Background: In China, alcohol consumption has increased significantly in recent decades. Little evidence exists, however, about temporal trends in levels and patterns of alcohol consumption and associated factors in adult populations. Methods: In 2004–08, the China Kadoorie Biobank recruited ~ 512,000 adults (41% men, mean age 52 years [SD 10.7]) from 10 (5 urban, 5 rural) geographically diverse regions across China, with ~ 25,000 randomly selected participants resurveyed in 2013–14. The self-reported prevalence and patterns (e.g., amount, beverage type, heavy drinking episodes) of alcohol drinking at baseline and resurvey were compared and related to socio-demographic, health and other factors. Results: At baseline, 33% of men drank alcohol at least weekly (i.e., current regular), compared to only 2% of women. In men, current regular drinking was more common in urban (38%) than in rural (29%) areas at baseline. Among men, the proportion of current regular drinkers slightly decreased at resurvey (33% baseline vs. 29% resurvey), while the proportion of ex-regular drinkers slightly increased (4% vs. 6%), particularly among older men, with more than half of ex-regular drinkers stopping for health reasons. Among current regular drinkers, the proportion engaging in heavy episodic drinking (i.e., > 60 g/session) increased (30% baseline vs. 35% resurvey) in both rural (29% vs. 33%) and urban (31% vs. 36%) areas, particularly among younger men born in the 1970s (41% vs. 47%). Alcohol intake involved primarily spirits, at both baseline and resurvey. Those engaging in heavy drinking episodes tended to have multiple other health-related risk factors (e.g., regular smoking, low fruit intake, low physical activity and hypertension). Conclusions: Among Chinese men, the proportion of drinkers engaging in harmful drinking behaviours increased in the past decade, particularly among younger men. Harmful drinking patterns tended to cluster with other unhealthy lifestyles and health-related risk factors.
Keywords: Alcohol; China; Patterns; Trends
Ling Yang and Zhengming Chen are Joint Senior Authors.
Globally, alcohol is the seventh leading risk factor for poor health, accounting for 4.2% of total disability-adjusted life years and 5.2% of deaths in 2016 [[
While several cross-sectional studies in China reported on drinking patterns during 1994–2007 [[
Longitudinal studies conducted in Western populations have shown that alcohol drinking varies with age throughout the life-course [[
Using data from the China Kadoorie Biobank (CKB) prospective cohort study, we aim to assess: 1) the temporal trends in the prevalence and patterns of alcohol drinking between the baseline survey in 2004–8 and a resurvey in 2013–14, and 2) the socio-demographic, health and other factors associated with changes in drinking prevalence and patterns over this time period.
The China Kadoorie Biobank (CKB) is a nationwide prospective blood-based cohort study involving 0.5 million participants recruited from 10 geographically defined regions across China, established to investigate genetic and non-genetic causes of many common chronic diseases in the Chinese population. Details of the CKB study design and survey methods have been described previously [[
Detailed information on questionnaire assessment of alcohol consumption in the CKB study has been described previously [[
The amount of pure alcohol consumed in grams per session was calculated according to the beverage type and amount drunk on the last time they drank alcohol (see Additional file 1: Tables S2 and S3 for information on data quality), based on the assumption of the following alcohol content by volume (v/v) typically seen in China [[
To describe changes in drinking status at the individual level between baseline and resurvey, six categories were defined: stable non-drinkers; starters; stable drinkers; stoppers; decreased-intake drinkers; and increased-intake drinkers (detailed definition shown in Additional file 1: Table S1).
To assess the presence of other health-related risk factors in participants, a risk factor index was derived by summing the individual scores of four major risk factors (1 = yes, 0 = no): regular smoking, lack of daily fresh fruit intake, low physical activity, and hypertension (see Additional file 1: Table S1 for detailed classifications).
