Retinal and Choroidal Perfusion Status in the Area of Laser Scars Assessed With Swept-Source Optical Coherence Tomography Angiography

Abstract

Purpose: To evaluate the perfusion status of the retina and choriocapillaris in the area of laser scars on swept-source optical coherence tomography angiography (OCTA) images of eyes previously treated with panretinal photocoagulation (PRP).

Methods: Cross-sectional exploratory analysis of swept-source OCTA images, which were retrospectively reviewed for laser scars. The appearance of the capillary networks in the area of previous laser were evaluated following a three-step grading system (normal/sparse/missing capillary network). The superficial and deep capillary plexus of the retina and the choriocapillaris were graded separately.

Results: A total of 3140 laser scars in 54 eyes of 31 patients (13 female, mean age 57 ± 12 years) were included in this analysis. In the retina, 6.8% of the superficial and deep capillary network in the area evaluated appeared normal, 58% and 56% sparse, and 35% and 37% missing. Capillary dropout in the retina was not restricted to the area of prior laser treatment. The choriocapillaris decorrelation signal was either sparse (61%) or completely missing (38%) within the laser scar area. The perfusion of the choriocapillaris appeared normal in the area adjacent to laser scars.

Conclusions: Capillary non-perfusion in the choriocapillaris was found within the laser scar area. Laser treatment seems to cause sustained non-perfusion of choriocapillaris in the area treated.

Proliferative diabetic retinopathy (PDR) is a sight-threatening microvascular complication of diabetes mellitus with a global prevalence of 6.8% of patients with diabetes.¹ Panretinal laser photocoagulation (PRP) has been the standard of care for PDR for decades.^2,3 Laser treatment aims to destroy ischemic tissue to reduce vascular endothelial growth factor (VEGF) production and enhance oxygen delivery from the choriocapillaris (CC) to the inner retina.^4,5 However, little is known about the choriocapillaris after laser treatment. The laser energy is mainly absorbed by the retinal pigment epithelium (RPE), where it converts to thermal energy, destroys RPE and photoreceptor cells, and injures the surrounding retina and CC. After an initial increase in VEGF expression due to inflammation, the injured RPE cells surrounding the laser spots release cytokines that inhibit neovascularization and endothelial cell growth.^6–8 Apart from in vitro experiments and histology studies, the effect of laser treatment has been extensively studied in vivo. The retina is relatively easily accessible with standard and advanced imaging methods. Hence, the thermal effects of argon laser photocoagulation on the neuroretina, photoreceptors and RPE have been followed in vivo with optical coherence tomography (OCT), fundus autofluorescence,⁹ polarization sensitive OCT,¹⁰ and adaptive optics scanning laser ophthalmoscopes.¹¹ In addition to a loss of photoreceptors and RPE cells, an edema of the outer retina has been observed in spectral-domain OCT images. 

Signals from the CC and the choroid are widely attenuated by the RPE, which impedes detailed imaging of this highly perfused tissue. Indirect signs, such as the choroidal thickness measured with enhanced depth OCT, allow a quantitative assessment of the central choroid but is devoid of structural details.¹² Further, current knowledge about the effect of retinal photocoagulation on the CC is based on histological studies in animal models, which have followed laser treatment for up to 1 month.^13–15 These results suggest that laser injury to the CC initially results in microthrombosis but that vascular integrity is restored entirely after 30 days.¹⁴ 

Angiography of the CC remains challenging due to dye leakage from fenestrated CC vessels in fluorescein angiography (FA) and the attenuation of the blue-green excitation signal by the RPE. Laser spots appear hyperfluorescent in FA images, which impedes an evaluation of the local perfusion status. Indocyanine green (ICG) is frequently used for imaging the choroidal circulation because it fluoresces better through the RPE. Additionally, ICG molecules are protein-bound, so diffusion through the CC is limited. The choriocapillaris appeared to be missing in ICG images after large laser spots of more than 3 mm where applied in an animal model.¹⁶ This effect exceeded the treated area by at least 1 to 2 mm. However, with laser spots of 200 to 500 μm in diameter, the CC appeared normal. This might be due to the low resolution of the ICG imaging technique. 

With the introduction of OCT angiography (OCTA), the separate visualization of each capillary plexus of the retina and choroid became accessible in vivo. The depiction of blood flow at different depths within a tissue seems to be particularly suitable for evaluating a highly perfused tissue such as the choriocapillaris. Swept-source (SS) technology uses faster scanning speed and longer wavelength than spectral domain OCTA (100 kHz vs. 70 kHz and 1050 nm vs. 840 nm, respectively), which allows for better lateral resolution and greater light penetration into deeper tissues beyond the RPE. The purpose of this study was to evaluate the perfusion status of the retina and the choriocapillaris in the area of laser scars with SS OCTA in patients previously treated for high-risk proliferative diabetic retinopathy. 

