Pediatric Melanoma: Risk Factor and Survival Analysis of the Surveillance, Epidemiology and End Results Database
http://www.100md.com
《临床肿瘤学》
the Department of Pediatrics, Johns Hopkins University School of Medicine, Baltimore
Biostatistics Branch and Genetics Epidemiology Branch/Division of Cancer Epidemiology and Genetics and Pediatric Oncology Branch/Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD
ABSTRACT
PURPOSE: To evaluate risk factors for the development of and factors influencing survival in pediatric melanoma.
PATIENTS AND METHODS: We evaluated 1,255 children (age < 20 years) and 2,673 young adults (age 20 to 24 years) with melanoma in the 2001 National Cancer Institute (NCI) Surveillance, Epidemiology and End Results (SEER) database. We estimated exposure to UV radiation based on Environmental Protection Agency (EPA) measurements.
RESULTS: The incidence of pediatric melanoma increased 46% (95% CI, 40 to 52) per year of age and 2.9% (95% CI, 2.1 to 3.6) per year from 1973 to 2001. Incidence rates were lower in black patients (–95%; 95% CI, –98 to –90) compared with white patients and in male patients (–39%; 95% CI –46 to –31) compared with females. Increased ambient UV radiation was associated with elevated risk (19% per kJ; 95% CI, 9 to 30). Children with melanoma had a 5-year melanoma-specific survival of 93.6% (95% CI, 91.9 to 94.9), which improved from 1973 to 2001. The hazard ratio of death from melanoma increased with male sex; older age; advanced disease; location of the primary other than extremities or torso; earlier year of diagnosis; and previous cancer.
CONCLUSION: The incidence of melanoma in the United States is increasing rapidly in children. Risk factors for pediatric melanoma include being white, being female, increasing age, and environmental UV radiation. Survival is decreased for children and adolescents with unfavorable prognostic factors (male sex, unfavorable site, and/or second primary or regional or distant metastasis). More effective therapeutic strategies are needed for these groups.
INTRODUCTION
The age-adjusted incidence of malignant melanoma is rapidly increasing in the United States. The rate of increase in all age groups was 2.8% per year from 1981 to 2001 and in children (age < 20 years) was 1.1% per year from 1975 to 2001.1 The diagnosis is often delayed in children since melanoma is rare (300 to 420 new cases per year), and benign lesions, especially the Spitz nevus, may mimic melanoma.2,3 The prognosis is good when there is prompt identification and wide local excision of early disease, but poor for those with advanced disease at presentation.4,5
Case-control studies in adults have identified multiple host and environmental factors associated with increased risk of malignant melanoma. Host factors include fair skin, white race, blond or red hair, light eye color, tendency to burn with UV radiation exposure, increased number of benign nevi, dysplastic nevi, family history of melanoma, and xeroderma pigmentosum.6,7 Environmental factors include sunburns, often as a child, and increased exposure to UV radiation. Proposed risk factors for pediatric melanoma include congenital, dysplastic, or increased number of nevi; inability to tan; blue eyes; facial freckling; family history of melanoma; disorders of DNA excision repair like xeroderma pigmentosum; acquired or congenital immunosuppression; and a previous history of malignancy.5,8
Accepted prognostic factors in adult melanoma include primary lesion thickness, ulceration, and nonextremity site; increased age; regional lymph node involvement; satellite or in-transit metastases; elevated serum lactate dehydrogenase level; and visceral or brain metastases.4,9 However, prognosis and prognostic factors in children are less well defined. In a recent review of more than 300 cases, the outcome for pediatric patients (5-year survival of 74%) was slightly worse than that of young adults, but these survival estimates have limitations.5 In a large European registry study of children, male sex and lesions on the trunk were associated with a worse prognosis.10 Advanced stage has been associated with poor prognosis in other pediatric studies.11,12 The goal of this study was to evaluate risk factors for the development of and factors influencing prognosis in pediatric melanoma.
PATIENTS AND METHODS
Database
The National Cancer Institute Surveillance, Epidemiology and End Results database consists of population-based tumor registries that together include 14% of the population of the United States.13 Those included in the registries are generally representative of the US population with respect to socioeconomic status and education, but the areas sampled are more urban, with a higher proportion of foreign-born individuals. Most cases (95%) have had continuous follow-up from diagnosis to November 1, 2003, or the time of death. We used SEER*Stat software to access the SEER 12 and the SEER 9 databases; both include data from 1973 to 2001.14-16
Case Identification
Selection criteria included diagnosis of malignant melanoma coded according to the International Classification of Childhood Cancer and age less than 25 years.17 Data extracted on each case included age, sex, year of diagnosis, SEER registry, SEER stage (in situ, localized, regional, or distant), extent of disease including Breslow thickness, ulceration, lymph nodes, and distant metastases, first primary, location of primary, grade, histology, race (white, black, Asian, Native American, or unknown), survival in months, vital status, and cause of death. Thickness, Clark level, and ulceration data were available starting in 1988. Cases were classified as pediatric (age < 20 years) or young adult (age 20 to 24 years) at diagnosis (Table 1).
UV Exposure Estimates
Environmental UV radiation measurements were performed in 1999 and 2000 at the Environmental Protection Agency's (EPA's) network of sites established for this purpose. The EPA provides the daily Diffey-weighted UV irradiance (DUV), the rate at which fair skin will redden, on their Web site (http://www.epa.gov/uvnet/). We calculated the mean of integrated DUV per day over 1 year at the sites within or close to registry areas, after excluding days with poor quality observations per guidelines on the Web site (Table 2). 18 For Connecticut, the mean DUV per day was estimated with a regression model comparing mean DUV to mean UV as measured by Robertson-Berger meters at ground level at weather stations in each registry.19
Analysis
Identified cases and associated data were exported to Intercooled Stata 8.2 for Windows (Stata Corp, College Station, TX) for all further analysis.20 Manual review and exploratory data analysis were used to confirm search criteria. We compared baseline characteristics by the 2 test for categoric variables and Student's t test for continuous variables.
