Sequence of Early Vascular Events after Photodynamic Therapy

Abstract

purpose. To identify early vascular changes in choroidal neovascularization (CNV) and in adjacent normal choroid, after photodynamic therapy (PDT).

methods. In a prospective study, 40 patients with predominantly classic CNV due to age-related macular degeneration (AMD) were treated with PDT performed with verteporfin. Verteporfin was administered intravenously at a dose of 6 mg/m² body surface area. A near infrared laser light dose of 50 J/cm², an irradiance of 600 mW/cm² and a wavelength of 692 nm was applied. A scanning laser system was used to perform confocal fluorescein angiography (FA) and indocyanine green angiography (ICGA) before treatment and regularly at 5 hours, 1 day, 1 week, and 3 months after PDT. Images were analyzed for CNV size and leakage area as seen by FA and ICGA. Collateral damage within the surrounding choroid was documented based on the hypofluorescence in early- and late-phase ICGA.

results. No immediate occlusion of the CNV complex was found angiographically, but a dynamic change over time was observed in the early perfusion patterns and late-phase hyper- and hypofluorescence. At 5 hours after treatment, large portions of the CNV lesion were still perfused. One day after PDT, CNV size in early FA and early ICGA reached its minimum, at 0.49 mm² (15.7%) and 0.78 mm² (31.1%) of the initial area, respectively. In late-phase FA and ICGA, however, an immediate massive exudation with a continuous increase in hyperfluorescence originated from the CNV and surrounding choroid, with a maximum in leakage area at 1 day. At 1 week PDT-induced exudation slowly resolved. Eyes in 36 patients showed some choroidal hypofluorescence by ICGA before treatment. A progressive increase of the hypofluorescent area surrounding the CNV was observed, which correlated with the size of the laser spot. Maximum hypofluorescence was noted at 1 week with an average size of 11.1 mm² in early- and late-phase ICGA.

conclusions. In contrast to findings in experimental animals, PDT in humans with classic CNV did not induce immediate thrombosis, but primarily caused a breakdown of vascular barriers. A characteristic sequence of vascular changes was observed with early, enhanced leakage from the CNV and normal choroid followed by nonperfusion later. Occlusion of the CNV lesions occurred 1 day after treatment, but closure of the adjacent choroidal vessels proceeded slowly over as long as 1 week.

Photodynamic therapy (PDT) has shown its therapeutic effectiveness in the treatment of predominantly classic choroidal neovascularization (CNV) in AMD, CNV due to pathologic myopia, and a subgroup of purely occult neovascular lesions in AMD. ^{1 2 3 4}  

The current concept of PDT is that it produces selective nonthermal photothrombosis. The intravenously applied photosensitizer (verteporfin) forms intravascular complexes with low-density lipoproteins (LDLs). Aggregates are bound by LDL-receptors expressed by the proliferative endothelium of CNV. ^{5 6 7} Growing pathologic vessels such as CNV express an up to 10-fold increase in LDL receptors. ^{8 9} The preferential binding of verteporfin to these receptors, its local photoactivation by a low-intensity laser beam (600 mW/cm²) at 692 nm, and the higher sensitivity of proliferating endothelial cells for any toxic stimulus should lead to a selective closure of CNV. The overlying neurosensory retina is preserved. Choriocapillary hypoperfusion usually recovers within 3 months after PDT. ^{10 11 12} Thrombosis occurs as a result of spatially confined damage to vascular endothelial cells ^{5 13} after the immediate production of reactive oxygen species with an extremely short diffusion distance of 0.1 μm. ¹⁴  

