A, Red-free fundus photograph illustrates acute hemorrhagic superotemporal branch retinal vein occlusion at baseline. B, Fluorescein angiography taken 36 months following presentation illustrates resolution of hemorrhage and presence of collateral vessels. The depth of the collateral vessels within the retina cannot be determined with fluorescein angiography. C, Projection-resolved stereo image pair depicting a volume rendering of superficial vascular plexus (yellow) and deep vascular complex (orange), illustrating the location of collateral vessels in the deep vascular complex.
A and B, 3 × 3-mm en face OCTA images (Optovue) with segmentation of the superficial vascular plexus (SVP) (A) and deep vascular complex (DVC) (B), respectively, illustrate reduced flow density at each level and collaterals only in the DVC. Horizontal (green) raster lines on en face images correspond to B-scan segments through the SVP (C) and DVC (D), with vertical (red) raster lines corresponding to a collateral vessel and highlight its location in the DVC from both en face and B-scan perspectives. C and D, Green lines in B-scan images demonstrate segmentation boundaries for the SVP and DVC, respectively, with red flow overlay.
Standard Spectralis OCT2 en face optical coherence tomography (OCTA) with segmentation of the superficial vascular plexus (SVP) (A) and deep vascular complex (DVC) (B) show reduced flow density at each level and collaterals confined to the DVC, confirming findings illustrated with the Optovue (Figure 4) and Topcon Triton (eFigure 2 in the Supplement) devices. Horizontal (green) raster lines correspond to B-scan segments through the SVP (C) and DVC (D) with vertical (blue) raster lines corresponding to the same collateral vessel as eFigures 1 and 2 in the Supplement confirms its location in the DVC from both en face and B-scan perspectives. C and D, Corresponding B-scan images with yellow flow overlay show the segmentation boundaries (red lines) used for the SVP (A and C) and DVC (B and D).
There is a more radiating pattern (red) in superficial vascular plexus (SVP) specialized for arterial inflow with arterioles supplying all 3 capillary beds. A more vortex-oriented pattern in the deep vascular complex (DVC) is specialized for venous outflow (blue). Note that all venous outflow courses through the DVC owing to minimal arteriovenous communication confined to the SVP. DCP indicates deep capillary plexus; GCL, ganglion cell layer; ICP, intermediate capillary plexus; IPL, inner plexiform layer; INL inner nuclear layer; NFL, nerve fiber layer, OPL, outer plexiform layer.
eFigure 1. Left Eye of Patient 10 With Collateral Vessels Following a superotemporal Branch Retinal Vein Occlusion
eFigure 2. Topcon Triton Imaging of Patient 10 With Collateral Vessels Following a Superotemporal Branch Retinal Vein Occlusion in the Left
eFigure 3. Dense B-scan OCTA (DB OCTA) En Face
Customize your JAMA Network experience by selecting one or more topics from the list below.
Freund KB, Sarraf D, Leong BCS, Garrity ST, Vupparaboina KK, Dansingani KK. Association of Optical Coherence Tomography Angiography of Collaterals in Retinal Vein Occlusion With Major Venous Outflow Through the Deep Vascular Complex. JAMA Ophthalmol. 2018;136(11):1262–1270. doi:10.1001/jamaophthalmol.2018.3586
Can the location and course of collateral vessels formed after retinal vein occlusion enhance our understanding of the venous outflow anatomy in the macula?
In this cohort study of 23 eyes in 23 patients with collateral vessel formation resulting from a retinal vein occlusion, all collateral vessels were noted to course through the deep vascular complex, and no collaterals were located exclusively in the superficial vascular plexus.
Venous drainage from the retinal capillary system may course predominantly through the deep vascular complex.
Analysis of collateral vessel formation following retinal vein occlusion may advance our understanding of the venous outflow anatomy in the macula.
To determine the location of collateral vessels with optical coherence tomography (OCT) angiography imaging.
Design, Setting, and Participants
Observational retrospective cohort study. Collateral vessel formation was studied with OCT angiography (OCTA) in patients with retinal vein occlusion (RVO). The study took place at 2 retinal practices (Vitreous Retina Macula Consultants of New York and Stein Eye Institute, University of California, Los Angeles), with patient records retrieved from March 2015 to August 2017. Data analysis was completed in November 2017.
