SEARCH

SEARCH BY CITATION

Keywords:

  • melanoma;
  • ethnic groups;
  • minority health;
  • statistical model;
  • New Zealand

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONFLICT OF INTEREST DISCLOSURES
  7. REFERENCES

BACKGROUND:

Thickness is the major prognostic indicator for patients with melanoma. In many countries, the incidence of thick melanoma has not decreased. To reduce mortality, knowledge of the characteristics associated with melanoma depth is needed.

METHODS:

To examine the relation between melanoma thickness and other factors in Europeans, Maori, and Pacific peoples, the authors analyzed the 14,802 melanoma registrations in New Zealand between 1996 and 2006. Notifications of invasive cutaneous melanoma from 1996 to 2006 were extracted from the New Zealand Cancer Registry. Ethnicity was categorized using the Statistics New Zealand prioritization algorithm. The geometric mean tumor thickness was calculated using log-transformed Breslow depth. Multivariate linear regression was used to examine the relation of predictor variables, their interactions, and melanoma thickness.

RESULTS:

Melanoma thickness increased by 1% per year from 1996 to 2006. Although melanoma is rare in Maori and Pacific peoples, after adjustment, melanoma thickness was significantly greater in those populations compared with Europeans. Among Europeans, melanoma thickness was associated significantly with age at diagnosis, year of registration, sex, histology subtype, and extent of disease at diagnosis. In Maori, thick melanomas were likely to be nodular and to have regional or lymph node spread. The most important determinant of melanoma thickness in Pacific peoples was body site.

CONCLUSIONS:

Differences in melanoma thickness among ethnic groups were not explained fully by tumor subtype, site, or extent of disease. The current results indicated that the thicker melanomas in darker skinned populations probably can be accounted for in part by more aggressive lesions. Research is needed to identify additional characteristics to explain ethnic variations in melanoma thickness. Cancer 2011. © 2010 American Cancer Society.

Melanoma incidence has increased in white-skinned populations in many countries over the past decade. New Zealand has 1 of the highest age-standardized registration rates of melanoma in the world (47.7 per 100,000 population in 2006), and melanoma registration has increased in New Zealand Europeans in the last 10 years.1 Although melanoma is much less common in darker skinned populations worldwide,2-4 melanoma incidence also has increased in New Zealand Maori over the same time.1

Breslow thickness is still the major predictor of prognosis in cutaneous melanoma. Little improvement has been made in the treatment of thick melanoma, and its detection at an early stage while still thin remains the best way to improve outcomes for this disease.5 In the last few decades, melanoma thickness has decreased in several countries; however, despite public health campaigns promoting prevention and early diagnosis in many countries, including New Zealand, the incidence of thick melanoma has not decreased.6-8 Worldwide, an increased incidence of thicker melanomas has been observed in some population groups.2, 9

In New Zealand, there is considerable variability in melanoma thickness at diagnosis for different ethnic groups.1, 8 It is unknown whether these differences within and among ethnic groups are caused by delays in diagnosis, different biologic behaviors or host responses for similar lesions in different ethnicities, or the development of inherently different histologies with different biology. Detailed investigation of the factors related to melanoma thickness and histologic subtypes in different ethnic groups may help to direct further work on the pathogenesis and host responses in melanoma and particularly in nodular and other thick melanomas. To reduce mortality from melanoma, knowledge of the characteristics associated with thick melanoma is also needed to better target early diagnosis and prevention campaigns. For the current study, we analyzed the relation between melanoma thickness and other histologic and clinical variables in Europeans, Maori, and Pacific peoples using data from 14,802 diagnoses of cutaneous melanoma registered by the New Zealand Cancer Registry between 1996 and 2006.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONFLICT OF INTEREST DISCLOSURES
  7. REFERENCES

The New Zealand Cancer Registry, part of the New Zealand Health Information Service, has collected data since 1948 about all incident malignant cancers excluding nonmelanoma skin cancer.10 Notification of pathology reports of new cancer cases has been mandatory since July 1994 after the introduction of the Cancer Registry Regulations in 1994.

