Seasonal Patterns of Skin Melanoma Incidence in Hawaii
M. WARD HINDS, MD, MPH, JAMES LEE, PHD, AND LAURENCE N. KOLONEL, MD, PHD
Abstract: We analyzed the seasonal pattern of diagnosis of malignant melanoma of the skin among Caucasians in Hawaii during 1960-1978. For all cases combined, a significant sine-wave pattern with a summertime peak was found. This pattern was most
pronounced for melanoma of head and neck and lower
extremities and for persons age 50 and over. These
findings support the hypothesis that solar ultraviolet
radiation may be a short-term promoter of some malignant melanomas of the skin. (Am J Public Health 1981;
71:496-499.)
Epidemiologic evidence strongly suggests that solar ultraviolet radiation is an etiologic agent for malignant melanoma of the skin'-. Measurements of solar ultraviolet radiation at various United States locations, including Hawaii,
show a sine wave pattern, with peaks in the summer months
and troughs in the winter months.6 Scotto and Nam recently
fitted the monthly incidence of skin melanoma in the Third
National Cancer Survey to sine wave models and found a
strong seasonal pattern with summertime peak for all melanoma in females and for upper extremity melanoma in
males.7 As noted by the authors of that study, it was impossible to determine if the observed seasonal pattern resulted
from possible promoting effects of ultraviolet radiation or
simply from enhanced recognition of skin changes during the
summer months when skin is less covered by clothing. Since
in Hawaii there is little seasonal variation in clothing styles
because of the tropical climate (the mean temperature in
Honolulu is only 6.50 F lower in January than in August), we
felt an examination of the seasonal patterns of skin melanoma incidence here could address the question. If a seasonal incidence pattern with a summertime peak could be demonstrated among Hawaii's residents, then a stronger case
would be made for solar ultraviolet radiation as a short-term
promoter of melanoma.
in Hawaii,5 only cases in Caucasians were analysed. During
1960-1978, 364 skin melanomas (215 in men and 149 in women) in Caucasians were reported to the HTR. The month of
initial diagnosis was known for 353 of these. Both month of
diagnosis and anatomic site were known for 351.
We followed in general the methods of Scotto and Nam7
and assumed a simple sine wave model:
Methods
The Hawaii Tumor Registry (HTR) provided the basic
data for the analyses. This has been a population-based,
statewide registry since 1960, and has been a member of the
Surveillance, Epidemiology and End Results (SEER) program of the National Cancer Institute since 1973.
Since there is no evidence that ultraviolet radiation is a
factor in development of melanoma among non-Caucasians
Address reprint requests to M. Ward Hinds, MD, MPH, Epidemiology Program, Cancer Center of Hawaii, University of Hawaii,
1236 Lauhala Street, Honolulu, HI 96813. Drs. Lee and Kolonel are
also affiliated with the Cancer Center. This paper, submitted to the
Journal July 21, 1980, was revised and accepted for publication December 1, 1980.
496
Yi
=
I
+Asin O+ k Xi i
kk
=
1, 2,
k
where Yi is the frequency of cases diagnosed in interval Xi
(two contiguous months in our model), k is the number of
intervals in the full cycle (6 two-month intervals in a year)
and A and 6 (the amplitude and phase angle, respectively)
are the two parameters describing the model.
Two-month intervals were used because of the small
number of cases. Estimates of the parameters A and 6 were
obtained by the maximum likelihood method.8 The expected
frequency of cases for interval Xi (i = 1, 2, . . ., k) was computed based on the model. The observed frequencies were
then fitted to the expected frequencies by the chi-square
goodness-of-fit test.9 If the model was not statistically rejected at the predetermined 5 per cent probability level, then
Edwards' method'0 was used to test whether or not the amplitude A differed statistically from zero.
Since we combined data for a 19-year period and the
incidence rate of melanoma in Hawaii has been increasing
over that period,5 we considered the possibility that a seasonal pattern might appear simply because of the long-term
upward trend-in incidence. Because we examined calendar
years, a long-term increase in incidence of melanoma in the
absence of any true seasonal pattern should lead to fewer
cases diagnosed in January than in December each year.
