You are on page 1of 8

Optical Coherence Tomography

Angiography of DME and Its Association with


Anti-VEGF Treatment Response
Junyeop Lee, MD, PhD,1 Byung Gil Moon, MD,2,3 Ah Ran Cho, MD,2,3 Young Hee Yoon, MD, PhD2,3

Purpose: To investigate the structural integrity of the superficial capillary plexuses (SCPs) and deep capillary
plexuses (DCPs) using optical coherence tomography (OCT) angiography (OCTA) in patients with diabetic
macular edema (DME) and its association with the response to antievascular endothelial growth factor (VEGF)
treatment.
Design: Retrospective, case-control study.
Participants: We included 51 DME eyes with a poor response to anti-VEGF agents and 32 age-matched
DME eyes with a good response to anti-VEGF treatment, along with 20 fellow eyes without DME from the
cases and controls.
Methods: The medical records, including OCTA and spectral-domain OCT (SD OCT), were reviewed and
compared between the groups. En face OCTA images of the SCP and DCP were obtained for each eye. An anti-
VEGF responder was defined by a reduction of more than 50 mm in central retinal thickness after 3 consecutive
anti-VEGF treatments. A poor responder was defined by a reduction of less than 50 mm or an increase in central
retinal thickness after 3 monthly injections.
Main Outcome Measures: We measured the vascular density and foveal avascular zone (FAZ) area and
counted the number of microaneurysms in each layer. The SD OCT images were compared with OCTA findings.
Results: Compared with non-DME eyes, DME eyes had a lower vascular density (P < 0.001) and larger FAZ
area (P < 0.001) in the DCP and more microaneurysms (P < 0.001) in both layers. Although there was no sig-
nificant difference in the SCP between anti-VEGF responders and poor responders, poor responders tended to
show greater damage and more microaneurysms in the DCP (P < 0.001) and a larger FAZ area (P < 0.001). The
topographic location of the disrupted synaptic portion of the outer plexiform layer (OPL) in SD OCT exactly
corresponded to the nonflow area of the DCP in OCTA.
Conclusions: Compared with DME eyes that responded to anti-VEGF treatment, poor responders show
significant damage to the integrity of the DCP, but not the SCP. The degree of OPL disruption in SD OCT cor-
responds well with the extent of DCP loss in DME eyes. The extent of DCP loss and the corresponding OPL
disruption could be useful predictors of responsiveness to anti-VEGF treatment. Ophthalmology 2016;123:2368-
2375 ª 2016 by the American Academy of Ophthalmology. This is an open access article under the CC BY-NC-
ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Supplemental material is available at www.aaojournal.org.

Diabetic macular edema (DME), a major cause of visual Fluorescein angiography (FA) is an important diagnostic
impairment in diabetic patients, affects up to 21 million of the tool for investigating leaking microaneurysms and pooling
93 million patients with diabetic retinopathy (DR) world- patterns in DME. However, FA cannot visualize the precise
wide.1 The retinal swelling in DME is caused by breakdown structure of deep retinal layers where most microaneurysms
of the blooderetinal barrier and leaking microaneurysms. and hyperpermeable retinal capillaries are located.5 Because
Because vascular endothelial growth factor (VEGF) is the spectral-domain (SD) optical coherence tomography (OCT)
primary factor for retinal vascular hyperpermeability, anti- provides cross-sectional images of the deep retinal layer,
VEGF agents are used as the primary therapy for DME, several studies have categorized different patterns of DME
effectively improving macular edema and vision in most pa- according to the combination of FA and OCT findings.6e9
tients.2 In contrast, in the case of anti-VEGFeresistant DME, These studies have performed morphologic analyses and
which suggests chronic inflammation, corticosteroids are an speculated about the pathogenesis of DME, but none of
alternative treatment option.3 Because the cost of anti- these earlier reports evaluated the association with the
VEGF agents is a major burden for patients with DME,4 it treatment response.
is necessary to identify factors in baseline examinations that Optical coherence tomography angiography (OCTA), a
can predict the treatment response to anti-VEGF agents. recently developed procedure, enables closer observation of

2368 ª 2016 by the American Academy of Ophthalmology http://dx.doi.org/10.1016/j.ophtha.2016.07.010


This is an open access article under the CC BY-NC-ND license ISSN 0161-6420/16
(http://creativecommons.org/licenses/by-nc-nd/4.0/). Published by Elsevier Inc.