In 2007, all notifications of primary cutaneous melanoma, from 1996 to 2006 with an International Classification of Diseases, 9th Revision code of 172 or an International Classification of Diseases, 10th Revision code of C43 (International Classification of Diseases for Oncology codes M8720/3, M8721/3, M8742/3, M8743/3, M8744/3, or M8745/3) were extracted. For each individual, the first primary melanoma that occurred in the time window was selected. For multiple melanomas that were diagnosed in the same individual on the same date, the deepest melanoma was chosen as the primary melanoma, because this was most likely to determine prognosis. Data on age at diagnosis, sex, ethnic group, date of diagnosis, Breslow thickness at diagnosis in millimeters,11 histologic subtype, body site, and extent of disease at diagnosis were extracted from the Cancer Registry record.

The New Zealand Health Information Service collects ethnicity for each registration from data providers, the National Minimum Dataset of hospital events, or the National Health Index (a database of unique identifiers and demographics).12 The ethnicity data collected by Statistics New Zealand in the census indicate the self-identified ethnicity in which individuals select up to 3 ethnic groups.13 Ethnicity data collected from the health sector use the same question as the census. Respondents are then categorized into prioritized ethnic groups using the Statistics New Zealand algorithm: Each respondent is allocated to a single ethnic group using the priority system (Maori first, then Pacific peoples, then Asians, then other groups except New Zealand Europeans, and finally New Zealand Europeans). Registrations with no ethnicity stated have been excluded from analyses using ethnic group.

Age at diagnosis was used as a categorical variable (ages <40 years, 40-49 years, 50-59 years, 60-69 years, 70-79 years, and ≥80 years) or as a continuous variable. For Pacific peoples, only 4 categories (ages <40 years, 40-49 years, 50-69 years, and ≥70 years) were used. The 4 major histologic subtypes of melanoma (Hutchinson melanotic freckle [HMF] or lentigo maligna, superficial spreading melanoma [SSM], nodular melanoma [NM], and acral lentiginous melanoma [ALM]) and melanoma not otherwise specified (MMNOS) were included in analyses of histologic subtype. Body site was classified as face, scalp or neck, trunk, upper limb including the shoulder, and lower limb including the hips and feet. Extent of disease at diagnosis was coded according to the US National Cancer Institute's Surveillance, Epidemiology, and End Results summary system as in situ, localized, regional (direct extension or lymph node involvement), or distant (metastatic).14 In situ melanomas and melanomas that were diagnosed on the basis of metastases (which do not usually have a depth recorded for the primary lesion) were excluded.

Statistical Methods

All statistics were computed using Stata version 10 (StataCorp, College Station, Tex).15 Because Breslow thickness was not distributed normally, it was log-transformed before regression analysis and was included as a continuous variable. Despite log-transformation, some very high values of Breslow thickness (≥37 mm) remained as outliers (identified using Studentized residuals). These 12 records (11 Europeans and 1 Pacific individual) were excluded from all analyses. The geometric mean (GM) tumor thickness (an approximation of the median tumor thickness) and 95% confidence interval (CI) were calculated.

A linear regression model was fitted to the data with log-transformed Breslow thickness as the dependent variable. Associations are reported as GM ratios (GMRs), which were calculated by taking the exponential of the relevant regression coefficients. P values were 2-sided (α = .05). Multiple linear regression models were used to examine the relation of predictor variables, their interactions, and melanoma thickness. Potential 2-way interaction terms were tested individually for their significance in each model. Final models included all potential confounding factors that were associated with tumor thickness in univariate analyses (P ≤ .10) as well as any significant interaction terms. All univariate and interaction terms were assessed for inclusion in the model by backward stepwise selection using the Akaike information criterion as the stay criterion.16 For models that were tied using Akaike information criterion, the most parsimonious model was selected. Using likelihood-ratio statistics as the criteria to assess the importance of explanatory variables produced the same results. The adjusted coefficient of determination (R2) was used to assess the proportion of variation in thickness explained by factors that were included in the model.17

