The COMS Randomized Trial of Iodine 125 Brachytherapy for Choroidal Melanoma, III: Initial Mortality Findings:  COMS Report No. 18 FREE

CHOROIDAL melanoma, a primary intraocular cancer, has the potential to extend locally through the scleral wall and into the orbit and to metastasize to the liver and other extraocular sites. No effective treatment has been found to date for metastatic choroidal melanoma.^1- 4 Death typically occurs within months after metastasis is detected clinically.^4- 5 Thus, the goal is to treat choroidal melanoma while it is confined to the eye. For nearly 100 years, enucleation was the only treatment for choroidal melanoma that was believed to protect the patient from metastasis and death. Enucleation can be performed by most ophthalmologists and, thus, is widely available. It remains the accepted treatment for large choroidal melanoma.^6- 8 However, alternatives to enucleation have been deemed desirable, particularly for smaller choroidal melanoma. The benefits of an eye-conserving treatment include retention of the eye and some vision and perceived cosmetic benefits that may result in better quality of life. In addition, removal of an eye that occasionally, albeit rarely, has been misdiagnosed as containing choroidal melanoma may be avoided. Moreover, choroidal melanoma that remains small in size is believed by most ophthalmologists and oncologists to have little potential for metastasis to extraocular sites.^9- 11 However, despite technical advances in the development of alternative treatments, evaluations of eye-conserving treatments with respect to mortality or metastatic tumor spread only recently have provided data from long-term follow-up of patients.^12- 14 In addition, eye-conserving treatments may be considerably more expensive than enucleation because of the facilities, materials, or expertise required to administer them or because of more frequent clinical examinations to monitor the tumor for regrowth after treatment.

The interest in alternatives to enucleation of eyes with choroidal melanoma was heightened in the 1970s by an analysis of data presented by Zimmerman and McLean et al^15- 17 that purported to show, contrary to expectation, that enucleation increased the risk of tumor metastasis. This hypothesis and the discussion it provoked^18- 21 emphasized the need for long-term follow-up studies of both enucleation and alternative treatments. At that time, radiotherapy was considered the best alternative to enucleation. Several isotopes and delivery systems had been developed and were in use for treatment of choroidal melanoma in the United States and elsewhere.

By the early 1980s, some ophthalmologists and radiation oncologists advocated radiotherapy as the preferred treatment for choroidal melanoma when the tumor was not so large that enucleation was the only reasonable approach or so small that the diagnosis of choroidal melanoma could not be made reliably without observation of tumor growth during clinical follow-up.^22- 24 However, with only short-term survival data available, many ophthalmologists were concerned that the patient's life would be shortened by any treatment that did not include removal of the eye.

In 1983 and 1984, the National Eye Institute of the National Institutes of Health, Bethesda, Md, sponsored several meetings of ophthalmologists and others interested in the treatment of choroidal melanoma, including proponents of radiotherapy, to evaluate and summarize the survival data available for patients treated with radiotherapy, to compare them with similar data from patients treated with enucleation, and to reach a consensus on the design of studies judged necessary to evaluate radiotherapy. Most ophthalmologists and oncologists who participated in these meetings believed that only a multicenter clinical trial in which patients with this rare cancer would be assigned randomly to eye-conserving radiotherapy or enucleation and followed for a minimum of 5 years for mortality would settle the controversy.

Thus, the Collaborative Ocular Melanoma Study (COMS) was designed primarily to compare radiotherapy with enucleation with respect to mortality among the subset of patients with choroidal melanoma for whom radiotherapy would be considered an acceptable treatment by the ophthalmologist and patient if it provided survival rates comparable to enucleation. When the COMS was designed in 1985, the 2 most widely used methods of radiotherapy for choroidal melanoma were charged particles^22,24 and brachytherapy by means of episcleral plaques to which radioactive material was affixed.^23,25- 27 Several factors were considered in the selection of a radioactive plaque and iodine 125 for radiation delivery in this COMS trial, as discussed elsewhere.^28- 29

Accrual of patients to the COMS clinical trial of ¹²⁵I brachytherapy began in February 1987 and was completed on July 31, 1998, after the number of patients required to meet the primary objectives of the trial had enrolled. The purpose of this report is to present mortality findings after a minimum follow-up of 2 years and a maximum follow-up of 13 years of the patients in the 2 treatment arms. Initial findings from a parallel study of health-related quality of life³⁰ will be available within the next 2 years. Initial findings from a randomized trial of preenucleation external beam irradiation for large choroidal melanoma, also conducted by the COMS Group, were published in 1998.^8,31- 32

Descriptions of many aspects of the COMS design and methods have been published previously.^8,31- 40 The COMS Manual of Procedures⁴¹ and the COMS Forms Book⁴² are available.

The Data and Safety Monitoring Committee appointed by the National Eye Institute director approved the study design and methods on August 27, 1986, prior to initiation of patient evaluation and enrollment. In addition, the institutional review board of each participating institution reviewed and approved the COMS protocol and the consent forms used at the respective clinical center prior to initiation of patient enrollment at individual centers.

Patients with suspected choroidal melanoma were referred by their ophthalmologists to 1 of 43 COMS clinical centers in the United States and Canada for confirmation of the diagnosis and evaluation of eligibility for either of 2 COMS randomized trials of radiotherapy. Details of examinations and methods of patient evaluation and data collection used in one or both COMS clinical trials have been reported elsewhere.^{8,31- 32,37- 40} In brief, a medical and ocular history was elicited before the patient underwent a detailed ophthalmic examination by a COMS ophthalmologist. The ophthalmic examination included ophthalmoscopic and biomicroscopic examination of the tumor and the fundus of each eye. The history and examination addressed the eligibility of the patient and provided baseline descriptive data. Best-corrected visual acuity was measured on a Bailey-Lovie chart in accord with a standard protocol.⁴¹ Standardized A-scan and contact B-scan echography were performed.³⁷ Wide-angle color photographs and frames during fluorescein angiography were obtained of all tumor margins, whenever possible, and of the macula and optic disc of both eyes.

