W-L. S. and H-C. C. contributed equally to this work.
Influences of tobacco and alcohol use on hepatocellular carcinoma survival
Article first published online: 28 MAR 2012
Copyright © 2012 UICC
International Journal of Cancer
Volume 131, Issue 11, pages 2612–2621, 1 December 2012
How to Cite
Shih, W.-L., Chang, H.-C., Liaw, Y.-F., Lin, S.-M., Lee, S.-D., Chen, P.-J., Liu, C.-J., Lin, C.-L. and Yu, M.-W. (2012), Influences of tobacco and alcohol use on hepatocellular carcinoma survival. Int. J. Cancer, 131: 2612–2621. doi: 10.1002/ijc.27508
- Issue published online: 25 SEP 2012
- Article first published online: 28 MAR 2012
- Accepted manuscript online: 23 FEB 2012 06:11AM EST
- Manuscript Accepted: 15 FEB 2012
- Manuscript Received: 17 SEP 2011
- National Science Council, Taiwan. Grant Numbers: NSC 97-3112-B-002-028, NSC 98-3112-B-002-021, NSC 100-2325-B-002-025
- alcohol consumption;
- cigarette smoking;
- hepatitis virus infection;
- hepatocellular carcinoma;
Prognosis of hepatocellular carcinoma (HCC) is generally poor. The role of modifiable lifestyle factors on HCC survival has been less studied. To examine whether prediagnosis smoking and alcohol affected HCC survival stratified by viral etiology, we conducted a prospective cohort study of 2,273 (1990 with viral hepatitis and 283 without) incident HCC cases aged 20–75 years who were enrolled between 1997 and 2004 from a Taiwanese multicenter study, and followed up through 2007. Information on habitual smoking and alcohol consumption was obtained at baseline through personal interview. After follow-up to a maximum of 10 years, 1,757 participants died and 1,488 (84.7%) were attributed to HCC. Prediagnosis smoking and alcohol worsened prognosis independent of each other and clinical predictors. The effects of both risky behaviors were limited to viral hepatitis-related HCC and more profound among those with early-stage HCC. Risk for HCC-specific mortality increased with increasing pack-years smoked and ethanol intake (all p < 0.001 for trend), with an additive effect shown for the two habits [hazard ratio (HR) for alcohol ≥46.2 g/day and ≥10 pack-years = 1.72, 95% confidence interval (CI) = 1.45–2.05]. For either habit, cessation reduced HCC-specific mortality, but a significant mortality benefit occurred 10 years after abstinence (quitting smoking ≥10 years vs. continuing smokers: HR = 0.77, 95% CI = 0.61–0.97; quitting drinking ≥10 years vs. continuing drinkers: HR = 0.74, 95% CI = 0.56–0.98). In conclusion, among patients with viral hepatitis-related HCC, prediagnosis smoking and alcohol have a deleterious effect on HCC survival. Quitting smoking or drinking alcohol could reduce the excess risk, but only after a long interval of cessation.
Hepatocellular carcinoma (HCC) is one of the world's leading cancer threats associated with death.1 The prognosis of advanced HCC remains poor. With marked improvement in the early detection and management of HCC during recent decades; however, up to around 60% of early-stage HCC patients may survive without recurrence for 5 or even 7 years.2–4 Established prognostic predictors include the degrees of liver damage, tumor burden and serum α-fetoprotein values.2–4 Although many studies have addressed these and other prognostic factors,2–5 there has been limited exploration of the influence of modifiable lifestyle factors on HCC prognosis.
Both alcohol consumption and cigarette smoking have been associated with increased risk of HCC in many epidemiological studies, although the results are somewhat inconsistent.6–12 Heavy drinking is a major cause of liver cirrhosis,13, 14 which underlies HCC and contributes to poor prognosis of HCC.2–4, 11 Tobacco-related carcinogens can initiate liver tumor formation and inactivate tumor suppressor genes via promoter hypermethylation or mutation in animal models,15–19 yet the information on the role of prediagnostic smoking in HCC survival is limited.20 Whether smoking cessation can improve survival rates remains unclear.
Using a multicenter cohort study after a maximum of 10 years of follow-up, we report a comprehensive evaluation of the association with survival rates in early- and late-stage HCC patients for prediagnostic tobacco and alcohol use, with a particular focus on the differences in association according to viral etiology and the effects of smoking cessation and alcohol abstinence.