The first experimental targets used to validate the effectiveness of selective photodynamic vaso-occlusion were neovascularization induced in the cornea and the choroidal vascular layer in rabbits. ^{14 15 16} Angiography performed as early as 1 hour after photosensitization with light and verteporfin demonstrated complete and reproducible occlusion of corneal neovascularization and the choriocapillaris. A similar immediate effect on perfusion was seen 1 hour after PDT in highly vascularized experimental melanomas implanted into the suprachoroidal space of the rabbit eye. ^{17 18} Similarly, the typical fluorescein angiographic appearance in a monkey model with laser-induced CNV 24 hours after PDT, revealed hypofluorescence within the treatment area without any apparent filling of neovascular tissue. ^{19 20 21} Based on these experimental observations, the concept of PDT was defined as immediate thrombosis and cessation of perfusion, particularly in neovascularization, which was preferentially targeted. The hypothesis of an immediate occlusive effect was further strengthened by angiographic results of phase I and II clinical trials that demonstrated homogeneous hypofluorescence throughout the treatment site at the first posttreatment fluorescein angiography (FA) evaluation scheduled at 1 week after PDT. Using appropriate parameters, 100% of treated lesions were nonperfused at 1 week, the surrounding choroid showed some degree of collateral hypofluorescence, and leakage from classic CNV was mostly absent. ²²  

However, the characteristic subjective symptoms and clinical findings in patients within the first days after PDT include an increase in metamorphopsia and increased intra- and subretinal fluid within the treated area, as seen by ophthalmoscope. These observations do not correlate with the hypothesis of immediate CNV and choriocapillary closure and sudden resolution of leakage. Furthermore, the substantial hypofluorescence observed by indocyanine green angiography (ICGA) 1 week after PDT, ^{10 11 22} which is consistent with the substantial occlusion of the choriocapillary layer observed histologically, would suggest a more pronounced decrease in vision shortly after PDT. These experimental and clinical observations remain controversial due to lack of information about the immediate photodynamic effects within treated human eyes. 

To elucidate vascular effects induced by PDT in human eyes, we designed a prospective angiographic study. The purpose of the trial was to identify the nature and sequence of events in terms of occlusion of CNV and choroid as well as enhancement and resolution of exudation. A confocal scanning laser ophthalmoscope (SLO) system was used for FA and ICGA imaging, which was performed at specified intervals according to a standardized protocol. Vascular changes were monitored for the neovascular lesions and for the adjacent normal choroid. 

Methods

The angiographic analysis was designed as a prospective, noncomparative, interventional trial. Indications for PDT were based on the specific approval and recommendations of the FDA and European authorities. The study protocol complied with the Declaration of Helsinki. A separate approval from the institutional ethics committee was obtained. All participants were informed in detail about the potential risks and complications of additional angiograms, particularly those taken 5 hours and 1 day after treatment. Patients agreed to the conditions of the additional examination and signed a written informed consent form. 

Photodynamic Therapy

Forty patients consecutively attending a tertiary referral center were included in the study. All eyes underwent best refracted visual acuity evaluation according to the Early-Treatment Diabetic Retinopathy Study (ETDRS) standards and demonstrated a visual acuity of 20/200 or better. The indication for treatment was based on the presence of CNV with involvement of the foveal avascular zone (FAZ) and a predominantly classic component based on the criteria defined by the Macular Photocoagulation Study group. ²¹ Classic CNV had to occupy at least 50% of the area of the entire neovascular lesion. The greatest linear dimension could not exceed 5400 μm. Patients who had received previous treatment of AMD such as photocoagulation, PDT, radiation or any other type of therapy were excluded, as were patients with any ocular disease other than CNV. Only one eye of each patient was included in the trial. 

Standard treatment parameters for PDT with verteporfin were used. Verteporfin was infused over 10 minutes at a dose of 6 mg/m² body surface area. Five minutes after the end of the infusion, a diode laser emitting at 689 nm with a slit lamp delivery system (Visulas; Zeiss Jena GmbH, Jena, Germany) was used to deliver 50 J/cm² at an intensity of 600 mW/cm² over 83 seconds. 

The greatest linear diameter (GLD) of the lesion was measured with a standardized transparent overlay adapted to the magnification factor of the imaging system. An additional 1000 μm was added to the GLD to provide a 500-μm circular security margin. The laser light was applied by a contact lens (Mainster standard, Mainster wide field; Ocular Instruments, Inc., Bellevue, WA). 

