To evaluate the agreement between clinical examination, spectral-domain ocular coherence tomography (SD OCT), and fluorescein angiography (FA) in diagnosing intraretinal microvascular abnormality (IRMA) and neovascularization elsewhere (NVE) and define the SD OCT features that differentiate NVEs from IRMAs.
Data were collected from 23 lesions from 8 diabetic patients, seen from July 2012 through October 2013 at Moorfields Eye Hospital, United Kingdom. Main outcomes were SD OCT features and FA leakage of IRMA and neovascular complex. The agreement between 3 evaluations was analyzed by Fleiss’ kappa.
The following 5 SD OCT features significantly differentiated IRMAs from NVEs: (1) hyperreflective dots in superficial inner retina ( P = .002); (2) the outpouching of internal limiting membrane (ILM) ( P = .004); (3) the breach of ILM ( P = .004); (4) the breach of posterior hyaloid ( P = .0005); (5) hyperreflective dots in vitreous ( P = .008). The agreement was moderate between 3 evaluations (κ = 0.48, P = 7.11 × 10 −5 ) but substantial between clinical and SD OCT evaluation (κ = 0.72, P = .00055). There was no significant agreement between OCT evaluation and FA leakage (κ = 0.249, P = .232).
SD OCT will be a valuable adjunct in evaluating IRMA and NVE, since it can verify the histopathologic correlate. SD OCT provides subtle anatomic insights and may be more accurate than clinical examination or leakage on FA, our current method of diagnosing this important endpoint, which has implications in future trial design for proliferative diabetic retinopathy prevention.
Diabetic retinopathy (DR) is the leading causes of blindness in working-age populations worldwide. Initially described by Jaeger in 1855, DR was mainly categorized into nonproliferative vs proliferative disease. Proliferative diabetic retinopathy or “diabetic retinitis proliferans” was first reported by Manz in 1876, and the initial descriptions of diabetic neovascularizations have been largely based on histopathologic description of new blood vessels that grow into the vitreous through a break of the internal limiting membrane (ILM).
The term intraretinal microvascular abnormality (IRMA) arose much later as a clinical definition in 1968 from the Airlie Classification of Diabetic Retinopathy. Histopathologic description of IRMA predates the clinical definition, but Airlie classification provided an early framework for a “common language” in staging DR in clinical practice and trials. In 1981, Diabetic Retinopathy Study report No. 7 provided standard photographs 8A and 8B and subsequently, IRMA became defined as tortuous intraretinal vascular segments in fields 4–7, varying in caliber from barely visible to 31 μm per Early Treatment of Diabetic Retinopathy Study (ETDRS). As histopathology is limited to examining a single time point of a lesion’s evolution, whether IRMAs are a direct precursor lesion of neovascularization elsewhere (NVE) has not been established, but the severity of IRMA was shown to be a risk factor for the progression into proliferative diabetic retinopathy (PDR). In fact, IRMA became one of the defining characteristics of end-stage nonproliferative diabetic retinopathy and therefore an important clinical endpoint.
During the landmark trials of the Diabetic Retinopathy Study and the ETDRS, IRMA and NVE were differentiated based on color stereoscopic photographs. Although fluorescein angiography (FA) was used in ETDRS to evaluate the degree of macular edema and the severity of DR, it was not employed for the definition of IRMAs. However, ETDRS did identify that the source of “fluorescein leakage” in DR included microaneurysms, dilated capillaries, and other evident vascular abnormalities such as IRMA and neovascularization. Furthermore, it was revealed that diffuse leakage in the retina was predictive of progression of DR. Thus, in the time of Airlie classification and consequent landmark trials, FA findings were to be used only as an adjunct to clinical examination and color photography, rather than as the source of defining stages of DR. However, textbooks often state that IRMAs have no or minimal leakage on FA and that this is often how they are differentiated from NVEs.
