Descriptive Epidemiology of Pancreatic Cancer in Japan

This paper aimed to explore descriptive epidemiological features of pancreatic cancer in Japan, by examining secular trends in mortality (1950-1995) and incidence rates (1975-1991), and also to discuss background factors possibly ascribable to the recent variation in mortality and incidence rates. The age-adjusted mortality rate (adjusted by the 1985 model population) has increased by approximately 9-fold in both males (from 1.4 to 12.5 per 100,000 population) and females (from 0.8 to 6.8 per 100,000 population) between 1950 and 1995. But it began to level off in both sexes since 1985. The age-adjusted incidence rate has remained plateau in both sexes, but likely showed a downward trend since 1988 in females. The older age groups showed the most steep increasing gradient in both age-specific mortality and incidence rates in both sexes, whereas a little or no apparent increase was observed in young age groups. Diagnostic improvement and increased cigarette consumption were believed to be the two major factors contributing to the increased mortality and incidence rates in Japan, but further epidemiological studies will be required to clarify possible background factors for the recent levelling-off of the mortality and incidence rates. J Epidemiol, 1998 ; 8 : 52-59.

Pancreatic cancer, which is well known due to extremely poor survival rate, is one of the major causes of death in many developed countries').Steadily increasing mortality and incidence rates have been observed worldwide during the past several decades21.In Japan, pancreatic cancer numbered 8,965 deaths in males and 7,054 deaths in females in 1995: being the fifth leading cause of cancer deaths, following cancer of the lung, stomach, colon, and liver in this order 3).Pancreatic cancer is a major public health issue with large clinical challenge in Japan, since its fatality rate is quite high with almost unknown etiology4).Beside this, pancreatic cancer mortality is known to have been increasing in Japan 5), whereas it has already begun to level off in the U.S.A. since 1970s, after a consistent increase for approximately 40 years, likewise in other developed countries2, [6][7][8].It is, therefore, of significance in Japan to examine whether the mortality and incidence rates are still continuously increasing or not, since chronological observation of mortality and incidence rates might possibly provide some new clues to the etiology of pancreatic cancer.Thus, in this communication, we will explore descriptive epidemiological features of pancreatic cancer in Japan, by examining secular trends in mortality  and incidence (1975)(1976)(1977)(1978)(1979)(1980)(1981)(1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991) rates, and also discuss background factors possibly ascribable to the recent variation in mortality and incidence rates.

MATERIALS AND METHODS
Data on all certified pancreatic cancer deaths from 1950 to 1995 were routinely provided from National Vital Statistics 3), which were annually published by the Ministry of Health and Welfare, including the population by sex and 5-year age group.Cancer of the pancreas has continuously been coded as 157 from 1950 to 1994 according to the International Classification of Diseases (ICD 6th-9th Revision), and changed to C25 in 1995 (ICD 10th Rivision) 3).Age-adjusted mortality rate was calculated by the direct method using the 1985 model population in Japan as the standard 9).Mean age at death was calculated by summing the product of each middle point of 5-year age group and the number of deaths in each 5-year group, and then by dividing the sum by the total number of deaths in each year.
Trends in age-specific mortality rate were also examined together with birth cohort analysis.Age and cohort effects on mortality were examined, using the matrices of deaths and population for every 5 years since 1975, and applying the multiplicative model proposed by Breslow and Day10).
Mortality by 2010 was projected based on mortality data from 1985-1995.First, the future age-specific mortality rate was estimated by applying linear regression.Then, age-adjusted mortality was calculated, and the number of deaths was estimated using future population estimations reported by the Institute of Population Problem in the Ministry of Health and Welfare 11) In Japan, annual cancer incidence rates have been estimated since 1975 by The Research Group for Population-Based Cancer Registration, based on data from several representative cancer registries.Annual incidence data (1975)(1976)(1977)(1978)(1979)(1980)(1981)(1982)(1983)(1984)(1985)(1986)(1987)(1988) were obtained from published sources 12-15), and latest ones (1989)(1990)(1991) from annual progress reports of The Research Group for Population-Based Cancer Registration16).The reliability of incidence estimation in Japan was described elsewhere in details12).In brief, the population covered by these registries was 19.3 million, which covered 15.7% of the entire population of Japan in 1988.The average DCO (proportions of cases registered by death certificate only) of these registries was 26.6%, and average LID (the ratio of incidence to deaths) was 1.59 'a.Age and cohort effects on the incidence were also assessed by the same method applied to the mortality