Thus, the sum of cases diagnosed in January for all 19 years
should be fewer than the sum of cases diagnosed in December for all 19 years. In fact, using the average monthly rate of
increase in melanoma diagnoses observed in Hawaii from
1960 to 1978, we calculated that the total number of cases in
January-February should be 5.8 per cent less than the total
number in November-December as a result of this long-term
upward incidence trend. Since this difference is so small, we
AJPH May 1981, Vol. 71, No. 5
SEASONAL MELANOMA PATTERNS IN HAWAII
TABLE 1-Cases of Skin Melanoma among Caucasian Residents of Hawaii, 1960-1978, by Calendar Period of Diagnosis and Anatomic Site
Anatomic Site of Diagnosis
Calendar Period
of Diagnosis
All Skin
Head & Neck
Upper
Extremities
Trunk
Extremities
January-February
March-April
May-June
July-August
September-October
November-December
TOTAL
48
51
71
65
69
49
353*
8
10
9
17
14
5
63
16
12
21
11
13
14
87
17
20
23
16
27
19
122
7
9
17
20
15
11
79
Lower
*Specific anatomic site unknown for 2 cases
chose not to correct for it in our analysis. It also should be
noted that this seasonal variation produced by a long-term
upward trend in incidence peaks in December, not in the
summer months as does solar ultraviolet radiation. Thus, a
fit between the patterns of melanoma diagnosis and solar radiation could not spuriously result from the upward incidence trend.
Results
We found no statistically significant differences at the 5
per cent probability level between males and females when
we examined the distribution of cases by time of year at diagnosis divided into 6 two-month intervals. This was true
whether we exanined all skin melanoma or melanoma of the
head and neck, upper extremities, trunk, or lower extremities. Therefore, all applications to the sine wave model
were carried out using male and female cases combined. The
distribution of these cases is shown in Table 1.
The sine wave model was not rejected by the chi-square
goodness-of-fit test for all skin melanoma nor for melanoma
occurring at any of the four specific anatomic sites. Since the
number of cases in each two-month interval was rather
small, this lack of model rejection does not necessarily mean
that the data follow a sine wave pattern, only that they are
consistent with one. However, the amplitude of the fitted
sine wave model was found to be significantly different from
zero both for all skin melanoma and for melanoma of the
lower extremities (Table 2).
We also examined the seasonal pattern of all skin melanoma for two age groups: 0-49 (N = 203) and age 50+ (N =
148). Again, we found the sine wave model was not rejected
by the chi-square goodness-of-fit test for either age group.
However, only for age 50 and over was the amplitude (6.7
cases) significantly greater than zero by Edwards' test (X2
6.05, p = 0.049).
In Figure 1, we have graphed the sine wave model for all
skin melanoma in males and females. As can be seen, the
data fit the model rather well, although the peak months for
observed cases are May-June rather than July-August as
predicted by the model.
AJPH May 1981, Vol. 71, No. 5
Discussion
Our analysis of seasonal pattern of diagnosis for skin
melanoma in Hawaii shows consistency with a sine wave
model having a peak in the summer months. The model predicts that the number of melanoma cases diagnosed in JulyAugust will be about 53 per cent more than the number diagnosed in January-February. Solar ultraviolet radiation
reaching the earth's surface follows the same seasonal pattern, with average radiation levels in Hawaii being more than
twice as high in August as in January.6 Reasons other than
a causal relationship between solar radiation and melanoma
diagnosis were first considered. Because clothing worn at
different times of the year in Hawaii varies little, and minimum clothing such as beachwear is worn commonly
throughout the year, it seems unlikely that the seasonal
pattern observed would be due simply to increased observation of the skin during summer months. We also considered
the possibility that the Caucasian population of Hawaii might
not be constant in all seasons, but this seems an unlikely explanation for a summertime peak, since most part-time residents spend winter in Hawaii and summer on the mainland.
Still, we examined this question, as well as that of other possible seasonal artifacts by looking at the number of Caucasian cases of stomach, lung, colon, and breast cancer diagnosed by month during 1960-78. We found no evidence for
any seasonal pattern of diagnosis for these cancer sites.
TABLE 2-Results of the Application of Edwards' Method to
Test for Amplitude = 0 against Amplitude # 0 in the
Sine Wave Model of Seasonal Pattern of Melanoma
Anatomic Site
of Melanoma
All skin
Head & Neck
Upper Extremities
Trunk
Lower Extremities
N
353t
63
87
122
79
Amplitude
(cases)
X2*
p Value
12.4
4.3
1.8
1.7
6.3
8.01
5.36
0.53
0.30
8.16
0.018
0.069
0.770
0.860
0.017
*Two degrees of freedom
tSpecific anatomic site unknown for 2 cases
497
HINDS, ET AL.