As part of the ophthalmoscopic examination, the ophthalmologist estimated the apical height and the longest basal diameter of the tumor, taking account of both clinical observations and local measurements from echograms. The ophthalmologist also provided a detailed description of tumor location based on distances from landmark intraocular structures, with particular attention to location relative to the optic disc to assess eligibility for enrollment.

Patients judged to be eligible based on ophthalmologic examination and medical history were referred to the local COMS radiation oncologist for discussion of radiation issues. The radiation oncologist, a medical oncologist, or an internist performed a cancer-oriented physical examination that included a detailed medical history, laboratory studies, and anterior-posterior radiography of the chest. A computed tomographic scan, magnetic resonance images, or biopsy of the liver or other organs was required by the COMS protocol only when necessary to rule out metastatic melanoma or other neoplasms following defined abnormalities in tests of liver function or other suspicious findings from the physical examination.

Echograms were forwarded to the COMS Echography Center (Mars Hill, NC) where study personnel reviewed them for consistency with the diagnosis of choroidal melanoma and evidence of extrascleral extension, measured the apical height of the tumor, and classified configuration, location, and internal reflectivity of the tumor. Wide-angle photographs and fluorescein angiograms were forwarded to the COMS Photograph Reading Center (Iowa City, Iowa) for confirmation of the diagnosis of choroidal melanoma. The clinical center ophthalmologist was notified whenever extrascleral extension was noted at the Echography Center and whenever the diagnosis was questioned by personnel at either of these 2 resource centers.

The choroidal melanoma had to be from 2.5 to 10.0 mm in apical height and no more than 16.0 mm in longest basal diameter. (Until November 1990, the lower and upper limits on apical height were 3.1 and 8.0 mm, respectively.) Patients with peripapillary tumors, ie, those with the proximal tumor border 2.0 mm or closer to the optic disc, were eligible only when the tumor was contained within a 90° angle, with the apex at the center of the optic disc, and when the enrolling ophthalmologist was confident that a radioactive episcleral plaque could be placed to cover the entire base of the tumor and a 2-mm margin beyond the tumor borders apart from the border proximal to the optic disc. Patients whose tumors were contiguous with the optic disc were ineligible. Whenever the ocular media were temporarily too opaque to assess eligibility, to make the necessary baseline observations, or to estimate dimensions that would be necessary for planning ¹²⁵I brachytherapy, at the option of the patient and ophthalmologist, a final decision regarding eligibility and enrollment could be postponed until sufficient clarity was achieved; otherwise, the patient was deemed ineligible.

Eligible patients had unilateral primary choroidal melanoma; were aged 21 years or older; had no coexisting disease that threatened survival for 5 years or longer; were judged by the examining oncologist or internist to be free of metastatic melanoma and other primary or metastatic cancers; had best-corrected visual acuity in the fellow eye of 20/200 or better; and were able to give informed consent, willing to adhere to local radiation safety guidelines, and able to return for posttreatment follow-up examinations. In a few instances, patients who had received a cancer diagnosis in the distant past, were documented to be free of recurrence, and were in good health apart from the choroidal melanoma were permitted to enroll in the clinical trial after consultation with and approval by the principal radiation oncologist or the designated medical oncologist. Otherwise, patients who had a history of cancer other than noninvasive nonmelanotic skin cancer or carcinoma in situ of the uterine cervix were ineligible. Previous treatment for choroidal or ciliary body melanoma in either eye, treatment of any condition secondary to the tumor, or fine-needle aspiration biopsy of the melanoma rendered a patient ineligible for the clinical trial. Patients with extrascleral tumor extension 2.0 mm or thicker detected during echography or clinical examination; diffuse, ring, or multifocal tumors; or tumors judged to be predominantly ciliary body melanoma were ineligible. Patients using immunosuppressive therapy that could not be discontinued indefinitely also were ineligible.

Eligible patients who gave signed consent were enrolled by means of a telephone call from the local clinic coordinator and ophthalmologist to the COMS Coordinating Center (Baltimore, Md). During the call, the patient was registered, eligibility was reviewed, and the next treatment assignment, enucleation or ¹²⁵I brachytherapy, was selected by computer from an encoded electronic file of treatment assignments prepared by Coordinating Center personnel and communicated to the clinical personnel. Randomly permuted blocks of assignments of randomly selected sizes were used to construct the treatment assignment files. A separate file of treatment assignments was maintained for each clinical center for each clinical trial. Official enrollment of the patient took place when the treatment assignment was communicated by Coordinating Center personnel to 2 members of the enrolling clinical center staff.

The COMS Radiation Therapy and Physics Committee designed a standard set of episcleral plaques made of gold alloy and patterned after individually fabricated plaques.^27,43 Sets of plaques that were 12, 14, 16, 18, and 20 mm in diameter were fabricated centrally (Trachsel Dental Studio, Rochester, Minn) or locally to the same standards after approval by the Radiation Therapy and Physics Committee.⁴¹ The soft plastic seed carrier was cemented to the plaque so that the ¹²⁵I seeds were placed in evenly spaced slots arranged in concentric rings with 1 mm of carrier between the seeds and the sclera. The size of the plaque was selected to provide a 2- to 3-mm margin beyond the base of the tumor; thus, the size of the plaque was 4- to 6-mm greater than the maximum basal diameter of the tumor. For tumors near the optic nerve, notched plaques of the same design or plaques with the lip removed adjacent to the optic nerve were permitted.

For COMS purposes, the dosimetry recommended by Task Group 43 of the American Association of Physicists in Medicine Radiation Therapy Committee was followed.⁴⁴ Seeds were assumed to be point sources without anisotropy. The effects of the gold plaque on scatter and attenuation were ignored, as was the shielding owing to the plaque rim. The plastic insert was assigned water equivalent absorption. Additional details are provided in the COMS Manual of Procedures.⁴¹

For choroidal melanoma 5 mm or greater in apical height, the dose was prescribed to the apex of the tumor. For tumors 2.5 to 4.9 mm high at the apex, the prescription point was 5 mm from the interior surface of the sclera. The protocol dose to the prescription point was 85 Gy, delivered at a rate of at least 0.42 Gy per hour but no more than 1.05 Gy per hour. Thus, the 85 Gy isodose surface passed through the prescription point, encompassed the entire tumor, and extended at least to the edge of the gold plaque.

