Survival and Surgical Outcomes for Pediatric Head and Neck Melanoma

Key Points

Question  How do survival, demographic, tumor, and treatment characteristics of pediatric head and neck melanoma compare with those of adult head and neck melanoma?

Findings  In this large cohort study, pediatric patients presented with different histologic subtypes, similar tumor depth, and more frequent nodal metastases but higher overall survival (5-year survival was 24% higher for pediatric vs adult patients).

Meaning  Pediatric head and neck melanoma presents differently than in adults; therefore, pediatric-specific survival estimates are more accurate than adult estimates for this patient population.

Importance  Melanoma in children is rare, accounting for approximately 2% of all pediatric malignant neoplasms. However, for the past 30 years, the incidence of melanoma in those younger than 20 years has been increasing. Location of the primary tumor has been shown to be an important prognostic factor, with melanomas of the scalp and neck conferring a worse prognosis than those originating at other sites.

Objective  To examine the survival, demographic, tumor, and treatment characteristics of pediatric head and neck melanoma.

Design, Setting, and Participants  We performed a retrospective cohort study using information from the National Cancer Data Base from January 1, 1998, to December 31, 2012, on pediatric (≤18 years) and adult (>18 years) patients with head and neck melanoma. Data analysis was conducted from August 1, 2015, to June 30, 2016.

Exposure  Pediatric age (≤18 years) at diagnosis of head and neck melanoma.

Main Outcomes and Measures  Survival differences were estimated using a Cox proportional hazards regression model. Surgical outcomes, including nodal sampling and margin status, were estimated with generalized linear models comparing pediatric and adult patients. Patient demographic, tumor, and treatment characteristics were estimated using t tests and χ² tests between pediatric and adult patients with head and neck melanoma for continuous and categorical data, respectively.

Results  Of the 84 744 patients with head and neck melanoma, 657 (0.8%) were 18 years or younger (mean [SD] age, 13.5 [4.7] years; 285 female and 372 male; 610 white). Pediatric and adult patients had similar demographics but different histologic subtypes (risk difference of pediatric vs adult patients: melanoma, not otherwise specified, 8.5% [95% CI, 4.7%-12.3%]; superficial spreading, 4.2% [95% CI, 0.89%-7.4%]; and lentigo maligna, –13.4% [95% CI, –14.1% to 12.6%]). Pediatric patients had tumors of similar mean depth to those in adult patients (pediatric, 1.54 mm; adult; 1.39 mm; absolute difference, 0.15 mm; [95% CI, –0.32 to 0.008]) and more frequent nodal metastases than did adult patients (risk difference of pediatric vs adult patients for stage T2, 23.9% [95% CI, 14.1%-33.6%]). Five-year survival among pediatric patients was higher for those with stage 1, 2, or 3 disease (absolute difference of pediatric vs adult patients: stage 1, 18% [95% CI, 9.7%-26.3%]; stage 2, 36% [95% CI, 25.3%-46.7%]; stage 3, 39% [95% CI, 26.8%-51.2%]; and stage 4, 2% [95% CI, –8.2% to 12.2%]).

Conclusions and Relevance  Although pediatric patients with head and neck melanoma present with similar tumor depth and more frequent nodal metastases than do adult patients, younger patients have higher overall survival.

Melanoma in children is rare, accounting for approximately 2% of all pediatric malignant neoplasms.¹ However, for the past 30 years, the incidence of melanoma in those younger than 20 years has been increasing by roughly 3% per year.^2-5 Location of the primary tumor has been shown to be an important prognostic factor, with melanomas of the scalp and neck conferring a worse prognosis than those originating at other sites.^6-8 Given the rarity of the disease in children, systematic study of presentation, prognostic factors, and clinical outcomes of pediatric head and neck melanoma remains scarce. Data from 33 patients at a single institution suggest that melanoma may present differently in pediatric patients than in adult patients, with more atypical clinical features (amelanotic and raised lesions), nodular histologic subtype, and thicker Breslow depth at the time of presentation.⁹ Another study of 39 pediatric patients with head and neck melanoma corroborated these findings: histologic subtypes of superficial spreading and nodular melanoma were more prevalent than the lentigo maligna histologic subtype most commonly observed in adults with head and neck melanoma.¹⁰ The lack of knowledge about common differences in presentation between pediatric and adult patients could result in delayed diagnoses of deadly lesions and overtreatment of those that are benign. Therefore, we compared the tumor characteristics and overall survival of a large cohort of pediatric and adult patients with head and neck melanoma.

