cosinoRmixedeffects: an R package for mixed-effects cosinor models

Area of Science:

• Chronobiology
• Biostatistics
• Wearable Technology

Background:

• Wearable devices monitor physiological parameters, some with circadian rhythms.
• Existing R packages like cosinor have limitations in analyzing longitudinal data and complex group comparisons.
• A need exists for advanced models to evaluate longitudinal changes in circadian patterns.

Purpose of the Study:

• To develop an R package for analyzing longitudinal periodic data using mixed-effects cosinor models.
• To extend the cosinor model to accommodate random effects and interactions with time-varying covariates.
• To facilitate ease of use and hypothesis testing for circadian parameter analysis.

Main Methods:

• Developed the cosinoRmixedeffects R package implementing mixed-effects cosinor models.
• Integrated syntax and functions from the emmeans package for estimated marginal means and contrasts.
• Employed bootstrapping for estimation and hypothesis testing of non-linear circadian parameters (MESOR, amplitude, acrophase).

Main Results:

• The cosinoRmixedeffects package enables model fitting, estimation, and hypothesis testing for mixed-effects cosinor models.
• The model supports linear and non-linear circadian parameters, including MESOR, amplitude, and acrophase.
• Demonstrated functionality by analyzing longitudinal heart rate variability (HRV) data, assessing differences between genders, BMI, and during SARS-CoV2 infection.

Conclusions:

• The cosinoRmixedeffects package offers a robust framework for analyzing longitudinal circadian data.
• It accommodates factors with multiple categories and complex interactions with circadian parameters.
• This tool enhances the analysis of wearable sensor data for understanding circadian rhythm dynamics.