Collective endpoint of visual acuity and contrast sensitivity function from hierarchical Bayesian joint modeling

Area of Science:

• Ophthalmology
• Biostatistics
• Clinical Trials

Background:

• Clinical trials often analyze multiple endpoints to detect treatment efficacy.
• High-dimensional data in trials can obscure subtle treatment effects.

Purpose of the Study:

• To develop a hierarchical Bayesian joint model (HBJM) for improved signal detection in clinical trials.
• To compute a novel five-dimensional collective endpoint (CE5D) integrating contrast sensitivity function (CSF) and visual acuity (VA).

Main Methods:

• Developed a hierarchical Bayesian joint model (HBJM) to analyze CSF and VA data hierarchically (population, individual, test).
• Generated joint posterior distributions for CE5D, incorporating CSF parameters (peak gain, peak frequency, bandwidth) and VA parameters (threshold, range).
• Applied the HBJM to a dataset of 14 eyes tested under four Bangerter foil conditions using quantitative VA and CSF procedures.

Main Results:

• The HBJM revealed strong correlations among CE5D components across all hierarchical levels.
• Variance of estimated components was reduced by an average of 72% with 15 qVA and 25 qCSF rows.
• CE5D demonstrated significantly higher sensitivity and accuracy in discriminating performance differences between foil conditions compared to original tests.

Conclusions:

• The HBJM effectively extracts information on CSF and VA parameter covariance, enhancing parameter precision.
• This framework increases statistical power for detecting vision changes by combining signals and reducing noise from multiple tests.
• The HBJM shows potential for improving statistical power in combining multi-modality data within ophthalmic clinical trials.