Pleasing some of the people none of the time

Editor—The editorial by Maddern et al was in large part a critical appraisal of our randomised trial of tension free vaginal tape.¹,² We endorse the view expressed by the authors that realistic funding for surgical trials is crucial to their meaningful outcome, and that contingency funding should be available so that trials are not underpowered. We accept that our failure to recruit up to the calculated sample size in this trial has an inevitable adverse effect on the power of its conclusions. We emphasised this in our publication and in a subsequent commentary.²,³

It must be recognised, however, that statistical power is a fickle concept. Our own trial sought to recruit 394 patients, on the assumption of 90% cure from colposuspension, and that 10% difference in cure would be clinically important. The trial achieved recruitment of only 344 patients, and, given that the cure from colposuspension was lower than that expected from the literature, the study had only 50% power to detect a 10% difference or 80% power to detect a 15% difference between the procedures under review.² The ongoing Medical Research Council trial of laparoscopic colposuspension seeks to recruit 290 patients on the assumption that colposuspension has a 80% cure and that a 15% difference in cure is clinically important.⁴ Surely this would mean that our own trial assumptions were not unreasonable at the time they were made.

Maddern et al express concern at the biases inherent in our study favouring tension free vaginal tape, on the basis of our failure to test assumptions in the analysis and our handling of missing data. We recognise that had we explored the range of possible scenarios—for example, that all missing data from one arm represented treatment failure and from the other arm treatment success, our conclusions may have been different. As the authors point out, however, women treated with tension free vaginal tape had shorter operating time, reduced hospital stay, less adverse impact on quality of life surrounding their surgery, and returned to normal activities more rapidly. Although the trade-offs that patients make in their decisions regarding surgical treatment are difficult to quantify, it is clear that more are choosing the less invasive approach, and that secondary trial outcomes may be of primary significance to patients.⁵

The broad definition of intention to treat is clear, but its full application is possible only when complete outcome data are available for all randomised subjects.⁶ The handling of post-randomisation withdrawals is undoubtedly crucial in the interpretation of cure in surgical trials.³,⁷ In our initial submission for publication, we were concerned to avoid bias favouring the new intervention resulting from differential withdrawal rates; hence we had undertaken what we described as a “modified intention to treat analysis,” including only those patients undergoing surgery. The advice from the editors and reviewers of our manuscript was that all patients randomised must be included in the analysis. It is therefore something of a frustration to us that Maddern et al now raise this criticism under an editorial banner.