Correlation Between Hyperreflective Foci and Clinical Outcomes in Neovascular Age-Related Macular Degeneration After Switching to Aflibercept

Abstract

Purpose: To assess the correlation between hyperreflective foci (HF) and visual and anatomical outcomes in treatment-resistant neovascular age-related macular degeneration (AMD) using spectral-domain optical coherence tomography (SD-OCT).

Methods: This was a prospective interventional case series. Thirty-three eyes of 30 consecutive patients with treatment-resistant neovascular AMD were enrolled. Intravitreal aflibercept injections were performed at week 0 (baseline), week 4, and week 8. Spectral-domain OCT images were obtained before each injection and 4 weeks after the third injection. The main focus was on the measurement of choroidal neovascularization (CNV) size in the cross-sectional area in the B-scan through the fovea, and HF number along line segments of 1- and 3-mm length passing through the fovea.

Results: Mean number of HF in the radius of 500 μm decreased from 8.36 ± 7.58 to 4.15 ± 3.39 (P = 0.02). Mean number of HF in the radius of 1500 μm was reduced from 21.30 ± 12.47 to 10.45 ± 6.34 (P < 0.001). Mean CNV area decreased from 0.35 ± 0.22 to 0.22 ± 0.16 mm² (P < 0.001). There was a significant positive correlation between HF reduction in the radius of 500 μm and decrease in central subfield thickness (CST) (r = 0.43, P = 0.01), but no statistically significant correlation was found between HF decline in the radius of 1500 μm and other parameters.

Conclusions: Switching from ranibizumab to aflibercept caused significant decrease in the number of HF 1 month after aflibercept upload, and HF decrease in the radius of 500 μm was correlated positively with the reduction in CST.

Hyperreflective foci (HF) are discrete, well-circumscribed lesions with greater reflectivity than the retinal pigment epithelium (RPE) band on spectral-domain optical coherence tomography (SD-OCT).¹ Macular hyperpigmentation in eyes with intermediate age-related macular degeneration (AMD) has high spatial correlation to SD-OCT-detected HF and often represents the same anatomical lesion.² Hyperreflective foci overlying drusen are likely to represent progression of disease with RPE migration into the retina and possible photoreceptor degeneration or glial scar formation.³ It was reported that proliferation and inner retinal migration of HF occurred during follow-up in eyes with intermediate AMD, and they were associated with higher incidence of geographic atrophy; therefore, HF proliferation and migration may serve as biomarkers for AMD progression.⁴ The role of HF has been investigated in several other retinal disorders like retinitis pigmentosa,⁵ central retinal vein occlusion,⁶ and branch retinal vein occlusion.⁷ These structures are also scattered throughout all retinal layers in diabetic macular edema (DME) and might represent the initial stages in the development of intraretinal hard exudates.⁸ It has been reported that the presence of HF in the outer retina is closely related to a disrupted external limiting membrane (ELM), inner segment/outer segment (IS/OS) line on SD-OCT images, and decreased visual acuity in DME.⁹ The HF amount seems to indicate the severity of DME. Fewer HF might represent better retinal tissue integrity, whereas the presence of many HF reflects tissue disintegrity and more severe DME conditions.¹⁰ 

Neovascular AMD carries a devastating visual prognosis if left untreated.¹¹ The use of intravitreal antivascular endothelial growth factor (VEGF) therapy is the standard of care for neovascular AMD. Intravitreal bevacizumab and ranibizumab reduce exudative fluid and improve best-corrected visual acuity (BCVA) in eyes with neovascular AMD.^12,13 However, in spite of promising results of multiple trials, many patients require continued monthly injections because of recurrent exudation, and others have an incomplete response or develop tachyphylaxis.¹⁴ Aflibercept is a new VEGF inhibitor with a high affinity for VEGF. It is a protein manufactured by fusion of the second binding domain of the receptor VEGFR1 and the third binding domain of the receptor VEGFR2 to the crystalline portion of IgG1.¹⁵ Intravitreal aflibercept is an effective and safe treatment for exudative AMD and not inferior to ranibizumab.¹⁶ Patients with neovascular AMD who exhibit recurrent or resistant intraretinal or subretinal fluid following multiple injections with either bevacizumab or ranibizumab may take advantage of switching to intravitreal aflibercept therapy.^17–19 

In this study we evaluated SD-OCT scans of eyes with treatment-resistant neovascular AMD after switching from ranibizumab to aflibercept, in order to determine the behavior of HF and their associations with visual and anatomical outcomes after three intravitreal aflibercept injections (aflibercept upload). 

