Café com Física

Double-organ Bias in Vision Research: Insights from Age Related Macular Degeneration (AMD) in a Portuguese Cohort

by Rita Coimbra (AIBILI - Association for Innovation and Biomedical Research on Light and Image, Coimbra, Portugal)

Portugal
Sala de Conferências (Departamento de Física FCTUC)

Sala de Conferências

Departamento de Física FCTUC

Universidade de Coimbra
Description

In ophtalmological clinical trials, data can be collected from either one or both eyes of a subject. Most of the statistical tests used assume that observations in a sample are independent, but measurements obtained from right and left eyes are usually correlated (double-organ bias).
Depending on the study hypothesis and clinical relevance, data can be selected from one eye or both eyes. We review the systematic selection process of just one eye, when subject is the unit analysis (one eye per individual) and the statistical methodologies to address the problem using eye as unit analysis (two eyes per individual). When data from both eyes are available, rejecting data from the fellow eye, reduces the potential power of the study, increases the number of subjects to be recruited and rejects valid data that will not be analysed. Even in high-impact-factor journals, statistical analysis are performed with only one, reducing the power of the study or with both eyes but without take into account the correlation between eyes. So this problematic should be clarified and addressed in ophthalmology.
We overview the various methodologies used in clinical trials at AIBILI (Association for Innovation and Biomedical Research on Light and Image), a research technology organization dedicated to the development and clinical research of health technologies. We focus on the Coimbra Eye Study (CES), an epidemiologic population-based study on the prevalence and incidence of Age-related Macular Degeneration (AMD) in a Portuguese population (NCT01298674, NCT02748824). We will present an example using a logistic regression model with Generalized estimated equation (GEE) to account for inter-eye correlations.

Organised by

Paulo Brás, Paulo Silva, Jaime Silva