Material and Methods
This is a prospective cohort study involving consecutive patients diagnosed with HCC seen at National Taiwan University Hospital, Chang Gung Memorial Hospital and Taipei Veterans General Hospital, three medical centers in Taiwan, between September 1997 and December 2004. The HCC diagnostic criteria were based on the findings of typical radiological features in at least two imaging modalities including ultrasound, computed tomography, magnetic resonance imaging and hepatic arterial angiography or histologically confirmed by needle biopsy or by a single positive image examination plus serum α-fetoprotein level ≥200 ng/mL. After the patients were informed about the study and had agreed to sign an informed consent form for participation, in-person interview using a structured questionnaire was carried out by trained interviewers. The questionnaire covered sociodemographic characteristics, personal and family disease histories (including cancer and hepatic disease), reproductive factors (if women) and selected lifestyle factors. We identified common comorbidities including diabetes and hypertension through patient self-report interview since 1999, leading to missing values on such comorbidities in 22.2% of the study subjects. Inpatient medical charts were reviewed to obtain information on hepatitis B and C virus markers, tumor size, number of lesions and serum α-fetoprotein levels. On average, <5% of the hospital patients approached for interview refused to participate.
Eligibility criteria for our study included age between 20 and 75 years at enrollment and no prior history of HCC or other cancers. A total of 2,962 HCC patients agreed to participate and completed questionnaires. Of these, 2,273 who met the eligibility criteria were included in the study. Reasons for exclusion were hospital admission due to HCC recurrence (n = 505; 73.3%), age <20 or >75 years (n = 47; 6.8%), prior history of other cancers (n = 21, 3.0%) and death within 1 month of hospital admission (n = 116; 16.8%). Participants' deaths occurring during follow-up until December 31, 2007 were ascertained through data linkage with the national death registry database. Classification of deaths was based on the primary cause according to the 9th version of the International Classification of Disease. All patients signed an informed consent, and our study was approved by the research ethics committee at the College of Public Health, National Taiwan University.
Assessment of tobacco and alcohol use
Prediagnosis tobacco and alcohol use were assessed at cohort entry using a standardized questionnaire. The main items of information collected included age at initiation, average quantity per day, frequency of consumption and age at cessation. Participants were classified as never smokers (drinkers), ex-smokers (drinkers) and current smokers (drinkers). Ex-smokers (drinkers) were those who were smokers (drinkers) and had stopped smoking (drinking) >1 years before hospital admission. Pack-years of smoking were calculated as number of cigarettes per day multiplied by number of years of smoking divided by 20. For alcohol consumption, each participant was also inquired about typical consumption of alcohol beverage. Daily alcohol intake among participants who reported consuming ≥1 drink/week for ≥1 year was calculated based on the frequency of consumption, the alcohol content of the beverage and the average quantity consumed.
The end point was death from HCC. The survival time for each patient was calculated from the date of enrollment to the date of death or December 31, 2007 (the date of the most recent linkage to the death registry). We censored all patients whose cause of death was not attributed to HCC. Cumulative survival rates were calculated by the Kaplan-Meier method, and survival curves were compared by the log-rank test. Multivariate Cox proportional hazards regression was used to estimate hazard ratios (HR) and 95% confidence intervals (CI). Each model included the following potential confounding variables as covariates: age, sex, known prognosis factors, history of liver cirrhosis, status of HBsAg and anti-HCV and habitual smoking or alcohol consumption as appropriate. Among patients who ever consumed alcohol, daily ethanol intake and cumulative exposure to ethanol were categorized into quartiles according to the distribution of the continuous variables. Categorical trends were tested by modeling the median value of each category on an ordinal scale. Analyses were also performed after stratifying by viral etiology or clinical stage. The null hypothesis of equal effect of smoking (or alcohol) in different strata were evaluated by including cross-product terms of smoking (or alcohol) and stratification factor in multivariable models. Tests for interaction to identify heterogeneity of the smoking (or alcohol) effect were computed based on likelihood ratio tests. To measure statistical additive interaction between smoking and alcohol, we evaluated the Rothman synergy index and calculated its 95% CI in judging whether a deviation from an additive model ought to be considered statistically significant.21 All statistical analyses were done using SAS v9.1.3 (SAS Institute, Cary, NC). A two-tailed p value of 0.05 or less was considered to be statistically significant.
