One hundred commercial wheat grain samples were collected during the 2015 sea- son across 78 municipalities in the states of Paraná (PR), Rio Grande do Sul (RS), and São Paulo (SP), Brazil. Separate subsamples were analyzed for the concentration of deoxynivalenol (DON), zearalenona (ZEA) and ochratoxin A (OTA) mycotoxins using two methods: UHPLC-MS/MS (refer- ence method) and a commercial enzyme-linked immunosorbent assay (ELISA) (AgraQuant®). The OTA mycotoxin was not found in the samples by both methods. DON and ZEA were detected in 55 % and 39 % of the samples by the reference method, with overall mean levels of 795.2 μg kg−1 and 79.78 μg kg−1, respectively. There was a signi cant and positive correlation (Spearman rank) between DON and ZEA estimates by the reference method (r = 0.77, p < 0.001). The DON levels estimated by the immunoassay agreed poorly with the reference, being largely overesti- mated. Based on a cut-off level of 1000 μg kg−1, the immunoassay correctly classi ed 57 sam- ples as true negatives and 15 as true positives. Only 28 were classi ed as false positives. For ZEA, the levels estimated by the two methods were in better agreement than for DON. Using the cut-off level of 200 μg kg−1, 96 % of the samples were classi ed correctly as true positives and only one sample was classi ed as false positive. The levels for both mycotoxins were mostly ac- ceptable for human consumption. Further studies should focus on multi-toxin methods compared with immunoassays to understand the reasons of overestimation and the role of immunoassays as a cost-effective solution for fast screening of mycotoxins in the food chain.