FA and ICGA Imaging

Confocal FA and ICGA were performed within 1 week before treatment as well as 5 hours, 1 day, 1 week, and 12 weeks after treatment. A scanning laser ophthalmoscope (Heidelberg Retina Angiograph [HRA], Heidelberg Engineering, Dossenheim, Germany) was used. Radiant exposures for 480 nm for excitation of fluorescein and 795 nm for excitation of ICG incident at the ocular surface were measured for the duration of a regular angiographic examination. Exposures were restricted to 30-second intervals for the early sequence taken at 1 minute for FA and ICGA and the late sequence taken at 10 minutes for FA or 20 minutes for ICGA. Irradiance and total light dose for each wavelength were calculated and compared with the light dose necessary to induce photochemical effects, based on the absorption spectrum of verteporfin and the decay of the plasma level of the sensitizer at 5 and 24 hours after administration. The calculated light exposure was below the light dose necessary for phototoxicity with the a factor of 10 ⁻² and 10⁻³. 

Patients received a bolus of 5 mL of a 10% fluorescein solution for FA and a bolus of 50 mg ICG for subsequent ICGA in a 5-mL solution (ICG-Pulsion; Medical Systems, Munich, Germany). The size of the scanning field was set at 30° × 30°. Single images were taken in rapid sequence during the early (1 minute) and late (10 or 20 minutes) phases. Images were digitized in frames of 512 × 512 pixels. 

Image Analysis

Evaluation parameters included features of the neovascular lesion and adjacent choroid. The CNV size was measured separately as it appeared early in FA and early in ICGA, according to the presence of a neovascular pattern. The area of leakage was determined as the area showing hyperfluorescence originating from the previous CNV site, but absent during early phases, separately for late-phase FA and ICGA. Changes in the adjacent choroid were documented as the size of the hypofluorescence area measured during the early- and late-phase ICGA. Areas were documented in square millimeters by manual perimetry (HRA software version 1.10, package R1-V1.08, 1998; Heidelberg Engineering). To avoid a potential bias, patients’ names were replaced by numbers, and images were evaluated in random chronological order. Each image was read by two independent observers. Individual readings for each aspect (CNV size, leakage, and hypofluorescent area) were averaged. Analysis of deviations indicated an interindividual variability below 5% (SD). 

For statistical evaluation, the Wilcoxon signed ranks test was used. P ≤ 0.05 was considered significant. In the graphic presentation of the results, the standard error of the mean was calculated for each data point. 

Results

CNV Size in Early FA and Early ICGA

All patients at baseline had a classic component larger than 50% of the total neovascular lesion size, as defined by early-phase FA. The neovascular net was detected by FA and ICGA in every eye before treatment with a distinct delineation. Lesions fluoresced more intensely on FA (Fig. 1A) than on ICGA (Fig. 1B) . The mean total CNV size was 3.12 mm² in FA and 2.51 mm² in ICGA. Five hours after treatment, major parts of the CNV were still perfused, which was more obvious in FA images with a reduction to about half of the initial size (1.41 mm²; 45.2%; P = 0.009), whereas ICGA showed less reduction in CNV size (2.16 mm²; 86%). Persistent portions of CNV were present in all FA images at 5 hours after PDT, although lesions appeared smaller (Fig. 1C) . CNV was seen in all ICGA images at 5 hours, with more distinct margins (Fig. 1D) . At 1 day, lesions were usually inapparent with both modalities (Figs. 1E 1F) . The average CNV size at 1 day was 0.49 mm² (15.7%, P < 0.001) in FA and 0.78 mm² (31.1%, P = 0.001) in ICGA. Further follow-up showed a continuous growth in CNV size. The increase was clinically minimal at 1 week, with a mean CNV size of 0.95 mm² (30.4%, P < 0.001) in FA and 1.04 mm² (41.4%, P < 0.001) in ICGA. At 3 months, however, the CNV size was consistently larger than at baseline in FA (4.00 mm², 128.2%) and ICGA (3.35 mm², 133.5%; Fig. 2 ). 

