Retina Society 2026: How blood biomarkers may predict GA speed

Geographic atrophy (GA) progression has traditionally been tracked through a single snapshot of disease activity, but new research suggests systemic complement levels—measured over time—may carry prognostic weight of their own. Talisa Forest (de Carlo), MD, assistant professor, medical director of imaging, and co-director of the AMD Registry in the Vitreoretinal Surgery service at the Sue Anschutz-Rodgers Eye Center, University of Colorado Anschutz, presented findings correlating dynamic systemic complement factors with artificial intelligence (AI)-assisted GA growth rate at the 59th Annual Scientific Meeting of The Retina Society, September 23-26, 2026, in Los Angeles, California.

In this Q&A conversation with Modern Retina, Forest shares the biomarker patterns that emerged, the multimodal imaging pipeline behind the AI-assisted growth measurement, and what would need to happen before this work could inform clinical decisions.

Note: Transcript edited lightly for clarity and length. The authors’ manuscript on this work is under revision at IOVS.

MR: What was the key relationship you found between dynamic systemic complement factors and AI-assisted GA growth rate? Are certain markers tracking with faster progression?

Talisa Forest (de Carlo), MD: Given the exploratory and comparative nature of this analysis, we focused on the pattern of associations across biomarkers, including consistency in direction and relative strength, rather than on whether individual correlations met conventional thresholds for statistical significance. Complement ratios C3b/C3 and C4b/C4 had posterior distributions centered near zero, and thus limited evidence for association. Factor I (posterior mean = -0.076, 95% Credible Interval (95% CrI): -0.347 to 0.198) and C3a/C3 (-0.138, 95% CrI: -0.439 to 0.182) had posterior distributions centered below zero, suggesting possible negative associations with SQRT GA growth rate. Factor H (0.092, 95% CrI: -0.195 to 0.373) and TP ratio sC5b-9/C5 (0.094, 95% Crl: -0.190 to 0.373) had posterior distributions centered above zero, suggesting potential positive associations with SQRT GA growth rate.

The alternative pathway ratios Ba/Factor B (0.350, 95% CrI: 0.021 to 0.644), and Bb/Factor B (0.203, 95% CrI: -0.086 to 0.475) demonstrated the largest magnitude correlations with posterior distributions largely above zero, consistent with a positive association with SQRT GA growth rate.

Therefore, subjects who tended to have faster GA growth rate also tend to have higher Ba/Factor B and Bb/Factor B levels over time.

MR: You specifically studied dynamic systemic complement factors rather than a single baseline measurement. Why did looking at how these factors change over time matter more than a one-time snapshot, and how frequently were you measuring them?

Talisa Forest (de Carlo), MD: We wanted to look at complement levels longitudinally instead of cross-sectionally to specifically look at the relationship between the trends of these biomarkers and GA growth. AMD is a chronic, progressive disease that develops over years. A one-time measurement provides only a snapshot and may be influenced by transient biological variability, including acute inflammatory responses, illness, or other short-term stressors in complement activity.

In contrast, repeated measurements allow us to assess whether complement activation is persistently elevated over time. Sustained complement dysregulation is likely more biologically relevant to AMD pathogenesis than a transient increase and may therefore provide a more robust indicator of the chronic inflammatory processes underlying disease progression.

Further, it is possible that complement activity may be more or less active during different stages of the GA progression (such as in earlier GA or once the disease has slowed down). Our group wants to understand how these potential dynamics are associated with the disease state.

MR: How is the AI-assisted growth rate measurement working here? What imaging modality is it built on, and how does it improve on the way GA progression has traditionally been measured and tracked?

Talisa Forest (de Carlo), MD: We use a segmentation algorithm trained on FAF images. We also developed a method to spatially-register IR/SLO and OCT images with the FAF. Then we had a VR surgeon review each segmentation (along with the co-registered IR and OCT as adjuvant guides) and adjust as needed. The improvement in traditional measurements is thus more related to the multimodal integration as each modality alone has their drawbacks for accurate detection of the GA borders.

MR: If systemic complement levels do correlate with GA growth rate, could this become a practical biomarker — something used to identify patients likely to progress faster, or to monitor response in patients on complement inhibitor therapy?

Talisa Forest (de Carlo), MD: First, I think that this work, while promising and exciting, is still exploratory and needs further investigation and external validation before being adapted to clinical use.

In the future, we hope that this work will help to uncover biomarker profiles (where it is a single test or a more comprehensive endotype) that help us to understand prognosis and who are the best treatment candidates. Along these lines, we also hope that this work can help support investigations into future complement therapeutics in GA.

MR: What is the next step for this research? Are you looking to validate this in a larger or prospective cohort, or explore whether modifying systemic complement activity actually changes the growth trajectory?

Talisa Forest (de Carlo), MD: There are many works in progress in the University of Colorado AMD Registry. We are expanding recruitment into our GA cohort so that we can study this specific study further. With larger sample sizes we hope to investigate joint models that incorporating longitudinal GA growth trajectories (including temporospatial growth modeling metrics), baseline characteristics (GA size, focality, foveal involvement) and systemic complement measures. This can help us to understand if complement may be more active during certain phases of the disease.

Ultimately, external validation would be important but may be difficult as there are rare other AMD registries in which they are prospectively collecting longitudinal blood samples. We are also separately, evaluating if anti-complement therapeutics change the growth trajectory of our GA patients using our AI team’s Gompertz modeled metrics. However, more patient data is needed given the somewhat limited sample size of patients on these therapeutics at our single hospital system.

Talisa Forest (de Carlo), MD

E: [email protected]

Forest (de Carlo) is assistant professor, medical director of imaging, and co-director of the AMD Registry in the Vitreoretinal Surgery service, Ophthalmology Department, at the Sue Anschutz-Rodgers Eye Center, University of Colorado Anschutz, in Aurora, Colorado.