We performed a retrospective cohort study comparing pediatric with adult invasive head and neck melanoma using the National Cancer Data Base (NCDB). Pediatric patients 18 years or younger were compared with adults older than 18 years. All of the data made available to us by the NCDB ranged from January 1, 1998, to December 31, 2012. No duplicate cases were included. Head and neck disease location was based on the NCDB primary site classifications, including skin of the lip (code C440); eyelid (C441); external ear (C442); cheek, chin, face, and forehead (C443); and scalp and neck (C444). This study was approved by the Seattle Children’s Hospital institutional review board, who waived the need for informed consent.

The NCDB began data collection in 1989 and is cosponsored by the American College of Surgeons Commission on Cancer and the American Cancer Society.¹¹ Currently the NCDB captures approximately 70% of all newly diagnosed cases of cancer and contains more than 30 million records. Data are collected from more than 1500 hospitals nationwide that are accredited by the Commission on Cancer. The intended use of the NCDB is for the longitudinal examination of cancer care in the United States to aid quality improvement and quality assurance efforts.¹² The database is estimated to include 42% of all childhood malignant neoplasms.¹³ Trained data abstractors use standard data definitions to extract data from patients’ records.

Demographic characteristics included sex, race/ethnicity, insurance status, household income, educational level, and rural vs urban place of residence. Trained abstractors documented the patients’ race/ethnicity based on the medical record and according to standardized classifications within the NCDB data dictionary.¹⁴ Insurance status was defined as private, governmental, or uninsured. Household income was divided into quartiles based on the patient’s home zip code. Educational level was based on the percentage of high school graduates in the patient’s home zip code. Rural vs urban place of residence was based on the population density and proximity to a metropolitan area of the patient’s home zip code.

Tumor characteristics included histologic subtype, stage of disease, node positivity, presence of ulceration, and tumor depth. Histologic subtype was based on the International Classification of Diseases for Oncology, Third Edition (ICD-O-3) codes¹⁵ and included malignant melanoma, not otherwise specified (code 8720/3), superficial spreading (8743/3), lentigo maligna (8742/3), acral lentiginous (8744/3), and other (8722/3, 8723/3, 8730/3, 8740/3, 8741/3, 8745/3, 8746/3, 8761/3, 8770-4/3, 8780/3, and 8790/3). The NCDB uses the ICD-O-3 coding manual, which does not specify a histologic subtype code for spitzoid melanoma; however, we included only invasive disease in our population. Disease staging was based on the National Comprehensive Cancer Network TNM classification.¹⁶ The NCDB reports a combined stage that uses the pathologic stage of disease unless these data are unavailable, in which case the clinical stage is used to obtain the most accurate staging data. Node positivity and presence of ulceration were binary variables. Tumor depth was measured in millimeters as a continuous variable.

Treatment factors included nodal sampling, margin status, and receipt of immunotherapy. Nodal sampling was categorized as performed or not performed. The NCDB records the number of regional lymph nodes that were removed and examined by a pathologist. The indication for removal, including sentinel node biopsy, elective lymph node dissection, or complete dissection for clinically node-positive disease, is not recorded. For the purpose of this study, we considered any patient who had lymph nodes examined by a pathologist to have undergone node sampling. Margin status was defined as either positive or negative. Immunotherapy was assessed as received or not received.

Data analysis was conducted from August 1, 2015, to June 30, 2016. To perform the analyses, pediatric patients 18 years or younger were compared with the adult population (>18 years) using univariate statistics with the χ² test for categorical data and t tests for continuous data. P < .05 was considered statistically significant for all analyses. An adjusted incidence rate ratio was estimated using generalized linear models with a log link function and Poisson distribution using robust standard errors. Five- and 10-year survival estimates were calculated using hazard functions and displayed using Kaplan-Meier survival curves. An adjusted Cox proportional hazards regression model was used to estimate differences in all-cause survival for pediatric and adult patients. Adjustment factors were determined a priori to adjust for differences between groups such as socioeconomic status and included sex, income, educational level, race/ethnicity, insurance, urban vs rural place of residence, and histologic subtype. Statistical analysis was completed using Stata statistical software, version 12 (StataCorp LP).

