Discussion
Gass’ classification system for BVMD, which was based on fundoscopic findings and developed decades prior to OCT, has been widely used to describe macular lesions in BVMD. However, despite its widespread adoption, the subjectivity of Gass staging makes the scheme difficult to employ in many cases. By contrast, OCT-based evaluation now enables a more detailed, cross-sectional view of structural biomarkers that are relatively easy to identify and may provide better prognostic information. In our study, grader agreement was excellent for OCT-based classification of macular lesions, and OCT-based categorisation was far more consistent than assigning Gass lesion type (90% overall grader agreement for OCT features vs 46% for Gass lesion type). OCT-based features also provide information that correlates closely with visual acuity. Specifically, the presence of a fibrotic pillar, FCE, IRF, CNV and atrophy were all associated with worse vision. Consistent with prior studies showing relatively little difference in BCVA between intermediate Gass stages,7 8 in our study, there were no significant differences between eyes graded as Gass types 2–4. Given the subjective nature of Gass typing and the stronger correlation of OCT-based structural features with BCVA, OCT-based categorisation may have a role for prognosis in addition to, or in place of, traditional Gass staging.10
To date, few studies have evaluated longitudinal structural changes on OCT in patients with BVMD. Previous case reports have shown that patients may not progress sequentially through Gass stages, and some eyes may even paradoxically reverse from higher (more advanced) stages to lower (less advanced) stages, even in the absence of any treatment.11 12 In our study, we encountered several examples (10.7% of eyes with longitudinal follow-up) where patients reverted from higher to lower Gass lesion types (eg, rebound of a vitelliform lesion after initial lesion collapse). Interestingly, 57.1% (4 eyes) had CNV responsive to intravitreal anti-VEGF treatment.
CNV is traditionally thought to be a late-stage manifestation of BVMD,13–17 but the detection of CNV is particularly challenging with fluorescein angiography (eg, pooling within vitelliform lesions often limits the identification of CNV-related leakage). The presence of SRF on OCT is a useful indicator of CNV in other conditions such as exudative AMD. However, in patients with BVMD, SRF is routinely seen within vitelliform lesions, and the presence of SRF is associated with better visual acuity whereas loss of SRF (ie, collapse of vitelliform lesion) is a poor prognostic sign. As such, the presence of SRF is not a reliable sign of CNV in BVMD, and other imaging-based biomarkers (eg, fibrotic pillar, FCE, IRF) may be more useful. The presence of FCE and IRF have been well-described as OCT-based markers for CNV in other conditions.25–27 29 Fibrotic pillars are a relatively distinct finding in patients with BVMD, and recently, OCTA has been used to demonstrate that these subretinal lesions often contain abnormal choroidal vascular networks.18 As shown in Case 3 (figure 3), these peaked subretinal lesions are highly-hyperreflective on OCT, and may collapse or involute, resulting in FCE, loss of SRF and decrease in visual acuity.
The relatively high prevalence of structural OCT features associated with CNV suggests that CNV may actually be an early rather than late-stage finding in patients with BVMD, and that CNV may play an important role in the transition from vitelliform lesions to atrophy. Individual examples such as Case 2 (figure 2) are thus particularly provocative and demonstrate that vitelliform lesions can rebound after initial collapse, violating the expected advancement through progressively higher Gass stages. As shown in this example case, the collapse of a vitelliform lesion was associated with the presence of FCE and CNV on OCT and OCTA. Treatment with anti-VEGF therapy resulted in regression of any visible CNV on OCTA, rebound of the vitelliform lesion with SRF and improvement in BCVA. These findings suggest that CNV may play an important role in the pathophysiology of BVMD, particularly pertaining to progression from earlier stages to vitelleruptive or atrophic lesion types. Early detection of the presence of CNV, including with newer imaging modalities such as OCTA, may allow for treatment that decreases the likelihood of vision loss. Studies with longitudinal OCTA-based are needed to further clarify the relationship between structural and vascular changes over time.
Our study had several limitations, including its retrospective nature and the lack of standardised volume scan protocols with regard to scan density or follow-up intervals. As such, we focused on qualitative OCT-based features that might be readily identifiable by a clinician. Our group has previously demonstrated that photoreceptor outer segment length varies in BVMD,20 including by genotype30; future studies may use retinal sublayer thickness analyses to investigate quantitative differences by genotype. Due to the relative nascency of OCTA imaging, we had longitudinal OCTA in only a few patients, including those with high clinical suspicion for CNV. More work is needed to correlate structural features (eg, pillars, FCE) with the presence or absence of a visible CNV on OCTA, perhaps with swept-source OCTA given its improved ability to study choroidal details. Similarly, though prior OCTA-based studies have demonstrated choriocapillaris flow loss in patients with BVMD,18 the pathophysiologic sequence of choroidal flow loss, FCE formation and CNV remains unclear. Leveraging of data from multimodal imaging and long-term follow-up studies may pave the way for a more comprehensive staging system that parallels the pathophysiology of BVMD and gives clinicians a better sense of prognosis.