Two enzyme-linked immunosorbent assay (ELISA) systems, one using natural excretory–secretory (ES) antigens and the other a synthetic glycan antigen (3,6-dideoxy-d-arabinohexose [tyvelose, TY]), were evaluated for the serological diagnosis of trichinellosis in swine. Sensitivity was estimated using samples (n = 113) collected 3–21 wk PI from 15 experimentally infected pigs, and specificity was estimated using samples (n = 397) from a population of Trichinella spp.–free pigs. Results were analyzed using 2 cutoff values recommended in international guidelines (Office Internationale des Epizooties [OIE]) and by the optimal cutoff level as determined by receiver–operator characteristic (ROC) analysis. The ROC-optimized TY-ELISA consistently performed better than all other combinations. None of the combinations of test and cut-off detected infected pigs sooner than 35 days; however, the ROC-optimized TY-ELISA identified 8 of 15 pigs earlier than the ES-ELISA and detected 2 pigs missed by all other tests. At 49 days PI the sensitivity and specificity of the ROC-optimized TY-ELISA were 94.3 and 96.7%, respectively, as compared with the ROC-optimized ES-ELISA at 84.9 and 96.0%, respectively. The ROC-optimized TY-ELISA was 100% specific at OIE-recommended cut-offs. This study indicates that the TY-ELISA is as good or better than the ES-ELISA for the detection of trichinellosis in swine.