Aim To describe the in vivo spatial and morphological vitreoretinal relationships associated with diabetic retinal neovascularisation using Fourier-domain optical coherence tomography (FD-OCT).
Methods Qualitative assessment of macula, retina and optic disc head FD-OCT (Topcon 3D OCT-1000) imaging of patients with treatment-naive and laser-treated proliferative diabetic retinopathy (PDR). The morphology and plane of retinal neovascularisation at the disc (NVD) and elsewhere in the retina (NVE) were examined, and the posterior vitreous relationships were evaluated. The FD-OCT characteristics of clinical versus subclinical PDR disease were correlated with conventional and wide-field Optos fundus fluorescein angiography.
Results 50 eyes of 50 patients were evaluated in this retrospective study. Retinal neovascularisation appears as a hyper-reflective complex, with NVE arising from inner retina with disruption through the internal limiting membrane to attach to the posterior hyaloid surface. FD-OCT detected subclinical hyper-reflective NVD complexes that were subvisible on colour fundus imaging. We describe retinoschisis, vitreoretinal adhesions and pegs, zones of separation, and intraretinal tractional elements in untreated PDR patients using high resolution FD-OCT.
Conclusions FD-OCT can non-invasively characterise retinal and optic nerve head neovascular complexes at different stages of the proliferative disease process. In clinical practice, FD-OCT can monitor the in vivo serial changes of retinal neovascularisation over time.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Retinal and optic nerve head neovascularisation are the hallmark of proliferative diabetic retinopathy (PDR) and responsible for the haemorrhagic complications that lead to significant visual loss in this disorder. In diabetics, the retinal neovascular outgrowths are characterised by a rete of new vessels arising from the venous side of the circulation, and penetrate the internal limiting membrane (ILM).1 The early new vessels lie flat and orientate parallel with the retinal surface. In response to the ongoing hypoxic stimulus, retinal ischaemia and elevated intraocular vascular endothelial growth factor levels, the new vessels expand around a central peduncle. Mature retinal neovascular complexes resemble ‘fronds’ and continue to grow in the shape of a flat cartwheel or umbrella.2 Shimizu and coworkers demonstrated that the mid-peripheral retina was far more prone to developing capillary non-perfusion than the posterior retina, with the highest rate of proliferative activity in the overlapping area between the ‘penumbra’ and outer shell of tissue.3
In current clinical practice, retinal neovascularisation in PDR is classified using the modified Airlie House Classification that defines four key factors on standard fundus photographs: the location of new vessels; the extent of fibrous proliferation; the plane of proliferation; and the presence of retinal elevation.4 New vessels at the disc (NVD) and new vessels elsewhere (NVE) are then graded and staged for activity or regression using clinical biomicroscopy and fundus fluorescein angiography (FFA). The natural history of severe and advanced NVD and NVE results in the development of tractional retinal detachment (TRD), avulsion of new vessels into the vitreous, perhaps retinoschisis, and haemorrhagic complications at the vitreoretinal interface.5 In the UK, the diabetic retinopathy screening programme captures the majority of patients with PDR who are classified R3 for urgent hospital referral.
Optical coherence tomography (OCT) is a non-invasive method for imaging the optic nerve head, macula and posterior pole retina, and is now used routinely for the diagnosis and monitoring of patients with macular oedema.6 ,7 In clinical trials, OCT parameters are used as outcome measures, and alternative OCT classifications have been developed for diabetic macular oedema (DMO).6 ,8 The microstructural changes at the vitreoretinal interface are understood from histopathological studies. The new vessels are thought to adhere to the posterior hyaloid surface and become elevated during involutional vitreous changes such as synchisis and synaeresis, whereas other studies have suggested that the neovascular outgrowths become incorporated into the cortical vitreous.9 ,10
In our study, we used Fourier-domain OCT (FD-OCT) to non-invasively capture high-resolution tomographic and topographic images of optic nerve head and retinal neovascularisation. Scanning laser ophthalmoscopy fundus imaging and wide-field FFA testing were used to correlate the FD-OCT features. The main objective of this study is to describe the in vivo OCT characteristics of optic nerve head and retinal neovascularisation, and evaluate the natural history of neovascular complexes before and following laser photocoagulation therapy.
