bs_bs_banner
Clinical and Experimental Ophthalmology 2015; 43: 735–741 doi: 10.1111/ceo.12557
Original Article Diagnostic accuracy of disorganization of the retinal inner layers in detecting macular capillary non-perfusion in diabetic retinopathy Luke Nicholson FRCOphth, Jayashree Ramu MBBS, Ioanna Triantafyllopoulou MBBS, Namritha V Patrao MBBS, Oliver Comyn MD MRCOphth, Philip Hykin MD FRCOphth and Sobha Sivaprasad MD FRCS NIHR Moorfields Biomedical Research Centre, London, United Kingdom
ABSTRACT Background: Disorganization of the retinal inner layers (DRIL) on optical coherence tomography (OCT) is thought to represent retinal capillary nonperfusion (CNP) in eyes with diabetic retinopathy. This study was designed to evaluate the ability of DRIL to accurately predict CNP. Design: Retrospective masked reliability and diagnostic accuracy study performed in the National Institute for Health Research (NIHR) Moorfields Biomedical Research Centre, London, UK Samples: Retinal images of patients with diabetic retinopathy Methods: The OCT images from 90 separate areas of angiographically confirmed perfused and nonperfused areas of the macula from 37 eyes of 31 patients were anonymized and coded. Two masked graders independently graded these OCT scans for the presence or absence of DRIL to determine the intergrader reliability. The diagnostic accuracy of DRIL in identifying CNP was evaluated from the results obtained. Main Outcome Measures: Sensitivity and specificity of DRIL in accurately detecting CNP Results: The intergrader agreement was high with a Cohen’s kappa of 0.909. DRIL was present in 84.4%
(38/45) of non-perfused retina and none in perfused retina (0/45). The sensitivity and specificity of DRIL in detecting angiographic evidence of CNP was 84.4% and 100%, respectively. The positive predictive value was 100% and the negative predictive value was 86.5%. Conclusions: The presence of DRIL is a reliable predictor of areas of macular CNP. However, DRIL is not a universal finding of non-perfusion, with some cases exhibiting absence of DRIL despite angiographic CNP. Key words: diabetic retinopathy, diagnostic study, retinal-imaging.
INTRODUCTION Macular capillary non-perfusion (CNP) is an important cause of visual impairment in patients with diabetes mellitus and is a predictor of peripheral CNP.1,2 Approximately 55% of patients presenting with diabetic macular oedema (DMO) have co-existent macular CNP.1 At present the gold standard for diagnosis of CNP is fluorescein angiography (FA). Areas of CNP on FA are seen as hypofluorescent areas within the normal ground-glass appearance of perfused retina. Co-existent DMO may mask the angiographic evidence of CNP. Angiographic evidence of macular CNP indicates a decrease or absence of retinal blood circulation to the inner retina, which in turn may compromise inner retinal
■ Correspondence: Mr Luke Nicholson, NIHR Moorfields Biomedical Research Centre, Clinical Research Facility, 162 City Road, Moorfields Eye Hospital, London, EC1V 2PD, UK. Email: [email protected] Received 10 April 2015; accepted 18 May 2015. Conflict of interest: None. Funding sources: None. © 2015 Royal Australian and New Zealand College of Ophthalmologists
736
Nicholson et al.
integrity. However the development of spectral domain optical coherence tomography (SD-OCT) has enabled better visualization of macular DMO such that, diagnosis and treatment of DMO is typically OCT-guided at least in part. This and the recognized risks of FA have led to a general decline in its use which is likely to be accentuated if spectral domain OCT features such as disorganization of the retinal inner layers (DRIL) are established as reliable indicators of CNP. The International Nomenclature for Optical Coherence Tomography [IN • OCT] Panel has recently published a useful standardized nomenclature system for normal anatomic landmarks seen on SD-OCT.3 The various layers of the retina are now well characterized on SD-OCT, and changes in the morphology of these layers serve as valuable diagnostic and prognostic parameters. The SD-OCT has been used to evaluate retinal CNP.4 Sun et al. described the concept of foveal DRIL on OCT as a surrogate for poor visual outcome.5 However, the diagnostic accuracy of DRIL in detecting macular CNP is unknown. This is of particular importance in the era of anti-vascular endothelial growth factor treatment for DMO as SD-OCT is the most common imaging tool now used in DMO management. The objective of this study was to assess the sensitivity and specificity of DRIL in detecting macular CNP and to determine factors associated with DRIL that may account for its diagnostic accuracy.
METHODS We performed a post hoc analysis of images obtained from participants of completed randomized clinical trials on DMO conducted in National Institute for Health Research (NIHR) Moorfields Biomedical Research Centre, London, UK. Local institutional review board approval was obtained and the study was conducted in accordance to the tenets of the Declaration of Helsinki. The FA and SD-OCT images of consecutive patients with severe non-proliferative or proliferative diabetic retinopathy were included. Data on patient demographics and treatment history were available.
Image acquisition The imaging protocols for acquisition of FA and SD-OCT were similar for every patient included in this study.
FA FA was performed with an intravenous infusion of 5 mL fluorescein sodium 20%. A series of 35 degree images were obtained with a Topcon TRC 50 DX Fundus camera (Topcon, Tokyo, Japan). Standard 35-degree 7-field colour fundus image and FA images were available for both eyes. A 35-degree macular centred colour fundus and fluorescein angiography image of both eyes of each patient were used for this analysis.
SD-OCT All individuals underwent macular SD-OCT scans in both eyes. These were performed using Spectralis HRA + OCT (Heidelberg Engineering, Germany) which combines high-resolution SD-OCT with scanning laser ophthalmoscopy. The posterior pole volume scan was performed in both eyes on a 20 × 20 degree cube, with 49 raster lines, each containing 1064 pixels. Only scans with good signal strength (signal-to-noise ratio, 20 dB or greater) were evaluated. All scans were centred at the fovea to avoid any fixation errors and all segmentation errors were manually corrected.
