ENDORECT: a preoperative score to accurately predict rectosigmoid involvement in patients with endometriosis

Could we construct and validate a preoperative score to predict rectosigmoid involvement in endometriosis (RE)?

We developed a simple preoperative score (ENDORECT) to predict RE.

Accurate preoperative classification is important to optimize the surgical approach for patients with endometriosis but there is currently no reliable first-line examination to determine RE.

This was a single-centre observational study including all women (N = 119) who underwent complete surgery for endometriosis between January 2011 and June 2016 in the Gynaecological Department of the University Hospital of Poissy Saint-Germain en Laye.

Of the 119 women, 47 had RE and 72 did not. Two-thirds of the patients were randomly selected to derive the predictive score based on multiple logistic regression with internal validation by bootstrap. We used information from a self-assessment questionnaire, digital and speculum examination, transvaginal ultrasound and MRI. The score was then applied to the remaining sample of patients for validation.

Four variables were independently associated with RE: palpation of a posterior nodule on digital examination (aOR=5.6; 95%CI [1.7-21.8]); a UBESS score of 3 on ultrasonography (aOR=4.9; 95%CI [1.4-19.8); RE infiltration on MRI (aOR=6.8; 95%CI [2-25.5]); and presence of blood in the stools during menstruation (aOR=5.2; 95%CI [1.3-24.7]). The ROC-AUC of the model was 0.86 (95%CI [0.77-0.94]) and the bootstrap procedure showed that the model was stable. The ENDORECT score was derived from these four criteria and three risk groups were identified: the high-risk group (score>17) had a probability of RE of 100% with an specificity (Sp) of 100%, postive likelihood ratio (Lr+)>10; the intermediate-risk group (score: 7-17) had a probability of RE of 42%; and the low-risk group (score=0), with a sensitivity (Se) of 97%, negative likelihood ratio (Lr-) of 0.07 and a probability of RE of 5%. In the validation cohort, a score >17 predicted RE with an Sp of 96, Lr+ of 9.2, and probability of RE of 83%. Patients in this sample with a score=0, had an Se of 100%, Lr- of 0 and a probability of RE of 0%.

The single-centre recruitment and over-representation of RE could constitute a referral bias.

The use of a preoperative predictive score could facilitate patient counselling and guide surgical management. Both MRI and transvaginal ultrasound provide independent information and are useful before surgery for RE.

No financial support was specifically received for this study. The authors declare no conflict of interest.

N/A.