For continuous predictors of melanoma depth, such as age at diagnosis, the GMR is the ratio of the GM of melanoma thickness for a 1-unit increase in the predictor expressed as a percentage. For associations with categorical variables, the parameter estimate is the GMR relative to the reference category. Records that were missing values for any predictor variable or for the dependent variable were excluded from each model. For a graphic display of categorical covariate interactions with age at diagnosis (as a continuous variable) adjusted for confounding factors, regression lines for back-transformed melanoma thickness on age were plotted for separate covariate subgroups.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONFLICT OF INTEREST DISCLOSURES
  7. REFERENCES

From 1996 to 2006, there were 19,149 first registrations of melanoma in New Zealand with an age range of 2 to 101 years. Ninety-nine percent (n = 18,998) of these melanomas were verified by histology. Approximately 8.6% (n = 1643) were missing data for thickness. Of those with thickness data (n = 17,506), 14.8% (n = 2584) were missing data for ethnicity. Missing data for thickness varied by ethnic group, with 9.1% missing thickness data among Europeans, 14% missing thickness data among Maori, and 22.9% missing thickness data among Pacific peoples. After excluding the 12 outliers and melanomas that were diagnosed on the basis of metastases, 14,646 Europeans, 131 Maori, and 25 Pacific peoples remained.

Melanomas occurred most commonly in Europeans aged 70 to 79 years; whereas, in Maori, almost 33% occurred in individuals aged <40 years (Table 1). Melanoma occurred in similar proportions in men and women among Europeans and Pacific peoples; however, among Maori, women accounted for 65.6% of diagnoses. Melanoma occurred most commonly on the trunk (31.1%) followed by the lower limb (27.7%) in Europeans; in Maori, there were similar proportions on the trunk and lower limb (29.8% and 32.8%, respectively); whereas, in Pacific peoples, most melanomas (60%) developed on the lower limb. Over 30% of melanomas in Pacific peoples were ALM; whereas, in Maori and Europeans, only 3.9% and 0.7%, respectively, were ALM. In Maori, 36.7% of registered melanomas did not have a melanoma subtype listed. In Europeans, 93.8% of melanomas were localized at diagnosis compared with only 70.8% in Pacific peoples.

Table 1. Geometric Mean Melanoma Thickness (mm) for Clinical and Histologic Factors by Ethnic Group
VariableEuropeans, n=14,646Maori, n=131Pacific Peoples, n=25a
%GM (95% CI)%GM (95% CI)%GM (95% CI)
  • GM indicates geometric mean; CI, confidence interval; HMF, Hutchinson melanotic freckle; SSM, superficial spreading melanoma; NM, nodular melanoma; ALM, acral lentiginous melanoma; MMNOS, melanoma not otherwise specified.

  • a

    Some subcategories have been collapsed, because there were too few records with Breslow thickness provided to calculate the GM.

