Time Trends of Incidence for Cutaneous Melanoma Among the Japanese Population : an Analysis of Osaka Cancer Registry Data , 1964-95

Cutaneous malignant melanoma occurs much less frequently among non-white populations than among white populations. Little is known of the descriptive epidemiology of melanoma among Japanese. We investigated time trends of incidence of invasive cutaneous malignant melanoma using data from the Osaka Cancer Registry (Japan) among 321 men and 313 women diagnosed between 1964-95. Average, annual, age-standardized incidence rates per 1,000,000 population were 2.45 (95% confidence interval (CI): 2.17-2.72) for men and 2.04 (95% Cl : 1.812.28) for women. The age-standardized rate ratio among men from 1964-71 as a reference was almost constant during the study period, whereas that among women increased up to 1.8 fold (95% CI : 1.25-2.56) in 1980-87 and seems to have reached a plateau recently. Among men, the ratio for head and neck lesions decreased to 0.5 fold (95% Cl : 0.26-0.99) in 1988-95. Among women, the ratio for lesions of the extremities steeply increased up to 4.7 fold (95% CI : 2.688.35) in 1980-87 from the reference period of 1964-71, whereas a slight increase for trunk lesions and no increase for head and neck lesions were noted during the same period. Possible explanations for the subsite-specific time trends are discussed. J Epidemiol, 1999 ; 9 : S129-S135.

Incidence of cutaneous malignant melanoma in many populations of European origin showed a steadily increasing trend between the early to mid-1960s and the early 1990s1-5).In many populations that are not of European origin, the trend is unclear because of the low incidence and the generally low number of cases1-5).Analyses of time trends in age-or subsitespecific incidence of cutaneous malignant melanoma among Japanese would be useful for both understanding the causal risk factors and evaluating prophylaxis in the observed population.
We report the time trends of incidence of cutaneous malignant melanoma in Osaka, Japan, including stratified analysis by sex, age group and anatomic subsite in the past three decades.