Collaterals identified with fundus photography and/or fluorescein angiography were analyzed with OCTA to determine their course through the superficial vascular plexus (SVP) and the deep vascular complex (DVC).
Main Outcomes and Measures
Collateral vessel pathways through the SVP and DVC were analyzed with cross-sectional and en face OCT and OCTA segmentation and color-coded volume renderings prepared from raw OCTA voxel data.
From 23 eyes (22 branch and 1 hemispheric retinal vein occlusion ) of 23 patients (mean [SD] age, 73  years), 101 collateral vessels were identified and analyzed (mean [SD], 4.4 [2.0]; range, 2-9 collateral per eye). On OCTA, the collaterals appeared as curvilinear dilated flow signals that connected veins across the horizontal raphe or veins on opposite sides of an occluded venous segment within the same retinal hemisphere. Of the 101 collaterals analyzed, all showed greater flow signal in the DVC, and all had some portion of their course identified within the DVC. No collaterals were found exclusively in the SVP. Volume renderings for 3 cases confirmed qualitatively that retinal collateral vessels course through the retina predominantly at the level of the DVC.
Conclusions and Relevance
Based on a limited number of cases, all collateral vessels associated with retinal vein occlusion were found to course through the DVC. The absence of collaterals isolated to the SVP supports a serial arrangement of the SVP and DVC, with venous drainage predominantly coursing through the DVC.
Retinal collaterals occur more frequently after branch retinal vein occlusion (BRVO) than after central retinal vein occlusion (CRVO). As described by Henkind and Wise,1 collaterals occur in existing capillary beds and link vessels of the same anatomical classification unlike vascular shunts that connect artery to vein.1,2 Collaterals may be identified as dilated and tortuous vascular segments developing within weeks of an RVO.1 Their occurrence is driven by hemodynamic and hydrostatic forces causing existing capillary channels to enlarge to bypass venous obstruction.1,3-5
Venous collaterals should be distinguished from retinal neovascularization that forms owing to endothelial proliferation driven by ischemia-induced angiogenic cytokines.6-9 Retinal neovascularization proliferates anteriorly to the retinal surface and is prone to leakage and bleeding; collaterals exist within the retinal capillary bed, linking the obstructed vessel with the nearest patent vessel of the same type.1,3,9
Some understanding of the capillary organization within the central retina originates from histologic study of nonhuman primates and human donor tissue.10-15 Optical coherence tomography angiography (OCTA) enables noninvasive depth-resolved visualization of retinal vascular flow as an alternative to dye-based angiography, which lacks depth resolution and ability to delineate deeper capillary networks.16,17 Optical coherence tomography angiography facilitates identification and segmentation of the deep vascular anatomy in normal eyes.18-20 Using projection-resolved OCTA, Campbell et al21 successfully resolved 3 main capillary plexuses within the macula and 4 plexuses around the optic nerve which were previously described ex vivo.21 Gattoussi and Freund22 used structural en face OCT to confirm vessel location identified with en face OCTA without the need for a projection-resolved OCTA algorithm, confirming the presence of discrete capillary planes bracketing the inner nuclear layer, illustrating different morphology within these 2 plexuses.22
A gap exists in understanding the route of venous drainage through the retinal capillary plexuses.23 Older literature has described deep capillary beds as venous in nature and superficial vascular plexuses (SVP) as arterial.24-27 However, OCTA reports assumed that venous drainage originates in parallel within the 3 main retinal plexuses (SVP, intermediate capillary plexus [ICP], and deep capillary plexus [DCP]), with each plexus having a separate and independent arterial inflow and venous outflow.21 Another study described transverse capillaries connecting arterioles to veins within the SVP.28 By contrast, other OCTA studies have proposed that the capillary plexuses are arranged in series with the main arterial supply in the SVP and main venous outflow originating in the ICP and DCP (collectively referred to as the deep vascular complex [DVC]).13,23,29 This in-series arrangement was further supported by a 2017 analysis of porcine immunolabeled retinal vessels.30 The purpose of our study was to explore venous outflow anatomy within the macula using an OCTA analysis of collateral formation in BRVO or hemispheric RVO.
