American Journal of Epidemiology
Copyright © 2002 by the Johns Hopkins Bloomberg School of Public Health
All rights reserved
Vol. 155, No. 7
Printed in U.S.A.
Cutaneous Melanin and Risk of Skin Cancer Abstract Dwyer et al.
Cutaneous Melanin Density of Caucasians Measured by Spectrophotometry
and Risk of Malignant Melanoma, Basal Cell Carcinoma, and Squamous Cell
Carcinoma of the Skin
Terence Dwyer,1 Leigh Blizzard,1 Rosemary Ashbolt,1 Juliet Plumb,1 Marianne Berwick,2 and James M.
Stankovich1
Recent advances have enabled quite accurate estimation by spectrophotometry of the density of cutaneous
melanin. The relation between skin cancers and this objective measure of skin phenotype is examined here. For
this purpose, a population-based case-control study of subjects aged 20–59 years of northern European
ancestry was conducted in Tasmania, Australia. Cases (n = 244) of cutaneous malignant melanoma during
1998–1999, and a sample of cases of basal cell carcinoma (n = 220) and squamous cell carcinoma (n = 195)
of the skin were identified from cancer registrations. Controls (n = 483) were selected from a comprehensive
population listing. Melanin at the upper inner arm was estimated from skin reflectance of light of 400 and 420
nm wavelengths. For melanoma, basal cell carcinoma, and squamous cell carcinoma, respectively, the odds
ratios comparing the least with the highest of four melanin categories were 6.2 (95% confidence interval (CI):
2.3, 16.6), 6.3 (95% CI: 2.6, 15.1), and 4.2 (95% CI: 1.7, 10.8) for men and 1.9 (95% CI: 1.0, 3.7), 1.4 (95% CI:
0.7, 3.0), and 0.7 (95% CI: 0.3, 1.7) for women. The gender differences were not due to disparities in site of
occurrence or (for melanoma) in thickness of the lesion. The authors conclude that, particularly for men,
cutaneous melanin density at the upper inner arm is a strong predictor of risk of skin cancer. Am J Epidemiol
2002;155:614–21.
case-control studies; melanins; melanoma; population; skin neoplasms
skin tone charts or prosthesis (1–3) or by personal assessment (4), and skin reflectance of light at the infrared end of
the visible light spectrum (5). These phenotype properties
have been selected for research because they appear to be
markers for the biologic differences that distinguish northern Europeans from lower risk populations. While the biologic basis for these measures has not usually been made
explicit, it is likely that they have been used as proxies for
the density or type of melanin in the skin.
We recently reported that melanin density in the epidermis could be estimated quite accurately using a spectrophotometer (6). The calculation is based on the difference in
reflectance by the skin of wave bands of light centered at
400 and 420 nm. The correlation coefficient for linear association of spectrophotometric measure of melanin and the
histopathologic measurement of melanin for the upper inner
arm was r 0.68.
The association of nevi in adolescents with this spectrophotometric measure of cutaneous melanin density has
since been investigated (7). The risk estimates we obtained
for melanin density were higher than those for other commonly used measures of phenotype such as hair or eye color
and were also higher than those for melanin type estimated
from hair samples.
In this report, we describe the relation between cutaneous
malignancy and skin phenotype assessed using this direct
measure of cutaneous melanin density. The study sample
consisted of 659 cases of malignant melanoma and non-
Previous research on the etiology of cutaneous malignant
melanoma (CMM), basal cell carcinoma (BCC), and squamous cell carcinoma (SCC) has suggested that skin phenotype and sun exposure are independent contributors to risk.
The very large, measured differences in risk between persons of African and European descent, for example, provide
strong evidence for the importance of skin phenotype.
However, in research involving measures of phenotype in
the population group most at risk of skin cancer, those of
northern European descent, only weak associations have
been detected. This may reflect a lack of precision in the
measures of phenotype that have been available or that the
wrong measure of phenotype was chosen.
The commonly used markers for skin phenotype have
been eye, hair, and skin color. The various measures of skin
color have included self-report, interviewer appraisal using
Received for publication December 22, 2000, and accepted for
publication September 26, 2001.
Abbreviations: BCC, basal cell carcinoma of the skin; CI, confidence interval; CMM, cutaneous malignant melanoma; ICD-9,
International Classification of Diseases, Ninth Revision; SCC, squamous cell carcinoma.
