A fluorescence polarization immunoassay previously described for deoxynivalenol (DON) screening in wheat was optimized for the rapid quantification of DON in durum wheat kernels, semolina, and pasta. A background signal was observed in both spiked and naturally contaminated samples, strictly depending on the testing matrix. After subtracting the background DON level for durum wheat (0.27 microg of DON per g), semolina (0.08 microg of DON per g), and pasta (0.04 microg of DON per g), an accurate quantification of DON was possible at levels greater than 0.10 microg/g for all matrices. Average recoveries from spiked samples (0.25 to 1.75 microg/g) were 98, 102, and 101% for wheat, semolina, and pasta, respectively. Comparative analyses of 35 naturally contaminated durum wheat samples, 22 semolina samples, and 26 pasta samples performed by both the fluorescence polarization method and high-pressure liquid chromatography/immunoaffinity cleanup showed a good correlation (r > 0.995). The fluorescence polarization method showed better accuracy and precision with respect to the high-pressure liquid chromatography method and is suitable for the rapid and quantitative determination of DON in durum wheat-based products at levels foreseen by existing or coming international regulations.