This study adhered to the tenets of the Declaration of Helsinki and complied with the Health Insurance Portability and Accountability Act of 1996. It was approved by the institutional review board committees at the Western institutional review board and University of California, Los Angeles. Patients provided written informed consent.
A retrospective medical record review was performed on patients with BRVO or HRVO, with macular involvement, undergoing OCTA at least 6 months after baseline at 2 retinal practices (K.B.F. and D.S.). Patients with BRVO and collaterals identified by color photography and/or fluorescein angiography (FA) imaged with OCTA between March 2015 and August 2017 were enrolled. Imaging was reviewed to identify the retinal arterial and venous systems and to identify collateral vessels as dilated and tortuous capillary vessels arising from venous beds connecting the occluded vein to an adjacent patent vein. Eyes that could not be assessed by OCTA owing to media opacity (OCTA signal strength <50) or segmentation failure owing to severe macular edema or disrupted perifoveal anatomy were excluded.
All patients underwent spectral-domain OCT (Spectralis HRA + OCT; Heidelberg Engineering), OCTA (RTVue XR Avanti; Optovue Inc), fundus photography, and/or FA. Fluorescein angiography was performed using the wide-field Optos system (Optos PLC) or Topcon TRC-50XF fundus camera (Topcon Medical Systems). When available, OCTA studies performed on other systems were reviewed.
The OCTA images were acquired with the RTVue XR Avanti (Optovue Inc) using ReVue, version 2016.2.0.35 at each site. Both 6 × 6-mm and 3 × 3-mm volumes centered on the fovea were reviewed, with 6 × 6-mm scans analyzed when both were available. These scan patterns were a 304 B-scan raster with 304 A-scans per B-scan. Two orthogonal volumes were acquired at each retinal location.31
When available, additional OCTA studies acquired on the Spectralis OCT2 (Heidelberg engineering), Topcon DRI OCT Triton (Topcon Medical Systems), and PLEX Elite 9000 were analyzed qualitatively. The Triton and PLEX Elite are swept-source OCT devices that use different proprietary algorithms.32-34
In 1 eye with BRVO, dense B-scan OCTA (DB OCTA) acquired with the Spectralis OCT2 device was reviewed. Dense B-scan OCTA is a custom scan pattern that acquires 150 to 300 structural B-scans over a scan line width of 10° to 30° and a total raster span of 0.1° to 1.0°.35 Single cross-sectional structural OCT images are created by applying a Gaussian filter across neighboring scans to increase signal to noise ratio and to smooth both the structural cross-section and the flow overlay.
Two independent readers reviewed the data from their respective institutions (B.L. and S.G.). Ambiguity was resolved by the senior grader at each institution (K.B.F. or D.S.). Collaterals, first identified after evaluation of fundus photography and/or FA, were then analyzed with OCTA to determine whether they coursed through the SCP, DVC, or both.
Optical coherence tomography angiography assessment of collateral location was subsequently made on cross-sectional imaging to ensure only flow overlay was accounted for and to eliminate projection artifact misinterpretation. The corresponding location was only then identified on en face imaging.
The Optovue ReVue software was used for the quantitative analysis. The SVP was defined as spanning the nerve fiber layer, ganglion cell layer, and superficial inner plexiform layer (IPL) layers, as previously described.21,36 Automated segmentation parameters included a 3-μm inner limiting membrane offset and 15-μm IPL offset that was manually adjusted if deviation from these anatomic landmarks caused by mild macular edema or retinal atrophy was detected. Analysis was made of every cross-sectional scan with flow overlay in each volume, and adjustment of segmentation where necessary to avoid projection artifact misinterpretation.
The DVC was defined as spanning the deeper IPL, inner nuclear layer, and outer plexiform layer. Automated segmentation parameters included an upper 15-μm IPL offset and lower 71-μm IPL offset.