Methods

This was a retrospective, cross-sectional exploratory analysis registered at the IRB (EKNr 1118/2018) and performed in accordance with the ethical standards stated in the Declaration of Helsinki. The research received no specific grant from any funding agency in the public, commercial, or nonprofit sectors. We reviewed OCTA images recorded with the Zeiss Plex Elite 9000 (Carl Zeiss Meditec AG, Jena, Germany) between July and November 2017 at the tertiary eye care center (Department of Ophthalmology, Medical University of Vienna, Austria). The SS OCTA device used operates at a center wavelength spectrum between 1040 and 1060 nm recording 100,000 A-scans per second to acquire OCTA volumes consisting of 500 × 500 A-scans and two frames per B-scan over an area of up to 12 × 12 mm. We included only images of good quality (9/10 or 10/10) and free from imaging artifacts that displayed lesions secondary to panretinal laser photocoagulation in the enface OCT fundus image. Scanning pattern were 12 × 12 mm centered on the fovea and 6 × 6 mm centered below the optic nerve head. Laser scars from focal or grid laser were not analyzed. Panretinal laser treatment had either been done with a conventional argon laser (Supra; Quantel Medical, Cournon d'Auvergne, France) or with the PAtterned SCAnning Laser (PASCAL; OptiMedica Corp., Santa Clara, CA, USA) system. PASCAL, a semi-automated neodymium-doped yttrium-aluminum-garnet laser, employs a shorter pulse duration of 10 to 20 ms compared to 100 to 200 ms in conventional argon laser. Both systems emit green light at a wavelength of 532 nm. Patient charts were reviewed to identify patients with diabetes and variables such as age, sex, A1c level, diabetes type and duration, history of anti-VEGF injections or vitreoretinal surgery, time since panretinal laser treatment, number of applied laser spots and the laser device used were recorded. 

Image Grading

Following images were de-identified and exported to ImageJ (v1.51r; National Institutes of Health, Bethesda, MD, USA) for image analysis: OCT fundus image from the “Enface Report” and the “Angioplex Images” superficial, deep, and choriocapillaris from the “Angiography Analysis Report.”¹⁷ After adjusting the size of the structural enface OCT fundus image to 1024 × 1024 pixels to match the Angioplex images, they were organized in a stack. First, the laser scars were identified and consecutively numbered on the enface OCT fundus image, and the appearance of the corresponding area was graded in the Angioplex images of the superficial retinal capillary plexus (SCP), the deep retinal capillary complex (DCC), and the CC individually (Fig. 1). The vascular layers were segmented using the segmentation algorithm of the Plex Elite OCTA: the SCP was segmented between ILM and the outer boundary of the inner plexiform layer, the DCC between inner and outer plexiform layer (segmenting the intermediate and deep capillary plexus together), and the CC between 29 and 49 μm below the retinal pigment epithelium.¹⁷ Segmentation errors were corrected manually if necessary. The vascular network at the area of the laser scar was evaluated according to a triple-class grading system: (1) normal, meaning that the capillary network in the area of laser application appears as dense as in healthy or surrounding areas, making it impossible to locate the laser scar within the vascular plane; (2) sparse, referring to a diminished capillary network in the area of laser application compared with the surrounding or healthy vascular network or a missing capillary network in some part of the laser lesion; or (3) missing capillary network, indicating that the capillary bed is completely absent in the entire area of laser application, giving it a “punched out” appearance if the surrounding network is intact, facilitating an exact localization of the laser scar in the corresponding vascular plane. Single arterioles or venules or blood vessels of greater caliber crossing the graded area were not considered. Examples of the gradings “sparse” and “missing” are shown in Figure 1. If the vascular networks at the area of a laser scar could not be graded reliably, the grading “NA” (not available) was chosen. 