Optical coherence tomography (OCT) is a noninvasive imaging modality that allows the evaluation of the vitreous cavity, retinal layers, retinal pigment epithelium, and choroid. The advent of spectral-domain (SD) OCT has allowed better sensitivity, increased depth of penetration, and higher resolution of each image obtained. Current commercially available SD OCT provides high-resolution images with an axial resolution of <5 μm. As a result, OCT parameters are increasingly used in various clinical trials. With commercially available OCT, it is now possible to evaluate the disruption of the ILM and the breach of the posterior hyaloid associated with NVE or neovascularization of disc (NVD). However, whether IRMA and NVE can be distinguished on SD OCT has not been established.
In this study, we perform detailed characterization of the SD OCT features of IRMA and NVE/NVD, with comparison to clinical and FA findings. In particular, we evaluate the ability of SD OCT to show breach of the posterior hyaloid in support of previous histopathologic descriptions of NVE, with the objective of refining disease feature definitions for use as clinical endpoints.
Subjects and Methods
Inclusion Criteria and Data Collection
Clinical and imaging data were collected retrospectively from patients attending medical retinal clinics at Moorfields Eye Hospital, London, United Kingdom from July 1, 2012 to October 31, 2013. All patients were assessed by medical retina specialists in the same institution. Approval for data collection and analysis were obtained from the Institutional Review Board at Moorfields Eye Hospital, London, United Kingdom and adhered to the tenets set forth in the Declaration of Helsinki.
Eight patients with a diagnosis of type 1 or 2 diabetes mellitus who had undergone concurrent FA and SD OCT scanning (Spectralis; Heidelberg Engineering, Heidelberg, Germany) for evaluation of PDR were included in the study. Patients with angiographic and SD OCT image sets of insufficient quality to allow grading of DM severity and segmentation of retinal and posterior hyaloid boundaries were excluded. No image manipulation was performed. Classification of IRMA and NVE were based on a clinical diagnosis using color and red-free photographs as part of the patients’ standard of care.
Acquisition and Analysis of Fluorescein Angiography
All angiographic images were acquired with a digital retinal camera system (Topcon TRC 50IX; Topcon Medical Systems, Inc, Paramus, New Jersey, USA). Macular centered FAs with peripheral sweeps were obtained.
Qualitative Analysis of Fluorescein Angiography Images
FA and any available fundus images were reviewed independently by 2 masked graders (C.L., A.L.). FA was interpreted as leakage or no leakage.
Acquisition and Analysis of Spectral-Domain Ocular Coherence Tomography Image Sets
SD OCT images sets were obtained using a standard, commercially available SD OCT device. In each case, both macular and extramacular raster scan acquisition protocol were performed, centered on the fovea and the NVE, respectively. SD OCT images at the NVEs were selected either with the vertical or horizontal scanning plane bisecting the NVE, and the image set size was adjusted accordingly in order to include the whole extent of the NVE, using equally spaced OCT B-scan sections, each composed of 50–100 averaged B-scans.
Qualitative Analysis of Spectral-Domain Ocular Coherence Tomography Images
All SD OCT image sets were reviewed independently by 2 masked graders without correlating FA or fundus images. Each image set was assessed for the presence of the following vitreoretinal features: (1) hyperreflective dots in superficial portion of inner retina without evidence of ILM breach; (2) the outpouching of ILM without disruption in the ILM layer; (3) the breach of ILM, defined as a disruption in the ILM; (4) the breach of the posterior hyaloid, defined as a connecting hyperreflective layer from the ILM to posterior hyaloid/vitreous cavity; (5) hyperreflective dots in the vitreous cavity; (6) presence of posterior vitreous detachment (PVD), defined as a fully detached posterior hyaloid seen as a thin hyperreflective layer above the ILM.