As alcohol drinking prevalence and behaviour differed significantly between men and women in CKB, all analyses were conducted separately in men and women. Temporal trends in alcohol drinking were assessed using cross-sectional data from all participants recruited at baseline (n = 512,891) and from participants attending the resurvey (n = 24,996). The crude prevalence of alcohol drinking and mean consumption were calculated at each survey. Among subgroups defined by socio-demographic factors and the risk factor index, the prevalence of alcohol drinking was directly standardized to the age and region structure of the study population. As drinking patterns were assessed in weekly drinkers, the relevant variables were standardized according to the age and region structure of weekly drinkers. Due to the small number of female weekly drinkers (n
Longitudinal analyses of changes in drinking status were conducted among men who attended both surveys (n = 9569). A similar standardization approach was applied as above to obtain the prevalence of each status change category standardised to the age and region structure of the longitudinal subset. The statistical association between factors and change in drinking status was tested using logistic regression among baseline non-drinkers and multinomial logistic regression among baseline drinkers. The reasons for stopping weekly drinking were assessed among previous weekly drinkers (i.e. ex-weekly and reduced-intake drinkers) in the resurvey (n = 1166). Crude percentages of each of the reasons reported were calculated overall, and by socio-demographics and health factors. Among men who were current weekly drinkers at both baseline and resurvey, a dependent sample t-test was used to test for the statistical significance of change in alcohol consumption at individual level over time. For a quality check of the self-reported alcohol data, associations of blood pressure with alcohol consumption were assessed using linear regression adjusted for age, study region, education, income, smoking category, physical activity and month of recruitment. Data from the CKB Data Release 10 was used in this study. All analyses were performed in SAS version 9.4.
Of the 512,891 participants recruited at baseline, the mean age was 52 years (SD 10.7), 41% were men and 56% were from rural areas. The subset of participants resurveyed (n = 24,996) was broadly representative of the baseline study population in terms of the distribution of baseline characteristics of sex, birth cohorts, area, education, and health and lifestyle factors (Additional file 1: Table S4). Though the response rates at resurvey were highly consistent across baseline drinking groups, there was slight variation across birth cohorts from 68% in the oldest and youngest cohorts to 79% among the middle cohorts (Additional file 1: Table S5). Between baseline and resurvey, there was an increase in household income in the study population (Table 1).
Cross-sectional characteristics of participants at baseline (2004–8) and resurvey (2013–14)
Characteristics* Men Women Baseline ( Resurvey ( Baseline ( Resurvey ( Socio-demographic characteristics Mean age, years 52.4 59.7 51.0 58.5 Birth cohorts, % < 1940 13.8 11.6 10.3 8.6 1940–1949 22.4 24.2 20.6 22.2 1950–1959 30.9 32.9 32.2 34.4 1960–1969 27.9 27.5 31.3 30.5 ≥ 1970 4.9 3.9 5.7 4.3 Area, % Rural 56.6 56.8 55.4 57.1 Highest education, % No formal education 8.9 9.2 25.3 27.1 Primary school 33.4 33.4 31.4 31.4 Middle or high school 49.9 49.2 38.8 37.4 Technical school/college or above 7.9 8.2 4.5 4.2 Household income (yuan/year), % < 10,000 26.0 8.2 29.8 9.0 10,000-19,999 28.3 11.2 29.6 13.0 20,000-34,999 25.4 17.9 24.2 20.2 35,000+ 20.2 62.7 16.5 57.8 Health and lifestyle factors Regular smoking, % 61.1 51.0 2.4 1.6 Daily fruit intake, % 23.0 39.0 31.8 48.8 Physical activity, mean MET hours/day 22.0 21.1 20.4 18.4 Self-reported good healtha, % 49.2 47.0 43.3 41.8 Prior diseaseb, % 22.6 31.7 22.1 33.9 Satisfied with lifec, % 69.7 82.4 67.7 80.5 Physical measurements Mean SBP, mmHg 132.8 136.9 129.9 136.3 Mean DBP, mmHg 79.2 79.6 76.8 77.5 Mean heart rate, beats/minute 77.7 76.6 79.7 78.2 Mean BMI, kg/m2 23.4 24.0 23.8 24.3 Mean WHR 0.9 0.9 0.9 0.9 Mean standing height, cm 165.2 164.7 154.1 153.5
BMI body mass index, S/DBP systolic/diastolic blood pressure, MET metabolic equivalent task, WHR waist-hip ratio *Characteristics were based on the characteristics of the participants collected at each of the baseline and the resurvey
At baseline, 33% of men but only 2% of women reported drinking alcohol at least weekly (i.e., current regular drinkers), whereas a third of men and women drank occasionally (Table 2). Among male weekly drinkers, mean alcohol intake was 50 g/session (based on the last drinking day), with 62% drinking daily or almost daily (6-7 days/week). The majority of male weekly drinkers engaged in heavy episodic drinking on special occasions (84%), with 30% doing so on the last drinking day, and 24% reported at least one indicator of problem drinking (e.g., drinking in the morning, unable to work due to drinking etc). Female weekly drinkers had lower mean consumption and were less likely to report heavy episodic drinking or problems related to their drinking. Increased alcohol consumption was associated with higher blood pressure at each survey (Additional file 1: Table S3). In both sexes, strong spirits were the most common beverage type consumed. Most weekly drinkers usually drank with meals (86% overall, in men and women), which was broadly consistent across study areas, except for one rural region where ~ 20% drank with meals (Additional file 1: Table S6).
