Mortality of hepatitis B surface antigenpositive blood donors in England and Wales

Transcription

1 International Epidemiological Association 2003 Printed in Great Britain International Journal of Epidemiology 2003;32: DOI: /ije/dyg039 Mortality of hepatitis B surface antigenpositive blood donors in England and Wales Paul D Crook, 1 Michael E Jones 2 and Andrew J Hall 1 Accepted 14 October 2002 Background Large population-based cohort studies in areas of high hepatitis B virus (HBV) prevalence have provided the evidence establishing hepatitis B surface antigen (HBsAg) carriage as a risk factor for hepatocellular carcinoma (HCC) and liver disease. Fewer studies have examined this in Western countries, where both HBV infection and carriage are less common and transmission patterns differ. This is the only prospective population-based study to examine this relationship in Europe. Methods Results Conclusion In all, 2681 male and 977 female blood donors in England and Wales, found to be HBsAg positive during routine blood-donation screening, were followed up from recruitment in to December 1999 and their cause-specific mortality was analysed. This was compared with that of the general population of England and Wales. During a mean of 22 years of follow-up, 17.4% of the 420 deaths were due to HCC or liver disease. There were 20 deaths from HCC in male HBsAg carriers, representing a significantly high standardized mortality ratio (SMR) compared to the male population of England and Wales of 26 (SMR = 26.26; 95% CI: ). The HCC incidence rate in males was 33.5 per person years and 4.4 per person years in females. Men had 8.5 (SMR = 8.50; 95% CI: ) and women had 3.9 times the risk of death from liver disease (SMR = 3.89; 95% CI: ). The risk of circulatory disease deaths was reduced in both males and females. There was a significant increased risk of non-hodgkins lymphoma that was not apparent in the first decade of follow-up. The increased risk of HCC and liver disease in men fell with follow-up. Hepatitis B surface antigen carriage is a significant risk factor in England and Wales for both liver disease and HCC mortality. However, this risk has declined with duration of follow-up. This could be due to natural reversion to HBsAg negativity or as a result of treatment and avoidance of other risk factors. The increased risk of non-hodgkins lymphoma seen in longer follow-up is likely to be related to HIV infection acquired subsequent to recruitment. Keywords Hepatitis B virus, hepatocellular carcinoma, chronic liver disease, cohort, population-based There are an estimated one million new hepatitis B virus (HBV) infections and new carriers of HBV each year in Europeans. 1 In 1994, the World Health Organization s (WHO) International Agency for Research on Cancer concluded that 1 London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK. 2 The Institute of Cancer Research, Epidemiology Section, Block D, Sutton SM2 5NG, UK. Correspondence: Andrew J Hall, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK. andy.hall@lshtm.ac.uk. there was sufficient evidence that chronic infection with HBV is carcinogenic to humans. 2 The classic cohort study in Taiwan showed the relative risk of hepatocellular carcinoma (HCC) for hepatitis B virus surface antigen (HBsAg) positive compared with HBsAg negative subjects to be Further populationbased cohort studies, including those using blood donors as a source of HBsAg carriers, have confirmed HBsAg to be a risk factor for HCC in other populations This prospective study, which examines the mortality of HBsAg-positive blood donors, was established in England and 118

2 HEPATITIS B SURFACE ANTIGEN-POSITIVE BLOOD DONOR MORTALITY 119 Wales in 1983, at a time when there was considerable evidence linking HBV to chronic liver disease and HCC in HBV endemic countries, such as China. However, it was doubtful whether this evidence could be generalized to different populations since aflatoxin is thought to play an important role in modifying risk of HCC in endemic areas. 21 In addition, the major route of transmission in these countries differs from that in Europe. In China, some 40% of carriers are determined by perinatal transmission from an infectious carrier mother, whereas in Africa and other endemic areas child-to-child transmission early in life ( 5 years of age) is the predominant determinant of carriage. The risk of HCC may vary by route of transmission as suggested by a comparison of cohort results in China and Senegal. 