07
70
A~~
g
-
O
0
0)
uJ
0
65
60-
LuJ
() 55
0
z
<
50
LuJ
MARAPR
MAYJUN
JULAUG
SEP-
OCT
NOVDEC
JANFEB
MARAPR
FIGURE 1-Predicted Sine Wave Model (solid line) for Diagnosis of
Melanoma of the Skin among Caucasians in Hawaii, Compared to Observed Cases (circles) Diagnosed in 1960-1978
The time of diagnosis of a cancer is admittedly not a
precise indicator of the onset of growth. Very slowly growing cancers might be present for years prior to diagnosis.
Nevertheless, it seems reasonable to assume that the onset
of rapid growth in a skin cancer would be soon followed by
diagnosis. To the extent that onset of rapid growth is not
soon followed (within 1-2 months) by diagnosis, it would become more difficult to demonstrate a seasonal pattern of diagnosis where a seasonal pattern of onset of rapid growth of
melanoma was present.
On the whole, we believe our findings support the hypothesis that solar radiation may be a factor in promoting the
short-term growth of malignant melanoma of the skin. Epidemiologic findings of two previous studies also support this
hypothesis.7' 11 An analysis of over 3,000 cases of skin melanoma in the Swedish Cancer Registry found a statistically
significant increase in the incidence among women during
the months of June to August compared with December to
February." The Swedish data showed, as did ours, that the
summertime peak in incidence was most pronounced for
melanoma of the lower extremities. No seasonal pattern of
melanoma incidence in Swedish men was demonstrated,
however. Scotto and Nam's analysis of 2,167 Third National
Cancer Survey melanoma cases found a statistically significant seasonal pattern with a summertime peak for month of
diagnosis of all skin melanomas in women, but not in men.
Their anatomic site-specific analysis showed the strongest
seasonal pattern for the lower extremities of women as well
as the upper extremities of women and men.7 Thus, a consistent pattern in these two earlier epidemiologic studies, plus
our own, is that melanoma of the lower extremities seems to
display the most pronounced seasonal pattern of diagnosis.
Our finding that men as well as women in Hawaii demonstrate a seasonal pattern of skin melanoma diagnosis is in
contrast to the other two reports, but might be explained by
our observation that the wearing of shorts is as common
among men as women in Hawaii. This would also be consistent with our previously reported finding that the incidence
498
rate of melanoma of the lower extremities in men was equal
to that in women in 1972-77.5
The animal studies of Kripke have suggested that ultraviolet radiation may be capable of promoting skin cancer
growth by alteration of immunologic mechanisms.'2 Her experiments have shown that ultraviolet-irradiated mice cannot reject transplants of syngeneic ultraviolet light-induced
skin tumors that are normally rejected by unirradiated mice.
This seems to be true for melanoma as well as more common
skin tumors.
We recognize that several questions can be raised about
the epidemiologic data. For instance, why is there no significant seasonal pattern of diagnosis for men in Sweden or in
the mainland United States areas covered by the Third National Cancer Survey?7'" Why does melanoma of the trunk
exhibit no seasonal pattern of diagnosis in either sex? Scotto
and Nam hypothesized that the intermittency of exposure of
the trunk might allow for sufficient tissue repair to delay the
detection of melanoma to other seasons of the year.7 While
this theory might explain the lack of a clear seasonal pattern
of diagnosis for trunk melanoma and possibly lower extremity melanoma in Swedish and US mainland men, it is insufficient to explain that lack for upper extremity melanoma
in Hawaii, where the wearing of short-sleeved shirts is almost universal.
In spite of the remaining questions, the epidemiologic
and animal exprimental evidence suggest that solar ultraviolet radiation may be a short-term promoting factor for malignant melanoma of the skin, especially of the legs. Malignant
melanoma incidence rates are increasing among Caucasians
in many areas of the world, probably related, in part, to increased leisure time spent in the sun." 3-5 " In Hawaii,
melanoma of the skin is an important site for invasive cancer, particularly in young adults. Among Caucasians age 1544, melanoma of the skin accounted for almost 14 per cent of
all invasive cancers during 1973-77 (Hawaii Tumor Registry). Scotto and Fears have demonstrated that a great proportion of solar ultraviolet radiation reaches the earth's surface between the hours of 10 am and 2 pm during the summer
months.6 Cancer control programs should direct their efforts
toward educating the public to avoid exposure to the sun
during those hours. If avoidance of peak solar radiation exposure can reduce growth promotion of skin melanoma, then
successful public education programs might reverse the
trend of increasing melanoma incidence within a few years'
time.