1
Menzies Centre for Population Health Research, University of
Tasmania, Hobart, Tasmania, Australia.
2
Memorial Sloan-Kettering Cancer Center, New York, NY.
Reprint requests to Prof. Terence Dwyer, Menzies Centre for
Population Health Research, University of Tasmania, GPO Box
252–23, Hobart 7001; Australia (e-mail: [email protected]).
614
Cutaneous Melanin and Risk of Skin Cancer 615
melanocytic skin cancer and 483 community controls aged
20–59 years of northern European ancestry.
MATERIALS AND METHODS
Subjects
The study population consisted of subjects aged 20–59
years of northern European ancestry who were residents of
Tasmania, Australia, and who had never been diagnosed
previously with a histologically confirmed CMM. The study
population was limited to the predominant ethnic group in
Tasmania (approximately 90 percent trace their ancestry to
the British Isles), which is the group at highest risk of skin
cancer. The restriction to those never diagnosed with CMM
was made to limit recall bias in reporting of previous sun
exposure. Recruitment procedures and study protocols were
approved by the Human Research Ethics Committee of the
University of Tasmania.
Melanoma cases. There were 285 cases (119 males and
166 females) of histologically confirmed primary in situ and
invasive CMM (International Classification of Diseases,
Ninth Revision (ICD-9) code 172) identified from registrations by the Tasmanian Cancer Registry for the period
January 1, 1998, to December 31, 1999, and 258 were interviewed. After interview, 13 cases were excluded as not
being of northern European ancestry (n 7) or having previously been diagnosed with CMM (n 6). The response
from eligible cases was 90.1 percent (245 of 272).
Nonmelanoma skin cancer cases. Histologically confirmed cases of BCC (ICD-9 code 173; International
Classification of Diseases for Oncology code 80903) and
SCC (ICD-9 code 173; International Classification of
Diseases for Oncology code 80703) were sampled from registrations by the Tasmanian Cancer Registry for the period
April 1, 1998, to March 31, 1999. They were selected at random from within strata of 5-year age group, sex, and month
of diagnosis to approximately match the historical age, sex,
and monthly incidence of CMM cases. This was done so that
all cases could be compared with the common set of controls.
In practice, all cases younger than age 35 years (BCC), 50
years (SCC males), or 60 years (SCC females) were
included. When there were multiple diagnoses on the same
selected date, each different lesion was recorded separately.
There were 268 cases of BCC selected, and 238 were
interviewed. After interview, 14 cases were excluded as not
being of northern European ancestry (n 12) or having previously been diagnosed with CMM (n 2). The response
from eligible BCC cases was 88.2 percent (224 of 254).
There were 240 cases of SCC selected, and after exclusion of two persons noted on the pathology form as having
had an organ transplant, 211 were interviewed. After interview, 12 cases were excluded as not being of northern
European ancestry (n 6) and/or having previously been
diagnosed with a CMM (n 8). The response from eligible
SCC cases was 88.1 percent (199 of 226).
Controls. Controls were selected from the roll of registered electors, a comprehensive listing of the population
maintained by the State Electoral Office of Tasmania. They
were selected at random from within strata of 5-year age
Am J Epidemiol Vol. 155, No. 7, 2002
groups and frequency matched to CMM cases. Selection as
a control did not preclude subsequent inclusion as a case (8).
One control was also subsequently included as a SCC case,
and another was subsequently included also as a CMM case.
In total, 644 persons were selected. Of these, 25 were not
approached because they had been diagnosed with a CMM
(n 9), were men with southern European or Asian surnames (n 10), lived on one of the small islands (n 2),
and/or had been interviewed as a BCC or SCC case (n 6).
Among the remainder, 503 were interviewed. After interview, 13 persons were excluded as either not being of northern European ancestry (n 12) or being age 60 years (n 1). The response from eligible controls was 80.7 percent
(490 of 607).
Measurements
Cutaneous melanin. The estimation of melanin density
from skin reflectance was based on previous results (6).
