Prevalence of atrial fibrillation

A detailed description of the recruitment procedure for the CoLaus|PsyCoLaus study and the follow-up process has been published [14]. Briefly, the CoLaus|PsyCoLaus study is a population-based cohort exploring the prevalence and determinants of cardiovascular risk factors in the population of Lausanne, Switzerland. Between 2003 and 2006, a nonstratified, representative sample was recruited based on the following inclusion criteria: (a) age 35–75 years, (b) willingness to participate, and (c) Caucasian origin. The first follow-up was performed between April 2009 and September 2012, 5.6 years on average after baseline; the second follow-up was performed between May 2014 and July 2017, 10.9 years on average after baseline. All study periods included an interview, a physical examination, and blood analysis. At the second follow-up, 12-lead electrocardiograms (ECG) were performed. Hence, this study was conducted using data from the second follow-up.

A resting 12-lead body surface ECG was performed using a portable ECG machine (CARDIOVIT MS-2015, Schiller Reomed® AG, Dietikon, Switzerland) with the patients in supine position, at a paper speed of 25 mm/s. Data on basic rhythm, ventricular rate, P waves, PQ interval, QRS width, and QT interval were collected. Atrial fibrillation was defined as irregular RR intervals and no discernible, distinct P waves. Typical AFL was defined as an atrial activity seen as a “saw tooth” pattern (also called “F waves”), especially in the inferior leads (II, III, aVF), with a variable ventricular rate (ratio of atrial-to-ventricular contraction ranging from 4:1 to 2:1). Atypical AFL was defined as an atrial rate of more than 180/min with no F waves in the inferior leads. All ECGs were interpreted by two physicians (DS, FB). In the case of disagreement, a senior cardiologist (JS) was consulted and his interpretation was retained.

For this study, we considered two main outcomes: “pure” AF (noted as “AF” in this study) and combined AF and typical AFL (noted as „AF + AFL“ in this study). Previous studies [8] grouped AF and typical AFL into the same category, as both arrhythmias share the same risk factors and complications [15].

Covariates were collected by self-administered questionnaires, interview, or physical examination. The detailed definitions of the covariates have been described elsewhere [16].

We collected personal and family history of cardiovascular events. Alcohol consumption was self-reported and categorized as alcohol drinkers (yes/no). Smoking status was defined as never, former, and current. Physical activity (PA) was assessed using a validated wrist-worn triaxial accelerometer (GENEActiv, Activinsights Ltd., UK) and classified in accordance with the World Health Organization (WHO) criteria [17].

Overweight was defined as a body mass index (BMI) of ≥ 25 and < 30 kg/m², and obesity as BMI ≥ 30 kg/m². Hypertension was defined as a systolic blood pressure (BP) of ≥ 140 mm Hg and/or diastolic BP ≥ 90 mm Hg and/or presence of an antihypertensive drug treatment.

Blood samples after an overnight fast included levels of total cholesterol, high-density lipoprotein (HDL)-cholesterol, triglycerides, glucose and glycated hemoglobin (HbA₁c), creatinine, high-sensitive C‑reactive protein (hs-CRP) and pro-brain natriuretic peptide (pro-BNP). Low-density lipoprotein (LDL)-cholesterol was calculated using the Friedewald formula. Dyslipidemia was defined as an HDL-cholesterol < 1 mmol/L in men and < 1.29 mmol/L in women and/or LDL-cholesterol ≥ 4.1 mmol/L (≥ 2.6 mmol/L if personal history of cardiovascular disease [CVD] or diabetes) and/or triglyceride ≥ 2.2 mmol/L, and/or presence of a hypolipidemic drug treatment. Diabetes was defined as a fasting plasma glucose level of ≥ 7 mmol/L and/or a HbA1c ≥ 48 mmol/mol 6.5% and/or presence of oral hypoglycemic or insulin treatment.

Obstructive sleep apnea (OSA) is a risk factors for AF or AFL [12]; thus, we performed a subgroup analysis using participants with OSA data (n = 1920), i.e., the Berlin Questionnaire [18]. This instrument was developed to identify people at high risk of OSA in the ambulatory setting. It is composed of three categories, based on different symptoms of sleep or sleepiness. High risk of OSA is defined as ≥ 2 positive categories.

We excluded patients (a) without ECG data and (b) with ≥ 1 missing covariate(s) mentioned above.

