Population-Based Analysis of Prognostic Factors and Survival in Familial Melanoma
http://www.100md.com
《临床肿瘤学》
the Departments of Dermatology, Oncological Sciences, Medical Informatics and Internal Medicine, Melanoma Program, Huntsman Cancer Institute, University of Utah, Salt Lake City, UT
ABSTRACT
PURPOSE: Familial melanoma patients are reported to present with thinner melanomas, to be younger at the time of diagnosis, and to have a greater likelihood of developing multiple primary tumors. We sought to determine whether melanomas that occur in a familial setting demonstrate different prognostic and survival statistics relative to sporadic melanoma.
PATIENTS AND METHODS: This population-based study used the Utah Cancer Registry and Utah Population Database to objectively evaluate prognostic and survival statistics of the familial melanoma population. From 1973 to 1999, there were 7,785 cases of invasive melanoma identified through the Utah Cancer Registry. These were linked to the Utah Population Database, resulting in 2,659 subjects with family-history information from which a familiality score could be calculated. Cases scored in the top ninth percentile were assigned as high familial risk, and the remaining 91% were considered low familial risk.
RESULTS: Multivariate logistic-regression analysis found no association between sex, Breslow depth, Clark level, or survival and the familial status. Age at first diagnosis of invasive melanoma was slightly lower in the high-familial-risk group (57 v 60 years; P = .03). High-familial-risk subjects had more melanomas diagnosed at age 30 or younger (12% v 6%; P < .001). A significant difference in the overall number of individuals with two or more primary malignant melanomas was not detected among the groups (P = .2).
CONCLUSION: These data suggest that melanomas occurring in the context of an underlying inherited susceptibility do not have a significantly different biologic behavior.
INTRODUCTION
Familial melanoma can be defined as an excess clustering of melanoma within families beyond that predicted based on shared lifestyles, similar environmental exposures, and the likelihood of sporadic melanoma. Numerous reports have documented such clustering,1-9 and germline mutations in three genes (CDKN2A, CDK4, and ARF) have been shown to cosegregate with disease in these families.10-15 Despite expanding knowledge about the genetics of cutaneous melanoma, there have been few attempts to define clinically relevant differences between familial and sporadic melanomas.
Among the few publications that have compared familial and sporadic melanoma, most investigations have concluded that hereditary melanomas occur in younger individuals and that the incidence of multiple primary melanomas is higher.7,9,16-19 However, even these findings have been debated. It has been reported also that familial melanoma, in contrast with sporadic melanoma, tends to occur in a random distribution, with less predominance on photoexposed areas of the body, and has improved survival.9,16 The discrepancies between these studies are likely caused by the difficulty of ascertainment that typically relies heavily on patient-reported family history. This study takes advantage of a unique genealogical database that permits an objective assignment of melanoma populations without ascertainment bias.
The Utah Population Database (Huntsman Cancer Institute, University of Utah, Salt Lake City, UT) is a University of Utah research resource that contains more than 7 million individual records from the Family History Library maintained by the Church of Jesus Christ of Latter-day Saints (Salt Lake City, UT), vital records from the Utah State Department of Health, and other statewide data sets. The resource includes family-history data on more than 3 million individuals, representing pedigrees that span up to 10 generations. The majority of families living in Utah are represented in this database, which has a special emphasis on genealogy records of the founders of Utah and their Utah descendants. Extensive investigation of the Utah population reveals that it is a noninbred population representative of the white population in the United States (predominately Northern European origin).20,21 The representative nature of the Utah Population Database can be explained by a variety of factors, namely, the large founding size of the Mormon population, the high rates of gene flow,21 and the fact that the Utah Population Database includes individuals within and outside the Latter-day Saints Church membership. The genealogy records were computerized in the mid-1970s22 and have been linked to other Utah data sets, including cancer records, birth and death certificates, driver's license records, and census records, and are a robust source of data for studies in cancer genetics. For example, several key cancer-predisposition genes have been discovered, in part, through the use of the Utah Population Database, including melanoma (CDKN2A11,23) and breast and ovarian cancers (BRCA124 and BRCA225). In addition, the Utah Population Database has been used for population-based assessment of cancer risk in relatives of subjects with several types of cancer.26,27
In comparison, the Swedish Family-Cancer Database is a larger cancer database, because it includes cancer cases beginning in 1958 for a larger population. However, it has a different structure in that it includes only two-generation families.28 The Icelandic Cancer Registry includes cases diagnosed beginning in 1955, and these have been linked to an extensive genealogical database; however, the Icelandic population is relatively small, and there were only 618 melanoma cases available for analysis from 1955 to 2002.29
In contrast to data derived from other cancer registries in the United States that are not linked to genealogical records, Utah Population Database–derived data permit assessment of the inheritance pattern of melanoma over multiple generations within the Utah population. The Utah Population Database has an advantage over traditional recruitment strategies in that it is a population-based registry, and therefore ascertainment bias is minimized and recall bias is eliminated. In this report, we used the Utah Population Database to identify familial melanoma cases.
Many definitions of "familial" have been applied to investigations of the genetics of melanoma. Some studies have considered any family with two or more affected members as familial,9 although more stringent criteria (two or more first-degree relatives or three affected family members) have also been applied. Because it has been shown that families containing a member with multiple primary melanomas18,30-32 or families containing a melanoma and a pancreatic carcinoma33-35 have an increased frequency of germline CDKN2A mutations, these groups might also be considered "familial."34,36
Because of the difficulty in establishing clear and accurate definitions for "familial" status, University of Utah researchers have directed significant effort toward the development of a reliable method for measuring excess relative risk of a disease attributable to familial factors.37 Our current study uses this method, a familial standardized incidence ratio, which has been applied successfully to breast and colon cancer cases within the Utah Population Database in the past.38,39 The familial standardized incidence ratio permits quantification of an individual's familial risk of disease, taking into account all known biologic relatives, their degree of relatedness to the subject, and the probability that they might have the disease of interest. By applying the familial standardized incidence ratio formula to all melanomas within the Utah Population Database, we were able to make an objective assignment of each melanoma into high-familial-risk or low-familial-risk groups. We compared the two groups by performing separate multivariate survival analyses using age at diagnosis, sex, site, Breslow depth, Clark level, ulceration status, lymph-node status, and familial status as predictors. Each of these analyses was adjusted for age and sex when appropriate. The data presented here represent one of the largest comparisons of familial and nonfamilial melanoma and, to our knowledge, is the only population-based evaluation of prognostic indicators with respect to overall and melanoma-specific survival. The ultimate goal of this article is to provide outcome data to clinicians who care for high-risk melanoma populations and to offer a more accurate assessment of prognosis and survival.
PATIENTS AND METHODS
Subject Identification
The Utah Cancer Registry has been a part of the Surveillance, Epidemiology, and End Results program of the National Cancer Institute since 1973 and has adopted statewide quality-control measures to ensure that every pathology laboratory and hospital in the state is surveyed for cancer diagnoses each year. In addition, regular reminders are sent to each licensed physician in the state to report all cancers as mandated by Utah law. Any melanomas that were diagnosed out of state or were sent out of state for pathologic review might not be captured by the registry. The impact of this potential underascertainment on melanoma reporting in Utah is unknown but is unlikely to lead to a systematic bias in prognostic or survival data.
The Utah Cancer Registry records have been linked to other records in the Utah Population Database. Approximately 55% of all cancer records and more than 48% of melanoma records have a match to family-history records. However, there are several categories of individuals that are under-represented in the familial analysis: (1) short-time residents who have few or no vital events in Utah and (2) women having lower record-linking probabilities than men because of name changes. To describe the number of previous generations available for study, selected Utah birth cohorts in the Utah Population Database were analyzed in terms of their generational depth.40 For individuals born in 1920, there is a mean of 4.1 ancestor generations, with 82% having grandparent information and 55% having five or more previous generations available. For the 1950 birth cohort, there is a mean of 4.9 ancestor generations, with 79% having grandparent information and 67% having five or more previous generations. This kinship structure provides the type of information needed for examining multiple classes of relatives and was used to identify familial melanoma cases.
The Utah Cancer Registry was queried for all melanoma cases diagnosed in Utah in the years 1973-1999. For multiple primary tumors, the file contains multiple records per person. In the time period studied, the Utah Cancer Registry recorded 7,785 cases of invasive melanoma and collected available information regarding survival time in months, cause of death (death unrelated to cancer, cancer death, and melanoma-specific death), age at first diagnosis, sex, Breslow depth, and Clark level. Because these data are derived from pathology reports and medical records, the data for each of prognostic indicators are incomplete for some of the records. These 7,785 cases were linked with the Utah Population Database. For a linked record to be eligible for this analysis, there had to be ancestral or descendent records in the family-history file. This requirement resulted in 2,659 subjects that had family information. These cases then were coupled to a surgical database containing more complete prognostic variables, including tumor ulceration and lymph-node status (in those cases in which the lymph nodes were sampled). This resulted in the identification of 676 cases of melanoma with both family-history information and complete prognostic data. Figure 1 depicts these data sources in a flow chart. The institutional review board at the University of Utah Health Sciences Center approved this investigation (No. 7916-00). Permission to use the Utah Cancer Registry and Utah Population Database was overseen by an independent board (Resource for Genetic and Epidemiologic Research at the University of Utah, Salt Lake City, UT).
