Rapid nucleic acid-based methods to detect human pathogens in foods are dependent on the reliability of the DNA or RNA extraction method used. Skim milk, non-fat dry milk, Cheddar and Brie cheese, and reconstituted whey powder were seeded with serially diluted (10(0)-10(7) cfu 10 ml(-1)) Escherichia coli O157:H7 and subjected to DNA extraction (i) directly from the food product using a solvent-based procedure and (ii) using a guanidinium isothiocyanate (GITC) procedure after previous bacterial concentration. Both the efficiency of DNA extraction and the overall PCR detection limits were evaluated. In almost all instances, the total DNA yield using the solvent method was greater than that obtained for the concentration method. However, the purity of the DNA obtained after bacterial concentration was significantly better than that obtained after organic extraction alone, PCR detection limits after each DNA recovery method varied with the specific food, ranging from 10(1) to 10(4) cfu ml(-1) for all products except whey powder. DNA yields and subsequent PCR detection limits for reconstituted whey powder were extremely poor, and neither procedural changes nor the addition of PCR enhancement agents were able to improve recovery and/or detection. It is concluded that the efficiency of DNA extraction is an extremely important and frequently overlooked variable impacting the overall detection limits of PCR-based detection strategies.