MRI versus laparoscopy to diagnose the main causes of chronic pelvic pain in women: a test-accuracy study and economic evaluation

Background: Chronic pelvic pain (CPP) symptoms in women are variable and non-specific; establishing a differential diagnosis can be hard. A diagnostic laparoscopy is often performed, although a prior magnetic resonance imaging (MRI) scan may beneficial. Objectives: To estimate the accuracy and added value of MRI in making diagnoses of (1) idiopathic CPP and (2) the main gynaecological causes of CPP. To quantify the impact MRI can have on decision-making with respect to triaging for therapeutic laparoscopy and to conduct an economic evaluation. Design: Comparative test-accuracy study with cost-effectiveness modelling. Setting: Twenty-six UK-based hospitals. Participants: A total of 291 women with CPP. Methods: Pre-index information concerning the patient’s medical history, previous pelvic examinations and ultrasound scans was collected. Women reported symptoms and quality of life at baseline and 6 months. MRI scans and diagnostic laparoscopy (undertaken and interpreted blind to each other) were the index tests. For each potential cause of CPP, gynaecologists indicated their level of certainty that the condition was causing the pelvic pain. The analysis considered both diagnostic laparoscopy as a reference standard for observing structural gynaecological causes and consensus from a two-stage expert independent panel for ascertaining the cause of CPP. The stage 1 consensus was based on pre-index, laparoscopy and follow-up data; for stage 2, the MRI scan report was also provided. The primary analysis involved calculations of sensitivity and specificity for the presence or absence of each structural gynaecological cause of pain. A decision-analytic model was developed, with a 6-month time horizon. Two strategies, laparoscopy or MRI, were considered and populated with study data. Results: Using reference standards of laparoscopic and expert panel diagnoses, MRI scans had high specificity but poor sensitivity for observing deep-infiltrating endometriosis, endometrioma, adhesions and ovarian cysts. MRI scans correctly identified 56% [95% confidence interval (CI) 48% to 64%] of women judged to have idiopathic CPP, but missed 46% (95% CI 37% to 55%) of those considered to have a gynaecological structural cause of CPP. MRI added significant value, over and above the pre-index information, in identifying deep-infiltrating endometriosis (p = 0.006) and endometrioma (p = 0.02) as the cause of pain, but not for other gynaecological structural causes or for identifying idiopathic CPP (p = 0.08). Laparoscopy was significantly more accurate than MRI in diagnosing idiopathic CPP (p < 0.0001), superficial peritoneal endometriosis (p < 0.0001), deep-infiltrating endometriosis (p < 0.0001) and endometrioma of the ovary (p = 0.02) as the cause of pelvic pain. The accuracy of laparoscopy appeared to be able to rule in these diagnoses. Using MRI to identify women who require therapeutic laparoscopy would lead to 369 women in a cohort of 1000 receiving laparoscopy unnecessarily, and 136 women who required laparoscopy not receiving it. The economic analysis highlighted the importance of the time horizon, the prevalence of CPP and the cut-off values to inform the sensitivity and specificity of MRI and laparoscopy on the model results. MRI was not found to be a cost-effective diagnostic approach in any scenario. Conclusions: MRI was dominated by laparoscopy in differential diagnosis of women presenting to gynaecology clinics with CPP. It did not add value to information already gained from history, examination and ultrasound about idiopathic CPP and various gynaecological conditions. Trial registration: Current Controlled Trials ISRCTN13028601. Funding: This project was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 22, No. 40. See the NIHR Journals Library website for further project information