Downloaded for Bowo Widiasmoko (bowo_wa@hotmail.com) at Universitas Kristen Duta Wacana from ClinicalKey.com by Elsevier on May 14, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
Lee et al #
OCT Angiography of DME

the blood flow of each retinal capillary layer.10 Abnormalities image quality, 25 to 30 frames were averaged for each B-scan. Of
in capillary flow density and microaneurysms from the deep the 25 B-scan images obtained, 1 image with the maximum height
capillary layer have been demonstrated in patients with DR, of the largest intraretinal cyst was exported to analyze the optical
aspects that could not be evaluated with FA and OCT.11e14 density ratio (ODR). We analyzed the central retinal thickness in
the standard 9-area Early Treatment of Diabetic Retinopathy Study
Because the diabetic vascular change occurs primarily in
grid and total macular volume.
the deep retinal layer, we hypothesized that the integrity of For the OCTA, AngioVue (Optovue Inc., Fremont, CA) was
the deep capillary plexus (DCP) might be associated with the used to obtain split-spectrum amplitude-decorrelation angiography
pathogenesis and treatment response of DME. We also as previously described by Spaide et al.10 The scanning area was
investigated baseline SD OCT findings associated with the captured in 3"3-mm sections and was centered on the fovea.
response of DME to anti-VEGF treatment. Optical coherence tomography angiography was performed after
the eyes were classified as anti-VEGF responders or poor re-
sponders to anti-VEGF treatment. Although there are multiple
Methods retinal capillary planes that communicate through the vertical
branches in the retina, we simply measured the superficial capillary
Patients plexus (SCP) in the ganglion cell layer and the DCP beneath the
inner plexiform layer (IPL). We used auto-segmentation software
This retrospective case-control study included 55 patients with type as follows: The en face image was segmented with an inner
2 diabetes who were diagnosed with DR using FA and underwent boundary 15 mm beneath the IPL, and the outer boundary was set at
comprehensive ophthalmologic examinations, including SD OCT 70 mm beneath the IPL to obtain images of the DCP. In the case of
and OCTA, in the Department of Ophthalmology at Asan Medical incorrect segmentation, we manually adjusted the boundary be-
Center, Seoul, Korea. The clinical severity of DR in each patient tween the SCP and the DCP. Multilayer-involving large intraretinal
was classified using the International Clinical Diabetic Retinopathy cysts frequently caused errors in the accurate detection of IPL,
and Diabetic Macular Edema Severity Scale.15 Eyes with any stage which is the reference layer for dividing the SCP and DCP. If the
of DR were included in this study. We included treatment-naïve border between the SCP and DCP was not located within the range
eyes and eyes that received previous anti-VEGF or panretinal of the IPL, we defined this as an inaccurate segmentation.
photocoagulation treatments more than 6 months before the base- Although the auto-segmentation set the retinal thickness of the
line. We defined DME as a central retinal thickness greater than DCP at 55 mm by default, we manually adjusted the offset values
300 mm. The cases were DME eyes that poorly responded to anti- of the inner and outer border of the DCP so that the segmentation
VEGF agents, and the controls were age-matched DME eyes with a lines covered the full thickness of the inner nuclear layer and outer
good response to anti-VEGF agents. A response to anti-VEGF plexiform layer (OPL). If the segmentation error with an irregularly
agents was defined as a reduction of more than 50 mm in central wavy border was not resolved by this manual extension of the DCP
retinal thickness after 3 consecutive anti-VEGF injections. A poor range, we used the flatten band tools and moved the outer border
response to anti-VEGF agents was defined as no reduction, a into the outer nuclear layer to accurately measure the DCP. In
reduction less than 50 mm, or an increase in central retinal thick- addition, we obtained images of the total capillary plexus (TCP) to
ness after 3 consecutive monthly anti-VEGF injections compared determine the entire retinal thickness via the flattening and manual
with the initial value observed 4 months previously. extension of the SCP outer border.
We allowed a 1-week variation for every monthly dosing interval.
Optical coherence tomography angiography was performed within 2
months of the last consecutive injection of anti-VEGF agent. We Image Analysis and Statistical Analysis
selected 20 DR eyes with no history of DME from among the fellow To quantify the OCTA images, we measured the vascular flow
eyes of 55 case and control patients. These non-DME eyes were density and foveal avascular zone (FAZ) area and counted the
defined by any visible microaneurysm on FA but no evidence of number of microaneurysms in each layer (Fig S1, available at
DME on SD OCT. The anti-VEGF agents used in this study included www.aaojournal.org). Saccular capillary ends and fusiform
bevacizumab (76 eyes), ranibizumab (4 eyes), and aflibercept (3 dilation 2-fold thicker than the capillary diameter were defined as
eyes). We did not include any patients with a history of a vitrectomy, microaneurysms in this study. The number of microaneurysms in a
other retinal vascular diseases, severe cataracts, or DME previously 3"3-mm area of each vascular layer was manually counted by 2
treated with intravitreal or periocular corticosteroids. Eyes that could retinal specialists (J.L. and B.G.M.). The FAZ area was defined as
not be scanned using SD OCT or with poor OCTA images with a an avascular central area, and the border of the FAZ was manually
signal strength index <40 due to media opacity or significant motion drawn by 2 retinal specialists (J.L. and B.G.M.) who were blind to
artifact resulting from poor patient cooperation also were excluded all clinical information. These specialists started with the raw
from the study. We evaluated the baseline systemic characteristics of images, did their own segmentation correction, and then analyzed
patients, including diabetic duration, presence of hypertension, and the images. The intraclass correlation coefficient was used to
glycosylated hemoglobin, within the preceding 3 months. This study determine the interobserver agreement for the manually counted
was approved by the Institutional Review Board and Ethics Com- number of microaneurysms and measured FAZ area. Interobserver
mittee of Asan Medical Center. agreement, in terms of the manual measurements of micro-
aneurysms and FAZ area, was satisfactory (intraclass correlation
Optical Coherence Tomography and Optical coefficients of 0.913 and 0.934, respectively). The FAZ area and
Coherence Tomography Angiography vascular flow density of images from a 3"3-mm area of each
vascular layer were calculated after conversion of the images to 8-
We performed SD OCT (Spectralis; Heidelberg Engineering, bit grayscale images using ImageJ software, as previously
Heidelberg, Germany) at every visit including the baseline exam- described.16
ination. A custom 20! "20! volume acquisition protocol was used In SD OCT images, we measured the ODR of the largest
to obtain a set of high-speed scans from each eye. With this pro- intraretinal cyst to the vitreous humor using ImageJ 1.48 software
tocol, 25 horizontal and central vertical cross-sectional B-scan (National Institutes of Health, Bethesda, MD) as previously
images were obtained, each composed of 512 A-scans. To improve described but with minor modifications.17 The entire intraretinal

