%0 Journal Article
%T Assessment of Statistical Methods Used in the Iranian Original Articles on Mammography
%A "A. Azizian
%A H. Bakhshandeh
%A N. Riyahi Alam
%A M. Guity "
%J Iranian Journal of Radiology
%D 2005
%I Tehran University of Medical Sciences
%X Introduction & Background: One of the most impor-tant objectives in studies about diagnostic tests is to determine their validity. Two basic criteria of validity are ¡°sensitivity¡± and ¡°specificity¡±, but they depend on cut points used in determining ¡°positive¡± and ¡°nega-tive¡± test results. They also used to measure the accu-racy of a test. However, accuracy depends on the prevalence of disease in population. Receiver operat-ing characteristic (ROC) curve is a plot of sensitivity versus its false positive rate for all possible cut points. Because sensitivity and specificity are independent of disease prevalence, the ROC curve will be independ-ent too. The area under curve (AUC) is a good sum-mary measure of a test¡¯s overall accuracy, because it does not depend on the prevalence of disease or the cut points used to form the curve. ROC analysis is the most accurate method used in mammographic stud-ies. Our experience showed that this method is not commonly a choice in our local journals. This article evaluates the proper use of this statistical method in our local published articles related to mammography. Patients & Methods: We searched Mammography keyword in ¡°IranMedex¡± database which contains eighty one of Iranian journals. We found 26 related articles, three of them were case reports and ex-cluded. Remained articles were divided in two groups: A- articles which required ROC analysis; B- articles which did not require ROC analysis. We placed 10 articles in group A and evaluated for their methods. Results: In group A, 7 articles (70%) had not used ROC analysis and applied only ¡°sensitivity¡± and ¡°specificity¡± or ¡°accuracy¡± for one cut point in their method. Conclusions: As the results show, our local published articles related to mammography are not frequently using ROC analysis in their methods. As it previously mentioned, ¡°sensitivity¡± and ¡°specificity¡± are not the adequate measures in some studies and must apply a proper method for those studies.
%U http://journals.tums.ac.ir/PdfMed.aspx?pdf_med=/upload_files/pdf/2182.pdf&manuscript_id=2182