Skin reflectance was measured with a handheld Minolta 508
spectrophotometer (Minolta Camera Company, Ltd., Osaka,
Japan) (figure 1). The equation used was:
MD400 10030.035307 0.009974 1R420 R400 2 4,
where MD400 is an estimate of the percentage of the epidermis of the skin at the upper inner arm that contains melanin,
and R400 and R420 denote the averages of three measurements
of reflectance at 400 and 420 nm made at that site, respectively. Approximately 2 percent of the estimates in this sample took small negative estimated values, but we did not
truncate at zero. As evidence of the high reproducibility of
measurements made by a single measurer, the percentage of
total variation in MD400 in this sample attributed to withinperson variation in the three measurements was 0.7 percent
(intraclass correlation 0.99, calculated from a one-way
random effects model using formula R1 of reference 9). The
melanin measurements are also reproducible by different
measurers. In a study of 20 subjects each measured by two
of the research assistants responsible for spectrophotometric
measurements at our institution, the measurers accounted
for only 2.1 percent of the variance in MD400 (intraclass
correlation 0.98 for the average of three measurements,
calculated from a two-way random effects model using formula R3 of reference 9).
Other measurements of skin reflectance were made at the
top of the buttocks (another unexposed site) and, for about
half of the subjects, on the fleshy part of the upper aspect of
the hand nearest the thumb (an exposed site). Melanin density at those other sites was calculated in the same way from
R400 and R420 at those sites.
Other study factors. The research assistants assessed
eye color, and subjects completed a standardized, interviewer-administered questionnaire, which included questions on ethnicity of grandparents, natural hair color, skin
type (10) by self-assessment, and usual sun exposure and
outdoor activities as a child, teenager, and adult. The questions on skin type asked about skin reaction to unaccustomed sun, tendency to burn when in the sun, and end-ofvacation tan. The questions on sun exposure were based on
616
Dwyer et al.
FIGURE 1. Spectrophotometric assessment of cutaneous melanin density, Tasmania, Australia, 1998–1999. The handheld Minolta 508
spectrophotometer is shown in A. It was used to measure reflectance by the skin of wave bands of light centered at 400 and 420 nm, from which
density of cutaneous melanin was estimated. B and C are views of 4 µm sections of skin biopsies fixed in 10 percent phosphate-buffered
formalin, embedded in paraffin, and stained using the Mason Fontana method for melanin. The view in B is from a subject with relatively little
melanin in the basal layers. The subject in C has more melanin.
the two questions that were established, using polysulphone
badge readings as the comparison method, to be the most
reliable and valid measures of habitual sun exposure by
teenagers in this climate (11). They asked about time spent
in the sun each day during vacations or weekends and frequency of outdoor activities.
Data analysis
Excluded from data analysis were three subjects with skin
conditions (epidermoid bullosa, xeroderma pigmentosum,
and psoriasis), six subjects who had a renal transplant, five
subjects who had recently used a solarium, one case with
CMM who had a chemically induced tan on the arm, and
one control who reported changes in skin color from medication with prednisolone for a liver complaint. These exclusions reduced the data set to 244 CMM cases, 220 BCC
cases, 195 SCC cases, and 483 controls.
Spearman rank correlation coefficients were calculated as
measures of association. Odds ratio estimates of the relative
risks of skin cancer (CMM, BCC, or SCC) were estimated
by logistic regression for categories of density of cutaneous
melanin (MD400), and 95 percent confidence intervals were
calculated from the standard errors of the estimated coefficients of the binary (zero of one) predictors used. Tests of
trend were undertaken by replacing the three binary predictors with a single linear predictor, taking rank scores for the
four categories. Binary (zero of one) predictors were
included for the 5-year age-groups used in matching other
than the two youngest age groups (the age groups 30–34 and
35–39 years were combined as the reference group because
of paucity of numbers). To adjust for self-reported sun exposure, we added a linear predictor taking rank scores for the
categories given in the questionnaire to the regression
model.
In site-specific analyses, lesions were classified by
whether they had occurred on generally exposed sites
(head, neck, and limbs, including the feet) or on the trunk.
Because some cases had multiple lesions of the same type
at different sites, one CMM case and 11 BCC cases were
included in analyses as both a head, neck, and limb lesion
and a trunk lesion.
Am J Epidemiol Vol. 155, No. 7, 2002
Cutaneous Melanin and Risk of Skin Cancer 617
RESULTS
Several characteristics of the samples of cases and controls are summarized in table 1. The study samples varied
somewhat in age, with the group of SCC cases, in particular,
containing fewer younger subjects. The density of cutaneous
melanin (MD400) at the upper inner arm correlated with measures of skin type and most strongly with end-of-summer
tan graded from “dark tan” to “practically no tan” (male
controls: r –0.44, p < 0.01; female controls: r –0.56, p
< 0.010).