Eighty-eight percent of the patients were positive for hepatitis B surface antigen (HBsAg) and/or antibodies against hepatitis C virus (anti-HCV). After a maximum follow-up of 10.3 years, there were 1,757 deaths. HCC was the most common cause of death and accounted for 84.7% of total deaths. Overall, the median survival and 1-, 3-, 5- and 7-year survival rates were 18 months, 56.9, 35.8, 27.2 and 22.1%, respectively. Older age at diagnosis, male sex, HBsAg positivity alone, low education level, cigarette smoking, alcohol consumption, high α-fetoprotein values, large tumor size, multiple lesions and exceeding Milan staging criteria (i.e., solitary tumor ≤5 cm, or <4 lesions and none >3 cm)2 were inversely related to the survival rates (Table 1).
We collected data before treatment in newly diagnosed patients. This resulted in 60.1% of the study subjects that were missing treatment data. In those with intention-to-treat data, surgical resection and percutaneous ethanol injection that are considered to be curable treatments of choice were intended for use in 11.7% of the patients and transcatheter arterial embolization was arranged in 78.0%. Treatment methods did not vary substantially across groups of smoking or alcohol consumption but were highly correlated with tumor size and number of lesions (all p < 0.0001).
After adjustment for clinical prognostic factors, history of liver cirrhosis, status of HBsAg and anti-HCV and alcohol consumption, smoking remained associated with increased risk of HCC-specific mortality (Table 2). The smoking effect appeared to be more profound in patients with viral hepatitis than in those without, although this difference between groups did not reach statistical significance. Among patients with viral hepatitis, the HR was 1.19 (95% CI = 0.99–1.42) for ex-smokers and 1.25 (95% CI = 1.09–1.44) for current smokers, as compared to never smokers. We also observed a significant trend in the relative risk estimate for HCC-specific mortality with increasing pack-years of smoking (p = 0.0002). The number of cigarettes smoked per day did not differ between continuing smokers and ex-smokers (p = 0.9252 by Student's t test) or vary by durations of abstinence among ex-smokers (p = 0.0891 by 1-way ANOVA). Compared to continuing smokers, a significant 23% reduction in the risk of HCC-specific mortality was observed among ex-smokers after 10 years of cessation (HR = 0.77, 95% CI = 0.61–0.97), when the excess risk due to smoking approached the level of a never smoker's risk (HR = 0.75, 95% CI = 0.66–0.86).
Thirty-four percent of the patients were drinkers, who had a median daily intake of pure ethanol of 46.1 grams. In patients with viral hepatitis, the elevated risk for deaths due to HCC increased with increasing ethanol intake (p = 0.0002 for trend) after adjusting for smoking and clinical prognosis factors. Compared to never drinkers, those who consumed 46.2–106.9 g of ethanol per day had a HR of 1.35 (95% CI = 1.10–1.65) and 107 g or more per day had a HR of 1.36 (95% CI = 1.12–1.66). We further calculated cumulative alcohol intake defined as the cross-product of the amount of daily ethanol intake and the duration of consumption. A significant increase in the risk of HCC-specific mortality was observed for a cumulative alcohol intake of ≥1,031 gram-years in patients with viral hepatitis. The amount of daily ethanol intake did not differ between continuing drinkers and ex-drinkers (p = 0.5604 by Student's t test) or vary by different durations of abstinence among ex-drinkers (p = 0.4385 by 1-way ANOVA). In a multivariable regression model with simultaneous adjustment for age, sex, clinical prognostic factors, history of liver cirrhosis, status of HBsAg and anti-HCV and cigarette smoking, ex-drinkers who quit drinking ≥10 years before the hospital admission compared to continuing drinkers showed a decreased risk of death from HCC (HR = 0.74, 95% CI = 0.56–0.98) (Table 3).