Definition of macular CNP on FFA Using FA, CNP was defined as the absence of retinal arterioles and/or capillaries and was detected by characteristics such as a pruned appearance of adjacent arterioles and a darker appearance of the choroid6 (Fig. 1). Perfused retina was defined as a ground-glass appearance of normal angiogram. The colour fundus images were reviewed for cotton wool spots and retinal haemorrhages that may mimic nonperfusion due to blocked fluorescence during FA and these were excluded. Figure 1. Left image – fluorescein angiogram of the right eye. Capillary non-perfusion (CNP) with marked hypofluorescence and loss of visible capillary perfusion on fluorescein angiography marked in green and area in pink is a laser scar which is excluded from analysis.
© 2015 Royal Australian and New Zealand College of Ophthalmologists
DRIL to detect capillary non-perfusion
737
Image processing
Factors associated with DRIL
A single investigator (LN) identified 45 separate areas of macular CNP on the FA images (excluding the foveal avascular zone) and marked, coded and anonymized the associated SD-OCT line scans of these identified areas from 37 eyes of 31 patients. Similarly, a perfused area in a corresponding topographic location equidistant from the foveal centre in the same eye was chosen as reference in each study eye (Fig. 2). The SD-OCT scans of these areas were also marked, coded and anonymized.
We studied three factors that may be associated with DRIL that may help understand its diagnostic accuracy. These include the retinal thickness of the area of DRIL, area of CNP associated with DRIL and the duration of CNP in areas of DRIL. The retinal thickness in the area of perfused and non-perfused areas was measured by repositioning the Early Treatment Diabetic Retinopathy Study (ETDRS) grid to these areas and measuring the central subfield thickness in these areas. The absolute difference in retinal thickness between perfused and non-perfused areas was calculated. The area of CNP was measured using the ‘measure image’ tool in the Topcon IMAGEnet 2000 software (Topcon, Tokyo, Japan). Using previous FA images, the minimum duration of CNP was estimated from the first visible presence of CNP in the studied area.
Definition of DRIL Presence of DRIL is defined as the inability to identify any boundaries between the ganglion cell–inner plexiform layer complex, inner nuclear layer, and outer plexiform layer (Fig. 3). This definition of DRIL has been adopted from Sun et al., with the only exclusion being the horizontal extent in microns was not measured.5 The definitions of the hyper/hypo reflective layers in the OCT images were based on the proposed lexicon for the anatomical landmarks in retinal OCT imaging by the IN-OCT concensus.3
Inter-grader and intra-grader reliability The sequencing of the 90 anonymized images was ‘shuffled’ so that images with perfusion and nonperfusion of each eye were not contiguous. Two masked investigators (JR, NP) graded the images for the presence of DRIL. Arbitration was provided by a senior clinician (SS). The intergrader agreement was assessed. The two masked investigators repeated the grading process of the same set of images 6 months later and the intra-grader agreement was assessed.
Figure 2. A, Area of capillary non-perfusion; B, area of capillary perfusion equidistant; X, mm from the fovea.
Evolution of DRIL with time and treatment To further study the evolution of DRIL, the presence of DRIL on SD-OCT in the exact same location was assessed at 12 ± 1 months (−12 months) previously. The SD-OCT must have been performed using the same OCT machine and parameters, that is, Spectralis HRA + OCT volume scan and 20 × 20 degree cube with 49 raster lines each containing 1064 pixels. The images must be referenced to allow
Figure 3. OCT image with clear evidence of normal inner retinal boundaries and DRIL present temporally.
© 2015 Royal Australian and New Zealand College of Ophthalmologists
738
Nicholson et al.
for direct comparison. Angiographic presence of CNP at or before 12 months prior was also confirmed. Data on macular treatment received during this 12-month period were collected. This was grouped to either intravitreal anti-VEGF, intravitreal dexamethasone implant (Ozurdex, Alergan, Irvine, California, USA) or no intravitreal treatment.
Statistical analysis Statistical analysis was performed using Microsoft Excel (Microsoft Corp.) and GraphPad Prism (GraphPad Software Inc., La Jolla, California, USA). Descriptive statistics were used to present the data. Sample size was determined based on the calculation recommended by Flahault et al.7 For a sensitivity of 0.90, with a 0.95 probability of the lower 95% confidence limit of 0.75, the approximate number of images required is 70. A 2 × 2 contingency table protocol was used to evaluate the sensitivity, specificity and positive and negative predictive values of DRIL in the diagnosis of CNP. Sensitivity represents the percentage of true positives in all patients with the reference angiogram and specificity represents the percentage of true negatives. Positive and negative predictive values reflect the percentage of time that a positive or negative test (respectively) accurately captures the diagnosis of CNP. The confidence intervals for the sensitivity and specificity were computed via an online calculator at http:// www.pedro.org.au/english/downloads/confidence -interval-calculator/ using the Wilson score method. The number needed to diagnose was derived from the formula 1/[sensitivity − (1-specificity)] and represents the number of tests that need to be performed for DRIL to accurately diagnose the presence of CNP. Cohen’s kappa (k) coefficients and per cent agreement between examiners were used as estimates of inter-grader and intra-grader reliability for diagnosis of DRIL. Kappa values were interpreted according to the guidelines proposed by Landis and Koch.8 Almost perfect agreement is described for values between 0.81 and 1.00, substantial agreement for values between 0.61 and 0.80, moderate for values between 0.41 and 0.60, fair for values between 0.21 and 0.40, and slight for values under 0.20. The association of DRIL with CNP was assessed using the 2 × 2 contingency table and the Fisher’s exact test for significance. Comparisons between retinal thickness of areas of non-perfusion and perfusion was analysed using a Wilcoxon rank-sum test. Comparison between the sub-field thickness, area of CNP on FA and duration of CNP in areas with DRIL and without DRIL were analysed using the Mann–Whitney U test.