We used Intercooled Stata 8.2 to calculate age-adjusted incidence rates and incidence rate ratios (IRR) using the 2000 US census population. Incidence rates and IRR were compared using the 2 test.21 We evaluated the relationship between rates of melanoma and UV radiation, race, sex, age, year of diagnosis, and registry by univariate and multivariate Poisson regression. We used the deviance statistic as a measure of the goodness of fit of the Poisson regression. The significance of indicator variables for race, registry, and adult age was evaluated by the Wald test. The final Poisson model included age, natural log of age, sex, race, UV radiation, and year of diagnosis as independent variables.
We analyzed melanoma-specific mortality using the Kaplan-Meier estimator for the entire group and stratified by variable of interest. We censored cases with other or unknown causes of death (33% of pediatric cases) at the time of death. We compared survival functions with the log-rank test for equality of survival functions. After excluding in situ melanoma, we used a Cox proportional hazards regression model to assess differences in melanoma-specific and all-cause mortality by age group, sex, SEER stage, thickness and location of primary, first primary, histology, and race. We used females with localized melanoma of the extremities and no previous cancer as the reference group. Models were compared with the partial likelihood ratio test. The proportional hazard assumption was examined using time dependent variables. All tests were two-tailed and P values less than .05 were considered significant. The final models included only pediatric cases (age < 20 years).
RESULTS
Case finding using the SEER*Stat program (National Cancer Institute, Bethesda, MD) identified 1,255 cases of melanoma (including 204 in situ cases) in patients younger than 20 years of age and 2,673 cases (including 408 in situ cases) in patients 20 to 24 years old from the 140,206 cases included in the SEER databases from 1973 to 2001. The incidence of melanoma in children (age < 20 years) increased 2.9% (95% CI, 2.1 to 3.6) per year from 1973 to 2001 after adjustment for age, race, sex, and ambient UV radiation in the multivariate Poisson regression. Young adults (age 20 to 24 years) and adolescents (age 10 to 19 years) had similar rates of increase (3.0% per year), while young children (age < 10 years) had a lower rate of increase (1.4% per year). The incidence of melanoma increased with age (46% per year; 95% CI, 40 to 52) and was significantly lower in black patients (–95%; 95% CI, –98 to –90) and Asians and Native Americans (–82%; 95% CI, –88 to –73) compared with white patients (Fig 1). Males had a significantly lower incidence (–39%; 95% CI, –46 to –31) compared with females; primarily due to different rates of melanoma of the extremities (Fig 2). Incidence also varied by registry (ie, residence at the time of diagnosis), but these differences were not significant in the multivariate model. The incidence rate of melanoma was positively correlated with environmental UV radiation (19% per kJ; 95% CI, 9 to 30).
There were significant differences in baseline characteristics of young children (age < 10 years) compared with adolescents and young adults (Table 1). Young children were more likely to be non-white and to have metastases; nodular or other histology; head, face, or neck primaries; thicker lesions; and history of cancer.
Kaplan-Meier estimates of melanoma-specific survival were similar by age group (Fig 3; P = .20). Survival significantly decreased with male sex (Fig 4; P < .005) or regional or distant metastasis (Fig 5, P < .001). Five-year melanoma-specific survival for pediatric cases (age < 20 years) was 100% for in situ disease, 96.1% (95% CI, 94.3 to 97.3) for localized disease, 77.2% (95% CI, 66.4 to 85.0) for regional disease, and 57.3% (95% CI, 33.0 to 75.6) for distant disease. Five-year overall survival was 88.9% (95% CI, 80.3 to 93.9) for young children (age < 10 years), 91.5% (95% CI, 89.5 to 93.1) for adolescents (age 10 to 19 years), and 90.9% (95% CI, 89.6 to 92.0 years) for young adults (age 20 to 24 years; P > .5) without significant differences by stage (data not shown). After excluding in situ melanoma, 5-year overall survival for pediatric cases (age < 20 years) was 90.9% (95% CI, 88.7 to 92.6).
Univariate Cox proportional hazard analysis showed significantly worse survival for males, patients with regional or unstaged disease; nodular histology; increasing thickness of the primary; primary of the head, face, neck, eye, orbit, CNS, genitals, or overlapping sites; earlier year of diagnosis; and previous cancer (Table 3). The final multivariate model identified similar associations. As information regarding thickness of the primary and nodular histology were not available before 1988, these variables were not included in the final model (Table 3).
DISCUSSION
We analyzed children and young adults with melanoma included in the SEER database between 1973 and 2001. Older age, more recent year of diagnosis, female sex, white race, and increased environmental UV radiation were all associated with a significant increase in the risk of melanoma. Our findings are consistent with earlier analyses of the SEER database,22,23 and a European population-based study of children.10 Notably, increasing age confers greater risk in white compared with black individuals. In the first year of life, the incidence of melanoma is similar by race, but it diverges by age 5 to 9 years and is more than 40-fold higher in white individuals by age 20 to 24 years. This results from the logarithmic increase in melanoma incidence in white patients after age 10 years. This pattern may reflect increased susceptibility to UV radiation exposure in white individuals, other environmental factors, and/or differences in genetic predisposition.
The incidence of melanoma is increasing rapidly in children, especially in adolescents. This increase is similar to that seen in young adults. This may reflect increased cumulative UV exposure during childhood or adolescence, greater awareness and more frequent diagnosis of melanoma (eg, versus atypical Spitz nevus), and/or other environmental factors. Review of the pathology of pediatric melanoma cases in the SEER registries could clarify whether the histologic characteristics have changed over time.
