Geographic Atrophy and Activity of Neovascularization in Retinal Angiomatous Proliferation

Abstract

Purpose: To investigate the association between geographic atrophy (GA) and neovascular activity in retinal angiomatous proliferation (RAP) during anti-VEGF treatment.

Methods: Ninety-one RAP eyes (74 patients) treated with anti-VEGF on an as-needed basis for at least 3 years were evaluated. Development of GA, area of GA, and injection numbers were assessed.

Results: Eighteen eyes that developed fibrous scar or massive hemorrhage were excluded. Forty-four eyes (60%) developed GA (GA group), and 29 eyes (40%) did not develop GA (non-GA group) at year 3. The mean injection number continuously decreased in the GA group (5.1, 3.1, and 1.9 at years 1, 2, and 3, respectively, P < 0.01, < 0.01), but did not decrease at year 3 in the non-GA group (4.6, 3.5, and 3.3 at years 1, 2, and 3, respectively, P < 0.01, = 0.64). In both groups, best-corrected visual acuity improved significantly at year 1 and declined to baseline level at year 3. During the entire follow-up (mean of 48.5 months), 57 eyes developed GA. In those eyes, number of injections per year before and after the development of GA was 4.5 and 2.1 (P < 0.01), and showed continuous decline after GA development as the area of GA progressed at a rate of 1.85 mm² per year.

Conclusions: The activity of RAP lesion requiring anti-VEGF treatment diminished as GA developed and progressed. Identification of GA and its progression provides further information to tailor anti-VEGF treatment for each patient.

Retinal angiomatous proliferation (RAP), which was also termed “type 3 neovascularization,” is a distinct subgroup of neovascular AMD characterized by the origination of new vessels from the retina or from the choroid with early formation of a retinal choroidal anastomosis.^1–3 The retinal circulation as well as choroidal circulation contributes to the vasogenic process. Owing to high vasogenic potential, RAP has a very poor functional prognosis without treatment, and it typically progresses to increasingly severe stages before ultimately developing a disciform scar.⁴ With the advent of anti-VEGF drugs, the treatment of RAP has benefited profoundly, as have other types of neovascular AMD.^5–10 

The progressive nature of neovascular AMD was well established through several extension studies.^11–13 There was an incremental decline in initial visual gain with less intensive treatment during further follow-up, leading to an overall decline in visual acuity (VA) up to nine letters after 7 years.¹³ In a recent study using database information of more than 1200 eyes, mean VA remained above the baseline level for approximately 6 years, which is better than the visual outcomes in previous extension studies.¹⁴ It may be attributable to more frequent injections in that study, in which a median of six injections were given over the first 12 months, followed by five injections per annum thereafter for 7 years. These results suggest that, in general, the number of injections required to control neovascular activity in AMD would not decrease over a long term. However, there are few studies reporting long-term outcomes of VEGF inhibition for RAP in terms of disease activity and treatment frequency. 

Geographic atrophy (GA) is a common feature of RAP, irrespective of treatment.^15–18 In one study, approximately 40% developed de novo GA after 2 years of anti-VEGF treatment.¹⁵ Another study revealed that development of GA or enlargement of preexisting GA was noted in almost all eyes during the posttreatment follow-up.¹⁶ Large soft and/or confluent drusen are a significant risk factor for developing GA, and almost always are present in the macular area in patients with RAP. However, they are seen infrequently in Asian patients with other subtypes of neovascular AMD, particularly polypoidal choroidal vasculopathy. This different baseline characteristic among our patients provided an opportunity to observe strikingly higher incidence of GA development in patients with RAP during anti-VEGF treatment, and to observe a trend of decreases in the required number of injections as follow-up period lengthened in those patients. 

The hypothesis is that the number of RPE cells, the main sites of VEGF production, would decrease in proportion to an increase of GA area, leading to decrease of neovascular activity. We carried out this study to investigate whether treatment frequency decreased as GA developed and progressed in patients with RAP who received anti-VEGF mono-treatment. 

Materials and Methods

This retrospective cohort study was performed at the Department of Ophthalmology in Seoul St. Mary's Hospital, The Catholic University of Korea. This study was approved by the hospital's institutional review board and conducted according to the Declaration of Helsinki. 

