The Collaborative Ocular Melanoma Study Group. Assessment of Metastatic Disease Status at Death in 435 Patients With Large Choroidal Melanoma in the Collaborative Ocular Melanoma Study (COMS)COMS Report No. 15. Arch Ophthalmol. 2001;119(5):670–676. doi:10.1001/archopht.119.5.670
A systematic review and assessment of disease-related mortality as part of standardized prospective patient follow-up and evaluation within a multicenter clinical trial have been lacking in previous studies of choroidal melanoma.
To describe disease status at death in patients with large choroidal melanoma treated and followed up in the Collaborative Ocular Melanoma Study(COMS).
Analysis of reviews of patient status at death performed by the COMS Mortality Coding Committee using available clinical and histopathologic information.
Setting and Patients
Reviews of deaths as of July 31, 1997, the cutoff date for reporting initial mortality findings.
Patients were treated by either enucleation preceded by external beam radiotherapy or enucleation only.
Main Outcome Measures
Disease status at the time of death and certainty associated with the coding of disease status, sites of metastasis, and availability of autopsy.
Of 1003 patients enrolled in the trial, 457 had died; the estimated median survival from time of enrollment was 7.4 years. Disease status at time of death had been reviewed for 435 deaths (95%). The autopsy rate was 6%. A total of 269 patients (62%) had histopathologically confirmed melanoma metastasis at the time of death, and metastasis was suspected in 92 additional patients(21%) on the basis of imaging and tests but without tissue confirmation. The common sites were liver (93%), lung (24%), and bone (16%); multiple sites were identified in 87% of patients with metastasis. The likelihood of 3 or more sites increased more than 4-fold when autopsy results were available.
Detailed mortality coding following a standard protocol provides the most accurate reporting to date of disease-related mortality in patients with choroidal melanoma and also identifies difficulties. Guidelines for the evaluation of future patients in clinical studies of choroidal melanoma are suggested.
METHODS FOR coding the disease status of a patient at the time of death are important in clinical trials of interventions intended to prolong survival or to prevent life-threatening complications. The effect of treatment on mortality and, in particular, disease-related mortality is often the primary outcome of interest. Assessment of disease-related mortality is essential in describing the differential impact of various treatments on the subsequent history of a disease and may lead to better strategies for evaluating and managing patients. For patients with primary cancers, specification of the sites of metastatic disease may provide useful information for the design of prophylactic or therapeutic interventions.
Although rare, choroidal melanoma is the most common intraocular cancer in adults,1 but little is known about the natural history of choroidal melanoma and the patterns of mortality after treatment. However, when metastasis occurs, the first site identified is usually the liver, and the time between diagnosis of metastasis and death is often short.2 Previous studies3- 16 of survival after treatment for choroidal melanoma have reported melanoma-specific or tumor-related mortality rates. Various terms have been used to describe deaths related to melanoma, including melanoma deaths, melanoma-related mortality, deaths from malignant choroidal melanoma, and deaths from confirmed metastatic uveal melanoma. In general, use of the term melanoma-related mortality has been based on varying sources of information: contact with the patient, his or her family, or physicians who regularly followed up the patient; death certificates; and/or other anecdotal information. Melanoma metastasis is typically evaluated by tests or methods with varying sensitivity and specificity: liver enzyme studies, liver scan, and/or biopsy or autopsy.10 In previous studies, both Seddon et al9 and Augsburger et al14 have categorized status at death, based on information regarding the patient's most recent follow-up contact, as dead of metastatic melanoma, dead of nonmelanoma cancer, or dead of a noncancer cause. However, using this grouping, a death was assigned as melanoma related even when the melanoma was not the patient's immediate or underlying cause of death. Others also have coded melanoma-related mortality on the basis of a history of metastatic melanoma11,15 rather than on a review of clinical and histopathologic materials available near the time of death.
As part of a meta-analysis of all-cause mortality following enucleation for choroidal melanoma, we discovered that the arithmetic difference between tumor-related and all-cause 5-year mortality rates in reports specifying both ranged from 6% to 25%.17 Part of this variability appeared to be due to differences in design, data collection, and patient follow-up among these studies. Previous studies often have had retrospective study designs in which patient information was retrieved after notification of his or her death was received. Follow-up of patients in these studies may not have been regular or complete, information leading to the diagnosis of metastasis may not have been collected routinely, and the sites and extent of metastasis may not have been documented. These problems can be minimized in a prospective randomized controlled clinical trial for which standard schedules and methods for patient follow-up and data collection provide mechanisms for monitoring disease status over time and retrieving clinical information and pathologic materials associated with a patient's death.
Methods for coding the metastatic status of patients at the time of death were developed as part of the Collaborative Ocular Melanoma Study (COMS). This article describes the methods and codings for patients in the COMS large tumor trial; reports the results as of July 31, 1997, the cutoff date for reporting the initial mortality findings; and discusses the implications of these findings for this trial and others.
