March 10, 2021

Nonruminant Nutrition Symposium II: Designing and Analyzing Nutrition Trials to Detect Small but Meaningful Differences

Summary: Nonruminant Nutrition Symposium II: Designing and Analyzing Nutrition Trials to Detect Small but Meaningful Differences
Dr. Caitlin Vonderohe, Postdoctoral Associate, Baylor College of Medicine

Today, several speakers presented on strategies to manage stress physiology in pigs and statistics use in non-ruminant nutrition at the Non-Ruminant Nutrition (II) symposium at the Midwestern section of the American Society of Animal Science meeting on March 9, 2021. The talks were kicked off by a presentation by Dr. Jay Johnson, recipient of the 2021 Young Investigator Award, who discussed pig response to transport and weaning stress, and recent work he has done to mitigate these stressors and strain to improve pig health and productivity. Dr. Johnson specifically presented work his group has done supplementing L-glutamine to pigs after weaning and exposed to simulated or actual transport stress. L-glutamine improved growth performance, feed intake, pig activity and other physiologic markers of stress, and some of these effects extended beyond the feeding period. Additional work, and future endeavors presented indicate that Dr. Johnson will continue to improve pig health, welfare and productivity into the future.

            The symposium switched gears after Dr. Johnson’s presentation, the remaining speakers discussed the challenges and solutions associated with statistical methods used in animal science research. The first of these talks, presented by Dr. Nick Serão focused on the training of graduate students in statics. There has been a gradual shift in focus from the actual statistics to SAS procedures and R program packages, which has adversely affected both statics use in data analysis and the training of graduate students. Dr. Serão then described common mistakes made in statistical analysis found in manuscripts, including correlations between explanatory and response variables, unsupervised machine learning algorithms for model based research, a lack of randomization during study design, and the removal of design effects from statistical models. Finally, Dr. Serão described a recent study that assessed graduate student perceptions of statistical training, and used a common exam to assess their training. He found a great disconnect between perception and reality; graduate students expressed familiarity and proficiency with statistical methods, but most failed the exam testing proficiency with those methods.

            The next speaker was Dr. Mike Tokach from Kansas State University who spoke about another facet of statistical analysis in animal science research, strategies to use to reduce variance during data collection in swine nutrition trials. Dr. Tokach emphasized that, while there are multiple times one can reduce variance when planning and executing a study, the underlying themes involve trustworthy, detail oriented people executing the study and collecting the data. Trustworthy data collection hinges on good planning, appropriate facilities and equipment, and strenuous data collection methods and management strategies.

            The final speaker in this symposium was Dr. Neil Paton, applied statistician, from Cargill Health and Nutrition. Dr. Paton described the history and significance of null hypothesis significance testing (NHST) and the use of the p-value in animal science. Although NHST has been used and popularized in animal science research since the 1930s, recent issues surrounding the repeatability and inconsistencies in the application of p-values have indicated a potential for researchers to explore additional ways to describe and analyze data. Most importantly, Dr. Paton asserted that genuine thought and care need to be applied to experimental design for optimal results and data generation.

Dr. Paton, Dr. Tokach and Dr. Serão assert that the field of statistics needs to be treated as a tool that can vastly improve and optimize the analysis of data, but also needs respect to ensure that this analysis is appropriate for the experiment. Investigators, journal editors, and graduate trainees should engage with statistics as a discipline and be open to novel methodologies to push animal science forward as a discipline.

An unedited recording of this sympoisum can be found on the meeting website.