Bayesian modelling of lung function data from multiple-breath washout tests

Area of Science:

• Pediatric Pulmonology
• Biostatistics
• Computational Biology

Background:

• Multiple-breath washout (MBW) tests are crucial for assessing infant lung function, particularly in longitudinal studies.
• A significant number of infant MBW tests do not meet current acceptability criteria, leading to data loss.
• Traditional analysis methods for MBW data may not fully utilize available information.

Purpose of the Study:

• To develop and validate a novel statistical model for analyzing infant MBW data.
• To improve the efficiency and utility of MBW tests in pediatric respiratory research.
• To enable the estimation of the lung clearance index (LCI) from incomplete or shorter MBW tests.

Main Methods:

• Development of a novel Bayesian statistical model tailored for infant MBW data.
• Application of the model to a large dataset of 1197 MBW tests from 432 infants in a birth cohort.
• Focus on Bayesian estimation of the lung clearance index (LCI).
• Model checking using posterior predictive distributions.

Main Results:

• The developed Bayesian model demonstrated an excellent fit to the infant MBW data.
• The model provides insights into the statistical properties of standard empirical analysis methods.
• The novel approach allows LCI estimation from tests of varying completeness, unlike standard methods.
• Previously unusable data from incomplete or shorter tests can now be utilized without significant loss of precision.

Conclusions:

• A model-based approach offers a more efficient and comprehensive analysis of infant MBW data compared to traditional methods.
• The Bayesian model enhances data utilization by accommodating tests with different degrees of completeness.
• This methodology supports the routine use of shorter MBW tests, improving data acquisition in pediatric studies.
• The study highlights the practical benefits of Bayesian hierarchical modeling in analyzing repeated measures data in developmental research.