The COMS is a set of multicenter randomized clinical trials designed to evaluate the effectiveness of radiotherapy in comparison or in addition to standard enucleation in prolonging survival of patients with choroidal melanoma.18 The COMS trial for large choroidal melanoma compared enucleation alone with enucleation preceded by external beam radiation. Large tumors were defined as those for which ophthalmologists agreed that the eye should be removed, ie, those greater than 10 mm in apical height or at least 16 mm in diameter, or tumors 8 to 10 mm in apical height but too close to the optic nerve for effective treatment with radiation without irreversible damage to the optic nerve. In addition to requiring accurate delineation by echography of the height of their unilateral primary choroidal melanoma, patients were eligible for the trial only if they had no previous diagnosis of metastasis or another primary cancer and also met the requirements of a systemic evaluation, which included a chest x-ray examination and liver function tests. Following enucleation, eyes were processed for histopathologic confirmation of the diagnosis and cell type at the central COMS Pathology Center (Madison, Wis).19 After treatment, patients had an annual examination, which included ophthalmologic examinations of the study eye and orbit and the fellow eye; a yearly physical examination; liver function tests; and a chest x-ray examination for the detection of possible melanoma metastasis. Whenever metastasis was suspected on the basis of these evaluations or clinical judgment, the protocol recommended confirmation by at least 1 diagnostic test, including computed tomographic scans, other imaging scans, and/or biopsy or cytologic tests. When a diagnosis of melanoma metastasis was made, a report of the diagnosis was submitted to the COMS Coordinating Center (Baltimore, Md), along with reports of all procedures that were performed to confirm the diagnosis. Biopsy or cytology slides or blocks of embedded tissue were sent to the COMS Pathology Center for histopathologic review.
The COMS Mortality Coding Committee (MCC) was established to review the clinical and pathologic information for deceased patients. This review determined the status of tumor presence and site(s) of spread at the time of a patient's death, documented the basis for such conclusions, and provided the level of certainty regarding the determination. The MCC consisted of 6 members: a general anatomical pathologist, a medical oncologist, an ocular pathologist, a biostatistician, an epidemiologist, and a coordinator from the coordinating center. The MCC met twice annually to review records, which included central histopathologic reviews and other clinical and histopathologic information. Before MCC meetings, both MCC pathologists independently reviewed each specimen sent to the COMS Pathology Center (central pathology) and assigned it to 1 of the following categories on a form submitted to the COMS Coordinating Center: diagnostic of metastatic malignant melanoma, consistent with metastatic malignant melanoma, not consistent with metastatic malignant melanoma, no tumor present, or tumor present but insufficient for diagnosis.
During MCC meetings, each patient's status at death was coded on a standard form20 as follows: (1) dead with melanoma metastasis(confirmed metastasis); (2) malignant tumor present, primary uncertain (suspected metastasis); (3) malignant tumor present, not metastatic melanoma; (4) no evidence of malignancy; or (5) inadequate evidence to establish presence of malignancy. If status was coded as either of the first 2 categories, then all metastatic sites and types of documentation also were recorded on the form.
The level of certainty associated with the coding was based on the availability of central or local pathology review of biopsy or autopsy materials. One of 3 categories was assigned to each death review: (1) confirmation of diagnosis of metastatic melanoma from central pathology review of autopsy or biopsy materials (high level of certainty), (2) tissue diagnosis of metastatic melanoma made by a local pathologist (moderate level of certainty), or (3) no tissue diagnosis available either locally or centrally (low level of certainty). A case was coded as dead with melanoma metastasis or malignant tumor, primary uncertain, only with histopathologic confirmation (either central or local). A more detailed description of COMS mortality coding procedures has been given elsewhere.21
The primary outcome of interest for this analysis was status at death by the level of certainty associated with coding. Secondary outcomes of interest were sites of metastasis, autopsy status, and availability of clinical and histopathologic materials. SAS software (SAS Institute Inc, Cary, NC) was used for all statistical analysis.
The COMS protocol and consent form were reviewed and approved by the institutional review boards of all 43 participating clinical centers before recruitment, and informed consent was obtained from all patients. Details of the large tumor study design, methods, and baseline characteristics have been reported elsewhere.18,22,23 Patient accrual began on November 12, 1986, and ended on December 15, 1994; initial mortality findings were published in 199823 and were based on follow-up as of July 31, 1997. This report focuses on patients in the large tumor trial who had died and for whom mortality codings had been performed by July 31, 1997.