REFERENCES
1. Anaise D, Steinitz R, Hur NB: Solar radiation: a possible etiologic factor in malignant melanoma in Israel. Cancer 1978;
42:299-304.
2. Gellin GA, Kopf AW, Garfinkel L: Malignant melanoma-a
controlled study of possibly associated factors. Arch Derm
1969; 99:43-48.
3. Houghton A, Munster EW, Viola MV: Increased incidence of
malignant melanoma after peaks of sunspot activity. Lancet
1978; 1:759-760.
4. Swerdlow AJ: Incidence of malignant melanoma of the skin in
England and Wales and its relationship to sunshine. Br Med J
1979; 2:1324-1327.
AJPH May 1981, Vol. 71, No. 5
SEASONAL MELANOMA PATTERNS IN HAWAII
5. Hinds MW, Kolonel LN: Malignant melanoma of the skin in
Hawaii, 1960-1977. Cancer 1980; 45:811-817.
6. Scotto J, Fears TR, Gori GB: Measurements of ultraviolet radiation in the United States and comparisons with skin cancer
data. DHEW Pub. No. (NIH) 76-1029, 1975, section 1.
7. Scotto J, Nam J-M: Skin melanoma and seasonal patterns. Am J
Epidemiol 1980; 111: 309-314.
8. Bliss CI: Statistics in Biology. Vol. II New York: McGraw Hill,
1970, pp 219-287.
9. Snedecor GW, Cochran WG: Statistical Methods. Ames: Iowa
State University Press, 1967, pp 236-238.
10. Edwards JH: The recognition and estimation of cyclic trends.
Ann Hum Genet 1961; 25:83-87.
11. Malec E, Eklund G: The changing incidence of malignant melanoma of the skin in Sweden, 1959-1968. Scand J Plast Reconstr
Surg 1978; 12:19-27.
12. Kripke M: Speculations on the role of ultraviolet radiation in the
development of malignant melanoma. JNCI 1979; 63:541-548.
ACKNOWLEDGMENTS
This research was funded in part by Contract No. NO1-CP53511 from the National Cancer Institute, DHEW.
1981 Graduate Summer Session in Epidemiology
University of Minnesota
The Sixteenth Graduate Summer Session in Epidemiology sponsored by the Epidemiology Section
of the American Public Health Association, the Association of Teachers of Preventive Medicine, and
the American College of Preventive Medicine will be presented at the University of Minnesota in
Minneapolis through the School of Public Health, Health Sciences Center and the Nolte Center for
Continuing Education during the three-week period from June 21 to July 11, 1981.
These summer graduate sessions are designed primarily for teachers in medical schools, but postdoctoral fellows, graduate students and residents in departments of preventive medicine and other
medical school departments may qualify. Similarily teachers, post-doctoral fellows and graduate
students in schools of public health, dentistry and veterinary medicine are eligible as are qualified
personnel of federal, state and local health agencies. The course is accredited (Category I) for the AMA
Physicians Recognition Award.
In general, the 1981 summer session will follow the pattern previously established. In addition to
the two basic courses in Fundamentals of Epidemiology and Fundamentals of Biostatistics, several of
the previously presented courses will be offered. These include: Epidemiology of Infectious Diseases,
Surveillance and Control of Communicable Diseases, Hospital Epidemiology and Infection Control,
Epidemiology of Cancer, Epidemiology of Cardiovascular Diseases, Advanced Statistical Methods in
Epidemiology, Occupational Epidemiology, Clinical Trials-Design, Operation and Analysis, and
Epidemiology of Injuries. A new course on environmental epidemiology will also be offered.
Tuition for the three-week session will be $500. Special rates for food and lodging in dormitories
have been arranged. Grants for tuition may become available but prospective registrants should explore
other sources of funding as soon as possible.
The course will be limited to 300 students with special course limits affecting certain courses. The
application deadline is May 1, 1981. A $50 deposit, credited to tuition, should accompany applications.
Receipt of the deposit will constitute evidence of intent to attend. Refund of deposit will be made if the
application is not accepted or if the applicant's plans change before May 1, 1981.
Further information and application forms for the 1981 session may be obtained by writing to Dr.
Leonard M. Schuman, Director, Graduate Summer Session in Epidemiology, University of Minnesota
School of Public Health, 1-117 Health Sciences Unit A, 515 Delaware St. SE, Minneapolis, MN 55455.
AJPH May 1981, Vol. 71, No. 5