The increased risk of melanoma in girls, particularly on the lower extremities, may be a result of increased UV exposure. In adults, sun-related behaviors differ between men and women.22 The increased rates of melanoma in adolescent and young women may reflect sunbathing or the widespread (> 25%) practice of indoor tanning.24
Increased exposure to UV radiation, measured by number of blistering sunburns (often during childhood), and reported time spent outdoors, confers a two- to five-fold increased risk of melanoma in case-control studies of adults.23,25-27 Animal models also support the role of early sunburn in the pathogenesis of melanoma.28 Although case-specific exposure was not directly evaluated in this study, there was a 19% increase in the incidence of melanoma in children per kJ of DUV.
The prognosis for young children, adolescents, and young adults with melanoma appears to be similar. Although the difference in survival was small across age groups in this study, mortality from melanoma in pediatric patients (age < 20 years) increased with older age in the multivariate analysis. Increasing age is a known negative prognostic factor in melanoma of all stages, but this effect is mostly in older adults.4 We also found increased risk of death in male children, those with regional or distant metastasis, primary sites other than the extremities or torso, increasing thickness of the primary, earlier year of diagnosis, and previous cancer.
Importantly, the thickness of the primary lesion represents one of the strongest prognostic factors in melanoma in adults. This association was evident in children, but the limited numbers of cases with complete data precluded full assessment of its contribution. In the multivariate analysis, other variables (eg, histology, extent of disease) may have accounted for much of the predictive value of thickness. For example, of the 50 patients with primary lesions more than 4 mm in depth and complete staging, 70% had regional or distant disease at presentation.
Melanoma-specific survival in children has improved by approximately 4% per year during the last 3 decades. It is difficult to explain this dramatic improvement. Although earlier diagnosis could be associated with improved survival, there has been no decrease in lesion thickness over the last decade. Furthermore, survival has improved for all stages of pediatric melanoma. The most notable improvement has been in the "unstaged" group, likely due to more complete staging. This group has become smaller over time and there has been a recent increase in the percentage of children with regional disease, which may reflect the increased use of sentinel node biopsy. The application of surgical treatment standards from adults to children with melanoma may have contributed to better local control and improvements in survival. It is unlikely that sentinel lymph node biopsy or systemic therapies have played a significant role because of the limited use of these approaches for pediatric melanoma. However, the prevalence of positive sentinel lymph nodes is higher in patients younger than 35 years than in older adults. This finding supports the recommendation for sentinel lymph node biopsy in adults younger than 35 years and additional indicators of poor prognosis even if their melanoma is less than 1 mm thick.29 This suggests that sentinel lymph node biopsy should be considered in children with the caveat that the best approach is unknown and requires further investigation.
There are important differences in young children (age < 10 years) with melanoma compared with adolescents and young adults that may reflect distinct tumor biology and/or host characteristics. Younger patients are more likely to have congenital nevi and, possibly, syndromes that predispose to cancer. Our data suggest that young children with melanoma are more likely to have poor prognostic features (eg, metastasis, thick primaries, high risk histology, history of cancer). Despite this, survival seems to be similar in adolescents and young adults.
Published large series of pediatric melanoma report 5-year survival rates of 74% to 80%.5,10 This is significantly worse that the 91% 5-year overall survival seen in our analysis after the exclusion of cases of melanoma in situ. Survival curves calculated from the SEER database are likely to be more accurate for several reasons. Unlike most hospital-based registries and retrospective case series, the SEER database contains many large population-based registries and should be representative of the spectrum of melanoma in the United States. Series from referral centers are likely to be biased by the selection of patients with more advanced disease. On the other hand, although patients with in situ melanoma were excluded from the survival analyses in this study, mortality may still have been underestimated by limiting the analyses to melanoma-specific death. Notably, 33% of the pediatric deaths were not attributable to melanoma, including 10% with missing or uninterpretable death certificates. However, including these deaths did not modify the strength of the associations noted above and only significantly changed survival for regional and distant disease (data not shown). Additionally, those in whom follow-up was incomplete may have been more likely to die. This form of bias may be particularly troublesome in a highly mobile population like that of the US. Fortunately, the SEER database has active follow-up to minimize this cause of information bias. As in all studies, children with melanoma in the SEER database may represent misclassification of benign lesions. The diagnosis of malignant melanoma is particularly challenging in children and concordance of expert dermatopathologists is frequently variable.30,31 The SEER database, while the standard for population-based cancer registries, has no central review of the diagnostic specimens. There are several additional limitations of these analyses. Incidence rates may have been underestimated in more recent years due to reporting delays. To maximize the inclusion of cases, there is a 2-year delay between the date of diagnosis and inclusion in the SEER database. However, reporting can be delayed for longer than 2 years for diseases, like melanoma, that are frequently diagnosed and treated entirely in the outpatient setting.32 The environmental measure of UV radiation is an imprecise measure of individual exposure. Information in regard to several putative risk factors for melanoma, including skin type, eye color, family history, number of blistering sunburns, benign nevi, and acquired or inherited immunodeficiency is not available in the SEER database. Finally, information on histology is frequently incomplete (not specified in 55% of the pediatric cases) and data about thickness, Clark level, and ulceration have only been included since 1988.
In summary, pediatric melanoma is an important and increasing problem. Overall, melanoma in childhood is similar to that in young adults, except that there are greater proportions of cases in males, unusual sites, and thicker primaries. Factors conferring risk of adult melanoma, including older age, white race, environmental exposure to UV radiation, and female sex, are also important in pediatric melanoma. We recommend further study of pediatric melanoma, preferably through the creation of a national cancer registry or use of the existing Rare Tumor Registry of the Children's Oncology Group. With local excision, 5-year survival for pediatric melanoma is excellent, except in the infrequent cases of regional or metastatic disease. For these children, effective systemic therapies are needed and are most likely to be identified through close cooperation of national oncology groups to develop trials that include both children and adults.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
NOTES
Supported by a National Cancer Institute K12 Training Grant, 2K12CA001709-11 (J.J.S.).