There was excellent inter-grader agreement between two masked investigators with a Cohen’s kappa of 0.909 and 95.56% observed agreements. The intra-grader agreement was 0.932 with a 96.67% agreement. Table 1 shows the 2 × 2 contingency table representing the grading outcomes of the 90 images for presence or absence of DRIL on OCT with the corresponding presence of CNP on fluorescein angiography. DRIL was significantly associated with CNP (P = 0.001). DRIL was absent on SD-OCT in 7/45 (15.6%) areas of CNP (Fig. 4). The sensitivity of DRIL as a predictive tool for CNP was 84.4% (95% CI 71.22% to 92.25%) with a specificity of 100% (95% CI 92.13% to 100%). The positive predictive value and negative predictive value were 100% and 86.5%, respectively. The number needed to diagnose was 1.18. Seven patients with angiographic evidence of CNP were graded as not having DRIL. These images were then re-examined in detail after arbitration, and it was confirmed that DRIL was absent in these cases.
Factors associated with DRIL The mean thickness of non-perfused retina was 333.27μ ± 87.62 and for the corresponding perfused retina, 311.13μ ± 56.07 (P = 0.21). For the 45 areas with CNP, the characteristics for areas with presence of DRIL and absence of DRIL are described in Table 2.
Evolution of DRIL with time and treatment From the 45 areas of angiographic CNP, 44 areas fulfilled the predetermined SD-OCT imaging requirements. All 44 areas selected for this study had evidence of angiographic CNP of 12 or more months, and 33 out of these 44 areas (75%) also had presence of DRIL of at least 12 months duration. Twenty-one of these 33 areas were not exposed to any intravitreal treatment while the rest of the 11 areas were exposed Table 1.
DRIL in detecting angiographic CNP Presence of DRIL on OCT Total Yes
No
38 0 38
7 45 52
RESULTS
Angiographic evidence Yes of capillary No non-perfusion Total
The mean age of the patients included was 61.94 ± 11.49 years. The male to female ratio was 28:3.
CNP, capillary non-perfusion; DRIL, disorganization of the retinal inner layers; OCT, optical coherence tomography.
45 45 90
© 2015 Royal Australian and New Zealand College of Ophthalmologists
DRIL to detect capillary non-perfusion
739
Figure 4. (a, c, e) fluorescein angiography with the green box corresponding to SD-OCT scanned area and single horizontal line corresponding to SD-OCT image shown on the right. Area of capillary non-perfusion marked in purple. (b, d, f) SD-OCT line scan with vertical line corresponding to border of non-perfusion with DRIL.
Table 2.
Comparison between DRIL present and DRIL absent areas of CNP
Absolute retinal thickness difference between perfused and non-perfused retina Area of non-perfusion Duration of non-perfusion
DRIL present (n = 38)
DRIL absent (n = 7)
P-value
60.53 μm ± 56.5
18.0 μm ± 19.22
0.01
9.43 mm2 27.68 months
3.88 mm2 20.57 months
0.17 0.81
CNP, capillary non-perfusion; DRIL, disorganization of the retinal inner layers.
to intravitreal treatment (11 to intravitreal Ozurdex and one with multiple injections of an anti-VEGF agent). There were four angiographic areas of CNP that were DRIL negative that developed DRIL over 12 months and one of these areas was also exposed to repeated injections of anti-VEGF agent. The remaining seven areas with angiographic evidence of CNP with absence of DRIL on OCT remained DRIL negative throughout the previous 12 months, and one of these areas was exposed to intravitreal Ozurdex.
DISCUSSION This study found that DRIL can be used to detect macular CNP in patients with diabetic retinopathy. The diagnostic sensitivity was 84.4% and specificity was 100%. In addition, we achieved a strong intergrader agreement suggesting that detection of DRIL is easy, reliable and an accurate parameter of CNP. The presence of DRIL is also a consistent finding with no evidence of reversal or normalization observed over a 12-month period.
© 2015 Royal Australian and New Zealand College of Ophthalmologists
740
Nicholson et al.