2369

Downloaded for Bowo Widiasmoko (bowo_wa@hotmail.com) at Universitas Kristen Duta Wacana from ClinicalKey.com by Elsevier on May 14, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
Ophthalmology Volume 123, Number 11, November 2016

cyst and the entire vitreous humor in the SD OCT image were However, the difference of the vascular density in the SCP
selected for the set measurement of the optical density. The between the DME and non-DME eyes did not quite reach
optical densities were extracted from the measured gray-level the level of significance. Next, we investigated whether the clinical
intensity of the corresponding region on a scale of 0 (pure black) stage of DR was associated with the flow density of the
to 255 (pure white). The selected areas were measured in pixels. DCP in DME. Compared with DME eyes with nonproliferative DR
The ODR was calculated from the measured optical densities (n ¼ 41), DME eyes with proliferative DR (n ¼ 42) had a larger mean
(Fig S2, available at www.aaojournal.org). The integrity of the FAZ area in the DCP (Table S5, available at www.aaojournal.org).
OPL was evaluated in all horizontal and vertical B-scan images. However, no other OCTA findings in the DCP and SCP were
We defined the OPL as being disrupted if the OPL was significantly different between the nonproliferative DR and
discontinued at any image among 25 B-scans and cysts from proliferative DR groups. Therefore, this discrepancy in the
both nuclear layers communicated across the OPL disruption. parameters of the DCP between the groups suggests that integrity of
All statistical analyses were performed using the Statistical the perifoveal DCP is involved in the pathogenesis of DME but is
Package for the Social Sciences (version 21.0; SPSS Inc., an IBM not associated with the clinical stage of DR.
Company, Chicago, IL). Continuous variables are presented as the
mean $ standard deviation. Comparisons between groups were Baseline Spectral-Domain Optical Coherence
evaluated using the Student t test, chi-square test, or Fisher exact Tomography Findings Associated with the
test, as appropriate. Analysis of variance with post hoc Tukey Response to AntieVascular Endothelial Growth
honest significant difference was used to compare the 3 groups. A
P value <0.05 was considered statistically significant. Factor Treatment
We investigated whether the baseline SD OCT findings were
associated with the response to anti-VEGF treatment in DME eyes.
Results First, we measured the ODR of the largest intraretinal cyst to the
vitreous. Compared with the good responder DME eyes, poor
Eye Characteristics responder DME eyes exhibited a higher ODR at both baseline and
A total of 103 eyes from 55 patients (32 male and 23 female) with after treatment (Table 6). Notably, in the poor responder DME
DR at different clinical stages were included in this study. The eyes, the ODR significantly increased after treatment with anti-
cases comprised 51 DME eyes with a poor response to anti-VEGF VEGF drugs (Table 6).
agents, and the age-matched controls comprised 32 DME eyes with Next, we closely observed the integrity of the OPL where the
a good response to anti-VEGF agents; the other 20 eyes were non- DCP is located. In total, 47 eyes of the 51 poor responder DME
DME fellow eyes of the cases and controls. Although the clinical eyes presented multiple instances of disruption of the synaptic OPL
stage of DR was more severe and a trend toward a higher preva- and thus intraretinal cyst communication between the inner and
lence of hypertension in the DME groups was evident, no other outer nuclear layers (Table 6, Fig 5). In contrast, only 1 eye of the
baseline demographics differed between the groups (Table 1). 32 anti-VEGF good responder DME eyes displayed these findings.
The topographic location of the disrupted OPL in SD OCT exactly
corresponded to the nonflow area of the DCP in OCTA (Fig 5F and
Association of the Deep Capillary Plexus with G). Notably, the disruption in the OPL was not recovered even
the Response to AntieVascular Endothelial after the edema improved, which suggests that there might be
Growth Factor Treatment permanent damage (Fig 5H). In patients with a relatively intact
DCP, the integrity of the OPL was preserved even in the
We first compared the OCTA findings between DME eyes that numerous retinal cysts in both the inner and outer nuclear layers
responded to anti-VEGF agents and those that poorly responded to (Fig 5AeD).
anti-VEGF agents. In both groups, the mean number of micro-
aneurysms and mean area of the FAZ were higher in the DCP than
in the SCP, but the mean vascular flow density was lower in the Discussion
DCP than in the SCP (Table 2, Figs 3 and 4). Compared with the
DME eyes considered to be anti-VEGF responders, poor responder Our current results suggest that microvascular damage in the
DME eyes exhibited more microaneurysms, lower vascular flow DCP is strongly associated with the treatment response to
density, a larger FAZ area in the DCP, and a lower flow density in
anti-VEGF agents in patients with DME. In contrast to
the TCP, whereas there were no significant differences between the
2 groups in these parameters in the SCP (Table 2). The discrepancy DME eyes responding well to anti-VEGF drugs, poor
in the DCP parameters between the 2 groups suggests that the responder DME eyes showed a significantly lower flow
integrity of the perifoveal DCP, but not that of the SCP, is density, larger number of microaneurysms, and larger area
associated with the treatment response to anti-VEGF agents. of the FAZ in the DCP. Although some previous OCTA
studies have reported more severe microvascular abnor-
Association of the Deep Capillary Plexus with malities in the DCP than in the SCP among eyes with DR or
the Occurrence of Diabetic Macular Edema DME,11,12 no study has previously compared the degree of
microvascular damage in relation to the treatment response
The aforementioned results encouraged us to compare the integrity to anti-VEGF agents.
of the DCP between DME and non-DME eyes. We selected 20 Although the exact pathogenesis of DME has not been
fellow eyes without a history of DME from 55 patients. There were
fully elucidated, the VEGF-mediated breakdown of the
significant differences in the number of microaneurysms and the
vascular flow density in the DCP and TCP between DME eyes and blooderetinal barrier seems to be the primary pathologic
non-DME eyes (Table 3). These differences were still significant process that triggers the subsequent inflammatory re-
even when the poor responder DME eyes were excluded to actions.18 Therefore, anti-VEGF therapy has been approved
avoid the possibility that the DME group might be biased toward as a primary modality in the treatment of DME via several
severe DME (Table S4, available at www.aaojournal.org). clinical trials.2,18 Nevertheless, there are still concerns about

