Choroidal Neovascular Membranes Treated With Photodynamic Therapy

The Treatment of Age-Related Macular Degeneration With Photodynamic Therapy (TAP) Study Group^1,2 reports 1 and 2 demonstrated a statistically significant reduction in moderate and severe vision loss in patients with predominantly classic subfoveal choroidal neovascular membranes (CNVMs) secondary to age-related macular degeneration(ARMD) treated with intravenous verteporfin (Visudyne; CIBA Vision Corp, Duluth, Ga) and photodynamic therapy (PDT). Patients were followed up every 3 months after PDT treatment with fluorescein angiography. If leakage was identified from the CNVM, patients underwent repeated treatment with verteporfin PDT.

Verteporfin PDT treatment was also associated with fewer eyes experiencing progression of classic CNVM beyond the area of the lesion identified at baseline. At the 12-month follow-up visit, 46% of the verteporfin-treated eyes vs 71.1% of the placebo-treated eyes had documented growth of the CNVM on fluorescein angiography.^1,2

We describe the clinical course and histopathologic findings in 5 patients with predominantly classic subfoveal CNVM secondary to ARMD who had progression of classic CNVM, despite treatment with verteporfin PDT (Figure 1). These patients underwent submacular surgery with excision of their CNVM. The excised membranes were examined using light and transmission electron microscopy.

The patients' clinical history, surgical report, and fluorescein angiograms were reviewed. Visual acuity was measured using an illuminated Early Treatment of Diabetic Retinopathy Study chart. The logarithm of the minimal angle of resolution (log MAR) units were calculated by obtaining the logarithm of the reciprocal of the Snellen visual acuity. The log MAR values were then converted back into Snellen visual acuity to report outcomes. The greatest linear dimension and 2-dimensional size of the CNVM lesion on digital fluorescein angiography were measured using the IMAGEnet (TOPCON America Corp, Paramus, NJ) measure function. These results are found in Table 1.

After signing an informed consent, patients underwent a conventional 3-port pars plana vitrectomy. In 1 case (patient 4), the posterior hyaloid was bimanually separated from the optic nerve and retina. The patients' most recent fluorescein angiograms were used to determine a suitable location for the retinotomy, usually temporal to the macula in an area devoid of large retinal vessels. A subretinal pic was used to create a retinotomy, lyse any retinochoroidal vessels or adhesions, and gently elevate the edge of the membrane from the underlying tissue. With the intraocular pressure elevated to 60 mm Hg, horizontal subretinal forceps were used to grasp the subretinal membrane and slowly deliver it through the retinotomy. The intraocular pressure was then slowly lowered back to normal while the macula was inspected for any subretinal bleeding. The membrane was removed from the eye, placed on a dry sponge, and placed in fixative. The orientation of the specimen was not specified in any of the cases. Each patient underwent a 360° scleral depressed examination with an indirect ophthalmoscope before having a complete fluid-air exchange performed.

Four (patients 1 through 4) of the 5 patients had their excised CNVM fixed in 10% formalin for paraffin embedding and light microscopy. In 1 case(patient 5), the CNVM was fixed in 2.5% glutaraldehyde and 2% formaldehyde in 0.1M cacodylate buffer and embedded in EPON for 1-µm sectioning and transmission electron microscopy.

The patients ranged in age from 75 to 86 years. They all had predominantly classic subfoveal CNVM secondary to ARMD. Their mean initial visual acuity was 20/300, greatest linear dimension was 2.68 mm, and CNVM lesion size was 3312 µm². They received intravenous verteporfin (6 mg/m² of body surface area) over a 10-minute period. Five minutes after the end of the infusion, a 689-nm laser was applied over the entire lesion for 83 seconds. At the first 3-month follow-up examination, the CNVM had grown in all of the cases and measured on average 8454 µm². Four(patients 1, 3, 4, and 5) of the 5 patients had repeated verteporfin PDT, while 1 underwent submacular surgery and excision of the CNVM. At the subsequent 3-month follow-up visit, the mean size of the PDT-treated CNVM was 11 310µm². Patient 1 demonstrated retraction of the membrane at one edge and growth at another edge, with an overall decrease in the size of the membrane. At this point, the remaining patients elected to undergo submacular surgery with CNVM excision. All of these patients underwent submacular surgery 3 to 4 months after their last PDT treatment (Table 1).