From a clinical perspective, the presence of DRIL is a very useful alternative to performing FA to identify CNP as DRIL is a definite indicator of nonperfused retina. There were no false positives. However, it is useful to consider the clinical implications of absence of DRIL in the presence of angiographic evidence of CNP (false negatives). With regard to false negatives, 7/45 (25.6%) angiographic evidence of CNP did not manifest as DRIL. Therefore, the absence of DRIL does not rule out CNP. This may suggest that the disorganization of retinal layers may only occur after a period of CNP or it is a sequelae of CNP. It is useful to understand the time to onset of DRIL from the development of CNP because DRIL is associated with poor visual outcome. Sun et al. reported the relationship between foveal disorganization of retinal inner layers (foveal DRIL) on SD-OCT and poor visual acuity in DMO.5 We believe macular ischemia and an enlarged foveal avascular zone may contribute to foveal DRIL and reduced visual acuity. This was not formerly assessed in the aforementioned paper. In our study, we have associated DRIL with nonperfusion and thus foveal DRIL may in fact be related to an enlarged foveal avascular zone or macular ischaemia. The association between macular ischaemia and poor visual acuity is known.1,9 Treatment trials on macular CNP should ideally be initiated before the onset of DRIL. We further explored the factors that may be associated with the presence or absence of DRIL in areas of macular CNP to better understand the evolution of DRIL. However, as the study was not set out to specifically study these factors, the numbers are limited and larger numbers may be required to confirm our observations. There is a significant relationship between larger absolute retinal thickness differences between perfused and non-perfused retina in DRIL positive and DRIL negative areas. Areas with DRIL on OCT were on average 60.53 μ thicker or thinner than corresponding perfused areas within the same eye. As for DRIL negative non-perfused areas, this difference only measured 18 μ. A thinner retina is associated with degenerate tissue10,11, and thicker retinas possibly relate to an ongoing ischaemic process.11 These two possibilities may be used to explain the finding we have observed in our study. DRIL is likely to be present when a significant level of ischaemia has been reached. In areas that do not present a significant difference in thickness between perfused and non-perfused capillaries, DRIL is likely to be absent despite angiographic evidence of non-perfusion. Therefore, it may be possible that the retina has not suffered significant changes to manifest DRIL, or adaptive processes may have been sufficient to maintain the integrity of the retinal structure. We
acknowledge the limitations in the study in not having microperimetry data to evaluate the function of these areas. A further limitation of the results and its application lies in the fact that the SD-OCT image needs to be of good quality for the clinician to confidently assess for DRIL. Although not statistically significant, we found a trend towards larger areas of non-perfusion being more likely to exhibit DRIL as opposed to smaller areas. The duration of retinal non-perfusion did not differ significantly between DRIL positive and DRIL negative images of angiographic non-perfused retina. We appreciate the method used in obtaining the duration of non-perfusion was highly dependent on previous imaging. Some subjects that presented late to the medical retina service may have had an area of non-perfusion that was only detected recently and provide a falsely shorter duration of CNP. Swept source OCT technology is one approach by which wide field OCT images may be obtained.12 Extrapolating the use of DRIL to the peripheral retina may enable us to identify peripheral nonperfusion with the use of wide field OCT. We acknowledge that non-invasive fluorescein angiography such as the speckle variance OCT will be extremely useful in identifying ischaemia13; however, currently it does not cover a large area of the macula and as yet is not widely used in clinical practice. DRIL is not only evident in CNP and its presence has been reported in acute retinal necrosis14 and closed globe trauma.15 It corroborates the theory that the disorganization is a manifestation of compromise to the inner retinal circulation. In summary, we have presented evidence to confirm the association between DRIL and CNP and suggest that DRIL is a predictive tool in identifying macular CNP but its absence does not rule out CNP.
REFERENCES 1. Sim DA, Keane PA, Zarranz-Ventura J et al. The effects of macular ischemia on visual acuity in diabetic retinopathy. Invest Ophthalmol Vis Sci 2013; 54: 2353–60. 2. Sim DA, Keane PA, Rajendram R et al. Patterns of peripheral retinal and central macula ischemia in diabetic retinopathy as evaluated by ultra-widefield fluorescein angiography. Am J Ophthalmol 2014; 158: 144– 153.e1. 3. Staurenghi G, Sadda S, Chakravarthy U, Spaide RF. International Nomenclature for Optical Coherence Tomography (IN•OCT) Panel. Proposed lexicon for anatomic landmarks in normal posterior segment spectral-domain optical coherence tomography: the IN•OCT consensus. Ophthalmology 2014; 121: 1572–8. 4. Unoki N, Nishijima K, Sakamoto A et al. Retinal sensitivity loss and structural disturbance in areas of
© 2015 Royal Australian and New Zealand College of Ophthalmologists
DRIL to detect capillary non-perfusion
5.
6.
7.
8.
9.
10.
capillary nonperfusion of eyes with diabetic retinopathy. Am J Ophthalmol 2007; 144: 755–60. Sun JK, Lin MM, Lammer J et al. Disorganization of the retinal inner layers as a predictor of visual acuity in eyes with center-involved diabetic macular edema. JAMA Ophthalmol 2014; 132: 1309–16. Chan CK, Ip MS, Vanveldhuisen PC, SCORE Study Investigator Group et al. SCORE Study report #11: incidences of neovascular events in eyes with retinal vein occlusion. Ophthalmology 2011; 118: 1364–72. Flahault A, Cadilhac M, Thomas G. Sample size calculation should be performed for design accuracy in diagnostic test studies. J Clin Epidemiol 2005; 58: 859– 62. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33: 159– 74. Lee DH, Kim JT, Jung DW, Joe SG, Yoon YH. The relationship between foveal ischemia and spectraldomain optical coherence tomography findings in ischemic diabetic macular edema. Invest Ophthalmol Vis Sci 2013; 54: 1080–5. Reznicek L, Kernt M, Haritoglou C, Kampik A, Ulbig M, Neubauer AS. In vivo characterization of ischemic
741
11.
12.
13.
14.
15.
© 2015 Royal Australian and New Zealand College of Ophthalmologists
retina in diabetic retinopathy. Clin Ophthalmol 2010; 5: 31–5. Murthy RK, Grover S, Chalam KV. Sequential spectral domain OCT documentation of retinal changes after branch retinal artery occlusion. Clin Ophthalmol 2010; 4: 327–9. Reznicek L, Klein T, Wieser W et al. Megahertz ultrawide-field swept-source retina optical coherence tomography compared to current existing imaging devices. Graefes Arch Clin Exp Ophthalmol 2014; 252: 1009–16. Xu J, Han S, Balaratnasingam C et al. Retinal angiography with real-time speckle variance optical coherence tomography. Br J Ophthalmol 2015; doi: 10.1136/ bjophthalmol-2014-306010. [Epub ahead of print]. Suzuki J, Goto H, Minoda H, Iwasaki T, Sakai J, Usui M. Analysis of retinal findings of acute retinal necrosis using optical coherence tomography. Ocul Immunol Inflamm 2006; 14: 165–70. Chen H, Lu Y, Huang H, Zheng J, Hou P, Chen W. Prediction of visual prognosis with spectral-domain optical coherence tomography in outer retinal atrophy secondary to closed globe trauma. Retina 2013; 33: 1258–62.