2370

Downloaded for Bowo Widiasmoko (bowo_wa@hotmail.com) at Universitas Kristen Duta Wacana from ClinicalKey.com by Elsevier on May 14, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
Lee et al #
OCT Angiography of DME

Table 1. Demographic Information on the Diabetic Macular are absorbed by the inner retinal tissue.21 Thus, previous
Edema and Nondiabetic Macular Edema Eyes studies that used FA to investigate the microvascular
structure in DR have limitations in that they did not
Poor Good precisely observe the DCP, which is where the diabetic
Responder Responder
DME DME No DME P vascular change primarily occurs.
(n [ 51) (n [ 32) (n [ 20) Value Since OCTA has recently become available, the impor-
tance of the DCP in several retinal diseases has been
Mean age (SD), yrs 59.3 (8.8) 57.5 (10.3) 60.5 (8.6) 0.483
Range, yrs 39e79 39e79 44e74
emphasized. Nevertheless, the functional roles of the DCP
Male, n (%) 31 (60.8) 17 (53.1) 13 (65.0) 0.663 in terms of oxygen supply to the outer retina are still
Duration of 12.6 (7.0) 11.7 (8.0) 10.4 (7.5) 0.522 obscure. Although the outer one third of the retina is mainly
diabetes, yrs supplied by the choriocapillaris, the DCP provides approx-
Duration of DR, mos 20.4 (19.2) 14.2 (17.4) 13.7 (17.8) 0.217 imately 15% of the oxygen supply of the photoreceptors.22
Positive history of 22 (43.1) 13 (40.6) 3 (15.0) 0.076 A steady decrease in the gradient of oxygen from the
hypertension,
n (%) choriocapillaris to the photoreceptors has been
HbA1c, % 7.9 (1.7) 7.6 (1.7) 8.5 (1.9) 0.171 experimentally proven.23 In the absence of the DCP,
Clinical severity of 0/26/25 0/15/17 4/12/4 <0.001 bipolar cells in the inner nuclear layer are vulnerable to
DR, n (moderate hypoxic cell death, and electroretinography has revealed
NPDR/severe decreased post-photoreceptor responses.24 On the basis of
NPDR/PDR)
these previous observations, we can assume that the DCP
provides oxygen to the inner and outer nuclear layers.
DME ¼ diabetic macular edema; DR ¼ diabetic retinopathy; HbA1c ¼ Therefore, the OPL and the region just below it may act
glycosylated hemoglobin; NPDR ¼ nonproliferative diabetic retinopathy;
PDR ¼ proliferative diabetic retinopathy; SD ¼ standard deviation. as a watershed zone and could be especially vulnerable to
ischemic damage from the DCP.25 In this context, the
OPL at the region where the flow of DCP is absent could
be disrupted.
the cost-effectiveness of anti-VEGF drugs in chronic or anti- There are concerns about the low reliability of OCTA
VEGF-resistant cases because multiple injections might and its poor visualization of the DCP in patients with
place a considerable burden on diabetic patients and the macular edema. Indeed, a lack of visualization of flow in
social economy.4 To avoid ineffective repeated injections, it OCTA does not necessarily mean that flow is completely
is important to screen for anti-VEGF-resistant DME in absent. Spaide20 recently reported that the extent of flow
baseline examinations. Although several studies have void is not exactly matched with the cystoid space. The
evaluated the hard exudate or inflammatory status to cystoid space may compress the vessels or, conversely,
identify variables predicting the treatment response,3,19 we fluid may preferentially accumulate in the region of low
focused on the structure of the retinal blood vessel itself flow in the DCP. These mutual effects suggest that the
because VEGF targets the endothelial cells of blood vessels.
The underlying mechanism of the association between
the low-flow density in the DCP and treatment resistance to Table 2. Comparisons of Optical Coherence Tomography
anti-VEGF agents remains to be clearly defined. There are Angiography Parameters in Diabetic Macular Edema According to
several possible explanations for this association. One pos- the Response to AntieVascular Endothelial Growth Factor
sibility is that protein-rich intraretinal cysts alter the ability Treatment
of anti-VEGF agents to diffuse from the vitreous to the