Overall 0.91 (0.89-0.92) 1.28 (1.05-1.55) 2.06 (1.26-3.38)
Age at diagnosis, y
 <4013.30.72 (0.69-0.74)31.30.78 (0.61-1.01)24.00.86 (0.41-1.77)
 40-4915.20.74 (0.71-0.76)14.51.48 (0.86-2.57)28.02.69 (0.88-8.16)
 50-5918.20.81 (0.79-0.84)16.01.35 (0.77-2.36)36.0}2.33 (0.80-6.76)
 60-6919.30.89 (0.86-0.93)22.11.83 (1.18-2.83)  
 70-7920.51.02 (0.98-1.05)10.71.22 (0.58-2.53)12.0}4.45 (0.69-28.72)
 ≥8013.41.43 (1.36-1.50)5.33.11 (0.98-9.87)  
Sex
 Men49.70.97 (0.95-0.99)34.41.49 (1.03-2.15)56.02.19 (1.07-4.49)
 Women50.20.84 (0.83-0.86)65.61.18 (0.93-1.48)44.01.91 (0.85-4.26)
Body site
 Face12.71.06 (1.01-1.11)8.41.29 (0.64-2.59)8.0}6.63 (0.61-71.71)
 Scalp/neck6.41.23 (1.15-1.32)6.93.01 (1.15-7.93)  
 Trunk31.10.82 (0.79-0.84)29.80.86 (0.60-1.22)24.00.76 (0.20-2.90)
 Upper limb22.10.87 (0.85-0.90)22.11.77 (1.16-2.68)8.02.10 (1.14-3.84)
 Lower limb27.70.91 (0.89-0.94)32.81.22 (0.88-1.70)60.02.62 (1.43-4.81)
Histology subtype
 HMF7.00.64 (0.60-0.68)3.10.33 (0.16-0.68)29.2}1.54 (0.41-5.82)
 SSM48.10.71 (0.70-0.72)38.30.85 (0.67-1.09)  
 NM11.52.95 (2.85-3.06)18.03.96 (2.83-5.53)20.86.63 (2.51-17.53)
 ALM0.71.46 (1.22-1.75)3.92.93 (1.08-7.97)33.31.86 (0.92-3.77)
 MMNOS32.60.85 (0.83-0.87)36.71.14 (0.81-1.61)16.70.69 (0.33-1.46)
Extent of disease
 Localized93.80.81 (0.80-0.82)85.90.97 (0.81-1.17)70.81.51 (0.82-2.78)
 Regional or lymph nodes6.24.61 (4.39-4.83)14.15.50 (3.71-8.17)29.25.03 (2.40-10.56)

The GM melanoma thickness was 0.91 mm in Europeans, 1.28 mm in Maori, and 2.06 mm in Pacific peoples (Table 1). Melanomas >4 mm thick consisted of 50% NM, 26% MMNOS, 18% SSM, 4% HMF, and 2% ALM (data not shown). The deepest melanomas among Europeans and Maori occurred in patients aged >80 years (median thickness, 1.43 mm and 3.11 mm, respectively), were located on the scalp and neck (median thickness, 1.23 mm and 3.01 mm, respectively), were of the nodular subtype (median thickness, 2.95 mm and 3.96 mm, respectively), and had regional or lymph node spread at diagnosis (median thickness, 4.61 mm and 5.50 mm, respectively). Among Pacific peoples, the thickest lesions occurred in patients aged >70 years (median thickness, 4.45 mm) or were located on the head and neck (median thickness, 6.63 mm).

In univariate linear regression, age at diagnosis, year of registration, sex, ethnic group, histologic subtype, body site, and extent of disease at diagnosis all were associated significantly with melanoma thickness (Table 2), and thickness increased significantly with age and year of registration (by 1% per year of age or year of registration; P < .001 for each). The median melanoma thickness was 41% greater in Maori compared with Europeans (P < .001); and, in Pacific peoples, it was more than double the median thickness in Europeans (GMR, 2.28; P < .001). Melanomas were significantly thinner in women compared with men (GMR, 0.87; P < .001).

Table 2. Univariate Analysis and Main Effects From Multiple Linear Regression Models for Geometric Mean Melanoma Thickness and Explanatory Variables
VariableNo.Crude GMRaPAdjusted GMRbP
  • GMR indicates geometric mean ratio (the ratio of the geometric mean relative to the reference group); ref, reference group; SSM, superficial spreading melanoma; HMF, Hutchinson melanotic freckle; NM, nodular melanoma; ALM, acral lentiginous melanoma; MMNOS, melanoma not otherwise specified; R2, coefficient of determination.

  • a

    Unadjusted GMR.

  • b

    All factors are adjusted for each other. The model also includes 3 two-way interaction terms: sex and age at diagnosis; body site and age at diagnosis; and histology subtype and extent of disease at diagnosis.