Clinical and Experimental Ophthalmology 2015; 43: 735–741 doi: 10.1111/ceo.12557
Original Article Diagnostic accuracy of disorganization of the retinal inner layers in detecting macular capillary non-perfusion in diabetic retinopathy Luke Nicholson FRCOphth, Jayashree Ramu MBBS, Ioanna Triantafyllopoulou MBBS, Namritha V Patrao MBBS, Oliver Comyn MD MRCOphth, Philip Hykin MD FRCOphth and Sobha Sivaprasad MD FRCS NIHR Moorfields Biomedical Research Centre, London, United Kingdom
ABSTRACT Background: Disorganization of the retinal inner layers (DRIL) on optical coherence tomography (OCT) is thought to represent retinal capillary nonperfusion (CNP) in eyes with diabetic retinopathy. This study was designed to evaluate the ability of DRIL to accurately predict CNP. Design: Retrospective masked reliability and diagnostic accuracy study performed in the National Institute for Health Research (NIHR) Moorfields Biomedical Research Centre, London, UK Samples: Retinal images of patients with diabetic retinopathy Methods: The OCT images from 90 separate areas of angiographically confirmed perfused and nonperfused areas of the macula from 37 eyes of 31 patients were anonymized and coded. Two masked graders independently graded these OCT scans for the presence or absence of DRIL to determine the intergrader reliability. The diagnostic accuracy of DRIL in identifying CNP was evaluated from the results obtained. Main Outcome Measures: Sensitivity and specificity of DRIL in accurately detecting CNP Results: The intergrader agreement was high with a Cohen’s kappa of 0.909. DRIL was present in 84.4%
(38/45) of non-perfused retina and none in perfused retina (0/45). The sensitivity and specificity of DRIL in detecting angiographic evidence of CNP was 84.4% and 100%, respectively. The positive predictive value was 100% and the negative predictive value was 86.5%. Conclusions: The presence of DRIL is a reliable predictor of areas of macular CNP. However, DRIL is not a universal finding of non-perfusion, with some cases exhibiting absence of DRIL despite angiographic CNP. Key words: diabetic retinopathy, diagnostic study, retinal-imaging.
INTRODUCTION Macular capillary non-perfusion (CNP) is an important cause of visual impairment in patients with diabetes mellitus and is a predictor of peripheral CNP.1,2 Approximately 55% of patients presenting with diabetic macular oedema (DMO) have co-existent macular CNP.1 At present the gold standard for diagnosis of CNP is fluorescein angiography (FA). Areas of CNP on FA are seen as hypofluorescent areas within the normal ground-glass appearance of perfused retina. Co-existent DMO may mask the angiographic evidence of CNP. Angiographic evidence of macular CNP indicates a decrease or absence of retinal blood circulation to the inner retina, which in turn may compromise inner retinal
■ Correspondence: Mr Luke Nicholson, NIHR Moorfields Biomedical Research Centre, Clinical Research Facility, 162 City Road, Moorfields Eye Hospital, London, EC1V 2PD, UK. Email: [email protected] Received 10 April 2015; accepted 18 May 2015. Conflict of interest: None. Funding sources: None. © 2015 Royal Australian and New Zealand College of Ophthalmologists
736
Nicholson et al.
integrity. However the development of spectral domain optical coherence tomography (SD-OCT) has enabled better visualization of macular DMO such that, diagnosis and treatment of DMO is typically OCT-guided at least in part. This and the recognized risks of FA have led to a general decline in its use which is likely to be accentuated if spectral domain OCT features such as disorganization of the retinal inner layers (DRIL) are established as reliable indicators of CNP. The International Nomenclature for Optical Coherence Tomography [IN • OCT] Panel has recently published a useful standardized nomenclature system for normal anatomic landmarks seen on SD-OCT.3 The various layers of the retina are now well characterized on SD-OCT, and changes in the morphology of these layers serve as valuable diagnostic and prognostic parameters. The SD-OCT has been used to evaluate retinal CNP.4 Sun et al. described the concept of foveal DRIL on OCT as a surrogate for poor visual outcome.5 However, the diagnostic accuracy of DRIL in detecting macular CNP is unknown. This is of particular importance in the era of anti-vascular endothelial growth factor treatment for DMO as SD-OCT is the most common imaging tool now used in DMO management. The objective of this study was to assess the sensitivity and specificity of DRIL in detecting macular CNP and to determine factors associated with DRIL that may account for its diagnostic accuracy.
METHODS We performed a post hoc analysis of images obtained from participants of completed randomized clinical trials on DMO conducted in National Institute for Health Research (NIHR) Moorfields Biomedical Research Centre, London, UK. Local institutional review board approval was obtained and the study was conducted in accordance to the tenets of the Declaration of Helsinki. The FA and SD-OCT images of consecutive patients with severe non-proliferative or proliferative diabetic retinopathy were included. Data on patient demographics and treatment history were available.
Image acquisition The imaging protocols for acquisition of FA and SD-OCT were similar for every patient included in this study.
FA FA was performed with an intravenous infusion of 5 mL fluorescein sodium 20%. A series of 35 degree images were obtained with a Topcon TRC 50 DX Fundus camera (Topcon, Tokyo, Japan). Standard 35-degree 7-field colour fundus image and FA images were available for both eyes. A 35-degree macular centred colour fundus and fluorescein angiography image of both eyes of each patient were used for this analysis.
SD-OCT All individuals underwent macular SD-OCT scans in both eyes. These were performed using Spectralis HRA + OCT (Heidelberg Engineering, Germany) which combines high-resolution SD-OCT with scanning laser ophthalmoscopy. The posterior pole volume scan was performed in both eyes on a 20 × 20 degree cube, with 49 raster lines, each containing 1064 pixels. Only scans with good signal strength (signal-to-noise ratio, 20 dB or greater) were evaluated. All scans were centred at the fovea to avoid any fixation errors and all segmentation errors were manually corrected.