The COMS protocol required that each ¹²⁵I seed be assayed separately using an ionization chamber that had been calibrated by the local radiation physicist with an ¹²⁵I seed calibrated at the National Institute of Standards and Technology (Gaithersburg, Md) or at an accredited dosimetry calibration laboratory for the type of ¹²⁵I seed used locally (model 6711 or 6702). Records of the constancy of the chamber factor were maintained for periodic review by physicists from the Radiological Physics Center (Houston, Tex).

Doses to critical ocular structures were calculated by the local COMS radiation oncologist and physicist on the basis of measurements, retinal drawings, and dimensions provided by the examining ophthalmologist. For COMS purposes, the dose to the sclera at the tumor base was estimated at a point on the central axis of the plaque 1 mm from the surface of the plaque. The dose to the optic nerve was calculated at the center of the optic disc; the dose to the macula was calculated at the foveola. The dose to the retina opposite the tumor was calculated 22 mm from the scleral surface at the base of the tumor as measured along the diameter of the globe that passed through the tumor apex.

Prior to approval to enroll patients in the COMS trial, the dosimetry software and calibration facilities at each participating center were tested, and adherence to the COMS protocol was documented for 1 or more test cases. In addition, the local COMS team of ophthalmic surgeon, radiation oncologist, and radiation physicist and the local radiation safety officer were required to collaborate in the preparation and application of an ¹²⁵I plaque to a non-COMS patient. For each COMS treatment, the following data were submitted to the Radiological Physics Center for review: all measurements of tumor location and dimensions; intraocular distances between the tumor and critical structures; a diagram or photograph of the plaque that showed the placement of the ¹²⁵I seeds and the distance from the seeds to the surface of the applicator; seed strength in units of air Kerma Strength and the date of assay; apparent seed activity in units of MCi and the date of assay; number of seeds used; total length of implant; and dose distribution perpendicular to the plaque through the tumor apex, with the tumor outlined on the distribution to show the tumor thickness and basal dimensions. Dosimetry was calculated independently at the Radiological Physics Center. Local dosimetry calculations were corrected as appropriate after confirmation with local COMS personnel. Protocol adherence was assessed by Radiological Physics Center personnel. In addition, all radiotherapy documentation was forwarded to the COMS principal radiation oncologist who reviewed the material and provided a final assessment of adherence to the brachytherapy protocol.

Procedures for plaque placement and removal have been described and illustrated elsewhere.^41,45 Transparent templates were available in the same sizes as the gold shields and were used to facilitate localization and subsequent placement of the radioactive plaque. Ocular and systemic complications observed during placement or removal of the radioactive plaque and during examinations in the immediate postoperative period were reported on standard forms that were sent to the Coordinating Center.

The COMS protocol specified that ¹²⁵I brachytherapy be completed within 4 weeks after enrollment of the patient. In those cases in which plaque placement was delayed by more than 4 weeks, the ophthalmologist was required to repeat the initial ophthalmologic and echographic examinations and to communicate to the radiation oncologist and physicist any changes in tumor dimensions and other intraocular measurements so that corrections could be made to the dosimetry calculations. Complications and observations at the time of plaque placement and removal, dosimetry data, and results of evaluations at the Radiological Physics Center and by the principal radiation oncologist were reported on standard forms⁴² sent to the Coordinating Center.

The COMS protocol specified that enucleation of eyes so assigned take place within 4 weeks after enrollment. Selection of general or local anesthesia was based on patient status, preferences of the patient and surgeon, and local circumstances. After a 360° limbal peritomy and detachment of the conjunctiva and the Tenon capsule, the surgeon inspected each quadrant for evidence of extraocular extension. Whenever extraocular tumor was present, the tumor was dissected beyond all visible margins; transection of any portion was avoided. Recti muscles, and oblique muscles whenever indicated, were cut from insertions with as little manipulation as possible; the optic nerve was cut posterior to the globe beyond any visible tumor extension along the nerve. After removing the eye, the surgeon closed peripheral rents in the Tenon capsule and inserted a nonmetallic implant. The posterior opening in the Tenon capsule was not closed routinely. The surgeon performed a tenonectomy whenever tumor was visible in the sclera or extraocular extension was observed.

Intraoperative and immediate postoperative ocular and systemic complications were reported on standard forms.⁴² After the orbital tissues had healed, patients were referred to an ocularist for construction and fitting of a prosthesis.

Enucleated eyes were processed either locally or centrally for central histopathologic assessment at the COMS Pathology Center (Madison, Wis); the procedures have been reported.^34- 36 Cell type was coded by members of the Pathology Review Committee using the modification of the Callender classification developed at the Armed Forces Institute of Pathology (Washington, DC).⁴⁶ A detailed report of histopathologic findings was sent to the clinical center after central assessment of each ocular tumor.

Patients were examined as necessary for clinical care during the postoperative period; they were scheduled for an examination for data collection purposes no later than 6 weeks after surgery. Subsequent clinical examinations for data collection purposes were scheduled 6 months after enrollment and at 6-month intervals thereafter until the 5-year examination. During the next 5 years, examinations at annual intervals were required; examinations at the midpoint between annual examinations were optional. Whenever a clinical examination was not conducted between annual examinations, the patient was contacted by telephone to ascertain current health status. As a patient approached the 10-year anniversary of enrollment, he or she was asked to consent to continued follow-up examinations and data collection by COMS personnel. Patients who agreed continued to be followed at annual examinations and received either an examination or a telephone interview at the midpoint between annual examinations.