Patients

We reviewed medical charts of 152 patients who were diagnosed as having RAP between 2008 and 2012 and who were treated with intravitreal anti-VEGF injections. The diagnosis of RAP was based on the characteristic features detected by funduscopic examination, fluorescein angiography (FA), and optical coherence tomography (OCT), and confirmed by the indocyanine green angiogram (ICGA) finding of a connection between the neovascular complex and the retinal vasculature. It is often difficult to determine whether the serous pigment epithelial detachment (PED) was associated with subretinal neovascularization (stage 2 RAP¹) or choroidal neovascularization (stage 3 RAP). In this study, RAP was classified as early stage (RAP without PED), which is the equivalent to stage 1, and as advanced stage (RAP with PED), which includes stages 2 or 3. 

Initially treatment-naïve RAP eyes treated with anti-VEGF monotherapy and followed longer than 3 years were included. All patients were treated based on as-needed dosing (pro re nata [PRN]) without three loading injections, using 0.5 mg ranibizumab (Lucentis; Genentech, Inc., South San Francisco, CA, USA) or 1.25 mg bevacizumab (Avastin; Genentech, Inc.). Selection between the two drugs was determined mainly by considering the national health insurance coverage and reimbursement schedules. Follow-up examinations were scheduled regularly at 1- to 3-month intervals, depending on lesion activity. Retreatment was given whenever intraretinal fluid, subretinal fluid, or PED was detected on follow-up OCT. Exclusion criteria were any of the following eye conditions: any kind of previous treatment for RAP lesion, GA at baseline, subfoveal fibrosis at baseline, other concomitant ocular diseases that could affect VA, and prior vitrectomy. 

All patients underwent a complete ocular examination, including Snellen best-corrected VA (BCVA), digital color fundus photographs, near infrared (NIR) fundus photographs, and OCT images, which were obtained at baseline and at each follow-up visit. Near infrared images were obtained using confocal scanning laser ophthalmoscope (Heidelberg Retina Angiograph [HRA]; Heidelberg Engineering, Heidelberg, Germany) with 787-nm wavelength light. Optical coherence tomography images were obtained with Stratus (Carl Zeiss Meditec, Dublin, CA, USA) or Cirrus OCT (Carl Zeiss Meditec) before November 2011, and subsequently with a Spectralis OCT (Heidelberg Engineering) with enhanced depth imaging (EDI) protocol. Fluorescein angiography and ICGA using HRA were performed at baseline and when required. 

Analysis of Images

Near infrared and OCT images were used for GA measurement. A high signal area on NIR and characteristics such as variable shape, sharply demarcated border, and visibility of underlying choroidal vessels with a punched-out appearance were required to define GA. Finally, an atrophic RPE layer was confirmed on corresponding OCT scans.¹⁹ Confirmed areas of GA were outlined on the NIR image using Image J (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health, Bethesda, MD, USA).^18,20 The dimension and the pixels of each NIR image were found under the original image property. The area of image in square millimeters divided by the total number of pixels gives a conversion factor in square millimeters per pixel. Image J calculates the number of pixels within the atrophic areas and the number of pixels was converted to square millimeters by multiplying the conversion factor (Fig. 1). 

Figure 1

 

Measurement of GA area. Confirmed areas of GA were outlined on the NIR image using Image J, which calculates the number of pixels within the atrophic areas. The area of atrophy is equal to the conversion factor multiplied by the pixels of atrophy. The numbers indicate areas of each segment of atrophy converted to mm². The areas indicated on NIR image by black arrows are indicated by corresponding white arrowhead on the OCT image. Absence of RPE in hyperreflective GA area identified using NIR is seen in the OCT image.

In this manner, the areas of GA were recorded at baseline and each year. Choroidal thickness was defined as the vertical distance from the hyperreflective line of the Bruch membrane to the innermost hyperreflective line of the chorioscleral interface based on 1:1-pixel image and was measured manually by using the Heidelberg Eye Explorer software version 5.8.3.0. Patterns of GA were categorized into multifocal or confluent according to their lesion characteristics on NIR image. Two graders (JB and JHL) independently evaluated the presence and area of GA. The supervising grader (WKL) evaluated in case of significant discrepancy. 