DCP, where the hyperpermeable retinal capillaries are Poor Good
Responder Responder P
located. The second possible explanation is that abundant DME (n [ 51) DME (n [ 32) Value
expression of VEGF in the ischemic deep retina in the
absence of the DCP restricts the efficacy of anti-VEGF SCP
Microaneurysms, n 0.7 (0.9) 0.3 (0.7) 0.070
agents. A third possibility is that inhibition of VEGF Vascular flow 21.81 (3.71) 21.94 (3.34) 0.884
might not be able to restore absent or broken vessels density, %
because it produces a selective tightening effect on endo- Foveal 0.33 (0.16) 0.34 (0.18) 0.954
thelial junctions. Finally, the DCP might participate in the avascular area, mm2
removal of excess fluid from the retina20; its absence could DCP
Microaneurysms, n 8.9 (4.1) 3.9 (1.2) <0.001
predispose the retina to the localized accumulation of fluid.
Vascular flow 10.62 (4.66) 17.39 (3.98) <0.001
Thus, a smaller DCP would reduce fluid absorption, even density, %
though capillary permeability is inhibited by anti-VEGF Foveal avascular area, mm2 0.87 (0.41) 0.57 (0.22) <0.001
agents. TCP
Fluorescein angiography is considered the gold standard Vascular flow 24.69 (4.39) 26.63 (4.05) 0.043
for evaluating microvascular structure and its functional role density (SD), %
in retinal blood vessels. However, the ability of FA to
visualize the DCP is limited by the superposition of the SCP DCP ¼ deep capillary plexus; DME ¼ diabetic macular edema; SCP ¼
and overlying leakage.21 Signals from excited sodium superficial capillary plexus; TCP ¼ total capillary plexus.
Data are mean (standard deviation) unless otherwise indicated.
fluorescein in the DCP cannot penetrate retinal layers but

2371

Downloaded for Bowo Widiasmoko (bowo_wa@hotmail.com) at Universitas Kristen Duta Wacana from ClinicalKey.com by Elsevier on May 14, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
Ophthalmology Volume 123, Number 11, November 2016

Figure 3. A diabetic macular edema (DME) eye showing a good response to antievascular endothelial growth factor treatment. Baseline fundus
photography (A) and fluorescein angiography (FA) (B) of the eye. Optical coherence tomography angiography showing multiple microaneurysms in the
deep capillary plexus (D), but none in the superficial capillary plexus (C). Yellow circles indicate microaneurysms that were not depicted on FA. Red circles
indicate a microaneurysm that is also seen on the FA images. Compared with the baseline spectral-domain optical coherence tomography (SD OCT)
(E), the follow-up SD OCT (F) showed improved macular edema after 3 bevacizumab injections.

OCTA information is not valueless but can be useful to calculating the TCP for the entire retinal thickness to
investigate the pathogenesis of DME if carefully overcome segmentation issues. In addition, OCTA was
interpreted. In this study, we carefully adjusted the performed after treatments when the macular edema
boundary between the SCP and DCP in the eyes with improved in most of the eyes. Therefore, the bias from the
severe DME and confirmed the difference in the DCP by cystoid space was minimized in this analysis. Our results

Figure 4. A diabetic macular edema (DME) eye showing a poor response to antievascular endothelial growth factor treatment. Baseline fundus
photography (A) and fluorescein angiography (FA) (B) of the eye. The optical coherence tomography angiography showed microaneurysms in both the
superficial capillary plexus (SCP) (C) and the deep capillary plexus (DCP) (D). A lower vascular flow density was seen in the large foveal avascular area in
the DCP compared with the relatively preserved vascular flow density in the SCP. Yellow circles indicate microaneurysms not depicted by FA. Red circles
indicate microaneurysms that were also seen on the FA images. In the follow-up spectral-domain optical coherence tomography SD OCT, macular edema
was not significantly improved after 3 bevacizumab injections (F) compared with the baseline SD OCT (E).

