REPORTS ON SEMINARS

(1) Seminar on “Statistical Methods in Cancer Epidemiology”

The U.S.-Japan Conference on “Statistical Methods in Cancer Epidemiology”, was held successfully at the Radiation Effects Research Foundation, Hiroshima, March 1-2, 1984 attended by many young epidemiologists as observers in addition to regular participants from U.S., and Japan.
The outlines of the meeting were as follows:
I. Introduction
The history of the U.S.-Japan Cooperative Cancer Research Programme was reviewed with special reference to cancer epidemiology and biostatistics by Dr. Hirayama. Then the role of statistician in cancer research was outlined by Dr. Miller allocating each participant to suitable place on the following scale:
II. Descriptive Studies
A method of test of significance for geographic cluster was proposed by Dr. Ohno using a simple chi-square test, validity of which was confirmed by a Monte Carlo simulation method.
Using this method, the geographical clustering of cancers of esophagus and cervix was shown significant while that of urinary bladder and breast was considered not to be significant.
A new statistical method to rearrange and simplify similar frequency areas on a disease map was introduced by Drs. Aoki, Mizuno.
A constellation graphical methods with the angle parameter to visualize disease clusterings in multidimensional data was proposed by Dr. Wakimoto.
III. Case-control Studies
Recent developments in the design and analysis of case-control studies was summarized by Dr. Breslow. Issues discusses include limited role and efficiencies of matching in design, a better method of estimating variances for the Mantel-Haenszel estimate, multivariate risk functions, regression diagnostics for matched case-control studies, “synthetic” case-control study and attributable risk.
The problems in using “death certificate controls” for cases of fatal disease was discussed by Dr. Blot. He calculated odds ratio (OR) by assuming death certificate controls as cases and living population controls as controls and found the effect of smoking could be underestimated if death certificate controls were used. Overestimate of the amount of cigarette smoking by next-of-kin and underestimate of the history of exposure to asbestos by next-of-kin were observed.
The association of smoking with eating and drinking habit was discussed from the points of view of biological and psychological association by Drs. Aoki, Mizuno and by Dr. Hirayama.
Panel discussion of methodologic issues in case-control studies followed.
IV. Cohort Studies
Major results of the Prospective Study (1966-81) involving over 260,000 population aged 40 years and over was presented by Dr. Hirayama. Besides results on the individual effect of smoking, drinking, and other dietary habits (intakes of beef and green-yellow vegetables etc.) on various sites of cancer, risks of cancer in people with life styles similar to SDAs (Seventh Day Adventists) and non-SDAs were compared by considering smoking, drinking, and intakes of beef and green-yellow vegetables simultaneously.
Recent developments in statistical methodology of the analyses of cohort data was presented by Dr. Breslow. Major points discussed include the standardized mortality ratio (SMR), Mantel-Haenszel estimator in cohort studies and comparison of various mathematical models in fitting exposure-relative risk curves.
Results of analyses of mortality and disease incidence among atomic bomb survivors was presented by Drs. Kato, Preston and Kopecky based on Life Span Study sample by using the multiplicative model and the additive model. The former model fitted better for most sites of cancer, but the latter model fitted better for acute leukemia.
Details of the record linkage system in the Osaka Cancer Registry and its applications in cancer epidemiology was presented by Dr. Fujimoto. Examples of cohort studies by record linkage were (1) second primary cancers among cervical cancer cases received radiotherapy, (2) follow-up of examinees of stomach cancer screening to study beneficial and adverse effects of screening and to study the natural history of stomach cancer, (3) follow-up of HBsAg(+) blood donors to study the risk of developing liver cancer and liver cirrhosis.
A new method of collating possibly same two persons by using statistical probability was presented by Dr. Ohtaki.
V. Clinical Studies and Screening
Several examples of application of various statistical methods to data of several epidemiological, clinical and experimental studies were presented by Dr. Yanagimoto.
Results of the studies of natural history of early stage stomach cancer was discussed by Dr. Ohshima. By following up unoperated endoscopically diagnosed early stage stomach cancer patients, the median time from the detection of early stage cancer to advanced cancer was found as 37 months and the median survival time of those patients as 77 months. These data were considered useful in estimating the efficacy of stomach cancer screening.
The second study was on the risk of developing liver cancer and liver cirrhosis among HBsAg(+) blood donors. The O (observed number)/ E (expected number) ratio of developing liver cancer was 7.12 while that of stomach cancer was 0.93.
The principles of screening was discussed in detail by Dr. Morrison. Statistical methods for estimating the possible lead time bias were shown. The importance of getting data to demonstrate the efficacy of screening by conducting a randomized trial before widely promoting cancer screening programmes was emphasized.
Panel discussions on methodologic issue in cohort and clinical studies followed.
VI. Conclusions
Views on the future directions for biostatistics and cancer epidemiology in Japan was presented by Dr. Shigematsu. The situation of developing biostatistics in Japan had not been substantially improved since the last U.S.-Japan meeting on the similar issue in 1978 and no faculties nor departments have been established yet which provide systematic teaching programmes for biostatistics. Meanwhile, much efforts were thought necessary to strengthen the cooperation among biostatisticians, physicians and biologists in conducting epidemiological studies on cancer.
The Meeting as this nature was felt strongly to be quite valuable especially for cancer researchers in Japan. The proceedings of the meeting will be published as a RERF Monograph.