The case groups differed from the control groups in measures of skin phenotype, including MD400. Male cases had less
melanin at the upper inner arm and buttocks than did male
controls. For women, the differences were significant only for
CMM cases at the upper inner arm and SCC cases at the buttocks. There were no differences in the mean concentrations
of melanin at the dorsum of the hand, an exposed site.
Odds ratio estimates of the relative risk for categories of
arm melanin are shown in table 2. The odds ratios shown in
the top panel of the table are adjusted only for age, the
matching factor. The associations with lower levels of
melanin for men are strong, with clear evidence of dose
response. For women, the risk estimates are much lower.
The lack of a strong association for women may be due to
negative confounding by sun exposure. This could happen if
the fairer subjects, who are most at risk, avoided the sun.
There was some evidence of this in our data. Measurements
of melanin on the dorsum of the hand, an exposed site, were
available for 406 cases and 319 controls. The distributions of
values for the 158 male and 161 female controls are shown in
figure 2. The means were almost identical (both 3.8 percent),
but more of the women had very high or very low values, and
the variance was 2.21 times greater (p < 0.01) for women than
for men. In comparison, the variation in arm melanin values
for the women in this sample was 1.07 times that of the men,
and the difference was not significant (p 0.66).
Adjustment for one measure of sun exposure—time spent
outdoors on activities as a teenager or an adult (“sun activities”)—nevertheless produced only minor increases in the
estimated risks of skin cancer (table 2). The increases using
other measures of sun exposure, or measures of sun exposure
for other periods of life, were less. The increase in risk on
adjustment was greatest for BCC and SCC because the sun
activities measure was linearly associated with increased risk
of BCC (men, p 0.24; women, p < 0.01) and SCC (men,
p < 0.01; women, p 0.02). It was not associated with risk
of CMM (men, p 0.53; women, p 0.35).
Table 3 shows the effect estimates for some of the other
measures of phenotype that have been used in skin cancer
research. The highest risk estimates were those for MD400
among men. Only for inability to tan, which was reported by
only 12 male controls, did an alternative measure of phenotype produce a higher risk estimate, and then it was only for
male cases of BCC. The relative risk estimates were generally lower for women, and none of the measures of their
phenotype had a consistent advantage over the others.
To this point, we have presented data for all body sites
combined. However, the distribution of skin cancers by
Am J Epidemiol Vol. 155, No. 7, 2002
body site differed for men and women. To examine whether
the differences in the effect estimates for men and women
were due to factors associated with body site, the odds ratios
for CMM and BCC were calculated separately for lesions
located on generally exposed sites (head, neck, and limbs)
and those on the trunk. For men, the odds ratios comparing
the extreme categories of MD400 were 5.5 (95 percent confidence interval (CI): 1.7, 18.1) for CMM, 6.6 (95 percent CI:
2.3, 19.0) for BCC on the head, neck, and limbs, 7.3 (95 percent CI: 1.5, 35.0) for CMM, and 5.5 (95 percent CI: 1.7,
18.2) for BCC on the trunk. For women, the odds ratios
were 1.9 (95 percent CI: 0.9, 3.7) for CMM, 1.3 (95 percent
CI: 0.5, 3.1) for BCC on the head, neck, and limbs, and 2.6
(95 percent CI: 0.6, 11.2) for CMM and 1.5 (95 percent CI:
0.6, 3.8) for BCC on the trunk. There were too few SCC on
the trunk of men (n 7) or women (n 13) to make meaningful comparisons.
For CMM, we were able to examine the risk estimates for
lesions of different thickness. In this sample, 38 percent (90
of 240) of the lesions with a Clark’s level classification were
in situ lesions. Combining them with 49 other lesions less
than 0.5 mm in thickness (40 of 49 of these were Clark’s
level 2), the odds ratios were calculated separately for
lesions less than 0.5 mm in thickness and those at least 0.5
mm in thickness. The results are shown in figure 3. The relative risk estimates were higher for the thicker lesions, but
the gender difference in relative risk was maintained
because similar percentages of male (39 percent (39 of 100))
and female (43 percent (60 of 141)) cases had lesions less
than 0.5 mm in diameter.
