External Beam Irradiation of Subfoveal Choroidal Neovascularization Complicating Age-Related Macular DegenerationOne-Year Results of a Prospective, Double-Masked, Randomized Clinical Trial

Objectives  To determine the effects of low-dose external beam irradiation compared with observation on the visual function of eyes with subfoveal choroidal neovascularization(CNV) complicating age-related macular degeneration (ARMD).

Design  Prospective, double-masked, randomized clinical trial. Patients randomized to the radiation group received external beam irradiation at a dose of 14 Gy in 7 fractions of 2 Gy. Patients randomized to the observation group received sham radiation.

Setting  Tertiary care retinal referral practice.

Patients  Individuals with classic, mixed, or occult subfoveal CNV secondary to ARMD.

Main Outcome Measures  Change in visual acuity from baseline to specified time periods. Secondary outcome variables were contrast sensitivity and fundus photographic/fluorescein angiographic progression.

Results  Forty-two eyes were randomized to observation; 41 eyes, to radiation. Baseline characteristics and demographics did not differ between groups. The median distance visual acuity (DVA) in radiation-treated eyes decreased from 20/80 at baseline to 20/320 (mean loss rate, 4.14 lines) at 1-year follow-up. The median DVA in observation group eyes decreased from 20/125 at baseline to 20/250 (mean loss rate, 3.39 lines) at 1-year follow-up. There were no statistically significant differences in changes in DVA, contrast sensitivity, or fluorescein angiographic progression from baseline between groups at any follow-up period.

Conclusions  At 1-year follow-up, low-dose external beam irradiation at 14 Gy in 7 fractions of 2 Gy is neither beneficial nor harmful for subfoveal CNV complicating ARMD.

LASER photocoagulation is a proven therapy for choroidal neovascularization(CNV) complicating age-related macular degeneration (ARMD).^1-7 The Macular Photocoagulation Study (MPS) has reported findings from randomized trials of laser treatment for subfoveal neovascular lesions secondary to ARMD.^4-6 Despite the long-term benefit of laser treatment of subfoveal lesions within the context of the MPS, the management of subfoveal CNV in ARMD remains controversial, as laser treatment results in immediate visual loss after foveal ablation.⁸ In addition, the treatment of CNV in ARMD is based on the accurate identification of CNV using fluorescein angiography. Most patients with exudative ARMD demonstrate occult subfoveal CNV, which does not have clinical and fluorescein angiographic features that meet the eligibility criteria of the MPS for laser photocoagulation.^9,10 Alternative, less destructive therapies that do not require angiographic identification of the exact boundaries of subfoveal CNV are therefore essential.

Recently, photodynamic therapy with verteporfin has been demonstrated to safely reduce the risk for visual loss in ARMD patients with predominantly classic, subfoveal CNV.¹¹ However, most verteporfin-treated patients demonstrated visual acuity (VA) loss at 1 year follow-up.¹¹ In addition, a recent pharmaceutical company press release noted that photodynamic therapy with verteporfin had no significant harmful or beneficial effect in ARMD patients with occult subfoveal CNV.¹² Although the results of photodynamic therapy with verteporfin are significant and encouraging, additional therapies are necessary to find a benefit for occult CNV and to improve on the photodynamic therapy outcomes for predominantly classic CNV.

Radiotherapy has been proposed as one such experimental treatment for subfoveal CNV complicating ARMD. Experimental in vitro^13-19 and in vivo^20-26 studies demonstrate that the retinal endothelium and choriocapillaris are radiosensitive. Clinical experience demonstrates that radiation is antiangiogenic and is toxic to vascular tumors and endothelial cells.^27-39 These properties of radiation have led to numerous nonrandomized studies evaluating its effect on exudative ARMD. Most preliminary studies suffer from an absence of adequate control groups or randomization, from short follow-up, and from a lack of strict clinical and fluorescein angiographic entry criteria and measurement.^40-51 Although sound scientific rationale exists for offering radiotherapy in a disease for which there is often inadequate or no treatment, there is no definitive proof of a beneficial or harmful radiotherapy effect. We therefore performed a prospective, double-masked, randomized, clinical trial to evaluate whether low-dose external beam irradiation can decrease the risk for VA loss in ARMD patients with subfoveal CNV.