22 Therefore, this study was primarily established to explore to what degree HBsAg was a risk factor for HCC in England and Wales in males. Results from an earlier analysis of this study have previously been published. 18 This paper extends follow-up of the cohort to 1999 a further 15 years of data. This increases the power to examine association with less-common causes of death and to examine how risks have changed by calendar period. Materials and Methods This is a prospective cohort study of HBsAg-positive blood donors comparing mortality rates in the cohort with the general population. Details from all blood donors in England and Wales found to be HBsAg positive during routine screening between 1970 and 1982 were collected. Blood donation was voluntary (i.e. unpaid) and to exclude those with a current or previous history of health problems, blood donors were typically screened with a questionnaire and then by blood pressure examination and a haemoglobin estimate. However, screening procedures were not uniform throughout the country and did not include details on sexual or travel history or injecting drug use. The tests used for the detection of HBsAg varied during recruitment into the study, both between regional transfusion centres and over time. Transfusion centres typically used immuno-electrophoresis up until 1974, changing to reverse passive haemagglutination (RPHA) in 1975 and an improved haemagglutination technique in late Donors found to be HBsAg positive were generally advised to seek medical advice. There was no follow-up testing to discover whether the HBsAg positive status persisted. Full name, date of birth, place of birth, sex, area of blood donation and date of detection of HBsAg were collected for each patient. The entry date into the study was taken as the date of HBsAg detection. Where place of birth information was missing, the surname was used to estimate ethnicity. Subjects identified were discarded if they lacked sufficient identifying details for follow-up. The National Health Service Central Register (NHSCR) was sent details of the remaining subjects for flagging. The NHSCR is a virtually complete population register of all residents of England and Wales continuously updated with information regarding vital status events, e.g. all notifications of deaths, emigrations, and cancer registrations but with no information on specific health problems such as human immunodeficiency virus (HIV), hepatitis C virus (HCV), liver cirrhosis or details of treatment. Follow-up information on the cohort members was obtained from this register. For each subject of the cohort who died, a copy of the death certificate was obtained which contained the coding for the underlying cause of death by the International Classification for Diseases (ICD) revision in use at the time of death. For the analysis, bridge coding of ICD revisions was conducted to give the corresponding Ninth Revision (ICD-9) categories. 23 The general population of England and Wales was used as the control population. National cause-specific mortality rates by age, sex and calendar year were obtained from the Office for National Statistics for 1971 to The cohort mortality data were compared to those of the general population. For each cohort member, person-years at risk were calculated by age, sex, and calendar year from the time the subject entered the study and finishing at the date of death, emigration, 85th birthday or 31 December 1999, whichever was earliest. Follow-up was truncated at age 85 as the quality of the information on the death certificates relating to cause of death is considered to be poor after this age. Expected deaths by cause in the cohort were calculated by multiplying the age-, sex- and year-specific person-years at risk within the cohort, by the corresponding national cause-specific mortality rates using the computer program, PYRS. 24 Standardized mortality ratios (SMR) were then calculated by dividing the number of deaths in the cohort with the expected number. Confidence intervals and two-sided P-values were calculated based on the Poisson distribution. The SMR were calculated for different calendar year and follow-up periods and for different ages at entry to establish if any of these factors modified the effect of HBsAg carriage. Ethical approval was obtained from the University of Southampton. Permission to use information in the NHSCR was obtained from NHS research and development. Results Subjects In all, 4118 