Study design

We conducted a cohort study with an 11‑year follow‑up, based on the Polish part of the HAPIEE (Health, Alcohol and Psychosocial Factors in Eastern Europe) project. The rationale for the study and the methodology of the whole project were described in detail elsewhere.¹⁸ Brief information relevant for this report is presented below.

Study sample

The study group was a random sample of 19 865 men and women selected from a population registry of permanent residents of Kraków aged 45 to 69 years, after stratification by sex, district, and 5‑year age groups. The response rate was 61%. After excluding those participants who did not agree for follow‑up, the study sample included 10 012 persons. All participants provided written consent to participate in the study. The study was approved by the bioethical committee at Jagiellonian University Medical College.

At baseline (2002–2005), trained nurses interviewed participants who completed an extensive, structured questionnaire, then underwent a physical examination in a clinic, and had a fasting blood sample taken. The examination procedure included 2 stages, and the participation rate for the clinical examination was approximately 10% lower than for the interview.

Assessment of alcohol use disorder

Alcohol use disorder was assessed by the CAGE questionnaire, a widely used and validated instrument in alcohol research.¹⁹ The questionnaire consists of the following 4 items: 1) Have you ever felt you should Cut down on your drinking?; 2) Have people Annoyed you by criticizing your drinking?; 3) Have you ever felt bad or Guilty about your drinking?; 4) Have you ever had a drink first thing in the morning to steady your nerves or to get rid of hangover (Eye opener)? The answers ‘no’ or ‘yes’ for each of the questions were coded as 0 or 1, respectively. The number of positive answers were summed. The score ranged between 0 and 4. The higher the score, the higher the probability of alcohol use disorder. A total score of 2 or greater was considered clinically significant for alcohol use disorder. In the current analysis, we adopted 2 approaches: 1) the estimation of the risk of CVD event for persons with CAGE score ≥2 compared with persons with a CAGE score of 0 to 1; and 2) the estimation of the risk of an incident of CVD for each number of points on the CAGE scale, with the reference category of 0 points.

Covariates

Covariates, measured at baseline, included age, education (vocational or lower, secondary, university), marital status (married or cohabiting versus single, separated, divorced, or widowed), smoking status (pack‑years), self‑reported history or presence of major cardiovascular chronic conditions (myocardial infarction, stroke; coded as yes versus no). Alcohol consumption was self‑reported by the participants using the graduated frequency questionnaire that included 9 mutually exclusive categories of frequency and amounts, in local units, of beer, wine, and spirit.¹⁹ Annual alcohol intake was assessed in grams of pure ethanol per year. Participants reporting no alcohol consumption (0 g of pure alcohol) in the past year were categorized as nondrinkers. Body mass index on clinical examination was calculated in kg/m². Hypertension was defined as blood pressure ≥140/90 mm Hg or receiving treatment for hypertension. Hypercholesterolemia was regarded as total cholesterol level ≥5 mmol/l or low‑density lipoprotein cholesterol level ≥3 mmol/l, or receiving lipid‑lowering treatment.^3,20 Diabetes was defined as fasting plasma glucose level ≥7 mmol/l or having diabetes diagnosed by a doctor.

Follow‑up

Data on deaths by cause were obtained using the mortality registry of residents of the city of Kraków, the Central Statistical Office, and by contacting the respondents’ families. The causes of deaths were coded according to the International Classification of Diseases and Related Health Problems, 10th Revision (ICD‑10). Deaths due to ICD‑10 codes from I00 to I99 were regarded as caused by CVD. New cases of nonfatal CVD including myocardial infarction, stroke, coronary artery bypass grafting, percutaneous coronary interventions, and unstable coronary artery disease (confirmed by coronary angiography) were identified in the respondents through 3 postal questionnaires (2005–2006, 2008–2010, and 2012–2013) and an re‑examination interview (2006–2008) and verified by the review of medical documentation. The identically worded questions as to whether the participant had had myocardial infarction, stroke, coronary angiography, coronary artery bypass grafting, or percutaneous coronary intervention were used in the postal questionnaires and during re‑examination. At the end of the follow‑up, the status of each respondent was determined and the exact survival time was calculated. The follow‑up was completed on December 31, 2014. For participants who were lost to follow‑up, the censorship date was the date of the last contact.

Statistical analysis

The distribution of categorical variables was presented as number and percentage and as mean (SD) or median (interquartile range) for continuous variables, as appropriate. The Cox regression was used to estimate hazard ratios (HRs) with 95% CIs for associations between the CAGE score and the risk of CVD, based on time‑on‑study as the time scale. Three models were fitted: 1) adjusted only for age and sex; 2) adjusted for age, sex, and smoking status; 3) adjusted for age, sex, smoking status, education, marital status, hypertension, hypercholesterolemia, diabetes, body mass index, and physical activity. In the supplementary tables, CAGE was also regarded as a continuous variable (P value for trend). The analysis was restricted to participants with complete records for all covariates.

Moderation analysis was performed to check whether the association between the CAGE score and the risk of CVD was homogeneous across age groups and districts. There was no interaction between the CAGE score and age as well as between the CAGE score and districts (χ² = 1.37, P = 0.85 and χ² = 1.32, P = 0.72, respectively). All analyses in the full study sample were repeated among alcohol consumers (after excluding abstainers). The STATA software, version 14 (StataCorp LP, College Station, Texas, United States) was used for all analyses.