Comparison of patient-controlled epidural infusion with nurse-administered epidural infusion
Winchester, UK
E-mail: rob.chambers{at}dsl.pipex.com
Editor—I read with interest the study comparing patient-controlled epidural infusion with nurse-administered epidural infusion for patients undergoing colonic resection.1 The authors correctly identify that the four-point verbal rating scale (VRS) used to assess postoperative pain would produce non-parametric or qualitative ordinal data and used the Mann–Whitney U-test for statistical analysis. The authors also describe the mean and standard deviation for the VRS scores which are parametric statistics and not appropriate for these data. Instead, the median and inter-quartile range should be used to describe the central tendency and spread of non-parametric data. The conclusion that mean summary pain scores are significantly lower in the patient controlled epidural analgesia (PCEA) group is therefore inaccurate. The Visual Analogue Scale, which utilizes a 100 mm continuous scale to assess pain, could be described as quantitative, ratio data and may be more appropriate for parametric analysis.
Portsmouth, UK
* E-mail: jeremy.nightingale{at}porthosp.nhs.uk
Editor—Thank you for the opportunity to respond to Dr Chambers' comments. As clearly indicated in the text and stated in the results section, median scores were reported and compared using a Mann–Whitney U-test. We feel that he may be confused by the fact that the ‘mean’ score of repeated measurements per patient was used as one of the outcome measures and thus formed the unit of measurement, and not the summary statistic for comparison. More precise wording may have helped, stating that the median of the mean scores per patient was found to be significantly lower in the PCEA group.
Dr Chambers also feels that presenting means and standard deviations is inappropriate. We disagree on three counts. First, means and standard deviations are often used to inform sample size calculations, even in situations in which non-parametric tests are thought to be necessary. We felt that publishing these figures might help others in the quest to design studies where the same measurement tool is used. Secondly, means and standard deviations are more easily combined through meta-analyses to provide evidence from several studies in systematic reviews. Finally, this study might be considered sufficiently large to justify the use of parametric tests to compare the summary outcome measures.
The Visual Analogue Scale was not employed in this study and therefore parametric analysis was not employed.
Reference
1 Nightingale JJ, Knight MV, Higgins B, Dean T. Randomized, double-blind comparison of patient-controlled epidural infusion vs nurse-administered epidural infusion for post operative analgesia in patients undergoing colonic resection. Br J Anaesth (2007) 98:380–4.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||