We appreciate Drs Mangano and Mangano’s interest in our article. We agree with their remark that, before undertaking any prospective clinical trial (either randomized or nonrandomized), a power analysis for sample size calculation should be performed according to the CONSORT guidelines. Calculating the sample size that would be appropriate to be used in a specific investigation increases the overall statistical power of the analysis, since not just statistically significant differences are assessed but, most importantly, clinically significant differences are identified.
The design characteristics of our clinical trial complied with most of the scientific criteria of evidence-based studies (prospective study, consecutive patients, split-mouth design, and so on). With regard to the power analysis, the calculations needed depend on the size of the effect in the population. Therefore, the first step in assessing a sample size is to estimate the size of the effect. If there are published trials similar to the one to be conducted, then the effects obtained in these studies can be used as a guide, and, by using typical statistical models for power analysis, the sample size can be calculated in advance. If there are no similar previously published studies, as was the case with our investigation, the corresponding calculations can be performed only by using a more simple analysis.
Thus, the calculation of the sample size in our study was based on the simple forms of the 1-sample t test (concerning ΔE) and the t test for dependent samples (concerning the L∗, a∗, and b∗ variables). In accordance with the split-mouth design, each tooth type would be observed under only 1 of the 2 regimens. This way, it can be easily verified that, by using an effect size (mean difference and standard deviation) equal to 0.8, an alpha type I error probability equal to 0.05, and 15 patients (or more precisely 15 teeth), the power to reject the null hypothesis is over 80% in both cases. During the 4-year period of the study, 34 patients satisfied the inclusion and exclusion criteria; at the end of the study, 8 patients dropped out, leaving 26 patients for the statistical analysis (or 13 teeth of each tooth type).
With respect to the CONSORT guidelines, indicating that more complicated statistical models should be used for data analysis than in our study, our trial did not compromise its relevance, since we clearly discussed differences between statistically and clinically significant findings in the article. Nevertheless, new trials on the same subject could use a typical power analysis for the sample size calculation according to the CONSORT guidelines, since they can use the data obtained in our study.