Author response: Laparoscopic-assisted transversus abdominis plane block versus port-site infiltration in appendicectomy: multicentre randomized clinical trial

Correspondence to: Gordon R. Daly (e-mail: gordondaly@rcsi.com)

Department of Surgery

RCSI University of Medicine and Health Sciences

123 St. Stephen’s Green

Dublin

D02 YN77

Ireland

BJS, https://doi.org/10.1093/bjs/znaf257, published 09 January 2026

Many thanks for allowing us the opportunity to respond to Salmonsen et al.’s comments. We appreciate the interest of the Danish colleagues in our study on transversus abdominis plane block1. However, there are several points that they have misinterpreted that require correction.

Firstly, they have clearly misunderstood the organization of general surgery in Ireland and our recruitment of patients, when they suggest that it should be representative of the hospital services in the capital area.

Secondly, they challenge our use of parametric statistical methods for the visual analogue pain scale and argue that a confidence interval that includes a negative value (-1.0) indicates a violation of normality. Unfortunately, their calculation is wrong, and the 95% confidence interval for the baseline mean score of 2.4 is 2.1-2.8 and does not include a negative value. In preparing the manuscript, we have of course considered the validity of our chosen analytical strategy. The parametric approach is relatively robust when there is a reasonable sample size and offers a range of attractive opportunities in adjusting for background characteristics and accounting for inter-personal correlations2 . In this analysis, we particularly exploited the longitudinal nature of data obtained for the same individual at different time points by estimating random effects models. These options are readily available and easy to interpret, in contrast to a non-parametric alternative. For these parametric models, there is no requirement that the dependent variable is normally distributed, but that the residuals are. Some of these attractive features could also be achieved by considering the categorical variables as ordinal variables. Statistical models for the longitudinal analysis of ordinal data are more complicated to estimate and interpret. When we tested these models, we found that the estimates were very similar and all confidence intervals overlapped.

Finally, when patients were discharged from the hospital, they did not provide data for the subsequent observations. When missing data are ignored in the random effects models, an assumption of missing by random is imposed. This would indeed give biased change scores. However, because the patients were only discharged when they had low pain scores, ignorance of the early discharged patients would be more problematic than assuming that they were pain-free at subsequent, missing data points. In the analysis, a no-pain score was assigned to three patients at the 12-hour observation and 15 patients at the 24-hour observation.