(2) Seminar on “The Role of Pathologists in Cancer Epidemiology”

At the opening, as background information for this workshop, Dr. Sugano explained the time-trends for cancers in Japan and also the difference in the incidence of major cancers in U.S. and Japan. In males, the incidence of most cancers is low, only that of stomach cancer is extremely high in Japan, whereas, in the United States, most cancers have a high incidence; the highest is lung cancer followed by prostate and colon cancer. In females, stomach cancer also shows a high frequency and other cancers are low in frequency in Japan, on the contrary, in the United States most cancers are more frequent, and breast cancer is particularly prominent. Time-trend data indicate that cancer incidence in Japan generally tends to “westernize.”
Dr. Urano reviewed the system and results of the National autopsy registration in Japan which has been conducted by the Japanese Pathological Society for the last 25 years (1958-1983). About 400,000 autopsy cases were registered and 230,000 cases (1974-1981) were computerized. It is available to display various items, such as main cause of death, age, sex, geographic location and so on. Participants were interested in this unique system in the world, but Dr. Higginson pointed out that selection bias in autopsies is unavoidable, so that adequate adjustment is necessary. The importance of histologic subtypes were also stressed. Dr. Stemmerman mentioned a possibility of offering the data of Hawaian Japanese autopsied at Kuakini Hospital.
Dr. Mori talked about the multiple etiology of hepatocellular carcinoma in Japan. The incidence of hepatoma is still very high, and most cases are related to hepatitis B virus infection accompanying liver cirrhosis and fibrosis. Using recombinant DNA technology and integration of HBV genome in hepatoma DNA was probed and the role of HBV genome in hepatocarcinogenesis was noted. He also pointed out a small number of hepatoma cases related to schistosomiasis japonicum (in the Fukuoka area), alcoholic cirrhosis (in urban areas, vinyl chloride, thorotrast, but no adenoma cases by contraceptive pills were found in Japan. Dr. Higginson discussed the co-carcinogenic role of aflatoxin B₁ and malnutrition.
Dr. Kitagawa discussed occupational lung cancer in Japan. He mentioned an etiological role of asbestos, chromate, mustard gas and generator gas. He emphasized the long latent period, cocarcinogenic effect with smoking and high relative risk in cancer development in these cases. He also mentioned the importance of precancerous lesions, such as pleural plaques in asbestos cases and squamous metaplasia in mustard gas and in chromate cases. Dr. Miller pointed out another two kinds of occupational lung cancers in Japan; lung cancers among copper smelterers in Kyushu and atomic-bomb survivors in Hiroshima and Nagasaki.
Dr. Sugano discussed the histopathology and time-trends of carcinoma of the stomach and colon. He emphasized the difference between two kinds of carcinomas, variable cancer and basic cancer; the former is much more influenced by the environment, while the latter is less influenced by the environment. In the stomach, the poorly differentiated cancer (diffuse type) is the basic, and well differentiated cancer (intestinal type) is the variable type. Further, he stated that microcarcinomas of the stomach mostly do not develop into a clinical cancer, but remain as a latent cancer. Dr. Stemmerman discussed the frequency of stomach cancer in aged autopsy cases in Hawaii Japanese; it is almost the same as that for Japanese in Japan. He also noted that cancer incidence is closely related to socioeconomic state. Dr. Miller pointed out the importance of genetic background of the basic cancer.
