Traditionally, the diagnosis of geographic atrophy (GA) secondary to age-related macular degeneration (AMD) has been firmly linked to fundus autofluorescence (FAF) – and that is for a good reason. Large trials have established FAF as the imaging modality easiest to employ to diagnose GA and monitor its progression [1, 2]. Advantages of FAF especially lie in the possibility of a quick acquisition and also intuitive interpretation – “dead tissue” shows a clearly demarcated hypo-autofluorescence, healthy tissue shows iso-autofluorescence, and metabolically challenged tissue shows hyperautofluorescence [2]. Enlargement of GA lesions can be monitored by increases in the area of hypo-autofluorescence, and disease activity, i.e., individual enlargement rates, can be estimated by the interpretation of the hyper-autofluorescent patterns found around GA lesions [2].

Using FAF to discern “alive” from “dead” (outer) retina and to estimate retinal function from a certain locus’ autofluorescence on the en-face image however dramatically neglects the fact that autofluorescence is mostly a signal of the retinal pigment epithelium (RPE), given the presence of lipofuscin within it [2]. In fact, microperimetry reveals residual retinal function in areas covered by GA on FAF, providing light sensitivity which is well above expected values for absolute scotoma [3]. Thus, microperimetry could potentially aid in the monitoring of GA as a complementary “vertical measure” as compared to the horizontal assessment of enlargement using FAF.

In summary, our historic understanding of autofluorescence represents a “false black and white dichotomy”, while in reality there is a multitude of shades of grey. This fact is not only reflected in the rise of quantitative autofluorescence [4], but also the increased use of optical coherence tomography (OCT) in the assessment of GA [5].

This paradigm shift from a purely FAF-based to a multimodal, OCT-enhanced assessment of GA is not only reflected in imaging research, but also in clinical trials investigating pharmaceutical compounds to slow GA growth. The OAKS and DERBY phase III clinical trial program investigating pegcetacoplan, as well as the GATHER1/2 program investigating avacincaptad pegol, both used foveal centerpoint involvement on OCT to define “foveal” and “extrafoveal” lesions. In theory, one can assume that sparing the fovea from GA growth will result in better visual acuity outcomes. In reality, recent trials however failed to demonstrate an advantage in best corrected visual acuity (BCVA) in the treatment vs. the sham groups at two years, despite anatomical success.

This structure-function paradox has not only haunted these recent RCT programs, but has also become evident in past trials, most notably CHROMA and SPECTRI investigating lampalizumab. Post-hoc analyses of these trials have shown that the relationship between „saved retina” by slowing GA growth and functional benefits is more complex than previously thought. For example, large and multifocal GA lesions in CHROMA and SPECTRI showed faster growth than small and unifocal ones; however, it were the small and unifocal lesions which are associated with a faster and more devastating loss in visual acuity [6, 7].

Recent research into GA cluster analysis and quantitative autofluorescence might explain this paradox. Monés and Biarnés already described different GA phenotypes depending on their association with either soft drusen or reticular pseudodrusen (RPD)/=subretinal drusenoid deposits (SDD). Wei and colleagues recently confirmed this notion of distinct pathways into GA by quantitative autofluorescence [4]. As in the Monés and Biarnés study, pathway one was associated with soft drusen and pigment epithelium detachments (PED), whereas pathway two was associated with the presence of RPD/SDD [4], consistent with current research suggesting that soft drusen and RPD/SDD represent different mechanisms of disease within the spectrum of AMD [8]. Interestingly, both pathways into GA had different measures of quantitative autofluorescence, with a stronger loss of quantitative autofluorescence in GA associated with soft drusen as compared to GA associated with RPD/SDD [4].

Applying these findings to our current knowledge of the relationship between structure and function in GA, one might deduce that GA lesions associated with RPD might have a lesser impact on visual acuity as they usually begin extrafoveally, owing to the mostly extrafoveal location of RPD/SDD, and start off in multifocal small patches, i.e. the ones associated with less BCVA decline as found in CHROMA and SPECTRI. And, reading into the findings of quantitative autofluorescence, they reside more on the brighter grey scale within the shades of grey from black to white. So, judging by our knowledge in the year 2024, they might be a lesser threat to BCVA than originally thought, while they were initially assumed more dangerous due to their faster growth.

The question, however, is: How do we define growth? Historically, growth of GA was defined as the horizontal enlargement of the area hypo-autofluorescent on FAF. Based on our current knowledge, we should additionally pay as much attention to the vertical loss of retinal tissue, to retinal thinning, within the atrophic area. Figuratively speaking: Does the forest fire spread laterally from tree to tree, or does it scorch trees more into the ground, commencing in the treetop and continuing down the stem?

Therefore, new terminology might be of clinical and research use to transfer our understanding of GA enlargement from a merely two-dimensional approach based on FAF to a three-dimensional approach in combination with OCT imaging. When we talk about GA growth, we should discern horizontal growth, as measured on FAF, from vertical growth, as measured on OCT as loss of outer retinal tissue; or, as measured on quantitative autofluorescence, a shift from a brighter grey autofluorescence pattern into a darker, more black pattern, i.e. more pronounced loss of lipofuscin in absolute terms.

Most importantly, the distinction between horizontal and vertical GA might help explain our current lack of a structure-function correlation in the GA trials noted above. Just because GA involves the foveal centre point, not all light-sensitive tissue might be lost there; and some retinal areas covered by GA might be more “dead” than others.