Prevalence and patterns of alcohol consumption at baseline (2004–2008) and resurvey (2013–2014), by sex
Men Women Baseline Resurvey Baseline Resurvey Overall Number of participants 210,259 9569 302,632 15,427 Drinking categories, % Abstainer 20.4 34.9 63.6 82.0 Ex-weekly 3.8 6.3 0.4 0.7 Reduced-intake 4.9 3.8 0.4 0.5 Occasional 37.7 26.4 33.5 14.9 Current weekly 33.2 28.6 2.1 1.9 Among current weekly drinkers Number of participants 69,904 2732 6248 292 Types consumed^, % Strong spirit (≥40% alcohol) only 41.5 41.8 47.6 39.4 Weak spirit (< 40% alcohol) only 19.8 16.0 11.4 12.0 Beer only 20.1 13.5 22.6 12.3 Rice wine or grape wine only 11.0 12.3 15.7 27.1 Mixed 7.7 16.3 2.7 9.2 Mean consumption, g/session^ 49.9 55.6 22.7 24.1 Number of drinking days per week, % 1–2 19.9 14.9 33.1 21.9 3–5 18.0 14.0 21.6 15.4 6–7 62.1 71.1 45.2 62.7 Types consumed on special occasions, % Strong spirit (≥40% alcohol) only 35.7 35.8 41.4 33.9 Weak spirit (< 40% alcohol) only 16.3 14.7 9.9 10.6 Beer only 10.9 8.4 16.7 11.0 Rice wine or grape wine only 6.6 9.3 12.3 25.7 Mixed 30.5 31.8 19.6 18.8 Mean consumption on special occasions, g/occasion 147.4 116.3 55.1 41.3 Drinking patterns, % Drinking outside meal. 14.1 17.6 13.8 20.9 Heavy episodic drinking^a 29.8 34.7 18.2 17.8 Heavy episodic drinking on special occasionsa 83.6 72.9 57.1 43.8 Problem drinking indicator(s)b 23.9 23.8 9.8 9.2 Flushing response after drinking 17.9 15.5 23.6 13.7 Mean age started weekly drinking 28.7 29.3 37.7 40.4
Among men, the prevalence of weekly drinking was lower at resurvey than at baseline (33% baseline vs. 29% resurvey), as was occasional drinking (38% vs. 26%) (Table 2). Similar differences were also evident in women for occasional drinking (34% vs. 15%). Among male weekly drinkers, however, there were increases in mean consumption on the last drinking day (50 g/session vs. 56 g/session), and in the prevalence of heavy episodic drinking (30% vs. 35%) and daily drinking (62% vs. 71%) during the same period. Further analyses among men who drank weekly at both baseline and resurvey showed that mean consumption increased by 3.7 g/session (p = 0.004) (Additional file 1: Table S7). While the most common beverage type was strong spirits only (consumed by ~ 42% of male weekly drinkers at both surveys), there was an increase in the proportion of drinkers consuming multiple beverage types (8% vs. 16%), who tended to have a higher intake per session than those drinking a single beverage type (Additional file 1: Table S8). Similar drinking patterns and temporal trends were observed in the subset of participants who were involved in both the baseline survey and resurvey (Additional file 1: Table S9).