Patients were recruited from one clinical center at the Department of Ophthalmology, Medical College of Georgia, Augusta. Patients were enrolled from February 3, 1995, through September 2, 1998. Patients with active subfoveal CNV secondary to ARMD were eligible for inclusion in the study. Patients included were older than 48 years and had not undergone previous laser therapy in the study eye. Patients were excluded if there was a history of ocular disease associated with CNV due to causes other than ARMD. To avoid a decreased threshold for radiation complications, patients with diabetes or other forms of retinal vascular disease and patients who received or were likely candidates for chemotherapeutic agents were excluded. Patients who previously received ocular, orbital, or periorbital radiation were excluded. Only one eye for any given patient was eligible. Baseline VA of the study eye was required to be no worse than 20/400. All patients were required to have clinical and fluorescein angiographic evidence of active classic, occult, or mixed CNV for which the CNV itself, or contiguous blood, was under the center of the foveal avascular zone. Patients with CNV eligible for subfoveal laser therapy according to MPS guidelines^4-6 were offered laser therapy, randomization to laser vs radiation therapy (data not reported because of low patient recruitment), or observation vs radiation therapy (present study). Patients with subfoveal CNV ineligible for laser therapy according to MPS guidelines^4-6 were offered randomization to observation vs radiation therapy (present study). Patients with classic or predominantly classic CNV were eligible for this study because photodynamic therapy with verteporfin was not proven or available at that time.

All patients underwent a general ophthalmologic and retinal examination, including tonometry, indirect ophthalmoscopy, and slitlamp biomicroscopy with a contact 90- or 78-diopter lens. The following was measured at each study baseline and follow-up visit: Best-corrected distance VA (DVA) was measured at 10 feet on a backlit Early Treatment Diabetic Retinopathy Study chart. The smallest line read with no less than 2 mistakes was recorded as the VA. If VA was poor, the chart was brought closer to the patient in 1-foot increments until the patient could read the largest line. Examination was performed to determine the presence of radiation-induced adverse effects such as eyelid erythema or dermatitis, madarosis, superficial punctate keratitis or dry eye, cataract progression, retinal vasculopathy, iris or retinal neovascularization, or optic disc edema or atrophy. Phakic eyes underwent assessment in the study and nonstudy eyes for extent of nuclear sclerosis, cortical cataract, and posterior cataract using a subjective 0 to 4 grading system. Phakic eyes underwent Scheimpflug slit imaging of the lens^52,53 in both eyes at preentry and at 6-month intervals to assess objectively for cataract development or progression (data to be analyzed at 4-year follow-up). Contrast threshold for large letters was measured at a distance of 1 m using the Pelli-Robson chart.⁵⁴ The contrast threshold was scored as the level of contrast required to read at least 2 of the 3 letters per contrast level presented on the eye chart. Color stereoscopic photography and fluorescein angiography using a 30° camera (FF4 Fundus Camera; Zeiss/Humphrey, Oberkochen, Germany) were performed.

A medical history was ascertained and data for possible contributing factors to the outcome variables were collected, including age, sex, hypertension or receiving an antihypertensive (yes or no), smoking status (no, quit, or currently), aspirin or warfarin sodium (Coumadin) intake (yes or no), and vitamin intake (yes or no). Baseline characteristics of CNV (classic, occult, or mixed), baseline eligibility for laser according to MPS guidelines (yes or no), and baseline size of CNV (MPS disc area) were graded by the Scheie Eye Institute Photograph Reading Center, Philadelphia, Pa, and were analyzed as possible contributing factors to the outcome variables.

Follow-up evaluations were performed at 3, 6, 12, and 24 weeks after enrollment and at 6-month intervals thereafter for a total of 4 years. The present study reports findings up to the 1-year follow-up. The patient, examining ophthalmologist, and ophthalmic technician were unaware of the assignment to observation or radiation treatment groups.