Additional adjustment for comorbidities such as diabetes and hypertension did not alter the results [e.g., in patients with viral hepatitis: HRs for ex-smokers and current smokers were 1.14 (95% CI = 0.92–1.40) and 1.26 (95% CI = 1.07–1.48), respectively; HRs for ex-drinkers and current drinkers were 1.08 (95% CI = 0.88–1.33) and 1.29 (95% CI = 1.09–1.51), respectively] but led to the exclusion of 22.2% of study subjects from analysis due to missing values. Educational level was strongly associated with tobacco and alcohol use in patients with and without viral hepatitis. We found that, after additional adjustment for education, the associations with HCC mortality for smoking- or alcohol-related variables in patients with viral hepatitis were slightly attenuated but remained significant (data not shown).
We further evaluated the additive interactive effect of both risky behaviors by cross-classifying patients with viral hepatitis by the amount of daily ethanol intake and pack-years of smoking (Table 4). Compared to patients who had daily ethanol intake <46.2 g and <10 pack-years of smoking, the HRs were 1.28 (95% CI = 0.94–1.73) for those who were heavy drinkers (defined as ≥46.2 g/day) alone and 1.28 (95% CI = 1.12–1.46) for those who were heavy smokers (defined as ≥10 pack-years) alone. This analysis suggested an additive interaction of a combined effect of smoking and alcohol (for patients who were both heavy drinkers and heavy smokers: HR = 1.72, 95% CI = 1.45–2.05). Nevertheless, the synergy index was not significantly different from 1.
Early-stage HCC meeting Milan staging criteria (solitary tumor ≤5 cm, or <4 lesions and none >3 cm)2, 22 was associated with better prognosis (Table 1). We next assessed whether the impact of smoking or alcohol differed by clinical stage. There was a significant effect of heavy smoking, irrespective of the stage. The effect of heavy drinking was only observed in early-stage patients, yet no statistically significant modification effect by stage was found (p = 0.1905) (Table 4).
We have conducted an in-depth evaluation of the impact of prediagnosis smoking and alcohol on HCC survival, and found that both risky behaviors worsened prognosis independent of each other and traditional clinical prognostic indicators, suggesting an additive effect of the two factors on enhancement of HCC-specific mortality. Globally, >75% of HCC cases occur in association with hepatitis B or C virus infection.1 In our study, the increased risk of mortality by gradient of exposure to cigarette smoke and alcohol seemed to be more pronounced in patients with viral hepatitis. While the relationship between smoking and an increased risk of HCC-specific mortality was evident for both patients with early- and late-stage cancers, the effect of alcohol consumption was only seen among early-stage HCC patients. In addition, smoking cessation or alcohol abstinence was associated with decreased risk of mortality. However, the beneficial effect appeared to occur after 10 or more years of cessation.
Socioeconomic difference may be associated with tobacco and alcohol use, and association between socioeconomic status and cancer survival has been observed in other malignancies.23 We found that the association for HCC-specific mortality and both risky behaviors persisted after additional adjustment for educational levels. Disparities in cancer outcome due to differences in socioeconomic status or education can be attributed to many factors, including access to medical care, differences in the quality of care after diagnosis and treatment options. Taiwan launched a national health insurance program in 1995, and provides health cares to all Taiwan citizens. This program covers inpatient care, outpatient care, laboratory tests, pharmaceuticals and expensive medical services for life threatening illness such as cancer. Hence, it can be hypothesized that socioeconomic status or education could not explain most of the associations observed for prediagnosis smoking and alcohol and HCC survival in our study.
Strengths of our study include the large number of HCC patients enrolled with known hepatitis B and C serostatus, which permitted assessment on the impact of prediagnosis smoking and alcohol consumption by etiology. The patients were recruited from three major medical centers in Taiwan, which allowed the inclusion of a wider range of patient groups and increased the generalizability of the study. Because of the extensive data on intensity and duration of tobacco and alcohol use, we could comprehensively examine the association with mortality for a relatively well-defined dose-dependent effect of tobacco and alcohol use, and the influence of cessation of such risky behaviors that are potentially clinically usable. Upto 10.3 years of follow-up offered the opportunity to study long-term effects.