DISCUSSION
This is the first direct evidence that melanin density of the
skin is a determinant of risk of melanoma and nonmelanocytic skin cancer. Our data show that cutaneous
melanin density at the upper inner arm is a strong predictor
of risk for all three types of skin cancer, particularly in
males. The strength of the association was greater than those
found previously in the high-risk group of northern
European descent for the phenotype measures of eye or hair
color or the various previous measures of skin color used in
research on humans. Studies conducted in Denmark (3),
Canada (1), the United Kingdom (12, 13), and Australia (2,
14, 15), where the subjects were principally of northern
European extraction, have produced odds ratios in the range
of 1.0–3.0 for these previously used markers of phenotype.
Particularly for males, the data here suggest that the
objective measure of melanin density we have developed (6)
is a stronger predictor than are those more subjective measures. The dose-response effects found further support the
inference that the association is causal. It was also a stronger
predictor than were measures of skin type, based on questions that form the basis of the Fitzpatrick classification
(10), the measure most commonly used by dermatologists.
In addition, the link is one that has biologic plausibility.
Melanin does have the capacity to absorb ultraviolet light,
and it is found in greater density in the skin of racial groups
that have the lowest incidence of skin cancers (16).
618
Men
Controls
Women
CMM*
SCC*
BCC*
Am J Epidemiol Vol. 155, No. 7, 2002
%
No. of
subjects
%
No. of
subjects
%
12
41
44
30
30
39
76
76
99
35
31
35
50
44
50
30
34
36
34
38
40
3
31
66
3
30
65
10
32
58
10
31
56
17
36
47
42
89
116
12
38
49
17
55
70
11
38
52
12
42
58
16
27
57
16
26
55
19
37
24
40
23
39
26
37
63
91
9
34
13
49
21
32
23
36
20
34
19
33
40
8
43
9
29
7
28
7
34
3
85
8
51
6
73
9
44
4
49
4
36
10
35
10
14
52
24
10
10
45
26
19
11
49
28
20
20
39
26
15
19
38
25
15
17
41
27
16
42
102
67
40
12
36
33
19
17
52
48
27
8
43
33
16
9
48
37
18
10
38
26
27
10
37
25
26
Mean
(SD)
No. of
subjects
Mean
(SD)
No. of
subjects
Mean
(SD)
No. of
subjects
Mean
(SD)
No. of
subjects
Mean
(SD)
No. of
subjects
Mean
(SD)
No. of
subjects
Mean
(SD)
1.51 (0.90)
0.74 (0.89)
3.64 (0.46)
108
108
66
1.44 (0.92)
0.68 (0.79)
3.73 (0.57)
97
96
49
1.74 (0.97)
0.83 (0.85)
3.75 (0.37)
251
240
160
2.08 (1.07)
0.92 (0.95)
3.79 (0.49)
144
143
100
1.82 (1.11)
0.78 (0.90)
3.65 (0.61)
112
112
59
1.94 (1.00)
0.78 (0.98)
3.45 (0.81)
97
97
56
1.91 (0.93)
0.71 (0.74)
3.71 (0.36)
31
42
35
12
42
45
9
27
64
10
29
68
24
30
18
34
36
10
36
10
14
52
24
10
No. of
subjects
%
No. of
subjects
%
No. of
subjects
Age (years)
20–39
40–49
50–59
30
34
36
70
79
83
30
25
45
30
25
45
29
39
32
Eye color
Brown
Hazel or green
Blue or gray
17
26
57
39
60
132
19
23
57
19
23
56
26
42
60
95
24
30
29
4
65
8
30
45
20
5
69
104
47
12
Cutaneous melanin
(%)‡
Upper inner arm
Buttocks
Hand
No. of
subjects
Mean
(SD*)
232
226
158
2.01 (1.03)
1.05 (0.88)
3.76 (0.33)
100
98
78
SCC
%
%
%
End of vacation tan
Dark
Medium
Light
Practically none
BCC
CMM
No. of
subjects
No. of
subjects
No. of
subjects
Hair color†
Black or dark
brown
Light brown
Mousy
brown/blonde
Red
Controls
No. of
subjects
* CMM, cutaneous malignant melanoma; BCC, basal cell carcinoma; SCC, squamous cell carcinoma of the skin; SD, standard deviation.
† Self-reported hair color as a teenager or young adult.