Age at diagnosis14,8021.01<.0011.00.712
Registration year14,8021.01<.0011.01<.001
Ethnic group
 European (ref)14,6461.00 1.00 
 Maori1311.41<.0011.23.002
 Pacific peoples252.28<.0011.46.015
Sex
 Men (ref)73361.00 1.00 
 Women74660.87<.0010.79<.001
Body site
 Trunk (ref)45671.00 1.00 
 Face18611.30<.0010.95.606
 Scalp/neck9381.53<.0011.00.975
 Upper limb32391.08<.0010.84.008
 Lower limb40771.12<.0010.91.108
Histology subtype
 SSM (ref)68861.00 1.00 
 HMF10040.90<.0010.52<.001
 NM16624.19<.0017.26<.001
 ALM1112.18<.0012.14.001
 MMNOS46721.20<.0011.08.328
Extent of disease
 Localized (ref)13,7491.00 1.00 
 Regional or lymph nodes9225.69<.0014.53<.001
Adjusted R2   0.3755 

The multivariate model (Table 2) included all 7 factors that were significant in univariate analysis along with 3 interaction terms (age at diagnosis and sex, age at diagnosis and body site, and histology subtype and extent of disease at diagnosis). After adjustment for all other factors and interaction terms, all variables that were identified in univariate analyses remained significantly associated with melanoma thickness, although only approximately 40% of the variability in thickness (R2 = 0.3755) was explained by these factors. After adjustment, the median melanoma thickness still was significantly greater in Maori and Pacific peoples compared with Europeans (GMR, 1.23 [P = .002] and GMR, 1.46 [P = .015], respectively). Overall, melanoma thickness increased by 1% per year (P < .001) from 1996 to 2006 (Table 2), resulting in an increase in median thickness of 11% over this period. Women still had significantly thinner melanomas compared with men (GMR, 0.79; P < .001). Scalp and neck melanomas were no longer significantly deeper than melanomas in other sites (GMR, 1.00). NMs were >7 times deeper than SSMs (P < .001), whereas HMFs were 48% thinner than SSMs (GMR, 0.52; P < .001). Melanomas that had regional or lymph node spread at diagnosis were 4.53 times thicker than localized melanomas (P < .001).

Univariate and Multivariate Associations of Tumor Thickness and Explanatory Factors for Each Ethnic Group

Because the difference in median thickness between Maori and Pacific peoples remained after adjustment for multiple factors and their interactions, models for the associations of thickness and explanatory factors were developed separately for each ethnic group. Europeans made up the vast majority of the total sample, so the results for Europeans were almost the same as those for the population overall (Table 3). The European multivariate model included all 6 factors that were significant in univariate analysis (age at diagnosis, sex, year of registration, histology subtype, body site, and extent of disease at diagnosis) and 4 interaction terms (age at diagnosis and sex, age at diagnosis and body site, histology subtype and extent of disease at diagnosis, and histology subtype and body site).

Table 3. Univariate Analysis and Main Effects From Multiple Linear Regression Models for Geometric Mean Melanoma Thickness and Explanatory Variables in Europeans, Maori, and Pacific Peoples
VariableEuropeansMaoriPacific Peoples
No.Crude GMRaPGMRbPNo.Crude GMRaPGMRcPNo.Crude GMRaPGMRdP
  • GMR indicates geometric mean ratio; ref, reference group; SSM, superficial spreading melanoma; HMF, Hutchinson melanotic freckle; NM, nodular melanoma; ALM, acral lentiginous melanoma; MMNOS, melanoma not otherwise specified; R2, coefficient of determination.

  • a

    The ratio of the geometric mean relative to the reference group.

  • b

    The model for Europeans includes all listed factors and 4 two-way interaction terms: sex and age at diagnosis; body site and age at diagnosis; histology subtype and extent of disease at diagnosis; and histology subtype and body site. All factors are adjusted for each other.

  • c

    The Maori model includes the interaction of age at diagnosis and body site. All factors are adjusted for each other.

  • d

    The Pacific model includes the interaction of histology subtype and body site. All factors are adjusted for each other.

