Swallowing therapy compared to standard care may not have had a beneficial effect on the incidence of pneumonia for patients in acute care. Author’s reply

Abstract

In reply to this correspondence, we thank the authors for their feedback on our review and respond to their comments [1, 2]. We acknowledge that there is a debate around fixedand random-effects models [3]. Cochrane does not provide a universal recommendation or suggests that risk of bias is a justification for choice of model. In the review protocol, we planned to use a fixed-effects model under the assumption that the true effect of dysphagia interventions (in size and direction) would be the same in every study and that observed differences among studies would be due to chance [4]. In keeping with our protocol, further analyses using a random-effects model was only completed when there was substantial heterogeneity (> 50%). This occurred for two outcomes in this review; quality of life and length of hospital stay and effects were not found to be different (Electronic supplementary material, Appendix J). Heterogeneity was 33% (acute care studies) and 15% (both acute and critical care studies), respectively, for pneumonia incidence, and hence, why additional analyses using a random-effects model was not conducted. Perhaps given the ongoing debate, sensitivity analysis could be conducted for all outcomes in future reviews. Regarding the authors’ second point on not offering the analysis of quality of studies, such analysis was completed using the risk of bias tool and judgements for each of the domains provided. While this detailed analysis was not reported in the main review paper, it is available in Appendix I of the Electronic supplementary material. Finally, the authors highlight that a trial sequential analysis was not completed in this review and suggest that this would have been “important to assess the risk of random errors due to sparse data and multiple testing of accumulating data and to calculate the required information size”. We understand the benefits that a trial sequential analysis (TSA) can bring to understanding in producing a required information size, and the limitations of undertaking it often on the primary outcome that requires pre-specified assumptions. We did not plan a TSA at protocol development stage, because our review followed the Cochrane methodology that did not recommend TSA methods (https:// metho ds. cochr ane. org/ sites/ defau lt/ files/ public/ uploa ds/ tsa_ expert_ panel_ guida nce_ and_ recom menda tion_ final. pdf ).