Definition of macular CNP on FFA Using FA, CNP was defined as the absence of retinal arterioles and/or capillaries and was detected by characteristics such as a pruned appearance of adjacent arterioles and a darker appearance of the choroid6 (Fig. 1). Perfused retina was defined as a ground-glass appearance of normal angiogram. The colour fundus images were reviewed for cotton wool spots and retinal haemorrhages that may mimic nonperfusion due to blocked fluorescence during FA and these were excluded. Figure 1. Left image – fluorescein angiogram of the right eye. Capillary non-perfusion (CNP) with marked hypofluorescence and loss of visible capillary perfusion on fluorescein angiography marked in green and area in pink is a laser scar which is excluded from analysis.
© 2015 Royal Australian and New Zealand College of Ophthalmologists
DRIL to detect capillary non-perfusion
737
Image processing
Factors associated with DRIL
A single investigator (LN) identified 45 separate areas of macular CNP on the FA images (excluding the foveal avascular zone) and marked, coded and anonymized the associated SD-OCT line scans of these identified areas from 37 eyes of 31 patients. Similarly, a perfused area in a corresponding topographic location equidistant from the foveal centre in the same eye was chosen as reference in each study eye (Fig. 2). The SD-OCT scans of these areas were also marked, coded and anonymized.
We studied three factors that may be associated with DRIL that may help understand its diagnostic accuracy. These include the retinal thickness of the area of DRIL, area of CNP associated with DRIL and the duration of CNP in areas of DRIL. The retinal thickness in the area of perfused and non-perfused areas was measured by repositioning the Early Treatment Diabetic Retinopathy Study (ETDRS) grid to these areas and measuring the central subfield thickness in these areas. The absolute difference in retinal thickness between perfused and non-perfused areas was calculated. The area of CNP was measured using the ‘measure image’ tool in the Topcon IMAGEnet 2000 software (Topcon, Tokyo, Japan). Using previous FA images, the minimum duration of CNP was estimated from the first visible presence of CNP in the studied area.
Definition of DRIL Presence of DRIL is defined as the inability to identify any boundaries between the ganglion cell–inner plexiform layer complex, inner nuclear layer, and outer plexiform layer (Fig. 3). This definition of DRIL has been adopted from Sun et al., with the only exclusion being the horizontal extent in microns was not measured.5 The definitions of the hyper/hypo reflective layers in the OCT images were based on the proposed lexicon for the anatomical landmarks in retinal OCT imaging by the IN-OCT concensus.3
Inter-grader and intra-grader reliability The sequencing of the 90 anonymized images was ‘shuffled’ so that images with perfusion and nonperfusion of each eye were not contiguous. Two masked investigators (JR, NP) graded the images for the presence of DRIL. Arbitration was provided by a senior clinician (SS). The intergrader agreement was assessed. The two masked investigators repeated the grading process of the same set of images 6 months later and the intra-grader agreement was assessed.
Figure 2. A, Area of capillary non-perfusion; B, area of capillary perfusion equidistant; X, mm from the fovea.
Evolution of DRIL with time and treatment To further study the evolution of DRIL, the presence of DRIL on SD-OCT in the exact same location was assessed at 12 ± 1 months (−12 months) previously. The SD-OCT must have been performed using the same OCT machine and parameters, that is, Spectralis HRA + OCT volume scan and 20 × 20 degree cube with 49 raster lines each containing 1064 pixels. The images must be referenced to allow
Figure 3. OCT image with clear evidence of normal inner retinal boundaries and DRIL present temporally.
© 2015 Royal Australian and New Zealand College of Ophthalmologists
738
Nicholson et al.
for direct comparison. Angiographic presence of CNP at or before 12 months prior was also confirmed. Data on macular treatment received during this 12-month period were collected. This was grouped to either intravitreal anti-VEGF, intravitreal dexamethasone implant (Ozurdex, Alergan, Irvine, California, USA) or no intravitreal treatment.
Statistical analysis Statistical analysis was performed using Microsoft Excel (Microsoft Corp.) and GraphPad Prism (GraphPad Software Inc., La Jolla, California, USA). Descriptive statistics were used to present the data. Sample size was determined based on the calculation recommended by Flahault et al.7 For a sensitivity of 0.90, with a 0.95 probability of the lower 95% confidence limit of 0.75, the approximate number of images required is 70. A 2 × 2 contingency table protocol was used to evaluate the sensitivity, specificity and positive and negative predictive values of DRIL in the diagnosis of CNP. Sensitivity represents the percentage of true positives in all patients with the reference angiogram and specificity represents the percentage of true negatives. Positive and negative predictive values reflect the percentage of time that a positive or negative test (respectively) accurately captures the diagnosis of CNP. The confidence intervals for the sensitivity and specificity were computed via an online calculator at http:// www.pedro.org.au/english/downloads/confidence -interval-calculator/ using the Wilson score method. The number needed to diagnose was derived from the formula 1/[sensitivity − (1-specificity)] and represents the number of tests that need to be performed for DRIL to accurately diagnose the presence of CNP. Cohen’s kappa (k) coefficients and per cent agreement between examiners were used as estimates of inter-grader and intra-grader reliability for diagnosis of DRIL. Kappa values were interpreted according to the guidelines proposed by Landis and Koch.8 Almost perfect agreement is described for values between 0.81 and 1.00, substantial agreement for values between 0.61 and 0.80, moderate for values between 0.41 and 0.60, fair for values between 0.21 and 0.40, and slight for values under 0.20. The association of DRIL with CNP was assessed using the 2 × 2 contingency table and the Fisher’s exact test for significance. Comparisons between retinal thickness of areas of non-perfusion and perfusion was analysed using a Wilcoxon rank-sum test. Comparison between the sub-field thickness, area of CNP on FA and duration of CNP in areas with DRIL and without DRIL were analysed using the Mann–Whitney U test.