‡ Male cases had less cutaneous melanin at the upper inner arm (CMM, p < 0.01; BCC, p < 0.01; SCC, p = 0.03) and buttocks (CMM, p = 0.02; BCC, p < 0.01; SCC, p = 0.04) than did male controls. For
women, the differences were significant only for CMM cases at the arm (p = 0.02) and SCC cases at the buttocks (p = 0.04).
Dwyer et al.
TABLE 1. Characteristics of participants in the case-control study of cutaneous malignant melanoma, basal cell carcinoma and squamous cell carcinoma of the skin in
Tasmania, Australia, 1998–1999
p = 0.25
0.4, 1.8
1.1, 5.3
0.3, 2.2
1.0
0.8
2.4
0.9
p = 0.10
14
24
47
12
p = 0.02
0.6, 2.1
0.8, 2.8
1.0, 3.8
1.0
1.1
1.5
2.0
23
35
49
37
p < 0.01
51
77
78
45
0.8, 5.1
1.2, 6.9
2.2, 15.7
1.0
2.1
2.8
5.9
p < 0.01
8
28
36
25
0.6, 3.5
1.3, 7.3
2.8, 17.1
1.0
1.4
3.1
7.0
p < 0.01
8
19
40
41
0.8, 5.9
1.6, 10.6
2.4, 17.8
Adjusted for age
and sun
exposure†
≥3.00
2.00–2.99
1.00–1.99
<1.00
Linear
trend
44
75
75
38
6
22
41
31
1.0
2.2
4.1
6.5
p < 0.01
Am J Epidemiol Vol. 155, No. 7, 2002
* CMM, cutaneous malignant melanoma; BCC, basal cell carcinoma; SCC, squamous cell carcinoma of the skin; OR, odds ratio; CI, confidence interval.
† Adjusted for time spent outdoors on activities (sport, spectator sports, gardening, walking, etc.) as an adult (CMM and BCC) or as a teenager (SCC).
18
33
39
22
1.0
1.3
1.7
1.8
0.7, 2.7
0.8, 3.3
0.8, 3.9
p = 0.31
p = 0.03
p < 0.01
p < 0.01
1.0
2.1
2.6
4.2
8
28
36
25
1.0
1.4
2.9
6.3
8
19
40
41
0.8, 5.8
1.6, 10.4
2.3, 16.6
1.0
2.2
4.1
6.2
6
22
41
31
44
75
75
38
Adjusted for age
≥3.00
2.00–2.99
1.00–1.99
<1.00
Linear
trend
OR*
0.6, 3.5
1.3, 6.8
2.6, 15.1
OR
95%
CI
OR
95%
CI*
No.
of
subjects
No.
of
subjects
No.
of
controls
Arm
melanin (%)
p = 0.41
0.3, 1.7
0.9, 4.3
0.3, 1.7
1.0
0.8
2.0
0.7
14
24
47
12
1.0
1.3
1.4
1.4
18
33
39
22
0.6, 2.0
0.8, 2.6
1.0, 3.7
1.0
1.1
1.4
1.9
23
35
49
37
51
77
78
45
OR
No.
of
subjects
95%
CI
No.
of
controls
No.
of
subjects
0.9, 5.2
1.1, 6.1
1.7, 10.8
95%
CI
OR
0.6, 2.5
0.7, 2.8
0.7, 3.0
95%
CI
No.
of
subjects
OR
SCC
BCC
No.
of
subjects
Women
CMM
SCC*
Men
BCC*
CMM*
TABLE 2. Age-adjusted odds ratios of cutaneous malignant melanoma, basal cell carcinoma, and squamous cell carcinoma of the skin, for density of cutaneous
melanin at the upper inner arm, Australia, 1998–1999
95%
CI
Cutaneous Melanin and Risk of Skin Cancer 619
FIGURE 2. Distribution of hand melanin among controls in the
Tasmanian case-control study of melanoma and nonmelanoma skin
cancer, Tasmania, Australia, 1998–1999.
The weaker association found between melanin density
and risk of each skin cancer in women cannot be explained
by differences in site of occurrence or thickness of lesions.
