Introduction. - Multidetector computed tomography coronarography is a promising tool, offering a non-invasive anatomic evaluation of coronary arteries. The great majority Of Studies conducted upon it are single-center studies, and have reported results based upon a highly selected patient population. Our aim was to determine its diagnostic accuracy among an unselected population in multicenter studies. Methods. - Fifty-two patients were included in a non-randomised, retrospective study. Patients underwent multidetector computed tomography coronarography (16, 40 and 64 slices), in ten different centers (community hospitals or private centres), for clinical suspicion of coronary stenoses. The diagnostic accuracy for detecting significant coronary stenoses (>=50\%) was determined in comparison with conventional coronarography. Results. - The sensitivity, specificity, positive and negative predictive values and diagnostic accuracy were 91.4\%, 17.7\%, 69.6\%, 50.0\%, and 67.3\%, respectively, in a patient-based analysis, and 55.3\%, 85.6\%, 30.3\%, 94.4\% and 82.5\% in a segment-based analysis. Conclusion.- The results of this study indicate that routine implementation of multidetector computed tomography coronarography is limited by a high false-positive rate, when performed among an unselected population of patients with a high pretest probability of having coronary stenoses, in centres with variable expertises. Its place within the range of diagnostic tools has yet to be determined by large multicenter studies, before being subject to precise recommendations framing its routine clinical application. (C) 2009 Elsevier Masson SAS. All rights reserved.