Findings of the study and dispute over AMN pathogenesis

In our study, we only found significant perfusion deficits in the choriocapillaris and choroid layer, but no changes in the blood perfusion of the retinal layers. Since this contradicts earlier findings which suspect a DCP origin, further insight into the different plexuses and their potential role in AMN is needed.

Although the precise anatomy of retinal capillary plexuses and their watersheds are under debate, the DCP is generally considered to supply the inner nuclear layer (INL) and outer plexiform layer (OPL). Any DCP circulation deficit, therefore, should rather affect these inner retinal layers as seen in PAMM, and indeed, this is the suspected pathomechanism for PAMM [24, 25]. Given this, it remains unclear how a perfusion deficit in the DCP should be able to cause both PAMM and AMN, two entities that not only affect different retinal layers, but also have different epidemiologies [26].

The hallmark of AMN type 2, in contrast, is less alterations in the inner retinal layers, but rather a disruption of the ellipsoid zone. Adaptive optics investigations have confirmed that structural alterations happen in the photoreceptor layer [27, 28], and so does our en face OCT analysis. Because the photoreceptors are nourished by the underlying choriocapillaris, no microvascular perfusion in the retina should cause these changes. If the underlying cause was a retinal perfusion deficit in the DCP, then the photoreceptor alterations could only be explained by retrograde transneuronal degeneration, i.e., propagation of an inner retinal defect onto the photoreceptors. However, no such behavior can be seen in retinal artery occlusion, and also, the early onset of alterations within hours is implausible for this theory.

Lately, a shift in reporting has occurred. A focal perfusion deficit in the choroidal layers could in theory much better explain the isolated alterations on the photoreceptor layer in AMN. As Bottin [29] and Munk [30] pointed out, the changes that are often attributed to the outer plexiform layer (i.e., domain of the DCP blood supply) are much rather swellings of the Henle fiber layer (HFL). The Henle fibers are the axons of the photoreceptors that connect the photoreceptor cells to the outer plexiform layer and are thus dependent on choroid blood flow. In AMN occurrences near the foveal depression, the sidewards tilt of these fibers can be visualized, as in Fig. 4C (yellow arrowheads). Thanos et al. [31] were the first to report on choroidal perfusion deficits that could be observed in OCTA analysis of AMN cases. This finding was later backed by other investigators [32, 33], although these findings did not go undisputed [9]. This theory much better explains the structural alterations that are mainly in the photoreceptors. Duan et al. proposed the double watershed theory, which presumes a concurrent relative perfusion deficit both from the DCP and the choroidal vasculature [34] that ultimately results in AMN lesions.

Interpretation of contradicting results from different studies

The contradicting findings of the authors suspecting a choroidal perfusion deficit versus the authors that have found DCP perfusion deficits within the same imaging modality (OCTA) [3, 8,9,10,11,12,13] remain at least partially unexplainable. One major factor, however, could be that on top of different OCTA devices with different wavelengths and image output, no standardized measurements of OCTA perfusion exist. Probably, the most important issue is the non-availability of standardized OCTA quantification software. Even though different authors have used similar principles, most of the used OCTA analysis software remains unavailable to other research groups, which hinders reproduction studies. We therefore chose to release our quantification software open source, so that future authors can use the same methodology for reproduction studies or other scientific questions.

Another factor is that different authors have used different controls (from surrounding area to unaffected partner eye to healthy controls). We specifically chose our control parameters within the same eye of the same patient to make the findings less prone to intra- and interindividual alterations of OCTA. VAD is known to vary not only in diseased but also healthy subjects [35], which limits the validity of healthy control groups. Furthermore, the influence of age, sex, axial length, etc. is not sufficiently established to allow for proper selection criteria of a control group [36]. Moreover, since the AMN lesion typically only involves a small fraction of the OCTA scan, it is unclear if the whole scan area or just the area of the AMN lesion should be compared, and if the latter, it remains unclear of how to properly transferring this region onto other eyes with different foveal and vascular architecture. Also for intraindividual comparison with the partner eye, this issue persists. Furthermore, we found no conclusive literature on the correlation of VAD in-between both eyes of the same subject; especially, we found no evidence that VAD should strongly correlate between left and right eye in general and for specific regions within the scan in special.

We therefore used a comparator within the same eye. Since the retinal vascular architecture is designed centripetally towards the fovea, our control parameters ATC and RS were chosen so that they should be mostly homogenous within the same eye. However, while we found significant results for the directly adjacent tissue control, the comparison to the ring segment nearly missed the significance threshold. Previous studies have shown that VAD measurements between different quadrants of the same ring segment can alter substantially [35], which might explain higher fluctuations in ring segment VAD and therefore might make reaching significance threshold more complicated. Nonetheless, also this comparator showed a clear trend towards a reduced choroidal perfusion in AMN.