However, it might be due to negative confounding with sun
exposure. The distribution of hand melanin for women compared with that of men suggests a wider range of sun exposure by women. That very fair women might have less sun
exposure than very fair men is supported by data we
obtained from Tasmanian adolescents, showing that fairer
females avoided the sun more than did males of similar skin
phenotype (11). However, in this study we were unable to
remove the differences in risk between males and females
by adjusting for their self-reported sun exposure. This leaves
the possibility that there are biologic differences in the
pathogenesis of skin cancers in men and women involving
other causes or differences in skin response to ultraviolet
radiation.
This demonstration that melanin density measured noninvasively by spectrophotometry at the skin’s surface can
predict risk of skin cancer, particularly melanoma, has
important research implications. This new, objective measure is free from observer bias and errors associated with
recall. Previous measures of skin phenotype used in epi-
Least
risk group
Greatest
risk group
BCC*
CMM*
SCC*
Description
No.
of
controls
Description
No.
of
controls
OR*,†
Men
Arm melanin
Eye color
Hair color‡
Skin reaction
End-of-vacation tan
≥3% melanin
Brown
Black/dark brown
Tan only
Deep tan
44
39
60
43
69
0–1% melanin
Blue or gray
Light blonde/red
Burn, then peel
Practically no tan
38
132
32
101
12
6.5
0.9
1.8
1.4
4.1
2.4,
0.5,
0.9,
0.7,
1.5,
17.8
1.7
3.7
2.9
11.6
7.0
2.0
3.0
3.1
12.1
2.8,
0.9,
1.4,
1.4,
4.7,
17.1
4.3
6.2
6.8
34.3
5.9
2.1
1.6
3.2
5.2
2.2,
0.9,
0.7,
1.4,
2.0,
15.7
4.7
3.4
7.4
14.1
Women
Arm melanin
Eye color
Hair color‡
Skin reaction
End-of-vacation tan
≥3% melanin
Brown
Black/dark brown
Tan only
Deep tan
51
42
63
42
42
0–1% melanin
Blue or gray
Light blonde/red
Burn, then peel
Practically no tan
45
116
46
132
40
2.0
1.5
4.5
1.6
1.8
1.0,
0.8,
2.2,
0.8,
0.8,
3.8
2.8
9.4
3.0
3.9
1.8
1.6
1.8
2.2
2.8
0.8,
0.8,
0.9,
1.0,
1.1,
3.9
3.4
3.5
4.8
7.1
0.9
1.2
2.0
3.6
4.4
0.3,
0.6,
0.9,
1.5,
1.7,
2.2
2.5
4.3
8.7
11.4
95%
CI*
OR†
* CMM, cutaneous malignant melanoma; BCC, basal cell carcinoma; SCC, squamous cell carcinoma of the skin; OR, odds ratio; CI, confidence interval.
† Adjusted for age and time spent outdoors on activities (sport, spectator sports, gardening, walking, etc.) as a teenager (SCC) or an adult (CMM and BCC).
‡ Hair color = self-reported hair color as a teenager or young adult.
95%
CI
OR†
95%
CI
Dwyer et al.
Measure of
phenotype
620
TABLE 3. Adjusted odds ratios of cutaneous malignant melanoma, basal cell carcinoma, and squamous cell carcinoma of the skin, for extreme categories of measures
of phenotype, Australia, 1998–1999
FIGURE 3. Odds ratio estimates of relative risk of melanoma for
cutaneous melanin density (percent), classified by thickness of
lesion, Tasmania, Australia, 1998–1999.
demiologic studies have either been subjective or, if apparently objective, only weakly associated with risk. Some
have been shown to be influenced by past experience of sun
exposure, making them unsuitable in research where the
goal is to separate the effects of sun exposure from phenotype or, alternatively, to estimate how phenotype modifies
the effect of sun exposure (17). This new measure appears
to provide significant advantages for use in future research
in this field.
It is also conceivable that the spectophotometric measure
of cutaneous melanin could have a clinical application.
High-risk persons could be identified in a valid and reproducible way and given advice on sun avoidance appropriate
to their risk status. A determination of its predictive value in
other populations would be desirable because in this study
we examined its use in a population that included only persons of northern European descent.
A further issue remaining is whether this phenotype measure will prove to be preferable to measures of genotype for
clinical screening or for use in research. A study by Valverde
et al. (18) has shown that the presence of mutations in the
MC1R gene is associated with an increase in risk of
Am J Epidemiol Vol. 155, No. 7, 2002
Cutaneous Melanin and Risk of Skin Cancer 621
melanoma. The odds ratio estimate was 3.9:1, suggesting
that measure of this genotype does not have advantages (for
males, at least) in prediction of risk over measurement of the
melanin density phenotype using our methods. It is considerably more costly. Further development in genotyping may
provide additional advantages.