Materials and Methods

Patient Selection

Patients with treatment-resistant neovascular AMD who previously received intravitreal ranibizumab outside of the eye hospital of Hannover Medical School were assigned to aflibercept therapy in this prospective case series. The study was subject to the ethical standards stated in the 1964 Declaration of Helsinki and approved by the Clinical Research Ethics Committee of Medical School of Hannover (Medizinische Hochschule Hannover) with informed consent obtained. Treatment-resistant neovascular AMD was defined as choroidal neovascularization (CNV) secondary to AMD, which was documented in the patients' records as subretinal fluid and/or intraretinal fluid/cysts after more than 6 months of monthly ranibizumab therapy. All patients had been treated with monthly 0.5-mg ranibizumab. The interval between the last ranibizumab and the first aflibercept injection was not less than 4 or more than 6 weeks. Exclusion criteria were (1) previous macular laser photocoagulation in the study eye; (2) prior photodynamic therapy; (3) presence of fibrosis or scar in the macular region; (4) history of vitrectomy surgery; (5) history of retinal angiomatous proliferation (RAP); and (6) history of polypoidal choroidal vasculopathy (PCV) in the study eye. Each eligible eye received intravitreal injections of aflibercept at the recommended dose of 2 mg (0.05 mL) every 4 weeks for 3 months (loading phase). Comprehensive ophthalmologic examination including best-corrected visual acuity (BCVA) measurements, applanation tonometry, and ophthalmoscopy was obtained on the day of aflibercept initiation and at subsequent follow-up visits until 4 weeks after the third aflibercept injection. 

Image Acquisition, Processing, and Analysis

Each patient underwent regular pre- and posttreatment SD-OCT (Spectralis HRA+OCT; Heidelberg Engineering, Heidelberg, Germany) examinations. The postinjection examinations were performed 4 weeks after the first injection. Spectral-domain OCT volume scans (20° × 20° approximately 6 × 6 mm) with 49 B-scans each spaced 120 μm apart and automatic real time (ART) of 16 were obtained at each session. The acquisition software version was 5.6.3.0. The same horizontal B-scan sections (in x-z plane) passing through the fovea were used during the follow-up to ensure matching sections for comparison. Images were displayed using the Heidelberg Eye Explorer software version 1.7.1.0. The cross-sectional CNV area in the B-scan through the fovea was measured by using the distance measurement tools in 1:1 pixel mode provided by the software (Fig. 1). In the same section of measured CNV size, evaluation of HF was performed. Hyperreflective foci were defined as focal, well-circumscribed structures with reflectivity similar to that of the RPE layer. The location of the fovea in the thickness map was used as reference point (Fig. 2). Since the central subfield thickness (CST) is defined as the average retinal thickness in the area enclosed in a 1000-μm-diameter circle centered at the center of the fovea,^20,21 two 500-μm-segment lines corresponding to the diameter of the circle were plotted from the reference point in opposite directions along the x-axis. The 1:1-μm mode was used at this stage to avoid measurement inaccuracies; then two other lines perpendicular to the aforementioned lines were drawn (Fig. 3a). Finally, measurement of the number of HF was made in the 1:1 pixel mode on a single B-scan along a line segment of length 1 mm (Fig. 3b). The magnification tool of the software was used if necessary. Using the same method, measurement of HF number was done on a single B-scan along a line segment of length 3 mm (Figs. 3c, 3d). Figure 4 shows labeled examples of dots counted as HF. The postoperative retinal structure was qualitatively interpreted as (1) dry, exhibiting no fluid, or (2) wet, revealing unchanged edema or incomplete fluid regression. The main parameters used for quantitative evaluation of macular thickness were CST and macular volume. 

Figure 1

 

Measurement of CNV size in 1:1 pixel mode using Heidelberg Eye Explorer software in the foveal section.

Figure 2

 

Finding the fovea in the thickness map and setting it as the reference point.

Figure 3

 

Measurement of HF amount using Heidelberg Eye Explorer software. (a) Two 500-μm-length segment lines were plotted from the reference point in the nasal and temporal directions; then two other segment lines perpendicular to the first lines were drawn in 1:1-μm mode. (b) Counting of HF in the radius of 500 μm was performed within the surrounding area in 1:1 pixel mode. The same method was used in the radius of 1500 μm (c, d).

Figure 4

 

Labeled examples of dots that are counted as HF in the radius of 500 μm (a) and 1500 μm (b).

Statistical Analysis

Data were collected and analyzed using SPSS software (version 20.0; IBM Corporation, Armonk, NY, USA). Quantitative variables were tested for normal distribution by the Kolmogorov-Smirnov test. Within-group changes from baseline were analyzed using paired t-test. Spearman rho test was used for correlation analysis. A P value of <0.05 was considered statistically significant. Means ± standard deviation were used for reporting the results. 