SUBJECTS AND METHODS
Osaka prefecture (1995 population: approximately 8.8 million) lies between latitudes 34 and 35 degrees north.Estimated ground ultra violet-B (UV-B) level in 1990 was reported to be 16300 Bin26).Main industries of Osaka are commerce and manufacture.Of all workers in 1985, the proportion of outdoor workers, such as farmers, fishermen or construction workers, was 11%.
Osaka Cancer Registry (OCR) has been operating since 1962, covering all residents in Osaka Prefecture.The register collects reports on newly diagnosed cancer patients from all medical institutions in Osaka Prefecture.Routinely these have been supplemented by death certificates gathered by the Osaka Prefectural Government.Detailed procedures and validity of the OCR have been described elsewhere 5 7).
From 1964 to 1995, 321 males and 313 females with invasive cutaneous malignant melanomas were ascertained through the OCR.Anatomic subsites coded by the registry were reclassified in this analysis into three subsites: head and neck (ICD-9: 1720-24), trunk (ICD-9: 1725), and extremities (ICD-9: 1726-27).To investigate time trends of the incidence rates, we divided the study periods into four 8-year periods: 1964-71, 1972-79, 1980-87 and 1988-95.The age-standardized incidence rates per 106 person-years were calculated for three age groups (15-44, 45-64 and 65-84 years).Since these three groupings include a wide range of ages, age standardization was performed using the world population as a standard.
In the 634 incident cases, we found two cases with an overlapping lesion (ICD-9: 1728) and 171 cases with subsite unspecified (ICD-9: 1729), which were inadequate for determination of subsite.Therefore, we used following procedure for estimating sex-, age-and period-specific incidence rates by anatomic subsite, under the assumption that in the same grouping according to sex, age and period at diagnosis, cases without diagnosis of anatomic subsite would have the same distribution as those with diagnosis of anatomic subsite.First, sex-, 5-year age and period-specific incidence rates were calculated for all cases of cutaneous melanoma, including those with an overlapping lesion or subsite unspecified.Second, sex-, age-(0-44, 45-64, 65 or more) and period-specific proportions of the three anatomic subsites among the cases were calculated.Finally, sex-, 5-year age-and period-specific incidence rates were multiplied by the corresponding sex-, age-and period-specific proportion to estimate sex-, age-and period-specific incidence rates by the anatomic subsite.Based on the estimated sex-, age-and period-specific incidence rates, we calculated the agestandardized incidence rate by the anatomic subsite using the world population as a standard.
To assess the trends of incidence in terms of chance variation, we calculated the age-standardized rate ratios of the incidence of melanoma using the rate in 1964-71 as a reference for 1972-79, 1980-87 and 1988-95.The 95% confidence interval of the incidence rate and its rate ratio were calculated based on the Poisson distribution.
Table 1 shows the number of incident cases and age-standardized incidence rates per 106 person-years of invasive cutaneous malignant melanoma, together with the proportion of cases registered by death certificate only (DCO%) and the incidence to mortality ratio (I/D) from January 1, 1964 to December 31, 1995.Throughout all observed periods, the agestandardized rates were extremely low, ranging from 1 to 3 for both sexes.Among males the rate remained almost constant, although a tentative decline was seen in 1972-79.Among females, an upward trend was observed in the period from 1964 to 87, after which the rate seemed to reach a plateau.DCO% decreased from 24% to 14% among males, while it remained rather constant among females.The I/D in 1972-79 were relatively lower than those in the other periods for both sexes.
Figure 1 shows time trends of age-standardized incidence rates for the three age groups.
An increasing trend was observed among males aged 45 to 64 between 1972 and 1995, whereas it seemed to decline for those aged 65 to 84 during 1964-95.Among females an upward trend was noted for 45to 64-year-olds.For the 15-44 year-old age group, the rate rose steeply from 1964-71 to 1980-87, then showed a decline in the most recent period.
Figure 2 presents the sex-and age-specific incidence of cases of cutaneous malignant melanoma by anatomic subsite.For both males and females, and in every age group, melanoma occurred most frequently on the lower limbs.In both sexes, the ratio of the number of cases with trunk lesions to those with head and neck lesions was greater than 1.0 among the 0-44 age group, and almost even among the 45-64 age group, whereas it was smaller than 1.0 among the age group 65 years or older.Figure 3 presents time trends in age-standardized incidence rates according to anatomic subsites.In males, a steady increase was noted in cases of the extremities, while a decreasing trend was shown in cases of the head and neck.For lesions of the trunk, the rates seemed to increase sharply from 1972-79 to 1980-87.Thus, the rate for lesions of the trunk caught up with that of the head and neck in 1988-95.In females, the rate for lesions of the extremities rose steeply during 1964 and 1987, and leveled off thereafter.A declining trend was noted for those of the head and neck, and a slight increasing trend for those of the trunk was shown in females.
Age-standardized rate ratios of the incidence of melanoma are presented in Table 2 using the rates in 1964-71 as a reference.The ratio for males in all age was almost constant during 1972-1995.For the 45-64 age group in males, a significant increase was observed in the period after 1980, whereas for the age groups 15-44 and 65-84 years old, the ratio significantly declined after 1972.The standardized rate ratios among cases of the head and neck were around 0.5 in the period 1972-95 for males, with a significant decline (95% CI: 0.26-0.99)occurring in 1988-95.In contrast, there was an increasing trend for melanomas of the trunk and extremities up to 1.5 fold for both although without statistical significance.For females in all age, the rate ratio was significantly elevated during the observed period, reaching 1.8 fold (95% CI: 1.3-2.6) in 1980-87, and seems to have reached a plateau in the most recent period.The largest increase in the rate ratio among females occurred in the 15-44 age group.For the 45-64 age group it rose steadily up to 1.7 fold (95% CI: 1.6-1.9) in 1988-95.For the 65-84 age group there was an increase up to 1.3 fold in the period 1980-87 and a slight decrease afterwards.The rate ratio for extremities showed a steep increase up to 4.2 fold (95% CI: 2.4-7.3) in the 1988-95 period.A slight increasing trend of melanoma of the trunk was observed in females.The ratio of cases of the head and neck in females declined to 0.7-0.8fold from the 1964-71 period to the 1972-95 period, but without statistical significance.

DISCUSSION
This report presents data on time trends in the incidence of cutaneous malignant melanoma in the Japanese population over the most recent three decades.Very low average, annual age-standardized incidence rates (less than 10 per 106) were

Trend of Cutaneous
Melanoma Incidence in Osaka S-133 Table 2. Age-standardized rate ratio of cutaneous malignant melanoma incidence when the period 1964-71 was defined as a reference in Osaka Cancer Registry.

