Re: ‘‘Meta-Analysis of Clinical Efficacy of Sildenafil,a Phosphodiesterase Type-5 Inhibitor on High Altitude Hypoxia and Its Complications” by Xu et al.

Dear Editor,

We have read the review article titled “Meta-analysis of clinical efficacy of sildenafil, a phosphodiesterase type-5 inhibitor on high altitude hypoxia and its complications” (Xu et al., 2014) published in High Altitude Medicine & Biology. The article provides noteworthy insight on the role of sildenafil in high altitude hypoxia and adverse affects associated with it.

Although it has been several years of publication of this meta-analysis, we have found a couple of statistical fallacies in the article, which we think should be brought to attention. The meta-analysis reports five studies (Ghofrani et al., 2004; Richalet et al., 2005; Faoro et al., 2007; Reichenberger et al., 2007; Bates et al., 2011) where the effects of sildenafil was compared with placebo to find out the inference of sildenafil in high altitude physiology. These studies reported the outcomes as arterial oxygen saturation, pulmonary artery systolic pressure, heart rate, Lake Louise Consensus AMS symptom score in terms of mean and standard deviation except one study (Ghofrani et al., 2004) that reported the outcomes as median and median difference. The authors have incorporated the median data in the article (Ghofrani et al., 2004) as mean in the meta-analysis, which we suppose is an error. The Cochrane Handbook for Systematic Reviews of Interventions states that medians are reported when the data are skewed and in skewed data median cannot be substituted by mean (Higgins et al., 2020). A method of estimating mean and variance from median has been derived by Hozo et al. (2005) using range and sample size, but it has not been used. Nevertheless, study by Ghofrani et al. does not have the smallest and largest value too for the estimation of range.

Another error we found is that the authors have reported the statistical error of mean value (SE) of arterial oxygen saturation given by another article (Reichenberger et al., 2007) as standard deviation (SD) in the meta-analysis. Standard deviation and standard error of mean are two different entities and cannot substitute each other. The relation between standard error of the mean and standard deviation is given by the simple formula SE = SD/√(sample size) (Altman and Bland, 2005).

Although these two statistical fallacies may not have major discrepancies in the results of the study, we think it should be brought to the attention of authors, editors and reviewers. This will help us to give more effort and inputs in coming days for in depth review of articles as well as consideration toward the statistical tools and measures used.