Collaterals were selected and analyzed if they met the qualitative criteria and quantitative threshold for microvascular tortuosity. Tortuosity was judged qualitatively by identifying vessels with conspicuous spiral twisting, kinking, and/or angulation of capillary vessels compared with normal capillary beds in an area unaffected by the BRVO. Tortuosity was quantified by measuring retinal capillaries within the SVP and DVC in areas of normal macular perfusion of fellow normal eyes using the measure function in Fiji software (ImageJ). One hundred well-defined normal retinal capillary segments in the SVP and DVC, respectively, were measured from a total of 6 normal eyes. Tortuosity of normal capillary segments was calculated as previously described by Lee et al37:
Normal Capillary Tortuosity = Sum of Lengths of Normal Retinal Capillary Segments / Sum of Euclidean Lengths of the Same Retinal Capillary Segments
This technique yielded a normal vessel tortuosity ratio (SD) of 1.120 ( 0.105) in the SVP, a vessel tortuosity ratio (SD) of 1.157 (0.112) in the DVC, and an overall ratio (SD) of 1.136 (0.110). Vessels first identified as collaterals on fundus photography and/or FA with tortuosity ratios greater than 1.33 (2 SDs greater than the mean) on OCTA were included in our study set as collateral vessels.
The OCTA analysis of collaterals was performed within the 6 × 6-mm OCTA volume scan. The 3 × 3-mm volume scan was assessed if the 6 × 6-mm volume scan was not available. Collaterals were studied 3-dimensionally by scrolling through en face OCTA projections to determine whether they coursed through the SVP, DVC, or both. We excluded analysis of collaterals involving the perifoveal vascular ring where the SVP terminates closer to the fovea than the DVC.18 Hence, the central oval area of 0.95 mm horizontally and 0.85 mm vertically were excluded from analysis.
For better appreciation of the anatomical associations of collateral vessels with retinal capillary plexuses and to illustrate the main study findings, volume renderings were prepared using raw OCTA voxel data from RTVue XR scans. This was performed on the first 3 high-signal strength index studies whose collateral vessels were within a 3 × 3-mm volume centered at the fovea. To reduce ambiguity in interpreting the depths of collateral vessels, projection artifacts were removed using the technique described by Zhang et al.38 Segmentation of the SVP and DVC was performed using a modified binarization method previously validated39 and a Canny edge40 detection operator implemented in MATLAB (MathWorks) so that each plexus could be color-coded. Final renderings and stereo pairs were prepared in ImageJ using the 3D viewer plugin.41
Twenty-three eyes of 23 patients having a diagnosis of BRVO (22 eyes) or hemispheric RVO (1 eye) and associated collaterals underwent OCTA imaging with the RTVue XR Avanti and displayed images of sufficient quality and signal strength for quantitative analysis. The analyzed cohort included 13 women (56.5%), with a mean (SD) age of 73 (11) years (Table). All patients were white, with the exception of patient 1, who was South Asian. Mean (SD) logMAR visual acuity at OCTA acquisition was 0.22 (0.135) and median time from RVO diagnosis or initial visit to imaging acquisition was 3.79 (25% and 75% interquartile range, 1.25 and 6.32) years. A 3 × 3-mm and/or 6 × 6-mm scan acquired with the RTVue XR Avanti was analyzed. The 3 × 3-mm scans were analyzed for 5 eyes, while 6 × 6-mm scans were analyzed for the remaining 18 eyes. One hundred one collateral vessels were identified from the 23 eyes. No collaterals were subsequently excluded by including the quantitative component of our methods. In addition, scans from 10 eyes (10 patients) imaged with the Triton DRI OCT, 4 eyes (4 patients) imaged with the PLEX Elite 9000, 3 eyes (3 patients) imaged with the Spectralis OCT2, and 1 eye imaged with the Spectralis dense B-scan OCTA were analyzed.
With en face OCTA, a variable number of large well-defined flow signatures corresponding to collaterals identified on fundus photographs and/or FA were visualized readily in DVC segments but identified inconsistently in SVP segments. Collaterals appeared as curvilinear dilated flow signals that connected veins across the horizontal raphe or veins on opposite sides of an occluded venous segment within the same retinal hemisphere. There was a mean (SD) of 4.4 (2.0) collateral vessels per eye (range, 2-9) (Table). All 101 collaterals showed more flow signal in the DVC and all had most of their course identified within the DVC (Figures 1 and 2; eFigure 1 in the Supplement). Additional tortuous microvascular flow signals identified predominantly in the DVC segmentation that did not meet inclusion criteria were a common finding. These flow signals likely represented networks of smaller collaterals that we were unable to isolate, analyze, and quantify.