2372

Downloaded for Bowo Widiasmoko (bowo_wa@hotmail.com) at Universitas Kristen Duta Wacana from ClinicalKey.com by Elsevier on May 14, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
Lee et al #
OCT Angiography of DME

Table 3. Comparisons of Optical Coherence Tomography Chronic inflammation with highly exudative DME might
Angiography Parameters in Diabetic Retinopathy with or without cause protein-rich intraretinal cysts that appear as hyper-
Diabetic Macular Edema reflective signals on SD OCT, as in other retinal diseases.17,29
DME No DME P
(n [ 83) (n [ 20) Value Study Limitations
SCP There are some noteworthy limitations of this retrospective
Microaneurysms, n 0.5 (0.9) 0 (0) <0.001 study. Our study series comprised patients with differing
Vascular flow 21.87 (3.55) 23.43 (3.43) 0.078 previous treatments and DME durations. Other confounding
density, %
Foveal avascular 0.34 (0.17) 0.31 (0.13) 0.551
factors including extrafoveal leaking microaneurysms or hard
area, mm2 exudates may have influenced the responsiveness to
DCP anti-VEGF agents. In addition, our quantification of the
Microaneurysms, n 6.2 (4.7) 0 (0) <0.001 microaneurysm number and FAZ area included a subjective
Vascular flow 13.23 (5.49) 20.75 (3.48) <0.001 process because no reliable software for this analysis is
density, % currently available. However, it must be noted that our
Foveal avascular 0.76 (0.38) 0.49 (0.14) <0.001
area, mm2 quantification methods demonstrated an excellent level of
TCP interobserver agreement. To obtain accurate flow maps in
Vascular flow 25.44 (4.34) 29.04 (4.13) 0.002 OCTA, patients must have good fixation to prevent signifi-
density (SD), % cant motion artifacts. Many patients with severe DME have
poor vision, restricting the usefulness of OCTA, and were
DCP ¼ deep capillary plexus; DME ¼ diabetic macular edema; SCP ¼ thus excluded from our analysis.
superficial capillary plexus; TCP ¼ total capillary plexus. In conclusion, we reveal that the integrity of the DCP is
Data are mean (standard deviation) unless otherwise indicated. important not only for the occurrence of DME but also for
the treatment response to anti-VEGF agents. Therefore, the
extent of DCP loss assessed by OCTA could be a useful
are further supported by Pearson correlation analysis diagnostic tool for predicting the treatment response to anti-
demonstrating that the density of the DCP was not VEGF agents. In addition, the disruption of the OPL in SD
significantly associated with retinal thickness or volume at OCT corresponds well with the extent of DCP loss and
baseline and after treatment (Table S7, available at thus could be an alternative predictor of anti-VEGF
www.aaojournal.org). Table 6. Comparisons of Spectral-Domain Optical Coherence
For the improvement of OCTA reliability, motion artifact Tomography Angiography Parameters in Diabetic Macular Edema
and reproducibility have been critical issues. In this study, According to the Response to Anti-Vascular Endothelial Growth
we overcame any limitations that may have derived from Factor Treatment
motion artifacts by excluding poor-quality images with a
low signal strength index. Stable fixation of patients and a Poor Good
precise centration technique at the follow-up examination Responder Responder P
DME (n [ 51) DME (n [ 32) Value
are necessary for OCTA reproducibility. Although the
excellent reproducibility of FAZ in healthy subjects has Baseline SD OCT findings
been reported,26 intraobserver and interobserver variability Central retinal 482.7 (149.3) 442.3 (164.6) 0.251
thickness, mm
in diseased eyes need to be assessed in future studies.
Total macular 11.8 (2.0) 11.0 (1.7) 0.066
In this study, as a predictor of the treatment responsive- volume, mm3
ness to anti-VEGF agents, we also showed OPL line ODR of intraretinal 3.6 (2.6)* 2.1 (1.3)y <0.001
disruption on SD OCT. Because OCTA is not yet widely cyst to vitreous
available, we concentrated on detecting the corresponding Positive in OPL 47 (90.4) 1 (3.1) <0.001
disruption, n (%)
feature of the DCP loss in SD OCT images in DME eyes.
Follow-up SD OCT findings
Given that the DCP is located at the junction between the after anti-VEGF treatments
inner nuclear layer and OPL, we hypothesized that blood Central retinal 493.7 (128.5) 328.9 (77.7) <0.001
flow in the DCP might be primarily responsible for the thickness, mm
integrity of the synaptic portion of the OPL where the syn- Total macular 11.5 (1.9) 9.9 (0.9) <0.001
apse between bipolar cells and photoreceptor cells occurs. volume, mm3
ODR of intraretinal 4.7 (3.4)* 2.2 (1.4)y <0.001
The disruption of the OPL in SD OCT topographically cor- cyst to vitreous
responded to the nonflow area in the DCP. Because the OPL Positive in OPL 49 (96.1) 2 (6.3) <0.001
acts as a fluid barrier,27 intraretinal cysts between the inner disruption, n (%)
and outer nuclear layers communicated through the
disrupted OPL. In addition, disruption of the layered retinal DME ¼ diabetic macular edema; ODR ¼ optical density ratio; OPL ¼
structure can predict visual acuity in eyes with DME.28 outer plexiform layer; SD OCT ¼ spectral-domain optical coherence
Thus, the association between DCP integrity and visual tomography.
prognosis will be evaluated in a future study. The other SD Data are mean (standard deviation) unless otherwise indicated.
*Poor responder ODR paired t test. P ¼ 0.046.
OCT finding associated with the response to anti-VEGF y
Good responder ODR paired t test. P ¼ 0.417.
treatment was the optical density of the intraretinal cysts.

