Results
From October 2009 to March 2020, 363 patients met inclusion criteria and were enrolled in the study. Among them, 241 and 122 women received surgery in Rouen and Bordeaux respectively. 312 women (86% of the total sample) were managed by one gynaecological surgeon (H.R.): 223 women in Rouen (92.5% of patients having surgery in Rouen) and 89 in Bordeaux (73% of those managed in Bordeaux). The rate of preventive stoma was 71.4% in Rouen (N=172) and 30.3% in Bordeaux (N=37). Rectovaginal fistula was recorded in 31 cases (8.5%): in 19 women managed in Rouen (7.9% of women having surgery in Rouen) and in 12 women undergoing surgery in Bordeaux (9.8% of those managed in Bordeaux).
Table 1 presents patient characteristics, the majority of which were comparable between the two groups. Patients managed in Bordeaux were older and had more frequent past history of open abdominal or pelvic surgery. Table 2 presents intraoperative findings, surgical procedures and main postoperative complications. Patients managed in Bordeaux had shorter operative time, larger rectal nodules, deep nodules more often involving both uterosacral ligaments and rectovaginal space, and more frequently underwent hysterectomy. Other surgical procedures and postoperative complications were comparable between the two groups, with the exception of bladder excision which was more frequent in patients managed in Rouen.
Table 3 presents the two groups of women stratified on performing or not performing stoma. The four groups compared were statistically different regarding rectal nodule size and vaginal infiltration (larger in the groups receiving stoma in both centres). The height of rectal stapled line was significantly lower in women undergoing stoma, particularly in those managed in Bordeaux. Women undergoing nodule excision involving sacral roots and the sciatic nerve were more likely to have a stoma. Conversely, the rectovaginal fistula rate was comparable between women managed in the two centres, with or without stoma. Rectovaginal fistula was recorded in respectively 9.4%, 10.8%, 10.1% and 7% in women managed without (8/85) and with stoma (4/37) in Bordeaux and in those managed without (7/69) and with stoma (12/172) in Rouen.
Table 4 presents the independent relationship between several risk factors and the likelihood of rectovaginal fistula. Performing rectal sutures at a height inferior to 8 cm above the anal verge led to a more than 3-fold increase in risk of rectovaginal fistula, independently of stoma confection, surgical procedure carried out on the rectum, extent of vaginal infiltration or associated excision of deep endometriosis involving pelvic nerves. When compared to women managed in Bordeaux without fistula, there was a tendency toward a lower probability of rectovaginal fistula in women receiving stoma in Rouen, however the P value did not reach statistical significance (P=0.075).

Discussion

Our study compared the risk of rectovaginal fistula in two groups of patients managed for rectovaginal endometriosis with differing policies concerning preventive stoma. Despite the stoma rate being more than halved in one group, no significant difference in risk of rectovaginal fistula following concomitant excision of rectum and vagina was found. Our study however showed that a rectal stapled line at a height of <8 cm was a major risk factor for fistula. This information is useful for colleagues involved in management of severe forms of endometriosis.
The major limit of our study relates to the lack of randomisation, resulting in the presence of women with and without stoma in each group. In a randomised trial, the assignment of a patient to the arm with or without stoma is based on a unique randomisation list drawn up by a statistician, resulting in two similar groups of women to compare. In our study, as in other non-randomised series7,21, performing stoma is based on colorectal surgeon decision and intraoperative events. The particularity of our study relates to centre-based policies, i.e. an overall intention to carry out a stoma in patients with concomitant rectal and vaginal repair in Rouen, although no stoma was ultimately performed in a third of cases, versus an overall intention to avoid performing a stoma in similar patients in Bordeaux, where a stoma was performed in one third of the cases. These differing approaches to the use of stoma led to a comparative study, the results of which may help in planning a future randomised trial.