Eligible patients underwent an extensive explanation of randomization and the experimental nature and possible complications of radiation treatment. Medical College of Georgia institutional review board approval for both the study and informed consent was obtained. Patients who were believed to satisfy all eligibility criteria, including signed, written informed consent, were assigned randomly at enrollment to receive radiation or to observation (sham radiation). The randomization incorporated blocking, which is recommended any time patient recruitment extends for a long period of time. Blocks of size 2 or 4 were assigned randomly, and a separate random permutation was used to assign the 2 treatments to the blocks. A randomization schedule was printed and sent to the radiology team, who then sequentially allocated the patients to the sham or actual radiation treatments. Radiation or sham treatments were performed as soon as possible after enrollment, but no later than 2 weeks after baseline examination and fluorescein angiography.

Patients randomized to the radiation group underwent conventional simulation followed by computed tomographic (CT) localization or by means of a CT simulator in preparation for use of a small treatment port. Patients underwent simulation in the supine position using a headrest and base plate (WFR/Aquaplast Corporation, Wyckoff, NJ). After position check, reference marks were applied to the mask, and an x-ray film was obtained with the simulator centered over the marker of the treatment eye at a 15° angle anterior-posterior to the coronal plane. Final reference marks were applied to the mask, followed by CT localization in the simulation position using the mask and markers. Final beam position was planned from CT images to ensure adequate coverage of the macula. The radiation group was treated with a 6-MV photon beam using a machine field setting of 3 cm wide by 4 cm long and a custom-fabricated treatment collimating device with a semicircular shape of 1.5 cm wide by 3.0 cm long (1.5-cm radius; 3.0-cm diameter). Seven fractions of 2 Gy each (total, 14 Gy) were administered during 7 consecutive business days in the radiation group. A dose of up to 14 Gy at 2 Gy per fraction was prescribed to the 90% isodose line.

The observation (sham) group did not undergo CT simulation. Patients randomized to the observation group underwent 1 sham session with the radiation oncologist (W.C.S.), who was not masked to group assignment. Sham sessions were administered with the mask. Both groups adhered to identical follow-up schedules to maintain the double-masked design. During review of the Medical College of Georgia institutional review board–approved informed consent, patients were informed that they would remain masked as to the group they were randomized to and that they may receive sham or fake radiation. Patients were informed that they would receive at most 7 treatments, but that they may receive fewer sessions. Patients were informed that they or their insurance carrier would not be charged for any of the costs associated with radiotherapy or treatment planning.

A special linear accelerator collimator was designed for this study to minimize the dose to the lens and other out-of-beam structures. It consists of a mounting assembly that attaches to the lowest part of the accelerator head and a lead alloy plug that fits snugly into the tube extending from the center of the mount (Figure 1 and Figure 2). The extended collimator assembly, similar to that used for stereotactic radiosurgery, is external to the accelerator head and provides additional protection because of the 10-cm added shielding thickness. Because it is closer to the patient, the beam edge defined by the external collimator has less penumbra and therefore a sharper dose falloff at the beam edge.

Two methods of dose measurement were used to determine doses to the macula, lens, optic nerve, and retina of both eyes for a typical setup. A small-volume ionization chamber was positioned in a block of polystyrene, which is equivalent to tissue in regard to radiation dose distribution. A typical mock-up of a patient treatment was used to determine doses at the position of each relevant structure. The same measurement was performed using thermoluminescent dosimeters in a polystyrene phantom. Finally, measurements of lens dose made using the study setup were compared with doses measured for a similar beam size (1.5 × 3.0 cm) using the linear accelerator without external collimators.

The primary outcome variable measured at each examination was DVA. Secondary outcome variables were contrast sensitivity (CS) and angiographic/photographic appearance. Distance visual acuity was coded as integers so that the difference between each line had a value of 1, and a difference of 3 integers (3 lines) corresponded to a doubling of the visual angle. Distance visual acuity at each time point was analyzed for group differences using the Wilcoxon rank sum test. Kaplan-Meier curves were used to test the difference between the treatment groups for the rate of decrease in DVA across the entire study time frame. Changes in DVA from baseline were analyzed using t tests if there were 2 groups undergoing testing, or analysis of variance(ANOVA) if there were more than 2 groups or when testing possible contributing factors.