Statistical analyses were performed using Stata version 15.0 for windows (Stata Corp, College Station, TX, USA). Descriptive results are expressed as number of participants (percentage) for categorical variables and as average ± standard deviation or median and [interquartile range] for continuous variables. Bivariate analyses were performed using chi-square or Fisher’s exact test for qualitative variables and Student’s t test, analysis of variance, or the Kruskal–Wallis test for quantitative variables. Multivariable analysis was performed using logistic regression and the results are expressed as odds ratio (OR) and 95% confidence interval (CI). Statistical significance was assessed for a two-sided test with p < 0.05.

Due to its paroxysmal nature, it was possible that some participants already presented with AF or AFL in the past. As it was unknown who had already presented with these conditions, we hypothesized that participants taking anticoagulants might have AF or AFL. Therefore, we decided to exclude them from the sensitivity analysis.

The institutional Ethics Committee of the University of Lausanne, which later became the Commission cantonale d’éthique de la recherche sur l’être humain (www.​cer-vd.​ch) approved the baseline CoLaus|PsyCoLaus study (reference 16/03) and the approval was renewed for the first (reference 33/09) and second (reference 26/14) follow-ups. The study was performed in agreement with the Helsinki Declaration and all participants gave their signed informed consent before joining the study.

On the initial 4881 participants, 4616 (94.6%) were retained for analysis. The exclusion criteria are indicated in Fig. 1 and the characteristics of the included and excluded participants are summarized in Supplementary Table 1. Included participants were younger, had lower levels of hypertension, diabetes, dyslipidemia, creatinine, hs-CRP, and pro-BNP, and less frequently reported a personal history of CVD. Included participants also reported higher levels of alcohol units consumed per week and were more physically active. Similar findings were obtained when exclusion criteria were extended to participants with at least one missing covariate(s) (Supplementary Table 2), except that included participants were more frequently overweight and had dyslipidemia and less frequently alcohol drinkers. To summarize, excluded patients were older and sicker, which might explain the absence of ECG data as those patients are harder to keep in a prospective study [19, 20].

Overall, 42 cases of AF were detected, resulting in an AF point prevalence of 0.9% (95% CI: 0.7–1.2). The point prevalence of AF by sex and age is presented in Fig. 2.

Furthermore, 10 cases of typical AFL were recorded, but no atypical AFL was detected. Therefore, the combined point prevalence of AF + AFL is 1.1% (95% CI: 0.4–1.5%). The point prevalence of combined AF + AFL by sex and age is presented in Fig. 3.

The bivariate comparisons of the clinical and sociodemographic characteristics between AF and non-AF cases are summarized in Table 1. Participants with AF were older, more often men (81%), presented more often with hypertension, dyslipidemia, or diabetes, had higher levels of BMI, and had more frequently a personal history of CVD. Participants with AF also had higher levels of creatinine, hs-CRP, and pro-BNP, were more often alcohol drinkers, reported higher alcohol consumption, and were less physically active. The AF participants were slightly more at risk of OSA but without reaching a statistically significant difference.

Similar findings were obtained when comparing participants with and without AF + AFL (combined outcome; Table 2) or when restricting the analysis to participants with all covariates, except that no difference was found regarding dyslipidemia (Supplementary Table 3).

Multivariable analysis identified increasing age and male gender as being positively associated with AF and AF + AFL (Table 3). The others risk factors associated with AF and AF + AFL also tended to be positively associated, but were not statistically significant.

In the subgroup of participants with OSA data, multivariate analysis including the Berlin Questionnaire showed overall similar results, except that a high risk of OSA was not statistically significantly associated with AF or AF/AFL (Supplementary Table 4).

This study has some limitations. First, our study potentially excludes participants with more comorbidities, since participation in population-based studies are selective and healthy people participate more frequently [20]. Thus, our study might have underestimated the prevalence of AF and AFL. However, this effect could be partly counterbalanced by the fact that our study included only participants of Caucasian origin (inclusion criteria at baseline) and it is known that individuals of non-European ancestry have a lower prevalence of AF [35]. Furthermore, the prevalence of AF and AFL was derived from a single time point ECG, and therefore we could only calculate point prevalence, which means we might have underestimated the true AF prevalence, as paroxysmal AF might have been missed. Furthermore, we did not detect atypical AFL in our study, which is also caused by atrial volume overload and we cannot make a statement about its prevalence. Smartphone-based ambulatory monitoring will probably be an attractive alternative in the future [36]. Finally, the small number of participants with AF reduced the statistical power and precluded the identification of some determinants of AF.