We would like to thank Mr Ismail for his comments, and appreciate his interest in the study.1,2 We agree that the study could have been performed using a different methodology with potentially different results. All the points raised are an intrinsic limitation of a randomised controlled trial (RCT). For example, if we used a different type of drain, the results might be different. Similarly, standardising the vault closure might have given a different result, but the results would then be applicable to only that specific type of closure. Using RCT methodology in the context of complex surgery always results in this problem. If the protocol is too specific (i.e. vaginal hysterectomy only performed for prolapse, only one type of drain, and one type of closure, etc.), the results then apply only to that specific set of circumstances and are less generalisable.
The rationale for adopting the specific methodology was to reflect typical practice in the UK so that the results can be generalised.
Although the use of a vaginal pack was not standardised, the majority of the patients included in the study had a vaginal pack inserted (95% in the ‘drain group’ and 97% in the ‘no drain group’), which reflects usual practice in UK.
Febrile morbidity is the most common presentation of a symptomatic vault haematoma,3 and was therefore chosen as the primary outcome measure. The reasons for using febrile morbidity rather than ultrasound detection of vault haematoma, and the corresponding power calculation, have been described in the introduction and discussion of the article.3 The vaginal hysterectomies were performed by ten surgeons in different theatres, and therefore having the envelope attached to the notes prior to going to theatre was a practical way of ensuring that all patients who agreed to participate were included in the study. It also allowed for a more efficient intention-to-treat analysis, as patients had an envelope allocated only if a vaginal hysterectomy was performed. We used a Student’s t-test for the analysis, which does assume that the underlying distribution of the variable is normal, as there was no evidence to suggest that the variables were not normally distributed. However, there is no difference in the results in the two groups, even on a Rank sum test, which does not assume any underlying distribution. The indications for hysterectomy and number of patients who returned to theatre in both groups were discussed in the results section of the article.2 The number of women who had concomitant pelvic floor repair was also similar in both groups.