Study population

The study participants belonged to the Cohort of Swedish Men (COSM), a population-based cohort established between 1997 and 1998. All men born between 1918 and 1952 and residing in Örebro and Västmanland counties in Sweden were asked to complete an extensive questionnaire regarding physical activity, diet, anthropometric traits, and other lifestyle factors. In total, 48 850 men answered the questionnaire.

We excluded those participants with incorrect or missing identification numbers (N=297), those who died before the start of follow-up (N=55), and those with implausible energy intakes (ie, 3 SDs from the log_e-transformed mean energy intake, N=535). We also excluded participants with pre-baseline cancer (except non-melanoma skin cancer, N=2592), HF or myocardial infarction (based on linkage of the cohort to the Swedish National Patient Register, N=2971) because the diagnosis of these conditions may lead to changes in dietary pattern and beverage consumption. After theses exclusions, the final study population comprised 42 400 men.

Assessment of diet and beverage

Diet and beverage consumption was assessed in 1997 with a food frequency questionnaire (FFQ), designed to assess the Swedish diet, in which participants reported their average consumption of 96 foods and beverage during the past year. For assessment of commonly consumed beverage, such as soft drinks and coffee, study participants were asked to report their daily or weekly intake of a standard 200 mL (one glass) serving. Sweetened beverage consumption was measured by asking the participants the following question: “How many soft drinks or sweetened juice drinks do you drink per day or per week”? Fruit juice was not included in our definition of sweetened beverages.

HF ascertainment

Dates of the first registered incident HF hospitalisation as well as dates of deaths from HF were ascertained from 1 January 1998 to 31 December 2010 through linkage of the cohort to the Swedish National Patient Register and the Cause of Death Register. The Swedish National Patient Register represents the inpatient register, which includes all hospital admissions that entailed at least one overnight stay, and the outpatient register which includes diagnoses registered during non-private specialised care. For identification of HF incidence we used International Classification of Diseases-10 codes I50 and I11.0. We included incident HF events recorded in the registers listed as either the primary or secondary diagnosis of hospitalisation or death. The study has been approved by the Regional Ethical Review Board at Karolinska Institutet, Sweden, and completion of the self-administered questionnaire was considered to imply informed consent.

Assessment of covariates

The questionnaire included information on participant's educational attainment, family history of myocardial infarction, diagnosis of hypertension, diagnosis of diabetes (which was complemented with information from the diabetes register and the Swedish National Patient Register), smoking and drinking habits, weight, height, physical activity, and diet. Body mass index (BMI) was calculated as the weight (in kilograms) divided by the square of height (in metres). Physical activity was assessed by reported time spent at each activity per day (eg, work, walking/bicycling, exercise, housework, sleeping) and was multiplied by its typical energy expenditure requirements.18 Information on diagnoses of stroke and angina was obtained from the Swedish National Patient Register.

Statistical analyses

Data handling and generation of descriptive statistics were performed in SAS (V.9.2; SAS Institute, Inc, Cary, North Carolina, USA). Descriptive statistics of the study population were presented as means and SDs for continuous variables, and n (%) for categorical variables. Baseline characteristics were standardised to the age distribution of the total study population at baseline. Stata software V.12.1 (StataCorp) was used to calculate age-standardised rates and fit Kaplan-Meier survival methods and Cox proportional hazards regression. Kaplan-Meier survival methods were used to examine HF-free survival according to categories of sweetened beverage consumption. HRs and 95% CIs were estimated using Cox proportional hazards regression models with attained age as the underlying time scale. All p values were two-sided, and a value of p>0.05 indicated the absence of a statistically significant effect. Follow-up was censored at the date of first event of HF, death or 31 December 2010, whichever occurred first. The proportional hazards assumption was examined by investigating Schoenfeld's residuals.

Sweetened beverage consumption was divided into five different categories. The categorisation for the four upper groups was based on the nearest whole or half value of quartiles of sweetened beverage consumption among those who consumed sweetened beverages. This was done in order to make the categorisations of servings per day more interpretable. The reference category was no consumption of sweetened beverage.

Covariates included in the multivariable-adjusted model were: educational attainment (primary school, high school, university), smoking (never, past, current (≤10, >10 cigarettes/day)), alcohol consumption (never/past/current (<5, ≥5 g/day)), total physical activity (MET (metabolic equivalent) in hours/day categorised into quartiles), family history of myocardial infarction, history of stroke, history of angina, hypertension, diabetes, BMI (<18.5, 18.5–24.9, 25–29.9, ≥30 kg/m²), coffee consumption (cups/day), fruit intake (servings/day), vegetable intake (servings/day), processed meat consumption (g/day), fish consumption (servings/day), and total energy intake (kcal/day). The outcome variable was incident HF event. We carried out trend tests by assigning the median value for each category of sweetened beverage consumption and entered this variable as a continuous variable in the model. We also investigated if smoking status, being overweight, and age >65 years were potential effect modifiers in the association between sweetened beverage consumption and HF incidence by including them as interaction terms in the model.

To assess the potential effect of reverse causality —that is, undiagnosed HF contributes to subsequent changes in consumption of sweetened beverage—we conducted a sensitivity analysis where participants with HF events that were diagnosed during the first 5 years of follow-up were excluded from the model.

Three separate sensitivity analyses in which we excluded participants with either diabetes, history of stroke, or participants with history of angina from the model were also performed since these diseases could have strong confounding effects on the association between sweetened beverage consumption and HF. Diabetic patients are at increased risk of HF and they may have reduced their sweetened beverage consumption after their diagnosis. We also conducted an analysis where diabetes was not included as a covariate since diabetes could potentially have a mediator effect—that is, lie in the causal pathway between sweetened beverage consumption and HF incidence.