Presented at the 40th Annual Meeting of American Society of Clinical Oncology, New Orleans, LA, June 5-8, 2004.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Ries L, Eisner M, Kosary C, et al: SEER Cancer Statistics Review 1975-2000. Bethesda, MD, National Cancer Institute, 2004, http://seer.cancer.gov/csr/1975_2001/
Fabrizi G, Massi G: Spitzoid malignant melanoma in teenagers: An entity with no better prognosis than that of other forms of melanoma. Histopathology 38:448-453, 2001
Pappo AS, Ries LAG, Herzog C, et al: Malignant melanoma in the first three decades of life: A report from the U.S. Surveillance, Epidemiology and End Results (SEER) program. 23:721, 2004 (abstr 7557)
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
Pappo AS: Melanoma in children and adolescents. Eur J Cancer 39:2651-2661, 2003
Kraemer KH, Lee MM, Andrews AD, et al: The role of sunlight and DNA repair in melanoma and nonmelanoma skin cancer: The xeroderma pigmentosum paradigm. Arch Dermatol 130:1018-1021, 1994
Tucker MA, Goldstein AM: Melanoma etiology: Where are we Oncogene 22:3042-3052, 2003
Whiteman DC, Valery P, McWhirter W, et al: Risk factors for childhood melanoma in Queensland, Australia. Int J Cancer 70:26-31, 1997
Ahmed I: Malignant Melanoma: Prognostic Indicators. Mayo Clin Proc 72:356-361, 1997
Conti EM, Cercato MC, Gatta G, et al: Childhood melanoma in Europe since 1978: a population-based survival study. Eur J Cancer 37:780-784, 2001
Saenz NC, Saenz-Badillos J, Busam K, et al: Childhood melanoma survival. Cancer 85:750-754, 1999
Milton GW, Shaw HM, Thompson JF, et al: Cutaneous melanoma in childhood: Incidence and prognosis. Australas J Dermatol 38:S44-S48, 1997
About SEER. 2004 (cited 2004 September 16). http:/seer.cancer.gov/about/
Surveillance Research Program, National Cancer Institute SEER*Stat software version 5.2.2. http://www.seer.cancer.gov/seerstat
Surveillance, Epidemiology, and End Results (SEER) Program SEER*Stat Database: Incidence - SEER 11 Regs + AK Public-Use, Nov 2003 Sub (1973-2001 varying). National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch: Bethesda; released April 2004. http://www.seer.cancer.gov
Surveillance, Epidemiology, and End Results (SEER) Program SEER*Stat Database: Incidence - SEER 9 Regs Public-Use, Nov 2003 Sub (1973-2001). National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch: Bethesda; released April 2004 (cited 2004 August 17). http://www.seer.cancer.gov
Kramarova E, Stiller C, Ferlay J, et al: International Classification of Childhood Cancer: IARC Technical Report No. 29. Lyon, France, IARCPress, 1996, pp 48
Environmental Protection Agency. Web site. http://www.epa.gov/uvnet/
Scotto J, Fears TR, Fraumeni JF Jr: Solar Radiation, in Shottenfeld D, Fraumeni JF Jr (eds): Cancer Epidemiology and Prevention (ed 2). New York, NY, Oxford University Press, 1996, pp 355-372
Stata Statistical Software: Intercooled Stata 8.2. College Station, TX, Stata Corp, 2003
Rosner B: Fundamental of Biostatistics (ed 5). Pacific Grove, CA, Duxbury, 2000
Hamre MR, Chuba P, Bakhshi S, et al: Cutaneous melanoma in childhood and adolescence. Pediatr Hematol Oncol 19:309-317, 2002
Fears TR, Bird CC, Guerry D IV, et al: Average midrange ultraviolet radiation flux and time outdoors predict melanoma risk. Cancer Res 62:3992-3996, 2002
Demko CA, Borawski EA, Debanne SM, et al: Use of Indoor Tanning Facilities by White Adolescents in the United States. Arch Pediatr Adolesc Med 157:854-860, 2003
Loria D, Matos E: Risk factors for cutaneous melanoma: A case-control study in Argentina. Int J Dermatol 40:108-114, 2001
Gilchrest BA, Eller MS, Geller AC, et al: The pathogenesis of melanoma induced by ultraviolet radiation. N Engl J Med 340:1341-1348, 1999
Kaskel P, Sander S, Kron M, et al: Outdoor activities in childhood: A protective factor for cutaneous melanoma Results of a case-control study in 271 matched pairs. Br J Dermatol 145:602-609, 2001
Noonan FP, Recio JA, Takayama H, et al: Neonatal sunburn and melanoma in mice. Nature 413:271-272, 2001
Sondak VK, Taylor JMG, Sabel MS, et al: Mitotic rate and younger age are predictors of sentinel lymph node positivity: Lessons learned from the generation of a probabilistic model. Ann Surg Oncol 11:247-258, 2004
Wechsler J, Bastuji-Garin S, Spatz A, et al: Reliability of the histopathologic diagnosis of malignant melanoma in childhood. Arch Dermatol 138:625-628, 2002
Farmer ER, Gonin R, Hanna MP: Discordance in the histopathologic diagnosis of melanoma and melanocytic nevi between expert pathologists. Hum Pathol 27:528-531, 1996
Clegg LX, Feuer EJ, Midthune DN, et al: Impact of reporting delay and reporting error on cancer incidence rates and trends: JNCI Cancer Spectrum. J Natl Cancer Inst 94:1537-1545, 2002(John J. Strouse, Thomas R)
Biostatistics Branch and Genetics Epidemiology Branch/Division of Cancer Epidemiology and Genetics and Pediatric Oncology Branch/Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD
ABSTRACT
PURPOSE: To evaluate risk factors for the development of and factors influencing survival in pediatric melanoma.
PATIENTS AND METHODS: We evaluated 1,255 children (age < 20 years) and 2,673 young adults (age 20 to 24 years) with melanoma in the 2001 National Cancer Institute (NCI) Surveillance, Epidemiology and End Results (SEER) database. We estimated exposure to UV radiation based on Environmental Protection Agency (EPA) measurements.
RESULTS: The incidence of pediatric melanoma increased 46% (95% CI, 40 to 52) per year of age and 2.9% (95% CI, 2.1 to 3.6) per year from 1973 to 2001. Incidence rates were lower in black patients (–95%; 95% CI, –98 to –90) compared with white patients and in male patients (–39%; 95% CI –46 to –31) compared with females. Increased ambient UV radiation was associated with elevated risk (19% per kJ; 95% CI, 9 to 30). Children with melanoma had a 5-year melanoma-specific survival of 93.6% (95% CI, 91.9 to 94.9), which improved from 1973 to 2001. The hazard ratio of death from melanoma increased with male sex; older age; advanced disease; location of the primary other than extremities or torso; earlier year of diagnosis; and previous cancer.
CONCLUSION: The incidence of melanoma in the United States is increasing rapidly in children. Risk factors for pediatric melanoma include being white, being female, increasing age, and environmental UV radiation. Survival is decreased for children and adolescents with unfavorable prognostic factors (male sex, unfavorable site, and/or second primary or regional or distant metastasis). More effective therapeutic strategies are needed for these groups.
INTRODUCTION
The age-adjusted incidence of malignant melanoma is rapidly increasing in the United States. The rate of increase in all age groups was 2.8% per year from 1981 to 2001 and in children (age < 20 years) was 1.1% per year from 1975 to 2001.1 The diagnosis is often delayed in children since melanoma is rare (300 to 420 new cases per year), and benign lesions, especially the Spitz nevus, may mimic melanoma.2,3 The prognosis is good when there is prompt identification and wide local excision of early disease, but poor for those with advanced disease at presentation.4,5
Case-control studies in adults have identified multiple host and environmental factors associated with increased risk of malignant melanoma. Host factors include fair skin, white race, blond or red hair, light eye color, tendency to burn with UV radiation exposure, increased number of benign nevi, dysplastic nevi, family history of melanoma, and xeroderma pigmentosum.6,7 Environmental factors include sunburns, often as a child, and increased exposure to UV radiation. Proposed risk factors for pediatric melanoma include congenital, dysplastic, or increased number of nevi; inability to tan; blue eyes; facial freckling; family history of melanoma; disorders of DNA excision repair like xeroderma pigmentosum; acquired or congenital immunosuppression; and a previous history of malignancy.5,8
Accepted prognostic factors in adult melanoma include primary lesion thickness, ulceration, and nonextremity site; increased age; regional lymph node involvement; satellite or in-transit metastases; elevated serum lactate dehydrogenase level; and visceral or brain metastases.4,9 However, prognosis and prognostic factors in children are less well defined. In a recent review of more than 300 cases, the outcome for pediatric patients (5-year survival of 74%) was slightly worse than that of young adults, but these survival estimates have limitations.5 In a large European registry study of children, male sex and lesions on the trunk were associated with a worse prognosis.10 Advanced stage has been associated with poor prognosis in other pediatric studies.11,12 The goal of this study was to evaluate risk factors for the development of and factors influencing prognosis in pediatric melanoma.
PATIENTS AND METHODS
Database
The National Cancer Institute Surveillance, Epidemiology and End Results database consists of population-based tumor registries that together include 14% of the population of the United States.13 Those included in the registries are generally representative of the US population with respect to socioeconomic status and education, but the areas sampled are more urban, with a higher proportion of foreign-born individuals. Most cases (95%) have had continuous follow-up from diagnosis to November 1, 2003, or the time of death. We used SEER*Stat software to access the SEER 12 and the SEER 9 databases; both include data from 1973 to 2001.14-16
Case Identification
Selection criteria included diagnosis of malignant melanoma coded according to the International Classification of Childhood Cancer and age less than 25 years.17 Data extracted on each case included age, sex, year of diagnosis, SEER registry, SEER stage (in situ, localized, regional, or distant), extent of disease including Breslow thickness, ulceration, lymph nodes, and distant metastases, first primary, location of primary, grade, histology, race (white, black, Asian, Native American, or unknown), survival in months, vital status, and cause of death. Thickness, Clark level, and ulceration data were available starting in 1988. Cases were classified as pediatric (age < 20 years) or young adult (age 20 to 24 years) at diagnosis (Table 1).
UV Exposure Estimates
Environmental UV radiation measurements were performed in 1999 and 2000 at the Environmental Protection Agency's (EPA's) network of sites established for this purpose. The EPA provides the daily Diffey-weighted UV irradiance (DUV), the rate at which fair skin will redden, on their Web site (http://www.epa.gov/uvnet/). We calculated the mean of integrated DUV per day over 1 year at the sites within or close to registry areas, after excluding days with poor quality observations per guidelines on the Web site (Table 2). 18 For Connecticut, the mean DUV per day was estimated with a regression model comparing mean DUV to mean UV as measured by Robertson-Berger meters at ground level at weather stations in each registry.19
Analysis
Identified cases and associated data were exported to Intercooled Stata 8.2 for Windows (Stata Corp, College Station, TX) for all further analysis.20 Manual review and exploratory data analysis were used to confirm search criteria. We compared baseline characteristics by the 2 test for categoric variables and Student's t test for continuous variables.
We used Intercooled Stata 8.2 to calculate age-adjusted incidence rates and incidence rate ratios (IRR) using the 2000 US census population. Incidence rates and IRR were compared using the 2 test.21 We evaluated the relationship between rates of melanoma and UV radiation, race, sex, age, year of diagnosis, and registry by univariate and multivariate Poisson regression. We used the deviance statistic as a measure of the goodness of fit of the Poisson regression. The significance of indicator variables for race, registry, and adult age was evaluated by the Wald test. The final Poisson model included age, natural log of age, sex, race, UV radiation, and year of diagnosis as independent variables.
We analyzed melanoma-specific mortality using the Kaplan-Meier estimator for the entire group and stratified by variable of interest. We censored cases with other or unknown causes of death (33% of pediatric cases) at the time of death. We compared survival functions with the log-rank test for equality of survival functions. After excluding in situ melanoma, we used a Cox proportional hazards regression model to assess differences in melanoma-specific and all-cause mortality by age group, sex, SEER stage, thickness and location of primary, first primary, histology, and race. We used females with localized melanoma of the extremities and no previous cancer as the reference group. Models were compared with the partial likelihood ratio test. The proportional hazard assumption was examined using time dependent variables. All tests were two-tailed and P values less than .05 were considered significant. The final models included only pediatric cases (age < 20 years).
RESULTS
Case finding using the SEER*Stat program (National Cancer Institute, Bethesda, MD) identified 1,255 cases of melanoma (including 204 in situ cases) in patients younger than 20 years of age and 2,673 cases (including 408 in situ cases) in patients 20 to 24 years old from the 140,206 cases included in the SEER databases from 1973 to 2001. The incidence of melanoma in children (age < 20 years) increased 2.9% (95% CI, 2.1 to 3.6) per year from 1973 to 2001 after adjustment for age, race, sex, and ambient UV radiation in the multivariate Poisson regression. Young adults (age 20 to 24 years) and adolescents (age 10 to 19 years) had similar rates of increase (3.0% per year), while young children (age < 10 years) had a lower rate of increase (1.4% per year). The incidence of melanoma increased with age (46% per year; 95% CI, 40 to 52) and was significantly lower in black patients (–95%; 95% CI, –98 to –90) and Asians and Native Americans (–82%; 95% CI, –88 to –73) compared with white patients (Fig 1). Males had a significantly lower incidence (–39%; 95% CI, –46 to –31) compared with females; primarily due to different rates of melanoma of the extremities (Fig 2). Incidence also varied by registry (ie, residence at the time of diagnosis), but these differences were not significant in the multivariate model. The incidence rate of melanoma was positively correlated with environmental UV radiation (19% per kJ; 95% CI, 9 to 30).
There were significant differences in baseline characteristics of young children (age < 10 years) compared with adolescents and young adults (Table 1). Young children were more likely to be non-white and to have metastases; nodular or other histology; head, face, or neck primaries; thicker lesions; and history of cancer.
Kaplan-Meier estimates of melanoma-specific survival were similar by age group (Fig 3; P = .20). Survival significantly decreased with male sex (Fig 4; P < .005) or regional or distant metastasis (Fig 5, P < .001). Five-year melanoma-specific survival for pediatric cases (age < 20 years) was 100% for in situ disease, 96.1% (95% CI, 94.3 to 97.3) for localized disease, 77.2% (95% CI, 66.4 to 85.0) for regional disease, and 57.3% (95% CI, 33.0 to 75.6) for distant disease. Five-year overall survival was 88.9% (95% CI, 80.3 to 93.9) for young children (age < 10 years), 91.5% (95% CI, 89.5 to 93.1) for adolescents (age 10 to 19 years), and 90.9% (95% CI, 89.6 to 92.0 years) for young adults (age 20 to 24 years; P > .5) without significant differences by stage (data not shown). After excluding in situ melanoma, 5-year overall survival for pediatric cases (age < 20 years) was 90.9% (95% CI, 88.7 to 92.6).
Univariate Cox proportional hazard analysis showed significantly worse survival for males, patients with regional or unstaged disease; nodular histology; increasing thickness of the primary; primary of the head, face, neck, eye, orbit, CNS, genitals, or overlapping sites; earlier year of diagnosis; and previous cancer (Table 3). The final multivariate model identified similar associations. As information regarding thickness of the primary and nodular histology were not available before 1988, these variables were not included in the final model (Table 3).
DISCUSSION
We analyzed children and young adults with melanoma included in the SEER database between 1973 and 2001. Older age, more recent year of diagnosis, female sex, white race, and increased environmental UV radiation were all associated with a significant increase in the risk of melanoma. Our findings are consistent with earlier analyses of the SEER database,22,23 and a European population-based study of children.10 Notably, increasing age confers greater risk in white compared with black individuals. In the first year of life, the incidence of melanoma is similar by race, but it diverges by age 5 to 9 years and is more than 40-fold higher in white individuals by age 20 to 24 years. This results from the logarithmic increase in melanoma incidence in white patients after age 10 years. This pattern may reflect increased susceptibility to UV radiation exposure in white individuals, other environmental factors, and/or differences in genetic predisposition.
The incidence of melanoma is increasing rapidly in children, especially in adolescents. This increase is similar to that seen in young adults. This may reflect increased cumulative UV exposure during childhood or adolescence, greater awareness and more frequent diagnosis of melanoma (eg, versus atypical Spitz nevus), and/or other environmental factors. Review of the pathology of pediatric melanoma cases in the SEER registries could clarify whether the histologic characteristics have changed over time.