Familial Risk and Number of First- and Second-Degree Relatives Calculations
The familial standardized incidence ratio is calculated by counting all the cases of disease observed among each subject's biologic relatives within a set of strata defined by age, sex, and genealogical distance and then computing the expected numbers based on the total Utah Population Database population incidence. The age- and sex-specific rates are combined across strata by using the expected numbers as weights (as for the conventional standardized incidence ratio41), and observed and expected counts are combined across genealogical strata by using the kinship coefficient42 as a weight. Thus, the experience of first-degree relatives is weighted twice as heavily as that of second-degree relatives, that of second-degree relatives is weighted twice as heavily as that of third-degree relatives, and so on.37 The resulting number expresses the relative risk of a specific disease among members of a subject's extended family and has been shown to be a broadly useful predictor of cancer risk.43
The familial standardized incidence ratio for melanoma is derived from the complete melanoma-risk experience of all observable biologic relatives, adjusted for age, sex, number, and shared inheritance with the subject.37 We used an automated search of the linked data from the Utah Cancer Registry and the Utah Population Database to calculate the familial standardized incidence ratio and to find all the relatives of each subject with melanoma. As described previously,37 we used a pedigree-search algorithm to determine the total number of relatives per subject with melanoma.
The familial standardized incidence ratio is a weighted ratio of the "observed" to "expected" incidence in the relatives, with expected incidence calculated from the population as a whole, by using a standard Poisson regression model. Thus, a familial standardized incidence ratio of 1.0 indicates that more cases of melanoma were observed in the subject's relatives than expected. The familial standardized incidence ratio is based on a greater-than-expected clustering of melanoma cases in a given family but is unable to establish whether the family carries a high-penetrance melanoma-predisposition gene. In fact, this method identifies melanoma clustering because of a combination of various genetic and environmental risks. One main limitation of the familial standardized incidence ratio is that the precision varies widely from subject to subject because the family sizes vary widely. The larger the family, the more precise the estimate. The familial standardized incidence ratio is also relatively insensitive to recessive interactions between genes. For melanoma, the calculation assumes an autosomal dominant pattern of inheritance, consistent with what is known for all three melanoma-predisposition genes (CDKN2A, CDK4, and ARF10-15). As we continue to evaluate each of these melanoma pedigrees individually, our understanding of the familial standardized incidence ratio model will be refined further.
The 2,659 Utah Population Database–linked melanomas were assigned a familial standardized incidence ratio score37 and ranked by this value. To rule out the possibility that our estimate of the appropriate cutoff for high familial risk was too high or low (and might thereby decrease our ability to detect differences between the groups), we also compared the groups having the highest scores (top 9%) with those having the lowest scores (bottom 9%). There was no difference in the conclusions drawn from these subset comparisons.
Statistical Analyses
We performed all statistical analyses by using either Statistica 6.0 (Statsoft, Tulsa, OK, 1995) or StatXact 5.0.3 (Cytel Software Corp, Cambridge, MA) statistical software. We used the Mann-Whitney U test to test for a difference in the age distribution between the high- and low-familial-risk groups. We used Fisher's exact test to test for differences in the distribution of categoric variables (such as sex, proportion of subjects diagnosed with melanoma 30 years of age, distribution of the number of melanomas, melanoma site, Clark level, presence of nodal involvement, and Breslow depth). Survival curves were estimated by using the Kaplan-Meier product-limit method. Comparisons of survival by familial indicator within Breslow-depth strata were performed by using the log-rank test. The proportional-hazards survival model was used to test for interaction between various prognostic variables (such as nodal involvement and Clark level) and familial risk. Age and sex were included as predictors in the proportional-hazards models.
RESULTS
Family History and Data-Set Definitions
The 2,659 subjects diagnosed with at least one melanoma were assigned a familial standardized incidence ratio; 241 (9%) were designated as "high familial risk" (absolute familial standardized incidence ratio score 1.0), and 2,418 (91%) were designated as "low familial risk" (familial standardized incidence ratio < 1.0). Melanomas occurring in the upper 9% of the 2,659 subjects were considered familial, and the remaining 91% were considered nonfamilial, which is in keeping with the 5% to 12% of melanomas reported in the literature to be familial.1,18,44-47 The distribution of familial standardized incidence ratio is presented in Table 1. The subset of 676 cases of melanoma with complete prognostic information (of the 2,659 cases of melanoma with linked information to Utah Population Database) was analyzed separately (Fig 1). Of these 676 melanomas, 55 (8%) were designated high familial risk, and 621 (92%) were considered low familial risk.
Descriptive Statistics of Data Sources
To confirm that our three data sets were similar and comparable, the mean age at diagnosis of the first melanoma, percentage of male and female subjects, and depth of tumor invasion (Breslow thickness) were calculated for each data set; the results are summarized in Table 2. The three main database sources included the Utah Cancer Registry, the Utah Cancer Registry linked with the Utah Population Database, and a surgical database. These three data sources were similar with respect to the variables described above.
Number of First- and Second-Degree Relatives in the High- and Low-Familial-Risk Groups
One method of measuring familial aggregation is to determine the number of affected first-degree (N1aff) and second-degree (N2aff) relatives. This is clinically relevant because it has been our experience that most patients are uncertain about melanoma family history past the second degree. As seen in Table 3, low-familial-risk subjects had a mean of 0.06 N1aff and 0.08 N2aff relatives, with a mean of 8.6 total first-degree relatives (N1) and 24.1 total second-degree relatives (N2). The high-familial-risk subjects had a mean of 0.30 N1aff and 0.12 N2aff relatives, with a mean of 9.1 N1 relatives and 26.4 N2 relatives. Thus, despite the similarity in the number of relatives in the high- and low-familial-risk groups, there are more cases of melanoma in the first- and second-degree relatives of the subjects designated as high familial risk. These data are consistent with an appropriate assignment of familial status by the familial standardized incidence ratio scoring method and also consistent with the fact that the familial standardized incidence ratio was found to be a better predictor of cancer risk than N1aff in a similar study of breast cancer subjects.48
Number of Primary Melanomas
The 241 subjects in the high-familial-risk group had 257 melanomas (average, 1.07 per person); and 2,418 subjects in the low-familial-risk group had 2,506 melanomas (average, 1.04 per person). A significant difference in overall number of individuals with two or more primary melanomas was not detected among the high- and low-familial-risk groups (P = .2; Table 4) despite a body of literature suggesting the contrary.9,16,18,19,45,47 These data suggest that the risk for sporadic melanoma patients to develop a second primary melanoma is similar to individuals with familial melanoma. This conclusion is in keeping with the known increased relative risk for melanoma in patients with a personal history of the disease.49
Age at Diagnosis
The age of first diagnosis of invasive melanoma was slightly lower in the high-familial-risk subjects (age, 57 years for the high-familial-risk group v 60 years for low-familial-risk group; P = .029). High-familial-risk subjects had a larger proportion of invasive melanoma diagnosed at age 30 years or younger (12% for high familial risk v 6% for low familial risk; P = < .001), as shown in Table 5. The odds ratio of high familial risk given an age of 30 years or younger at the time of initial diagnosis is 2.0 (95% CI, 1.3 to 3.1; P = < .001).
Survival and Familiality
High-familial-risk status was not identified as an independent prognostic indicator by the Cox proportional-hazards model (Table 6). However, differences between the high- and low-familial-risk groups might be observed for other prognostic variables. Therefore, the question of association between high-familial-risk melanoma and the selected prognostic variables was addressed further. Multivariate regression analysis evaluating the effect of age, sex, site, Breslow depth, Clark level, year of diagnosis, ulceration status, lymph-node status, and familial status on overall and melanoma-specific survival was performed.
Our data confirmed that age, sex, Breslow depth, Clark level, and nodal involvement, but not familial-risk status, were important independent prognostic indicators for overall and melanoma-specific survival in both our high- and low-familial-risk populations. When the influence of familial status on the various prognostic indicators was considered, the additional risk conferred by familiality was not significant. Hazard ratios (relative risk) for overall and melanoma-specific survival in the entire group of Utah Population Database melanoma patients versus the high-familial-risk patients did not differ (Tables 6 and 7). Thus, despite the earlier age of onset of melanoma in the high-familial-risk population, there is no apparent change in prognosis when adjusted for tumor thickness. Tables 8 and 9 present all-causes mortality and melanoma-specific mortality.
