I read the article by Robinson et al.1 with interest. The authors should be congratulated in completing this study on a substantial sample of women. However, I would be grateful if one or two points in the results and the statistical analysis could be clarified.
The important question is whether ultrasound measurements of bladder wall thickness can be adequately diagnostic. Central to this is: How much do these thickness measurements overlap between groups diagnosed urodynamically?
The authors give results including those below, and conclude that ‘Examination of the 95% CIs reveals no overlap in those women with a diagnosis of detrusor instability and in those with a diagnosis of GSI’:
This appears to be a non-sequitur. It is perfectly possible to have different means with non-overlapping confidence intervals, but still have so much overlap between the groups that many individual measurements will not be diagnostic. It is essential to distinguish between the 95% CI and the spread of values in the group. The 95% CI normally refers to the uncertainty in estimating the mean of the whole patient group from a limited sample (i.e. there is a 95% chance that the ‘real’ mean lies in this range). The bigger the sample, the smaller this range becomes. In the limit of measuring the entire population, there is no uncertainty in the mean at all. This is quite different to the spread of measurements in the sample, which does not get smaller as the sample size is increased.
If the intervals quoted above are indeed the 95% CIs, then it is easily shown that the spread should be around five times larger for the samples in this study. The authors helpfully include a reference to an earlier paper from the same centre2, where it is clear (Fig. 3) that corresponding measurements have a range much nearer 7 mm than 1.4 mm. It would then follow that the overlap between the groups is significantly greater than we hoped, and the diagnostic value is correspondingly less.
Ultimately, it would be very helpful to quote the specificity and the sensitivity of the ultrasound approach for appropriate choices of threshold bladder wall thickness, such as the 6.0 mm the authors suggest. This would be firmer ground on which to judge the usefulness of the ultrasound measurement.