The increased risk of melanoma in girls, particularly on the lower extremities, may be a result of increased UV exposure. In adults, sun-related behaviors differ between men and women.22 The increased rates of melanoma in adolescent and young women may reflect sunbathing or the widespread (> 25%) practice of indoor tanning.24
Increased exposure to UV radiation, measured by number of blistering sunburns (often during childhood), and reported time spent outdoors, confers a two- to five-fold increased risk of melanoma in case-control studies of adults.23,25-27 Animal models also support the role of early sunburn in the pathogenesis of melanoma.28 Although case-specific exposure was not directly evaluated in this study, there was a 19% increase in the incidence of melanoma in children per kJ of DUV.
The prognosis for young children, adolescents, and young adults with melanoma appears to be similar. Although the difference in survival was small across age groups in this study, mortality from melanoma in pediatric patients (age < 20 years) increased with older age in the multivariate analysis. Increasing age is a known negative prognostic factor in melanoma of all stages, but this effect is mostly in older adults.4 We also found increased risk of death in male children, those with regional or distant metastasis, primary sites other than the extremities or torso, increasing thickness of the primary, earlier year of diagnosis, and previous cancer.
Importantly, the thickness of the primary lesion represents one of the strongest prognostic factors in melanoma in adults. This association was evident in children, but the limited numbers of cases with complete data precluded full assessment of its contribution. In the multivariate analysis, other variables (eg, histology, extent of disease) may have accounted for much of the predictive value of thickness. For example, of the 50 patients with primary lesions more than 4 mm in depth and complete staging, 70% had regional or distant disease at presentation.
Melanoma-specific survival in children has improved by approximately 4% per year during the last 3 decades. It is difficult to explain this dramatic improvement. Although earlier diagnosis could be associated with improved survival, there has been no decrease in lesion thickness over the last decade. Furthermore, survival has improved for all stages of pediatric melanoma. The most notable improvement has been in the "unstaged" group, likely due to more complete staging. This group has become smaller over time and there has been a recent increase in the percentage of children with regional disease, which may reflect the increased use of sentinel node biopsy. The application of surgical treatment standards from adults to children with melanoma may have contributed to better local control and improvements in survival. It is unlikely that sentinel lymph node biopsy or systemic therapies have played a significant role because of the limited use of these approaches for pediatric melanoma. However, the prevalence of positive sentinel lymph nodes is higher in patients younger than 35 years than in older adults. This finding supports the recommendation for sentinel lymph node biopsy in adults younger than 35 years and additional indicators of poor prognosis even if their melanoma is less than 1 mm thick.29 This suggests that sentinel lymph node biopsy should be considered in children with the caveat that the best approach is unknown and requires further investigation.
There are important differences in young children (age < 10 years) with melanoma compared with adolescents and young adults that may reflect distinct tumor biology and/or host characteristics. Younger patients are more likely to have congenital nevi and, possibly, syndromes that predispose to cancer. Our data suggest that young children with melanoma are more likely to have poor prognostic features (eg, metastasis, thick primaries, high risk histology, history of cancer). Despite this, survival seems to be similar in adolescents and young adults.
Published large series of pediatric melanoma report 5-year survival rates of 74% to 80%.5,10 This is significantly worse that the 91% 5-year overall survival seen in our analysis after the exclusion of cases of melanoma in situ. Survival curves calculated from the SEER database are likely to be more accurate for several reasons. Unlike most hospital-based registries and retrospective case series, the SEER database contains many large population-based registries and should be representative of the spectrum of melanoma in the United States. Series from referral centers are likely to be biased by the selection of patients with more advanced disease. On the other hand, although patients with in situ melanoma were excluded from the survival analyses in this study, mortality may still have been underestimated by limiting the analyses to melanoma-specific death. Notably, 33% of the pediatric deaths were not attributable to melanoma, including 10% with missing or uninterpretable death certificates. However, including these deaths did not modify the strength of the associations noted above and only significantly changed survival for regional and distant disease (data not shown). Additionally, those in whom follow-up was incomplete may have been more likely to die. This form of bias may be particularly troublesome in a highly mobile population like that of the US. Fortunately, the SEER database has active follow-up to minimize this cause of information bias. As in all studies, children with melanoma in the SEER database may represent misclassification of benign lesions. The diagnosis of malignant melanoma is particularly challenging in children and concordance of expert dermatopathologists is frequently variable.30,31 The SEER database, while the standard for population-based cancer registries, has no central review of the diagnostic specimens. There are several additional limitations of these analyses. Incidence rates may have been underestimated in more recent years due to reporting delays. To maximize the inclusion of cases, there is a 2-year delay between the date of diagnosis and inclusion in the SEER database. However, reporting can be delayed for longer than 2 years for diseases, like melanoma, that are frequently diagnosed and treated entirely in the outpatient setting.32 The environmental measure of UV radiation is an imprecise measure of individual exposure. Information in regard to several putative risk factors for melanoma, including skin type, eye color, family history, number of blistering sunburns, benign nevi, and acquired or inherited immunodeficiency is not available in the SEER database. Finally, information on histology is frequently incomplete (not specified in 55% of the pediatric cases) and data about thickness, Clark level, and ulceration have only been included since 1988.
In summary, pediatric melanoma is an important and increasing problem. Overall, melanoma in childhood is similar to that in young adults, except that there are greater proportions of cases in males, unusual sites, and thicker primaries. Factors conferring risk of adult melanoma, including older age, white race, environmental exposure to UV radiation, and female sex, are also important in pediatric melanoma. We recommend further study of pediatric melanoma, preferably through the creation of a national cancer registry or use of the existing Rare Tumor Registry of the Children's Oncology Group. With local excision, 5-year survival for pediatric melanoma is excellent, except in the infrequent cases of regional or metastatic disease. For these children, effective systemic therapies are needed and are most likely to be identified through close cooperation of national oncology groups to develop trials that include both children and adults.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
NOTES
Supported by a National Cancer Institute K12 Training Grant, 2K12CA001709-11 (J.J.S.).
