Cigarette Smoking and Liver Cancer Risk: An Evaluation Based on a Systematic Review of Epidemiologic Evidence among Japanese

Background: Emerging epidemiologic data suggest that cigarette smoking may increase the risk of primary liver cancer. We evaluated this association based on a systematic review of epidemiologic evidence among Japanese populations. Methods: Original data were obtained from MEDLINE searches using PubMed, complemented with manual searches. The evaluation was performed in terms of the magnitude of association (‘strong’, ‘moderate’, ‘weak’ or ‘no association’) in each study and the strength of evidence (‘convincing’, ‘probable’, ‘possible’ or ‘insufficient’), together with biological plausibility as previously done by the International Agency for Research on Cancer. Results: A total of 12 cohort studies and 11 case–control studies were identified. Nine cohort studies (two with adjustment for hepatitis B and C virus infections and seven without it) reported weak to strong positive associations between smoking and liver cancer, with dose–response relationships shown in three studies. Five case–controls studies (three with the virus adjustment and two without it) demonstrated such positive associations, with a dose–response relationship shown in only one study, while in six case–control studies, the observed associations were judged to be of the lowest magnitude or inverse due to the lack of any dose–response relationship. Conclusion: We conclude that cigarette smoking ‘probably’ increases the risk of primary liver cancer among the Japanese.Potentialconfoundingbyhepatitis virusinfectionandvirus–smoking interactions need to be addressed in future studies.


INTRODUCTION
Primary liver cancer is one of the most common cancers in Japan (1). Its primary prevention remains to be a major concern for both clinicians and epidemiologists, since patients with this tumor still present poor prognosis (1,2). More than 90% of primary liver cancers in Japan are known to be hepatocellular carcinomas (2), which are mostly attributable to chronic infection with hepatitis C virus (HCV) and hepatitis B virus (HBV) (2,3). However, emerging evidence suggests that hepatocarcinogenesis is a multistage process, in which environmental factors other than hepatitis viruses may play additional roles (4). One of such candidates is cigarette smoking, which has not yet attracted much attention of clinicians or the public. Recently, the International Agency for Research on Cancer listed liver cancer as a tobacco-related malignancy (5). In this context, the objective of the present study was to review and summarize epidemiological findings on cigarette smoking and liver cancer among Japanese populations. This work was conducted as part of a project of systematic evaluation of the epidemiological evidence regarding lifestyles and cancers in Japan (6).

METHODS
The details of the evaluation method have been described elsewhere (6). In brief, original data for this review were identified by MEDLINE searches using PubMed, complemented by manual searches of references from relevant articles where necessary. All epidemiologic studies on the association between cigarette smoking and liver cancer incidence or mortality among the Japanese from 1963 to 2005, including papers in press if available, were identified using the search terms 'smoking', 'liver', 'hepatocellular', 'cohort', 'followup', 'case-control', 'Japan' and 'Japanese' as keywords. Papers written in either English or Japanese were reviewed, and only studies on Japanese populations living in Japan were included. The individual results were summarized in the tables separately by a study design as cohort or case-control studies.
The evaluation was made based on the magnitude of association and the strength of evidence. First, the former was assessed by classifying relative risk (RR) in each study into the following four categories, while considering statistical significance (SS) or no statistical significance (NS): (i) 'strong' (symbol ### or """) when RR < 0.5 (SS) or RR > 2.0 (SS); (ii) 'moderate' (symbol ## or "") when RR < 0.5 (NS), 0.5 < RR < 0.67 (SS), 1.5 < RR < 2.0 (SS) or RR > 2.0 (NS); (iii) 'weak' (symbol # or ") when 0.5 < RR < 0.67 (NS), 0.67 < RR < 1.5 (SS) or 1.5 < RR < 2.0 (NS) and (iv) 'no association' (symbol À) when 0.67 < RR < 1.5 (NS). When RRs for three or more exposure levels were reported, that for the highest level was employed for this classification. In the case of multiple publications of analyses of the same or overlapping datasets, only data from the largest or most updated results were included. After this process, the strength of evidence was evaluated in a similar manner to that used in the WHO/FAO Expert Consultation Report (7), in which evidence was classified as 'convincing', 'probable', 'possible' and 'insufficient'. We assumed that biological plausibility corresponded to the judgment of the most recent evaluation from the International Agency for Research on Cancer (5). Notwithstanding the use of this quantitative assessment rule, an arbitrary assessment cannot be avoided when considerable variation exists in the magnitude of association between the results of each study. The final judgment, therefore, was made based on a consensus of the research group members, and it was therefore not necessarily objective. When we reach a conclusion that there is 'convincing' or 'probable' evidence of an association, we conduct a meta-analysis to obtain summary estimates for the overall magnitude of association.
A summary of the magnitude of association for the cohort studies and case-control studies is shown in Tables 3 and  4, respectively. Among all 12 cohort studies, five (9,(13)(14)(15)19) reported strong positive associations of cigarette smoking with liver cancer in either sex or for both sexes combined (Tables 1 and 3); of the five studies, three (9,13,15) demonstrated clear dose-response relationships. Moderate, but not strong, positive associations were found in three cohort studies (10,11,18), and a weak association in one cohort study (17), without any presentation of doseresponse relation. In the remaining three (8,12,16), virtually no association was observed. Among the seven cohort studies in which mostly healthy subjects were followed, six (9,11,14,(17)(18)(19) revealed at least weak positive associations, whereas three (10,13,15) out of the five follow-up studies of patients with chronic liver disease showed such positive associations.
Among all 11 case-control studies, five (20,(26)(27)(28)(29) reported weak to strong positive associations with cigarette smoking, with a dose-response relationship presented in only one study (20) (Tables 2 and 4). In the remaining six studies (21)(22)(23)(24)(25)30), the observed associations were judged to be null or inverse due to the lack of dose-response relationship, although around 2-to 4-fold risk excess in light to moderate exposure categories was observed in five of them (21)(22)(23)(24)(25). In the nine case-control studies employing hospital or community controls, three (27)(28)(29) demonstrated at least weak positive associations, whereas both case-control studies using controls of HBV carriers or patients with chronic liver disease (20,26) afforded such positive associations.
In the cohort studies, cigarette smoking was almost consistently associated with elevated liver cancer risk. Information and selection biases may not be serious issues in those studies. However, potential confounding by chronic HBV and HCV

