Cancer Research Activities of Cancer Epidemiology in Japan

Cancer epidemiology in Japan has been led by the large-scale population based cohort studies. Permanent registration system has made follow-up easier and more efficient. Low cancer incidence and mortality rates led the researchers to investigate possible preventive factors in diet and life style, in addition to the study of risk factors. The on-going large sized prospective cohorts should be a fruitful resource for future cancer epidemiological studies. Various studies of cancer epidemiology are introduced from a methodological aspect. cancer Epidemiology, cancer Registry, Cohort study, Mass-screening, cancer prevantion J Epidemiol, 1996 ; 6 : S19-S29.

Cancer epidemiology in Japan has been supported by the large volume of registries1-3). Mortality data have accumulated continuously for more than 100 years except for a few years of interruption during the Second World War. This system contributed to build the population-based and organ-specific cancer registries4). Geographically Japan is a long country from North to South. Cultural and geographic differences contribute to significant variation in mortality and life style between different regions of the country5,6). Cross-sectional study comparing SMRs and different life style and environmental factors often sheds light on factors relating etiology of cancer. Accurate registry data are available for such analysis. Different cancer mortality rates in the world were also observed 7). Recent knowledge of cancer prevention suggests that the low incidence and the mortality rate of cancer in Japan could be a result from dietary intake of cancer preventive chemicals. This hypothesis could be confirmed by two current on-going large population based prospective studies8). Recent development in clinical epidemiology is another feature in Japan. Multiinstitutional case-control studies and randomized clinical trials have been often performed by alert clinicians who recognize the importance of epidemiology. However, the number of biostaticians in this field is rather insufficient.
Despite of this shortage, mathematical models are proposed in theoretical epidemiology for tasks such as estimating the effect of intervention and to calculating power to detect risks, etc.9). This chapter discusses the features of some current Japanese cancer epidemiological studies.

DEMOGRAPHIC DATA FOR CANCER EPIDEMIOLOGY
In Japan, permanent registration system has been conducted since 1869. All death certificates should be reported to the Information Bureau of Vital Statistics, Ministry of Health and Welfare, and annual data have been published as Vital Statistics1). This gives a great advantage for descriptive epi-demiology2-6). After the Second World War, causes of death had dramatically changed from infectious diseases such as tuberculosis to chronic diseases. Cancer has steadily increased to became the primary cause of death since 1950s (Fig. 1). Stomach cancer has been the primary cause of cancer in both males and females until the lung cancer death rate in males exceeded the stomach cancer death rate in 199310,11). Mortality rates of stomach and uterine cancer have decreased steadily during the last 20 years (Fig. 2). Instead, colon and lung cancer in both males and females, and breast cancer in females have increased. In addition, liver cancer in males showed a birth-cohort effect, probably due to the poor nutritional state by the Second World War during their teenag12).
According to the advance of cancer therapy, collection of   cancer registry, which has been continued for more than 30 years, produced many epidemiological work by linking hospital data and registered patients17-20).
After stomach cancer consistently stayed as the primary cause of cancer death for many years, in 1965, clinicians began to register their patients in order to study method of diagnosis, treatment and survival rates4). Since then several other cancers were also registered. By now, 18 site-specific cancer registries are carried out by organization or research groups of expert clinicians (Table 1). Stage specific survival rates are recently published13)