Age at diagnosis14,6461.01<.0011.003<.0011311.02.0010.99.147251.030.054 
Registration year14,6461.01<.0011.01<.0011311.05.118 250.930.325 
Sex
 Men (ref)72771.00 1.00 451.00  141.00  
 Women73690.87<.0010.79<.001860.79.257  110.870.779  
Body site
 Trunk (ref)45221.00 1.00 391.00 1.00 61.00 1.00 
 Face18491.30<.0010.90.256111.50.2820.81.912110.500.06521.57}.002
 Scalp/neck9281.51<.0010.97.73993.51.0020.49.48417.220.117  
 Upper limb32081.07.0010.83.004292.06.0080.31.08322.750.2805.93.046
 Lower limb40191.12<.0010.92.200431.43.1450.18.007153.440.0337.69.025
Histology subtype
 SSM (ref)68311.00 1.00 491.00 1.00 61.00 1.00 
 HFM9990.90<.0010.74<.00140.39.0640.39.04216.840.07821.57.002
 NM16344.17<.0013.76<.001234.64<.0012.38.00155.670.0070.69.711
 ALM982.08<.0011.45<.00153.44.0081.60.26581.590.3770.65.577
 MMNOS46211.20<.0011.11<.001471.34.1421.05.77740.590.4091.87.346
Extent of disease
 Localized (ref)13,6221.00 1.00 1101.00 1.00 171.00  
 Regional or lymph nodes8975.68<.0014.63<.001185.65<.0012.84<.00173.340.023  
 Adjusted R2   0.3744    0.4615    0.6551 

Among Europeans, the association of melanoma thickness with age at diagnosis was modified by both sex (P for interaction <.001) and body site (P for interaction = .001). After adjustment for year of registration, histology subtype, and extent of disease at diagnosis, melanomas were thinner in women compared with men up to age 80 years, but thickness increased more with age among women than among men (approximately 0.007 mm and 0.005 mm per year of age from ages 20 to 80 years, respectively) (Fig. 1). After adjustment, melanomas on the scalp and neck were the thickest lesions before age 80 years and increased in thickness by 0.0063 mm per year of age, from a median of 0.75 mm at age 20 years to 1.13 mm at age 80 years (Fig. 2). The median thickness of lower limb melanomas increased most with age (by 0.0078 mm per year of age), from 0.66 mm at age 20 years to 1.13 mm at age 80 years.

thumbnail image

Figure 1. Illustration of geometric mean melanoma thickness (mm) in Europeans by age at diagnosis for both sexes adjusted for year of registration, histology subtype, extent of disease at diagnosis, and body site.

Download figure to PowerPoint

thumbnail image

Figure 2. Illustration of geometric mean melanoma thickness (mm) in Europeans by age at diagnosis for different body sites adjusted for year of registration, sex, histology subtype, and extent of disease at diagnosis.

Download figure to PowerPoint

Although increasing age was associated with increased thickness among Maori in univariate analyses (P = .001), no effect was observed after adjustment for covariates (P = .147) (Table 3). In Maori, melanomas on the scalp and neck or upper limb were significantly thicker than those on the trunk (GMR, 3.51 [P = .002] and GMR, 2.06 [P = .008], respectively) in univariate analysis only. In multivariate analysis, lower limb melanomas were significantly thinner than truncal melanomas (GMR, 0.18; P = .007). Thickness was associated significantly with histology subtype and extent of disease at diagnosis in both univariate and multivariate analyses: NMs were significantly thicker than SSMs (adjusted GMR, 2.38; P = .001), HMFs were significantly thinner than SSMs (adjusted GMR, 0.39; P = .042), and the adjusted median thickness of melanomas with lymph node or regional spread was 2.8 times greater than that for localized melanomas (P < .001). In multivariate regression analyses, almost 50% of the variability in melanoma thickness (R2 = 0.4615) among Maori was accounted for by factors that were included in the model.

