Reply




We thank Mahroo and associates for their interest in our recent article, “Choice of Analytic Approach for Eye-Specific Outcomes: One Eye or Two?” We agree that adjusting for lack of independence may be necessary when there are multiple measurements from individual eyes, because measurements obtained within an eye tend to be correlated. The approach taken in analyzing such data, however, will depend on the research question. If the research interest lies at the level of the eye, then a per-eye summary is needed and no within-eye correlation adjustment is required. This would be the case if, for example, the outcome variable represents the presence or absence of at least 1 retinal break in the eye. If, however, the interest lies in describing features of a characteristic that may occur at multiple locations within each eye (such as a benign ocular tumor), then it is likely that a within-eye correlation adjustment will be needed, because each repeated observation in an eye may provide less information than an observation in a different eye. Indeed, if the correlation is close to 1, then obtaining multiple measurements adds very little information.


Our article considered analytic approaches appropriate when measurements were available from both eyes of a subject at a specific time point. In this case, and when multiple measurements are available from each eye at a single time point, ignoring the correlation may result in standard errors and P values that are too small. If, however, the study design involves obtaining repeated within-eye measurements over time and the interest lies in assessing changes over time, ignoring the correlation between measurements is likely to waste information and result in standard errors that are too large.

Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

Jan 12, 2017 | Posted by in OPHTHALMOLOGY | Comments Off on Reply

Full access? Get Clinical Tree

Get Clinical Tree app for offline access