2373

Downloaded for Bowo Widiasmoko (bowo_wa@hotmail.com) at Universitas Kristen Duta Wacana from ClinicalKey.com by Elsevier on May 14, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
Ophthalmology Volume 123, Number 11, November 2016

Figure 5. Comparison of the integrity of the outer plexiform layer (OPL) according to the response to antievascular endothelial growth factor treatment.
AeD, A good responder diabetic macular edema (DME) eye. EeH, A poor responder DME eye. The disrupted OPL in spectral-domain optical coherence
tomography (G and H, yellow dotted lines, arrows) exactly corresponded to the nonflow area (F, red dotted lines) of the deep capillary plexus in optical
coherence tomography angiography. H, Disruption of the OPL was not repaired, even after the edema improved with treatment. C and D, The integrity of
the OPL was preserved (yellow dotted lines), even in the numerous retinal cysts in the eye of a patient with good responder DME.

responsiveness in DME eyes. Our present findings could be 6. Byeon SH, Chu YK, Hong YT, et al. New insights into the
useful for further studies evaluating the longitudinal changes pathoanatomy of diabetic macular edema angiographic patterns
in the DCP after anti-VEGF treatment. and optical coherence tomography. Retina 2012;32:1087–99.
7. Wang HY, Chhablani J, Freeman WR, et al. Characterization
of diabetic microaneurysms by simultaneous fluorescein
angiography and spectral-domain optical coherence tomogra-
References phy. Am J Ophthalmol 2012;153:861–7.
8. Horu T, Murakami T, Nishijima K, et al. Relationship between
fluorescein pooling and optical coherence tomographic
1. Yau JW, Rogers SL, Kawasaki R, et al. Global prevalence and reflectivity of cystoid spaces in diabetic macular edema.
major risk factors of diabetic retinopathy. Diabetes Care Ophthalmology 2012;119:1047–55.
2012;35:556–64. 9. Ito H, Horii T, Nishijima K, et al. Association between fluo-
2. Nguyen QD, Brown DM, Marcus DM, et al. Ranibizumab for rescein leakage and optical coherence tomographic character-
diabetic macular edema: results from 2 phase III randomized istics of microaneurysms in diabetic retinopathy. Retina
trials: RISE and RIDE. Ophthalmology 2012;119:789–801. 2013;33:732–9.
3. Lazic R, Lukic M, Boras I, et al. Treatment of anti-vascular 10. Spaide RF, Klancnik J, Cooney MJ. Retinal vascular layers
endothelial growth factor-resistant diabetic macular edema with imaged by fluorescein angiography and optical coherence to-
dexamethasone intravitreal implant. Retina 2014;34:719–24. mography angiography. JAMA Ophthalmol 2015;133:45–50.
4. Gonder JR, Walker VM, Barbeau M, et al. Costs and quality of 11. Ishibazawa A, Nagaoka T, Takahashi A, et al. Optical coher-
life in diabetic macular edema: Canadian Burden of Diabetic ence tomography angiography in diabetic retinopathy: a pro-
Macular Edema Observational Study (C-REALITY). spective pilot study. Am J Ophthalmol 2015;160:35–44.e1.
J Ophthalmol 2014;2014:939315. 12. Agemy SA, Scripsema NK, Shah CM, et al. Retinal vascular
5. Horii T, Murakami T, Nishijima K, et al. Optical coherence perfusion density mapping using optical coherence tomogra-
tomographic characteristics of microaneurysms in diabetic phy angiography in normals and diabetic retinopathy patients.
retinopathy. Am J Ophthalmol 2010;150:840–8. Retina 2015;35:2353–63.

2374

Downloaded for Bowo Widiasmoko (bowo_wa@hotmail.com) at Universitas Kristen Duta Wacana from ClinicalKey.com by Elsevier on May 14, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.
Lee et al #
OCT Angiography of DME

