We thank Manyonda et al.1 for their interest in our trial2 and their comments on our speculations on possible reasons for the difference between our results2 and theirs3 on the effect of total versus subtotal hysterectomy on urinary incontinence.
We suggested that it could be due to a difference in sample size. Manyonda et al.1 find that we have misinformed the readers of BJOG because we state that their trial3 was smaller than ours.2 Thakar et al.3 randomised 279 women to one of the two operations. We randomised 319 women.2 Both trials suffered from losses to follow up. Thakar et al.3 lost 21 patients from the total and 12 from the subtotal hysterectomy group and included 246 patients in their final analyses, although one table (Table 3) only counts 119/117 and 121/120 patients in the two groups. In our trial,2 18 women in the total and 24 in the subtotal group were lost so our final analyses included 277 women.
Manyonda et al.1 state that their power calculation was based on the Battacharya et al. trial4 and that a larger sample size would not have clarified the issue further. Calculation of sample size is always based on assumptions. Looking at Thakar et al.'s assumptions,3 we wonder if they are correct. Thakar et al.3 uses 55% complaining of stress incontinence in the control group. To us it appears that only 44% of the women in the hysterectomy group had subjective complaints of stress incontinence in the Battacharaya et al. trial.4 So the control event proportion does not seem correct in the Thakar et al.3 sample size calculation. If we recalculate the sample size from the assumptions made by Thakar et al.3 (type I error: 5%, type II error: 10%, prevalence of stress incontinence in control treatment: 55%, difference between treatments: 18%), we reach a total number of 316 patients.
We also question if the Bhattacharya et al. trial4 is the best to refer to for sample size calculation. Large variations are found in the proportion of urinary incontinence after hysterectomy. Bhattacharya et al.4 reported 65% when both stress and urge incontinence are considered, whereas Milsom et al.5 reported only 21% when stress and urge incontinence are considered. The higher you assume the proportion of the outcome to be, the less patients you will need in your trial. The choice made by Thakar et al.3 may have made their trial too small.
We apologise that we misquoted Thakar et al.3 by stating that 35% were of Afro-Caribbean origin. This should have been 25%. However, 25% is still a large proportion. If these women have a lower incidence of stress urinary incontinence,6 this might influence the chance of finding a difference in the trial by Thakar et al.3 Manyonda et al.1 argue that the ethnic diversity of their study population3 renders their results more relevant to the international community. Manyonda et al.1 mix two topics, the sample size and the external validity. If a trial is too small, then internal validity is questionable. Without internal validity it becomes irrelevant to speak about external validity.
Manyonda et al.1 criticise us2 for not differentiating between urge and stress incontinence. We measured urge as well as stress incontinence in our questionnaire, which can be seen on the Internet site referred to in our paper.2 In our trial report,2 we chose to analyse urinary incontinence as one group. More detailed analyses—separating types of incontinence and reporting other lower urinary tract symptoms—will be reported soon.7
With regard to blinding, we are well aware of the bias lack of blinding may introduce.8,9 However, we do not think it is a likely explanation of the discrepant findings. Studies8,9 have demonstrated that lack of blinding would tend to exaggerate the effect of the experimental intervention (i.e. subtotal hysterectomy). In our trial, the control intervention, total hysterectomy, is the treatment with the better result. Another aspect regarding blinding in trials on subtotal versus total hysterectomy is the fact that it is hard/impossible to blind such trials. The patients are able to find out which treatment they have received by self-examination of the vagina. The doctors could consult the patient's record in the hospital. How do you inform patients about the possibility of vaginal bleeding after subtotal hysterectomy? And how do you treat and inform them when they consult your clinic in panic because of such bleeding post-operatively without mentioning that it is nothing to worry about because they had a subtotal hysterectomy? We think it is quite simple to ‘break the blinding’ in such trials.
One of the methodological differences between the trial by Thakar et al.3 and our trial2 is the use of sealed envelope technique for randomisation. As pointed out by Altman et al.,10 this is not a method which can be recommended because of the bias it may introduce. The paper by Thakar et al.3 did not provide enough information about this point. However, we are happy to be informed now1 that the discrepancy in the result could probably not arise from this difference in method.
Finally, Manyonda et al.1 find that their trial3 is more rigorous as they assessed pelvic organ function at three time points. Our trial assessed the pelvic organ function at four time points (pre-operatively and at 2, 6 and 12 months follow up) as written in the Material and Methods section.2 The results of sexual function and bladder function measured at all four time points are in press.7,11
In conclusion, we had no intention of misinforming. We only tried to discuss potential causes for the apparent discrepant findings. We leave judgement of the robustness and the rigorousness of the methodology and results of the two trials to readers and future Cochrane reviewers.