Diagnostic Accuracy of Transvaginal Sonography in Infertile Patients with Endometrial Polyps

Document Type : Letter to The Editor


Department of Epidemiology and Reproductive Health, Royan Institute for Reproductive Biomedicine, ACECR, Tehran, Iran


We are interested the article by Niknejadi et al. (1) highly scientific and provocative however there seem to be a lack of clarity on some of the methodological issues; 1. In the materials and methods section, it was mentioned that: “in cases that had normal endometrial findings on transvaginal sonography (TVS), a hysteroscopy was not scheduled due to ethical considerations” whereas in the results, a total of 466 cases with negative TVS were presented. It is not mentioned how these patients were found. If they have other indications for hysteroscopy, it may add bias to your research which may affect the results (verification bias). The bias would be present since both the TVS positive and negative patients would come from different patient populations that have different indications as a result of their various medical conditions. For example, most likely the TVS negative patients in this study came from patients who presented with in vitro fertilization (IVF) failure, intrauterine insemination (IUI) failure and abnormalities seen in clinical exams, to name a few. The TVS positive cases would beare the people diagnosed during routine TVS for infertility evaluation. Although your research is retrospective, it would be better to consider such differences between subjects (2-6). 2. Considering that TVS negative patients do not routinely undergoto hysteroscopies; therefore one can conclude that TVS is a highly sensitive test, thus positive patients will not be missed. Laparoscopy is a highly specific test needed solely for confirming the diagnosed problem and can 88 Royan Institute International Journal of Fertility and Sterility Vol 4, No 2, Jul-Sep 2010 filter out healthy patients (2). It seems that in your report, the sensitivity of TVS is not very high. On the other hand, if we assume that there are some degrees of verification bias in your study, the real sensitivity even will be lower than calculated