In 1998, bluetongue (BT) was introduced in northern Africa and then extended to northern latitudes including the French island of Corsica. Following the outbreaks in Corsica in 2000 and 2001, cross-sectional studies and surveillances have been set up in Corsica and also in the southern part of mainland France, a disease-free area but considered at high risk because of its proximity. The surveillance was based on regular blood sampling of susceptible species and antibody detection by a commercial competitive ELISA kit (cELISA). The performance of this cELISA was evaluated on both field results obtained during the 2001 surveillance campaigns and experimental results. ROC analyses were carried out using RT-PCR results as gold standard for determining the infection status of animals. From all these sets of data, cut-off values optimising the diagnostic accuracy of the test were computed. Their values ranged around the manufacturer's 50% threshold from 41% to 63%. The area under the ROC curve obtained from field data was 0.843 (95% CI: 0.762-0.923). In all our results, it appeared also that the specificity of the cELISA test was always perfect if the cut-off was at least at 80%. This cELISA test does not seem sufficient to diagnose BT disease in animals with BT-like symptoms. However, complementary data are needed to better estimate sensitivity and specificity values of this BT test for its use either as a diagnostic tool in infected areas or as a screening test in BT-free areas. The use and validity of RT-PCR results as gold standard are discussed. As the lack of suitable data strongly limited the applicable analyses, a discussion based on the OIE recommendations about test evaluation is initiated. © 2006 Elsevier B.V. All rights reserved.