Preoperative imaging of deep endometriosis: pitfalls of a diagnostic test before surgery
Deep endometriosis, sensitivity, Bayesian, specificity, test accuracy
Published online: Jan 08 2021
The usefulness of a test is determined by the clinical interpretation of its sensitivity and specificity. The pitfalls of a test with a surgical endpoint are described in this article, taking the diagnosis of deep endometriosis by imaging as an example, without discussing the management of deep endometriosis. Laparoscopy is not a 100% accurate “gold standard”. Since it is not performed in women without symptoms, results are valid only for the group of women as specified in the indication for surgery. The confidence limits of accuracy estimations widen when accuracy is lower and when observations are less. Since positive and negative predictive values are inaccurate when prevalence of the disease is low, prevalence figures in the group of women investigated should be available. The accuracy of imaging should be stratified by clinically important aspects such as localisation and size of the lesion. The use of other variables as soft markers during ultrasonographic examination should be specified. It should be clear whether the accuracy of the test reflects symptoms and clinical examination and imaging combined, or whether the accuracy of the added value of imaging which requires Bayesian analysis. When imaging is used as an indication for surgery, circular reasoning should be avoided and the number of symptomatic women not undergoing surgery because of negative imaging should be reported. In conclusion, imaging reports should permit the clinician to judge the validity of the accuracy estimations of a diagnostic test, especially when used as an indication for surgery and when surgery is the gold standard to diagnose a disease.