13. Couturier A, Mane V, Bonnin S, et al. Capillary plexus 22. Birol G, Wang S, Budzynski E, et al. Oxygen distribution and
anomalies in diabetic retinopathy on optical coherence to- consumption in the macaque retina. Am J Physiol Heart Circ
mography angiography. Retina 2015;35:2384–91. Physiol 2007;293:H1696–704.
14. Matsunaga DR, Yi JJ, De Koo LO, et al. Optical coherence 23. Wangsa-Wirawan ND, Linsenmeier RA. Retinal oxygen: funda-
tomography angiography of diabetic retinopathy in human mental and clinical aspects. Arch Ophthalmol 2003;121:547–57.
subjects. Ophthalmic Surg Lasers Imaging Retina 2015;46: 24. Lee J, Kim KE, Choi DK, et al. Angiopoietin-1 guides
796–805. directional angiogenesis through integrin alphavbeta5
15. Wu L, Fernandez-Loaiza P, Sauma J, et al. Classification of signaling for recovery of ischemic retinopathy. Sci Transl Med
diabetic retinopathy and diabetic macular edema. World J 2013;5:203ra127.
Diabetes 2013;4:290–4. 25. Sarraf D, Rahimy E, Fawzi AA, et al. Paracentral acute middle
16. Hartig SM. Basic image analysis and manipulation in ImageJ. maculopathy: a new variant of acute macular neuroretinopathy
Curr Protoc Mol Biol 2013;Chapter 14:Unit14.15. associated with retinal capillary ischemia. JAMA Ophthalmol
17. Baek J, Park YH. Optical density ratio in the subretinal fluid: 2013;131:1275–87.
differentiating chronic central serous chorioretinopathy and 26. Carpineto P, Mastropasqua R, Marchini G, et al. Reproduc-
polypodial choroidal vasculopathy. Am J Ophthalmol ibility and repeatability of foveal avascular zone measurements
2015;159:386–92. in healthy subjects by optical coherence tomography angiog-
18. Das A, McGuire PG, Rangasamy S. Diabetic macular edema: raphy. Br J Ophthalmol 2016;100:671–6.
pathophysiology and novel therapeutic targets. Ophthalmology 27. Antcliff RJ, Hussain AA, Marshall J. Hydraulic conductivity
2015;122:1375–94. of fixed retinal tissue after sequential excimer laser ablation:
19. Jeon S, Lee WK. Effect of intravitreal bevacizumab on dia- barriers limiting fluid distribution and implications for cystoid
betic macular edema with hard exudates. Clin Ophthalmol macular edema. Arch Ophthalmol 2001;119:539–44.
2014;8:1479–86. 28. Sun JK, Lin MM, Lammer J, et al. Disorganization of the
20. Spaide RF. Volume-rendered optical coherence tomography of retinal inner layers as a predictor of visual acuity in eyes with
diabetic retinopathy pilot study. Am J Ophthalmol 2015;160: center-involved diabetic macular edema. JAMA Ophthalmol
1200–10. 2014;132:1309–16.
21. Mendis KR, Balaratnasingam C, Yu P, et al. Correlation of 29. Kashani AH, Cheung AY, Robinson J, Williams GA. Longi-
histologic and clinical images to determine the diagnostic tudinal optical density analysis of subretinal fluid after surgical
value of fluorescein angiography for studying retinal capillary repair of rhegmatogenous retinal detachment. Retina 2015;35:
detail. Invest Ophthalmol Vis Sci 2010;51:5864–9. 149–56.

Footnotes and Financial Disclosures


Originally received: February 17, 2016. Author Contributions:
Final revision: June 24, 2016. Conception and design: Lee, Yoon
Accepted: July 10, 2016. Data collection: Lee, Moon, Cho, Yoon
Available online: September 6, 2016. Manuscript no. 2016-331.
1 Analysis and interpretation: Lee, Moon, Yoon
Department of Ophthalmology, Yeungnam University, College of Medi-
cine, Daegu, Korea. Obtained funding: Yoon
2
Department of Ophthalmology, Asan Medical Center, University of Overall responsibility: Lee, Yoon
Ulsan, College of Medicine, Seoul, Korea. Abbreviations and Acronyms:
3
Asan Diabetes Center, Asan Medical Center, University of Ulsan, College DCP ¼ deep capillary plexus; DME ¼ diabetic macular edema;
of Medicine, Seoul, Korea. DR ¼ diabetic retinopathy; FA ¼ fluorescein angiography; FAZ ¼ foveal
Presented in part at: Floretina 2015, December 13, 2015, Florence, Italy. avascular zone; IPL ¼ inner plexiform layer; OCT ¼ optical coherence
tomography; OCTA ¼ optical coherence tomography angiography;
Financial Disclosure(s): ODR ¼ optical density ratio; OPL ¼ outer plexiform layer;
The author(s) have made the following disclosure(s): Y.H.Y.: Grants e SCP ¼ superficial capillary plexus; SD ¼ spectral-domain; TCP ¼ total
Ministry of Science, ICT, and Future Planning, Republic of Korea;
capillary plexus; VEGF ¼ vascular endothelial growth factor.
Consultant/speaker e Alcon; Advisory board/consultant/investigator/
speaker e Allergan; Advisory board/consultant/investigator/speaker e Correspondence:
Bayer, outside the submitted work. Young Hee Yoon, MD, PhD, Department of Ophthalmology, Asan Medical
Center, University of Ulsan, College of Medicine, 88 Olympic-ro 43-gil,
Supported by grants from the Ministry of Science, ICT and Future Plan-
ning, Republic of Korea (NRF-2013R1A2A2A01068457). Songpa-gu, Seoul 138-736, Korea. E-mail: yhyoon@amc.seoul.kr.

2375

Downloaded for Bowo Widiasmoko (bowo_wa@hotmail.com) at Universitas Kristen Duta Wacana from ClinicalKey.com by Elsevier on May 14, 2018.
For personal use only. No other uses without permission. Copyright ©2018. Elsevier Inc. All rights reserved.