Food products and ingredients are frequently tested for the presence of undeclared allergenic food residues (including milk) using commercial enzyme-linked immunosorbent assays (ELISAs). However, little is understood about the efficacy of these kits with thermally processed foods. This study evaluated the performance of three milk ELISA kits with a model food processed by several methods. A model food (pastry dough squares) was spiked with nonfat dry milk at several concentrations. The pastry squares were processed by boiling (100 degrees C for 2 min), baking (190 degrees C for 30 min), frying (190 degrees C for 2 min), and retorting (121 degrees C for 20 min with 17 psi overpressure). Samples were analyzed with three commercial ELISA kits: Neogen Veratox Total Milk, ELISA Systems beta-lactoglobulin, and ELISA Systems casein. The detection of milk residues depended upon the type of processing applied to the food and the specific milk analyte targeted by the ELISA kit. Poor recoveries were obtained in all processed samples (2-10% of expected values) using the beta-lactoglobulin kit. Better recoveries were obtained in boiled samples (44 and 59%, respectively) using the total milk and casein kits. However, these kits performed poorly with baked (9 and 21%) and fried (7 and 18%) samples. Moderate recoveries were observed in retorted samples (23 and 28%). The decreased detection in processed samples is likely due to protein modifications, including aggregation and Maillard reactions, which affect the solubility and immunoreactivity of the antigens detected by the ELISA methods. The observed decreases in ELISA detection of milk are dramatic enough to affect risk-assessment decisions. However, a lower detection of milk residues does not necessarily indicate decreased allergenicity. These ELISA kits are not acceptable for all applications, and users should understand the strengths and limitations of each method.