Results

From June 2013 until March 2014, 40 eyes of 36 consecutive patients with treatment-resistant neovascular AMD were enrolled in this study. Seven eyes were excluded during the follow-up period: three cases due to the inappropriate picture quality and four cases of significant pigment epithelial detachment in the absence of measurable CNV. In the end, 33 eyes of 30 consecutive patients with treatment-resistant neovascular AMD were evaluated. Mean age of the patients was 76.3 ± 9.6 years (range, 56–94). Fourteen patients (46.7%) were male and 16 patients (53.3%) were female. There were 20 (60.6%) right eyes and 13 (39.4%) left eyes. Mean number of intravitreal ranibizumab injections was 8.27 ± 6.01 (range, 6–30). At baseline, mean BCVA was 48.03 ± 17.50 Early Treatment Diabetic Retinopathy Study (ETDRS) letters, and it significantly improved to 58.65 ± 17.80 ETDRS letters (P < 0.001) 1 month after the third injection. The number of HF in the radius of 500 and 1500 μm decreased 1 month after the third injection in 28 (85%) and 29 (88%) eyes, respectively. Mean number of HF in the radius of 500 μm decreased significantly from 8.36 ± 7.58 to 4.15 ± 3.39 (P = 0.02). Mean number of HF in the radius of 1500 μm decreased significantly from 21.30 ± 12.47 to 10.45 ± 6.34 (P < 0.001). Mean CST decreased from 381.2 ± 124.9 to 281.4 ± 83.9 μm (P < 0.001). Mean macular volume was 9.57 ± 1.75 mm³ and declined to 8.01 ± 1.03 mm³ (P < 0.001). The CNV area increased in 1 eye (3%), remained unchanged in 2 eyes (6%), and decreased in 30 eyes (90%). Mean baseline CNV area was 0.35 ± 0.22 mm², and it decreased significantly to 0.22 ± 0.16 mm² at the final visit (P < 0.001). Figure 5 shows an example of these changes. A dry macula was achieved in 21 eyes (64%), and 12 eyes (36%) showed decreased or unchanged edema. There was a weak significant positive correlation between HF reduction in the radius of 500 μm and decrease in CST (r = 0.43, P = 0.01), but no statistically significant correlation was found between HF decline in either the radius of 500 μm or the radius of 1500 μm, and BCVA improvement, CNV size change, and macular volume reduction. There was a positive significant correlation between baseline CST and baseline HF in the radius of 500 and 1500 μm (r = 0.50, P = 0.003, and r = 0.35, P = 0.045, respectively), but correlations between remaining baseline and final parameters were not significant. No ocular adverse effects or systemic complications occurred during the study period. 

Figure 5

 

Right eye of a 78-year-old female patient. SD-OCT represents subretinal edema and several HF in the surrounding area. CNV size is 0.63 mm² (a). One month after the third injection, the retina was dry and HF amount significantly decreased. CNV size was also reduced to 0.42 mm² (b).

Discussion

The RPE is characterized in vivo by its abundant melanosomes, melanolipofuscin, and lipofuscin granules, all of lysosomal lineage and all potential subcellular contributors to SD-OCT reflectivity.²² Three RPE grades contain epithelial and nonepithelial components (sloughed, shedding, and intraretinal). Four are epithelial only (nonuniform, very nonuniform, bilaminar, and vacuolated). Two are nonepithelial only (dissociated and entombed).²³ Entombed RPE is a major finding in CNV eyes. Sloughed, dissociated, and bilaminar morphologies are other abundant RPE grades in CNV eyes. Vacuolated RPE is uncommon in CNV. According to the published histology and clinical imaging, many histologic RPE grades are transferrable to SD-OCT. Sloughed and intraretinal morphologies are excellent candidates for the subretinal and intraretinal HF repeatedly observed on SD-OCT and interpreted variably as dead RPE and melanin-containing macrophages,²⁴ activated RPE cells,⁴ and inflammatory cells (e.g., retinal microglia).²⁵ Retinal pigment epithelium–derived cells (subducted, melanotic, and entubulated morphologies)²⁶ are outside the RPE layer and not attached to a basal lamina or basal laminar deposit. Melanotic cells are exclusively found in eyes with CNV scars and defined by a variable number of very dark, spherical melanosomes of different sizes. Entubulated morphologies are detected within the lumen of outer retinal tubulation.²⁷ These originate mainly from sloughed RPE in intact retina peripheral to the atrophic area and account for the hyperreflective spots distinguished by SD-OCT in these formations. Hyperreflective foci migrate into inner retinal layers, and their quantity and density increase first perifoveally and then foveally in a 2-year period marked by incidence of geographic atrophy.⁴ 