During pars plana vitrectomy, 4 (patients 1, 2, 3, and 5) of the 5 patients had a posterior vitreous detachment present, and 1 required peeling of the posterior hyaloid. In 2 cases (patients 1 and 4), the CNVM was stiff and enlarged the retinotomy on extraction through the retina. None of the patients had laser photocoagulation of the edges of the retinotomy site, and all of these sites remained flat after surgery. No other complications such as subretinal hemorrhage, peripheral retinal tear, or retinal detachment were identified.

A light micrograph from case 4 of the central aspect of the CNVM demonstrated large red blood cell–filled vessels within a thick, predominantly fibrous membrane (Figure 2A). In contrast, the peripheral edges of the CNVM were filled with smaller red blood cell–filled vessels within a thin, more cellular membrane (Figure 2B). Among all cases, retinal pigment epithelial (RPE) cells were seen on 1 side of the membranes and were sometimes found layered within a membrane. Patient 5 revealed a single layer of cuboidal, slightly pigmented, polarized RPE cells covering the presumed neurosensory retinal surface of the membrane (Figure 3). None of the membranes had evidence of giant cells.

A transmission electron micrograph from patient 5 revealed a stroma containing abundant collagen fibrils and fibroblasts with dilated rough endoplasmic reticulum and nuclei with dispersed chromatin (Figure 4A). Myofibroblasts with hemidesmosomes and pigment-laden macrophages with filopodia were present, as well as inflammatory cells such as lymphocytes. Proliferating RPE cells were rare in the stroma of this CNVM. Capillary endothelial cells and adjacent pericytes sometimes shared the same reduplicated basement membrane (Figure 4B). In addition, larger vessels surrounded by smooth muscle cells were identified. These vessels were filled with numerous enlarged endothelial cells that severely reduced or eliminated the vessel lumen (Figure 4C).

One side of the CNVM had multiple layers of RPE cells filled with vesicles, dilated rough endoplasmic reticulum, and irregular pigment granules. Long-spaced collagen, an RPE basement membrane (Figure 5A), and, occasionally, capillaries were identified in the subepithelial space. This was considered to be the native RPE layer that covered the Bruch membrane, although no collagenous or elastic components of the Bruch membrane or choroidal vessels were identified. Lightly pigmented cuboidal cells covered the side of the membrane presumably in contact with the neurosensory retina. These cells possessed apical stunted microvillous processes, tight junctions, single melanosomes and melanosomes jacketed by lipofuscin, basally concentrated mitochondria, crude basal infoldings, and a rudimentary basal lamina consistent with polarized RPE cells (Figure 5B).

We describe 5 patients who had recurrent growth of a classic subfoveal CNVM associated with ARMD, despite treatment with verteporfin PDT. Three (patients 2, 3, and 5) of the 5 patients had initial visual acuity worse than 20/200, the lower limit in the Treatment of Age-Related Macular Degeneration With Photodynamic Therapy study.¹ Growth of the CNVM was associated with worsening vision. Pars plana vitrectomy and excision of the CNVM were performed in all of these patients. In 2 of the patients(patients 1 and 4), the membranes were stiff and produced a moderate enlargement of the retinotomy site as they were extricated from the subretinal space. No other complications were associated with surgery. The mean postoperative visual acuity was 20/500, after a mean of 4 months' follow-up.

Fluorescein angiography in all of these patients revealed persistent growth of the CNVM beyond the area of the lesion documented on baseline angiography. The central body of the membrane tended to stain on the late frames of the angiogram, suggesting the presence of fibrous tissue. In contrast, the edges of the enlarging membrane were often hyperfluorescent early and leaked late in the angiogram, consistent with actively growing, highly permeable vessels.

