Pair-feeding designs are commonly used in experimental animal research to equalize food intake across groups with the main goal of isolating treatment effects from the effect of food intake on the outcome of interest. Typically, researchers implicitly assume that food intake is comparable between groups in such designs, and it is often not accounted for in the statistical analysis. However, not accounting for food intake raises the question of whether pair-feeding designs inadvertently may sometimes increase the risk of false positive results.

A recent simulation study by Najam et al., published in Obesity (“Pair-Feeding Study Designs Can Create Biases and Inflate Type I Error Rates: A Simulation Study”), explored this very question. The study found that both individual and group pair-feeding designs could substantially inflated false positive (type I error) rates (from 12 to as high as 71%) in models that did not adjust for food intake, across various sample sizes and study durations. This inflation might be attributed to the induced differences in food intake distributions between pair-fed and non-pair-fed groups. Importantly, the study suggested that adjusting for food intake reduced false positive (type I error) rates back to the expected 5% in many of the cases studied. These methods were illustrated and the findings placed in context by reanalysis of data from a previously published pair-feeding study involving pigs subjected to heat stress and yeast supplementation. In that case, adjusting for food intake as a covariate in the statistical analysis changed the originally reported results .

This work highlights a previously unrecognized oversight in pair-feeding studies. Assuming equivalent food intake without statistical adjustment can bias treatment effect estimates and lead to excessive false positive findings under some plausible circumstances. Najam et al.’s work suggests that food intake be measured and included as a covariate in the statistical analysis. By quantifying this potential bias through simulations, Najam et al. provide researchers with valuable guidance for enhancing rigor in studies utilizing pair-feeding designs and ensure more accurate isolation of intervention effects independent of food intake.