Contrast sensitivity values recorded from the Pelli-Robson chart are the base-10 logarithm of CS, which has the advantage that equal steps on this scale correspond to equal effects. Treatment differences in logarithm of CS were analyzed using t tests. Percentage of contrast is reported, which is the reciprocal of CS.

Fluorescein angiograms and color photographs were reviewed and graded in a masked manner with respect to randomization group. Baseline fluorescein characteristics, as outlined above, were graded for type, laser eligibility, and size by the reading center in a masked fashion. Grading of follow-up angiograms and all color photographs were performed by one of us (D.M.M.) in a masked fashion. Fluorescein angiograms and color fundus photographs were graded as to CNV size (≤1, >1 to ≤ 2, >2 to ≤3.5, >3.5 to ≤4, >4 to ≤6, >6 to ≤9, and >9 MPS disc areas); CNV size plus blood, elevated blocked fluorescence, or serous pigment epithelial detachments (CNV size + all) (≤1, >1 to ≤2, >2 to ≤3.5, >3.5 to ≤4, >4 to ≤6, >6 to ≤9, and >9 MPS disc areas); classic CNV size (none, ≤1, >1 to ≤2, >2 to ≤3.5, >3.5 to ≤4, >4 to ≤6, >6 to ≤ 9, and >9 MPS disc areas); hemorrhage(none, ≤25% of clinical macula, and >25% of clinical macula, arcade to arcade); and subretinal fibrosis (none, ≤25% of clinical macula, and >25% of clinical macula, arcade to arcade). For grading of follow-up angiograms and photographs, when there was an increase in size of the membrane and/or increase in hemorrhage or subretinal fluid from baseline, the membrane was graded as worsened. When there was no change in the size of the membrane without increased hemorrhage or subretinal fluid, the membrane was graded as stable. In occult lesions obscured by blood, if the extent of the membrane was visualized further after clearing of blood, the membrane was graded as stable. When there was a decrease in the degree of leakage or membrane size with improvement in hemorrhage or subretinal fluid, the membrane was graded as improved.

We used χ² testing when analyzing categorical data unless the cells sizes were too small; if so, Fisher exact tests were used.⁵⁵

Randomized, clinical trials investigating treatments of CNV with a classic component indicate that approximately 27%¹¹ to 30%⁴ of observed eyes exhibit 6 lines or more of visual loss at 1 year. If a 30% 6-line loss rate was reduced to 10% in the radiation treatment group, then a sample size of 50 in each group would be needed to have 80% power to detect this 20% difference when α = .05. Randomized, clinical trials investigating treatments for occult CNV without a classic component indicate that approximately 55% of observed eyes exhibit 3 lines or more of visual loss at 1 year.¹² If a 55% 3-line loss rate was reduced to 30% in the radiation group, then a sample size of 50 in each group would be needed to have 80% power to detect this 20% difference when α = .05. Data were reviewed semiannually by two of us (D.M.M. and M.H.J.) for evidence of treatment benefit or harm. In November 1998, a decision was made to stop enrollment into the trial and continue follow-up, since the group differences at 1 year were very small and were not approaching 20%.

Results of ionization chamber and thermoluminescent dosimeter measurement agreed within 4% of the prescribed dose for all points of measurement within the mock patient setup. Table 1summarizes the actual doses received at each point by ionization chamber and thermoluminescent dosimeters along with the toxic dose for the structure. The beam path is shown as a 2-dimensional isodose distribution in Figure 3. Exact points of measurement are annotated on the CT scan of Figure 4along with ion chamber doses expressed as a percentage of the prescribed dose to the ipsilateral macula.

Dose measurements using the external collimator constructed for the study were compared with the dose without added external collimation (1.5 × 3.0 cm). The doses to ipsilateral and contralateral lens were reduced to 50% of the original dose using the collimator.

A total of 83 eyes were assigned randomly to radiation treatment or to observation (sham radiation). Forty-one eyes were randomized to the radiation group, and 42 were assigned to the observation group. The number of eyes followed up for specific periods is outlined in Table 2. Of all 83 randomized eyes, 70 (84%) were examined at 1 year after enrollment.