Among men, the patterns and trends of alcohol drinking differed by birth cohort (Fig. 1a-1d). At baseline, weekly drinking prevalence was highest in men born in the 1950s–60s (36%) (Fig. 1a). Over the study period, the prevalence of weekly drinking decreased in most birth cohorts, except for the youngest cohort born in the 1970s who had a slightly higher prevalence of weekly drinking at resurvey (27% vs. 32%), which may be due to the effects of age (Additional file 1: Figure S1). Among male weekly drinkers in both surveys, drinking frequency was higher among older birth cohorts (Fig. 1b), but mean consumption per session and the prevalence of heavy episodic drinking were higher among younger birth cohorts and increased over time in these groups (58 g/session vs. 77 g/session and 41% vs. 47% respectively in men born in the 1970s) (Fig. 1c, d).
Graph: Fig. 1 Alcohol drinking characteristics in male weekly drinkers in 2004–8 and 2013–4, by birth cohort. Prevalence and mean were adjusted for regions. Size of boxes is proportional to the sample size of the respective birth cohort. Error bars are 95% confidence intervals. Mean consumption per session (g/session) and heavy episodic drinking was based on alcohol intake data reported on the last time the participants drank. Heavy episodic drinking is defined as drinking > 60 g of pure alcohol in one session for men. All men at baseline (n = 210,259) and resurvey (n = 9569) were included in (a). All male weekly drinkers at baseline (n = 69,904) and resurvey (n = 2732) were included in (b-d)
Drinking patterns also differed by study area and socio-economic status in men at both surveys. The prevalence of weekly drinking was higher in urban areas, while the proportion of daily drinkers and men reporting problem drinking indicators among those who drank at least weekly was higher in rural areas and lower socio-economic groups (Additional file 1: Tables S10 and S11). Over time, a more marked increase in mean alcohol consumption was observed among urban drinkers. The prevalence of men who reported at least one problem drinking indicator increased in urban areas and particularly in younger urban drinkers, but generally declined in rural areas (Additional file 1: Table S12). Furthermore, younger generations also had a higher tendency to drink beer or multiple beverage types, with a notable rise in the proportion of mixed-beverage drinkers between baseline and resurvey (10% vs. 25% in those born in the 1970s) (Additional file 1: Figure S2).
Having multiple health-related risk factors (i.e. regular smoking, low physical activity, low fresh fruit intake, hypertension) was correlated with alcohol drinking, with weekly drinking prevalence, drinking frequency, mean consumption and heavy episodic drinking prevalence all increasing with the number of risk factors in men at both surveys (Fig. 2a-d). When hypertension was removed from the risk factor index, the correlation persisted between unhealthy lifestyles and alcohol drinking (Additional file 1: Figure S3).
Graph: Fig. 2 Alcohol drinking characteristics in male weekly drinkers in 2004–8 and 2013–4, by health-related risk factor index. Prevalence and mean were adjusted for age and regions. Size of boxes is proportional to the sample size of the respective risk factor index group. Error bars are 95% confidence intervals. Mean consumption per session (g/session) and heavy episodic drinking was based on alcohol intake data reported on the last time the participants drank. Heavy episodic drinking is defined as drinking > 60 g of pure alcohol in one session for men. Risk factor index was derived by summing the individual scores of each of the four risk factors (0 if no, 1 if yes): regular smoking, lack of daily fruit intake, hypertension, low physical activity. All men at baseline (n = 210,259) and resurvey (n = 9569) were included in (a). All male weekly drinkers at baseline (n = 69,904) and resurvey (n = 2732) were included in (b-d)
Among 9569 men who attended both surveys, during 2004–8 and 2013–14, 40% of them continually drank and 18% did not drank alcohol on both occasions, 23% stopped drinking and 4% started drinking since baseline (Table 3). Men who were older, living in rural areas or with poor self-reported health status at resurvey were more likely to remain non-drinking as well as to have stopped drinking. The results were consistent though less apparent when analyses used baseline rather than resurvey characteristics (Additional file 1: Table S13), suggesting that changes in health status and ageing over the study period was correlated with stopping drinking. Similarly, among the 1166 ex-weekly or reduced-intake drinkers in the resurvey, over half of them reported existing physical illness as their main reason for stopping weekly drinking, particularly in those who were older or had prior diseases (Table 4). Other reported reasons for stopping drinking were money (24%), future health concerns (6%), and other reasons not related to financial or health concerns (19%).
