Abstract
Murine experiments are powerful tools for studying the effects of chronic medication regimens on aging. However, analytical complexity often induces researchers to simplify experimental design, ignore potential covariates, and employ common, but less powerful, statistical approaches to make many independent comparisons. We present an approach to longitudinally modelling treatment effects accounting for within mouse correlation, adjusting for covariates fixed in time (treatment and cohort) and those that vary over time (age, weight, and number per cage). In this study, nine treatment groups of 30 mice initiated treatment at 12 months of age and continued until 24 months with measures every three months. We chose a line-crossing metric calculated during spontaneous activity in the open field as an exemplar mobility outcome. We demonstrate the statistical methods in a step-by-step manner on the use this state-or-the-art statistical modeling by showing covariance fitting, coding for time fixed and varying covariates, statistical contrasts and graphical output. Preplanned comparisons between treatments were performed with contrasts. Final model selection was based on measures of goodness-of-fit. We found the best fit was with a first-order autoregression covariance, and that line crossing declined with advancing age (p<.001), increasing weight (p<.001), and there was a cohort effect (p<.001). A significant treatment by age interaction (p<.01) indicated that most treatments accelerated the age-related decline in performance. Several, pre-planned contrasts showed significant differences, especially those between comparisons of polypharmacy with high Drug Burden Index (cumulative anticholinergic and sedative drug exposure) and single drug treatments