Cox regression survival analysis with compositional covariates: Application to modelling mortality risk from 24-h...

Area of Science:

• Biostatistics
• Epidemiology
• Public Health

Background:

• Survival analysis, particularly Cox regression, is vital for assessing mortality risk associated with exposures.
• Standard Cox models struggle with compositional covariates (e.g., physical activity, sitting, sleep) due to violated assumptions.
• Compositional data's intrinsic interdependencies require specialized analytical approaches.

Purpose of the Study:

• To develop a Cox regression model formulation that appropriately handles compositional covariates.
• To enable scientifically meaningful interpretations and standard statistical inference for compositional exposure data.
• To apply this novel method to analyze mortality hazards linked to daily activity composition.

Main Methods:

• Introduced a Cox regression model formulation using log-ratio coordinates to address compositional data constraints.
• Applied the model to analyze mortality risk associated with the composition of physical activity, sitting time, and sleep.
• Utilized data from the U.S. National Health and Nutrition Examination Survey (NHANES).

Main Results:

• The log-ratio coordinate formulation effectively accommodates compositional covariates in Cox regression.
• The method facilitates standard statistical inference and yields interpretable results for complex exposures.
• Demonstrated practical application in a public health context using NHANES data.

Conclusions:

• The proposed Cox regression formulation offers a robust solution for analyzing survival data with compositional covariates.
• This approach enhances the accuracy and interpretability of mortality risk assessments in public health.
• Log-ratio coordinates provide a powerful tool for understanding the impact of behavioral compositions on health outcomes.