We also considered the effect of familial risk as a continuous predictor rather than arbitrarily choosing a specific cutoff value. When the analysis was performed in this manner, the hazard ratio (per unit familial standardized incidence ratio; N = 2,659) for overall survival was 0.94 (P = .57) and for melanoma-specific survival was 1.2 (P = .34). When specific prognostic indicators are considered with familial risk as a continuous predictor, the additional risk conferred by familiality was not significant for any of the prognostic indicators (data not shown). Finally, the high-familial-risk group was compared with a "very-low-familial-risk group" consisting of the 244 subjects with a familial standardized incidence ratio of less than 0.12 (approximately 9% of the subjects). Again, no significant difference was identified in overall or melanoma-specific survival between the high- and very-low-familial-risk groups (hazard ratios for overall and melanoma-specific survival, 0.79 [P = .12] and 1.004 [P = .99], respectively).
Other Survival Data
A recent publication by Doubrovsky and Menzies50 suggests that subjects with three or more primary melanomas have improved survival. A similar evaluation of our data set failed to show a significant improvement in survival in the subjects with two or more primary lesions (P = .36; n = 94). However, when the analysis was done for three or more primary lesions, there was a tendency toward improved survival, although this did not reach statistical significance (P = .87; n = 9). As mentioned previously, age is an independent prognostic variable, with younger age at diagnosis ( 30 years) being associated with improved survival (P = .02); however, no additional survival benefit was seen when familial risk was considered.
DISCUSSION
This study represents the first and largest population-based assessment of the prognostic and survival statistics of familial versus sporadic melanoma. The identification of differences in the biologic behavior of familial and nonfamilial melanoma is of potential clinical importance and might provide meaningful insight into the pathogenesis of the disease. Several factors have made it difficult to analyze differences between familial and sporadic melanomas. Accurate identification of subjects with a true family history of melanoma versus a sporadic melanoma is difficult and has often relied on self-reported family histories, usually from the affected patient. This is problematic because (1) subjects rarely know the medical history of relatives more distant than second degree,37 (2) subjects confuse married-in family members with biologic relatives,37 (3) affected individuals may be more likely to report cancer among relatives than controls with the same family history,37,51 (4) many patients do not understand the difference between melanoma and nonmelanoma skin cancer and may incorrectly report a positive family history, and (5) ascertainment bias is inherent in any study in which recruitment is based on self-report or physician referral, because these sources cannot account for all melanomas within a population. For these reasons, it is difficult to be confident in the assignment of an individual to a sporadic or familial group. The population-based design of this study using the Utah Population Database minimizes these difficulties and biases.
A few large, population-based studies of familial cancer have demonstrated excessive risk of cancer among family members having close (first- or second-degree) relatives with the neoplasm of interest.26,52,53 The data presented here represent the first study of familial melanoma in which the collective influence of close and distant relatives was used to assess the biologic behavior of melanoma in familial and sporadic settings.
In this study, subjects with invasive malignant melanoma were classified with respect to familial risk by the familial standardized incidence ratio, derived from the complete risk experience of all observable biologic relatives.37,48 Use of the familial standardized incidence ratio requires a large genealogical database with links to population-based disease registries,37 resources readily available for the study of malignant melanoma occurring in Utah. As such, this investigation provides an objective, population-based, comprehensive comparison of melanoma occurring in high- and low-familial-risk settings.
Because the familial standardized incidence ratio is a continuous variable, a cutoff value was established to assign high- and low-familial-risk subjects. As with any cutoff, it is possible that a few melanoma subjects are misclassified, but the familial standardized incidence ratio 1.0 classified 9% of the subjects as high familial risk, in keeping with previous estimates of the proportion of melanomas reported to have a familial component (5% to 12%).1,18,44-47 Because the familial standardized incidence ratio is weighted more heavily for closer relatives, high familial standardized incidence ratio values ( 1.0) can be seen in (1) large families with multiple occurrences of malignant melanoma, even among distant relatives, and (2) small families with one or few affected relatives. The number of affected first- and second-degree relatives is a straightforward measure of familial aggregation and forms the basis of most familial studies. The significantly larger number of affected first- and second-degree relatives seen in the high-familial-risk group directly correlates with the familial standardized incidence ratio, as expected, because the calculation incorporates this count in addition to all other observable biologic relatives. Our analysis suggests that the biologic behavior of familial melanoma is not significantly different from nonfamilial melanoma with regard to sex, Breslow depth, Clark level, distribution, ulceration status, or lymph-node status. Moreover, our conclusions did not change when the data were analyzed in two other ways: (1) using familial risk as a continuous predictor rather than establishing a specific cutoff value and (2) comparing the high-familial-risk group (upper 9% of all familial standardized incidence ratio values) with a very-low-familial-risk group (lowest 9% of all familial standardized incidence ratio values).
Barnhill et al1 reported an older average age at diagnosis among a small number of familial melanoma subjects. Other authors have reported a younger age at diagnosis in familial subjects that ranged in age from an average of 33 to 47 years.7,16,17,19,51 In our analysis, the average age at first diagnosis was slightly lower in the familial melanoma group (57 v 60 years), primarily because of a higher proportion of subjects with melanoma diagnosed at age 30 years or younger in the high-familial-risk group. These data strongly suggest that melanoma in a familial setting may have an earlier biologic onset.
Balch et al54 have published a large, comprehensive analysis of melanoma in the United States that demonstrated that overall and disease-specific survival in melanoma patients was significantly influenced by well-known prognostic factors including subject age, sex, depth of tumor invasion (both the Breslow measurement and Clark level), and lymph-node metastasis. We have confirmed that the Utah melanoma population demonstrates similar prognostic and survival characteristics and that the familial subset of this population is no different. Thus, although familiality is the single greatest risk factor for development of melanoma, it is not an independent prognostic indicator and does not seem to influence the overall or melanoma-specific survival.
In summary, we observed that melanomas within a high-familial-risk setting are diagnosed at a slightly younger age, and a larger proportion are diagnosed at or before the age of 30 years, suggesting that familial melanoma has an earlier onset than sporadic melanoma. This earlier age of onset is not associated with a difference in survival but does confer a two-fold risk of having a hereditary pattern of melanoma. The overall number of primary melanomas was similar in the high- and low-familial-risk groups. Although the number of cases was small, it was noted that subjects with three or more primary melanomas had a five-fold increased risk of having an inherited pattern of melanoma. Taken together, our data imply that the biologic behavior of melanoma is the same regardless of whether it occurs in a familial or sporadic setting and suggests that the familial melanoma disease model will be more broadly applicable to sporadic melanoma.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Acknowledgment
We thank the Huntsman Cancer Foundation for database support provided to the Pedigree and Population Resource of Huntsman Cancer Institute.
NOTES
Supported by the Dermatology Foundation Leaders Society Dermatologist Investigator Research Fellowship and Clinical Career Development Award (S.R.F.), National Institutes of Health Grants No. K23 RR17525-01 (S.R.F.) and CA102422 (L.A.C.-A.), the Doris Duke Charitable Foundation (S.A.L.), Fellowship-to-Faculty Transition Award from the University of Utah, funded in part by the Howard Hughes Medical Institute (S.A.L.), the Huntsman Cancer Foundation (S.A.L.), the Tom C. Mathews Jr Familial Melanoma Research Clinic at Huntsman Cancer Institute, Huntsman General Clinical Research Center Public Health Service Grant No. MO1 RR00064, National Cancer Institute Cancer Center support Grant No. 5P30CA420-14, and the Utah Cancer Registry, funded by Contract No. NCI-CN-67000 from the National Cancer Institute, with additional support from the Utah Department of Health and the University of Utah.
Presented in part at the 2nd International Melanoma Research Congress, Phoenix, AZ, November 13-16, 2004.
The current affiliation for C.W. is the New Mexico Tumor Registry, University of New Mexico, Albuquerque, NM.