Among Maori, melanomas on the scalp and neck and on the limbs increased in thickness with age at diagnosis, whereas the thickness of melanomas on the face and trunk decreased with age (Fig. 3). The greatest increase in thickness with age was observed for melanomas on the lower limb, which increased from 0.59 mm at age 20 years to 2.90 mm at age 80 years.

thumbnail image

Figure 3. Illustration of geometric mean melanoma thickness (mm) in the Maori population by age at diagnosis for different body sites adjusted for histology subtype and extent of disease at diagnosis.

Download figure to PowerPoint

Among Pacific peoples, the increase in thickness with age was of borderline significance (3% per year of age; P = .054) (Table 3), but no effect of age was observed after adjustment. In Pacific peoples, the median thickness of HMFs was >21 times greater than that of SSMs; however, although the difference was statistically significant (P = .002), this was based on only 1 very thick HMF. Melanomas on the head and neck were the thickest lesions, but this was based on only 2 melanomas. Most melanomas among Pacific peoples occurred on the lower limb (all 7 ALMs and 3 of 5 NMs were on the lower limb), and these lesions were >7 times thicker than truncal melanomas (P = .025). In Pacific peoples, about 66% of the variation in melanoma thickness was accounted for by body site and histology subtype.

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONFLICT OF INTEREST DISCLOSURES
  7. REFERENCES

Maori and Pacific peoples in New Zealand have a lower incidence of melanoma than New Zealand Europeans, but their melanomas tend to be thicker and more extensive at diagnosis.1, 8 To our knowledge, it has not been determined previously whether the histologic and clinical factors associated with melanoma thickness vary by ethnicity. The current large, population-based study of routinely collected national data from 14,802 patients who had melanoma diagnosed between 1996 and 2006 provides the first published details of factors associated with melanoma thickness in the predominantly light-skinned New Zealand Europeans and 2 darker skinned minority ethnic groups—Maori and Pacific peoples.

Since 1996, melanoma thickness has increased significantly by about 1% per year in Europeans. After adjustment for multiple factors and interactions, melanomas in Maori and Pacific peoples, although uncommon, still were significantly thicker than melanomas in Europeans. Melanoma thickness in Europeans was associated significantly with age at diagnosis, year of registration, men, histology subtype, and extent of disease at diagnosis. Significant interactions were identified between age and sex, age and body site, histology subtype and extent of disease at diagnosis, and histology subtype and body site. The increased melanoma thickness with increasing age among Europeans varied by sex and body site of melanoma, and the greatest increases with age occurred among women and in patients who had melanomas of the lower limb. Among Maori, there was a nonsignificant increase in thickness over time, and their thick melanomas were most likely to be nodular and to have regional or lymph node spread at diagnosis. The most important factor associated with melanoma thickness among Pacific peoples was body site; melanomas on the lower limbs were almost 8 times thicker than those on the trunk.

Although melanoma was under-reported in New Zealand before 1995,18 the current study was restricted to 11 years of registrations after the introduction of the Cancer Registry Act, minimizing the number of under-reported cases (only 0.2% of melanoma registrations are identified from death certification alone). By law, the New Zealand Cancer Registry receives pathology reports nationally; and, although pathologic parameters can be prone to interobserver variation, an Australian study19 indicated that there was excellent agreement for Breslow thickness. However, some data were missing, which may have influenced some of the results obtained. The collection and classification of ethnicity by health personnel are standardized, although the potential for misclassification still exists; 14% of records were missing ethnicity.

We observed a significant association between age at diagnosis and melanoma thickness in multivariate analysis among Europeans but not among Maori or Pacific peoples. Other studies worldwide7, 20-23 have reported inconsistent results for this association. In many studies, thickness has been dichotomized to handle highly non-normal data, but this may lead to bias, loss of power, and loss of information in analyses.24, 25 Although the choice of categorization may affect any association,25, 26 we also observed an increase in the risk of thick melanoma with increasing age at diagnosis when thickness was dichotomized (≤2 mm vs >2 mm) in multivariate logistic regression (data not shown). Our analyses used melanoma thickness as a continuous log-transformed variable, thus maximizing information and power, and minimizing potential bias from categorization.

