Shannon Reid1, Chuan Lu2, George Condous1,3. 1. Acute Gynaecology, Early Pregnancy and Advanced Endosurgery Unit, Nepean Medical School, Nepean Hospital, University of Sydney, Penrith, New South Wales. 2. Department of Computer Sciences, University of Wales, Aberystwyth, UK. 3. Norwest Private Hospital, Bella Vista, New South Wales, Australia.
Abstract
INTRODUCTION: A negative "sliding sign" during transvaginal sonography (TVS) is associated with pouch of Douglas (POD) obliteration at laparoscopy in women with suspected endometriosis. The aim of the current study was to develop and validate mathematical ultrasound models to determine whether a combination of TVS markers could improve the prediction of POD obliteration as compared with the TVS "sliding sign" alone. MATERIAL AND METHODS: Multicenter prospective observational study. In all, 189 women (100 in the training set and 89 in the test set) with suspected endometriosis underwent pre-operative TVS and laparoscopy. More than 50 historical, clinical and TVS end points were recorded for analysis. Univariate/multivariate analysis was performed to determine significant TVS variables associated with POD obliteration at laparoscopy. Two logistic regression models were developed on a training set: POD1 - posterior compartment deep infiltrating endometriosis, right ovarian fixation, negative "sliding sign" (i.e. POD obliteration); and POD2 model - unilateral ovarian fixation, unilateral endometrioma, negative "sliding sign." The two models were then applied to a test set to predict POD obliteration. POD1/POD2 performance was evaluated using receiver operating characteristic curves (probability cut-off value at 0.5) and compared with "sliding sign" alone for prediction of POD obliteration. RESULTS: Respective values for POD1 vs. POD2 vs. "sliding sign" alone for training/test sets: accuracy - 96/96% vs. 93/97% vs. 93/97%; sensitivity - 93/88% vs. 83/88% vs. 83/88%; specificity - 97/97% vs. 97/99% vs. 97/99%; positive predictive value - 93/88% vs. 93/94% vs. 93/94%; negative predictive ratio - 97/97% vs. 93/97% vs. 93/97%; positive likelihood ratio - 32.7/32.2 vs. 29.2/63.5 vs. 29.2/63.5; negative likelihood ratio - 0.07/0.12 vs. 0.17/0.12 vs. 0.17/0.12. CONCLUSIONS: Incorporation of TVS markers such as bowel endometriosis, endometrioma and ovarian fixation into mathematical models does not appear to improve the prediction of POD obliteration as compared with the TVS "sliding sign" alone.
INTRODUCTION: A negative "sliding sign" during transvaginal sonography (TVS) is associated with pouch of Douglas (POD) obliteration at laparoscopy in women with suspected endometriosis. The aim of the current study was to develop and validate mathematical ultrasound models to determine whether a combination of TVS markers could improve the prediction of POD obliteration as compared with the TVS "sliding sign" alone. MATERIAL AND METHODS: Multicenter prospective observational study. In all, 189 women (100 in the training set and 89 in the test set) with suspected endometriosis underwent pre-operative TVS and laparoscopy. More than 50 historical, clinical and TVS end points were recorded for analysis. Univariate/multivariate analysis was performed to determine significant TVS variables associated with POD obliteration at laparoscopy. Two logistic regression models were developed on a training set: POD1 - posterior compartment deep infiltrating endometriosis, right ovarian fixation, negative "sliding sign" (i.e. POD obliteration); and POD2 model - unilateral ovarian fixation, unilateral endometrioma, negative "sliding sign." The two models were then applied to a test set to predict POD obliteration. POD1/POD2 performance was evaluated using receiver operating characteristic curves (probability cut-off value at 0.5) and compared with "sliding sign" alone for prediction of POD obliteration. RESULTS: Respective values for POD1 vs. POD2 vs. "sliding sign" alone for training/test sets: accuracy - 96/96% vs. 93/97% vs. 93/97%; sensitivity - 93/88% vs. 83/88% vs. 83/88%; specificity - 97/97% vs. 97/99% vs. 97/99%; positive predictive value - 93/88% vs. 93/94% vs. 93/94%; negative predictive ratio - 97/97% vs. 93/97% vs. 93/97%; positive likelihood ratio - 32.7/32.2 vs. 29.2/63.5 vs. 29.2/63.5; negative likelihood ratio - 0.07/0.12 vs. 0.17/0.12 vs. 0.17/0.12. CONCLUSIONS: Incorporation of TVS markers such as bowel endometriosis, endometrioma and ovarian fixation into mathematical models does not appear to improve the prediction of POD obliteration as compared with the TVS "sliding sign" alone.