A total of 1003 patients were enrolled in the COMS large tumor trial. Patients were predominantly white (97%), 57% were male, and the mean age of patients at the time of enrollment was approximately 60 years. Approximately 64% of patients were diagnosed as having metastasis within 6 years of enrollment; the estimated median time from diagnosis of metastasis to death was slightly greater than 6 months. A total of 457 patient deaths had been reported by July 31, 1997,23 and the estimated median survival from time of enrollment was 7.4 years. Of these 457 deaths, 435 patient files(95%) had been reviewed by the MCC. Cases were coded as soon as available information had been assembled, resulting in varying times to coding, ie, within 6 months (146 [33%]), 6 to 12 months (164 [38%]), 12 to 24 months (69[16%]), and more than 24 months (56 [13%]) after the patient's date of death. Status at death by level of certainty is shown in Table 1. Slides were reviewed either both centrally and locally(high level of certainty) (n = 259) or locally only (moderate level) (n = 39) in 298 cases; no tissue diagnosis (low level) was available for 137 cases(31%). Overall, melanoma metastasis was confirmed (histopathologically confirmed metastasis) in 269 cases (62%) and suspected (malignant tumor present but uncertain primary tumor) in an additional 92 cases (21%), for a total of 361 cases. A malignant tumor other than choroidal melanoma was the cause of death in 22 cases (5%). There was no evidence of metastasis or another cancer in 38 cases (9%) for whom sufficient clinical information was available; 14 cases(3%) lacked adequate clinical information for definitive coding. Of 259 cases with histopathologic or cytologic slides or specimens reviewed centrally, 93% were coded as dead with histologically confirmed melanoma metastasis; in contrast, 74% of the 39 cases with local review only were coded as such(data not shown). By definition, none of the 137 cases without tissue diagnosis could be coded as dead with melanoma metastasis.
The autopsy rate was less than 10% in most of the 43 clinical centers; only 2 centers had autopsy rates exceeding 20%: 25% and 50% in centers reporting 12 and 4 deaths, respectively. Sixteen centers had autopsy information for at least 1 death. University centers had higher autopsy rates than other types of centers. Overall, autopsy results were available for only 28 (6%) of the 435 patients: 26 (7%) of the 361 patients with either histopathologically confirmed or suspected metastasis and 2 (3%) of the 74 patients without metastasis. Table 2 depicts the distribution of these 361 cases by autopsy status and the substantially larger proportion of histopathologically confirmed deaths with melanoma metastasis of patients with autopsies.
Table 3 shows the distribution of coded sites of metastasis, based on available information, by autopsy status of patients. The most common sites were liver (93%), lung (24%), bone (16%), skin or subcutaneous tissue (11%), and lymph nodes (10%). In cases with autopsies, liver (100%) and lung (50%) involvement exceeded that in cases without autopsies(93% and 22%, respectively). Multiple sites of metastasis were prevalent (Table 4). For the 361 patients with confirmed or suspected metastasis, 204 (57%) had evidence of metastasis in a single site, 73 (20%) in 2 sites, 47 (13%) in 3 sites, and 37 (10%) in 4 or more sites. Of the 204 cases with only 1 site of metastasis, the liver was involved in 193 patients (95%) (data not shown). Of the 336 patients with liver as a site of recurrence, only 193 (57%) had liver as the only site (Table 5).
Many types of records were reviewed by the MCC to make their assessments(Table 6). Official certification of death was available for nearly all patients. A physical examination report or clinical summary, liver function results, and computed tomographic scan reports close to the time of death were available in 74%, 68%, and 62% of the cases, respectively. X-ray reports and liver biopsy reports were available less often, ie, in 38% and 32% of the cases, respectively. Multiple biopsy or scan reports were often available for the same patient. As expected, greater availability of clinical and pathologic materials for review resulted in more definitive codings of status at death and sites of metastasis. The availability of materials for review and the distribution of codings did not differ substantially by treatment arm.
Initial mortality comparisons indicated no statistically significant difference in overall survival by treatment.23 The Kaplan-Meier estimate of 5-year all-cause mortality was 40%.23 On the basis of the histopathologically confirmed deaths from metastatic melanoma, the estimate of 5-year melanoma-related mortality was 27%,23 an arithmetic difference of 13% between tumor-related and all-cause mortality rates in this clinical trial. A less rigid definition of death with melanoma metastasis, ie, inclusion of malignant tumors with uncertain primary origin, would decrease this difference further to 5%.
