Predictive Value of Ultrasound Imaging for Diagnosis and Surgery of Deep Endometriosis: A Systematic Review

Abstract

Objective: To calculate the predictive value and thus the clinical usefulness of transvaginal ultrasound (US) imaging for the management of deep endometriosis, knowing that the positive predictive value (PPV) varies with the prevalence and probably with the volume and location of the disease.

Data sources: After registration on PROSPERO (CRD42022366323), PubMed was searched for all reports describing the diagnostic accuracy of US imaging for deep endometriosis published between January 1, 2000, and October 20, 2022.

Methods of study selection: The 536 articles on "endometriosis AND US And diagnosis" were hand searched, and 30 reports describing sensitivity and specificity of deep endometriosis were found. Besides sensitivity and specificity, the prevalence, localization, and size of deep endometriosis lesions were collected.

Tabulation, integration, and results: Prevalences of deep endometriosis were reported only twice as 12% and 32% by ultrasonographers. In women undergoing surgery, prevalences vary between 40% and 100% because of the variable inclusion criteria. Specificity is higher than sensitivity for all locations: rectovaginal (97% [86-100] vs 74% [31-95], p = .0002), rectosigmoid (97% [63-100] vs 88% [37-97], p = .0082), vesicouterine (100% [97-100] vs 63% [22-100], p = .0021), and uterosacrals (91% [77-99] vs 68% [18-83], p = .0005). Notwithstanding improved equipment, accuracy did not vary over the last 20 years. Sensitivities or specificities have not been stratified by the size of the lesion, and thus, the lower detection limits are not known. In the absence of blinding, the usefulness for surgery could not be established.

Conclusion: The reported sensitivities and specificities of transvaginal US are not only those of imaging but include symptoms and clinical examinations. In referral centers, the reported PPVs are high (94%-100%) given that prevalences are >10% and specificities are >95%. However, the extrapolation of the clinical use before surgical interventions should be considered with care, given that PPVs for smaller lesions and the lower detection limit are unknown and surgeons were not blinded to US results.

Keywords: Accuracy; Deep endometriosis; Predictive value; Surgery; Ultrasound.