Crude and age-adjusted mortality
Table 1 presents annual numbers of pancreatic cancer death, its proportions in total cancer deaths, crude and age-adjusted mortality rates, mean ages at death from 1950 to 1995, with crude and age-adjusted incidence rates from 1975 to 1991.A consistently increasing annual number of pancreatic cancer deaths was apparent: from 526 deaths in 1950 to 16,019 deaths in 1995.The proportion in total cancer deaths steadily increased in both sexes, being 7-fold in males and 9-fold in females between 1950 and 1995.Male mortality rates always predominated female ones.Mortality rates apparently and linearly increased from 1950 to 1985.Age-adjusted mortality rate has increased by approximately 9-fold in both males (from 1.4 to 12.5 per 100,000 population) and females (from 0.8 to 6.8 per 100,000 population) between 1950 and 1995, but ageadjusted mortality had begun to level off in both sexes since 1985 (Figure 1).A gradual increasing trend in mean age at death was observed in both sexes.Mean age at death increased from 56.0 years old in 1950 to 69.3 years old in 1995 in males, and from 56.4 years old to 74.0 years old in females, correspondingly.

Age-specific mortality rate
Figure 2 shows secular trends in age-specific mortality rate by five-year age group.The two oldest age groups (80-84 years and 85+ years) showed the most steep increasing gradient in both sexes, whereas a little or no apparent increase was observed in such younger age groups as 40-44 and 45-49 years old.In the intermediate age groups, increasing trends were also not apparent in both sexes, particularly in the recent two decades.

Birth-cohort analysis on mortality rate
Birth cohort analysis on the mortality rate showed an apparently increasing risk of death with an advancing birth-cohort in both sexes (Figure 3).The mortality rate sharply increased before 1916 birth cohort, but not thereafter, with notable levelling-off trend in both sexes after 1916 birth cohort.

Age and Cohort effect on the mortality rate
Age and cohort effects on the mortality rate are demonstrated in Figure 4. Risk of pancreatic cancer deaths progressively increased with an advancing age, particularly after 60 years old in both sexes.Linearly increasing effect of birth cohort was clearly expressed in males born after 1891, but in females only those born between 1891 and 1911.Unaltered cohort effect was apparent in females bom after 1911.Period effect was also examined for mortality, with no apparent increase after 1985 in both sexes (data not shown)

Prediction of mortality trend by 2010
Figure 5 shows the prediction of mortality by 2010.The number of deaths in 2010 was estimated as 13,821 in males and 12,015 in females: being 1.5-times and 1.7-times of those in 1995, respectively.The levelling-off of the age-adjusted mortality seems to continue in future years, being approximately 12.8 per 100,000 population in males and 7.2 per 100,000 population in females in 2010.

Crude and age-adjusted incidence rate
Table 1 and Figure 1 also present secular trends in crude and age-adjusted incidence rates.Male incidence rate predominated female one, likewise in mortality.Crude and age-adjusted incidence rates have steadily increased since 1975 in almost parallel with mortality trend.Crude incidence rates increased from 6.1 per 100,000 population in 1975 to 13.4 in 1991 in males, and 4.6 to 10.0 in females.Age-adjusted incidence rate correspondingly increased from 8.7 to 12.5 in males and from