Cross-Sectional Study :
Accurate statistical data of various areas in Japan could be used for ecological study. National Natritional Survey has been carried out on 7000 households each year since 194621).
This survey provides valuable information about the changes of Japanese diet. Due to the large geographical area covered by the country, the climate in each region and life-style of the inhabitants show considerable difference. Standardized cancer mortality ratios are thus diverted. This variation provides an advantage for conducting ecological studies22-24). For example, frequency of stomach cancer shows three fold difference between prefectures.
Various risk factors were proposed for stomach cancer, and recentry helicobacter pylori infection has been investigated as an important etiological risk factor for stomach cancer25-28).
Unusual cluster of special type of diseases could sometimes lead to the discovery of causative agent. Adult T-cell leukemia/lymphoma is such an example29). Geographic clustering of a peculiar form of leukemia led to the discovery of human T-lymphotropic virus type I29). Anthropological study on this viral infection has been done worldwide by Dr. Tajima and his group.
Cohort Study : As cancer may take many years to develop, a large-sized prospective cohort study is the most preferable method to find etiologic factors. "Six prefecture cohort study", which started in 1965 clarified the risk of active and passive smoking and preventive effect of green-yellow vegetables30,31). Cohort of atomic bomb survivors provided useful data for determining the association between the dose of irradiation and the risk of cancer 32,33) Currently three large population based cohort studies are being conducted8) ( Table 2). The outcomes of interest of the multipurpose Koseisho cohort study not only includes cancers but also cardiovascular diseases. It was started in 1990 and in 1993 with nearly 120,000 participants in 11 different health center districts, and will continue until year 2002. Other cohort study consisted of special occupational group is also being investigated currently. Follow-up study on retired soldiers who received sigmoidoscopy showed an association of increased large adenomatous polyp formation in the colon and alcohol consumption 34,35)  (Fig.3). The effectiveness of screening was measured mostly by case-control studies and each program is shown to be effective (Table 3)73-80. The high stomach cancer mortality rate in Akita prefecture decreased accordingly with increased proportion of population been screened. New biomarkers used to identify high risk group are employed in some studies ; i.e. helicobacter pylori antibody and plasma pepsinogen I and II81. Other new detection methods are under study ; mammography for breast cancer, helical computed tomography for lung cancer, sigmoidoscopy for colon cancer, and rectal ultrasonography or prostate-specific antigen for prostate cancer 83).
Neuroblastoma screening in infants has also been introduced in 1984, and the treatment method for Stage IVs is under debate now 82).

STUDIES ON CANCER PREVENTION
It is noteworthy that the totaled cancer incidence among Japanese has been relatively low, except for the stomach cancer. As described earlier, the standardized mortality rates showed great variation in Japan. Lowest area and highest area

NETWORK FOR MULTIDISCIPLINARY STUDIES
As described earlier, there exist many statistical data in Japan in addition to the death certificates. Nutritional survey has been conducted every year, population census is renewed every 5 years, and statistics of various industrial indices are revised annually. Some of these data even require special permission for use from the government. Nonetheless these are valuable data source for epidemiological studies, especially for biostaticians to deduct statistical models97.98). Environmental Agency also adds to the data set, the measurement of air pollution, water pollution, ultraviolet ray, and etc.99). These data are also applicable for epidemiological studies. Cross-sectional or ecological study is always effective as the first step of epidemiological study. With this regard, analysis performed by linking this large stack of information may open new field in epidemi-ology100) Autopsy records include more than 40,000 autopsies performed every year. This additional information also provides histological data of the tumor101). Multiple primary cancer cases can be detected by linking hospital registries l02-106) Drug-related carcinogenesis has also listed, and used in the clinical epidemiological study107-109). In Japan, randomized clinical trials are carried out under strict regulations, and epidemiologists are called to join the studies in order to assist clinicians. However, this area is not described here.

MODELS
To deduct statistical model is always important in theoretical epidemiology. Non-parametric Markov model needs a lot of epidemiological data in order to calculate transitional rates of risk9). Other models including cancer growth need basic knowledge in biology110-113)

EPILOGUE
The number of epidemiologists in Japan occupied about 2% of all cancer researchers. Funding has also stayed at two percent of all cancer fund from the government for many years. Since the establishment of Japanese Epidemiological Association in 1993, however, the number of members in the association has increased continuously. Limited by only one existing public health school in Japan, most epidemiologists are trained in the medical school. Therefore, more than 90% members in this organization have medical degree. It can be said that physician is a term that generally characterize the members in the Japanese Epidemiological Society.
The fields in epidemiology is expanding. We can distinguish four areas in epidemiological development ; namely epidemiology for acute infection (first generation), for chronic diseases (second generation), for clinical evaluation (third generation), and for accumulation of large volume of information for epidemiological research (fourth generation). Computer network can open a new field of research in the near future.