There was excellent inter-grader agreement between two masked investigators with a Cohen’s kappa of 0.909 and 95.56% observed agreements. The intra-grader agreement was 0.932 with a 96.67% agreement. Table 1 shows the 2 × 2 contingency table representing the grading outcomes of the 90 images for presence or absence of DRIL on OCT with the corresponding presence of CNP on fluorescein angiography. DRIL was significantly associated with CNP (P = 0.001). DRIL was absent on SD-OCT in 7/45 (15.6%) areas of CNP (Fig. 4). The sensitivity of DRIL as a predictive tool for CNP was 84.4% (95% CI 71.22% to 92.25%) with a specificity of 100% (95% CI 92.13% to 100%). The positive predictive value and negative predictive value were 100% and 86.5%, respectively. The number needed to diagnose was 1.18. Seven patients with angiographic evidence of CNP were graded as not having DRIL. These images were then re-examined in detail after arbitration, and it was confirmed that DRIL was absent in these cases.
Factors associated with DRIL The mean thickness of non-perfused retina was 333.27μ ± 87.62 and for the corresponding perfused retina, 311.13μ ± 56.07 (P = 0.21). For the 45 areas with CNP, the characteristics for areas with presence of DRIL and absence of DRIL are described in Table 2.
Evolution of DRIL with time and treatment From the 45 areas of angiographic CNP, 44 areas fulfilled the predetermined SD-OCT imaging requirements. All 44 areas selected for this study had evidence of angiographic CNP of 12 or more months, and 33 out of these 44 areas (75%) also had presence of DRIL of at least 12 months duration. Twenty-one of these 33 areas were not exposed to any intravitreal treatment while the rest of the 11 areas were exposed Table 1.
DRIL in detecting angiographic CNP Presence of DRIL on OCT Total Yes
No
38 0 38
7 45 52
RESULTS
Angiographic evidence Yes of capillary No non-perfusion Total
The mean age of the patients included was 61.94 ± 11.49 years. The male to female ratio was 28:3.
CNP, capillary non-perfusion; DRIL, disorganization of the retinal inner layers; OCT, optical coherence tomography.
45 45 90
© 2015 Royal Australian and New Zealand College of Ophthalmologists
DRIL to detect capillary non-perfusion
739
Figure 4. (a, c, e) fluorescein angiography with the green box corresponding to SD-OCT scanned area and single horizontal line corresponding to SD-OCT image shown on the right. Area of capillary non-perfusion marked in purple. (b, d, f) SD-OCT line scan with vertical line corresponding to border of non-perfusion with DRIL.
Table 2.
Comparison between DRIL present and DRIL absent areas of CNP
Absolute retinal thickness difference between perfused and non-perfused retina Area of non-perfusion Duration of non-perfusion
DRIL present (n = 38)
DRIL absent (n = 7)
P-value
60.53 μm ± 56.5
18.0 μm ± 19.22
0.01
9.43 mm2 27.68 months
3.88 mm2 20.57 months
0.17 0.81
CNP, capillary non-perfusion; DRIL, disorganization of the retinal inner layers.
to intravitreal treatment (11 to intravitreal Ozurdex and one with multiple injections of an anti-VEGF agent). There were four angiographic areas of CNP that were DRIL negative that developed DRIL over 12 months and one of these areas was also exposed to repeated injections of anti-VEGF agent. The remaining seven areas with angiographic evidence of CNP with absence of DRIL on OCT remained DRIL negative throughout the previous 12 months, and one of these areas was exposed to intravitreal Ozurdex.
DISCUSSION This study found that DRIL can be used to detect macular CNP in patients with diabetic retinopathy. The diagnostic sensitivity was 84.4% and specificity was 100%. In addition, we achieved a strong intergrader agreement suggesting that detection of DRIL is easy, reliable and an accurate parameter of CNP. The presence of DRIL is also a consistent finding with no evidence of reversal or normalization observed over a 12-month period.
© 2015 Royal Australian and New Zealand College of Ophthalmologists
740
Nicholson et al.
