-molecular-weight polymer, was found to be the most appropriate matrix with which to accomplish the separation of the mixture components. 1H NMR and PVP-assisted DOSY analyses were performed to characterize the metabolic profile of serum samples from acrylamide (ACR)-treated rats and study the mechanism of ACR-induced cardiovascular toxicity. Biomarker identification by 1H NMR and PVP-assisted DOSY yielded similar results. However, multivariate statistical analysis showed that, compared with 1H NMR, PVP-assisted DOSY achieved a better discrimination of the serum samples obtained from rats treated with different doses of ACR. A total of 11 differential metabolites were identified. An enrichment analysis of the metabolic pathways showed that, in rats treated with a high dose of ACR, the glutamate and glutathione metabolic pathways and the energy metabolism pathway of the tricarboxylic acid cycle were mainly affected.