Bayesian modeling of embryonic growth using latent variables.

Area of Science:

• Biostatistics
• Developmental Biology
• Reproductive Science

Background:

• Individuals progress through developmental states, with transition rates influenced by covariates.
• Estimating these transition rates is crucial for understanding growth dynamics.
• Cross-sectional data with unknown initiation times present challenges for growth modeling.

Purpose of the Study:

• To develop a Bayesian discrete-time multistate growth model for analyzing cross-sectional data.
• To link developmental progress covariates to an underlying latent growth variable.
• To investigate the association between latent growth and the probability of future events, specifically pregnancy loss.

Main Methods:

• Developed a Bayesian discrete-time multistate growth model.
• Incorporated covariates measuring developmental progress linked to a latent growth variable.
• Utilized a Markov chain Monte Carlo (MCMC) algorithm for posterior computation.

Main Results:

• Found evidence supporting a previously hypothesized association between slow early pregnancy growth and increased risk of spontaneous abortion.
• The model successfully linked developmental progress and latent growth to state transition rates.
• Demonstrated the utility of the model in a study of embryonic development and pregnancy loss.

Conclusions:

• Slow latent growth early in pregnancy is associated with a higher risk of spontaneous abortion.
• The developed Bayesian growth model provides a robust framework for analyzing complex developmental processes from cross-sectional data.
• This research offers new insights into the biological mechanisms underlying pregnancy loss.