At each scheduled follow-up examination of patients in the ¹²⁵I brachytherapy arm, the ophthalmologist conducted a detailed examination that included assessment of tumor margins for evidence of expansion. Standardized A-scan and contact B-scan echography were performed to ascertain current apical height of the tumor and to detect any extrascleral tumor growth. Suspected growth of choroidal melanoma during follow-up after the immediate posttreatment period mandated more frequent follow-up examinations. Treatment failure, as documented by growth of the tumor, extrascleral extension, or severe ocular pain, was managed at the discretion of the ophthalmologist. The reason for treatment failure and any additional treatment were reported to the Coordinating Center.

For each patient whose eye had been enucleated, the ophthalmologist examined the socket and eyelids for local tumor recurrence. The fit of the prosthesis was checked and the patient was referred to an ocularist for modifications as necessary. The fellow eyes of all patients were examined for any new ophthalmologic problems. The best-corrected visual acuity of the fellow eye and tumor-containing eye when present was measured following the same protocol as at baseline. A cancer-oriented physical examination was performed; blood tests and chest radiography, and other tests, as required, were performed to judge whether distant metastasis of the melanoma had occurred.

Documentation of metastasis, including reports of findings from diagnostic procedures and biopsy slides when available, was provided for central review. Deaths of patients were reported promptly, either by mail or telephone pending receipt of the death certificate, and logged immediately at the Coordinating Center. In May 1997, a search of the National Death Index database (National Center for Health Statistics, Department of Health and Human Services, Rockville, Md) was undertaken to ascertain the vital status of a few patients who had been reported to be lost to follow-up by the responsible clinical center personnel.

Information concerning events prior to death was evaluated by the Mortality Coding Committee to ascertain whether metastatic melanoma was present at the time of death. The committee members were masked to the treatment arm of each deceased patient. Documentation of the course of the final illness, including laboratory and diagnostic studies and clinical observations, was obtained from the local caregivers by local COMS clinical personnel and forwarded to the Coordinating Center for review and masking of information that would identify the patient or reveal the initial treatment for choroidal melanoma. Whenever possible, autopsy and biopsy slides of suspected malignancies were obtained for central review. Metastatic melanoma was diagnosed only when confirmed by autopsy, biopsy, or cytology findings. A detailed description of the procedures followed by the Mortality Coding Committee has been reported elsewhere.³⁹

Because it was judged unlikely that a similar randomized trial for choroidal melanoma patients would be carried out by the COMS Group or other investigators, particular attention was given to assuring that the COMS was designed and conducted in a manner to produce data of high quality. Quality assurance activities focused on diagnostic accuracy, enrollment of eligible patients, centralization of random treatment assignment and other key functions in specialized resource centers and expert committees, adherence to treatment protocols, completeness of patient follow-up, and careful review of all records relating to cause of death.

The COMS Quality Assurance Committee was responsible for quality assurance activities and for monitoring protocol adherence and data quality. Each resource center was represented by a member on the Quality Assurance Committee and had specific responsibilities for data quality, including training and certification of COMS personnel in their study roles and in data recording and submission procedures. Deficiencies in materials or data submitted were called to the attention of the COMS protocol monitor who chaired the Quality Assurance Committee.

Following an initial site visit to each center by members of the Executive Committee, site visits at 2- to 3-year intervals, or more frequently when indicated, were made to each clinical center by personnel from the Coordinating Center. During these site visits, local COMS operations were observed, protocol adherence was reviewed, and samples of data reported to the Coordinating Center were compared with findings in patient charts. Site visits to clinical centers by personnel located at other COMS resource centers were scheduled as indicated to solve problems, train new COMS personnel, or calibrate replacement equipment and facilities. At least 1 follow-up site visit was made to each resource center approximately 5 years after the initial visit by members of the Executive Committee or Data and Safety Monitoring Committee or by outside experts.

Beginning in September 1987, an independent Data and Safety Monitoring Committee reviewed accumulated data twice each year, with at least 1 review each year taking place at a full committee meeting. Initially, the committee reviewed mortality data from the 2 treatment groups combined; all other data, including complications, were reviewed with the treatment arm identified. At annual meetings in 1990, 1992, 1995, and 1997, the committee reviewed mortality data presented separately by treatment arm. At the time of the November 1997 review when the minimum accrual goal had been met, the committee recommended that accrual continue until July 31, 1998, that no additional patients be enrolled after that date but that follow-up of all patients continue in order to increase statistical power. This recommendation was made after reviewing information on survival by treatment arm and several analyses of the conditional likelihood of future changes in survival trends in each treatment arm. At the November 1999 meeting, the committee recommended that follow-up continue until the last patient enrolled was eligible for a 5-year examination and that the earliest patients enrolled be followed clinically for a maximum of 15 years, with the final study examinations to take place at the annual examinations between August 1, 2002, and July 31, 2003. At the December 13, 2000, meeting, with the 5-year survival outcome known for more than 80% of all patients, the committee approved publication of current mortality findings and reaffirmed the recommendation made in November 1999 that follow-up of patients continue.

A range of acceptable sample sizes was developed during the design phase of the trial.^33,41 Sample size estimates were based on the assumption that 70% of the patients assigned randomly to the enucleation arm would be alive 5 years after enrollment⁴⁷ and that the study should be able to detect or rule out a relative difference in 5-year all-cause mortality of 25% (eg, 30% vs 22.5% or 30% vs 37.5%). The maximum target sample size of 2400 patients was based on a type I (α) error of .01 (2-sided), a type II (β) error of .10 (ie, 90% power), and a projected 5-year rate of loss to follow-up for vital status of 5% of patients enrolled. The minimum acceptable sample size of 1250 patients was designed to provide 80% power and a type I error of .05, typical choices for randomized trials of treatments for other cancers, with no change in the other assumptions.

Data from clinical evaluations and from review of echograms, photographs, sections of enucleated eyes, and status at death were recorded on standard forms⁴² that were forwarded to the Coordinating Center for transcription to computer files. Automated and manual edits of the data for completeness, consistency, and protocol adherence; integration of all data for each patient; and data analysis and reporting were the responsibility of the Coordinating Center. Data received by September 30, 2000, have been included in this analysis.