The fluorescein angiographic findings in these patients are consistent with the histopathologic analyses of the CNVM. The central area of the membranes was thick and fibrous, while the edges of the membrane were diaphanous and cellular. Furthermore, the central area of the membrane often harbored large red blood cell–filled vessels, while the periphery of the membrane was usually filled with smaller red blood cell–filled vessels.

Ghazi et al³ found histologic evidence of endothelial degeneration and vascular occlusion in a surgically excised membrane 27 days after PDT treatment. In contrast, we did not find any evidence of acute vascular thrombosis or injury in our membranes excised more than 90 days after PDT treatment. This difference may be due to a longer interval between PDT and surgical excision, resulting in more mature CNVMs.

Our surgically excised CNVM specimens showed evidence of vascular repair. Capillaries with reduplication of the basement membrane were seen. This finding has also been documented in surgically excised CNVM not previously treated with PDT, ⁴ in CNVM previously treated with PDT 27 days before surgery, ³ and in cynomolgus monkey CNVM and normal choriocapillaris previously treated with PDT.⁵ This may be consistent with a significant antecedent vascular injury such as PDT and a subsequent endothelial revascularization of the ghost vessel, leading to reduplication of the capillary basement membrane. Larger vessels surrounded by smooth muscle cells had numerous swollen endothelial cells that narrowed the vascular lumen. Although similar endothelial cells have been described within large vessels surrounded by smooth muscle in CNVM, ⁴ we believe that the number of vessels and the degree of lumen narrowing present are remarkable. This finding may be consistent with endothelial cell revascularization within larger, more mature vessels 3 months after PDT-induced endothelial damage.

Retinal pigment epithelial cells were intimately associated with all of these membranes. As seen in other histologic studies, ^4,6,7 the RPE cells were found predominantly on 1 side of the CNVM but were also found layered within the membranes. It is difficult to orient surgical specimens of subretinal CNVM using only light microscopy. In patient 5, ultrastructural analysis revealed long-spacing collagen associated with the presumed native RPE cells. The CNVM in this patient appears to have grown over the native RPE cells in a fashion consistent with a type 2 membrane.⁶

In patient 5, a single layer of pigmented polarized RPE cells with apical tight junctions, apical microvilli, basally concentrated mitochondria, and a rudimentary basement membrane covered the neurosensory retinal surface of the membrane. Encapsulation of the CNVM by RPE cells has been seen in animal models of CNVM treated with PDT^5,8 and has been reported in surgically excised CNVM in patients with ARMD.⁷ Our findings in patient 5 indicate that RPE cells in patients with ARMD are capable of covering the neurosensory surface of a PDT-treated CNVM.

Taken together, we speculate that PDT-induced occlusion of CNVM vessels, and possibly the underlying choriocapillaris, ⁸ result in a subretinal environment that is more hypoxic than it is at baseline. Some vessels may not recover from this vascular insult, ³ while others, spurred on by such vascular mediators as vascular endothelial growth factor, ⁹ may be able to revascularize themselves via endothelial cell proliferation. Larger caliber vessels, with more rapid blood flow, may be more resistant to the injurious effects of PDT. This would be consistent with the clinical finding that small CNVMs have better outcomes after PDT compared with larger CNVMs.^1,2

Patients treated with verteporfin and PDT for subfoveal, predominantly classic CNVM secondary to ARMD may progress, despite repeated PDT treatment. Progressive growth of the CNVM is associated with loss of vision. Choroidal neovascular membranes previously treated with PDT appear to be composed of large vessels within a predominantly fibrous body and smaller vessels in the more cellular periphery of the CNVM. In some cases, RPE cells may cover the neurosensory retinal surface of PDT-treated CNVM.

Dr Arroyo is a recipient of a K-23 Physician Training Award from the National Institutes of Health, Bethesda, Md.

Corresponding author and reprints: Jorge G. Arroyo, MD, Retina Service, Department of Ophthalmology, Massachusetts Eye and Ear Infirmary, 243 Charles St, 12th Floor, Boston, MA 02114 (e-mail: jgarroyo@meei.harvard.edu).