In this large prospective study, current regular alcohol drinking was associated with significantly increased risks of liver cancer and several non-neoplastic chronic liver diseases in Chinese men, with significant dose-response relationships between alcohol intake and risks of liver cancer, liver cirrhosis, ALD, and NAFLD. The associations were consistent across subgroups defined by other risk factors (e.g. smoking, BMI, HBV infection, diabetes), although there was a suggestion of stronger associations of alcohol intake with ALD among those who had higher BMI, or who had the alcohol flushing response. For a given level of weekly intake, drinking daily and having a longer duration of regular drinking were associated with excess risks of ALD, while drinking without meals was associated with excess risks of liver cancer, liver cirrhosis, and ALD. Among the few female current regular drinkers, the risk of ALD was significantly elevated compared with abstainers, but the associations with other disease types were less clear.
A J-shaped association of alcohol consumption with chronic liver disease has been observed in previous prospective studies of Western populations [28, 29], with similar findings also shown in the present study of Chinese adults when the analyses included all alcohol drinking categories. These associations are likely to be affected by reverse causality and confounding, as pre-existing poor health or social disadvantages may lead to alcohol cessation or abstinence and, indeed, in our study abstainers and ex-regular drinkers tended to be older and had poorer baseline health. Furthermore, removing early follow-up years and participants with poor baseline health attenuated the apparently “protective” effects of occasional drinking compared with not drinking but increased the risks associated with regular drinking. To help reliably assess the dose-response relationships between the amount of alcohol intake and liver diseases, we focused the main analyses among male current regular drinkers and showed significant dose-response associations between alcohol consumption and total liver disease incidence as well as mortality in Chinese men (Additional file 1: Tables S19-S20), consistent with previous studies in high-income populations [28, 30].
Although alcohol consumption has been concluded to be causally related to liver cancer by WHO-IARC , evidence from China, where liver cancer has been historically primarily related to chronic HBV infection , is limited. A meta-analysis of 18 case-control studies from China reported a significant odds ratio of 1.56 for hepatocellular carcinoma in drinkers versus non-drinkers (3812 cases) , but no clear dose-response relationships were reported in prospective studies, which were conducted almost three decades ago when alcohol consumption was lower in China [13, 14, 16]. In our study, which was conducted in more recent decades, there were clear dose-response relationships between alcohol intake and liver cancer risk in Chinese men, largely independent of HBV infection status.
Alcohol intake has been consistently shown to be associated with increased risks of liver cirrhosis and ALD in Western and high-income Asian populations [4,5,6,7, 30]. In China, previous prospective studies have also reported excess liver cirrhosis mortality in male heavy drinkers [13, 15]. With linkage to the national health insurance system which covered almost all hospitalised events, our study now provides reliable prospective evidence of significant dose-response relationships between alcohol intake and incident liver cirrhosis and ALD in Chinese men.
In populations with high prevalence of obesity, fatty liver disease particularly NAFLD has emerged as a major chronic liver disease. Two meta-analyses of 18 and 16 individual studies, respectively, suggested that moderate drinking (i.e. < 20 g/day) was associated with lower risk of fatty liver disease (diagnosed and defined in various ways across different studies) compared with non-drinking [33, 34]. However, these findings were based mostly on cross-sectional studies and were inconsistent across different populations. Conversely, the longitudinal Whitehall II Study (n = 5407, 30-year follow-up) reported a twofold higher risk of fatty liver disease (defined as having a fatty liver index ≥ 60) associated with sustained heavy drinking versus stable low-risk drinking (> 168 vs. 8–168 g/week for men, > 112 vs. 8–112 g/week for women), with no reduced risks for stable low-risk drinking compared with stable non-drinking . With a much larger size of study population than previous studies, we provide new prospective evidence for a positive dose-response relationship between alcohol intake and NAFLD in Chinese male drinkers. Unlike some of the previous studies, the diagnosis of NAFLD in the present study was based on hospital admission rather than ultrasound test and there were strong positive associations between NAFLD and adiposity. Moreover, the diagnosis of NAFLD did not overlap with ALD (with less than ten having a diagnosis of both diseases during follow-up), suggesting the finding was unlikely to be due to misclassification of ALD.
Several possible mechanisms linking alcohol consumption with different types of liver diseases have been proposed . Alcohol use, particularly heavy consumption, can stimulate changes in lipid metabolism resulting in fatty liver. Moreover, heavy alcohol consumption may induce an inflammatory response resulting in alcoholic hepatitis (or steatohepatitis if accompanied by hepatic lipid deposition) which, if persistent and severe, can eventually lead to fibrosis and sclerotic changes in the liver resulting in liver cirrhosis, and ultimately liver failure or liver cancer. Previous studies showed that heavy alcohol consumption may interact with chronic HCV infection, possibly via increased viral replication and altered immune response, to aggravate liver disease progression and increase mortality risk [10, 11, 37]. Similar pathways have been assumed for HBV infection , but there was no clear evidence of interaction between alcohol consumption and HBV infection on risks of liver diseases in our study.