Figure 1

 

3 × 3 mm OCTA image (Plex Elite, Zeiss) of a patient with proliferative diabetic retinopathy (DR) at an area previously treated with panretinal photocoagulation. Laser scars contrast clearly from the surrounding retina in the enface OCT fundus image. (A, B) They were manually outlined in red (B). The overlay was used on the corresponding OCTA images of the superficial (G) and deep (H) capillary complex as well as of the choriocapillaris (I) for the grading of the vascular networks. Laser scars 1 and 2 are examples of (1) missing and (2) sparse choriocapillaris, respectively. (C–F) Blowup of B-scan without (C, E) and with (D, F) color-coded flow information of laser scar 1 (C, D) and laser scar 2 (E, F) at the area highlighted with the green line in A. The purple dashed line represents the inner (29 μm below the RPE) and outer (49 μm below the RPE) boundary of the CC segmentation. (G, H) Enface OCT angiography view of the superficial (G) and deep (H) capillary complex with a B scan below showing the corresponding segmentation boundaries in purple. The green line indicates the location of the B scan. The retinal capillary networks are clearly damaged secondary to advanced DR, but capillary dropout does not seem to relate to the laser scars. Laser area 1 and 2 were both graded as “sparse” because the capillary network is still visible but clearly diminished. (I) In the plane of the choriocapillaris (CC), the missing capillary network at the laser scars show a punched-out appearance and bigger vessels of the underlying choroid become visible. Laser scar 1 represents an example for the grade “missing.” In the area of laser scar 2, the capillary network is only missing in some parts of the laser scar. Nevertheless, a defect in the CC network is clearly visible. The choriocapillaris of laser scar 2 was graded as “sparse.”

Statistics

The main variables of interest were the effects in the choriocapillaris, superficial capillary plexus, and deep capillary complex, which describe a three-level ordinal grading of laser spots in these three different vascular planes. The main aim of this analysis was to investigate the potential influence of a set of prespecified covariates—sex, age, type of diabetes, duration of diabetes, A1c level, laser system used, time since laser treatment, total number of laser spots, prior anti-VEGF injections, and prior vitreoretinal surgery—on the ordinal grading of CC. First, descriptive statistics (including means, standard deviations, relative and absolute frequencies) were computed for all variables of interest. Second, relations of interest were depicted visually. Third, an ordinal regression using a cumulative link mixed model, accounting for the multiple observations per patient, was fitted to explain rating of CC by the set of covariates using the following strategy: For all covariates except for the known confounders of DR^18,19 (sex, age, type of diabetes and duration of diabetes), an ordinal logistic regression model was fitted to explain CC by the covariate. Covariates that were found to be significant were included in a final ordinal logistic regression model alongside the above-stated confounders of DR. P values < 0.05 were considered significant and P values < 0.1 were considered borderline significant. Because of the exploratory nature of this analysis, no multiple testing corrections were applied. All analyses were performed using R 3.6.0 and the ordinal package. 

Results

A total of 3140 laser scars in 54 eyes of 31 patients (13 female, mean age 57 ± 12 years) were included in this analysis, which was based on 53 12 × 12 mm OCTA scans centered on the fovea and one 6 × 6 mm OCTA scan centered below the optic nerve head. Patients' characteristics are summarized in Table 1. Thirty-one eyes were pseudophakic, 49 eyes had received anti-VEGF treatment, and 3 eyes had a history of vitreoretinal surgery. Twelve patients had diabetes type 1, and 21 patients were using insulin. 

Table 1

 

Descriptive Statistics of Variables Evaluated in 54 Eyes of 31 Patients

 

At the level of the superficial and deep retinal plexus of the retina, the capillary network in the areas evaluated appeared normal in 6.8%, sparse in 58% and 56%, and missing in 35% and 37% of all scars analyzed. Capillary dropout in the retina was not restricted to the area of prior laser treatment but also involved the neighboring areas (Figs. 1G, 1H). The CC decorrelation signal was either sparse (61%) or completely missing (38%) within the laser scar area, with well circumscribed edges to the surrounding CC (Fig. 1I). Of the laser scars, 0.6% showed no apparent damage to the CC network. Perfusion of the CC adjacent to laser scars appeared normal, giving scars a punched-out appearance on the corresponding vascular plane. In the area of missing CC, the blood flow signal from bigger vessels in the underlying choroid became visible (Fig. 1I). The number of laser lesions evaluated for each patient—including sex, age, type of laser device, and perfusion grading of the CC—is plotted in Figure 2. The univariate ordinal logistic regressions showed a significant correlation between CC perfusion and time since laser treatment (longer time since laser treatment corresponded to greater CC defect; P = 0.009 and laser type PASCAL led to less CC defect; P = 0.007). All other covariates showed no significant correlation (Table 2). 

Figure 2

 

Three-step grading of the CC in the area of the laser scar in each patient: 0 = missing CC, 1 = sparse CC, 2 = normal CC, NA = not available, secondary to an imaging artifact, age, sex, number of laser scars, and the type of laser used become visible in this graph.

