Should We Rely on the Findings of Each Published Randomized Controlled Study?

which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Nonalcoholic fatty liver disease (NAFLD) is likely the most common cause of chronic liver disease in many countries with controversies relative to its optimal treatment (1). Despite my interest to the results of Haji-aghamohammadi et al. study (2), on this topic there are shortcomings which should be taken into account before going to their results and implement the findings in clinical practice. First, the multivariate analysis-of-variance (MANOVA) is a generalization of ANOVA allowing multiple dependent variables analysis. Here, multiple dependent variables are considered but not at the same time in the analysis. So, the authors should use ANOVA instead of MANOVA. There is no indication for MANOVA according to the observed results and tables. In addition, even when the basic values are not different (maybe due to low power of the statistical test because of low sample size), it is better to do ANOVA with considering those basic values as covariates. It is due to the fact that there are remained differences which may not be statistically different but exist. Moreover, by considering baseline characteristics, personal differences could have been considered better than the way used in this paper. Additionally, most differences are not significant when there is low sample size. So, non-significant baseline differences can be due to low sample size. There are also some comparisons between subgroups like changes in AST, ALT, FBS, insulin level, and HOMA index in silymarin group in comparison with other groups. Post hoc analysis is needed for such comparisons with three different groups to show which subgroup difference(s) have caused significant difference between three groups at all. They did not mention how they have found these findings neither in methods nor in analysis. According to our post-hoc analysis, many of these differences are not significant despite what the authors have mentioned. Moreover, when we do multiple comparisons , we should use methods like Bonferroni or Holm