The second limit relates to the heterogeneity of techniques used to remove rectal nodules. Our study is observational and employs data prospectively recorded in a cohort of patients managed for endometriosis in two tertiary referral centres. As deep endometriosis is a complex disease, there is a large variation in length, width, depth and height of rectal infiltration, requiring an individual surgical approach using either full thickness disc excision or colorectal resection22. However, both approaches resulted in the presence of a rectal stapled line, juxtaposed with vaginal repair. This condition significantly increases the risk of rectovaginal fistula, independently of the technique used to remove the nodule.
The third limit relates to the “before and after” design of our study. Patients were managed in Rouen prior to those in Bordeaux, thus it is likely that the surgeon practicing in both centres benefited from more experience during the second time period. This hypothesis is supported by shorter operative times recorded in Bordeaux.
Our study presents several strengths. The comparability of the two groups is ensured by the presence of one gyneaecological surgeon in both centres, who was in charge of management of the majority of patients. Surgeon recruitment and technique did not vary between the two centres with colorectal surgeons in Bordeaux and Rouen employing the same techniques to remove the rectum, i.e. disc excision and segmental resection, despite their differing approach concerning the use of preventive stoma. All surgeons involved in this study had extensive experience in endometriosis, which would logically favour good postoperative outcomes. Patients were prospectively enrolled in a cohort and benefited from rigorous follow-up and detailed recording of pre-, intra- and post-operative data. Data were managed by dedicated research technicians, avoiding patient lost to follow up and lending support to the accuracy of the data and the validity of our results.
Our study showed that the prevalence of rectovaginal fistula was comparable between the four groups stratified on the centre, and use or not use of stoma. When compared to women managed without stoma in Bordeaux, the centre with a policy favouring a restrictive use of stoma, women receiving stoma in Rouen , a centre with large use of stoma, were found to have a tendency towards a lower rectovaginal fistula rate (P=0.075), after adjustment for rectal stapled line suture, surgical procedure on the rectum, size of vaginal infiltration and management of pelvic nerves (the latter item being an indicator for more complex surgery). Even though a significant difference might be revealed by a larger cohort study, expected benefits should be weighed against disadvantages related to routine stoma use. We previously showed that women with deep endometriosis benefiting from preventive stoma, had a 8.6% risk of undergoing further surgery to manage specific stoma-related complications such as bowel occlusions, haemoperitoneum or leakage of bowel suture at the stoma opening, repair of incisional hernia or stoma prolapse12. Furthermore temporary stoma has been shown to increase the risk of colorectal anastomosis stenosis in women undergoing segmental resection for endometriosis of the rectosigmoid11. All patients with stoma are required to undergo additional surgery to close the stoma and restore the digestive tract, and may engender further stoma-related disadvantages such as aesthetic harm, residual pain, stoma prolapse, incisional hernias or subcutaneous infections. The benefits from a presumed decrease in risk of rectovaginal fistula should be weighed against the above-mentioned unfavourable consequences.
The occurrence of rectovaginal fistula is a major unfavourable outcome which impacts patient postoperative well-being. Though the immediate consequences are not usually life-threatening, as stools are evacuated through the vagina and not inside the pelvis and abdomen, rectovaginal fistula repair may be challenging. Simple deviation of stools through a stoma only incidentally allows rectovaginal fistula repair, as the presence of vaginal opening in contact with a discontinuous rectal stapled line may be an obstacle to natural healing1. Rectovaginal fistula repair requires several additional procedures in almost half the patients, while stoma closure cannot be carried out before 10 months on average1. These additional procedures, such as secondary segmental resection or delayed colo-anal anastomosis1,2, are much more complex than the repair of incisional hernia of a stoma scar or endoscopic dilatation of colorectal anastomosis stenosis. These reasons lend support to a reduction in risk of rectovaginal fistula by performing preventive stoma, as has been demonstrated in low rectal resection for rectal cancer 4-6.
