Highlights
The correlation in geographic atrophy (GA) enlargement rates between fellow eyes was only modest, with Pearson coefficients ranging from 0.11 to 0.51 depending on the transformation method used. GA lesion size substantially influenced the correlation strength, with larger lesions showing weaker correlations compared to smaller lesions. Researchers should exercise caution when designing trials or interpreting analyses that assume highly correlated GA progression between eyes.
Background: The Clinical Context of Geographic Atrophy
Geographic atrophy represents an advanced form of age-related macular degeneration (AMD), characterized by progressive degeneration of the retinal pigment epithelium, photoreceptors, and choriocapillaris. This condition represents a leading cause of irreversible vision loss among elderly populations in developed nations, with prevalence projected to increase substantially as demographic shifts favor older age groups.
In patients with bilateral GA, clinicians and researchers have traditionally assumed that enlargement rates between fellow eyes would be highly correlated, reflecting shared systemic factors, genetic predisposition, and environmental exposures. This assumption has underpinned the design of numerous clinical trials investigating potential treatments for GA, with some studies proposing to use the untreated fellow eye as an internal control to reduce sample sizes and improve statistical efficiency. Understanding whether this assumption holds true carries significant implications for both trial methodology and our fundamental understanding of GA pathophysiology.
Study Design and Population
This investigation comprised a post hoc analysis of data from the Age-Related Eye Disease Study 2 (AREDS2), a multicenter study conducted across retinal specialty clinics throughout the United States. The analysis included participants from the AREDS2 trial who demonstrated bilateral GA at study enrollment.
The primary exposure under investigation was the presence of GA in the fellow eye. The main outcome measure was the Pearson correlation coefficient for 2-year GA enlargement rates in eye pairs. GA enlargement rates were derived from planimetry of annual fundus photographs and expressed in three different formats: untransformed rates, square root-transformed rates, and perimeter-adjusted rates. The analysis was performed examining correlations both overall and within clinically relevant strata, including baseline GA area, GA location, focality, and reticular pseudodrusen (RPD) status.
Study data were analyzed from May 2025 through January 2026, ensuring appropriate temporal alignment with the analytical procedures.
Key Findings: Variable and Often Weak Correlations
The analysis included 386 eyes from 193 AREDS2 participants with bilateral GA. The cohort had a mean age of 75.5 years with a standard deviation of 7.3 years, and 61.1% of participants were female (118 individuals).
The principal finding demonstrated that correlations in GA enlargement rates varied substantially depending on the rate transformation method employed. When using untransformed rates, the correlation coefficient was moderate at 0.51 (95% CI, 0.41-0.61). However, this correlation weakened considerably with alternative transformations: square root-transformed rates yielded a correlation of 0.38 (95% CI, 0.25-0.49), representing a weak to moderate relationship, while perimeter-adjusted rates produced only a very weak correlation of 0.11 (95% CI, -0.03 to 0.25).
Influence of Baseline GA Area
The analysis revealed a striking pattern regarding the influence of baseline GA area on correlation strength. For untransformed rates, eyes with small GA lesions demonstrated a correlation of 0.63 (95% CI, 0.41-0.78), while eyes with large GA lesions showed markedly weaker correlation at 0.30 (95% CI, 0.07-0.50). This pattern persisted across transformation methods, though with reduced magnitude differences. For perimeter-adjusted rates specifically, the correlation was essentially negligible in large GA eyes at 0.03 (95% CI, -0.21 to 0.26), compared to 0.22 (-0.08 to 0.49) in small GA eyes.
Impact of GA Location
When examining GA location, correlations using untransformed rates were strongest for extrafoveal GA, intermediate for subfoveal GA, and weakest for discordant pairs (where one eye had GA at a different location than the fellow eye). The square root-transformed rates demonstrated a similar pattern of correlation strength by location.
Effects of Focality and RPD Status
Analysis of focality revealed that correlations using untransformed rates were similar across unifocal, multifocal, and discordant pairs. However, when square root-transformed rates were employed, the correlation was strongest for unifocal GA, weakest for multifocal GA, and intermediate for discordant pairs.
Regarding RPD status, correlation was strongest in the absence of RPD, weakest when RPD was present, and intermediate for discordant pairs. These differences were smaller when square root transformation was applied.
Expert Commentary: Implications for Research and Clinical Practice
The findings from this analysis carry substantial implications for the field of GA therapeutics and clinical trial design. The assumption that bilateral GA progresses in a highly correlated manner appears to be largely unfounded when examined rigorously.
The discrepancy between correlation coefficients across different transformation methods deserves careful interpretation. The untransformed rates, which demonstrated the highest correlation (0.51), likely reflect an artifact of baseline symmetry in GA characteristics rather than true biological correlation in progressive enlargement. When more sophisticated measures accounting for lesion geometry were employed—particularly perimeter-adjusted rates—the correlation diminished dramatically to 0.11, suggesting that genuine linear enlargement rates show only minimal correspondence between fellow eyes.
These findings challenge the validity of trial designs that rely on untreated fellow eyes as internal controls, as the expected high correlation does not reliably exist. Researchers planning interventional studies for GA should consider this variability when calculating sample size requirements and should avoid assuming that fellow-eye data can substitute for independent control populations.
The differential correlation by GA area size may reflect biological differences in the underlying disease processes driving progression in early versus advanced stages of atrophy. Larger GA lesions may be subject to greater environmental or stochastic influences, while smaller lesions might be more tightly regulated by systemic genetic factors.
Conclusion: A Call for Revised Assumptions
This post hoc analysis of the AREDS2 cohort demonstrates that correlation in GA enlargement rates between fellow eyes is modest at best and highly variable depending on methodological approach and patient characteristics. The strength of correlation is influenced by GA area, location, focality, and RPD status, with larger lesions, discordant anatomical features, and presence of RPD all associated with weaker correlations.
These findings underscore the importance of methodological rigor in GA research and caution against assumptions of high inter-eye correlation. Clinical trials investigating GA treatments should not rely on untreated fellow eyes as primary controls without careful consideration of these sources of variability. Future research should explore the biological mechanisms underlying this variability and develop improved methods for accounting for inter-eye differences in disease progression.
The modest correlations observed in this study also suggest that GA progression, even within the same individual, may be influenced by local factors specific to each eye rather than solely by systemic determinants. This insight opens avenues for investigating microenvironmental factors that may drive differential progression rates between eyes.
Funding and Trial Registration
This analysis was conducted as part of the AREDS2 Research Group. Full funding disclosure and conflict of interest information are available in the original publication.
References
von der Emde L, Vance E, Mukherjee S, Hou J, Agrón E, Siddiq F, Domalpally A, Chakravarthy U, Chew EY, Keenan TDL, AREDS2 Research Group. Modest and Variable Correlations Between Geographic Atrophy Enlargement Rates in Fellow Eyes in the AREDS2 Study. JAMA ophthalmology. 2026-04-02. PMID: 41926108.