The current affiliation for A.T. is the Division of Biostatistics, Department of Epidemiology and Preventive Medicine, University of California at Davis, Davis, CA.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Barnhill RL, Roush GC, Titus-Ernstoff L, et al: Comparison of nonfamilial and familial melanoma. Dermatology 184:2-7, 1992
Clark WH Jr, Reimer RR, Greene M, et al: Origin of familial malignant melanomas from heritable melanocytic lesions: The B-K mole syndrome. Arch Dermatol 114:732-738, 1978
Greene MH, Goldin LR, Clark WH Jr, et al: Familial cutaneous malignant melanoma: Autosomal dominant trait possibly linked to the Rh locus. Proc Natl Acad Sci U S A 80:6071-6075, 1983
Greene MH, Sanders RJ, Chu FC, et al: The familial occurrence of cutaneous melanoma, intraocular melanoma, and the dysplastic nevus syndrome. Am J Ophthalmol 96:238-245, 1983
Greene MH, Clark WH Jr, Tucker MA, et al: High risk of malignant melanoma in melanoma-prone families with dysplastic nevi. Ann Intern Med 102:458-465, 1985
Kopf AW, Mintzis M, Grier RN, et al: Familial malignant melanoma. Cutis 17:873-876, 1976
Kopf AW, Hellman LJ, Rogers GS, et al: Familial malignant melanoma. JAMA 256:1915-1919, 1986
Reimer RR, Clark WH Jr, Greene MH, et al: Precursor lesions in familial melanoma: A new genetic preneoplastic syndrome. JAMA 239:744-746, 1978
Wallace DC, Beardmore GL, Exton LA: Familial malignant melanoma. Ann Surg 177:15-20, 1973
Hussussian CJ, Struewing JP, Goldstein AM, et al: Germline p16 mutations in familial melanoma. Nat Genet 8:15-21, 1994
Kamb A, Shattuck-Eidens D, Eeles R, et al: Analysis of the p16 gene (CDKN2) as a candidate for the chromosome 9p melanoma susceptibility locus. Nat Genet 8:23-26, 1994
Randerson-Moor JA, Harland M, Williams S, et al: A germline deletion of p14(ARF) but not CDKN2A in a melanoma-neural system tumour syndrome family. Hum Mol Genet 10:55-62, 2001
Rizos H, Darmanian AP, Holland EA, et al: Mutations in the INK4a/ARF melanoma susceptibility locus functionally impair p14ARF. J Biol Chem 276:41424-41434, 2001
Zuo L, Weger J, Yang Q, et al: Germline mutations in the p16INK4a binding domain of CDK4 in familial melanoma. Nat Genet 12:97-99, 1996
Rizos H, Puig S, Badenas C, et al: A melanoma-associated germline mutation in exon 1beta inactivates p14ARF. Oncogene 20:5543-5547, 2001
Anderson DE: Clinical characteristics of the genetic variety of cutaneous melanoma in man. Cancer 28:721-725, 1971
Goldstein AM, Fraser MC, Clark WH Jr, et al: Age at diagnosis and transmission of invasive melanoma in 23 families with cutaneous malignant melanoma/dysplastic nevi. J Natl Cancer Inst 86:1385-1390, 1994
Burden AD, Newell J, Andrew N, et al: Genetic and environmental influences in the development of multiple primary melanoma. Arch Dermatol 135:261-265, 1999
Anderson DE, Smith JL Jr, McBride CM: Hereditary aspects of malignant melanoma. JAMA 200:741-746, 1967
Jorde LB: Consanguinity and prereproductive mortality in the Utah Mormon population. Hum Hered 52:61-65, 2001
McLellan T, Jorde LB, Skolnick MH: Genetic distances between the Utah Mormons and related populations. Am J Hum Genet 36:836-857, 1984
Skolnick MH: The Utah genealogical database: A resource for genetic epidemiology, in MH Skonick (ed): Banbury Report No. 4: Cancer Incidence in Defined Populations. New York, NY, Cold Spring Harbor Laboratory, 1980, pp 285-297
Cannon-Albright LA, Goldgar DE, Meyer LJ, et al: Assignment of a locus for familial melanoma, MLM, to chromosome 9p13-p22. Science 258:1148-1152, 1992
Skolnick MH, Cannon-Albright LA, Goldgar DE, et al: Inheritance of proliferative breast disease in breast cancer kindreds. Science 250:1715-1720, 1990
Wooster R, Neuhausen SL, Mangion J, et al: Localization of a breast cancer susceptibility gene, BRCA2, to chromosome 13q12-13. Science 265:2088-2090, 1994
Goldgar DE, Easton DF, Cannon-Albright LA, et al: Systematic population-based assessment of cancer risk in first-degree relatives of cancer probands. J Natl Cancer Inst 86:1600-1608, 1994
Cannon-Albright LA, Thomas A, Goldgar DE, et al: Familiality of cancer in Utah. Cancer Res 54:2378-2385, 1994
Hemminki K, Li X, Plna K, et al: The nation-wide Swedish family-cancer database—updated structure and familial rates. Acta Oncol 40:772-777, 2001
Amundadottir LT, Thorvaldsson S, Gudbjartsson DF, et al: Cancer as a complex phenotype: Pattern of cancer distribution within and beyond the nuclear family. PLoS Med 1:e65, 2004
Burden AD, Vestey JP, Sirel JM, et al: Multiple primary melanoma: Risk factors and prognostic implications. BMJ 309:375, 1994
Titus-Ernstoff L, Duray PH, Ernstoff MS, et al: Dysplastic nevi in association with multiple primary melanoma. Cancer Res 48:1016-1018, 1988
Tucker MA, Boice JD Jr, Hoffman DA: Second cancer following cutaneous melanoma and cancers of the brain, thyroid, connective tissue, bone, and eye in Connecticut, 1935-82. Natl Cancer Inst Monogr 68:161-189, 1985
Lynch HT, Brand RE, Hogg D, et al: Phenotypic variation in eight extended CDKN2A germline mutation familial atypical multiple mole melanoma-pancreatic carcinoma-prone families: The familial atypical mole melanoma-pancreatic carcinoma syndrome. Cancer 94:84-96, 2002
Parker JF, Florell SR, Alexander A, et al: Pancreatic carcinoma surveillance in patients with familial melanoma. Arch Dermatol 139:1019-1025, 2003
Goldstein AM, Fraser MC, Struewing JP, et al: Increased risk of pancreatic cancer in melanoma-prone kindred with p16INK4 mutations. N Engl J Med 333:970-974, 1995
Hansen CB, Wadge LM, Lowstuter K, et al: Clinical germline genetic testing for melanoma. Lancet Oncol 5:314-319, 2004
Kerber RA: Method for calculating risk associated with family history of a disease. Genet Epidemiol 12:291-301, 1995
Slattery ML, Berry TD, Kerber RA: Is survival among women diagnosed with breast cancer influenced by family history of breast cancer? Epidemiology 4:543-548, 1993
Slattery ML, Kerber RA: The impact of family history of colon cancer on survival after diagnosis with colon cancer. Int J Epidemiol 24:888-896, 1995
Mineau GP, Fraser A: Kinship structure: Families past and present, in C Zick, KR Smith (eds): Utah at the Beginning of the New Millennium–A Demographic Perspective. Salt Lake City, UT, University of Utah Press, In press
Breslow NE, Day NE: Statistical methods in cancer research. Volume II—The design and analysis of cohort studies. IARC Sci Publ (82):1-406, 1987
Malecot G: The Mathematics of Heredity. Paris, France, Masson, 1948
Kerber RA, O'Brien E: A cohort study of cancer risk in relation to family histories of cancer in the Utah population database. Cancer 103:1906-1915, 2005
Goldstein AM, Tucker MA: Genetic epidemiology of familial melanoma. Dermatol Clin 13:605-612, 1995
Grange F, Chompret A, Guilloud-Bataille M, et al: Comparison between familial and nonfamilial melanoma in France. Arch Dermatol 131:1154-1159, 1995
Greene MH, Fraumeni JFJ: The hereditary variant of malignant melanoma, in WHJ Clark, LI Goldman, MJ Mastrangelo (eds): Human Malignant Melanoma. New York, NY, Grune and Stratton, 1979, pp 139-166
Wallace DC, Exton LA, McLeod GR: Genetic factor in malignant melanoma. Cancer 27:1262-1266, 1971
Boucher KM, Kerber RA: Measures of familial aggregation as predictors of breast-cancer risk. J Epidemiol Biostat 6:377-385, 2001
Kefford RF, Newton Bishop JA, Bergman W, et al: Counseling and DNA testing for individuals perceived to be genetically predisposed to melanoma: A consensus statement of the Melanoma Genetics Consortium. J Clin Oncol 17:3245-3251, 1999
Doubrovsky A, Menzies SW: Enhanced survival in patients with multiple primary melanoma. Arch Dermatol 139:1013-1018, 2003
Duke D, Castresana J, Lucchina L, et al: Familial cutaneous melanoma and two-mutational-event modeling. Cancer 72:3239-3243, 1993
Hemminki K, Zhang H, Czene K: Familial and attributable risks in cutaneous melanoma: Effects of proband and age. J Invest Dermatol 120:217-223, 2003
Hemminki K, Lonnstedt I, Vaittinen P: A population-based study of familial cutaneous melanoma. Melanoma Res 11:133-140, 2001
Balch CM, Soong SJ, Gershenwald JE, et al: Prognostic factors analysis of 17,600 melanoma patients: Validation of the American Joint Committee on Cancer melanoma staging system. J Clin Oncol 19:3622-3634, 2001(Scott R. Florell, Kenneth)
ABSTRACT
PURPOSE: Familial melanoma patients are reported to present with thinner melanomas, to be younger at the time of diagnosis, and to have a greater likelihood of developing multiple primary tumors. We sought to determine whether melanomas that occur in a familial setting demonstrate different prognostic and survival statistics relative to sporadic melanoma.