The distributions of baseline factors are outlined in Table 3 and were examined for differences between the radiation and observation groups. Variables were similarly distributed between groups. However, patients randomized to the radiation group were more likely to have hypertension and/or treatment with antihypertensive medication.

The VA distributions for patients examined 3 weeks, 6 weeks, 12 weeks, 6 months, and 12 months after randomization are outlined in Table 4. Although we measured DVA at 10 feet, we report the results using the more readily recognized and meaningful VA score with a numerator as 20 feet (ie, a measurement of 10/40 is reported as 20/80). The median DVA in radiation-treated eyes decreased from 20/80 at baseline to 20/320 at 1-year follow-up. The median DVA in observation group eyes decreased from 20/125 at baseline to 20/250 at 1-year follow-up. There were no statistically significant differences for median VA between groups at any follow-up period.

The distribution of change in DVA from the initial visit is outlined in Table 5 for patients examined 3 weeks, 6 weeks, 12 weeks, 6 months, and 12 months after randomization. There were no statistically significant differences in changes in DVA from baseline between groups at any follow-up period. At 1-year follow-up, eyes in the radiation and observation groups lost a mean of 4.14 and 3.39 lines, respectively, of DVA. Possible contributing factors at baseline such as age, sex, VA, status of contralateral nonstudy eye (exudative or nonexudative), smoking status, anticoagulant intake, hypertension history, and vitamin intake were analyzed for effect on VA outcomes and were found not to affect VA changes at 1 year(data not shown).

The percentage of eyes in the radiation and observation groups with a decrease in VA of 3 lines or more (moderate visual loss) and 6 lines or more (severe visual loss) is illustrated in Figure 5. There were no statistically significant differences in the development of moderate or severe visual loss between groups at any follow-up period.

The contrast threshold was compared between groups (Table 6). There were no statistically significant differences in the development of severe contrast loss between groups at any follow-up period.

Fluorescein angiography was performed at all visits for 26 (62%) of 42 patients in the observation group and for 33 (80%) of 41 patients in the radiation group. Mild allergic symptoms developed in 1 patient in the observation group who did not receive follow-up angiography. Observation and radiation groups did not differ at baseline with regard to various fluorescein angiographic features (Table 3). For both groups, VA line loss rates at 1-year follow-up were not statistically different for eyes with classic, mixed, or occult CNV or with laser-eligible CNV at baseline(data not shown).

Sizes of CNV, of CNV + all, and of classic CNV at baseline did not influence VA loss rates (data not shown). Growth rates of CNV did not differ between the randomization groups. The observation and radiation groups demonstrated a mean increase in CNV size at 1-year follow-up of 1.21 and 1.83 categories, respectively (P = .13). The observation and radiation groups demonstrated a mean increase in size of CNV + all at 1-year follow-up of 1.24 and 1.86 categories, respectively (P = .15). The observation and radiation groups demonstrated a mean increase in classic CNV size at 1-year follow-up of 1.75 and 2.06 categories, respectively (P = .69).

At baseline, 20 of 42 observation group eyes and 23 of 41 radiation group eyes demonstrated some blood on fundus photography (Fisher exact test, P = .51). For both groups, the presence or absence of blood at baseline did not influence VA loss rates (data not shown). There was an exception in that, for eyes with blood at baseline, 6-line loss rates were higher for eyes in the radiation group (ANOVA, P= .04). The extent or development of bleeding did not differ between groups in eyes where blood was absent at baseline (data not shown).

At baseline, 4 of 42 observation group eyes and 2 of 41 radiation group eyes demonstrated some fibrosis on fundus photographic findings (Fisher exact test, P = .68). For both groups, the presence or absence of fibrosis at baseline did not influence VA loss rates (data not shown). The extent or development of fibrosis did not differ between groups in eyes where fibrosis was absent at baseline (data not shown).