(
): 95% confidence interval noted throughout the observed periods, although differences in sex-, age-, period-, and subsite-specific incidence have appeared.The very low incidence rates here reported were similar to, or even lower than, the incidence rates in other populations of a non-European origin 1,8).For males, extremities (mainly the lower limbs) were the most frequent sites of melanoma in all age groups, while the trunk was the least frequent site.These findings differ from the recent subsite distribution of melanoma in many populations of non-Hispanic white males, which show that most cases are of the trunk s-b0) This study showed that the incidence rates of cutaneous melanoma remained almost constant during the most recent three decades in males, mainly because the decrease in the incidence of head and neck lesions offset the trend toward a rise in incidence of lesions of the trunk and extremities.A plausible reason for such a decrease in males is that occupational sun exposure among Osaka residents has been reduced due to the decrease in outdoor work during the recent three decades.The significant decrease in the age-standardized rate ratio for oldest age group, aged 65-84, from 1964-71 to 1972-95 may support our speculation, because occupational sun exposure to a chronically sun-exposed subsite is considered to be associated with the development of melanoma among the elderly 11).The Meteorological Agency of Japan has begun to examine the stratospheric ozone concentration in four areas, including Tsukuba (36 degree north) since 1958 and Kagoshima (32 degree north) since 1964.They found no significant change of ozone concentration in the two areas during the recent three decades 12), suggesting no significant increase in intensity of solar UV-B radiation in the Osaka area during this period.
Our study demonstrated the increasing trend of incidence of cutaneous melanoma in females during the past three decades.The overall increase resulted primarily from a dramatic rise in the rate of melanoma of the extremities.In contrast, the incidence rate for the head and neck was revealed to be 0.7-0.8fold from 1964-71 to 1972-95.These different patterns in rates by body site might be associated with different levels of sun exposure, which may be related to the change in clothing styles among Japanese women.However, the most prevalent subsite of melanoma of the extremities was reported to be the planta (a subtype of acral lentiginous melanoma) in Japanese 13.14).Therefore, further analyses of the subsite-specific incidence using a more detailed classification than the ICD classification are needed to determine whether sun exposure influenced the trend toward increases in the incidence of melanoma of the extremities in females.
The analysis of OCR data on cutaneous melanoma contains some methodological limitations.First, cancer reporting in Osaka is not mandated by law, which would allow for the underreporting of melanoma cases.The DCO% in 1964-71 among males was 5%-10% higher than that in the later period, suggesting that the amount of underreporting among males was greater in the period of 1964-71 than in the later period.This, however, would not affect the lack of an upward trend of the incidence rate of melanoma among males that is shown in our study.In females the DCO% was almost constant, around 16%-19%, throughout the observed period.This suggests that the amount of underreporting was not substantially different among each observed period.Second, improvement of diagnostic techniques for cutaneous melanoma might have increased the incidence during the observed period.This influence would have affected the incidence in all subsites equally.However, our results clearly showed different time trends of the incidence of melanoma by subsite for both males and females, indicating that improvement of diagnostic technique during the observed period did not substantially affect the sitespecific time trend of the incidence of melanoma .Third, we made the assumption in calculation of subsite-specific incidence rates that in the same grouping according to sex, age and time period of diagnosis, there would be equal distribution between cases without a report of the subsite and those with such a report.It is unrealistic to speculate that physicians would determine whether to report melanoma cases to OCR according to the subsite.Therefore, the assumption will not result in a severe bias as far as failure to report a particular subsite selectively is concerned.
In conclusion, although overall incidence rates of cutaneous malignant melanoma in Osaka were very low during the most recent three decades, the rate in females increased up to 1.7 fold from 1964-71 to 1988-95, primarily resulting from the dramatic rise in rates for the extremities.Among males, the rate for head and neck lesions decreased while the overall rate remained almost constant.Further continuous surveillance of the incidence of cutaneous melanoma among the Japanese population is needed to assess the hazard of solar UV exposure to skin, which will possibly change as a result of changes in lifestyle and the atmospheric environment in the future.

Figure 3 .
Figure 3. Trends in age-standardized incidence rates of cutaneous malignant melanoma by anatomic subsites from the Osaka Cancer Registry.

Table 1 .
Incident cases and age-standardized incidence rate of cutaneous malignant melanoma in Osaka Cancer Registry .
a age-standardized incidence rate per 106 person-years b percent of the registered cases from death certificate only c incidence to mortality ratio ( ): 95% confidence interval Trend of Cutaneous Melanoma Incidence in Osaka S-131 Figure 1.Trends of age-standardized incidence rates of cutaneous malignant melanoma in three age groups from the Osaka Cancer Registry.Figure 2. Incident cases of cutaneous malignant melanoma by anatomic subsites, Osaka, 1964-95.