PATIENTS AND METHODS: This population-based study used the Utah Cancer Registry and Utah Population Database to objectively evaluate prognostic and survival statistics of the familial melanoma population. From 1973 to 1999, there were 7,785 cases of invasive melanoma identified through the Utah Cancer Registry. These were linked to the Utah Population Database, resulting in 2,659 subjects with family-history information from which a familiality score could be calculated. Cases scored in the top ninth percentile were assigned as high familial risk, and the remaining 91% were considered low familial risk.
RESULTS: Multivariate logistic-regression analysis found no association between sex, Breslow depth, Clark level, or survival and the familial status. Age at first diagnosis of invasive melanoma was slightly lower in the high-familial-risk group (57 v 60 years; P = .03). High-familial-risk subjects had more melanomas diagnosed at age 30 or younger (12% v 6%; P < .001). A significant difference in the overall number of individuals with two or more primary malignant melanomas was not detected among the groups (P = .2).
CONCLUSION: These data suggest that melanomas occurring in the context of an underlying inherited susceptibility do not have a significantly different biologic behavior.
INTRODUCTION
Familial melanoma can be defined as an excess clustering of melanoma within families beyond that predicted based on shared lifestyles, similar environmental exposures, and the likelihood of sporadic melanoma. Numerous reports have documented such clustering,1-9 and germline mutations in three genes (CDKN2A, CDK4, and ARF) have been shown to cosegregate with disease in these families.10-15 Despite expanding knowledge about the genetics of cutaneous melanoma, there have been few attempts to define clinically relevant differences between familial and sporadic melanomas.
Among the few publications that have compared familial and sporadic melanoma, most investigations have concluded that hereditary melanomas occur in younger individuals and that the incidence of multiple primary melanomas is higher.7,9,16-19 However, even these findings have been debated. It has been reported also that familial melanoma, in contrast with sporadic melanoma, tends to occur in a random distribution, with less predominance on photoexposed areas of the body, and has improved survival.9,16 The discrepancies between these studies are likely caused by the difficulty of ascertainment that typically relies heavily on patient-reported family history. This study takes advantage of a unique genealogical database that permits an objective assignment of melanoma populations without ascertainment bias.
The Utah Population Database (Huntsman Cancer Institute, University of Utah, Salt Lake City, UT) is a University of Utah research resource that contains more than 7 million individual records from the Family History Library maintained by the Church of Jesus Christ of Latter-day Saints (Salt Lake City, UT), vital records from the Utah State Department of Health, and other statewide data sets. The resource includes family-history data on more than 3 million individuals, representing pedigrees that span up to 10 generations. The majority of families living in Utah are represented in this database, which has a special emphasis on genealogy records of the founders of Utah and their Utah descendants. Extensive investigation of the Utah population reveals that it is a noninbred population representative of the white population in the United States (predominately Northern European origin).20,21 The representative nature of the Utah Population Database can be explained by a variety of factors, namely, the large founding size of the Mormon population, the high rates of gene flow,21 and the fact that the Utah Population Database includes individuals within and outside the Latter-day Saints Church membership. The genealogy records were computerized in the mid-1970s22 and have been linked to other Utah data sets, including cancer records, birth and death certificates, driver's license records, and census records, and are a robust source of data for studies in cancer genetics. For example, several key cancer-predisposition genes have been discovered, in part, through the use of the Utah Population Database, including melanoma (CDKN2A11,23) and breast and ovarian cancers (BRCA124 and BRCA225). In addition, the Utah Population Database has been used for population-based assessment of cancer risk in relatives of subjects with several types of cancer.26,27
In comparison, the Swedish Family-Cancer Database is a larger cancer database, because it includes cancer cases beginning in 1958 for a larger population. However, it has a different structure in that it includes only two-generation families.28 The Icelandic Cancer Registry includes cases diagnosed beginning in 1955, and these have been linked to an extensive genealogical database; however, the Icelandic population is relatively small, and there were only 618 melanoma cases available for analysis from 1955 to 2002.29
In contrast to data derived from other cancer registries in the United States that are not linked to genealogical records, Utah Population Database–derived data permit assessment of the inheritance pattern of melanoma over multiple generations within the Utah population. The Utah Population Database has an advantage over traditional recruitment strategies in that it is a population-based registry, and therefore ascertainment bias is minimized and recall bias is eliminated. In this report, we used the Utah Population Database to identify familial melanoma cases.
Many definitions of "familial" have been applied to investigations of the genetics of melanoma. Some studies have considered any family with two or more affected members as familial,9 although more stringent criteria (two or more first-degree relatives or three affected family members) have also been applied. Because it has been shown that families containing a member with multiple primary melanomas18,30-32 or families containing a melanoma and a pancreatic carcinoma33-35 have an increased frequency of germline CDKN2A mutations, these groups might also be considered "familial."34,36
Because of the difficulty in establishing clear and accurate definitions for "familial" status, University of Utah researchers have directed significant effort toward the development of a reliable method for measuring excess relative risk of a disease attributable to familial factors.37 Our current study uses this method, a familial standardized incidence ratio, which has been applied successfully to breast and colon cancer cases within the Utah Population Database in the past.38,39 The familial standardized incidence ratio permits quantification of an individual's familial risk of disease, taking into account all known biologic relatives, their degree of relatedness to the subject, and the probability that they might have the disease of interest. By applying the familial standardized incidence ratio formula to all melanomas within the Utah Population Database, we were able to make an objective assignment of each melanoma into high-familial-risk or low-familial-risk groups. We compared the two groups by performing separate multivariate survival analyses using age at diagnosis, sex, site, Breslow depth, Clark level, ulceration status, lymph-node status, and familial status as predictors. Each of these analyses was adjusted for age and sex when appropriate. The data presented here represent one of the largest comparisons of familial and nonfamilial melanoma and, to our knowledge, is the only population-based evaluation of prognostic indicators with respect to overall and melanoma-specific survival. The ultimate goal of this article is to provide outcome data to clinicians who care for high-risk melanoma populations and to offer a more accurate assessment of prognosis and survival.
PATIENTS AND METHODS
Subject Identification
The Utah Cancer Registry has been a part of the Surveillance, Epidemiology, and End Results program of the National Cancer Institute since 1973 and has adopted statewide quality-control measures to ensure that every pathology laboratory and hospital in the state is surveyed for cancer diagnoses each year. In addition, regular reminders are sent to each licensed physician in the state to report all cancers as mandated by Utah law. Any melanomas that were diagnosed out of state or were sent out of state for pathologic review might not be captured by the registry. The impact of this potential underascertainment on melanoma reporting in Utah is unknown but is unlikely to lead to a systematic bias in prognostic or survival data.
The Utah Cancer Registry records have been linked to other records in the Utah Population Database. Approximately 55% of all cancer records and more than 48% of melanoma records have a match to family-history records. However, there are several categories of individuals that are under-represented in the familial analysis: (1) short-time residents who have few or no vital events in Utah and (2) women having lower record-linking probabilities than men because of name changes. To describe the number of previous generations available for study, selected Utah birth cohorts in the Utah Population Database were analyzed in terms of their generational depth.40 For individuals born in 1920, there is a mean of 4.1 ancestor generations, with 82% having grandparent information and 55% having five or more previous generations available. For the 1950 birth cohort, there is a mean of 4.9 ancestor generations, with 79% having grandparent information and 67% having five or more previous generations. This kinship structure provides the type of information needed for examining multiple classes of relatives and was used to identify familial melanoma cases.
The Utah Cancer Registry was queried for all melanoma cases diagnosed in Utah in the years 1973-1999. For multiple primary tumors, the file contains multiple records per person. In the time period studied, the Utah Cancer Registry recorded 7,785 cases of invasive melanoma and collected available information regarding survival time in months, cause of death (death unrelated to cancer, cancer death, and melanoma-specific death), age at first diagnosis, sex, Breslow depth, and Clark level. Because these data are derived from pathology reports and medical records, the data for each of prognostic indicators are incomplete for some of the records. These 7,785 cases were linked with the Utah Population Database. For a linked record to be eligible for this analysis, there had to be ancestral or descendent records in the family-history file. This requirement resulted in 2,659 subjects that had family information. These cases then were coupled to a surgical database containing more complete prognostic variables, including tumor ulceration and lymph-node status (in those cases in which the lymph nodes were sampled). This resulted in the identification of 676 cases of melanoma with both family-history information and complete prognostic data. Figure 1 depicts these data sources in a flow chart. The institutional review board at the University of Utah Health Sciences Center approved this investigation (No. 7916-00). Permission to use the Utah Cancer Registry and Utah Population Database was overseen by an independent board (Resource for Genetic and Epidemiologic Research at the University of Utah, Salt Lake City, UT).