A third aspect explaining different findings could be that the DCP quantifications are erroneous due to OCT alterations by the AMN. Even though we have argued above that we attribute the alterations rather to the HFL than to the OPL, proper OPL alterations have been reported by many authors; if such exist, they might cause light scattering and reduced OCTA signal. More likely, HFL hyperreflectivity might cause erroneous segmentation of the OPL slab, which consequently could alter VAD measurements especially of the DCP. Even though we acknowledge that the VAD measurements in the choroidal layers might be equally disputable (see next subchapter), this might explain the different findings.

A last consideration regarding the contradicting findings could be the mostly small sample size of the studies (in the studies supporting a DCP origin, only 1 larger study with 16 eyes [12], the rest smaller studies with 3 [11], 2 [10], or 1 eye [3, 8, 13]). All studies supporting a choriocapillary origin of AMN also were based on smaller sample sizes (3 eyes in both [31, 33], 9 eyes in [32]). Moreover, many studies only used descriptive terms that reported of hypoperfusion without proper quantification. Interestingly, all previous studies reporting on a choriocapillary perfusion deficit did not perform proper VAD quantification or statistical analysis. Our work thus is the first to quantify and statistically undermine the theory of a choriocapillary hypoperfusion origin of AMN.

Susceptibility of choroidal OCTA measurements to artifacts

A non-negotiable factor that limits the findings of the advocates of a choroidal perfusion deficit is the limited usability and accuracy of OCTA for imaging the very. This has been a major factor of dispute over the validity of these findings [9]. A study quantifying choriocapillaris vasculature in case of reticular pseudodrusen found that both larger retinal vessels and pseudodrusen would cause projection artifacts and could alter choriocapillaris quantification measurements [37]. Similarly, AMN-related photoreceptor changes could cause erroneous measurements similar to pseudodrusen. In our study, we found no significant differences in OCTA quantification of the choroidal layers whether areas of large retinal vessels (as identified by the frangi filter in SVP) were excluded in OCTA analysis or not (data not shown). However, the alterations in the photoreceptor layer could be very well argued to cause light scattering and diminished signal from underlying layers in OCTA.

The single B-scans, however, do not show any signs of reduced signal of the choroidal layers underlying the photoreceptor layers (see for example Fig. 4C). In theory, larger vessels can cast a shadow on the underlying retinal layers, whereas RPE atrophy is known to cause relative hyperreflectivity of the underlying choroid (because of missing blockade). In our patients, we found no visible artifact-related fluctuations in choroidal structure in regular OCT. If, however, OCTA might be more susceptible than structural OCT for light scattering artifacts is a question that cannot conclusively be debunked. We do, however, argue that the validity of the reported DCP underperfusion might equally be challenged (see above).

Interpretation of en face OCT findings

The en face OCT analysis conducted in this study shows that the IR hyporeflectivity correlates better to OCT alterations in the photoreceptor layer rather than the OPL, a finding that has not been reported before. One argument for a perfusion deficit in the DCP is that such would cause swelling of the OPL and thus hyperreflectivity in IR, similar to what can be seen in retinal artery occlusion. Our findings contradict this notion by clearly showing that the IR hyporeflectivity is caused by photoreceptor alterations instead. We believe this to further weaken the DCP perfusion deficit thesis.

Limitations and strengths of this study

Even though we believe to have put substantial weight on the conjecture of a choroidal perfusion deficit origin of AMN, we agree that also our evidence is not substantial enough to close the case. However, when disregarding the contradictory findings whether DCP or CC perfusion is reduced, we find that the pure theory behind the choroidal origin sounds more plausible than a DCP origin: Why should mainly the photoreceptors be altered in case of a DCP origin, when the photoreceptors are not perfused by it, and why are these alterations not present in the much more severe retinal artery occlusion?

One limitation of this study is the debatable quality of the choroidal OCTA measurements. Another limitation is that we solely discuss perfusion deficits as possible origins of AMN and disregard other thinkable pathomechanisms, e.g., autoimmune processes or other types of inflammation. Also, we only included patients with a SARS-CoV-2 infection into the study. As discussed earlier, SARS-CoV-2 has often been shown to cause AMN [7]; however, we would argue that this is similar to the previously described other viral infections causing AMN. We therefore do not believe that SARS-CoV-2-related AMN is fundamentally different from other AMN, and we found no literature that would indicate otherwise. Thus, we presume that the study’s findings might also be generalized to non-SARS-CoV-2 AMN cases. However, our data cannot conclusively prove this assumption. Likewise, any impact of oral contraceptives on the results of the study cannot be ruled out.

A strength of our study is the unbiased inclusion of all patients in a consecutive case series with proper statistical and not only descriptive analysis. Also, we believe that our approach of releasing our software open source will allow for reproduction studies and/or larger multicenter validation studies.