Changes in drinking status by socio-demographic characteristics and health factors among men from baseline to resurvey
Characteristics* N Non-drinkers (abstainers, ex-weekly drinkers) at baseline^ Drinkers (reduced-intake, occasional and weekly drinkers) at baseline# Stable non-drinker % Starter % Stable drinker % Stopper % Decreased-intake drinker % Increased-intake drinker % All men 9569 18.3 4.2 40.0 22.9 8.2 6.3 Socio-demographic characteristics Birth cohorts < 1940 1106 33.0 4.3 24.0 30.6 3.9 4.3 1940–1949 2311 23.1 4.3 32.4 28.3 6.9 5.2 1950–1959 3149 16.3 3.8 43.7 22.1 8.2 6.0 1960–1969 2632 11.3 4.6 47.3 18.1 10.7 8.0 ≥ 1970 371 8.3 3.7 50.3 16.0 11.7 10.0 Area Rural 5433 21.9 4.3 36.3 24.5 6.9 6.1 Urban 4136 13.6 4.1 44.8 20.9 10.0 6.6 Highest education Primary or below 4076 20.1 4.2 36.9 24.1 8.6 6.1 Middle or above 5493 16.7 3.9 42.1 22.5 8.4 6.4 Household income (yuan/year) < 35,000 3573 21.5 3.9 37.0 23.9 6.8 7.0 35,000+ 5996 16.8 4.4 40.8 22.9 8.9 6.2 Health factors Self-reported health statusa Good 4498 16.7 4.4 43.2 20.0 8.4 7.3 Poor 5071 19.9 4.0 37.0 25.6 8.1 5.4 Prior diseaseb No 6533 17.0 3.9 42.6 21.3 8.5 6.7 Yes 3036 21.4 5.1 34.3 26.4 7.6 5.1 Risk factor index scorec 0 469 18.1 5.4 36.7 25.5 7.9 6.3 1 1962 18.3 4.6 36.8 24.9 9.6 5.8 2 3559 18.4 4.7 40.1 23.0 7.4 6.4 3+ 3579 17.4 3.5 43.1 21.1 8.2 6.7
Prevalence at subgroup levels is adjusted for age and regions as appropriate *Except for age and regions, characteristics were based on the characteristics of the participants collected at the resurvey
Reasons for stopping drinking by socio-demographic characteristics and health factors among previous weekly drinkers at resurvey (2013–2014)
Characteristics* N Reasons Existing illness (%) Money (%) Future health^ (%) Other# (%) Overall 1166 51.9 23.6 5.8 18.7 Socio-demographic characteristics Gender Men 970 53.1 24.3 6.3 16.3 Women 196 45.9 19.9 3.6 30.6 Birth cohorts < 1940 174 56.9 17.8 5.7 19.5 1940–1949 408 57.4 21.1 5.4 16.2 1950–1959 381 47.2 27.3 7.1 18.4 ≥ 1960 203 45.3 26.6 4.4 23.6 Area Rural 687 60.3 21.5 6.3 11.9 Urban 479 39.9 26.5 5.2 28.4 Highest education Primary or below 589 60.1 20.4 7.0 12.6 Middle or above 577 43.5 26.9 4.7 25.0 Household income (yuan/year) < 35,000 487 58.5 23.2 5.3 12.9 35,000+ 679 47.1 23.9 6.2 22.8 Health factors Self-reported health statusa Good 408 42.9 26.0 7.1 24.0 Poor 758 56.7 22.3 5.1 15.8 Prior diseaseb No 618 44.8 27.0 6.5 21.7 Yes 548 59.9 19.7 5.1 15.3
*Except for age and regions, characteristics were based on the characteristics of the participants collected at the resurvey
This large study examines recent patterns and trends of alcohol drinking and factors associated with stopping drinking in urban and rural areas of China. Overall, there was a modest decline in weekly drinking prevalence between 2004 and 2013 as the study population became older, but a modest increase in mean consumption, drinking frequency and heavy episodic drinking prevalence among men who drank alcohol weekly. Younger age and having multiple health-related risk factors were important correlates of heavy drinking patterns. Older and less healthy men were more likely to stop drinking, suggesting that existing illness was a major reason for stopping drinking in this Chinese population.
