Animal models of cardiac arrest: A systematic review of bias and reporting

Abstract

Aim of the review: Animal models are essential in advancing resuscitation research but are susceptible to various biases compromising internal validity, which may explain unsuccessful transition to human clinical trials. This study aimed to assess risk of bias in animal studies of cardiac arrest.

Data sources: This study was based on a previous systematic review of all animal cardiac arrest studies published between March 8, 2011 and March 8, 2016 in PubMed and EMBASE. For this study, we focused on interventional studies and selected a random sample of 50 pig and 50 rat studies. We used a modified version of the SYRCLE's risk of bias tool for animal studies. Bias assessment was performed by two independent reviewers.

Results: 92% of pig studies and 88% of rat studies used randomization to assign interventions, but the methodology was unknown or insufficiently reported in 60% and 68% of the studies, respectively. Correct timing of randomization was lacking or unclear in over half of the studies. 40% of pig studies and 28% of rat studies reported insufficient baseline characteristics. When possible, blinding was not performed/reported in 68% of rat studies and 31% of pig studies. Blinding of outcome assessors was missing or inadequately reported in 65% of pig studies and 60% of rat studies. 80% of all studies lacked a sample size calculation, while 60% of pig and 80% of rat studies omitted a specified primary outcome.

Conclusion: This study indicates insufficient reporting and methodological shortcomings in animal models of cardiac arrest.