446
Cigarette smoking and liver cancer risk

450
Cigarette smoking and liver cancer risk    infections was not addressed in most studies. Since, in Japan, individuals with either or both infections may have more than 100 times higher risk than those without either (3,31), only a slight change in smoking habit among such infected individuals could result in a substantial distortion of associated RRs. Alcohol consumption, another potential confounder, was not adequately controlled in some studies. In addition, the lack of dose-response relationship in three-quarters of the cohort studies has made our conclusion more conservative.
As for the case-control studies, the data have been controversial. In some studies, the recruitment of hospital controls, which possibly included those with smoking-related diseases, may have biased the RRs towards unity. Confounding issues by hepatitis virus infection and alcohol drinking were the same as those in the cohort studies. The absence of dose-response relation in majority of the case-control studies appears very perplexing. Among cases, symptoms resulting from preexisting liver disease or physicians' advice on their health can lead to lifestyle changes including a reduction in number of cigarettes smoked per day. This might be responsible for elevated risks among light to moderate smokers observed in most case-control studies. However, the situation was similar in the cohort studies where smoking habit many years before the development of liver cancer was evaluated. Some unknown biological implications might exist in these non-linear relations.
An interaction issue between hepatitis viruses and cigarette smoking (i.e. possible difference in risk increase due to smoking according to hepatitis virus infection) should also be considered. Since the great majority of patients with hepatocellular carcinoma in Japan is known to be chronically infected with HBV or HCV (2,3), the following question naturally arises: 'Does smoking increase the risk of hepatocellular carcinoma among people without either HBV or HCV infection?' This question has not fully been addressed, probably due to the difficulty in conducting epidemiologic studies on this subject and its low practical implication in the prevention of liver cancer. It seems biologically implausible that cigarette smoking, without any hepatitis virus infection or heavy alcohol consumption, causes chronic liver disease, thereby playing a major role in hepatocarcinogenesis. On the other hand, the evaluation of the risk for smoking among people infected with HBV or HCV will be easier to be performed and will provide more practical information. It is noteworthy that, based on such evaluations, a limited number of cohort or case-control studies demonstrated clear doseresponse relationships between smoking and liver cancer risk (13,15,20).
Finally, the authors consider that it will be problematic to perform a meta-analysis to obtain a summary estimate for the overall magnitude of association, since such an estimate may not be applicable to general populations of the Japanese due to the above interaction issue. Therefore, the planned meta-analysis was not conducted in this particular evaluation. In addition, the authors cannot exclude the possibility of publication bias and missing relevant epidemiologic studies, although they have long been knowledgeable about the situation of such studies in Japan.

EVALUATION OF THE EVIDENCE ON CIGARETTE SMOKING AND LIVER CANCER RISK AMONG JAPANESE
From these results and based on assumed biological plausibility as previously done by the International Agency for Research on Cancer (5), we conclude that cigarette smoking 'probably' increases the risk of primary liver cancer among the Japanese. Potential confounding by hepatitis virus infection and virus-smoking interactions need to be addressed in future studies.