Increasing melanoma thickness over time in European New Zealanders was not explained fully by variation in age, sex, histologic subtype, body site, or extent of disease or by interactions that were included in the model, and it is unlikely that these increases were caused by changes in the coding of Breslow thickness. The reasons for this increase in melanoma thickness since 1996 require further research.

NMs account for a disproportionate number of thicker lesions with a poorer prognosis.7, 23, 27, 28 The ABCD29 system of identification (in which A indicates asymmetry; B, border irregularity; C, color variegation; and D, diameter >6 mm) has encouraged the early diagnosis of flat lesions, such as SSM and HMF, but nodular tumors may not reveal the well known signs of classic melanoma, and the ABCD(E) criteria (in which E indicates evolving lesions) may be insufficient to detect them.30 In New Zealand, NMs account for approximately 51% of all melanomas >4 mm thick, leaving 49% that may be more amenable to early detection.

The majority of NMs internationally are self-detected.27, 31, 32 Because their rapid evolution may limit the opportunity for detection by physicians,30 it is likely that the major burden of detection will fall on patients and their families.33 This may be improved only with greater community and physician knowledge of this subtype. Although there is some evidence that deeper melanomas often are self-detected more in New Zealand Europeans,34 whether this occurs similarly in other ethnic groups is unknown.

ALM, which occurs on the unexposed skin of the palms and soles and under the nails, is an uncommon tumor and accounts for only approximately 3% of melanomas in white-skinned individuals.35 In darker skinned ethnicities ALM, although rare, often is the most common melanoma subtype,36 and >75% of ALMs occur on the lower limb.35 In the current study, ALMs accounted for 33% of melanomas in Pacific peoples, and all occurred on the lower limb. Internationally, ALMs are thicker and more advanced at diagnosis and usually have a poorer prognosis than many other melanomas.35 ALMs may differ from other melanomas in their biologic behavior and may display genetic differences compared with melanomas that occur on sun-exposed skin.37, 38

Although some studies have suggested that melanomas on the head and neck are thicker than others, there is no consensus about the association of thickness and site.22, 23, 39 In many studies, associations between site and melanoma thickness suggested in univariate analyses did not persist in multivariate analyses.7, 28, 40

Some researchers have described a later stage of cancer presentation in minorities with melanoma and have implied delayed presentation as a probable explanation.41-43 However, several studies21, 31, 44 did not identify any association between the thickest tumors and patient or physician delay. Rather than pathway delays, it is more likely that the poorer prognosis can be accounted for in part by aggressive, rapidly growing lesions that occur more commonly in minority ethnicities.

The most important marker for melanoma prognosis, tumor thickness, is a crucial endpoint for melanoma campaigns, but increasing the early detection of melanomas is a considerable challenge. So, how could it be improved? Many melanomas are self-detected; thus, expanded education about melanoma subtypes targeted to high-risk and minority ethnic groups is warranted. The ABCD(E) mnemonic already is advocated to physicians and the public, and it remains to be determined whether it promotes earlier diagnosis. Physicians could increase their index of suspicion of skin lesions in older individuals and also of foot lesions in darker skinned individuals. In Europeans and Maori, considerable variation in melanoma thickness was because of factors that were not included in our model. Further research is needed to identify the additional characteristics of individuals and their melanomas to explain ethnic variations in melanoma depth. Research into the genetic, biochemical and molecular differences between thin and thick melanomas, and among melanoma subtypes is needed along with detailed investigation of the distinguishing presenting features of thick melanomas in New Zealand. The development of an individual New Zealand risk predictor that considers multiple risk factors and significant interactions among them, particularly for thick melanoma, could help identify the appropriate high-risk groups for surveillance and targeted prevention activities to reduce death from this disease.

CONFLICT OF INTEREST DISCLOSURES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONFLICT OF INTEREST DISCLOSURES
  7. REFERENCES

M.J.S. and B.C. were supported by the Director's Cancer Research Trust.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONFLICT OF INTEREST DISCLOSURES
  7. REFERENCES