Table 2

 

Univariate Logistic Regression Including Covariates That Might Have an Impact on the Perfusion of the CC at the Area of Prior Laser Treatment

 

The multiple ordinal regression model, including time since laser treatment and laser type, showed at least borderline significant correlations between CC perfusion and patient age (older age corresponded to greater CC defect; P = 0.009), diabetes duration (longer diabetes duration corresponded to less CC defect; P = 0.061), time since laser treatment (longer time since laser treatment corresponded to greater CC defect; P = 0.015) and laser type (PASCAL leads to less CC defect; P = 0.063) (Table 3). 

Table 3

 

Multiple Ordinal Regression Model Including Covariates, Which Are Known Confounders in Diabetic Retinopathy or were Significantly Correlated With CC Perfusion in the Univariate Regression Analysis

 

Sex, A1c level, and type of diabetes were not significantly correlated to the CC perfusion status. All 3 CC perfusion patterns were found in 9 (29%) patients, suggesting that laser spots do not heal uniformly in individual patients. 

Discussion

OCTA imaging illustrated the perfusion status of the superficial capillary plexus and the deep capillary complex of the retina as well as the choriocapillaris separately in vivo, even in areas previously treated with laser. Extended retinal ischemia in 93% of the areas evaluated reflects advanced damage secondary to DR in our patients. The capillary decorrelation signal in the superficial capillary plexus was similar but not exactly coherent with the signal in the deep vascular complex. However, capillary nonperfusion in the retina is commonly seen in DR and was not restricted to the area of laser scars. Further analyses were disregarded because the effect of laser treatment on the retina was restricted to the outer retinal layers.²⁰ Contrary to being seen in the retina, capillary nonperfusion in the CC was confined to the area of the laser scar. Laser treatment seems not only to destroy the photoreceptors and the retinal pigment epithelium but also cause a sustained nonperfusion of CC in the area treated. Shadowing from the focally scarred RPE is very unlikely to be responsible for the missing blood flow signal in the CC because blood flow within the choroid beneath remains visible. Damaged CC accorded with the results from human histology, which showed a persistent defect of CC at 6 and 14 months after panretinal photocoagulation (PRP) treatment.^21,22 The appearance of the CC network recorded with OCTA resembles the one recorded with a scanning electron microscopy (Fig. 3). In 99.4% of all laser lesions analyzed, the CC was sparse or completely missing, leaving only 0.6% without any visible impact on the CC. Similarly, only 2 laser lesions displayed evidence of microvascular repair in a patient who had previously received 1843 burns.²¹ Restored capillaries in areas previously treated with laser appeared different from the surrounding CC. They were typically small in diameter and had a plump non-fenestrated endothelium.²² The sparse appearance of the CC in 61% of laser lesions evaluated with OCTA could either correspond to an incomplete destruction or a partial reperfusion of initially damaged CC. However, OCTA allows evaluation only of perfusion, not the high resolution structural details of the capillaries. Adaptive optics technology might give us more in vivo insight of the CC in future and allow regular CC and newly formed vessels to be distinguished.²³ 

Figure 3

 

(A) Blow-up of laser scar 1 in Figure 1. Details of the choriocapillaris cannot be resolved due to resolution limits. Nevertheless, the dense capillary network is clearly interrupted, showing the blood flow signal from an underlying bigger vessel from the choroid. The OCTA image resembles the scanning electron microscopic appearance (B), reprinted with permission from Wilson DJ, Green WR. Argon laser panretinal photocoagulation for diabetic retinopathy. Scanning electron microscopy of human choroidal vascular casts. Arch Ophthalmol. 1987;105:239–242. Copyright 1987 American Medical Association.

In cats, the initial damage to the CC resembles that in humans, showing microthrombi occluding the capillaries. However, thrombolysis, fibrinolysis, and recanalization were observed within 2 to 4 days after laser injury, leading to an almost normal appearance of CC within 1 month.¹⁴ Contradictory results have been published about the impact in humans of lasers on the CC. While one report found that the CC “appeared intact” 4 years after PRP treatment, another found a clearly disrupted CC in areas of photocoagulation.^21,24 The CC was described as largely fibrosed and nonperfused, while untreated areas showed normal CC. In our cohort, the CC decorrelation signal was completely (38%) or partially (61%) missing in the area of laser application. More severe CC defects were correlated with older age. In general, capillary density decreases with age, and microvascular repair slows down.^25,26 By contrast, patients with a longer duration of diabetes had a less severe CC dropout. If the CC in these patients is more robust to laser damage or an enhanced healing process is responsible for this observation, this still needs to be investigated. A pro-inflammatory milieu and high levels of angiogenic factors in patients with diabetes accentuate the latter. However, it remains a hypothesis for this study because these factors were not measured in our patients, and we did not follow the laser scars over time. 

