Three-dimensional ultrasonography in the diagnosis of deep endometriosis

Abstract

In the use of 'tenderness-guided' transvaginal ultrasound, is the diagnostic accuracy of three-dimensional (3D) ultrasonography better than two-dimensional (2D) ultrasonography in the identification of deep endometriosis? Three-dimensional ultrasonography has a significantly higher diagnostic accuracy in the diagnosis of posterior locations of deep endometriosis without intestinal involvement, such as the uterosacral ligaments, vaginal and rectovaginal endometriosis. The only previous study of the diagnosis of posterior compartment endometriosis reported an poor sensitivity of 3D ultrasonography for uterosacral and sigmoid colon involvement. This diagnostic test study included 202 patients scheduled for surgery because of clinical suspicion of deep pelvic endometriosis and was carried out between January 2009 and September 2012. Modified transvaginal ultrasonography was performed on all of the women by a single examiner. Two locations of deep endometriosis were considered: intestinal involvement and other posterior lesions (including vaginal location, rectovaginal septum and uterosacral ligaments). Once the 2D ultrasonography had been performed, the 3D acquisition was performed and the obtained volume was stored. To avoid the risk of recall bias, the same operator evaluated the 3D volumes 6 months after the last examination using virtual navigation to provide a presumptive diagnosis of the presence and localization of deep endometriosis. In addition, to evaluate the reproducibility of 3D, two operators with different levels of expertise performed a retrospective review of 3D volumes from a random sample of 35 patients, twice, 1 week apart to also assess intraobserver agreement. The diagnostic performance of both tests was expressed as area under the receiver-operating characteristics curve (AUC), sensitivity, specificity, positive and negative predictive values, positive (LR+) and negative (LR-) likelihood ratios, with their respective 95% confidence interval (CI). Reproducibility was evaluated using kappa statistics. Surgery revealed deep endometriosis in 129 patients. The AUCs for endometriosis of intestinal location were similar for both ultrasound techniques. The AUCs for endometriosis of other posterior locations were significantly different (0.891, 95% CI 0.839-0.943 for 3D versus 0.789, 95% CI 0.720-0.858 for 2D; P = 0.0193). For the intestinal involvement, the specificity, sensitivity, positive and negative predictive value, and LR+ and LR- were 93% (89-95%), 95% (88-98%), 89% (83-92%), 97% (93-99%), 13, and 0.06, respectively, for 2D ultrasound and 97% (93-99%), 91% (84-94%), 95% (88-98%), 95% (91-96%), 25, and 0.09, respectively, for 3D ultrasound. For other posterior locations, the specificity, sensitivity, positive and negative predictive value, and LR+ and LR- were 88% (82-93%), 71% (64-77%), 83% (75-90%), 79% (74-83%), 6.10, 0.32, respectively, for 2D ultrasound and 94% (89-97%), 87% (81-91%), 92% (86-96%), 90% (85-93%), 14.0, 0.14, respectively, for 3D ultrasound. Intraobserver agreement was substantial for both examiners (kappa 0.8754, for operator A and 0.7087, for operator B, respectively). Interobserver agreement was also substantial. The disadvantages of 3D ultrasound to be considered are the necessity of newer ultrasonographic equipment and that fewer sonographers completely know the 3D technique. There are also some limitations within this study. First, an expert examiner performed the real-time ultrasound and 3D volume acquisitions. Second, the same operator also performed the 3D evaluations but at least 6 months after the last acquisition to avoid a possible recall bias. The diagnostic performance obtained in the present study is superior to the accuracy reported in other studies of 3D ultrasonography, but not superior to all other published articles of 2D ultrasonography. The reported high diagnostic accuracy of 3D ultrasound could be widely generalizable because good reproducibility was demonstrated even with an operator with less expertise. This study was supported in part by the Regione Autonoma della Sardegna (project code CPR-24750).