Analysis of Correlated Two Eyes Data for Ophthalmic Clinical Research

Abstract

Objective: To describe the proper statistical approach in ophthalmic research for analyzing correlated data for contralateral eyes of the same subjects. Methods: For data from a paired design and a design for two eyes that are commonly used in ophthalmic research, we described the appropriate statistical approaches for analyzing the correlated data for both eyes of the same subjects. As an example, we compared the intraocular pressure between patients < 60 vs. ≥60 years of age, using the inappropriate and appropriate statistical analysis approaches. Results: The inappropriate analysis of data for one eye leaded only to a biased or inefficient estimate of the difference between two groups. The analysis of data for both eyes of the same subjects without accounting for inter-eye correlation had a smaller p-value (P=0.02) than the analysis that accounted for inter-eye correlation (P=0.098). Conclusions: Appropriate statistical analytical approaches should be used to account for the inter-eye correlation. The statistical analysis using data for only one eye or ignoring the inter-eye correlation can lead to an incorrect conclusion. Key words: inter-eye correlation; statistical analysis; ophthalmic data; paired data.