CHROMA and SPECTRI suggested that, amongst all functional tests, only microperimetry showed an association with anatomical changes of GA at two years [7]. Specifically, these data demonstrated that change in GA lesion area most strongly correlated with the number of absolute scotomatous points [9]. In addition, Pfau and colleagues observed residual retinal function in areas covered by GA on FAF, providing light sensitivity well above expected values for absolute scotoma [3]. Most recently, a post-hoc analysis of data from the OAKS study, presented at the EURETINA 2023 by Usha Chakravarthy, suggests that an increase of absolute scotomatous points as measured by microperimetry might provide the best measure of a functional benefit of complement inhibition slowing GA growth [10].

Coining the term “foveal light insensitivity”, this post-hoc analysis defined an exploratory end point of “absolute foveal scotoma”, indicated by an absence of any light sensitivity in the central (“foveal”) 4 loci tested on microperimetry [10].

Interestingly, in spite of approximately two thirds of study eyes in OAKS having subfoveal involvement of GA on FAF [11], foveal light sensitivity in at least one of the 4 central loci was still present in 82% of studied eyes [10], strengthening the idea of a “horizontal” foveal involvement due to GA, but an incomplete “vertical” involvement. Given the idea that horizontal progression of GA in the fovea in those eyes was not possible anymore, in contrast vertical progression, indicated by loss of microperimetric light sensitivity in the fovea, was still a patient-relevant functional end point.

For patients and caregivers contemplating a potential therapeutic intervention, maintenance of foveal light sensitivity is of utmost importance. In contrast, the development of “foveal light insensitivity”, defined as the first occurrence of light insensitivity (−1 dB) at all 4 central stimuli, indicating absolute scotoma in these 4 central loci, can be used as exploratory end point. In the post-hoc analysis presented by Usha Chakravarthy, the development of “foveal light insensitivity” was investigated from OAKS to compare its incidence between sham and treatment arms. (As a side note, DERBY did not have microperimetry testing).

At baseline, 18.3% (n = 111) of patients already had foveal light insensitivity with 4 central scotomatous points; 81.7% (n = 493) patients still had foveal light sensitivity (≤ 3 central scotomatous points at baseline) and thus qualified for inclusion in the analysis. Excluding patients with no post-baseline microperimetry assessment, a total of 453 patients (PM, n = 155; PEOM, n = 152; sham pooled, n = 146) with foveal light sensitivity were included in a time-to-event analysis investigating the conversion of all 4 central points to scotoma.

Foveal light insensitivity occurred in 54 (34.8%), 59 (38.8%), and 64 (43.8%) patients in the PM, PEOM, and sham pooled arms, respectively. Interestingly, PM significantly reduced the hazard to develop foveal light insensitivity vs sham by 34% (HR vs sham: 0.66; 95% confidence interval [CI]: 0.46, 0.96; p = 0.0282). PEOM similarly reduced the hazard to develop foveal light insensitivity vs sham (HR vs sham: 0.64; 95% CI: 0.45, 0.92; p = 0.0164).

This post-hoc analysis indicates that pegcetacoplan treatment, both monthly and bimonthly, maintains foveal light sensitivity better than sham treatment. Thus, treatment might also be functionally relevant in eyes with foveal involvement on FAF. Foveal involvement is not equivalent to entirely dead foveal tissue, and tracing and slowing GA growth into the retinal (mid)periphery might therefore not be the ultimate goal in GA treatment, but preservation of foveal tissue, i.e. the slowing of vertical GA progression, might be of similar or even higher functional relevance.

Thus, a better understanding of the foveal degenerative processes reducing foveal light sensitivity, termed “vertical” GA as suggested here, should be a key point of investigation. Notably, the investigation of vertical GA is a much more complex task than the observation of FAF patterns, and might require a multimodal, and possibly an artificial intelligence (AI)-based approach. First, improvements in OCT analysis are needed, along with a uniform consensus on terminology, as seen in the Classification of Atrophy Meeting (CAM) program [5]. Also, specific OCT biomarkers indicating vertical progression have to be investigated and validated; these could include simple biomarkers like foveal retinal thickness, hypertransmission or degenerative fluid; or more advanced and precise biomarkers like the presence of the ELM (Fig. 1), the hyporeflective wedge-shaped band, or relative ellipsoid zone reflectivity as a novel innovative concept [12]. Second, AI assisted analyses providing a quantification of RPE and photoreceptor cell count (and possible protection under therapy) will help to zoom in from the bigger picture into a nearly cellular level [13]. Third, quantitative FAF might help to distinguish the multiple shades of grey indicating different levels of tissue loss in areas of hypo-autofluorescence. And fourth, AI algorithms might ultimately help to predict retinal sensitivity from these “shades of grey” in FAF – so called “inferred retinal sensitivity” [14] – in addition to microperimetric testing.

Fig. 1: Example of vertical and horizontal progression of GA in one eye.
figure 1

Panels A and C demonstrate the horizontal progression traditionally measured on autofluorescence. The corresponding OCT scans (B, D), obtained on the same day as the autofluorescence in A and C demonstrate vertical progression of GA. Within 8 months, horizontal progression is seen by an increase of GA area from 13.59 to 13.95 mm2, probably affecting visual function only in a limited manner. In contrast, vertical progression on OCT shows retinal thinning from 225 to 195 µm, and, even more importantly, loss of residual foveal external limiting membrane, probably affecting visual function more than horizontal growth of GA as measured on autofluorescence.

Full size image

In conclusion, complement inhibition provides an exciting opportunity to study retinal degeneration in GA in AMD. Arriving from an era dictated by two-dimensional FAF, we should now take the opportunity and start thinking – and analysing our imaging data – in three dimensions. The concept of horizontal (FAF) and vertical (OCT, quantitative FAF, microperimetry) GA might help us to further our understanding of the complex structure-function relationship in GA.