Your correspondents1 raise some interesting issues and ask for more clarity about the sample. In terms of the population, the practice involved in the study has a list of 9000 patients with the age distribution the same as the national average and a full mix of socioeconomic groups with emphasis on 3 and 4. The selection of women was by random number generation, with a specific number being randomly selected from from each age band.
The response rate at 50% of those eligible is in keeping with what is typically found from surveys and, indeed, on the more positive side. We would obviously agree that greater participation is desirable, but many bodies of literature are now dependent on such samples. The original sample size as in all studies was calculated to take account of non-returners. In fact we assumed that 40% would not return and also took account of the fact that further women would require exclusion because of the criteria. The requirement for acceptable power was for 39 per group and although some groups are slightly lower than desirable, it is the patterning that is of particular interest.
It is clear from the means and standard deviations that the distributions of the groups are closely matched; certainly there is no indication of difference suggestive of deficit linked with the menopause. This is clearly stated towards the end of the paper: ‘In terms of power, while it could be argued that a larger sample may have detected significant differences, the means reported in the descriptive data are very similar (the majority being <1) and where differences are suggested by the data these are not in the predicted direction’. A key issue here is that it's crucial that results that question received wisdom or the status quo in not finding differences as expected should be published. We are all well aware of typical publication bias toward positive results and how this may distort perspectives.
Finally, although your correspondents quip ‘that women with the real memory problems simply forgot to return the questionnaire’ all the sample were sent reminders, and the measurement concerned perceived memory problems. The fact that these are strongly associated with emotional distress rather than age or menopause is of importance as it allows an alternative understanding of aetiology.
- © British Journal of General Practice, 2004.