A cohort of treatment-naive patients with PDR were studied in two prospective clinical trials, the Manchester Pascal study (MAPASS) and the Manchester Targeted Retinal Photocoagulation (MTRAP) study at Manchester Royal Eye Hospital (MREH) between 23 June 2008 and 2 April 2011.11–13 The current study investigated secondary endpoint measures using FD-OCT and fundus imaging that were part of the MAPASS and MTRAP trials. The patient selection method was based on the location of the retinal and optic nerve head neovascularisation. We only included patients with neovascular complexes that could be captured on the standard disc and macula OCT scans that are outlined in the OCT protocol below. Patients with peripheral NVE complexes were outwith the OCT imaging protocol.
The patient inclusion criteria for the MAPASS and MTRAP trials included: diabetic patients older than 18 years of age; glycosylated haemoglobin (HBA1C) level of ≤10%; no previous laser, intraocular drug therapy or surgery to the study eye; blood pressure less than 180/110; and absence of any systemic medication known to be toxic to the retina. The study was approved by the Central Manchester Research Ethics Committee. Data and safety monitoring was provided by an independent panel at the University of Manchester and MREH. The study adhered to the tenets of the Declaration of Helsinki, and the study was registered at Central Manchester University Hospitals NHS Foundation Trust.
Optical coherence tomography
The FD-OCT (3D OCT-1000, Topcon Instruments, Newbury, UK) was performed in the week before treatment, and at intervals up to 6 months following laser treatment. The scans were 6 mm in length, with resolution 512×128, and each ETDRS macular grid was placed at the macula and at the optic disc head. High-resolution single scan imaging was captured using the overlapping 50 line scan mode. A certified and blinded ophthalmic photographer captured serial scans. At each visit, the study eye underwent two OCT scans using internal and external fixation to provide better reproducibility. In all studies, scans with a Q factor less than 50 were excluded from the final analysis. Each line scan across the optic nerve head scans was analysed to assess the vitreous status and configuration of the NVD.
Fundus photography and angiography
Certified and masked ophthalmic photographers at MREH undertook standard ETDRS colour fundus photography and conventional FFA imaging. In addition, Optos colour scanning laser ophthalmoscopy images of the fundus (Optos P200MA, Dunfermline, Scotland) and Optos wide-field FFA (optomap fa, Dunfermline, Scotland) were undertaken. The FFA confirmed the diagnosis of PDR with neovascular leakage. Two masked senior retina specialists at MREH graded baseline FFA images to establish a modified Airlie House classification for untreated PDR grade. The FD-OCT images were assessed and graded by two retina specialists (MMKM, PES). The colour and FFA photographs were then used to map the NVD and NVE complexes to the FD-OCT images.
In all, 50 eyes of 50 patients were retrospectively evaluated for this study, with 34 male and 16 female patients. The baseline characteristics included a mean age of 44 years (range 29–58) and average HBA1C of 9.0% (SD 1.2). There were 32 type 1 diabetic patients and 18 type 2 diabetics, with the majority of Caucasian origin (92%).
Stages of neovascularisation at the disc (NVD)
Three eyes demonstrated NVD on FFA without classical features of neovascularisation on clinical biomicroscopy and colour photographs (figure 1A). This group is designated subclinical NVD. The patients underwent FFA initially due to the suspicion of NVD or severe pre-PDR. The FD-OCT scan showed hyper-reflective elements present within the optic disc cup that were attached to the outer aspect of the posterior hyaloid which formed a layer across the central part of the optic disc cup. The posterior hyaloid was attached in all cases to the disc margin.
Early, mild NVD
Early, active NVD complexes were less than a third disc diameter in size. In 12 eyes, there was a mild NVD complex present on colour fundus photography, and FFA confirmed NVD leakage in all cases. FD-OCT showed hyper-reflective structures containing tiny hypo-reflective spaces, and the NVD formed tiny hyper-reflective loops as shown on FD-OCT. Mild, early NVD was seen on FD-OCT to arise from the inner aspects of the neuroretinal rim within the optic disc cup (figure 1D,F). The loops were attached to the outer aspect of the posterior hyaloid across the optic disc head (figure 1I).
Active NVD and vitreous traction
Eight eyes showed vitreoretinal adhesions with elevation of the NVD complex. FD-OCT showed different patterns of growth across the posterior hyaloid surface (figure 2). The flat NVD form a thick hyper-reflective complex on the outer aspect of the posterior hyaloid surface that extends above the optic disc head and retinal surface. The larger NVD complex extends along the posterior hyaloid sometimes with elevation of the NVD in the presence of PVD. The posterior hyaloid was hyper-reflective and with a linear appearance on the scans. The hyper-reflective complex disrupts the posterior hyaloid and lies within the vitreous cavity. This type of elevated NVD complex was associated with hyper-reflective opacities within the vitreous that represent vitreous haemorrhage (figure 2E).
In mature NVD complexes, the neovascular tissue shows a fibrotic component on biomicroscopy. FD-OCT imaging showed circular hyper-reflective complexes on the peripapillary retinal surface, adherent to the ILM. The posterior vitreous was partially separated from the retina in four eyes with a posterior hyaloid surface that appeared to be corrugated on the scans. The branching network of NVD complexes seen on biomicroscopy and FFA were visualised using FD-OCT. Hyper-reflective loops and complexes were shown to extend from the base of the optic disc cup with attachments along the inner aspects of the disc and peripapillary ILM surface.
Following successful retinal laser treatment, 10 eyes with NVD complexes inactive on biomicroscopy and FFA were examined using FD-OCT. The postlaser FFA demonstrated reduced leakage in all cases, with mild persistent leaks observed in larger NVD complexes that were greater than half a disc diameter in size. The FD-OCT showed hyper-reflective structures between the posterior hyaloid and the optic disc head. The NVD complex was present on FD-OCT without any extension into the vitreous body, apparently being completely wrapped by the posterior hyaloid (figure 3).
Following complete neovascular regression confirmed on FFA, the NVD complex produced small areas of disruption through the posterior hyaloid that may be related to tangential or anteroposterior traction forces centred across the optic disc. Subsequently, we observed tractional avulsion of inactive regressed NVD across the posterior hyaloid that resulted in either vitreous haemorrhage or retrohyaloid haemorrhage (figure 3). In larger NVD complexes, the posterior hyaloid remained attached to the retina and disc margin, and the hyper-reflective NVD complex seemed to have become avulsed through the posterior hyaloid face on biomicroscopy. A free-floating hyper-reflective complex was detected within the vitreous on FD-OCT.
Stages of neovascularisation elsewhere
Ten eyes demonstrated NVE that were captured using macular and nasal retina FD-OCT scans. The active NVE complexes showed preretinal hyper-reflective complexes that formed branching loops across the retinal surface (figure 4C). The flat hyper-reflective NVE complexes were incorporated within the inner retinal layers. The hyper-reflective complex arose from the outer plexiform layers, extending through the inner nuclear layer, inner plexiform layer, ganglion cell layer and then crossed the ILM to attach to the outer aspect of the posterior hyaloid surface (figure 4D). In cases of elevated NVE, the hyper-reflective vitreo-neovascular complex was thickened, with contraction and straightening of the posterior hyaloid surface. These eyes showed localised pegs with separation of the posterior hyaloid cortex from the surrounding retinal surface (figure 4E,F). The hyper-reflective neovascular complex extended across the posterior hyaloid surface with disruption into the vitreous body. In eyes with smaller and localised NVE hyper-reflective complexes, there was firm attachment of the posterior hyaloid to the preretinal NVE complex (figure 4A,B).
In advanced cases of large active NVE complexes, haemorrhagic complications may develop and the underlying retina is not assessable by FFA or biomicroscopy. We identified a patient who developed a sub-ILM haemorrhage and the FD-OCT was helpful in localising the location of the haemorrhage. A hyper-reflective layer delineating the extent of the haemorrhage was seen below the ILM, with masking of the OCT signal in the deeper retinal layers due to blockage from haemorrhage. In the same eye, there was a preretinal haemorrhage detected in the temporal macula, with a homogenous hyper-reflective elevation at the vitreoretinal interface above the ILM and FD-OCT detected a PVD over the affected area and also suggested the haemorrhage could be sub-ILM.
The abnormal adhesions produced by preretinal NVE complexes were evaluated using FD-OCT segmentation maps. The FD-OCT software enables ILM mapping, and the contraction of the branching NVE complexes across the posterior hyaloid produced ILM corrugations (figure 4G). On FD-OCT, the NVE complexes span across the preretinal surface, and the rete of vessels become adherent to the ILM at variable points with penetration to the inner retinal layers.
In five eyes, FD-OCT imaging of inactive NVE complexes was analysed. In areas of vitreous separation, the hyper-reflective complexes remained attached to the ILM and fragments of the neovascular complex were observed attached to the elevated posterior hyaloid (figure 5). The NVE complexes showed a broad area of adhesion to the posterior hyaloid with thickening of the posterior vitreous cortex. In eyes with fibrosed NVE, the inner retina and ILM appeared irregular and distorted in areas of vitreo-neovascular detachment. The fibrosed NVE complex appeared as a homogeneous hyper-reflective structure that encompassed all the inner retinal layers with a broad adhesion to the overlying vitreous surface.
Tractional retinal detachment
There were two cases of TRD related to extensive NVE that were unresponsive to retinal laser therapy. The FD-OCT showed extensive vitreoretinal adhesions and pegs, and traction from NVE complexes at different levels throughout the retina. The retinal detachment showed a concave configuration on FD-OCT scans with vitreo-neovascular adhesions and pegs producing disorganisation of the inner and outer retinal layers. Hypo-reflective cystic spaces were observed in areas of TRD, and one eye had associated disruption of the posterior hyaloid surface and intravitreal haemorrhage. The FD-OCT showed tractional retinoschisis in these cases and areas of full-thickness incorporation of the NVE complex with ‘glove-like’ attachment of the NVE to the posterior hyaloid surface.
This study has demonstrated the microstructural and spatial vitreoretinal relationships of neovascular complexes at different stages of PDR disease. The main FD-OCT features in PDR disease include: (1) In subclinical NVD and early PDR, the neovascular complex forms an initial hyper-reflective loop; (2) In NVE, the rete of vessels become incorporated within the inner retina and appear as a homogenous hyper-reflective complex; (3) Enlarging NVE complex breaks through the ILM and grows axially along the preretinal plane; (4) In areas of vitreoretinal attachment, the NVE adheres to the posterior hyaloid which seems to surround it; (5) In large NVE complexes, the NVE and the posterior hyaloid surface seem to become disrupted by the tractional forces exerted by the vitreous with development of retrohyaloid or intravitreal haemorrhage; (6) Enlarging NVD follows a similar spatial growth and natural history to NVE with adhesion to the posterior hyaloid surface and disruption of the complex and haemorrhage observed in both untreated and regressed NVD; and (7) Treated NVD and NVE complexes can be associated with vitreoretinal adhesions which can result in avulsion of neovascularisation from the retinal surface or tractional retinal complications.
An important finding of this study was the ability of FD-OCT to detect subclinical NVD in diabetics. Diabetic patients often have significant cardiac and renal disease, and together with unstable glycaemic control, not all patients are suitable to undergo FFA investigations. In such cases, FFA is deferred and clinical biomicroscopy alone is used to confirm the diagnosis of PDR. We would suggest that FD-OCT be used as a non-invasive tool to diagnose suspicious NVD complexes in suspected PDR disease. Although the FD-OCT provides no index of the activity of PDR disease, detection of NVD would assist clinicians with the decision to perform urgent retinal laser therapy in suspected cases.
In the UK, the NHS Diabetic Eye Screening Programme is a national population-based optometrist-led screening service for diabetic patients. Currently, fundus photography is used to detect and grade retinopathy, with R3 designated PDR disease. FD-OCT is currently being considered for DMO (M1) detection in centres within the UK.14 To date, there are no previous clinical studies that have reported the FD-OCT or spectral appearances of retinal neovascularisation at different stages of PDR disease and following laser treatment. Although FD-OCT is used in routine clinical practice to evaluate DMO, we would recommend that FD-OCT be considered as an additional tool to help primary and secondary retinopathy graders to improve the sensitivity of R3 detection in diabetic retinopathy screening programmes. The additional FD-OCT imaging would help arbitration graders decide on the borderline R3 cases where collateral vessels and anomalous optic disc margin capillary networks can masquerade as NVD.