The weakness of our study is the small sample size of certain categories in the subgroup of patients without viral hepatitis, which may affect the analysis in such patients by a lack of power for a small effect. In addition, smoking and alcohol consumption were generally associated with comorbidities and about one fifth of subjects in our study lack data on comorbidities such as diabetes mellitus and hypertension, which may influence the decision of whether or not to undergo surgery. Although we used complete-case analysis with additional adjustment for such comorbidities and found similar results, it should be noted that the results could be biased when the data are not missing completely at random.24
Chronic heavy alcohol drinking, defined as ≥60–80 g/day for prolonged period, is a cause of liver cirrhosis and an important risk factor for progression through cirrhosis to HCC.7, 9, 11–14 However, there is controversy about the magnitude of HCC risk associated with low to moderate alcohol intake.7–9, 11–13 In our study, at ethanol intakes of 46 g/day or more, the data are strongly supportive of an association between alcohol consumption and poor survival in HCC patients. The relationship between cigarette smoking and the development of HCC has been examined in many studies in both low- and high-incidence HCC areas and the results have been mixed,6–11 but evidence for a role of smoking in the etiology of HCC has been noted in two large prospective cohort studies conducted in areas where HBV infection is common.6, 7
Consistent with our study, a previous cohort study based on data collected as part of an insurance plan with a mean follow-up of 3 years has also indicated a deleterious effect of both smoking and alcohol consumption on survival in HCC. However, that study did not take hepatitis virus infection and clinical predictors into consideration.20 No studies have so far addressed the effect of discontinuation of smoking or alcohol intake. Our study is the first attempt to assess the long-term effects of smoking and alcohol cessation in regard to HCC, although smoking cessation has been repeatedly found to have beneficial effects on subsequent lung cancer mortality in observational cohort studies and randomized clinical trials with follow-up ranging from 8 to 31 years.25–28 We observed a significant 23% decrease in the risk of HCC-specific mortality among ex-smokers after >10 years of abstinence, whereas for lung cancer mortality, the significant reduction in the excess risk was observed within the first 5 years.25, 27
Our study adds to the growing evidence on the association between tobacco and alcohol use and mortality.25, 27–30 In contrast to consumption of alcohol beverage, which can lead to alcoholic liver disease,13 hepatologists have traditionally paid scant attention to the adverse effects of smoking on liver-related morbidity or mortality. Cigarette smoke contains a wide variety of toxic substances such as nicotine-derived nitrosamine ketone (NNK), which has been shown to produce an increased incidence of HCC in animal models.15, 31 NNK can induce genetic and epigenetic alterations responsible for initiation and progression of cancer in vitro and in vivo.15–19 The exposure to cigarette smoke has been associated with increased oxidative stress and production of proinflammatory cytokines that are key players in carcinogenesis32 and can result in chronic liver disease progression.33, 34 In view of our results combined with mounting evidence from experimental models, suggesting that exposure to cigarette smoke may accelerate the progression of HCC, there is a need to consider smoking as a prognostic factor in medical care and clinical trials of HCC patients.
Prognosis of HCC is generally poor.2–4 In our study, only 27.2% of patients overall survived for 5 years. However, around 35% of the patients were found at an early-stage HCC. These patients present early enough to be treated with curative intent and had a relatively good prognosis with a 5-year survival of 40.9%. We assessed the effect of smoking and alcohol consumption on HCC-specific mortality by clinical stage and found a higher risk in relation to both risky behaviors for early-stage than late-stage patients. This means that prevention against smoking and alcohol use might have a significant impact on the mortality burden of HCC. Although the excess risk of HCC-specific mortality due to either smoking or alcohol use could be eliminated by cessation, the risk remained elevated even in those who ceased smoking or alcohol for 9 years before diagnosis, implicating the importance of quitting such risky behaviors at an early age. Thus, smoking and alcohol-cessation intervention should be incorporated into screening project aimed to identify hepatitis virus carriers for regular follow-up, and should be offered for hepatitis patients undergoing antiviral therapy.
In conclusion, our study supports that prediagnosis smoking and alcohol are inversely associated with HCC survival in hepatitis B or C virus carriers, but has limited statistical power to conclusively determine an association for HCC-specific mortality with these two factors in HCC patients without viral hepatitis. In addition, a substantial reduction in risk for mortality in patients with viral hepatitis was found 10 or more years after quitting smoking or alcohol use before the cancer diagnosis. Hepatologists and primary-care physicians should be aware of the elevated risk in relation to smoking and alcohol use and should help hepatitis patients change risk behaviors, especially for smokers. Future studies are warranted to investigate the effectiveness of HCC screening program in conjunction with smoking-prevention and cessation intervention in the reduction of the mortality burden of HCC in hepatitis virus carriers.