Animal experimental research design in critical care

Justin S. Merkow, Janine M. Hoerauf, Angela F. Moss, Jason Brainard, Lena M. Mayes, Ana Fernandez-Bustamante, Susan K. Mikulich-Gilbertson, Karsten Bartels

Limited translational success in critical care medicine is thought to be in part due to inadequate methodology, study design, and reporting in preclinical studies. The purpose of this study was to compare reporting of core features of experimental rigor: blinding, randomization, and power calculations in critical care medicine animal experimental research. We hypothesized that these study design characteristics were more frequently reported in 2015 versus 2005. We performed an observational bibliometric study to grade manuscripts on blinding, randomization, and power calculations. Chi-square tests and logistic regression were used for analysis. Inter-rater agreement was assessed using kappa and Gwet's AC1. A total of 825 articles from seven journals were included. In 2005, power estimations were reported in 2%, randomization in 35%, and blinding in 20% (n = 482). In 2015, these metrics were included in 9, 47, and 36% of articles (n = 343). The increase in proportion for the metrics tested was statistically significant (p < 0.001, p = 0.002, and p < 0.001). Only a minority of published manuscripts in critical care medicine journals reported on recommended study design steps to increase rigor. Routine justification for the presence or absence of blinding, randomization, and power calculations should be considered to better enable readers to assess potential sources of bias.