All patients were analyzed with the treatment arm to which they were assigned randomly at time of enrollment, regardless of subsequent findings or treatment received ("intent-to-treat" analysis strategy). The Wilcoxon rank sum test⁴⁸ has been used to compare distributions of continuous variables. The χ² test for trend in ordered categories⁴⁹ was used to compare distributions of categorical variables in 3 or more qualitatively or quantitatively ordered categories. The χ² test for homogeneity⁴⁸ was used to compare dichotomous distributions and distributions in which the categories have no inherent qualitative or quantitative order. Time to death was estimated using the product-limit method of Kaplan and Meier.⁵⁰ Survival rates by treatment arm were compared using the log-rank test^51- 52; no adjustment has been made to account for 5 reviews of outcome data by the Data and Safety Monitoring Committee. All statements of probability (P values) are 2-sided. Patients who refused follow-up in the COMS beyond the 10-year examination were censored following that examination. For analysis of time to death with metastatic melanoma, deceased patients without histopathologically confirmed metastatic melanoma were censored at time of death. Patients not reported as deceased and not discovered to have died during the May 1997 search of the National Death Index database were assumed to be alive as of the cutoff date for this analysis. The proportional hazards model of Cox⁵³ has been used to adjust for the simultaneous effects of baseline covariates to account for minor imbalances between treatment arms at enrollment. SAS computer software (SAS Institute, Cary, NC) was used for proportional hazards analyses and for calculating risk ratios and confidence intervals (CIs).

When accrual ended in July 1998, 1317 patients had enrolled: 46% of 2882 patients with choroidal melanoma who were judged by the local study investigators to be eligible for the trial, and 26% of all 5046 patients reported to have choroidal melanoma of eligible size. Reasons for ineligibility and other reasons patients did not enroll have been reported elsewhere.⁴⁰ The most frequently reported reasons for ineligibility were tumor proximity or contiguity to the optic disc, previous diagnosis of another primary cancer, and tumor location predominantly in the ciliary body.

Of the 1317 patients who enrolled, 660 were assigned randomly to enucleation, and 657 were assigned to ¹²⁵I brachytherapy. After all baseline data had been received at the Coordinating Center, 12 patients (7 assigned to enucleation and 5 assigned to ¹²⁵I brachytherapy) were discovered to have been ineligible for the clinical trial based on strict application of COMS criteria. Seven of these 12 patients (4 assigned to enucleation and 3 assigned to brachytherapy) had an earlier diagnosis of cancer. One patient assigned to brachytherapy had earlier photocoagulation to a retinal detachment associated with the tumor; another was receiving corticosteroids at enrollment. One patient assigned to enucleation had extrascleral extension greater than 2 mm visible on echogram; another had visual acuity in the fellow eye worse than 20/200. The remaining patient assigned to enucleation was judged ineligible based on central review of echograms and photographs and was diagnosed to have a choroidal neovascular lesion rather than choroidal melanoma. The data from each of the 12 patients found to be ineligible after enrollment have been analyzed with those of the treatment group to which the patient was assigned at time of enrollment.

The numbers of patients in each treatment group by length of mortality follow-up through September 30, 2000, are given in Table 1. All patients had been followed for vital status for 2 years or longer; 1274 (97%) had been followed for at least 3 years. Based on time since enrollment, 1072 patients (81%) were eligible for 5 or more years of mortality follow-up; 416 patients (32%) were eligible for 10 or more years of follow-up.

Sociodemographic characteristics of patients, characteristics of choroidal melanoma, and ocular status at time of enrollment are given in Table 2, Table 3, Table 4, Table 5, and Table 6. Mean ages of patients at the time of enrollment were 60.4 years and 59.1 years for those assigned to enucleation and brachytherapy, respectively (P = .13). The treatment arms were well balanced on all other baseline characteristics.

Most patients enrolled within 1 month of the initial diagnosis of choroidal melanoma (Table 3); the median interval from diagnosis to enrollment was 20 days. However, 82 patients assigned to enucleation (13%) and 105 patients assigned to brachytherapy (16%) enrolled 6 months or longer after diagnosis. In each treatment arm, 45% of the tumors were 2.5 mm to 4.0 mm in apical height (Table 3). The mean apical height of the tumors was 4.8 mm; mean length of the longest diameter of the tumor base was 11.4 mm. Among 728 patients(55%) the entire tumor was located posterior to the equator; the anterior tumor boundary extended into the ciliary body in the eyes of only 143 patients(11%) (Table 4). In 632 patients(48%) the tumor apex was in the quadrant temporal to the fovea; the tumor apex was in the quadrant nasal to the fovea for only 159 patients (12%). The median distance between the closest tumor margin and the edge of the optic disc was 4.0 mm; 216 patients (16%) had choroidal melanoma that extended closer than 2.0 mm to the proximal edge of the optic disc. A larger percentage of the patients had choroidal melanoma located over or close to the center of the foveal avascular zone: 190 (14%) over the center, and 345 (26%) within 2.0 mm of the center (Table 4). The median distance between the tumor and the center of the foveal avascular zone was 3.0 mm.

Features of the tumors are displayed in Table 5. A dilated feeder vessel to the tumor was visible at baseline in 46 eyes (3%); a nonrhegmatogenous retinal detachment associated with the tumor was reported for 720 eyes (55%), and a rhegmatogenous detachment for 6 eyes (<1%). Most of the tumors (1014 [77%]) were dome shaped; only 207(16%) were collar-button shaped. Internal reflectivity was low, low-medium, or medium for 1111 tumors (84%). Epibulbar pigment was reported by the ophthalmologist for 13 patients (1%). The ophthalmologist reported drusen on the tumor surface for 91 (25%) of 350 patients enrolled by 1990 when reporting of this observation was halted (data not shown).

Among the study eyes (tumor-containing eyes), 255 (19%) had other vision-limiting or vision-threatening conditions diagnosed in the past (Table 6). The most frequently reported condition was cataract in 165 study eyes (13%). Among fellow eyes, 227 (17%) had conditions that limited or threatened vision; cataract also was the most common condition in fellow eyes, accounting for 166 fellow eyes (13%). Nevertheless, almost all patients were phakic in both eyes (Table 6).