HBsAg-positive blood donors were entered into the study. Only 3938 had adequate identifying details and of these, 3665 (93%) could be traced by the NHSCR. A further six had emigration dates prior to entry into the study and were discarded. Another subject was discarded as the date of birth appeared inaccurate. This left 3658 in the final analysis, of which 2681 (73%) were male and 977 (27%) were female. The median age of entry was 29.1 years (range ) for males and 28.5 years (range ) for females. Place of birth was obtained in 78% of subjects, (67% from the UK and 11% from overseas). In the rest of the subjects the place of birth was estimated from their name. Using this method 81% of all subjects were classified as born in the UK. Of those classified as born overseas, the largest group, of 165 subjects, was from the Indian Subcontinent. There were 141 individuals from South East Asia. The cohort members were followed up for a total of person-years, a mean of 22.4 years per subject. Those subjects surviving until the end of the study were followed up for 17.0 to 29.4 years. All-cause mortality During the follow-up, 422 (11.5%) subjects died and 96 (2.6%) subjects emigrated. At the study closure date on 31 December 1999, 3140 (85.8%) were alive. There were 12 subjects who were censored when they reached the age of 85 years and two

3 120 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY of these had died by the closure date. These two deaths were not included in the person-years SMR analysis. Of the 420 analysed deaths, 363 (86%) were in males and 57 (14%) in females. Causes of death are shown in Table 1. Five per cent of deaths were due to HCC and 12.4% were due to non-malignant liver disease. No information was available regarding the cause of death of the four subjects who died overseas. The primary results of the study are displayed in Table 2. These are SMR comparing mortality rates in the HBsAg-positive blood donor cohort with the general population of England and Wales, adjusting for age and calendar year. The all-cause mortality rate was increased in male HBsAg-positive blood donors but was not statistically significantly higher than the general population (SMR = 1.10; 95% CI: ). In females, the all-cause SMR was statistically significantly reduced by 26% (SMR = 0.74; 95% CI: ). Hepatocellular carcinoma mortality Those who died of HCC were significantly older than the rest of the cohort at entry (median age 39.1 years versus 29.0 years, P = 0.010). The subjects died of HCC after a mean of 13.3 years follow-up time (range ). The mean age at death was 50.6 years (range ). The death certificates of the HCC deaths were examined further and five certificates noted liver cirrhosis, one noted chronic active hepatitis and one noted oesophageal varices as causes. Only one certificate listed HCV as the underlying cause of the HCC. The mortality rate due to HCC in all males was statistically significantly higher in HBsAg-positive blood donors, over 26 times that of the general population (SMR = 26.26; 95% CI: ). Only one death occurred in the all female cohort compared with the 0.10 expected. This increased rate was not Table 1 Underlying cause of death of the study cohort statistically significant. The SMR for HCC in all males according to age at entry and follow-up period are presented in Table 3. With continuing follow-up the SMR for HCC has fallen, from 51 in the first decade of follow-up to 22 in The SMR were calculated separately for each birth origin classification (Table 2). The death certificates of all the HCC cases were examined for place of birth. In two HCC cases originally estimated as being born in England and Wales, the death certificates identified the place of birth to have been overseas, one subject being born in India, the other in Hong Kong. In all, seven subjects with HCC were born overseas, three in India and one from each of Cyprus, Germany, Hong Kong and Nigeria. The HCC incidence rate in all males was 33.5 per person years (95% CI: ) and 4.4 per person years in all females (95% CI: ). The HCC incidence rate in those men classified as born overseas was 33.9 per person years. Deaths from non-neoplastic liver disease There were 52 deaths from non-neoplastic liver disease in the cohort, 47 (90.4%) in males and 5 (9.6%) in females. The median age of entry for those with liver disease deaths was 43.2 years (range ). Liver disease deaths occurred after a mean of 14.2 years of follow-up (range ). The mean age