Statistical Analysis

Statistical analysis was performed with a commercial program (SPSS for Windows, version 19.0.1; SPSS, Inc., Chicago, IL, USA); P less than 0.05 was considered statistically significant. Snellen VA was converted to logMAR units for statistical analysis. A paired t-test and Wilcoxon signed rank test were used to compare the mean number of injections, GA area, BCVA, and subfoveal choroidal thickness at each time point with the baseline and previous values. Continuous variables between each time point and groups were compared using paired t-test after confirmation of normal distribution using Kolmogorov–Smirnov test. Categorical variables between groups were compared using the χ² test. 

Results

A total of 91 eyes from 74 patients with RAP were included in this study. Both eyes were involved in 17 patients (18%). All patients were Korean, and the mean age was 74.2 ± 6.4 years (range, 57–95 years). Sixteen patients were male, and 58 were female. All patients were followed for at least 3 years (mean follow-up: 48.5 ± 13.4 months, range 36–84 months), and the mean number of follow-up visits was 17.0 ± 5.7, range 5–30). Twelve eyes (13%) were treated with bevacizumab, 21 eyes (23%) with ranibizumab, and 58 eyes (64%) with both drugs. Baseline logMAR BCVA was 0.49 ± 0.21 (range, 0.1–3.0) (Snellen equivalent: 20/62). Subfoveal choroidal thickness measurement at baseline was available in 54 eyes, and the mean was 139.5 ± 55.9 μm (range 28–322 μm). Subretinal fibrosis and/or massive hemorrhage developed in 18 eyes (20%), and these eyes were excluded from the analysis. 

Comparison Between GA Group and Non-GA Group During 3 Years of Follow-Up

Newly developed GA was found in 44 (60%) of 73 RAP eyes during 3 years of follow-up (GA group): 8 (cumulative incidence [CI]: 0.11) in the first year, 21 (CI: 0.40) in the second year, and 15 (CI: 0.60) in the third year. Of those eyes, 30 (68%) eyes were multifocal pattern and 14 (32%) eyes were confluent pattern. Twenty-nine (40%) eyes did not develop GA during the 3-year follow-up period (non-GA group). There was no difference in anti-VEGF drug types (bevacizumab, ranibizumab, and both) between groups (P = 0.27). Retinal angiomatous proliferation stage at baseline did not differ between groups: eyes with advanced stages were 31 (70%) of 44 and 19 (66%) of 29 in GA and non-GA groups, respectively (P = 0.66). 

The overall number of injections did not differ between the two groups (a mean of 10.1 ± 5.8 in the GA group and 11.4 ± 5.5 in the non-GA group, P = 0.25). In the GA group (44 eyes), the numbers of injections were 5.1 ± 2.0, 3.1 ± 2.4, and 1.9 ± 2.21 in years 1, 2, and 3, respectively (all P < 0.01 compared with previous year). In the non-GA group (29 eyes), the number of injections decreased from 4.6 ± 1.7 in the first year to 3.5 ± 2.0 in the second year (P < 0.01), but the number remained steady at 3.3 ± 2.9 in the third year (P = 0.64). The number of injections did not differ between groups in the first and the second years (P = 0.32, P = 0.51, respectively), but the number was lower in the GA group in the third year (P = 0.02) (Fig. 2, left). 

Figure 2

 

Comparison of annual numbers of injections, VA, and subfoveal choroidal thickness in retinal angiomatous proliferation eyes that developed GA (GA group) and in which GA did not develop (non-GA group) during 3 years. (Left) Graph showing number of injections per year. Number of injections decreased each year in the GA group but was steady after the first year in the non-GA group. (Middle) Graph showing changes of logMAR VA. Visual acuity at year 1 in both groups; however, it showed no significant differences in year 3 from baseline. Visual acuity at baseline and at year 3 did not differ between groups, although VA in years 1 and 2 were better in the non-GA group. (Right) Graph showing choroidal thickness. Choroidal thickness was significantly thinner in the GA group and decreased each year in both groups. Error bar indicates SE. *Paired t-test compared with previous year; †paired t-test compared with baseline; ‡t-test between GA and non-GA groups.