Dr. Yatani presented findings in clinical and latent carcinoma of the prostate. The incidence of clinical cancer is very high in Americans and western people, but very low in Japanese. When comparison is made of rates for latent cancer among nations, the frequency of age-specific in situ or minimal invasive latent cancer is the same in western countries and Japan. On the contrary, invasive latent cancer increased with age in western pepole, but not in Japanese. Therefore, it is postulated that both the western and Japanese prostate is already initiated (cancerized), but promotion or progression is different; the western prostates are strongly promoted, whereas the Japanese are far less so. Dr. Matsumoto discussed that there are some similarities from the endocrinological viewpoint between prostate cancer and breast cancer. In relation to Dr. Yatani’s presentation on less susceptible in Japanese prostatic cancer, sensitivity to treatment was discussed. Dr. Stemmerman pointed out the poor prognosis in patients with the lowest cholesterol intake.
Dr. Matsumoto discussed a comparative study of breast cancer in Japanese and Americans. Breast cancer probably provides one of the most impressive contrasts between the two countries. He showed that the incidence is very high in Americans, whereas it is very low in Japanese. The dissimilarity appears to be due to a difference in the frequency of postmenopausal cancer. This cancer shows a high incidence and high positive rate for estrogen receptor (ER) in Americans, and is etiologically related to an excess endogenous estrogen which is closely related to obesity. Therefore postmenopausal fat women are at high risk for breast cancer. Dr. Stemmerman said that lactation in Japanese and Americans is different. In situ carcinoma of the breast is three-fold higher in Japanese than in other races, and the frequency of clinical breast cancer in Hawaiian Japanese is approaching that of Americans.
Dr. Hanaoka discussed a most interesting disease, adult T-cell lymphoma (ATL), which is etiologically related to human retrovirus (ATLV). It is endemic in Southwestern Japan. Both males and females are affected. Cutaneous lesions and hypercalcemia are commonly associated. Leukemic cells derive from mature T-cells with nuclear pleomorphism in which the HTLV genome exists as an integrated clone. In Uwajima, an endemic area, a high frequency of healthy carriers has been observed within families, which may indicate familial transmission of this virus. Also seroconversion by blood transfusion was observed. Dr. Sugano mentioned that according to Dr. Yoshida’s recent work, ATLV is identical to HTLV, which Dr. Gallo identified from cutaneous T-cell lymphoma from Caribbean territories. There were some reports that a similar virus is related to acquired immunodeficiency syndrome (AIDS), as well as to similar virus(es) of widely infected monkeys in Japan, East-South Asia and Africa, but not in South America. The transmission of this virus was extensively discussed.
Dr. John Higglnson (UAREP, Bethesda) described with eloquence the relationship between epidemiology and pathology, as well as the relationship of both to nutritional, metabolic and biochemical studies. He favors the epidemiologic approach as it ralates to the difficult problem of sorting out the role of diet and other features of life-style with respect to the more common cancers, and he introduced the importance of histologic subtypes in epidemiologic research.
Dr. John W. Berg (Denver) expanded on histologic refinement by noting that cigarette-smoking induces subtypes of lung cancer in proportions that are different in males vs females and whites vs blacks. Cigarette-smoking by females has produced a greater increase in oat cell cancers, whereas in males it has caused a rise primarily in squamous cell cancers. In Black males, the excess in recent years over that expected from studies of white males is due either to an increase in squamous cell cancers and/or low frequencies of other types of lung cancer. We do not know why only one cell type is selectively increased. Studies of other cancers by subtype are needed, but descriptive epidemiology has gone out of fashion Dr. Berg said.