at death was 54.8 years (range ). The death certificates were examined further for underlying causes and contributing factors; 27 (51.9%) listed cirrhosis, 5 (9.6%) listed chronic active hepatitis, 7 (13.5%) listed alcohol and only one (1.9%) listed HCV as an underlying cause. Both sexes had a significantly raised SMR for liver disease (Table 1). The rate of liver disease mortality was 8.5 times higher in all males and 3.9 times higher in all females than the Cause of death (ICD-9 a code) Males Females Total Causes of particular interest Hepatocellular carcinoma (155.0) 20 (5.5%) 1 (1.8%) 21 (5.0%) Non-malignant liver disease (070, ) 47 (12.9%) 5 (8.8%) 52 (12.4%) Other causes by ICD-9 category Infective and parasitic diseases ( ) 18 (4.9%) 2 (3.5%) 20 (4.8%) AIDS (079.9, 279.1, 042.9) b 5 (1.4%) 0 5 (1.2%) Viral hepatitis (070) 13 (3.6%) 2 (3.5%) 5 (1.2%) All neoplasms ( ) 133 (36.6%) 27 (47.4%) 160 (38.1%) Lung cancer (162) 42 (11.6%) 8 (14.0%) 50 (11.9%) Non-Hodgkin s lymphoma (200, 202) 11 (3.0%) 0 11 (2.6%) Endocrine, nutritional, metabolic diseases and immunity disorders ( ) 7 (1.9%) 0 7 (1.7%) Mental disorders ( ) 3 (0.8%) 1 (1.8%) 4 (1.0%) Diseases of nervous system and sense organs ( ) 6 (1.6%) 1 (1.8%) 7 (1.7%) Diseases of the circulatory system ( ) 88 (24.2%) 11 (19.3%) 99 (23.6%) Ischaemic heart disease ( ) 22 (6.1%) 3 (5.3%) 25 (6.0%) Cerebrovascular disease ( ) 13 (3.6%) 4 (7.0%) 17 (4.0%) Diseases of the respiratory system ( ) 29 (8.0%) 4 (7.0%) 33 (7.9%) Diseases of the digestive system ( ) 36 (10.0%) 4 (7.0%) 40 (9.5%) Liver disease (not acute hepatitis) ( ) 34 (9.4%) 3 (5.3%) 37 (8.8%) Diseases of the genitourinary system ( ) 1 (0.3%) 1 (1.8%) 2 (0.5%) Diseases of the skin and subcutaneous tissue ( ) 0 1 (1.8%) 1 (0.2%) Injury and poisoning ( ) 34 (9.4%) 4 (7.0%) 38 (9.0%) Congenital anomalies ( ) 0 1 (1.8%) 1 (0.2%) Symptoms, signs, and ill-defined conditions ( ) 4 (1.1%) 0 4 (1.0%) Died abroad with no details as to the cause 4 (1.1%) 0 4 (1.0%) Total a International Classification of Diseases, Ninth Revision (ICD-9). b Includes codes not part of the Infective and parasitic diseases ICD-9 category.

4 HEPATITIS B SURFACE ANTIGEN-POSITIVE BLOOD DONOR MORTALITY 121 Table 2 Standardized mortality ratios (SMR) by International Classification of Diseases, Ninth Revision (ICD-9) classification categories by sex a Male Female Cause of death (ICD-9 codes) Obs. Exp. SMR 95% CI P-value Obs. Exp. SMR 95% CI P-value All causes ( ) All causes All causes except liver disease and HCC b HCC stratified by place of birth (155.0) England and Wales Overseas Total Non-malignant liver disease stratified by place of birth (070: ) England and Wales Overseas Total All neoplasms (includes HCC) ( ) All malignant neoplasms (includes HCC) ( ) Malignant neoplasms minus HCC Non-Hodgkin s lymphoma (200, 202) Malignant neoplasm without specification of site (199) Infective and parasitic ( ) AIDS c (079.9, 279.1, 042.9) Viral hepatitis (also included in non-malignant liver disease above) (070) Circulatory system ( ) Ischaemic heart disease ( ) Cerebrovascular disease ( ) Digestive system ( ) Mental disorders ( ) Injury and poisoning ( ) a Obs. = observed; Exp. = expected number of deaths based on England and Wales age-, sex- and calendar-year specific rates; SMR = (Obs./Exp.) = standardized mortality ratio; P-value = two-sided, based on Poisson distribution to 1 d.f. b Hepatocellular cancer. c Includes codes not part of the Infective and Parasitic diseases ICD-9 category. general population (SMR = 8.50; 95% CI: in males: SMR = 3.89; 95% CI: in females). The SMR for all males for liver disease according to age at entry and follow-up period are presented in Table 3. With continuing follow-up the SMR for liver disease in men has fallen from 10 in the first decade to 6 in The SMR according to country of origin are displayed in Table 2. The incidence rate for liver disease mortality in all males was 78.8 per person years (Table 3. 