Familial Risk and Number of First- and Second-Degree Relatives Calculations
The familial standardized incidence ratio is calculated by counting all the cases of disease observed among each subject's biologic relatives within a set of strata defined by age, sex, and genealogical distance and then computing the expected numbers based on the total Utah Population Database population incidence. The age- and sex-specific rates are combined across strata by using the expected numbers as weights (as for the conventional standardized incidence ratio41), and observed and expected counts are combined across genealogical strata by using the kinship coefficient42 as a weight. Thus, the experience of first-degree relatives is weighted twice as heavily as that of second-degree relatives, that of second-degree relatives is weighted twice as heavily as that of third-degree relatives, and so on.37 The resulting number expresses the relative risk of a specific disease among members of a subject's extended family and has been shown to be a broadly useful predictor of cancer risk.43
The familial standardized incidence ratio for melanoma is derived from the complete melanoma-risk experience of all observable biologic relatives, adjusted for age, sex, number, and shared inheritance with the subject.37 We used an automated search of the linked data from the Utah Cancer Registry and the Utah Population Database to calculate the familial standardized incidence ratio and to find all the relatives of each subject with melanoma. As described previously,37 we used a pedigree-search algorithm to determine the total number of relatives per subject with melanoma.
The familial standardized incidence ratio is a weighted ratio of the "observed" to "expected" incidence in the relatives, with expected incidence calculated from the population as a whole, by using a standard Poisson regression model. Thus, a familial standardized incidence ratio of 1.0 indicates that more cases of melanoma were observed in the subject's relatives than expected. The familial standardized incidence ratio is based on a greater-than-expected clustering of melanoma cases in a given family but is unable to establish whether the family carries a high-penetrance melanoma-predisposition gene. In fact, this method identifies melanoma clustering because of a combination of various genetic and environmental risks. One main limitation of the familial standardized incidence ratio is that the precision varies widely from subject to subject because the family sizes vary widely. The larger the family, the more precise the estimate. The familial standardized incidence ratio is also relatively insensitive to recessive interactions between genes. For melanoma, the calculation assumes an autosomal dominant pattern of inheritance, consistent with what is known for all three melanoma-predisposition genes (CDKN2A, CDK4, and ARF10-15). As we continue to evaluate each of these melanoma pedigrees individually, our understanding of the familial standardized incidence ratio model will be refined further.
The 2,659 Utah Population Database–linked melanomas were assigned a familial standardized incidence ratio score37 and ranked by this value. To rule out the possibility that our estimate of the appropriate cutoff for high familial risk was too high or low (and might thereby decrease our ability to detect differences between the groups), we also compared the groups having the highest scores (top 9%) with those having the lowest scores (bottom 9%). There was no difference in the conclusions drawn from these subset comparisons.
Statistical Analyses
We performed all statistical analyses by using either Statistica 6.0 (Statsoft, Tulsa, OK, 1995) or StatXact 5.0.3 (Cytel Software Corp, Cambridge, MA) statistical software. We used the Mann-Whitney U test to test for a difference in the age distribution between the high- and low-familial-risk groups. We used Fisher's exact test to test for differences in the distribution of categoric variables (such as sex, proportion of subjects diagnosed with melanoma 30 years of age, distribution of the number of melanomas, melanoma site, Clark level, presence of nodal involvement, and Breslow depth). Survival curves were estimated by using the Kaplan-Meier product-limit method. Comparisons of survival by familial indicator within Breslow-depth strata were performed by using the log-rank test. The proportional-hazards survival model was used to test for interaction between various prognostic variables (such as nodal involvement and Clark level) and familial risk. Age and sex were included as predictors in the proportional-hazards models.
RESULTS
Family History and Data-Set Definitions
The 2,659 subjects diagnosed with at least one melanoma were assigned a familial standardized incidence ratio; 241 (9%) were designated as "high familial risk" (absolute familial standardized incidence ratio score 1.0), and 2,418 (91%) were designated as "low familial risk" (familial standardized incidence ratio < 1.0). Melanomas occurring in the upper 9% of the 2,659 subjects were considered familial, and the remaining 91% were considered nonfamilial, which is in keeping with the 5% to 12% of melanomas reported in the literature to be familial.1,18,44-47 The distribution of familial standardized incidence ratio is presented in Table 1. The subset of 676 cases of melanoma with complete prognostic information (of the 2,659 cases of melanoma with linked information to Utah Population Database) was analyzed separately (Fig 1). Of these 676 melanomas, 55 (8%) were designated high familial risk, and 621 (92%) were considered low familial risk.
Descriptive Statistics of Data Sources
To confirm that our three data sets were similar and comparable, the mean age at diagnosis of the first melanoma, percentage of male and female subjects, and depth of tumor invasion (Breslow thickness) were calculated for each data set; the results are summarized in Table 2. The three main database sources included the Utah Cancer Registry, the Utah Cancer Registry linked with the Utah Population Database, and a surgical database. These three data sources were similar with respect to the variables described above.
Number of First- and Second-Degree Relatives in the High- and Low-Familial-Risk Groups
One method of measuring familial aggregation is to determine the number of affected first-degree (N1aff) and second-degree (N2aff) relatives. This is clinically relevant because it has been our experience that most patients are uncertain about melanoma family history past the second degree. As seen in Table 3, low-familial-risk subjects had a mean of 0.06 N1aff and 0.08 N2aff relatives, with a mean of 8.6 total first-degree relatives (N1) and 24.1 total second-degree relatives (N2). The high-familial-risk subjects had a mean of 0.30 N1aff and 0.12 N2aff relatives, with a mean of 9.1 N1 relatives and 26.4 N2 relatives. Thus, despite the similarity in the number of relatives in the high- and low-familial-risk groups, there are more cases of melanoma in the first- and second-degree relatives of the subjects designated as high familial risk. These data are consistent with an appropriate assignment of familial status by the familial standardized incidence ratio scoring method and also consistent with the fact that the familial standardized incidence ratio was found to be a better predictor of cancer risk than N1aff in a similar study of breast cancer subjects.48
Number of Primary Melanomas
The 241 subjects in the high-familial-risk group had 257 melanomas (average, 1.07 per person); and 2,418 subjects in the low-familial-risk group had 2,506 melanomas (average, 1.04 per person). A significant difference in overall number of individuals with two or more primary melanomas was not detected among the high- and low-familial-risk groups (P = .2; Table 4) despite a body of literature suggesting the contrary.9,16,18,19,45,47 These data suggest that the risk for sporadic melanoma patients to develop a second primary melanoma is similar to individuals with familial melanoma. This conclusion is in keeping with the known increased relative risk for melanoma in patients with a personal history of the disease.49
Age at Diagnosis
The age of first diagnosis of invasive melanoma was slightly lower in the high-familial-risk subjects (age, 57 years for the high-familial-risk group v 60 years for low-familial-risk group; P = .029). High-familial-risk subjects had a larger proportion of invasive melanoma diagnosed at age 30 years or younger (12% for high familial risk v 6% for low familial risk; P = < .001), as shown in Table 5. The odds ratio of high familial risk given an age of 30 years or younger at the time of initial diagnosis is 2.0 (95% CI, 1.3 to 3.1; P = < .001).
Survival and Familiality
High-familial-risk status was not identified as an independent prognostic indicator by the Cox proportional-hazards model (Table 6). However, differences between the high- and low-familial-risk groups might be observed for other prognostic variables. Therefore, the question of association between high-familial-risk melanoma and the selected prognostic variables was addressed further. Multivariate regression analysis evaluating the effect of age, sex, site, Breslow depth, Clark level, year of diagnosis, ulceration status, lymph-node status, and familial status on overall and melanoma-specific survival was performed.
Our data confirmed that age, sex, Breslow depth, Clark level, and nodal involvement, but not familial-risk status, were important independent prognostic indicators for overall and melanoma-specific survival in both our high- and low-familial-risk populations. When the influence of familial status on the various prognostic indicators was considered, the additional risk conferred by familiality was not significant. Hazard ratios (relative risk) for overall and melanoma-specific survival in the entire group of Utah Population Database melanoma patients versus the high-familial-risk patients did not differ (Tables 6 and 7). Thus, despite the earlier age of onset of melanoma in the high-familial-risk population, there is no apparent change in prognosis when adjusted for tumor thickness. Tables 8 and 9 present all-causes mortality and melanoma-specific mortality.