As given in Table 6, more than two thirds of patients had excellent visual acuity in the fellow eye; only 23 patients had fellow eye visual acuity worse than 20/50. Although nearly one third of patients had a best-corrected visual acuity of 20/20 or better in the study eye at baseline, visual acuities spanned a wide range, with 64 patients (5%) having visual acuity worse than 20/500 (10/250) in the study eye.

Although patients with coexisting medical conditions expected to lead to death within 5 years, as judged by the examining physician, were excluded, more than two thirds of participating patients had 1 or more coexisting medical conditions at the time of enrollment (Table 7). Cardiovascular and circulatory conditions were the most common, with 290 patients in the enucleation arm (44%) and 256 patients in the brachytherapy arm (39%) reporting such conditions. The most common finding during baseline examination of the skin was cutaneous moles, reported for 33% of the patients in the enucleation arm and for 30% of the patients in the brachytherapy arm(data not shown). Fewer than 20 patients in each treatment arm (<3%) were reported to have vitiligo, subcutaneous nodules, or unusual pigmentation.

Nearly two thirds of the participating patients had 1 or more blood relative who had cancer (data not shown). Fourteen patients, 5 in the enucleation arm and 9 in the brachytherapy arm, reported having a relative who had been diagnosed with choroidal melanoma. Sixty-seven patients, 37 assigned to enucleation and 30 assigned to brachytherapy, reported a relative who had cutaneous melanoma.

Of the 657 patients assigned to ¹²⁵I brachytherapy, 650 were treated in the manner assigned and within the time limits specified by the COMS protocol. One patient delayed brachytherapy for 7 months; another patient delayed brachytherapy until after fine-needle aspiration biopsy and was treated at a center not participating in the COMS. A third patient who was discovered to have metastasis soon after enrollment was observed without treatment until death 6 months later. After random assignment to brachytherapy, 2 patients decided to undergo enucleation, and 1 decided to have teletherapy. For 1 additional patient, the COMS surgeon judged during surgery that a plaque could not be placed to cover the tumor base; subsequently, the eye was removed.

Of the 660 patients assigned to enucleation, 646 were treated promptly as assigned. One patient delayed enucleation for 14 months until the tumor had grown and the eye had become blind and painful. One patient initially refused enucleation; he returned to the clinical center nearly 6 years after enrollment to have the eye removed. One additional patient refused enucleation after randomization and was observed. Another patient did not undergo enucleation after a misdiagnosis of choroidal neovascularization was discovered during central review of echograms and photographs. After random assignment to enucleation, 7 patients decided to receive brachytherapy; 2, proton beam radiation; and 1, cryotherapy and eye-wall resection outside the COMS. Thus, all but 21 (1.6%) of 1317 patients were treated promptly as assigned at time of enrollment. Furthermore, 2 of these 21 patients received the assigned treatment after an initial delay; 1 additional patient had the assigned treatment after a prolonged delay, and 1 received the assigned treatment at a center not participating in the COMS.

Table 8 summarizes ¹²⁵I brachytherapy and dosimetry for the 650 patients who were so assigned and treated promptly. Summary distributions are presented separately by the location of the prescription point. The values represent the minimum (0 percentile); lower bound of the central 90% (5th percentile); lower quartile (25th percentile); median (50th percentile); upper quartile (75th percentile); upper bound of the central 90% (95th percentile); and maximum (100th percentile) of each distribution. As expected, the median number of ¹²⁵I seeds used for smaller tumors (prescription point located 5 mm from the inner surface of the sclera) was less than for larger tumors (prescription point at tumor apex): 13 vs 20. The distributions of the duration of the implant were similar, apart from a small shift of 4 to 6 hours in the median and the 25th and 75th percentiles of the distributions. The extreme duration of 458 hours resulted from an initial miscalculation of the dose rate that was discovered after the radioactive plaque had been inserted. Two patients in each subgroup had a dose rate lower than that prescribed by the brachytherapy protocol; for 3 patients, 1 whose prescription point was 5 mm from the sclera and 2 with the prescription point at the tumor apex, the dose rate was higher than that specified by the protocol. In both subgroups, more tumors received a radiation dose that was lower than that specified by the COMS protocol than received a dose that was greater than protocol (Table 8). In the subgroup of smaller tumors (prescription point 5 mm from the inner surface of the sclera), undertreatment typically resulted from use of the tumor apex as the prescription point

Central reviews of dosimetry and protocol adherence were completed for all 650 eyes treated with brachytherapy. Personnel at the Radiological Physics Center judged the size of the plaque used to be incorrectly selected in 17(3%) of the cases, with 4 cases ruled to have minor deviations (ie, plaque diameter more than 6.0 mm but not more than 8.0 mm greater than the longest basal diameter of the tumor) and 13 judged to be major deviations (plaque diameter less than 4.0 mm or more than 8.0 mm greater than the longest basal diameter of the tumor). With respect to dose to the prescription point, 17 minor deviations (dose delivered differed by more than 5% but no more than 10% of the target dose of 85 Gy) and 21 major deviations were reported. Dose rates outside the range prescribed by the protocol were rare; only 9 were noted during central review. Overall, 592 tumors (91%) were treated per protocol specifications. Of the remainder, a minor deviation was noted for 24 (4%) and a major deviation for 33 (5%). In the remaining case, the central seed positions were left unfilled. Although the COMS protocol specified an even distribution of sources, uneven distribution was not defined specifically as a protocol deviation. Nonetheless, the higher scleral dose relative to the prescription dose was noted as an anomaly during central review.

Table 9 summarizes complications and observations reported at the time of enucleation surgery or at the time of plaque application or removal. For 2 patients in the enucleation arm, extrascleral tumor extension 2 mm or thicker was observed at the time of surgery. For 5 additional patients, tumor extension to the scleral surface or extrascleral extension less than 2-mm thick was discovered at the time of surgery; a tenonectomy was performed for these 5 patients and for 3 others for other reasons. Other complications applicable to patients in both treatment arms were reported with similar frequency.

