The sum of all parts – a reply

Abstract

We read with great interest the remarks by Marhofer and Hopkins [1] on our systematic review, meta-analysis and trial sequential analysis on intravenous dexamethasone after spinal anaesthesia [2]. One issue raised by Marhofer and Hopkins relates to 'the skewed distributions of the primary outcome data,' which suggests that this outcome in our study, the 24 h morphine equivalent consumption, was skewed while our meta-analysis was calculated with non-skewed data. The distribution of data can be nonskewed (following a normal Gaussian distribution) or it can be skewed, indicative of a non-normal distribution; for a non-skewed distribution mean and standard deviation and for a skewed distribution a median, interquartile range and range are reported. It is inherent to the statistical algorithm underlying all meta-analyses to calculate with mean and standard deviation and there are formulae to derive these variables from skewed data. This is also true for the Cochrane reviews that Marhofer and Hopkins quote. They criticise these formulae, stating 'their use increases the chance of a false positive finding' and quote a study from 2014 [3]. However, this study does not say this, it rather improves an existing and widely used formula [4]. In our Kantonsspital Baden,