95% CI: ). Males born in England and Wales had a higher liver disease mortality incidence rate (91.3 per person years) than overseas males (26.2 per person years). The liver disease mortality incidence rate in females was 22.1 per person years (95% CI: ). Other disease mortality There was a significantly high SMR for the all malignancies category in males (Table 2). However, when deaths due to HCC were removed from the analysis the SMR, although high, was no longer significantly so. There was a non-significant increase in lung cancer (SMR = 1.33; 95% CI: ). Men had 3.5 times higher risk of death from non Hodgkins lymphoma (NHL) than expected. This increase was significant (SMR = 3.48; 95% CI: ). Human immunodeficiency virus or AIDS were not noted on the death certificates as the underlying causes for NHL. In addition, the SMR for malignant neoplasm without specification of site was significantly raised in males (SMR = 2.13; 95% CI: ). There was no significant increased mortality risk in any other malignancy. As a result of the increased mortality from viral hepatitis, the SMR for infective and parasitic diseases was significantly raised in males. Examining HIV and AIDS mortality is complicated as ICD codes changed markedly over the study period. Examination of the death certificates identified five male deaths from AIDS. The SMR in males for the ICD-9 codes assigned to these deaths was found to be significantly high (SMR = 7.42; 95% CI: ). There was a significantly high SMR in males for digestive system disease. However, all but two deaths in this category were due to liver disease. The SMR for circulatory diseases was significantly low in both males (SMR = 0.62; 95% CI: ) and females (SMR = 0.45; 95% CI: ). There were fewer deaths than expected from both ischaemic heart disease and cerebrovascular disease in both males and females. However, the SMR was only significantly reduced in male ischaemic heart disease (SMR = 0.59; 95% CI: ) i.e. males had a 41% reduction in mortality from ischaemic heart disease. Discussion This is the only prospective cohort study to follow-up populationbased HBsAg carriers in England and Wales to establish the

5 122 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Table 3 Incidence rates and standardized mortality ratios (SMR) for hepatocellular cancer (HCC) and non-malignant liver disease in the cohort for different age at entry and follow-up periods in all males a Incidence rate/ Obs. Exp. SMR 95% CI person years 95% CI HCC Age at entry (years) Follow-up period (years) Calendar time Total Non-malignant liver disease Age at entry (years) Follow-up period (years) Calendar time Total a Abbreviations as in Table 2. incidence and relative risk of HCC. This follow-up now exceeds 20 years per carrier. It shows a strong and significant association between HBsAg carriage and HCC in men. The mortality rate for males was 26 times that of the general population. This increased rate of HCC was not limited to immigrants. The rate ratio in those classified as born in the UK was 23. The limited number of women in the study limits statistical power and the one observed death from HCC does not result in a statistically significant result although it is consistent with a tenfold increased risk. The only other study to examine relative risk in women was in Japan where they identified a relative risk of Non-malignant chronic liver disease is an important cause of death in HBsAg carriers, accounting for 12.4% of deaths in our cohort, greater than those from lung cancer (11.9%). This study found a strong association between HBsAg carriage and liver disease with significantly high rate ratios of 8.5 in males and 3.9 in females. The high SMR for liver disease was not due to deaths in subjects born overseas. Alcohol and HCV both cause liver disease and HCC. However, they were not measured in this study and could not therefore be controlled for in the analysis. However, HCV was noted as the underlying cause of only one of the liver disease deaths and alcohol was an underlying cause in 13.5% of the deaths. This suggests that they are unlikely to have had a major influence on these results. The rate ratio of 26 for the association between HBsAg carriage and HCC for males compares to similar studies in the US (where the pattern of transmission is similar) which report rate ratios of 10 and ,15 A much higher rate ratio of 148 was found in another US study, but this was amongst a particular ethnic subset of the population, Alaskan Natives. 7 The results from studies in Asia vary considerably even within countries. Rate ratios ranging from 5 to 103 have been reported from China, Taiwan and Japan. 3,8 13,16,17 Studies from Asia have reported much higher crude HCC incidence rates, as high as 651 cases per person years, in HBsAg carriers in China. 