There are a number of FD-OCT and spectral-domain OCT devices used in ophthalmic practice. Spectral-domain OCT are based on Fourier transformation, so every spectral-domain device is a Fourier-domain. Currently swept source OCT devices, also based on Fourier-domain transformation, can provide a longer line that could help in the study of diabetic retinopathy. We have demonstrated the in vivo microstructural alterations that develop in patients with PDR disease, and similar appearances would be expected using other systems. The development of worsening grades of PDR disease, from subclinical to large NVE and NVD complexes, was previously reported using FFA and pathological studies.15 The use of OCT to show the vitreoretinal adhesions and tractional retinal morphological changes before and following laser photocoagulation provides clinicians with a better understanding of the complex natural history of PDR disease. Despite successful laser therapy, there are mechanical alterations at the level of the ILM and retina. OCT is able to accurately map posterior hyaloid attachments at the posterior pole, optic disc and macula that lead to haemorrhagic complications in well-lasered diabetic eyes, as well as help differentiate subclinical TRDs from tractional retinoschisis. In these cases, further laser may not be useful; vitrectomy surgery could be an option for recurrent tractional haemorrhages that are sight-threatening or visually disabling, while tractional retinoschisis does not respond to surgery with improvement in vision.
A study limitation includes the inability to image far beyond the vascular arcades with current commercially available FD-OCT devices. However, our study has demonstrated a spectrum of FD-OCT features representing early and severe PDR disease that form the majority of patients in clinical practice. Additional FD-OCT data from eyes with ‘table-top’ TRD at the macula would be helpful to understand this condition further, as our study had a low number in this subgroup. A further limitation of the FD-OCT device is the inability to image NVE complex within the inner retina in much higher resolution. Future OCT imaging studies using higher resolution devices and longer scans may provide further insights into the NVD and NVE in vivo pathogenesis. Another limitation of this study is that it did not include FD-OCT imaging of normal optic disc and retina as a control group as this was not deemed necessary, since these normative imaging data are well recognised and there were no cases of anomalous disc conditions included in this study.16 Cho and coworkers recently reported OCT features of retinal neovascularisation. However, in our study with a larger number of eyes, we are able to visualise FD-OCT features of NVD and NVE complexes at different stages of the PDR disease process with treatment-naive and laser-treated eyes with PDR.17
An important theoretical application of FD-OCT in relation to PDR disease is the microstructural delineation of tractional neovascular complexes, and the high-resolution imaging provided by FD-OCT can guide the preoperative vitreoretinal surgical plan in cases of TRD. The initial key surgical step for TRD cases involves the intraoperative identification of the cleavage plane between the posterior hyaloid or a large NVE complex/adhesion and the posterior pole and macula. This study may demonstrate that vitreoretinal adhesions and pegs, zones of separation, and intraretinal tractional elements are important spatial landmarks for avoiding iatrogenic retinal breaks and safely performing surgery. Future OCT systems may allow for the preoperative detection of vitreoschisis in this patient group, allowing perhaps for a more careful and safe intraoperative separation of the outer posterior hyaloid layer from the retina. In cases of tractional NVD with or without vitreous haemorrhage, the surgical decision to either segment or delaminate NVD complexes can be difficult to make. Preoperative and perhaps, intraoperative, FD-OCT evaluation may be a helpful surgical adjunct especially for eyes with TRD involving the macula to identify planes of cleavage.
In an era of intravitreal therapy for diabetic retinopathy, patients are now undergoing preoperative intravitreal bevacizumab (Avastin, Genentech) to reduce the risk of intraoperative haemorrhage.18 A potential complication is an immediate postbevacizumab fibrotic response with development of worsening TRD.19 In cases with clear media and TRD, FD-OCT may provide useful preoperative and postoperative outcome markers of disease, and the application of FD-OCT could be used in clinical studies as an outcome measure.
This study highlights the potential role of FD-OCT in the detection and monitoring of optic nerve head and retinal neovascularisation in PDR as well as the potential of FD-OCT in the preoperative assessment of patients requiring pars plana vitrectomy.
Contributors Substantial contributions to the intellectual content of the paper as follows: MMKM and PES: Design and hypothesis of study, data acquisition, data analysis and interpretation, patient recruitment and data acquisition. Substantial contributions to the written content of the paper as follows: MMKM and PES: Drafting of the manuscript, and critical revision of the manuscript for important intellectual content.
Funding This research has been supported by the Manchester Academic Health Sciences Centre (MAHSC) and the NIHR Manchester Biomedical Research Centre and the Manchester Vision Regeneration (MVR) Lab.
Competing interests None.
Ethics approval Central Manchester Research Ethics Committee, UK.
Provenance and peer review Not commissioned; externally peer reviewed.