Leakage Area in Late FA and Late ICGA

Hyperfluorescence in late-phase FA and ICGA at baseline was considered equivalent to leakage from the neovascular components of the lesion. Late-phase exudation originating from the classic component and, to a lesser extent, from occult portions was most pronounced in FA imaging (Fig. 3A) , but increased late-phase hyperfluorescence suggestive of extravascular ICG accumulation was also noted in ICGA (Fig. 3B) . The mean area of late hyperfluorescence before PDT was 6.76 mm² in FA and 3.30 mm² in ICGA. Five hours after treatment, the hyperfluorescent area increased, especially in the area beyond the remaining CNV net as defined by early-phase angiography. The area showing late leakage measured 8.09 mm² in FA (119.7%, P = 0.153) and 8.76 mm² in ICGA (265.5%, P = 0.001) on average. Extensive leakage originating from the entire area covered by the treatment spot also started 5 hours after PDT and reached its maximum at day 1 on FA (Fig. 3C) and, strikingly, also on ICGA (Fig. 3D) . The area demonstrating leakage at day 1 was similar in both modalities and delineated the size of the laser spot. The mean leakage area increased to 10.02 mm² in FA (148.2%, P = 0.354) and 12.49 mm² (378.5%, P < 0.001) in ICGA. FA subsequently revealed resolution of leakage with some residual activity in FA images (Fig. 3E) and no significant hyperfluorescence in ICGA (Fig. 3F) . At 1 week the mean hyperfluorescent area decreased to baseline in FA (6.97 mm², 103.1%, P = 0.89) and significantly below baseline in ICGA (1.92 mm², 58.2%, P = 0.002). No further progression, compared with baseline, was found in FA at 12 weeks (6.55 mm², 96.9%). In contrast, the area of late hyperfluorescence in ICGA showed a moderate increase compared with the baseline area (4.15 mm², 126%; Fig. 4 ). 

Collateral Hypofluorescence in Early and Late ICGA

Before PDT, a hypofluorescent area was typically present surrounding the neovascular membrane in 82% of patients in the early phase and in 79% of patients in the late phase of ICGA. Pretreatment hypofluorescence was irregular in size and more pronounced in early ICGA (Fig. 5A) compared with late ICGA (Fig. 5B) . The mean area of hypofluorescence was 4.49 and 3.86 mm², respectively. Posttreatment hypofluorescence was minimally enhanced in intensity and size at 5 hours (early: 5.75 mm², 128.1%; late: 5.23 mm², 135.5%; Figs. 5C 5D ). At 1 day after PDT, the increase in the hypofluorescent area became significant in early-phase ICGA (P = 0.004). Mean area was 8.89 mm² (198.0%) in early- and 6.09 mm² (157.8%) in late-phase angiography. Intensive and extended choroidal hypofluorescence was obvious at 1 week, with nonperfusion of small- and medium-caliber vessels (Fig. 5E) . The hypofluorescent defects did not fill with ICG in late-phase angiograms, except for some areas with focal leakage (Fig. 5F) . The increase in hypofluorescent area compared with baseline was highly significant (P < 0.001) in the early (11.04 mm², 245.9%) and in the late (11.16 mm², 289.1%) phases of ICGA. The diameter of the hypofluorescent spot at 1 week correlated with the diameter of the treatment beam. Partial recovery was seen at 3 months. Hypofluorescent areas were 7.99 (180.0%) and 6.75 mm² (174.9%), respectively (Fig. 6) . 

Discussion

PDT performed with verteporfin has recently entered the spectrum of therapeutic strategies in neovascular AMD. Several randomized, placebo-controlled trials have shown the functional benefit, including significantly better maintenance of visual acuity within 3 lines of the baseline value and significantly greater avoidance of severe visual loss more than 6 lines, with maintenance of contrast sensitivity and integrity of the central visual field, in patients with classic subfoveal CNV resulting from AMD, who received PDT with verteporfin by comparison with placebo-treated patients. ^{1 2 3 4 23} Although PDT is now widely used, many important issues regarding the mechanisms of PDT-induced vascular effects are still unresolved. Controversies are related to the choice of optimal treatment parameters, appropriate retreatment intervals and the number of retreatments that are appropriate. The differential effects on classic neovascular components compared with occult lesions has raised further discussions on the type of changes that occur within CNV and normal choroidal vasculature. A combination therapy using antiangiogenic compounds such as anecortave acetate and VEGF inhibitors (Slakter JS, Singerman LJ, Yannuzzi LA, et al., ARVO Abstract 2909, 2002; Singerman LJ, ARVO Abstract 2908, 2002) would be another option to improve outcomes; however, these substances demonstrate a characteristic spectrum of effects that may enhance or even counteract photodynamic changes. 

