We read with interest the article of Margot (2015) about our paper published 10 years ago (Roman, Soriano, Fuentes, Galvez, & Fernandez, 2004), and we wish to comment on some points.

First, our study arose simply from a perception among nurses and physicians at our Gastrointestinal Bleeding Unit that the number of admissions was higher on days with a full moon. When we designed the study in 1995, we found only one article, a paper by Perez-Cuadrado, Montero, and Bello (1986), addressing the issue and describing a higher number of admissions because of gastrointestinal bleeding in full moon. The aim of our study was therefore very simple: to determine whether the number of admissions to our bleeding unit was higher on full moon days than on non-full moon days. From the 738 days of the period studied (25 lunar cycles), there were 25 full moon days and 713 non-full moon days. We found the mean number of admissions on full moon days (M = 1.04, SD = 0.93) was higher than the mean on non-full moon days (M = 0.59, SD = 0.78, p = .007, using Mann-Whitney test). Margot states "Roman et al. suggested that the number of hospital admissions related to gastrointestinal bleeding was somehow influenced by the phases of the Earth's moon." This is not exact. Our conclusion in the abstract reads as follows: "The results of this study suggest an increase in the number of admissions related to gastrointestinal hemorrhage in our bleeding unit during the full moon, especially in men and in patients experiencing variceal hemorrhage. However, the wide variation in the number of admissions throughout the lunar cycle could limit interpretation of the results. Therefore, further studies are needed to clarify the possible influence of the moon on gastrointestinal hemorrhage."

Second, Margot considers the number of days included in the study is not clear and criticizes our "assignment of hospital admissions to one of the 29 days" of the lunar cycle. We will try to clarify these aspects. The study period included 25 complete lunar cycles (a total of 738 days). Full moon days were the days so-named on the local lunar calendar, and non-full moon days were the remaining days. Therefore, there was a total of 25 full moon days (considered day number 29 of the lunar cycle in our study), and the remaining 713 days were non-full moon days. As Margot states, the lunar cycle has a variable duration with a mean of 29.53 days and some cycles of 30 instead of 29 days. During the study period, 13 cycles had 30 days, and these 13 "extra" days explain the difference between the 25 cycles x 29 days = 725 days and the 738 days finally included in the study. There were no admissions on these 13 "extra" non-full moon days. Moreover, in no case were the admissions of two consecutive days assigned to one full-moon day, as Margot suggests. We omitted day 30 in the figure, showing the cumulative total number of admissions per day of the lunar cycle to avoid confusion, because there were only 13 Day 30s, compared to 25 days in each of the remaining days, 1 to 29. This design was chosen to express the results in a way as close as possible to clinical practice, but we acknowledge that it could make analysis of the data difficult if someone wanted to compare each day of the lunar cycle. We emphasize, however, that our aim was not to do this but to compare number of admissions on full moon days with all the non-full moon days, and we consider that this analysis was performed reliably. We disagree about the long series of shortcomings that Margot suggests and consider our data were correctly managed for the purpose of the article.

The third issue refers to our statistical analysis. We used the nonparametric Mann-Whitney test instead of Student's t test because of the nonnormal distribution of the data as evaluated by the Kolmogorov-Smirnov test. Margot states the use of Mann-Whitney test was incorrect, because the dependent variable (number of admissions per day) was not continuous or ordinal. Indeed, it was not continuous, but it was ordinal. Moreover, rank ordering was performed automatically by the SPSS program, and the mean rank was 470.04 in full moon days and 365.95 in non-full moon days. These data were not provided in our article because they are not usually specified in clinical research and are not easy to understand. We consider Margot's statement that "when the probability distributions of the two groups are not identical, the Mann-Whitney test cannot be used" is too restrictive and not shared by other authors (Conover, 1980). Moreover, Levene's test did not detect significant differences in the variances of the number of admissions between the full moon and non-full moon days. We also conducted other statistical evaluations, such as the Student's t test, regression analyses, and the median test, and the differences between the two groups were also statistically significant in all of them. Margot's analysis using a Poisson process can be an interesting method to explore variability in the number of admissions along the lunar cycle, but we emphasize again that this was not the objective of our study.

Fourth, Margot states that the day of the week is a confounding factor not accounted for in our study that could explain the differences observed between full moon days and non-full moon days. As commented by Margot, in studies analyzing traffic accidents, "what the authors had initially observed and incorrectly ascribed to a lunar influence was merely day-of-week variability." However, we consider that, if an increase in traffic accidents during the weekends could have been expected, there is no apparent reason to explain a possible relationship between the day of the week and the number of admissions because of an event such as gastrointestinal bleeding. Moreover, we have reanalyzed our data, and we did not find significant differences in the number of admissions between the different days of the week.

Fifth, Margot argues against our speculation concerning the possible effect of the full moon on splanchnic hemodynamics. We wish to emphasize that this was only speculation and was presented as such in the Discussion. Margot also writes that the absence of an increase in hospital admissions at new moon further invalidates this interpretation. However, we ourselves made the same comment in our article. We underscore the descriptive nature of our study. We did not state that a higher number of admissions on the days with a full moon was a consequence of a moon effect. Furthermore, this higher number of admissions does not mean that patients bled more frequently on full moon days, as the time between the first symptom of bleeding and the hospital admission is highly variable, as reported in other articles from our bleeding unit (Villanueva et al., 2006) and from other groups (Ingrand et al., 2006). Moreover, we did not pretend at any time to determine the mechanisms by which the moon could have affected the number of admissions for gastrointestinal bleeding.

Sixth, the author points out the need for "reproducibility and predictability" of studies evaluating the lunar effects. We, of course, agree on this point, but we have only found two other studies evaluating the possible relationship between full moon and gastrointestinal bleeding, and their findings were contradictory (Perez-Cuadrado et al., 1986; Tahri, Amouri, Fekih, El-Euch, & Krichen, 2003). Margot reviews a high number of studies demonstrating the absence of a correlation with lunar phases in a different setting, in relation to birth rates. However, although it seems clear that deliveries show no relationship with lunar phases, this does not necessarily mean that the full moon does not have any relationship with any other physiological or pathological situations, such as gastrointestinal bleeding.

Finally, we agree with Margot regarding the value of interdisciplinary approaches, to investigate not only possible effects of the lunar cycle but also in any area of research in health sciences. However, if our objective is to contribute to improvements in the healthcare of our patients, it is essential that the clinical environment in which a hypothesis is set and the exact question that clinical researchers pretend to answer are carefully taken into account.

References