He described his beginning use of numerical taxonomy (cluster analysis) to measure “demographic distances” between groups of subtypes classified by age, sex, race and time (Berg, J. W.: The epidemiologic meaning of histology in lung cancer In Correya P. et al: Lung Cancer. Verlag Chemie International Inc. In press).
Dr. Grant N. Stemmerman (Honolulu) spoke in part about differences in subtypes of cancer among Japanese who migrated to Hawaii as compared with those who remained in Japan. For example, the age-standardized annual rate for epithelial cancer of the ovary was 6.61 per 100,000 per year in Hawaii vs 1.79 in Miyagi Prefecture, Japan. Female breast cancer among the Japanese was of a different character than that among U.S. women. The Japanese more often had in situ lesions or nuclear grade 3, for example. Trends in cancer occurrence among Hawaiian Japanese women showed that the age-adjusted mean rate for breast cancer rose from 25.49 per 100,000 in 1962-65 to 47.63 in 1978-81, for uterine corpus cancer the rise was from 8.74 to 16.41, and for ovarian cancer, from 8.47 to 12.71. Substantial declines occurred in each sex in the age-adjusted mean rates for cancers of the colon and stomach. These observations dramatically reflect as yet unknown environmental influences on cancer occurrence.
Dr. J. Bruce Beckwith (Seattle) described two important recent advances in the differential diagnosis of Wilms’ tumor, achieved through review of the pathology of a large case-series (the National Wilms’ Tumor Survey). The histology of the tumors from patients who did not respond to therapy revealed not nephroblastoma, but previously unknown entities clear-cell sarcoma of the kidney which frequently metastasizes to bone, and rhabdoid tumor which has now occurred 17 times with medulloblastoma as a second primary tumor
The other new observation concerns the location of dysplastic renal tissue, which is perilobular in unilateral nephroblastoma or when hemihypertrophy is present, but is usually intralobular as well as perilobular when Wilms’ tumor is bilateral or occurs with aniridia. (We noted that these differences parallel those in graphs of the age at diagnosis, which on a semi-log plot are curvilinear for unilateral or hemihypertrophy-related Wilms’ tumor, but are linear for bilateral or aniridia-related Wilms’ tumor (Knudson and Strong, JNCI 48:313-324, 1972).
Dr. David T. Purtilo (Omaha) spoke on the reciprocal relationship between lymphoproliferative and autoimmune disorders among U.S. whites vs Japanese, and in males vs females. The differences may be due to various factors that regulate the proliferation and function of the immune system, including estrogen and testosterone, genetics, diet (especially animal lipids and beef protein in the U.S. as contrasted with fish and vegetables in Japan), and exposures to Epstein-Barr virus, adult T-cell leukemia virus and various chemicals. Autoimmune disorders are attributable to factors that cause B cells to produce autoantibodies. The emergence of malignant lymphoproliferative disease may be atrributable to agents that induce molecular or cytogenetic rearrangement of immunoglobulin genes and oncogenes. (Ed. note: The scope of the differences in frequencies of lymphoproliferative vs autoimmune disorders can be evaluated in Hawaii where medical care and diagnoses whould be similar for various racial groups. Incidence data for several autoimmune diseases are available in Japan, and are less easily available in the U.S. It would be of interest to determine if the frequencies of lymphoproliferative and autoimmune diseases in China are more similar to those in Japan or to those in the U.S.).



SEMlNAR AGENDA AND PARTICIPANTS

(1) SEMINAR ON STATISTICAL METHODS IN CANCER EPIDEMIOLOGY
Hiroshima, Japan, March 1-2, 1984

AGENDA