Of the 84 744 patients with invasive head and neck melanoma, 657 (0.8%) were 18 years or younger. The median age in the pediatric cohort was 15 years with an interquartile range of 12 to 17 years. The mean (SD) age of the pediatric patients was 13.5 (4.7) years. The median age in the adult cohort was 68 years (interquartile range, 55-78 years); the mean (SD) age was 65.4 (16.1) years. The risk difference (RD) in the proportion of patients who were female between the pediatric (43.3%) and adult populations (28.5%) was 14.9% (95% CI, 11.1%-18.7%). More pediatric patients had private insurance vs adult patients (RD, 37.9%; 95% CI, 34.9%-40.9%), but there were no significant differences between the groups in terms of race/ethnicity, income, educational level, and rural vs urban place of residence (Table 1).

Pediatric patients were more frequently identified as having malignant melanoma, not otherwise specified (RD, 8.5%; 95% CI, 4.7%-12.2%), or superficial spreading (RD, 4.2%; 95% CI, 0.9%-7.4%). In addition, pediatric patients presented more often with stage 3 disease (RD, 16.5%; 95% CI, 12.7%-20.3%) and similar tumor depth to that seen in adult patients (pediatric, 1.54 mm; adult, 1.39 mm; absolute difference, 0.15 mm [95% CI, –0.32 to 0.008]), and less frequently with histologically classified ulceration (RD, –4.6%; 95% CI, –8.4% to –0.8%). Pediatric patients were also more likely to undergo immunotherapy relative to adult patients (RD, 13.3%; 95% CI, 10.4%-16.2%) (Table 1).

Pediatric patients, regardless of stage, were more likely to have undergone nodal sampling compared with adult patients (RD, 25.2%; 95% CI, 21.7%-28.8%) (Table 2). Surgical management was less likely to produce positive margins in pediatric patients compared with adult patients for stage T1 (RD, –3.3%; 95% CI, –5% to –1.8%); however, there was no statistically significant difference in margin status after resection for any other T stage (Table 2).

When adjusting for covariates, including sex, income, educational level, race/ethnicity, insurance, urban vs rural place of residence, histologic subtype, and T stage, pediatric patients were 35% more likely to have undergone nodal sampling than were adult patients (relative risk, 1.35; 95% CI, 1.28-1.42). They were also almost twice as likely to have positive nodes found at resection relative to adult patients (relative risk, 1.98; 95% CI, 1.72-2.27). There was no difference in margin status between the groups (relative risk, 0.72; 95% CI, 0.47-1.11). Five- and 10-year unadjusted survival rates by stage were significantly higher for pediatric patients compared with adult patients, except for those who presented with stage 4 disease (RD for pediatric vs adult overall 5-year survival, 24%; 95% CI, 13.4%-34.6%; RD for pediatric vs adult overall 10-year survival, 39%; 95% CI, 27.3%-50.7%) (Table 3 and Figure). The adjusted risk of mortality was significantly lower for pediatric patients compared with adult patients (hazard ratio, 0.30; 95% CI, 0.22-0.39) when controlling for sex, income, educational level, race/ethnicity, insurance, urban vs rural place of residence, and histologic subtype.

To our knowledge, this investigation is the largest study to date of pediatric head and neck melanoma, and the first to directly compare outcomes between pediatric and adult patients. The pediatric patients in this study presented with more advanced disease than did adult patients. The mean depth of tumor invasion was similar in pediatric and adult patients, but the frequency of ulceration noted histologically was significantly lower in pediatric patients than in adult patients. In addition, nearly 30% of pediatric patients presented with stage 3 or 4 disease compared with almost 15% of adult patients. There are several hypotheses to explain these findings. Melanoma is less common in pediatric patients, lowering the index of suspicion among primary care physicians. Furthermore, performing a skin biopsy on a young child may require sedation or general anesthesia, which could raise a clinician’s threshold to perform a biopsy, resulting in delay.

Previous studies have suggested that pediatric melanoma has a higher proportion of clinically atypical features, specifically amelanotic, pedunculated, and raised nodular lesions.^9,17 We observed a higher proportion of melanoma, not otherwise specified, and “other” histologic subtypes in pediatric patients, which may indicate a less straightforward diagnosis.