Among brachytherapy patients, the most common complications during the first 5 years after treatment were loss of visual acuity³⁸ and growth of the tumor and other indications that led to enucleation. Among all patients treated with ¹²⁵I brachytherapy, regardless of baseline visual acuity, the 5-year cumulative rate and 95% CI for visual acuity of 20/200 or worse in the study eye at 2 consecutive examinations was 63% (95% CI, 59%-68%); for visual acuity of 5/200 or worse, the rate was 45% (95% CI, 41%-50%). By 5 years after ¹²⁵I brachytherapy, the cumulative rate of enucleation was 12% (95% CI, 10%-15%). A detailed description of complications and additional treatments during follow-up will be reported separately.

Complications following enucleation were similar to those already reported for eyes with larger choroidal melanoma.³² Apart from ptosis and problems related to the implant and prosthesis, few complications were reported during the first 5 years after enucleation.

The Pathology Review Committee received sections for evaluation from 647 eyes of patients assigned to enucleation. The tumor in 2 (<1%) of the 647 eyes was judged to be adenocarcinoma, as reported previously.³⁴ Among the remaining 645 eyes, tumors with both epithelioid cells and spindle cells accounted for 551 (85%), epithelioid cells only for 19 (3%), and spindle cells only for 75 (12%). As already noted, 1 additional eye assigned to enucleation was discovered after enrollment to have been misdiagnosed, and the eye was not removed.

The vital status of each patient had been updated during the 6-month period prior to the analysis cutoff date for all but 54 patients—33 assigned to enucleation and 21 assigned to ¹²⁵I brachytherapy. One of these patients had been diagnosed with metastatic melanoma 2 years earlier. (Shortly after the cutoff date for this analysis, 15 of the 21 brachytherapy patients and 17 of the 33 enucleation patients were confirmed to be alive; 2 enucleation patients were reported to have died.) Only 5 patients had been reported to be lost to follow-up by clinical center personnel.

As of the analysis date, 188 patients (28%) assigned to enucleation and 176 patients (27%) assigned to ¹²⁵I brachytherapy were known to have died (P = .48). Cumulative mortality by time since enrollment is displayed in Figure 1. The estimated 5-year cumulative mortality rates and 95% CIs were 19% (95% CI, 16%-23%) for patients in the enucleation arm and 18% (95% CI, 15%-21%) for patients in the brachytherapy arm. Adjustment for time since enrollment using a Cox proportional hazards model with treatment arm as the only covariate confirmed the nearly identical mortality rates in the 2 treatment arms, with a risk ratio of 0.93 (95% CI, 0.76-1.14) for brachytherapy relative to enucleation. This risk ratio and CI bounds are equivalent to a difference in mortality rates of 1.3% in favor of brachytherapy bounded by 4.6% better to 2.7% worse following brachytherapy than following enucleation.

Survival rates in the 2 treatment arms were examined within many subgroups of patients defined on the basis of baseline demographic, medical, and tumor characteristics: age, sex, ethnicity, educational attainment, smoking history, history of cardiovascular disease and other conditions, apical height and longest basal diameter of the choroidal melanoma, location and configuration of the tumor, and time from diagnosis to enrollment in the clinical trial. Among patients who reported graduation from high school without additional years of education and patients without any coexisting disease at baseline, unadjusted survival rates were somewhat better following brachytherapy than following enucleation; among patients with coexisting cardiovascular disease at baseline, unadjusted survival rates were somewhat better following enucleation than following brachytherapy (P<.05 for all 3 univariate comparisons).

In a multivariate Cox proportional hazards survival model in which treatment arm and baseline characteristics were included as covariates, age at time of enrollment, apical height of the tumor, longest diameter of the tumor base, distance from the proximal tumor border to the optic disc, tumor shape judged from B-scan echograms, smoking status, and coexisting medical conditions had independent and statistically significant effects on the length of survival(P<.05). Adjustment for all baseline covariates listed in the previous paragraph yielded an estimated risk ratio of 0.99 (95% CI, 0.80-1.22). These risk ratios translate into a difference in mortality of 0.2%, with this estimate bounded by 3.8% better to 4.2% worse following brachytherapy than following enucleation. Thus, neither the adjusted nor unadjusted mortality rates differed between the 2 treatment arms to a statistically or clinically significant degree in this clinical trial. The 95% CIs for the risk ratios suggest that the true relative difference in mortality, if any, is less than 25%.

Of the 364 deaths from all causes, 345 (95%) had been reviewed and classified by the Mortality Coding Committee. Findings from their reviews are summarized in Table 10. Fewer than half the decedents in each treatment arm were judged to have metastatic melanoma at time of death; all but 1 of these 159 patients were judged by the Mortality Coding Committee to have died because of the melanoma metastasis. The liver was the most frequently reported site of metastasis, with multiple sites reported for many patients. Metastatic melanoma was suspected to be present at death in 43 additional patients based on clinical findings, but histopathologic confirmation was not available. For 88 decedents (26%), there was no evidence of malignancy at time of death.

In Figure 2, estimated cumulative rates of death with histopathologically confirmed melanoma metastasis are shown by treatment arm and time since enrollment. Five-year rates of death with histopathologically confirmed metastatic melanoma were 11% (95% CI, 8%-13%) after enucleation and 9% (95% CI, 7%-11%) after ¹²⁵I brachytherapy(P = .56). Only age, longest tumor basal diameter, and location of the tumor relative to the optic disc were independent and statistically significant predictors of death with histopathologically confirmed metastatic melanoma (P<.05). Adjustment for all baseline covariates yielded an estimated risk ratio for this outcome of 0.91(95% CI, 0.66-1.24). Thus, based on the bounds of the 95% CI, the rate of death with histopathologically confirmed melanoma metastasis is unlikely to be 25% worse following brachytherapy than following enucleation.