From a clinical perspective, the presence of DRIL is a very useful alternative to performing FA to identify CNP as DRIL is a definite indicator of nonperfused retina. There were no false positives. However, it is useful to consider the clinical implications of absence of DRIL in the presence of angiographic evidence of CNP (false negatives). With regard to false negatives, 7/45 (25.6%) angiographic evidence of CNP did not manifest as DRIL. Therefore, the absence of DRIL does not rule out CNP. This may suggest that the disorganization of retinal layers may only occur after a period of CNP or it is a sequelae of CNP. It is useful to understand the time to onset of DRIL from the development of CNP because DRIL is associated with poor visual outcome. Sun et al. reported the relationship between foveal disorganization of retinal inner layers (foveal DRIL) on SD-OCT and poor visual acuity in DMO.5 We believe macular ischemia and an enlarged foveal avascular zone may contribute to foveal DRIL and reduced visual acuity. This was not formerly assessed in the aforementioned paper. In our study, we have associated DRIL with nonperfusion and thus foveal DRIL may in fact be related to an enlarged foveal avascular zone or macular ischaemia. The association between macular ischaemia and poor visual acuity is known.1,9 Treatment trials on macular CNP should ideally be initiated before the onset of DRIL. We further explored the factors that may be associated with the presence or absence of DRIL in areas of macular CNP to better understand the evolution of DRIL. However, as the study was not set out to specifically study these factors, the numbers are limited and larger numbers may be required to confirm our observations. There is a significant relationship between larger absolute retinal thickness differences between perfused and non-perfused retina in DRIL positive and DRIL negative areas. Areas with DRIL on OCT were on average 60.53 μ thicker or thinner than corresponding perfused areas within the same eye. As for DRIL negative non-perfused areas, this difference only measured 18 μ. A thinner retina is associated with degenerate tissue10,11, and thicker retinas possibly relate to an ongoing ischaemic process.11 These two possibilities may be used to explain the finding we have observed in our study. DRIL is likely to be present when a significant level of ischaemia has been reached. In areas that do not present a significant difference in thickness between perfused and non-perfused capillaries, DRIL is likely to be absent despite angiographic evidence of non-perfusion. Therefore, it may be possible that the retina has not suffered significant changes to manifest DRIL, or adaptive processes may have been sufficient to maintain the integrity of the retinal structure. We
acknowledge the limitations in the study in not having microperimetry data to evaluate the function of these areas. A further limitation of the results and its application lies in the fact that the SD-OCT image needs to be of good quality for the clinician to confidently assess for DRIL. Although not statistically significant, we found a trend towards larger areas of non-perfusion being more likely to exhibit DRIL as opposed to smaller areas. The duration of retinal non-perfusion did not differ significantly between DRIL positive and DRIL negative images of angiographic non-perfused retina. We appreciate the method used in obtaining the duration of non-perfusion was highly dependent on previous imaging. Some subjects that presented late to the medical retina service may have had an area of non-perfusion that was only detected recently and provide a falsely shorter duration of CNP. Swept source OCT technology is one approach by which wide field OCT images may be obtained.12 Extrapolating the use of DRIL to the peripheral retina may enable us to identify peripheral nonperfusion with the use of wide field OCT. We acknowledge that non-invasive fluorescein angiography such as the speckle variance OCT will be extremely useful in identifying ischaemia13; however, currently it does not cover a large area of the macula and as yet is not widely used in clinical practice. DRIL is not only evident in CNP and its presence has been reported in acute retinal necrosis14 and closed globe trauma.15 It corroborates the theory that the disorganization is a manifestation of compromise to the inner retinal circulation. In summary, we have presented evidence to confirm the association between DRIL and CNP and suggest that DRIL is a predictive tool in identifying macular CNP but its absence does not rule out CNP.
REFERENCES 1. Sim DA, Keane PA, Zarranz-Ventura J et al. The effects of macular ischemia on visual acuity in diabetic retinopathy. Invest Ophthalmol Vis Sci 2013; 54: 2353–60. 2. Sim DA, Keane PA, Rajendram R et al. Patterns of peripheral retinal and central macula ischemia in diabetic retinopathy as evaluated by ultra-widefield fluorescein angiography. Am J Ophthalmol 2014; 158: 144– 153.e1. 3. Staurenghi G, Sadda S, Chakravarthy U, Spaide RF. International Nomenclature for Optical Coherence Tomography (IN•OCT) Panel. Proposed lexicon for anatomic landmarks in normal posterior segment spectral-domain optical coherence tomography: the IN•OCT consensus. Ophthalmology 2014; 121: 1572–8. 4. Unoki N, Nishijima K, Sakamoto A et al. Retinal sensitivity loss and structural disturbance in areas of
© 2015 Royal Australian and New Zealand College of Ophthalmologists
DRIL to detect capillary non-perfusion
5.
6.
7.
8.
9.
10.
capillary nonperfusion of eyes with diabetic retinopathy. Am J Ophthalmol 2007; 144: 755–60. Sun JK, Lin MM, Lammer J et al. Disorganization of the retinal inner layers as a predictor of visual acuity in eyes with center-involved diabetic macular edema. JAMA Ophthalmol 2014; 132: 1309–16. Chan CK, Ip MS, Vanveldhuisen PC, SCORE Study Investigator Group et al. SCORE Study report #11: incidences of neovascular events in eyes with retinal vein occlusion. Ophthalmology 2011; 118: 1364–72. Flahault A, Cadilhac M, Thomas G. Sample size calculation should be performed for design accuracy in diagnostic test studies. J Clin Epidemiol 2005; 58: 859– 62. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33: 159– 74. Lee DH, Kim JT, Jung DW, Joe SG, Yoon YH. The relationship between foveal ischemia and spectraldomain optical coherence tomography findings in ischemic diabetic macular edema. Invest Ophthalmol Vis Sci 2013; 54: 1080–5. Reznicek L, Kernt M, Haritoglou C, Kampik A, Ulbig M, Neubauer AS. In vivo characterization of ischemic
741
11.
12.
13.
14.
15.
© 2015 Royal Australian and New Zealand College of Ophthalmologists
retina in diabetic retinopathy. Clin Ophthalmol 2010; 5: 31–5. Murthy RK, Grover S, Chalam KV. Sequential spectral domain OCT documentation of retinal changes after branch retinal artery occlusion. Clin Ophthalmol 2010; 4: 327–9. Reznicek L, Klein T, Wieser W et al. Megahertz ultrawide-field swept-source retina optical coherence tomography compared to current existing imaging devices. Graefes Arch Clin Exp Ophthalmol 2014; 252: 1009–16. Xu J, Han S, Balaratnasingam C et al. Retinal angiography with real-time speckle variance optical coherence tomography. Br J Ophthalmol 2015; doi: 10.1136/ bjophthalmol-2014-306010. [Epub ahead of print]. Suzuki J, Goto H, Minoda H, Iwasaki T, Sakai J, Usui M. Analysis of retinal findings of acute retinal necrosis using optical coherence tomography. Ocul Immunol Inflamm 2006; 14: 165–70. Chen H, Lu Y, Huang H, Zheng J, Hou P, Chen W. Prediction of visual prognosis with spectral-domain optical coherence tomography in outer retinal atrophy secondary to closed globe trauma. Retina 2013; 33: 1258–62.