DISCUSSION
The present epidemiological study on pancreatic cancer in Japan provided the following major fmdings.(1) Male mortali-ty rates were always higher than female ones.Mortality rate had steadily increased from 1950-1985 in both sexes, but, when age-adjusted, the mortality showed levelling-off after 1985.The two oldest age groups (80-84 years and 85+ years) experienced the most steep increasing gradient in age-specific mortality and incidence rate.(2) The incidence rate in 1975-1991 also showed almost parallel trend with the mortality.(3) Age and cohort effects on mortality and incidence were unique by sex.Risk of both pancreatic cancer incidence and mortality progressively increased when age advanced.The cohort effect on mortality and incidence steadily increased in males.In females, however, the cohort effect was unaltered in those born after 1911 in mortality and decreased in those born after 1916 in incidence.( 4) Age-adjusted mortality rate would continously level off in future years.
In general, when mortality rate has varied over time, the following factors 17) should be considered: (1) changes in the conception and recognition of the given disease, (2) changes in the diagnostic and detection techniques, (3) changes in the procedures for classification of causes of death, (4) changes in the classification code of causes of death, (5) changes in accuracy of reporting age at death, (6) errors in enumerating the population, (7) changes in age distribution of the population, (8) changes in survivorship due to changed fatality rate, (9) changes in the incidence of the disease due to genetic/environmental factors.These factors may be responsible for mortality variations only when they are substantially large.Since neither has there been big changes in the classification code for pancreatic cancer deaths since 1950, nor in the procedures for classification of causes of death, nor in accuracy of reporting age at death and fatality as well.Beside these background, errors in enumerating the population have been believed to be minimal in Japan, and age-adjusted mortality was calculated to remove the effect by changing age distribution of Japanese population.Then, changes in the conception and recognization of pancreatic cancer, changes in the diagnostic and detection techniques, changes in survivorship due to changed fatality rate, and changes in the incidence due to genetic/environmental factors should be taken into consideration.
To the apparently increased mortality rate in the past several decades in Japan, advanced diagnostic and detection techniques would partly be ascribable.Since 1981, introduction with the subsequent popularity of such non-invasive imaging techniques as ultrasound (US) and computed tomography (CT) have certainly and greatly improved the detection rate of pancreatic cancer and its diagnostic accuracy as well18).This has certainly resulted in an increased number of the patients detected, which consequently contributed to an apparently increased incidence rate and also to the mortality rate, since no appreciably improved survivorship has been achieved4).Among genetic/environmental factors, cigarette smoking seems only one factor confirmed to be positively associated with pancreatic cancer by accumulated epidemiological evidence19-21).Continued high smoking rate during past several decades in Japan (the highest adult smoking rate was 82.3% in males and 15.7% in females22) might probably be one determinant of increased incidence rate of pancreatic cancer.Other risk factors possibly incriminated for pancreatic cancer included consumption of alcohol and coffee, intake of fat and meats, personal history of chronic pancreatitis and diabetes mellitus, and occupational exposures23).It is however that epidemiological associations between these risk factors and pancreatic cancer carcinogenesis are still controversial.Increased consumption of alcohol and meats in the recent Japanese lifestyles might possibly be related to the increased incidence rate, coupled with high cigarette smoking rate.It might, therefore, be hypothesized that diagnostic improvement and increased cigarette consumption would be largely responsible for the increased pancreatic cancer incidence and mortality, with much higher rates in males than in females, during the past several decades in Japan.
Likewise in other developed countries 7,8,24), age-adjusted mortality and incidence rate of pancreatic cancer began to leveloff after 1985 in both sexes in Japan.Such levelling-off phenomenon was noted much earlier in the U.S.A. and Canada, where mortality rate peaked in the 1970s24).In the U.S.A., according to the examination25) of the mortality trend and cigarette smoking, it was concluded that cancer mortality rate well paralleled with prevalence of cigarette smoking in both sexes during the past several decades.In Japan also, this levellingoff phenomenon might possibly be ascribable to changing smoking habits.Adult smoking rate did begin to decline gradually year after year since 1980s, and it was 60.4% and 13.3% in males and females in 1992 26), respectively.Another factor, which might possibly be relevant to the levelling-off phenomenon in Japan, would be the gradually increasing intake of fruits and vegetables, since reduction of pancreatic cancer risk by increased consumption of fruits and/or vegetables has epidemiologically been documented27-29).Vitamin C, which is rich in many fruits and vegetables, was also known to have an independent protective effect on pancreatic cancer carcinogenesis 30).
In summary, age-adjusted mortality and incidence rates had sharply increased in the past decades in Japan, and levelled off in both males and females since 1985.Diagnostic improvements and cigarette consumption might have largely been ascribable to this variation in mortality and incidence.Further epidemiological investigations, which will focus on background factors contributing to the recent and future levellingoff trend, would be warranted to disclose some clues to the etiology of pancreatic cancer in the developed countries including Japan.

Table 1 .
Mortality and incidence of pancreatic cancer in Japan.**Rateper 100 ,000 population, adjusted by the 1985 population in Japan.Age and cohort effect on the incidence rate Figure8depicts age and cohort effects on the incidence rate.Age effect increased with an advancing age similarly in both sexes.Cohort effect gradually increased in males born after * Proportion in total cancer death .**Rateper 100 ,000 population.*YearYear Figure 2. Trends in age-specific mortality rate for pancreatic cancer in Japan (1950-1994).Age AgeFigure 3. Mortality rate of pancreatic cancer by birth cohort.Age-specific incidence rate Figure7shows secular trends in age-specific incidence rate by 5-year age group.Likewise in mortality, age-specific inci-1896, whereas remained plateau or decreased a little in females born after 1926.