In the GA group, the mean logMAR BCVA was 0.47 (Snellen equivalent: 20/59) at baseline and 0.41, 0.42, and 0.47 (20/51, 20/53, and 20/59; P = 0.01, 0.08, and 0.97 compared with the baseline) at years 1, 2, and 3, respectively. In the non-GA group, this value was 0.40 at baseline (Snellen equivalent: 20/50) and 0.30, 0.30, and 0.40 (20/40, 20/40, and 20/50; P = 0.01, 0.03, and 0.92 compared with the baseline) at years 1, 2, and 3, respectively (Fig. 2, middle). Geographic atrophy involved the foveal center in 10 eyes (23%) in the GA group, and the mean logMAR VA at year 3 in these eyes was 0.73 (Snellen equivalent: 20/107). Existence of fluid, requiring retreatment, at approximately year 3 (34 to 38 months from the initial treatment) was observed in 13 eyes (29%) in the GA group and 19 eyes (65%) in the non-GA group (P < 0.01). 

Initial subfoveal choroidal thickness measurement using spectral-domain OCT EDI protocol was available in 30 eyes in the GA group and 17 eyes in the non-GA group. The mean subfoveal choroidal thickness at baseline was 121.9 ± 37.8 μm in the GA group and 175.1 ± 68.7 μm in the non-GA group, which had a significant difference (P < 0.01). Subfoveal choroidal thickness decreased each year in both groups: 109.2 μm, 99.7 μm, and 96.6 μm in the GA group and 159.5 μm, 154.5 μm, and 151.4 μm in non-GA group at years 1, 2, and 3, respectively (all P < 0.05) (Fig. 2, right). 

Number of Injections and GA Area in Eyes That Developed GA During Total Follow-Up Period

A total of 57 eyes developed GA during a mean of 48.5 ± 13.4 months of follow-up; 13 of 29 non-GA eyes had newly developed GA after 3 years. In these eyes, the mean number of injections per year was 4.5 ± 2.0 times before the development of GA and 2.1 ± 1.7 times after the development of GA (P < 0.01). The mean follow-up period was 28.6 ± 11.1 months before the development of GA and 21.6 ± 16.9 months after GA (P = 0.22). Calculated mean interval for treatment was 2.7 and 6.0 months before and after development of GA, respectively. Number of injections decreased each year after the development of GA: 3.0 ± 2.3, 1.9 ± 2.0, and 1.4 ± 1.6 at years 1, 2, and 3, respectively (P < 0.01, < 0.01). Areas of GA after the development of GA were 2.38 ± 2.99 mm², 4.41 ± 4.70 mm², and 5.93 ± 4.96 mm² at years 1, 2, and 3, respectively (P < 0.01, < 0.01; progression rate: 1.85 mm² per year) (Fig. 3). Figure 4 shows a typical case in which required injection number was decreased as the GA area enlarged. 

Figure 3

 

Change of GA area and number of injections each year after the development of GA in RAP eyes that developed GA during total follow-up period. Bar graph showing increasing GA area each year and polygonal line showing decreasing number of injections each year after the development of GA. Error bar indicates SE. *Paired t-test compared with previous year; †Wilcoxon signed rank test compared with previous year.

Figure 4

 

A case of RAP with GA. Baseline color fundus photograph (A), FA (B), and ICGA (C) of the left eye of a patient with RAP. Serial follow-up infrared images demonstrate development and enlargement of GA during follow-up and decreasing annual number of injections activity after the development of GA (D–F).

Discussion

The constitutive expression of VEGF and its receptors in normal adult neuroretina and choriocapillaris suggests that it plays an important role in the survival of neuronal as well as vascular elements.²¹ Thus, inhibiting VEGF for a long term could conceivably lead to GA. The Comparison of Age-related Macular Degeneration Treatments Trials (CATT) Research Group reported that anti-VEGF treatment might increase the risk of new GA, and RAP as well as monthly dosing were included in the independent risk factors.¹⁷ The anatomical and pathophysiologic changes associated with a prominent feature of RAP, greater extent and density of drusen than the other forms of exudative AMD, may provide more appropriate milieu for GA development.^3,22 Our study results support this. Of 91 eyes, 11%, 40%, and 60% had de novo GA at years 1, 2, and 3, respectively, a finding that is compatible with other studies of RAP patients.^15,16 The incidence at 2 years (40%) is approximately twice that of the incidence (18%) of the CATT study, which includes all forms of neovascular AMD.¹⁷ 