In the present study we measured the HF number within two regions. Since there is less retinal tissue along a line segment of length 1 mm passing through the fovea, we speculated that HF amount in the central subfield may not completely reflect their behavior; therefore, we additionally counted these foci along a line segment of length 3 mm passing through the fovea, which includes the parafoveal region. The results of this study revealed that the mean number of HF in the radius of 500 μm decreased significantly 1 month after three intravitreal aflibercept injections, and this reduction had a rather weak but significant correlation with the reduction of CST. Baseline HF number in that area was positively correlated with baseline CST and macular volume, but no correlation was found between the final number of these foci and final CST or final macular volume. The baseline HF number in the radius of 1500 μm had also a rather weak but significant correlation with baseline CST. Despite the significant reduction of HF in the radius of 1500 μm, we could not find any significant correlation with the evaluated parameters. Although the size of CNV decreased after treatment as well, no significant correlation was found between HF reduction and CNV size decline. 

To the best of our knowledge, this is the first study to evaluate the correlation between CNV size and HF amount in the cross-sectional OCT B-scans after switching from ranibizumab to aflibercept in patients with treatment-resistant neovascular AMD. Akagi-Kurashige and associates reported that eyes with HF at baseline had more active CNV and more severe blood–retinal barrier damage.²⁸ A small percentage of cases without HF in their study (18.7%) had milder CNV and less severe blood–retinal barrier damage, but they did not measure the CNV size. Initial presence of HF in the foveal neurosensory retina was associated with poor final visual acuity in their study. Framme et al.²⁹ studied changes in the foveal and parafoveal area before and 1 month after the third ranibizumab injection in 61 treatment-naïve eyes with neovascular AMD. The amount of small dense particles (SDPs, equivalent to HF) decreased significantly after therapy, and SDP reduction correlated weakly but positively with reduction of retinal thickness (r = 0.268, P = 0.040). Unlike what was seen in our study, they found a positive correlation between SDP reduction and improvement in BCVA (P = 0.006). Contrary to the results of Akagi-Kurashige et al.,²⁸ Framme et al.²⁹ found that the baseline amount of SDPs correlated positively with the increase in BCVA (P = 0.005). They hypothesized that initial large number of HF may indicate a higher grade of inflammation, but the presence of a high number improves the effect of ranibizumab therapy in neovascular AMD, and they concluded that the amount of SDPs could be a predictive factor for the treatment outcome. Kang et al.⁷ evaluated HF in 97 patients who were treated with intravitreal bevacizumab for macular edema secondary to branch retinal vein occlusion. The eyes were divided into three groups based on the location of HF on SD-OCT: HF in outer retinal layers, HF in inner retinal layers, and no HF. Final BCVA was worse in eyes with HF in the outer retinal layer group than in the other two groups (P < 0.001), although baseline BCVA was not different between them. We did not investigate the presence of HF in different retinal layers, as the small sample size after subdivision could limit the statistical analysis. 

The strengths of this study are its prospective design, measurement of the CNV size, and definition of the area for HF evaluation. Limitations are a relatively small sample size and effects of previous treatment with ranibizumab on the retinal structure. It would be preferable to count HF on at least three representative equally spaced line segments of 1 mm, and five such of 3 mm, so that the three segments of 1-mm length are a subset of those of 3-mm length. Furthermore, the whole size of CNV cannot be measured on a single OCT B-scan. Since almost all of the patients had been treated with ranibizumab outside of our hospital before switching to aflibercept, we encountered missing data including CST before switching to aflibercept and behavior of HF under ranibizumab therapy. Due to short follow-up duration, we did not investigate the relationship between the mentioned parameters and clinical improvement or recurrence either. The current version of Heidelberg Eye Explorer software allows calculating the area of irregularly shaped CNVs, but it is not able to detect HF automatically; therefore we counted the HF number subjectively. 

In summary, SD-OCT serves as a valuable tool for evaluation of changes in retinal structure during treatment and follow-up of patients with neovascular AMD. Hyperreflective foci are observed frequently in the foveal region of patients with treatment-resistant AMD. Switching from ranibizumab to aflibercept caused a predominant decrease in the number of HF within the area of 1- and 3-mm diameter 1 month after aflibercept upload. Despite the significant improvement in BCVA, reduction of CNV size, decrease in CST, and decline in macular volume, only the correlation between HF reduction in the radius of 500 μm and CST decrease was statistically significant. Further studies with a longer follow-up period are needed to evaluate the time-dependent changes. Specific software could be developed for objective assessment of HF. 

Acknowledgments

The authors thank the German Ministry of Lower Saxony for its scientific and cultural organization to support our retinal imaging studies in the ophthalmology department of Hannover Medical School financed by Niedersachsen Vorab. 

Disclosure: K. Abri Aghdam, None; A. Pielen, None; C. Framme, None; B. Junker, None