None of the study eyes displayed collaterals isolated to the SVP. Two eyes had a single collateral isolated to the SVP segmentation but were located within the perifoveal vascular ring and thus excluded from analysis (patients 8 and 10). One eye had segmentation artifacts that prevented accurate localization of 2 collateral vessels which were excluded from analysis (patient 3). Eyes with OCTA imaging acquired on other devices displayed collaterals predominantly in the DVC and confirmed a distinct lack of collateral vessels located exclusively within the SVP (Figures 3; eFigures 2 and 3 in the Supplement). In 1 eye (patient 3) with unreliable segmentation of 2 collaterals on the RTVue XR Avanti device, Triton DRI OCTA clearly and precisely isolated a predominantly DVC component of these 2 collaterals. In 2 eyes (patients 4 and 7) with unreliable segmentation of 1 collateral on the Triton DRI OCTA, the RTVue XR Avanti device demonstrated that each of these vessels had a predominant DVC component. Collateral vessel length was consistently greater in the DVC when they traversed both the SVP and DVC. Projection-resolved volume renderings for 3 cases (patients 5, 6, and 10) demonstrated the course of collateral vessels as predominantly within the DVC (Figures 1; eFigure 1 in the Supplement).
In our OCTA analysis of 101 collateral vessels identified in 23 eyes with longstanding BRVO or HRVO, we found all collaterals to be primarily located within the DVC. Excluding the perifoveal vascular ring where the SVP and DVC merge,12,18,21 no collaterals confined exclusively to the SVP were detected. These results suggest direct venous drainage from the SVP may be limited within the perifoveal macula. The absence of venous collaterals identified entirely within the SVP suggests the preponderence of retinal venous drainage course through the DVC and that a component of the retinal capillary plexus may be arranged in series. If the SVP were to have independent venous outflow, one would expect collaterals isolated to the superficial level, but collaterals confined exclusively to the SVP were never identified. While we cannot discount the possibility that our findings are associated with an SVP that is not morphologically adapted to develop collateral flow or that ischemia-induced cytokine release could increase resistance to superficial venous outflow,42-44 we believe evidence exists to the contrary.1,23,29,30
Our findings are consistent with those of prior investigators who suggested that the DVC is the principle venous outflow system for the retinal capillary plexuses1,13,20,23,26,29,45 Further, some have noted that a vortex arrangement identified in the more posterior DCP portion of the DVC suggests that, unlike the SVP, the DVC may show structural adaptation for venous drainage.28,46 These morphologic variations indicate the retinal capillary plexus may be at least partly arranged in series.23,29,30
Our findings have implications regarding mechanisms of disease involving the posterior retina. For instance, in some eyes with acute CRVO associated with a prolonged arteriovenous transit on FA, clinical examination shows a characteristic pattern of macular perivenous retinal whitening.47,48 En face structural OCT performed in these eyes typically illustrates a perivenular pattern of paracentral acute middle maculopathy indicating macular ischemia involving the deep retina (specifically the DVC) that spares the superficial layers.29,45,46 These findings imply a predominantly in-series vascular arrangement in which the prolonged arteriovenous transit leads to increased oxygen extraction in the SVP and disproportionately hypoxic conditions when the blood reaches the more distal capillaries of the DVC in the middle retina, a phenomenon described by McLeod as misery perfusion.49 However, we acknowledge that other factors should be explored, including mechanisms for vascular autoregulation, that may also contribute to this pattern.42
A number of studies have illustrated that macular edema typically develops in association with DVC ischemia.50-52 The DVC may be a key factor in the development of bypass channels to permit unencumbered venous outflow of blood. Disruption or damage to the DVC owing to ischemia or occlusion may set the stage for macular edema development given the absence of a venous drainage pathway leading to extravasation of blood and fluid into surrounding retinal tissues. The importance of venous outflow coursing through the DVC may also have significance when considering disorders, such as nonproliferative diabetic retinopathy and macular telangiectasia type 2, that manifest more severe pathology at this level.53,54 Additionally, owing to its venous nature, the DVC may be particularly vulnerable to injury in systemic conditions that increase central venous pressure including sleep apnea, congestive cardiac failure, and pulmonary hypertension.