It has been suggested that smoking , obesity [38, 39], and diabetes  may synergistically interact with alcohol consumption to increase the risks of advanced liver diseases, but the existing epidemiological evidence is inconsistent [28,29,30, 38,39,40]. In our study, although the associations of alcohol consumption with most major chronic liver diseases appeared consistent between subgroups defined by these risk factors, stronger associations of alcohol intake with ALD were observed in never smokers, men with higher BMI, or men reporting flushing, compared with their counterparts. However, due to the limited number of ALD events, these findings should be interpreted with caution. It is possible that obesity and alcohol consumption may synergistically worsen hepatic insulin resistance and cause necro-inflammation, leading to progressive liver injury . In animal studies, aldehyde dehydrogenase 2 (ALDH2) deficiency has been shown to aggravate liver inflammation and fibrosis through increased exposure to acetaldehyde and its associated adducts after alcohol consumption ; although epidemiological studies reported a reduced alcoholic cirrhosis risk in individuals with ALDH2 deficiency , this is likely due to the substantially lower alcohol consumption in individuals with ALDH2 deficiency which inhibits heavy drinking.
This is the first prospective study to report on the associations between drinking patterns and risks of many different types of chronic liver diseases in a Chinese population. The UK Million Women Study (MWS) and the Danish Diet, Cancer and Health Study have shown that daily drinking was associated with increased risks of liver cirrhosis and ALD compared with non-daily drinking (HR range 1.61–3.65), given the amount consumed [6, 7]. Two small studies in Finland and the US (< 130 cases each) have also reported positive associations between HED frequency and risks of liver diseases (HR range 2.78–4.76 for ≥ weekly vs. never/rare engagement in HED), but the findings were prone to residual confounding by the total volume and frequency of alcohol consumption [44, 45]. Our findings seem to support the adverse effects of daily drinking on ALD and total liver disease over and above total alcohol intake. There may be increased hepatic damage with consistent exposure to alcohol, compared with taking breaks from drinking (i.e. “liver holidays”) which allows recovery of the normal metabolic function of the liver . Nevertheless, the findings should not be taken to dispute the potential harmful effects of HED, as HED results in a high amount consumed and has both acute (e.g. injuries, acute alcohol poisoning) and long-term consequences (e.g. hypertension, alcohol dependence), and chronic persistent HED can exacerbate liver injury . Our findings in men were generally consistent with those reported in the MWS, which showed a 31% lower risk of liver cirrhosis and ALD (1560 cases) among women who usually drank alcohol with meals than those who did not . In the absence of food, there may be faster absorption of alcohol from the intestine, leading to higher blood alcohol concentrations and greater detrimental effects of drinking . Our observed associations between duration of regular drinking and risks of ALD and total liver disease also concur with previous studies [49, 50].
The chief strengths of this study include the prospective study design with a large sample size, comprehensive adjustments for potential confounders, and large numbers of well-characterised liver disease events traced via comprehensive follow-up systems. The exclusions of individuals with prior diseases and early follow-up reduced reverse causality. Also, the repeat alcohol measures allowed adjustment for regression dilution bias . Our study also has several limitations. First, alcohol exposure data was self-reported, but they were positively correlated with blood pressure and gamma-glutamyl transferase as expected and consistent with the causal associations seen with genotype-predicted alcohol intake , suggesting good data quality. It is still possible, however, that heavy drinking may be under-reported, which could potentially lead to underestimation of the HED-associated disease risk. Second, it is possible that the observed association with NAFLD and other liver diseases may partly be due to some ALD cases being misclassified as NAFLD (e.g. the patient’s alcohol intake was under-reported or not stated in clinical notes); however, the strong positive associations between NAFLD and adiposity and lack of overlap in different disease diagnoses during follow-up suggested that our findings were unlikely to be due to misclassification of ALD. Third, although careful adjustments were made, residual confounding by measured or unmeasured factors might remain. For example, measurement error in baseline alcohol intake may lead to inadequate adjustment for total alcohol intake in the associations of drinking patterns with disease risks. Where possible, subgroup analyses were performed to further minimise residual confounding. Fourth, although the associations between alcohol consumption and liver diseases differed little by HBsAg sero-status, the on-site HBsAg rapid test used in CKB has relatively low sensitivity for lower serum HBsAg levels [51, 52]. Fifth, the lack of information on HCV infection precluded adjustment for HCV infection in our analyses and investigation of potential interactions between alcohol consumption and HCV infection on disease risks; however, given the low prevalence of HCV infection in China (< 1%)  it should not have materially affected our main findings. Sixth, it is possible that flushing response may diminish in intensity after a long or heavy drinking history, potentially leading to some misclassifications which might underestimate any heterogeneity in alcohol-liver disease risk relations by ALDH2 deficiency. Nevertheless, overall the flushing response was strongly associated with ALDH2-rs671 genotype among current regular drinkers in CKB . Finally, given the low prevalence of grape wine drinkers in China [9, 18], we were unable to investigate the separate associations for grape wine, which has been reported to have less detrimental effects on alcoholic liver cirrhosis than other beverage types . Likewise, we lacked sufficient statistical power for analyses in women and on ALD across subgroups.