By contrast with earlier studies in experimental animal models, ^{14 15 16 17 18 19 20 21} our present results in human subjects with neovascular AMD show that PDT does not induce an immediate thrombosis of the entire CNV complex. At 5 hours after treatment, the main neovascular structures were still perfused, whereas more peripheral parts of the neovascular complex were already closed. Within 1 day, a complete closure of the membrane occurred in most patients. One week after PDT, new vessels already showed regrowth, which continued at least until 3 months. 

Hence, thrombosis occurs over a prolonged period with progressive closure from the periphery to the center. Confocal ICGA imaging has shown that this process is incomplete in about half of the treated eyes, with persistence of the central portion of the CNV. ¹⁰ However, closure of the neovascular complex proceeds much faster than closure of adjacent, normal choroidal vessels. The photodynamic selectivity appears to be time dependent with faster endothelial damage occurring within neovascular channels than within physiological vascular walls early after PDT, but with occlusion of normal choroidal vessels appearing fully developed as late as 7 days after treatment. 

The second unexpected finding consists of a vascular barrier dysfunction directly induced by the PDT treatment. A similar effect has been described early after photocoagulation of CNV in AMD and also after transpupillary thermotherapy of CNV. ^{24 25} This finding raises serious doubts regarding a differential or selective effect unique to PDT. Many patients report compromised vision and an increase in metamorphopsia within the first few days after PDT. Early angiography after PDT demonstrated an increase of hyperfluorescence in the late phase of both FA and ICGA. Hyperfluorescence typically increased in intensity and size in both procedures. At 5 hours after treatment an increase in hyperfluorescence originated from the remaining CNV. A dramatic change in exudation occurred 1 day after PDT, when hyperfluorescence spread to the entire treated area. Massive leakage originating from the surrounding, previously unaffected choroid was seen with both fluorescein and ICG. 

There are two major hypothesis: The first is that there is an increase of leakage due to damage to the vascular endothelial cells and pericytes, first of the CNV and subsequently of the surrounding normal choriocapillary network. Fluorescein and ICG leakage areas and the treatment spot are identical in size, highlighting the fact that choroidal leakage is most likely directly induced by photochemical reactions within the entire light-exposed field. Progressive closure of the choriocapillaris leads to the reduced leakage of ICG from the choroid during long-term follow-up. 

The second explanation for accumulation of subretinal fluid is functional damage to the retinal pigment epithelium (RPE) by the PDT treatment. The activity of the RPE-dehydrating pump could be reduced, allowing accumulation of fluid that would normally be transported outward to the choroid. Both effects, most likely in combination, can lead to an increase of subretinal fluid appearing as intensive hyperfluorescence angiographically. 

A noteworthy phenomenon is the prolonged development of choroidal thrombosis. Comparisons between 1-week angiograms with FA and ICGA clearly highlight the difference in the detection of choroidal perfusion changes. Although no relevant hypofluorescence was present in the FA images, ICGA delineated a distinct area with intensive and homogeneous choroidal hypofluorescence (Figs. 5E 5F) . The use of FA alone may be responsible for an underestimation of damage to the surrounding, normal choroids in the clinical trials that exclusively used FA imaging. 

PDT-induced hypoperfusion and partial occlusion of the choriocapillary layer has been shown histologically in human eyes. ¹¹ Choroidal hypoperfusion can be clearly demonstrated by ICGA. ²⁶ Hypoperfusion of the choroid is a characteristic feature in AMD. ²⁷ However, PDT-induced hypofluorescence differed from the irregular perfusion changes seen in AMD, with distinct borders and marked hypofluorescence in late-phase ICGA. Perfusion of the choroid decreased slowly after PDT and was reduced to its minimum at 1 week. These relatively slow changes in perfusion of the physiologic choroid may explain why there is typically no substantial loss of vision. A marked recovery of the choroidal vasculature was seen in ICGA at 3 months, as demonstrated by a decrease in size and intensity of the hypofluorescence. Even if transient, choroidal hypoperfusion with its resultant tissue hypoxia may represent an angiogenic stimulus responsible for recanalization and progression of CNV and the need for repeated retreatments. Further studies using techniques such as three-dimensional angiography and optical coherence tomography are necessary to further elucidate the effect of PDT with verteporfin on the CNV and choroid in human eyes. 

PDT currently represents the only modality that achieves a reproducible inactivation of subfoveal CNV, together with improved preservation of visual acuity. It offers an important benefit in treatment outcome in comparison to photocoagulation. ²⁸ However, optimization of the treatment is warranted with respect to improved visual outcome and reduction in the number of treatments. Novel strategies such as antiangiogenic intervention, currently undergoing clinical trials, may be combined with PDT to achieve additive effects. Whether antiangiogenesis in combination with PDT is more effective than the use of antiangiogenesis alone remains to be determined. A complete understanding of PDT-induced mechanisms is necessary to design promising strategies for such potential combination approaches. 

 Submitted for publication June 18, 2002; revised November 6 and November 14, 2002; accepted November 23, 2002.

 Disclosure: S. Michels, None; U. Schmidt-Erfurth (P)

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

(A) FA before PDT. A classic CNV membrane extended under the geometric center of the FAZ. Subretinal hemorrhage was seen at the nasal portion. (B) ICGA before PDT. The neovascular net was well demarcated. (C) FA 5 hours after PDT shows that the central portion of the CNV complex was still perfused. (D) ICGA 5 hours after PDT. The central portion of the lesion was delineated by angiography. (E) FA 1 day after PDT. The neovascular channels were not detectable by early-phase FA. (F) ICGA 1 day after PDT. At this presentation, CNV features were absent during early frames.

Figure 2.

 

CNV size (in square millimeters) by early FA and ICGA. The neovascular complex was reduced in size but was still perfused at 5 hours after treatment. At 1 day after PDT, the CNV was angiographically absent in most eyes. The neovascular net showed recovery, however, especially in FA at the 1-week interval followed by further growth at 3 months. Data are expressed as the mean ± SE.

Figure 3.

 

(A) FA before PDT. The predominantly classic component of the lesion exhibited intensive leakage, with an occult portion visible at the inferior portion. (B) ICGA before PDT. Late hyperfluorescence delineated the entire lesion area with classic and occult portions. (C) FA 1 day after PDT. Extensive leakage covered the site of photosensitization, regardless of the previous site of classic CNV. (D) ICGA 1 day after PDT. Late hyperfluorescence within the area of light exposure was intensive and prominent, consistent with ICG leakage. (E) FA 1 week after PDT. Extravasation had largely resolved, with residual focal leaks remaining active, no collateral hypofluorescence was detectable. (F) ICGA 1 week after PDT. Leakage was replaced by homogeneous choroidal hypofluorescence, indicative of nonperfusion.

Figure 4.

 

Leakage area (in square millimeters) in late FA and ICGA. Exudation of fluid from the neovascular net was increased in FA 1 day after PDT and then declined. ICGA exhibited an immediate and intense increase in leakage as early as 5 hours after PDT, reaching a maximum at 1 day after treatment. Leakage was minimal at 1 week and was progressively increased during the 3-month follow-up. Data are expressed as the mean ± SE.

Figure 5.

 

(A) Early ICGA before PDT. The neovascular net contrasted well with diffuse background hypofluorescence. (B) Late ICGA before PDT. Hyperfluorescence indicated the site of the CNV lesion and surrounding hypofluorescence was minimal. (C) Early ICGA 1 day after PDT. Collateral hypofluorescence was poorly demarcated, although discretely enhanced choroidal vessels were still visible. (D) Late ICGA 1 day after PDT. Collateral hypofluorescence was present only in the center of the lesion. (E) Early ICGA 1 week after PDT. Intensive hypofluorescence was characterized by absence of choroidal perfusion. (F) Late ICGA 1 week after PDT. Collateral hypofluorescence was markedly increased in size and intensity, had well-demarcated borders, and persisted throughout the entire angiographic sequence.

Figure 6.

 

Hypofluorescent area (in square millimeters) in early and late ICGA. ICGA showed in the early and late phases a continuous increase in area of hypofluorescence up to 1 week after PDT with partial recovery at the 3-month follow-up. Data are expressed as the mean ± SE.