The findings on drinking prevalence in our study are broadly consistent with previous nationwide cross-sectional surveys at particular time point in China during 2002–2011 [[
Previous studies in China conducted in the 2000s reported that harmful drinking patterns peaked between the ages of 30–60 years in men [[
As in the present study, previous studies in England and Hong Kong also reported the clustering of heavy drinking with smoking, low fruit and vegetables intake and low exercise [[
One important step in alcohol policy development is to understand the triggers for drinking cessation in Chinese populations. Our findings that existing illness was the most reported reason for stopping weekly drinking are generally consistent with a study of 14,000 participants in China in the 1990s [[
Although not designed to be nationally representative, the diverse geographical regions, large sample size in CKB study and rich data on alcohol drinking patterns covered present a good source of data to assess recent alcohol consumption trends for Chinese adults born in 1920s–1970s. Furthermore, to our knowledge, it is the first study to explore longitudinal changes in alcohol drinking among Chinese adults. However, although the expected associations between alcohol intake and blood pressure observed in this study have provided support for the data quality, potential reporting bias of the self-reported alcohol intake data may still exist. Though older participants were slightly less likely to participate in the resurvey, the resurvey response rates were highly consistent across baseline alcohol drinking groups, suggesting that our findings on temporal trends of alcohol drinking were unlikely to be substantially biased by differential resurvey response rates. However, our results may still be partially influenced by the "sick-quitter" effect, i.e. older drinkers had quit alcohol drinking between the study period and the remaining younger heavy drinkers in the weekly drinker group drove up the mean consumption at resurvey. Furthermore, the measure of alcohol use based on "last time drinking" may have limitations subject to time variation [[
In summary, among men who drank regularly, the proportion engaging in harmful drinking behaviours has increased in China over the past decade, particularly among younger men, and heavy intake was associated with having other unhealthy lifestyles and health-related risk factors. As alcohol is a modifiable risk factor with reversible health hazards after stopping drinking in the long term, encouraging heavy drinkers to stop drinking before illness develops could have significant benefits to public health in China over the next decades.
PI analysed the data and drafted the manuscript. IM, LY and ZC contributed to the conception of this paper, interpretation of the results and the revision of manuscript. LL and ZC designed the study. LL, ZC, IM, LY, YG, ZB, HD, YC, YT, ZG, SW and YH contributed to data acquisition. All authors critically reviewed the manuscript and approved the final submission.
Ethical approval was obtained from the Ethical Review Committee of the Chinese Centre for Disease Control and Prevention (Beijing, China) and the Oxford Tropical Research Ethics Committee, University of Oxford (UK), and all participants provided written informed consent.
Not applicable.
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Graph: Supplementary tables and figures. (DOCX 118 kb)
We acknowledge the participants, the project staff, the China National Centre for Disease Control and Prevention, and its regional offices for access to death and disease registries. The Chinese National Health Insurance scheme provides electronic linkage to all hospital admission data.
The baseline survey was supported by the Kadoorie Charitable Foundation in Hong Kong. The long-term continuation of the study is supported by the UK Wellcome Trust, the Chinese Natural Science Foundation and the Chinese Ministry of Science and Technology. The UK Medical Research Council, British Heart Foundation and Cancer Research UK provided core funding to the Oxford Clinical Trial Service Unit. The funders had no role in the study design, data collection and analysis, decision to publish or preparation of the manuscript.
The datasets used and analysed during the current study are available from the corresponding author on reasonable request.
By Pek Kei Im; Iona Y. Millwood; Yu Guo; Huaidong Du; Yiping Chen; Zheng Bian; Yunlong Tan; Zhendong Guo; Shukuan Wu; Yujie Hua; Liming Li; Ling Yang and Zhengming Chen
Reported by Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author