The COMS large tumor trial has provided the largest reported series of patients with metastatic ocular melanoma and the only series resulting from a randomized controlled clinical trial with standardized follow-up and systematic data collection and mortality coding. The assessment of disease-related mortality for any condition depends on available clinical and histopathologic information and the criteria established by the coders. In the COMS clinical trial for large choroidal melanoma, regularly scheduled follow-up examination and systematic collection and review of all records pertinent to death and events preceding death permitted us to categorize accurately patient status at time of death with respect to metastasis after a diagnosis of choroidal melanoma. The mortality findings of our study have high external validity and should be applicable to all patients with large choroidal melanoma who satisfy the COMS eligibility criteria.22 In our study, 62% of deaths were associated with histopathologically confirmed metastasis; metastasis was suspected in the absence of tissue confirmation in an additional 21%. Overall, 43% of patients with confirmed or suspected metastasis had more than a single site of involvement; 20%, 13%, and 10% had 2, 3, or 4 or more sites of involvement, respectively. One unusual case of solitary cardiac metastasis was observed.24 We observed, however, when autopsy information was available, that 77% of patients had more than 1 site of metastasis; 39% had 4 or more sites. Although our autopsy rate of 6% is low, it is comparable to current estimates.25
Other investigators have described patterns of metastatic spread in smaller series of patients with choroidal melanoma (Table 7). Although, as observed in our study, the liver is the most frequent site of metastasis, the sites and extent of dissemination vary across case series, reflecting differences in patient populations, study designs, diagnostic procedures, and data collection over time and across studies. In other series, the proportion of cases with liver metastasis ranged from 56% in cases diagnosed by biopsy or clinical criteria to 100% in autopsy cases(Table 7). In our series, the liver was a site of metastasis in 93% of total deaths with metastasis and in 100% of the 26 cases with available autopsy information. Although an annual chest x-ray examination was required in the COMS, the relative frequency of diagnosis of lung metastasis was 24% (50% in autopsy cases); this proportion varied from 20% to 55% in the other series. Our findings are similar to those of the series by Rajpal et al30 of 35 cases(33 with autopsy); the variation may be due to differences in the patient populations. The series by Rajpal et al and other series consisted of small numbers of patients who were referred for treatment of metastatic disease and were not necessarily representative of all patients with metastasis. Of patients with confirmed or suspected metastasis at time of death in our series, fewer than 40% were reported by the clinical centers to have received treatment for metastatic disease at the time of diagnosis (data not shown). As of our cutoff date for this analysis, most of such patients had received chemotherapy. We will investigate the association between treatment and patterns of metastasis as our trial progresses.
The progression of choroidal melanoma and its propensity to metastasize to the liver, confirmed in this study, differ from progression and metastatic spread of cutaneous melanoma, which is primarily to the lymph nodes and lung.2,34,35 Our findings suggest guidelines for evaluation of future patients with ocular melanoma. The COMS protocol specifies annual follow-up, including a routine medical examination, liver function studies, and chest x-ray examination. Any suggestion of metastasis based on these evaluations is followed by diagnostic procedures, including computed tomographic scan or other imaging procedures and, if indicated, a biopsy.36 These steps follow the recommendations of Pach and Robertson.37 A recent assessment by Eskelin et al38 supports this approach at baseline and follow-up. However, Hicks et al39 concluded that all single liver function tests had poor sensitivity for metastatic disease and recommended routine liver ultrasonography at both presentation and follow-up. New genetic, histopathologic,40,41 metabolic, and molecular biologic predictors of metastasis may be more useful prognostic indicators than screening or imaging tests.
Once metastatic ocular melanoma is detected, no effective method of systemic treatment has been identified,42- 44 although one small series of patients with curative resection of metastasis exhibited improved survival.45 Treatment of only the liver by surgical resection or chemoembolization probably has not been effective because, as demonstrated in our patients, tumor dissemination elsewhere is also likely. The high rate of metastasis of choroidal melanoma, despite treatment of the primary tumor, and the short time from the clinical diagnosis of metastasis to death from metastatic melanoma32 suggest that adjuvant systemic treatment at the time of initial treatment of the choroidal melanoma may prove a more effective approach to reducing death due to melanoma metastasis. Identification of promising adjuvant therapies for metastatic melanoma of the choroid and evaluation in randomized controlled clinical trials are needed.
Accepted for publication December 15, 2000.
The COMS has received support from the National Eye Institute and the National Cancer Institute through cooperative agreements EY06253, EY06257, EY06258, EY06259, EY6260, EY06264, EY06265, EY06266, EY06268, EY06269, EY06270, EY06274, EY06275, EY06276, EY06279, EY06280, EY06282, EY06283, EY06284, EY06287, EY06288, EY06289, EY06291, EY06839, EY06843, EY06844, EY06848, EY06858, and EY06899 with the National Institutes of Health, Bethesda, Md.
Nancy Robinson, BA, and Michele Zimbric, BS, coordinated central review of autopsy, biopsy, and cytology materials by the MCC pathologists.
Corresponding author: Marie Diener-West, PhD, COMS Coordinating Center, Wilmer Clinical Trials and Biometry, 550 N Broadway, Ninth Floor, Baltimore, MD 21205 (e-mail: firstname.lastname@example.org).
Reprints: COMS Coordinating Center, Wilmer Clinical Trials and Biometry, 550 N Broadway, Ninth Floor, Baltimore, MD 21205.