It should be noted that 30.3% of patients managed in Bordeaux ultimately received a preventive stoma, colorectal surgeons estimating the risk of rectovaginal fistula to be too high. In Rouen, 28.6% of patients received no stoma, due to an estimated low risk of rectovaginal fistula. It results that an average of 40% of patients who had a stoma in Rouen were unlikely to have had a stoma performed in Bordeaux. Similarly 40% of those managed without stoma in Bordeaux would have probably had a stoma performed in Rouen. As patient characteristics and surgical procedures were comparable between the two centres, any difference in prevalence of rectovaginal fistula between the two groups is likely to stem exclusively from comparison between these 40% of patients, for whom the decision to perform a stoma or not, differed according to the centre where they were managed. In our series, this involved 50 patients in Bordeaux and 96 patients in Rouen and though these patient numbers do not provide sufficient statistical power to demonstrate a statistically significant difference, our data give an indication of expected rectovaginal fistula rates in the presence or absence of stoma. More specifically, the hypothesis of 7% rectovaginal fistula in women with large indications for stoma appears reasonable, being similar to that observed in women receiving stoma in Rouen, and to that reported in another series with a 96% stoma rate8. A higher rate of rectovaginal fistula in women with stoma in Bordeaux (10.8%) is also logical, as the risk of fistula was considered high enough to perform a stoma, by surgeons who would not routinely perform them. In a future randomised trial, the expected rate of rectovaginal fistula in women allocated to the arm without stoma, should be higher than 10.1% (the fistula rate in women in Rouen with no stoma). A randomisation process would allocate patients with high risk of rectovaginal fistula to the control group (30.3% of patients who had a stoma in Bordeaux), logically resulting in fistula rates higher than those recorded in our series. An important question concerns whether or not the fistula rate in women without stoma could exceed 16%, thereby surpassing the cumulative rates of rectovaginal fistula in women with stoma (7%) and of the surgical procedures required following complications incurred from routine use of stoma (8.6%)12. To our knowledge, to date, no other data in the literature provides answers to this question, highlighting the need for a randomised trial to compare rectovaginal fistula in women managed for deep rectovaginal endometriosis with or without stoma.
With regards to whether the rate of fistula in our series is valid and consistent with that observed in series published by other authors, only a few studies in the literature have focused on series of women managed for low rectal endometriosis and concomitant colpectomy, in which risk of rectovaginal fistula could be comparable to ours. Firstly in a series of 100 women undergoing low colorectal resection for deep rectovaginal endometriosis by open route, Dousset et al reported concomitant vaginal infiltration in 64% of cases, a 96% stoma rate and 6% anastomotic leakage (8). Secondly in a series of 44 women undergoing low rectal resection and colpectomy, Belghiti et al recorded a rectovaginal fistula rate of 15% in women with preventive stoma and 27% in women without stoma7. On the basis of this, our fistula rate observed in 363 patients undergoing colpectomy and excision of rectal endometriosis by either disc excision or segmental resection, appears reasonable.
In conclusion, this study did not reveal statistically significant differences in terms of risk of rectovaginal fistula between women with rectovaginal endometriosis managed respectively by a generalised or restrictive use of preventive stoma. However, our data underlines the higher risk of rectovaginal fistula independently related to a low rectal stapled line, specifically up to 8 cm above the anal verge. Our results also suggest that a policy of restrictive use of stoma in up to 30% of women with juxtaposed rectal and vaginal sutures does not significantly increase rectovaginal fistula risk when compared to a much larger employ of preventive stoma. A further randomised trial would enable practitioners to better weigh up the benefits in terms of rectovaginal fistula and related complications versus the risk of stoma-related complications.
Study funding/competing interest(s): The North-West Inter Regional Female Cohort for Patients with Endometriosis (CIRENDO) is financed by the G4 Group (The University Hospitals of Rouen, Lille, Amiens and Caen) and ROUENDOMETRIOSE Association. No financial support was received for this study. The authors declare no competing interests related to this study.
Acknowledgements: The authors are grateful to Amélie Bréant and Sophie Marinette for the management of data and Helene Braund for her help in editing the manuscript.