Mortality rates among patients treated with eye-conserving ¹²⁵I brachytherapy were similar to rates among patients whose eyes were enucleated for up to 12 years of follow-up, regardless of whether deaths from all causes or only deaths with metastatic melanoma are considered. The 5-year cumulative mortality rate of 19% following enucleation was lower than the projected rate of 30% based on published data available at the time the COMS was designed.⁴⁷ This difference is likely attributable to the prospectively established eligibility criteria for the clinical trial, which were intended to include only patients with choroidal melanoma who otherwise had a good prognosis for survival for 5 years or longer. The power of this randomized trial was greater than 80%, sufficient to conclude that neither treatment is likely to increase or decrease mortality rates by as much as 25% relative to the other. Thus, these findings apply to patients who meet COMS eligibility criteria and who are suitable candidates for either enucleation or ¹²⁵I brachytherapy as delivered in the COMS. Because a single method of eye-conserving radiotherapy (brachytherapy) and a single isotope(¹²⁵I) were evaluated in this clinical trial, caution is advised when considering extrapolating survival rates to other methods of radiotherapy delivery and other isotopes. As a result of the standard episcleral plaques and dosimetry designed for the COMS, ¹²⁵I brachytherapy has become widely available. In particular, the ophthalmologists and radiation oncologists who participated in the COMS have acquired considerable expertise in the use of this treatment.

As reported here and elsewhere,⁴⁰ the most frequent reason that patients with tumors of eligible size were ruled ineligible was proximity of the tumor to the optic nerve, which would have hampered correct placement of a radioactive plaque and delivery of the protocol radiation dose to the prescription point. Only 16% of the patients in this clinical trial had choroidal melanoma that extended closer than 2 mm to the optic disc (Table 3). Nearly half the patients had choroidal melanoma with the apex temporal to the fovea (Table 4). Thus, caution is recommended when considering extrapolation of COMS findings to peripapillary tumors that would not meet the eligibility criteria for this trial.

In the COMS design and conduct, a great deal of attention was devoted to assuring that patients received high-quality medical care and that data of excellent quality were collected. Personnel at the resource centers played key roles in assuring that the study was conducted to high standards. They participated in development of the COMS protocol and in the training of clinical personnel in study responsibilities. Other quality assurance practices that contributed to the excellent record of diagnostic accuracy (>99%) included central review of baseline echograms and photographs and tumor sections from enucleated eyes, a detailed review of materials for each patient judged ineligible after enrollment, and presentations at the annual meetings of participating physicians and other personnel. Similarly, factors that contributed to the excellent adherence to the brachytherapy protocol (91% of eyes treated per protocol) included central confirmation of dosimetry and discussions of each deviation from the brachytherapy protocol at the annual meetings of COMS personnel. The COMS record is exceptionally good among multicenter studies of radiotherapy.^54- 57 A key element of this achievement was the decision to adopt a standard plaque design and to collect a standard set of documentation for each patient treated with ¹²⁵I brachytherapy. Adherence to all aspects of the COMS protocol by personnel at each participating center was monitored by designated investigators who reviewed performance data on a regular schedule. In addition, important aspects of protocol adherence were reviewed annually with all investigators. These measures resulted in 99% of the patients satisfying strict application of all eligibility criteria and nearly complete follow-up of all patients.

Our findings support advising patients who have choroidal melanoma of appropriate size and who meet other COMS eligibility criteria that the choice of ¹²⁵I brachytherapy is unlikely to compromise survival. However, it is important that such patients have realistic expectations regarding the possibility of visual loss that they are likely to experience and of subsequent enucleation owing to tumor growth or other complications.³⁸ A detailed analysis of risk factors for changes in visual acuity during the first 3 years of follow-up after ¹²⁵I brachytherapy has been reported by the COMS Group; greater apical height of the tumor and closer proximity to the fovea were the strongest predictors of early loss of visual acuity.³⁸ In a parallel study,³⁰ quality of life is being assessed among patients who enrolled in this COMS trial in 1995 or later and survivors who had enrolled earlier; 3-year comparisons of treatment arms are expected to be published within the next 2 years. Findings regarding quality of life will be useful to patients and physicians who are faced with making a management choice when choroidal melanoma is diagnosed and treatment is indicated but the tumor is not yet so large that enucleation is required.

In summary, the COMS Group has documented in a carefully designed and conducted randomized trial of ¹²⁵I brachytherapy vs enucleation among 1317 patients suitable for either treatment that survival rates were essentially the same over a follow-up period of 2 to 12 years. Possibly the most encouraging finding is that survival of patients who met COMS criteria and were treated and followed according to the COMS protocol is better than projected when the COMS was designed in 1985, with 5-year survival rates of better than 80% in both treatment arms. The combined 5-year survival rate was 82% (95% CI, 80%-84%). Among patients treated with ¹²⁵I brachytherapy, 85% retained their eyes for 5 years or longer; 37% had visual acuity better than 20/200 in the irradiated eye 5 years after treatment. These findings are applicable to patients who meet the criteria used to select patients for the randomized trial; ophthalmologists should be cautious when considering extrapolation of the findings to other patients with choroidal melanoma. Follow-up of COMS patients will continue through July 2003 to provide data through 5 years of follow-up for all patients and to refine estimates of survival rates, complication rates, and quality of life outcomes during 5 to 15 years after treatment. When the COMS is completed, ophthalmologists will be in a better position to answer with confidence the question posed in 1986: "Do I take the eye out or leave it in?"⁵⁸

Accepted for publication April 16, 2001.

The COMS has received support from the National Eye Institute and the National Cancer Institute, National Institutes of Health, Bethesda, Md, through cooperative agreements EY 06253, EY 06257, EY 06258, EY 06260, EY 06264, EY 06265, EY 06266, EY 06268, EY 06269, EY 06270, EY 06274, EY 06275, EY 06276, EY 06279, EY 06280, EY 06282, EY 06283, EY 06284, EY 06287, EY 06288, EY 06289, EY 06291, EY 06839, EY 06843, EY 06844, EY 06848, EY 06858, and EY 06899.

The contributions by the participating patients and their families and friends are immeasurable and are gratefully acknowledged by the COMS investigators.

Corresponding author and reprints: Barbara S. Hawkins, PhD, COMS Coordinating Center, 550 N Broadway, Ninth Floor, Baltimore, MD 21205-2010.