Presented at the 40th Annual Meeting of American Society of Clinical Oncology, New Orleans, LA, June 5-8, 2004.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Ries L, Eisner M, Kosary C, et al: SEER Cancer Statistics Review 1975-2000. Bethesda, MD, National Cancer Institute, 2004, http://seer.cancer.gov/csr/1975_2001/
Fabrizi G, Massi G: Spitzoid malignant melanoma in teenagers: An entity with no better prognosis than that of other forms of melanoma. Histopathology 38:448-453, 2001
Pappo AS, Ries LAG, Herzog C, et al: Malignant melanoma in the first three decades of life: A report from the U.S. Surveillance, Epidemiology and End Results (SEER) program. 23:721, 2004 (abstr 7557)
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
Pappo AS: Melanoma in children and adolescents. Eur J Cancer 39:2651-2661, 2003
Kraemer KH, Lee MM, Andrews AD, et al: The role of sunlight and DNA repair in melanoma and nonmelanoma skin cancer: The xeroderma pigmentosum paradigm. Arch Dermatol 130:1018-1021, 1994
Tucker MA, Goldstein AM: Melanoma etiology: Where are we Oncogene 22:3042-3052, 2003
Whiteman DC, Valery P, McWhirter W, et al: Risk factors for childhood melanoma in Queensland, Australia. Int J Cancer 70:26-31, 1997
Ahmed I: Malignant Melanoma: Prognostic Indicators. Mayo Clin Proc 72:356-361, 1997
Conti EM, Cercato MC, Gatta G, et al: Childhood melanoma in Europe since 1978: a population-based survival study. Eur J Cancer 37:780-784, 2001
Saenz NC, Saenz-Badillos J, Busam K, et al: Childhood melanoma survival. Cancer 85:750-754, 1999
Milton GW, Shaw HM, Thompson JF, et al: Cutaneous melanoma in childhood: Incidence and prognosis. Australas J Dermatol 38:S44-S48, 1997
About SEER. 2004 (cited 2004 September 16). http:/seer.cancer.gov/about/
Surveillance Research Program, National Cancer Institute SEER*Stat software version 5.2.2. http://www.seer.cancer.gov/seerstat
Surveillance, Epidemiology, and End Results (SEER) Program SEER*Stat Database: Incidence - SEER 11 Regs + AK Public-Use, Nov 2003 Sub (1973-2001 varying). National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch: Bethesda; released April 2004. http://www.seer.cancer.gov
Surveillance, Epidemiology, and End Results (SEER) Program SEER*Stat Database: Incidence - SEER 9 Regs Public-Use, Nov 2003 Sub (1973-2001). National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch: Bethesda; released April 2004 (cited 2004 August 17). http://www.seer.cancer.gov
Kramarova E, Stiller C, Ferlay J, et al: International Classification of Childhood Cancer: IARC Technical Report No. 29. Lyon, France, IARCPress, 1996, pp 48
Environmental Protection Agency. Web site. http://www.epa.gov/uvnet/
Scotto J, Fears TR, Fraumeni JF Jr: Solar Radiation, in Shottenfeld D, Fraumeni JF Jr (eds): Cancer Epidemiology and Prevention (ed 2). New York, NY, Oxford University Press, 1996, pp 355-372
Stata Statistical Software: Intercooled Stata 8.2. College Station, TX, Stata Corp, 2003
Rosner B: Fundamental of Biostatistics (ed 5). Pacific Grove, CA, Duxbury, 2000
Hamre MR, Chuba P, Bakhshi S, et al: Cutaneous melanoma in childhood and adolescence. Pediatr Hematol Oncol 19:309-317, 2002
Fears TR, Bird CC, Guerry D IV, et al: Average midrange ultraviolet radiation flux and time outdoors predict melanoma risk. Cancer Res 62:3992-3996, 2002
Demko CA, Borawski EA, Debanne SM, et al: Use of Indoor Tanning Facilities by White Adolescents in the United States. Arch Pediatr Adolesc Med 157:854-860, 2003
Loria D, Matos E: Risk factors for cutaneous melanoma: A case-control study in Argentina. Int J Dermatol 40:108-114, 2001
Gilchrest BA, Eller MS, Geller AC, et al: The pathogenesis of melanoma induced by ultraviolet radiation. N Engl J Med 340:1341-1348, 1999
Kaskel P, Sander S, Kron M, et al: Outdoor activities in childhood: A protective factor for cutaneous melanoma Results of a case-control study in 271 matched pairs. Br J Dermatol 145:602-609, 2001
Noonan FP, Recio JA, Takayama H, et al: Neonatal sunburn and melanoma in mice. Nature 413:271-272, 2001
Sondak VK, Taylor JMG, Sabel MS, et al: Mitotic rate and younger age are predictors of sentinel lymph node positivity: Lessons learned from the generation of a probabilistic model. Ann Surg Oncol 11:247-258, 2004
Wechsler J, Bastuji-Garin S, Spatz A, et al: Reliability of the histopathologic diagnosis of malignant melanoma in childhood. Arch Dermatol 138:625-628, 2002
Farmer ER, Gonin R, Hanna MP: Discordance in the histopathologic diagnosis of melanoma and melanocytic nevi between expert pathologists. Hum Pathol 27:528-531, 1996
Clegg LX, Feuer EJ, Midthune DN, et al: Impact of reporting delay and reporting error on cancer incidence rates and trends: JNCI Cancer Spectrum. J Natl Cancer Inst 94:1537-1545, 2002(John J. Strouse, Thomas R)