At 1-year follow-up, 21 of 33 observation group eyes and 30 of 37 radiation group eyes were graded as worsened using fluorescein angiographic and/or fundus photographic findings; no statistically significant differences in fluorescein angiographic and/or fundus photographic deterioration were found (P = .10) (Figure 6).

No evidence of acute or subacute toxic effects of radiation was observed. None of the patients in the observation and radiation groups experienced phosphenes during the sham or radiotherapy sessions. Radiation-related optic neuropathy or retinopathy was not observed in either group. Two patients received a diagnosis of diabetes mellitus after enrollment. Radiation-related retinopathy and diabetic retinopathy did not develop in any patient.

Cataract progression was defined as the clinical grade (0-4) increasing by 1 or more grade at 1 year after baseline for nuclear sclerosis, cortical changes, and posterior subcapsular cataract (PSC). There were 28 and 12 phakic eyes in the radiation and observation groups, respectively. At 1-year follow-up, 8 of 28 phakic eyes and 3 of 12 phakic eyes demonstrated nuclear sclerosis cataract progression in the radiation and observation groups, respectively. At 1-year follow-up, 7 of 28 phakic eyes and 3 of 12 phakic eyes demonstrated cortical cataract progression in the radiation and observation groups, respectively. All 12 phakic eyes in the observation group demonstrated no evidence of PSC at baseline, and PSC had not developed in any of these eyes at 1-year follow-up. In the radiation group, 23 phakic eyes demonstrated no evidence of PSC at baseline; some PSC developed in 1 of the 23 eyes at 1-year follow-up. In the radiation group, 5 phakic eyes demonstrated PSC at baseline; none of these 5 eyes demonstrated PSC progression at 1-year follow-up. At 1-year follow-up, there were no differences in the rate of cataract progression between radiation and observation groups for nuclear sclerosis, cortical changes, and PSC (Fisher exact 2-tail test, P≤.99 for all 3 conditions).

One patient in each group with MPS laser-eligible CNV that had progressed after enrollment elected subfoveal laser photocoagulation ablation. Rhegmatogenous retinal detachment with vitreous hemorrhage developed in 1 patient in the radiation group during the study period; a large, nonclearing vitreous hemorrhage developed in another radiation group patient.

The results of our randomized study suggest that low-dose external beam irradiation at a dose of 14 Gy in 7 fractions of 2 Gy is not beneficial for subfoveal CNV complicating ARMD. No difference in VA, CS, or fluorescein angiographic progression was found between the observation and radiation groups at any follow-up period. Most patients in the radiation group demonstrated evidence of angiographic and/or fundus photographic progression of CNV without a diminished extent of fibrosis or subretinal hemorrhage, compared with patients in the observation group. No significant complications, such as radiation retinopathy, optic neuropathy, or cataract, were identified in radiation-treated patients. Although longer follow-up is necessary to determine if such complications develop, radiotherapy at this dose does not appear to be harmful at 2-year follow-up.

Radiation therapy, a treatment with known antiangiogenic properties, has been investigated as a modality to prevent severe visual loss in exudative ARMD. Since the report by Chakravarthy and coworkers⁴⁰ in 1993, many patients worldwide have received varying forms of radiation for this disease. Nonrandomized, uncontrolled studies^40-51 indicate that radiation therapy for CNV does not have significant short-term adverse effects, does not cause immediate visual loss, and does not require complete angiographic visualization. Conclusions of most reports studying radiotherapy are limited by short follow-up, small numbers of cases, retrospective data collection, absence of standardized VA measurement, absence of strict angiographic and visual entry criteria, and absence of an appropriate randomized control group.^40-51 Most of these uncontrolled, nonrandomized studies have used external beam irradiation with standard fractions of approximately 2 Gy to a total dose of 10 to 20 Gy. Some investigators have reported minimal or no therapeutic external beam irradiation effect,^48-50 whereas others have reported a moderate benefit with standard fractions.^40-46

Higher fractions and doses of external beam irradiation⁴¹ and other modalities such as brachytherapy^42,51 or proton beam irradiation⁴⁷ also have been examined. Nonstandard fractions of external beam irradiation (6- to 8-Gy fractions)⁴¹ and proton beam irradiation (8- and 14-Gy fractions)⁴⁷ have been proposed as beneficial therapy in uncontrolled, nonrandomized studies. Some evidence of angiographic regression of CNV has been observed with higher fractions, especially after proton beam irradiation.