9 This increased incidence is likely to be due to the high proportion of infections that occur due to perinatal infection and to other co-factors such as aflatoxin. Few population-based cohort studies examine the risk of nonneoplastic liver disease mortality in HBsAg carriers. The rate ratio in males for liver disease mortality of 8.5 found in our cohort is higher than a similar study in the US where a rate ratio of 1.4 was reported for liver cirrhosis mortality. 15 Several points need to be considered in interpreting this study. Blood donors are a self-selected healthier group of people who are further screened for ill health and one expects a considerable healthy donor effect. 25 Therefore, for each particular cause of mortality one would generally expect selection bias to lead to a reduction in the relative risk. Indeed this was observed for all-cause mortality in women and for cardiovascular disease in both sexes. Despite the healthy donor effect there was a non-significant increase in lung cancer deaths in our cohort. This suggests that the cohort smoked more than the general population and could not explain this reduction in cardiovascular deaths. A second bias inherent in this design is misclassification. The controls (the general population) also include HBsAg-positive

6 HEPATITIS B SURFACE ANTIGEN-POSITIVE BLOOD DONOR MORTALITY 123 subjects. However, since the prevalence of HBsAg carriage in the UK is around 0.5% 26 this will not have a major influence on the estimate. More importantly, subjects were entered into the study on the basis of just one positive HBsAg test defining HBsAg carriage, rather than positive test results over a period of 6 months. Since an estimated 1% of HBsAg carriers clear the antigen per annum, 27 and a few subjects may have had acute infection, a proportion of the cohort may no longer be carriers. This is probably the reason behind the fall in SMR with length of follow-up from in the first 10 years to in years If these potential misclassifications had any effect one would expect that they would result in the SMR for liverrelated diseases being smaller and therefore our raised results are unlikely to be explained by these biases. Misclassification in the birth origin may also have occurred in the 22% of subjects where no place of birth was originally obtained and where birth origin was estimated from the surname. Indeed a non-differential misclassification was identified when the death certificates of those dying of HCC were examined. One would expect such bias in this study to have increased the SMR for HCC in HBsAg carriers from England and Wales and have decreased the SMR in overseas carriers. Early in its history HBsAg carriage was suggested as a cause of other malignancies. This study is important in having one of the longest durations of follow-up and therefore the ability to detect such associations. There was an increased SMR in men for only two other malignancies. The increased rate of malignant neoplasm, without specification of site, is likely to be a result of undetected primary HCC. The rise in NHL in males is more interesting. There was a significantly raised male mortality related to AIDS in this study. HIV and HBV have similar routes of transmission and therefore one would expect that the acquisition of HBV might increase the probability of later infection with HIV due to continued risk behaviour. HIV is a well-known cause of NHL and it seems most likely that those recruited into this cohort had an increased frequency of HIV infection leading to the increased rate of NHL amongst males. Examination of the death certificates did not reveal HIV as the cause in any of the cases, although HIV is generally under-reported on death certificates. Longer follow-up has permitted us to examine how the SMR changes with length of follow-up. For both HCC and liver cancer it has fallen. This may be due to the expected seroconversion rate to HBsAg negative of about 1% per annum. However, it could also be explained by treatment effects and avoidance of known co-factors such as alcohol. Unfortunately no information on these is available for the cohort. This study provides further evidence that HBsAg carriage is strongly associated with HCC mortality in men, and liver disease mortality in both men and women in England and Wales. The results of the study provide a measure of the risk associated with chronic HBV carriage in the UK. The results are consistent with declared WHO policy that all countries implement universal HBV immunization. The UK is now one of the few countries in Europe and north America not to do so. Acknowledgements AJ Hall received the Thomas Hunt Memorial Grant of the British