Our study was intended to reveal that neovascular activity and treatment frequency decreased as GA developed and progressed in eyes with RAP. The number of injections per year decreased continuously during 3 years in the GA group; however, this trend was not observed in the non-GA group. In addition, in the GA group, the mean number of injections per year was significantly higher before GA development than after GA development (4.5 vs. 2.1). Calculated mean interval of treatment was significantly shorter before GA development than after (2.7 vs. 6.0). As the area of GA increased, the required number of injections was decreased. These findings may suggest the activity of neovascular lesion requiring treatment diminished as GA developed and progressed. A possible explanation for this is that GA accompanied reduction of RPE cells, resulting in decrease of VEGF secretion. Retinal pigment epithelium cells produce a large proportion of VEGF that is responsible for new vessel formation and activation.²³ Also, decreased secretion of VEGF can impair choriocapillaris survival, as stated earlier, and may lead to further atrophy of choroidal vessels. Experimental studies showed that RPE destruction leads to choriocapillaris atrophy.^24,25 As a result, decreased perfusion from the choroid to neovascular complex may also contribute to the decreased activity. 

Notably, the mean subfoveal choroidal thickness at baseline was thinner in the GA group. This finding is in agreement with a previous study by Cho and associates¹⁵ that reported thin choroid as a risk factor of development of GA in RAP. It has been well documented that choroidal blood flow and volume in patients with GA is impaired compared with age-matched controls.²⁶ Impaired choroidal perfusion may result in decreased oxygenation of outer retina, decreased metabolic exchange, and choriocapillary atrophy, which leads to GA development.^15,24,27 Choroidal thinning seems to be both a result and cause of GA. 

Furthermore, subfoveal choroidal thickness continuously decreased not only in the GA group but also in the non-GA group, suggesting that factors other than the loss of RPE cells might be involved in the process. There is accumulating evidence that intravitreally injected anti-VEGF has a pharmacologic effect on the underlying choroid as well as the neovascular membrane.^28,29 In that case, anti-VEGF therapy could inhibit choriocapillaris survival directly, and this effect would be manifested by reduction of subfoveal choroidal thickness. 

Significant gain of VA was observed at year 1 in both groups and began gradually to slip to the baseline level at year 2 or 3, which is compatible with the results of other studies.^7,10 As we treated patients on an as-needed basis, the loss of initial visual gain can be explained with undertreatment. This may reflect the inherent nature of the disease even during the course of treatment, such as development of neovascularization-related macular atrophy or structural destruction of fovea due to long-standing or repetitive exudative changes.³⁰ In the GA group, GA did not involve the foveal center during 3 years in most cases (77%), and the mean VA did not worsen below the baseline. Visual acuity may decrease further with longer follow-up in cases in which GA extends to the foveal center. 

Our study had some inherent limitations owing to its retrospective design. Patients were not treated with a standardized protocol and regular follow-up period. Nonetheless, we collected a large number of cases, and the treatment was administered by a single retinal specialist (WKL) with a constant retreatment criteria. Two different anti-VEGF drugs were used in this study. And PRN protocol might increase the opportunity of repetitive fluid accumulation compared with proactive protocol, which induces atrophic changes. However, the main purpose of our study is not to know the incidence and risk factors of GA, but to know the relationship between GA and activity of RAP lesion requiring treatment. These limitations may not affect the main result of our study. Also, there might be a loss to follow-up bias in the data from the fourth year. These data were used only for the analysis of changes of treatment frequency according to GA development and progression. Finally, interpretation of GA in wet AMD is more complicated than that in dry AMD owing to secondary atrophic changes associated with neovascular process and diverse clinical presentation of the disease itself. We excluded eyes that developed fibrous scarring or massive hemorrhage at baseline and during follow-up. Atrophic changes would develop around the disciform scars and at the area corresponding to previous hemorrhage that should be distinguished from GA.^31,32 Although autofluorescence imaging, a useful imaging modality for the analysis of GA, was not used in this study, we used NIR in combination with OCT, thus allowing clear identification of the borders of GA. 

In conclusion, the development and growth of GA decreases activity of neovascular lesion and number of anti-VEGF injections required in eyes affected by RAP. These findings have clinical implications in determining the optimal anti-VEGF dose for each patient. 

Acknowledgments

Disclosure: J. Baek, None; J.H. Lee, None; J.Y. Kim, None; N.H. Kim, None; W.K. Lee, Novartis (C, S), Bayer (C, S), Allergan (C, S), Alcon (C, S), Santen (C, S)