Two randomized, published studies comparing radiation with observation have used higher nonstandard fractions.^56,57 Bergink and coworkers⁵⁶ randomized 74 patients with classic, mixed, or occult subfoveal CNV to observation vs external beam irradiation. Four fractions of 6 Gy (total dose, 24 Gy), a significantly higher total dose and fraction size than in our study, were used. At 1-year follow-up, 52% of the observation group vs 32% of the radiated group lost 3 or more lines of VA (P = .03). Six or more lines of VA loss were observed in 41% of the observation group vs 9% of the radiation group (P = .002). A greater beneficial treatment effect was observed for mixed or occult CNV than for classic CNV. Although radiotherapy seemed to have a stabilizing effect, a significant proportion of irradiated eyes lost VA, and irradiated eyes demonstrated progressive growth of CNV.⁵⁶ Char and coworkers⁵⁷ performed a small randomized trial of 27 eyes with classic and occult subfoveal CNV. Patients were randomized to observation vs external beam irradiation(single fraction of 7.5 Gy). The mean VA line loss was 1.9 in radiation group eyes vs 5.5 in observation group eyes (P = .05). Similar to the study by Bergink and coworkers,⁵⁶ Char and coworkers⁵⁷ found no difference in angiographic progression between eyes in radiation and observation groups.

The results of our study show that the likelihood of a beneficial effect using a 14-Gy dose in 2-Gy fractions is very low. However, a significant proportion of our study eyes demonstrated occult CNV as opposed to other radiotherapy studies, which have required enrolled eyes to demonstrate a component of classic CNV. Many of the analyses examining possible differences among demographic, angiographic, and photographic subgroups deal with small numbers and had low power to find differences. Limitations exist in that a larger sample size might identify a small harmful or beneficial treatment effect for various subgroups (ie, classic vs occult CNV). The lack of treatment benefit in our randomized study may also be due to the use of sham radiation in the observation group. We have demonstrated that our technique of sham radiation keeps patients guessing as to which group they were randomized.⁵⁸ The absence of patient masking may create bias in studies examining subjective psychovisual outcome variables such as VA. Uncontrolled radiotherapy studies that found a small beneficial effect at similar dose and fraction sizes also may have been influenced by such a phenomenon in patients who know they are receiving a "treatment." The Radiation Therapy for Age-related Macular Degeneration Study Group⁵⁹ demonstrated an absence of treatment benefit in a randomized, multicenter radiotherapy trial using 8 fractions of 2 Gy (total, 16 Gy). Their study, similar to ours, used sham radiation and had a large proportion of eyes demonstrating occult CNV.

The absence of any efficacy in our randomized trial, despite the initially promising retrospective studies, highlights the need for caution in evaluating radiotherapy or other experimental treatments for exudative ARMD. However, in light of the beneficial visual outcomes demonstrated in randomized radiotherapy trials using higher fractions and doses, the favorable angiographic response observed with higher doses and fraction sizes, and the sound scientific and clinical rationale for radiotherapy, we believe that radiotherapy should not be eliminated as a potential therapy. Although ours and other studies using a safe and low dose with standard fractions of external beam irradiation did not show VA improvement or fluorescein angiographic regression, use of nonstandard fractions, higher total doses, and extramacular sparing techniques (ie, brachytherapy, charged particle irradiation, and gamma ray knife) may be beneficial. It is hoped that ongoing and future randomized trials will provide useful information for determining whether a therapeutic window for radiotherapy exists.

Accepted for publication July 31, 2000.

Supported in part by an Unrestricted Departmental Award from Research to Prevent Blindness, New York, NY, and by grants from the Knights Templar Educational Foundation of Georgia, Macon.

Corresponding author and reprints: Dennis M. Marcus, MD, Medical College of Georgia, Department of Ophthalmology, 1120 15th St, Augusta, GA 30912 (e-mail: dmarcus@mail.mcg.edu).