We also considered the effect of familial risk as a continuous predictor rather than arbitrarily choosing a specific cutoff value. When the analysis was performed in this manner, the hazard ratio (per unit familial standardized incidence ratio; N = 2,659) for overall survival was 0.94 (P = .57) and for melanoma-specific survival was 1.2 (P = .34). When specific prognostic indicators are considered with familial risk as a continuous predictor, the additional risk conferred by familiality was not significant for any of the prognostic indicators (data not shown). Finally, the high-familial-risk group was compared with a "very-low-familial-risk group" consisting of the 244 subjects with a familial standardized incidence ratio of less than 0.12 (approximately 9% of the subjects). Again, no significant difference was identified in overall or melanoma-specific survival between the high- and very-low-familial-risk groups (hazard ratios for overall and melanoma-specific survival, 0.79 [P = .12] and 1.004 [P = .99], respectively).
Other Survival Data
A recent publication by Doubrovsky and Menzies50 suggests that subjects with three or more primary melanomas have improved survival. A similar evaluation of our data set failed to show a significant improvement in survival in the subjects with two or more primary lesions (P = .36; n = 94). However, when the analysis was done for three or more primary lesions, there was a tendency toward improved survival, although this did not reach statistical significance (P = .87; n = 9). As mentioned previously, age is an independent prognostic variable, with younger age at diagnosis ( 30 years) being associated with improved survival (P = .02); however, no additional survival benefit was seen when familial risk was considered.
DISCUSSION
This study represents the first and largest population-based assessment of the prognostic and survival statistics of familial versus sporadic melanoma. The identification of differences in the biologic behavior of familial and nonfamilial melanoma is of potential clinical importance and might provide meaningful insight into the pathogenesis of the disease. Several factors have made it difficult to analyze differences between familial and sporadic melanomas. Accurate identification of subjects with a true family history of melanoma versus a sporadic melanoma is difficult and has often relied on self-reported family histories, usually from the affected patient. This is problematic because (1) subjects rarely know the medical history of relatives more distant than second degree,37 (2) subjects confuse married-in family members with biologic relatives,37 (3) affected individuals may be more likely to report cancer among relatives than controls with the same family history,37,51 (4) many patients do not understand the difference between melanoma and nonmelanoma skin cancer and may incorrectly report a positive family history, and (5) ascertainment bias is inherent in any study in which recruitment is based on self-report or physician referral, because these sources cannot account for all melanomas within a population. For these reasons, it is difficult to be confident in the assignment of an individual to a sporadic or familial group. The population-based design of this study using the Utah Population Database minimizes these difficulties and biases.
A few large, population-based studies of familial cancer have demonstrated excessive risk of cancer among family members having close (first- or second-degree) relatives with the neoplasm of interest.26,52,53 The data presented here represent the first study of familial melanoma in which the collective influence of close and distant relatives was used to assess the biologic behavior of melanoma in familial and sporadic settings.
In this study, subjects with invasive malignant melanoma were classified with respect to familial risk by the familial standardized incidence ratio, derived from the complete risk experience of all observable biologic relatives.37,48 Use of the familial standardized incidence ratio requires a large genealogical database with links to population-based disease registries,37 resources readily available for the study of malignant melanoma occurring in Utah. As such, this investigation provides an objective, population-based, comprehensive comparison of melanoma occurring in high- and low-familial-risk settings.
Because the familial standardized incidence ratio is a continuous variable, a cutoff value was established to assign high- and low-familial-risk subjects. As with any cutoff, it is possible that a few melanoma subjects are misclassified, but the familial standardized incidence ratio 1.0 classified 9% of the subjects as high familial risk, in keeping with previous estimates of the proportion of melanomas reported to have a familial component (5% to 12%).1,18,44-47 Because the familial standardized incidence ratio is weighted more heavily for closer relatives, high familial standardized incidence ratio values ( 1.0) can be seen in (1) large families with multiple occurrences of malignant melanoma, even among distant relatives, and (2) small families with one or few affected relatives. The number of affected first- and second-degree relatives is a straightforward measure of familial aggregation and forms the basis of most familial studies. The significantly larger number of affected first- and second-degree relatives seen in the high-familial-risk group directly correlates with the familial standardized incidence ratio, as expected, because the calculation incorporates this count in addition to all other observable biologic relatives. Our analysis suggests that the biologic behavior of familial melanoma is not significantly different from nonfamilial melanoma with regard to sex, Breslow depth, Clark level, distribution, ulceration status, or lymph-node status. Moreover, our conclusions did not change when the data were analyzed in two other ways: (1) using familial risk as a continuous predictor rather than establishing a specific cutoff value and (2) comparing the high-familial-risk group (upper 9% of all familial standardized incidence ratio values) with a very-low-familial-risk group (lowest 9% of all familial standardized incidence ratio values).
Barnhill et al1 reported an older average age at diagnosis among a small number of familial melanoma subjects. Other authors have reported a younger age at diagnosis in familial subjects that ranged in age from an average of 33 to 47 years.7,16,17,19,51 In our analysis, the average age at first diagnosis was slightly lower in the familial melanoma group (57 v 60 years), primarily because of a higher proportion of subjects with melanoma diagnosed at age 30 years or younger in the high-familial-risk group. These data strongly suggest that melanoma in a familial setting may have an earlier biologic onset.
Balch et al54 have published a large, comprehensive analysis of melanoma in the United States that demonstrated that overall and disease-specific survival in melanoma patients was significantly influenced by well-known prognostic factors including subject age, sex, depth of tumor invasion (both the Breslow measurement and Clark level), and lymph-node metastasis. We have confirmed that the Utah melanoma population demonstrates similar prognostic and survival characteristics and that the familial subset of this population is no different. Thus, although familiality is the single greatest risk factor for development of melanoma, it is not an independent prognostic indicator and does not seem to influence the overall or melanoma-specific survival.
In summary, we observed that melanomas within a high-familial-risk setting are diagnosed at a slightly younger age, and a larger proportion are diagnosed at or before the age of 30 years, suggesting that familial melanoma has an earlier onset than sporadic melanoma. This earlier age of onset is not associated with a difference in survival but does confer a two-fold risk of having a hereditary pattern of melanoma. The overall number of primary melanomas was similar in the high- and low-familial-risk groups. Although the number of cases was small, it was noted that subjects with three or more primary melanomas had a five-fold increased risk of having an inherited pattern of melanoma. Taken together, our data imply that the biologic behavior of melanoma is the same regardless of whether it occurs in a familial or sporadic setting and suggests that the familial melanoma disease model will be more broadly applicable to sporadic melanoma.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Acknowledgment
We thank the Huntsman Cancer Foundation for database support provided to the Pedigree and Population Resource of Huntsman Cancer Institute.
NOTES
Supported by the Dermatology Foundation Leaders Society Dermatologist Investigator Research Fellowship and Clinical Career Development Award (S.R.F.), National Institutes of Health Grants No. K23 RR17525-01 (S.R.F.) and CA102422 (L.A.C.-A.), the Doris Duke Charitable Foundation (S.A.L.), Fellowship-to-Faculty Transition Award from the University of Utah, funded in part by the Howard Hughes Medical Institute (S.A.L.), the Huntsman Cancer Foundation (S.A.L.), the Tom C. Mathews Jr Familial Melanoma Research Clinic at Huntsman Cancer Institute, Huntsman General Clinical Research Center Public Health Service Grant No. MO1 RR00064, National Cancer Institute Cancer Center support Grant No. 5P30CA420-14, and the Utah Cancer Registry, funded by Contract No. NCI-CN-67000 from the National Cancer Institute, with additional support from the Utah Department of Health and the University of Utah.
Presented in part at the 2nd International Melanoma Research Congress, Phoenix, AZ, November 13-16, 2004.
The current affiliation for C.W. is the New Mexico Tumor Registry, University of New Mexico, Albuquerque, NM.
