Septum resection versus expectant management: paucity of good quality data

Abstract

Sir, We read with keen interest the article by Rikken et al. (2021), titled ‘Septum resection versus expectant management in women with a septate uterus: an international multicentre open-label randomized controlled trial’. We applaud the efforts of the authors to tackle this difficult subject and recognize the challenges associated with recruiting into a study of this nature. Some members of the Association of Gynaecological Endoscopy Surgeons (AGES) of Nigeria felt compelled to write to the Editor to contribute our thoughts. Although not the first randomized controlled trial (RCT) on the subject, this study gains originality because it appears to be the only RCT so far that compared surgical with expectant management in women with uterine septum who have a history of subfertility, pregnancy loss, or preterm birth. The fact that this study went on for about 10 years to reach the sample size of 80 reflects just how difficult it was to execute. Of the 39 women randomized to septum resection, 12 had a live birth (31%) compared with 14 out of 40 (35%) who were randomized to expectant management (relative risk [RR] 0.88, [95% CI 0.47–1.65]) The authors, therefore, concluded that hysteroscopic septum resection does not improve reproductive outcomes in women with a septate uterus. Whilst we congratulate the authors in their attempt to provide an answer to this vexing question, we believe that some aspects of the study design and, not least, the study size weakens the reliability of its conclusion. Firstly, it appears that the authors, based on a priori power calculation on two reviews (Homer et al., 2000; Chan et al., 2011a). However, we found that neither of these provided reliable live birth data for expectant management of uterine septum. The systematic review by Chan et al. (2011b) reported clinical pregnancies (without time limit) in 40 out of total of 47 (85%) women with uterine septum compared with 50 out of 52 (99%) women with no anomalies. No live birth data was included. The article by Homer et al. (2000) found that, after treatment, the cumulative live birth rate (LBR) was 39% at 36 months for women with uterine septum associated with infertility; and 62% for those associated with miscarriages. Therefore, if the authors had planned to group all women with uterine septum together, whether presenting with subfertility, pregnancy loss, or preterm births, what was their rationale for presuming LBR of 35% in expectant versus 70% in surgical arm, with a power of 80%? This decision gives a misleading impression that the authors worked back towards a convenient sample size, rather than a statistically adequate sample size. As it has turned out, a post hoc power calculation seems justified; so, using LBRs of 31% versus 35%, it appears that these researchers ideally needed about 415 in each arm of the study (with an alpha error of 5% and power of 90%) to prove that the difference between 31% and 35% was not due to chance. Even if power was graded down to 80%, as was done in this study, they would still have needed 310 in each arm of the study to eliminate systemic random bias. Furthermore, the inclusion of subfertile women, irrespective of causative co-factors (male, anovulation, tubal, etc.) and women with history of one miscarriage are of dubious value, particularly in a study with such small sample size. Perhaps the authors anticipated low recruitment at onset, but in our opinion, this is not a justification for not eliminating other factors before including subfertile women. The small sample size makes subgroup analysis by reproductive factor of little value. It was also not clear what proportion of pregnancy on each arm were spontaneous conception or assisted conception. Finally, we believe that the follow-up period of 12 months from randomization, was inadequate. This is because conceptions occurring after 9 months following second-look (control) hysteroscopy were not likely to have been included for analysis. Not surprising therefore that LBR of 31% versus 35% were significantly lower than figures described in the review by Homer et al. (2000) for women with miscarriages, who form 63% of those studied. We felt that this low LBR was most probably due to attrition bias, where the expectant group started trying 3 months earlier and from incomplete follow-up. Women randomized to surgery also had second-look (control) hysteroscopy 6–8 weeks postoperatively, therefore this leaves them only 8–9 months to achieve a pregnancy. In conclusion, although we congratulate the authors for contributing to the body of literature on this important subject, we do not believe that their work resolves the issue of whether to operate or not to operate on a uterine septum. We believe that each case should be managed on its own merits following a frank and open discussion with the patient, who must understand the pros and cons of intervention and the paucity of good quality data upon which they can make an informed decision. None