The SD OCT and FA features of IRMA vs NVE were analyzed with Fisher exact test. The concordance between clinical examination, SD OCT evaluation, and FA leakage were analyzed with Fleiss’ kappa. The intra- and intergrader correlation was assessed by kappa test. Significance was defined as P value <.05. Multiple comparisons were adjusted by the Bonferroni correction. All statistical analysis was performed using R ( www.r-project.org/ ).
Twelve eyes (8 patients) were included, and a total of 23 lesions were examined. Six lesions in 1 patient were followed for 14 months. The baseline demographics and clinical characteristics of the study patients are summarized in Table 1 . The mean age was 46.1 years (SD = 15.1) and 6 patients were male. Two patients had type 1 diabetes. Two patients had previous panretinal photocoagulation. Out of 18 eyes, 7 had a single lesion and 2 had more than 2 lesions.
|Patient||Age||Sex||Type DM||Lesion No.||Laterality||Clinical Dx||OCT Dx||FA Leakage a||Previous PRP|
Characteristic Spectral-Domain Ocular Coherence Tomography Features in Clinically Diagnosed Intraretinal Microvascular Abnormalities vs Neovascular Complex (Neovascularization Elsewhere or of Disc)
Clinically diagnosed IRMA and NVE were based on the clinician’s best judgment at the time of evaluation, assisted by color photographs when available. SD OCT images were graded without prior knowledge of clinical diagnosis. No patient had a complete PVD on OCT. All 5 SD OCT features significantly differentiated IRMAs from NVEs even after adjustment for multiple comparisons ( Table 2 ). First, hyperreflective dots in the inner retina, without breach of the ILM ( Figure 1 , Top left), were seen in 70% (7/10) of clinically diagnosed IRMAs but in none (0/13) of the NVEs ( P = .002). Second, outpouching of the ILM without disruption of this layer was observed in 80% (8/10) of clinically diagnosed IRMAs but in only 7.7% (1/13) of NVEs ( P = .004) ( Figure 1 , Top middle). Third, disruption of the ILM without breach of the posterior hyaloid ( Figure 1 , Top right) was observed in 20% (2/10) of clinically diagnosed IRMAs and in 92.3% (12/13) of NVEs ( P = .0007). Fourth, breach of the posterior hyaloid was seen in 20% (2/10) of clinical IRMAs and 100% (13/13) of NVEs ( Figure 1 , Bottom left). Several lesions had multiple areas of breach and a horizontal growth pattern ( Figure 1 , Bottom middle). Lastly, hyperreflective dots in the vitreous were observed adjacent to 10% (1/10) of clinically diagnosed IRMAs and 69.2% (9/13) of NVEs ( P = .002) ( Figure 1 , Bottom right).
|IRMA (n)||NVE (n)||P Value||Adjusted P Value|
|Hyperreflective dots in inner retina||7||0||.00049||.0024|
|Posterior hyaloid breach||2||13||.000092||.00046|
Assessment of Intraretinal Microvascular Abnormalities vs Neovascular Complex (Neovascularization Elsewhere or of Disc)
Based on clinical examination, 10 of 23 lesions (43.5%) were IRMAs and 13 of 23 (56.5%) neovascular complexes (2 NVDs, 11 NVEs). Using the SD OCT evidence of the breach of the ILM as the defining criterion of NVE, 8 of 23 (34.8%) were IRMAs and 15 of 23 (65.2%) were neovascular complexes (2 NVDs, 13 NVEs). Figure 2 (Top left) shows an example of a clinically defined IRMA that has a clear disruption of the ILM on OCT ( Figure 2 , Top right), but that did not leak on FA ( Figure 2 , Bottom). Five out of 10 clinically defined IRMAs (50%) and 12 out of 13 clinically defined NVDs or NVEs (92.3%) showed leakage on FA. Figure 3 shows 3 clinically defined IRMAs and their SD OCTs, respectively ( Figure 3 , Top left and Bottom row). There is diffuse leakage from all 3 lesions on FA ( Figure 3 , Top middle and Right).