The current affiliation for A.T. is the Division of Biostatistics, Department of Epidemiology and Preventive Medicine, University of California at Davis, Davis, CA.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Barnhill RL, Roush GC, Titus-Ernstoff L, et al: Comparison of nonfamilial and familial melanoma. Dermatology 184:2-7, 1992
Clark WH Jr, Reimer RR, Greene M, et al: Origin of familial malignant melanomas from heritable melanocytic lesions: The B-K mole syndrome. Arch Dermatol 114:732-738, 1978
Greene MH, Goldin LR, Clark WH Jr, et al: Familial cutaneous malignant melanoma: Autosomal dominant trait possibly linked to the Rh locus. Proc Natl Acad Sci U S A 80:6071-6075, 1983
Greene MH, Sanders RJ, Chu FC, et al: The familial occurrence of cutaneous melanoma, intraocular melanoma, and the dysplastic nevus syndrome. Am J Ophthalmol 96:238-245, 1983
Greene MH, Clark WH Jr, Tucker MA, et al: High risk of malignant melanoma in melanoma-prone families with dysplastic nevi. Ann Intern Med 102:458-465, 1985
Kopf AW, Mintzis M, Grier RN, et al: Familial malignant melanoma. Cutis 17:873-876, 1976
Kopf AW, Hellman LJ, Rogers GS, et al: Familial malignant melanoma. JAMA 256:1915-1919, 1986
Reimer RR, Clark WH Jr, Greene MH, et al: Precursor lesions in familial melanoma: A new genetic preneoplastic syndrome. JAMA 239:744-746, 1978
Wallace DC, Beardmore GL, Exton LA: Familial malignant melanoma. Ann Surg 177:15-20, 1973
Hussussian CJ, Struewing JP, Goldstein AM, et al: Germline p16 mutations in familial melanoma. Nat Genet 8:15-21, 1994
Kamb A, Shattuck-Eidens D, Eeles R, et al: Analysis of the p16 gene (CDKN2) as a candidate for the chromosome 9p melanoma susceptibility locus. Nat Genet 8:23-26, 1994
Randerson-Moor JA, Harland M, Williams S, et al: A germline deletion of p14(ARF) but not CDKN2A in a melanoma-neural system tumour syndrome family. Hum Mol Genet 10:55-62, 2001
Rizos H, Darmanian AP, Holland EA, et al: Mutations in the INK4a/ARF melanoma susceptibility locus functionally impair p14ARF. J Biol Chem 276:41424-41434, 2001
Zuo L, Weger J, Yang Q, et al: Germline mutations in the p16INK4a binding domain of CDK4 in familial melanoma. Nat Genet 12:97-99, 1996
Rizos H, Puig S, Badenas C, et al: A melanoma-associated germline mutation in exon 1beta inactivates p14ARF. Oncogene 20:5543-5547, 2001
Anderson DE: Clinical characteristics of the genetic variety of cutaneous melanoma in man. Cancer 28:721-725, 1971
Goldstein AM, Fraser MC, Clark WH Jr, et al: Age at diagnosis and transmission of invasive melanoma in 23 families with cutaneous malignant melanoma/dysplastic nevi. J Natl Cancer Inst 86:1385-1390, 1994
Burden AD, Newell J, Andrew N, et al: Genetic and environmental influences in the development of multiple primary melanoma. Arch Dermatol 135:261-265, 1999
Anderson DE, Smith JL Jr, McBride CM: Hereditary aspects of malignant melanoma. JAMA 200:741-746, 1967
Jorde LB: Consanguinity and prereproductive mortality in the Utah Mormon population. Hum Hered 52:61-65, 2001
McLellan T, Jorde LB, Skolnick MH: Genetic distances between the Utah Mormons and related populations. Am J Hum Genet 36:836-857, 1984
Skolnick MH: The Utah genealogical database: A resource for genetic epidemiology, in MH Skonick (ed): Banbury Report No. 4: Cancer Incidence in Defined Populations. New York, NY, Cold Spring Harbor Laboratory, 1980, pp 285-297
Cannon-Albright LA, Goldgar DE, Meyer LJ, et al: Assignment of a locus for familial melanoma, MLM, to chromosome 9p13-p22. Science 258:1148-1152, 1992
Skolnick MH, Cannon-Albright LA, Goldgar DE, et al: Inheritance of proliferative breast disease in breast cancer kindreds. Science 250:1715-1720, 1990
Wooster R, Neuhausen SL, Mangion J, et al: Localization of a breast cancer susceptibility gene, BRCA2, to chromosome 13q12-13. Science 265:2088-2090, 1994
Goldgar DE, Easton DF, Cannon-Albright LA, et al: Systematic population-based assessment of cancer risk in first-degree relatives of cancer probands. J Natl Cancer Inst 86:1600-1608, 1994
Cannon-Albright LA, Thomas A, Goldgar DE, et al: Familiality of cancer in Utah. Cancer Res 54:2378-2385, 1994
Hemminki K, Li X, Plna K, et al: The nation-wide Swedish family-cancer database—updated structure and familial rates. Acta Oncol 40:772-777, 2001
Amundadottir LT, Thorvaldsson S, Gudbjartsson DF, et al: Cancer as a complex phenotype: Pattern of cancer distribution within and beyond the nuclear family. PLoS Med 1:e65, 2004
Burden AD, Vestey JP, Sirel JM, et al: Multiple primary melanoma: Risk factors and prognostic implications. BMJ 309:375, 1994
Titus-Ernstoff L, Duray PH, Ernstoff MS, et al: Dysplastic nevi in association with multiple primary melanoma. Cancer Res 48:1016-1018, 1988
Tucker MA, Boice JD Jr, Hoffman DA: Second cancer following cutaneous melanoma and cancers of the brain, thyroid, connective tissue, bone, and eye in Connecticut, 1935-82. Natl Cancer Inst Monogr 68:161-189, 1985
Lynch HT, Brand RE, Hogg D, et al: Phenotypic variation in eight extended CDKN2A germline mutation familial atypical multiple mole melanoma-pancreatic carcinoma-prone families: The familial atypical mole melanoma-pancreatic carcinoma syndrome. Cancer 94:84-96, 2002
Parker JF, Florell SR, Alexander A, et al: Pancreatic carcinoma surveillance in patients with familial melanoma. Arch Dermatol 139:1019-1025, 2003
Goldstein AM, Fraser MC, Struewing JP, et al: Increased risk of pancreatic cancer in melanoma-prone kindred with p16INK4 mutations. N Engl J Med 333:970-974, 1995
Hansen CB, Wadge LM, Lowstuter K, et al: Clinical germline genetic testing for melanoma. Lancet Oncol 5:314-319, 2004
Kerber RA: Method for calculating risk associated with family history of a disease. Genet Epidemiol 12:291-301, 1995
Slattery ML, Berry TD, Kerber RA: Is survival among women diagnosed with breast cancer influenced by family history of breast cancer? Epidemiology 4:543-548, 1993
Slattery ML, Kerber RA: The impact of family history of colon cancer on survival after diagnosis with colon cancer. Int J Epidemiol 24:888-896, 1995
Mineau GP, Fraser A: Kinship structure: Families past and present, in C Zick, KR Smith (eds): Utah at the Beginning of the New Millennium–A Demographic Perspective. Salt Lake City, UT, University of Utah Press, In press
Breslow NE, Day NE: Statistical methods in cancer research. Volume II—The design and analysis of cohort studies. IARC Sci Publ (82):1-406, 1987
Malecot G: The Mathematics of Heredity. Paris, France, Masson, 1948
Kerber RA, O'Brien E: A cohort study of cancer risk in relation to family histories of cancer in the Utah population database. Cancer 103:1906-1915, 2005
Goldstein AM, Tucker MA: Genetic epidemiology of familial melanoma. Dermatol Clin 13:605-612, 1995
Grange F, Chompret A, Guilloud-Bataille M, et al: Comparison between familial and nonfamilial melanoma in France. Arch Dermatol 131:1154-1159, 1995
Greene MH, Fraumeni JFJ: The hereditary variant of malignant melanoma, in WHJ Clark, LI Goldman, MJ Mastrangelo (eds): Human Malignant Melanoma. New York, NY, Grune and Stratton, 1979, pp 139-166
Wallace DC, Exton LA, McLeod GR: Genetic factor in malignant melanoma. Cancer 27:1262-1266, 1971
Boucher KM, Kerber RA: Measures of familial aggregation as predictors of breast-cancer risk. J Epidemiol Biostat 6:377-385, 2001
Kefford RF, Newton Bishop JA, Bergman W, et al: Counseling and DNA testing for individuals perceived to be genetically predisposed to melanoma: A consensus statement of the Melanoma Genetics Consortium. J Clin Oncol 17:3245-3251, 1999
Doubrovsky A, Menzies SW: Enhanced survival in patients with multiple primary melanoma. Arch Dermatol 139:1013-1018, 2003
Duke D, Castresana J, Lucchina L, et al: Familial cutaneous melanoma and two-mutational-event modeling. Cancer 72:3239-3243, 1993
Hemminki K, Zhang H, Czene K: Familial and attributable risks in cutaneous melanoma: Effects of proband and age. J Invest Dermatol 120:217-223, 2003
Hemminki K, Lonnstedt I, Vaittinen P: A population-based study of familial cutaneous melanoma. Melanoma Res 11:133-140, 2001
Balch CM, Soong SJ, Gershenwald JE, et al: Prognostic factors analysis of 17,600 melanoma patients: Validation of the American Joint Committee on Cancer melanoma staging system. J Clin Oncol 19:3622-3634, 2001(Scott R. Florell, Kenneth)