The surgical treatment of pediatric head and neck melanoma tended to be more aggressive than that for adult patients. More pediatric patients underwent lymph node sampling than did adult patients, either by sentinel node biopsy or neck dissection. In addition, significantly more pediatric patients received immunotherapy compared with adult patients. The NCDB does not document the incidence of satellite lesions or in-transit metastases, so we were unable to characterize these differences between pediatric and adult patients.

Finally, we found that clinical outcomes in pediatric patients were better than those in adult patients. Pediatric patients had double the frequency of positive nodes compared with adult patients. Five- and 10-year survival in pediatric patients was almost double that seen in adult patients. This finding contradicts data from a study of pediatric patients with melanoma (not limited to the head and neck) in which pediatric patients were matched with adults with melanoma of equivalent depth of invasion¹⁸; the study showed no significant difference in survival between groups. The NCDB does not document disease-specific mortality, so it would make sense that all-cause mortality would be greater in adult than pediatric patients. However, the fact that pediatric patients, more than half of whom presented with stage 2 or higher disease, had rates of 10-year survival approaching 90% is notable.

The NCDB was established in 1989 to document presentation, treatment, and outcomes of all cancer diagnoses in the United States. At present, the database captures 70% of all new cancer cases and includes more than 30 million records.¹¹ Prior studies of pediatric head and neck melanoma were typically based on single-institution experiences and thus subject to the bias of a single institution’s approach to diagnosis and treatment.^9,12,13 The data in our study were compiled from 657 pediatric patients and more than 85 000 adult patients from dozens of hospitals throughout the United States and therefore were less subject to such bias.

There are limitations to our study. It was a retrospective observational cohort study, so while association may be assessed, causation may not be inferred. The data were limited in that we were unable to distinguish between all-cause and disease-specific mortality, which would be useful when comparing pediatric and adult patients. We were also unable to determine whether there was a difference in margins resected (1-2 and >2 cm) for pediatric and adult patients as a result of a significant amount of missing data. The NCDB lacks a specific histologic subtype code for spitzoid melanoma, which could bias the pediatric population toward less aggressive disease; however, we performed a sensitivity analysis excluding patients with melanoma, not otherwise specified, and melanoma, other, and found similar outcomes. We were also unable to calculate the incidence of melanoma in our patient populations since the NCDB is not a population-based registry. The type of facility was not available so we were unable to determine if pediatric patients were treated at children’s hospitals or general adult hospitals.

Our study describes the clinical outcomes of the largest population to date of pediatric patients with head and neck melanoma. In comparing pediatric and adult patients, we found that pediatric patients presented at a more advanced stage of disease, with tumors of similar depth, but overall, their outcomes were superior. Possible explanations for more advanced disease at diagnosis include atypical clinical and histologic features and delay in biopsy. Despite a prevalent supposition that children do not die of melanoma, our research found this belief to be false. Whether this finding is associated with melanoma itself or complications of treatment is yet to be determined. Further studies will be needed to better elucidate the pathophysiological differences responsible for better survival outcomes in pediatric patients with head and neck melanoma.

Corresponding Author: Morgan K. Richards, MD, MPH, Division of Pediatric General and Thoracic Surgery, Seattle Children’s Hospital, 4800 Sand Point Way NE, PO Box OA.9.220, Seattle, WA 98105 (mkr19@u.washington.edu).

Accepted for Publication: July 15, 2016.

Published Online: September 8, 2016. doi:10.1001/jamaoto.2016.2630

Author Contributions: Drs Richards and Gow had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Richards, Czechowicz, Goldin, Gow, Doski, Langer, Beierle, Parikh.

Acquisition, analysis, or interpretation of data: Richards, Goldin, Gow, Doski, Goldfarb, Nuchtern, Vasudevan, Gupta.

Drafting of the manuscript: Richards, Czechowicz, Gow.

Critical revision of the manuscript for important intellectual content: Richards, Goldin, Gow, Doski, Goldfarb, Nuchtern, Langer, Beierle, Vasudevan, Gupta, Parikh.

Statistical analysis: Richards, Gow, Goldfarb.

Administrative, technical, or material support: Goldin, Gow, Doski.

Study supervision: Goldin, Gow, Goldfarb, Nuchtern, Parikh.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.