Half our patients were treated with the PASCAL system, and the other half received conventional argon laser treatment. The laser energy of both laser systems is absorbed mostly by the retinal pigment epithelium, resulting in damage to the surrounding tissue between the CC and the outer retinal layers. Eventually, the lateral expansion of scars after argon laser treatment considerably exceeds the expansion of those applied with the PASCAL system.²⁷ Extended elimination of hypoxic tissue might contribute to better neovascularization regression.²⁸ However, laser energy not only spreads laterally but also vertically. The shorter pulse duration of the PASCAL (20 ms compared to 200 ms with a conventional argon laser) results in a shorter period of hyperthermia and, consequently, in less thermal diffusion and less thermal damage in neighboring tissues such as the CC.¹³ In our analysis, patients treated with PASCAL displayed fewer CC perfusion defects compared to patients treated with conventional laser. 

Until now, the impact of CC integrity on the efficacy of PRP treatment remains concealed. Following PDR treatment, more destructive laser methods had a favorable outcome on risk reduction of disease progression; however, they were accompanied by considerable side effects.^2,29 It still needs to be investigated if CC destruction should be intended or avoided for the optimal treatment outcome. PRP has a sustained impact on retinal and choroidal perfusion leading to a pO₂ increase in the retina and a pO₂ decrease in the choroid.⁴ The rarefication of the CC network in areas treated with PRP may lead to increased choroidal blood flow in the foveal area, as shown with laser Doppler flowmetry.³⁰ With OCTA, the perfusion status of the retina and CC can now be visualized noninvasively and might become a biomarker for the clinical effectiveness of panretinal laser treatment. That said, a longitudinal follow-up of laser lesions in vivo is necessary to investigate vascular damage and reperfusion processes and the impact of different laser systems and setting on this potential new biomarker. 

In OCT, the use of enhanced-depth imaging and the implementation of longer scanning wavelengths enabled better tissue penetration. Although indirect signs of choroidal involvement, such as choroidal thickness and volume, have repeatedly been used as disease markers in diabetic retinopathy, detailed imaging of the CC requires good tissue penetration as well as high axial and lateral resolution.^31–34 We used a SS OCTA technology with a center wavelength of approximately 1050 nm, which allows light to penetrate deeper tissue. Aware of the limited lateral resolution of 20 μm, we refrained from quantitative measurements in our study and adhered to a three-class qualitative grading system instead. The implementation of adaptive optics technology might allow for better characterization of laser effects in the future. 

There are several limitations that need to be acknowledged: First, the mean number of laser scars evaluated was 42 with a wide range from 1 to 158 in each eye. Due to the retrospective character of this study, we only included PRP laser scars within the field of view of the wide-field (12 × 12 mm) OCTA image. We did not evaluate grid or focal laser scars because of the limited lateral resolution of wide-field OCTA images. A smaller scanning pattern (3 × 3) or adaptive optics technology might be appropriate to evaluate macular laser scars in detail. The correlation with systemic factors needs to be considered with caution due to the imbalance of laser scar numbers in each patient. However, we could show that areas previously treated with a laser displayed a characteristic abrupt loss of CC. Laser treatments were performed with the PASCAL or argon laser, and both showed an impact on the CC in the area of laser scars. We would like to emphasize that this analysis was not designed to directly compare the two laser systems, but our data suggest there might be a difference in the healing response that needs to be further investigated in a prospective study. Finally, OCTA technology captures blood flow only at a certain velocity. Variable interscan time analysis (VISTA) would be needed to exclude lower or higher blood flow velocities in these areas.³⁵ The similarity to the histological finding of missing CC at this distinct area of previous laser treatment remains striking. However, variable blood flow velocities might be present, especially in residual and newly formed capillaries. 

To conclude, PRP results in apparent damage not only to the outer retina but also to the CC in the area previously treated with a laser. On OCTA images, CC perfusion defects appear sharply demarcated from the untreated CC and are clearly visible. The gradual healing response of the CC in humans still needs to be investigated to discover if partial perfusion of the CC relates to remaining or newly built capillaries. 

Acknowledgments

Disclosure: S.G. Karst, None; H. Beiglboeck, None; R. Scharinger, None; E.L. Meyer, None; C. Mitsch, Carl Zeiss (R), Askin (R); C. Scholda, None; U.M. Schmidt-Erfurth, Alcon Laboratories (C, R), Bayer Healthcare (C, R), Novartis (C, R), Allergan (C, R), Boehringer (C, R)