Digestive Association in support of this work. We thank the Blood Transfusion Centres and the staff at the National Health Service Central Register. References 1 Van DP, Vellinga A. Epidemiology of hepatitis B and C in Europe. Acta Gastroenterol Belg 1998;61: International Agency for Research on Cancer. Hepatitis Viruses. Lyon: International Agency for Research on Cancer, Beasley RP, Hwang LY. Overview on the epidemiology of hepatocellular carcinoma. In: Hollinger FB (ed.). Viral Hepatitis & Liver Disease. Baltimore: Williams and Wilkins, 1991, pp Beasley RP, Lin CC. Hepatoma risk among HBsAg carriers. Am J Epidemiol 1978;108: Beasley RP, Hwang LY, Lin CC, Chien CS. Hepatocellular carcinoma and hepatitis B virus. A prospective study of men in Taiwan. Lancet 1981;ii: Nomura A, Stemmermann GN, Wasnich RD. Presence of hepatitis B surface antigen before primary hepatocellular carcinoma. JAMA 1982; 247: McMahon BJ, Alberts SR, Wainwright RB, Bulkow L, Lanier AP. Hepatitis B-related sequelae. Prospective study in 1400 hepatitis B surface antigen-positive Alaska native carriers. Arch Intern Med 1990; 150: Fukao A. An epidemiological study on relationship between hepatitis B virus and hepatocellular carcinoma. Nippon Shokakibyo Gakkai Zasshi 1985;82: Tu JT, Gao RN, Zhang DH, Gu BC. Hepatitis B virus and primary liver cancer on Chongming Island, People s Republic of China. Natl Cancer Inst Monogr 1985;69: Ding ZG. Epidemiological study on relationship between hepatitis B and liver cancer a prospective study on development of liver cancer and distribution of HBsAg carriers and liver damage persons in Guangxi. Zhonghua Liu Xing Bing Xue Za Zhi 1988;9: Sakuma K, Saitoh N, Kasai M et al. Relative risks of death due to liver disease among Japanese male adults having various statuses for hepatitis B s and e antigen/antibody in serum: a prospective study. Hepatology 1988;8: Yeh FS, Yu MC, Mo CC, Luo S, Tong MJ, Henderson BE. Hepatitis B virus, aflatoxins, and hepatocellular carcinoma in southern Guangxi, China. Cancer Res 1989;49: Mori M, Hara M, Wada I et al. Prospective study of hepatitis B and C viral infections, cigarette smoking, alcohol consumption, and other factors associated with hepatocellular carcinoma risk in Japan. Am J Epidemiol 2000;151: Prince AM, Alcabes P. The risk of development of hepatocellular carcinoma in hepatitis B virus carriers in New York. A preliminary estimate using death-records matching. Hepatology 1982;2:15S 20S. 15 Dodd RY, Nath N. Increased risk for lethal forms of liver disease among HBsAg-positive blood donors in the United States. J Virol Methods 1987;17: Oshima A, Tsukuma H, Hiyama T, Fujimoto I, Yamano H, Tanaka M. Follow-up study of HBs Ag-positive blood donors with special reference to effect of drinking and smoking on development of liver cancer. Int J Cancer 1984;34: Tokudome S, Ikeda M, Matsushita K, Maeda Y, Yoshinari M. Hepatocellular carcinoma among female Japanese hepatitis B virus carriers. Hepatogastroenterology 1987;34: Hall AJ, Winter PD, Wright R. Mortality of hepatitis B positive blood donors in England and Wales. Lancet 1985;i: Dragosics B, Ferenci P, Hitchman E, Denk H. Long-term follow-up study of asymptomatic HBsAg-positive voluntary blood donors in Austria: a clinical and histologic evaluation of 242 cases. Hepatology 1987;7:

7 124 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY 20 Sherman M, Peltekian KM, Lee C. Screening for hepatocellular carcinoma in chronic carriers of hepatitis B virus: incidence and prevalence of hepatocellular carcinoma in a North American urban population. Hepatology 1995;22: Ross RK et al. Urinary aflatoxin biomarkers and risks of hepatocellular carcinoma. Lancet 1992;339: London WT, Evans AA, Buetow K et al. Molecular and genetic epidemiology of hepatocellular carcinoma: studies in China and Senegal. Princess Takamatsu Symp 1995;25: Office of Population Censuses and Surveys. Comparison of the 8th and 9th International Classification of Diseases. Incorporating short list bridgecoding, 1978 (a sample) for England & Wales. 1978: 24 Coleman M, Douglas A, Hermon C, Peto J. Cohort study analysis with a FORTRAN computer program. Int J Epidemiol 1986;15: Monson RR. Occupational Epidemiology. Boca Raton, FL: CRC Press, 1980, pp Sobeslavsky O. Prevalence of markers of hepatitis B virus infection in various countries: a WHO collaborative study. Bull World Health Organ 1980;58: Sampliner RE. The duration of hepatitis B surface antigenemia. Arch Intern Med 1979;139: