The BDJ strives hard to achieve the highest standards of publication, both in clinical outlook and statistical content, and yet the statistical analysis and presentation of results in the paper by Delilbasi et al1 on gustatory function in postmenopausal women are largely either inappropriate or inadequate. As a consequence, I believe that the conclusions should be viewed with caution (perhaps with a pinch of salt?).

The authors have made no attempt to justify their choice of sample size using power calculations. It should be remembered that lack of significance, when it arises, may well be a consequence of low power. Although the authors have avoided some biases by randomising the order in which the stimulator solutions were taken and the order of locations for painting, each solution was given in increasing concentrations rather than at random.

Smoking is associated with increased (worse) taste thresholds but no indication is given of the proportions of male and female smokers. The type of analysis of variance used has not been described nor have the circumstances in which it has been employed. The authors regard an 'acceptance of a probability of p<0.05 as significant' but this is misleading since if p < 0.05 then the null hypothesis is rejected.

The postmenopausal women were compared to age-matched males but the non-parametric Mann-Whitney U test used to investigate differences between the gender groups treats the observations in the two groups as independent rather than paired. And why use a non-parametric test, which is useful for skewed data and essentially compares medians, and then report the result as a difference in means (where is the estimate? for which concentration?) when the mean is not a sensible measure of central tendency for skewed data? The authors indicate, incorrectly, that the relevant mean scores are contained in figures 1 and/or 2. In fact, figure 1 gives the results for each concentration separately and labels the vertical axis as '% of first identification (mean + se)'.

This is confusing since the percentage of individuals in a particular gender group first identifying a stimulus is a single figure and not a mean. In figure 2 the six oral locations are marked individually on the horizontal axis although the legend explains that the diagram shows the mean taste intensity ratings given to each compound summed over six oral locations.

Does each shaded portion of a bar for a given location then represent the mean taste intensity rating averaged over the compounds (where is its standard error or confidence interval?) for males or females?

The authors report that 35% of the women noticed failure in tasting sweet, salty, sour and bitter substances as strongly as before the menopause, but since taste sensitivity declines with age and complete amenorrhoea time is up to 15 years in these women, this could be a reflection of age and not of menopausal status.

Dr. Cagri Delilbasi, one of the authors of the paper responds: We are thankful to Ms Aviva Petrie for her assessment and advice about our article 'Evaluation of gustatory function in postmenopausal women' in the BDJ. There are some points we want to emphasize. In this study, to avoid some biases, the order in which the solutions were taken and the order of the locations for painting were randomized; however, the solutions were given in increasing concentration.

When the physiology of taste is considered, high concentration of a solution may reduce the person's ability to recognize a lower concentration of that solution. When we searched similar threshold studies about taste perception, we noticed that the solutions were mostly given in increasing order. When conducting such studies, one should consider 'physiologically' not only theoretically. We agree with Ms Petrie that it would be better if figure legends were more descriptive and clear for the reader to better understand what we wanted to say. The statistical methods used in the study were chosen by our statistician. Of course it is open to criticism as there are different ways to analyze the results.

There are many external factors that may affect taste perception such as smoking, dietary habits, denture use and alcohol consumption. In order to make our study more objective, we restricted the inclusion criteria for the study as mentioned in the 'Subjects and Methods' section. However, if there is a small sample size, it is not always possible to rule out all the factors that could influence the study. This is not a matter for only this study. We asked two questions to female subjects to learn their self-assessment of change in taste perception and dietary habits before and after menopause. Although age can influence this assessment besides many factors, the purpose of asking these questions was just to have information about the subjective evaluation of the participants.

The answers supported the results of the taste tests. We want to thank again to Ms Petrie for her careful evaluation of our paper. These kind of critics are very valuable because they help scientists to better prepare future studies.

C. Delilbasi

By email