6.
7.
8.
ACKNOWLEDGMENTS
The National Health and Medical Research Council of
Australia funded this study. Financial assistance with the
program of research was received from The Medical
Benefits Fund of Australia Limited, a registered health benefits organization.
The authors acknowledge the contribution of Dr. AnneLouise Ponsonby, who commented on an earlier version of
this paper. They thank also the research nurses who undertook the field measurements.
9.
10.
11.
12.
13.
14.
REFERENCES
1. Elwood JM, Gallagher RP, Hill GB, et al. Pigmentation and
skin reaction to sun as risk factors for cutaneous melanoma:
Western Canada Melanoma Study. Br Med J 1984;288:
99–102.
2. Green A, MacLennan R, Siskind V. Common acquired naevi
and the risk of malignant melanoma. Int J Cancer 1985;35:
297–300.
3. Østerlind A, Tucker MA, Hou-Jensen K, et al. The Danish
case-control study of cutaneous melanoma. I. Importance of
host factors. Int J Cancer 1988;42:200–6.
4. Chen Y, Dubrow R, Holford TR, et al. Malignant melanoma
risk factors by anatomic site: a case-control study and polychotomous logistic regression analysis. Int J Cancer 1996;67:
636–43.
5. Holman CDJ, Armstrong BK. Pigmentary traits, ethnic origin,
Am J Epidemiol Vol. 155, No. 7, 2002
15.
16.
17.
18.
benign nevi, and family history as risk factors for cutaneous
malignant melanoma. J Natl Cancer Inst 1984;72:257–66.
Dwyer T, Muller KH, Blizzard L, et al. The use of melanin
spectrophotometry to estimate melanin density in Caucasians.
Cancer Epidemiol Biomarkers Prev 1998;7:203–6.
Dwyer T, Prota G, Blizzard L, et al. Melanin density and
melanin type predict melanocytic naevi in 19–20 year olds of
northern European ancestry. Melanoma Res 2000;10:387–94.
Rothman KJ, Greenland S. Case-control studies. In: Rothman
KJ, Greenland S, eds. Modern epidemiology. 2nd ed.
Philadelphia, PA: Lippincott-Raven, 1998:99.
Armstrong BK, White E, Saracci R. Principles of exposure
measurement in epidemiology. Oxford, England: Oxford
University Press, 1992:101, 103.
Fitzpatrick TB. Editorial: the validity and practicality of sunreactive skin types I through VI. Arch Dermatol 1988;124:
869–71.
Dwyer T, Blizzard L, Gies PH, et al. Assessment of habitual
sun exposure in adolescents via questionnaire—a comparison
with objective measurement using polysulphone badges.
Melanoma Res 1996;6:231–9.
Elwood JM, Williamson C, Stapleton PJ. Malignant melanoma
in relation to moles, pigmentation, and exposure to fluorescent
and other lighting sources. Br J Cancer 1986;53:65–74.
Bataille V, Bishop JA, Sasieni P, et al. Risk of cutaneous
melanoma in relation to the numbers, types and sites of naevi:
a case-control study. Br J Cancer 1996;73:1605–11.
Grulich AE, Bataille V, Swerdlow AJ, et al. Naevi and pigmentary characteristics as risk factors for melanoma in a highrisk population: a case-control study in New South Wales,
Australia. Int J Cancer 1996;67:485–91.
English DR, Armstrong BK, Kricker A, et al. Demographic
characteristics, pigmentary and cutaneous risk factors for
squamous cell carcinoma of the skin: a case-control study. Int
J Cancer 1998;76:628–34.
Hill, MZ. The function of melanin or six blind people examine
an elephant. Bioessays 1992;14:49–56.
Blizzard L, Dwyer T, Ashbolt R. Changes in self-reported skin
type associated with experience of sunburning in 14–15 year
old children of northern European descent. Melanoma Res
1997;7:339–46.
Valverde P, Healy E, Jackson I, et al. Variants of the
melanocyte-stimulating hormone receptor gene are associated
with red hair and fair skin in humans. (Letter). Nature Genet
1995;11:328–30.