This study was limited to an analysis of 23 eyes, with only a single HRVO case. It is possible that with a greater number of cases, we may have identified collateral formation not principally in the DVC or cases with collaterals exclusively in the SVP. Some eyes received intravitreal anti–vascular endothelial growth factor injections that may have modified collateral vessel formation, although, to our knowledge, this effect has not been demonstrated.55 Moreover, it is possible, although unlikely, that OCTA technology lacks the sensitivity to detect low flow in very small collateral vessels in the SVP. As with any OCTA analysis intended to define the precise location of the intraretinal flow signal, there is the potential for layer segmentation errors and a need to distinguish projection artifacts from true flow in each retinal layer.56 Most of our analysis did not use a projection removal algorithm, so it is possible that DCP layer segmentation may have been partly confounded by more superficial vessels. However, the volume renderings for 3 cases were projection resolved and supported our interpretations of the data. While we excluded eyes with macular edema producing inaccurate OCTA segmentation not correctable with manual adjustment, we did not exclude eyes with retinal thinning associated with ischemia that necessitated manual adjustment. To minimize limitations, we analyzed both the en face structural OCT slabs and the cross-sectional scans with flow overlays and confirmed the locations for each of the 101 individual collateral vessels studied.
The digital illustration in Figure 4 is an updated interpretation of the arrangement of the retinal capillary plexuses based on most recent available evidence, including data from this study, and indicates a radiating capillary plexus in the SVP that is specialized for arterial flow and a vortex pattern in the DVC that is specialized for venous outflow (Figure 4).
Our analysis of 101 collateral vessels identified in 23 eyes with longstanding BRVO or HRVO illustrated that these compensatory channels were primarily located in the DVC (ICP and DCP). Importantly, no collaterals were exclusively confined to the SVP. Our findings support a predominant serial arrangement of the retinal vascular system in which venous drainage courses through the DVC adapted for this purpose.23,30 These observations may help explain specific disease patterns including perivenular paracentral acute middle maculopathy occurring in CRVO, macular edema originating in the DVC, and various retinal vascular disorders showing preferential involvement of the DVC including nonproliferative diabetic retinopathy. Further investigation into the mechanism of venous outflow at the level of the DCP may uncover novel insights into the pathogenesis of other macular diseases.
Corresponding Author: K. Bailey Freund, MD, Vitreous Retina Macula Consultants of New York, 460 Park Ave, New York, NY 10022 (firstname.lastname@example.org).
Accepted for Publication: July 3, 2018.
Published Online: August 23, 2018. doi:10.1001/jamaophthalmol.2018.3586
Author Contributions: Drs Freund and Sarraf had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Freund, Sarraf, Leong, Garrity.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Freund, Sarraf, Leong, Garrity, Vupparaboina.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Garrity.
Obtained funding: Freund.
Administrative, technical, or material support: Freund, Sarraf, Vupparaboina.
Supervision: Freund, Sarraf.
Other - image preparation: Dansingani.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Freund is a consultant to Genentech, Optos, Optovue, Heidelberg Engineering, and Spark Therapeutics and he receives research support from Genentech/Roche. Dr Sarraf is a consultant for Amgen, Bayer, Genentech, Novartis, Novelution, and Optovue and receives research grants from Allergan, Genentech, Heidelberg Engineering, Optovue, and Regeneron. No other disclosures were reported.
Funding/Support: The Macula Foundation Inc, New York, New York, funded the costs of the digital illustration.
Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We thank Ving Lac for his contribution in creating the digital illustration of our interpretation of the retinal vascular structure and flow for which financial compensation was provided by The Macula Foundation Inc. He has no affiliations.
Create a personal account or sign in to: