The incidence of both atrial fibrillation (AF) and obesity is increasing in the community, and lifestyle intervention is recommended. We aimed to test whether the predictive effect of body mass index (BMI) and weight change from age 25 years to midlife on incident AF were influenced by physical fitness. In 1972 to 1975, 2,014 healthy middle-age men conducted a bicycle exercise electrocardiographic test as a part of a cardiovascular survey program, defining physical fitness as work performed divided by body weight. During 35 years of follow-up, 270 men developed AF, documented by scrutiny of the health files in all Norwegian hospitals. Risk estimation was analyzed using Cox proportional hazard models and tested for age-adjusted physical fitness above and below the median. The mean BMI of 24.6 kg/m 2 defined a lean baseline cohort. The men with a baseline BMI of ≥28 kg/m 2 (11%) compared to a BMI <28 kg/m 2 had a 1.68-fold risk of AF (95% confidence interval 1.14 to 2.40) and men reporting weight gain of ≥10 kg (24%) compared to weight loss (11%) of 1.66-fold (95% confidence interval 1.00 to 2.89), respectively. The dichotomy into men with age-adjusted physical fitness above and below the median, demonstrated statistically significant risk associations only for men with low fitness. The overall risk of AF was reduced by 23% in the fit men. In conclusion, within our lean baseline cohort of healthy middle-age men, a BMI of ≥28 kg/m 2 and weight gain of ≥10 kg from age 25 to midlife were long-term predictors of incident AF in men with physical fitness below the population median. The fit men had an overall slightly reduced risk of AF.
The increasing incidence of both atrial fibrillation (AF) and obesity has become major challenges in the general population. Several studies in the past decade have established a significant association between these disorders. The Danish Diet, Cancer and Health Study found an increased risk of AF among both overweight and obese men and among obese women, independent of intermediate occurrence of diabetes or heart disease. The Swedish Primary Prevention Study reported that a body surface area at age 20 years, midlife body mass index (BMI), and weight gain from age 20 years to midlife were independent long-term predictors of AF. A recent review exploring the health risks of normal-weight versus obese subjects in relation to physical activity showed that cardiovascular mortality was lower in those with a high BMI and good aerobic fitness compared to those with a normal BMI and poor fitness. No study has evaluated the predictive effect of BMI and weight change on incident AF after adjustments for measured physical capacity. The present cardiovascular survey has previously shown that systolic blood pressure in the upper normal range is a long-term predictor of AF in initially healthy middle-age men after adjustments for age and BMI. In the present study, we aimed to test whether the midlife BMI and self-reported weight change from the age of 25 years to midlife predicted incident AF, and second, whether this predictive effect was influenced by physical fitness.
Methods
From 1972 to 1975, 2,014 apparently healthy men aged 40 to 59 years from 5 governmental institutions in Oslo practicing annual or biennial health examinations of their employees were included in a prospective cardiovascular survey. The participants were defined as healthy if they had no evidence of heart disease or diagnosed hypertension, diabetes mellitus, thyroid disorders, cancer, advanced pulmonary, renal, or liver disease, or other serious disorders. In all, 2,341 men fulfilled the health criteria, and 2,014 (86%) agreed to participate. They underwent clinical examination, blood tests, chest radiography, electrocardiography at rest, and symptom-limited bicycle exercise electrocardiography. The examinations took place in the morning hours (7:00 to 11:00 a.m. ), and the men were requested to abstain from smoking for ≥8 hours and from eating for 12 hours. All participants were asked whether they had consumed alcohol in the previous 2 days before the examination as a measure of their propensity for alcohol consumption. Physical fitness was measured using an Elema bicycle (Siemens-Elema, Solna, Sweden) with an initial load of 100 W, increased by 50 W each 6 minutes. All were encouraged to exercise to exhaustion or until termination by the investigator for safety reasons. Submaximal termination involved 91 men (4.5%) and has been previously reported in detail. The total work capacity was calculated as the sum of work performed at all workloads, and physical fitness was defined as the total exercise work capacity in kilojoules divided by the body weight in kilograms. A detailed description of the inclusion process and examination procedures has been previously reported. All the participants provided informed consent.
The BMI at baseline was calculated by dividing the body weight (in kilograms to 1 decimal) by the body height (in meters to 2 decimals) squared. The World Health Organization has classified the BMI into normal weight (BMI 18.5 to 24.9 kg/m 2 ), overweight (BMI 25 to 29.9 kg/m 2 ), and obesity (BMI ≥30 kg/m 2 ). Because of a lean study population with few obese men, we used a BMI of ≥28 kg/m 2 , representing 11% of the total cohort, to define our upper BMI category. The limit of 28 kg/m 2 is also used in other lean populations, such as Asian people. The self-reported weight change since 25 years of age was classified at baseline by the following groups: weight loss, weight gain 0 to 4.9 kg, weight gain 5.0 to 9.9 kg, and weight gain ≥10 kg.
Morbidity data, including AF, hypertension, coronary heart disease, heart failure, cerebral insults, diabetes mellitus, and chronic obstructive pulmonary disease, were consecutively obtained from 3 clinical surveys (1979 to 1982, 1989 to 1990, and 1995 to 1996), 1 questionnaire survey (1987) and 2 nationwide searches of all patient records at all Norwegian hospitals (in 1995 to 1996 and 2005 to 2008, respectively). The hospital medical records were manually scrutinized, and data were obtained from all available sections in the journals (e.g., history, electrocardiographic readings, outpatient notes, and letters from general practitioners). Mortality data were obtained from the Database of Statistics Norway, completed up to December 31, 2007, and none was lost to follow-up. Cardiovascular mortality included sudden death, myocardial infarction, stroke, ruptured abdominal aneurysm, and pulmonary embolism.
The registration of AF was performed by 2 of the investigators (J.B., J.E.) according to a description of paroxysmal, persistent, or permanent AF or atrial flutter in the hospital records, confirmed by electrocardiogram, or infrequently by documentation of cardioversion. A few cases of atrial flutter were included owing to the same code number as AF in the International Classification of Diseases , 10th revision. Episodes of AF related to acute myocardial infarction, recent coronary artery bypass grafting, or valve replacement were not included. The participants with diagnosed AF were censored solely at the first event of AF.
Differences in the baseline data between groups were tested using Student’s t test or the Pearson chi-square test, according to data type. The Kendall rank test and Cochran Armitage trend test were used to assess the correlation (trend) between categories of BMI and weight change and baseline data. Cox regression models were used to study the associations between selected predictors and incident AF on univariate and multivariate analyses. Because of the close correlation, the univariate significant blood pressure components canceled each other out on concurrent multivariate analysis; therefore, we analyzed the blood pressure variables separately in the model and then finally entered systolic blood pressure to the full multivariate model. We tested the Cox model for interaction between physical fitness and BMI, and no interaction was found (p = 0.7). All results were adjusted for age, systolic blood pressure, current smoking, total cholesterol, and blood glucose and, at a final step, physical fitness. Moreover, Cox analyses were performed separately for men with age-adjusted physical fitness above and below median to assess a possible interaction between BMI and weight change and high or low fitness. Finally, in a subsequent model, Cox proportional hazard analyses were performed after censoring at the first occurrence of diabetes mellitus or cardiac disease (myocardial infarction, coronary artery bypass grafting, aortic valve replacement, heart failure) before an AF event. We calculated the 95% confidence intervals (CIs) throughout the analyses. The p values were 2-tailed and considered significant if p <0.05. The data were analyzed using the statistical package JMP, version 9 (SAS Institute, Cary, North Carolina).
Results
The average age at inclusion was 50 years, and 44% were current smokers ( Table 1 ). Few men (<4%) were obese, and the low mean BMI of 24.6 kg/m 2 characterized a lean study cohort. The blood pressure, heart rate, total cholesterol, and blood glucose increased with greater BMI values ( Table 2 ), and physical fitness and the prevalence of current smokers demonstrated an inverse correlation. Increasing weight before baseline demonstrated a similar highly significant correlation for trend with respect to the same variables ( Table 3 ).
| Characteristic | All (n = 2,014) | Sinus Rhythm (n = 1,744) | Outcome Atrial Fibrillation (n = 270) | p Value |
|---|---|---|---|---|
| Age at baseline (years) | 50 ± 5.5 | 50 ± 5.5 | 50 ± 5.2 | NS |
| Body height (cm) | 177 ± 6 | 177 ± 6 | 178 ± 6 | NS |
| Body weight (kg) | 77 ± 10 | 76 ± 10 | 79 ± 10 | <0.001 |
| Body mass index (kg/m 2 ) | 25 ± 2.8 | 24 ± 2.8 | 25 ± 2.7 | <0.01 |
| <25 | 1221 (60%) | 1,075 | 146 | |
| 25–27.9 | 580 (29%) | 500 | 80 | |
| 28–29.9 | 137 (7%) | 108 | 29 | |
| >30 | 76 (4%) | 61 | 15 | P trend <0.01 |
| Weight change from age 25 years | ||||
| Weight loss | 224 (11%) | 205 | 19 | |
| Weight gain 0–4.9 kg | 778 (39%) | 685 | 93 | |
| Weight gain 5.0–9.9 kg | 517 (26%) | 440 | 77 | |
| Weight gain ≥10 kg | 495 (24%) | 414 | 81 | P trend <0.01 |
| Current smoker | 882 (44%) | 778 (45%) | 104 (39%) | NS |
| Alcohol consumption ⁎ | 448 (22%) | 385 (22%) | 63 (23%) | NS |
| Systolic blood pressure (mm Hg) | 130 ± 18 | 130 ± 18 | 133 ± 18 | <0.05 |
| Diastolic blood pressure (mm Hg) | 87 ± 10 | 87 ± 10 | 89 ± 10 | <0.001 |
| Pulse pressure (mm Hg) | 43 ± 11 | 43 ± 11 | 44 ± 12 | NS |
| Heart rate at rest (beats/min) | 61 ± 10 | 61 ± 10 | 61 ± 10 | NS |
| PR interval (ms) | 172 ± 25 | 171 ± 24 | 172 ± 26 | NS |
| Exercise at maximum heart rate (beats/min) | 163 ± 14 | 163 ± 14 | 163 ± 13 | NS |
| Exercise at maximum systolic blood pressure (mm Hg) | 216 ± 22 | 216 ± 23 | 217 ± 21 | NS |
| Physical fitness (kJ/kg) | 1.94 ± 0.8 | 1.94 ± 0.8 | 1.89 ± 0.8 | NS |
| Total cholesterol (mmol/L) † | 6.7 ± 1.2 | 6.7 ± 1.2 | 6.6 ± 1.3 | NS |
| Blood glucose (mmol/L) ‡ | 4.4 ± 0.6 | 4.4 ± 0.6 | 4.4 ± 0.6 | NS |
⁎ Alcohol consumption in previous 2 days before examination.
† Total cholesterol: mmol/L × 38.7 = mg/dl.
| Variable | BMI (kg/m 2 ) | p Value | |||
|---|---|---|---|---|---|
| <25 (n = 1,221) | 25–27.9 (n = 580) | 28–29.9 (n = 137) | ≥30 (n = 76) | ||
| Age (years) | 49.6 ± 5.5 | 50.3 ± 5.4 | 50.3 ± 5.9 | 49.9 ± 5.6 | 0.007 |
| Body height (cm) | 177.0 ± 6.1 | 176.4 ± 6.2 | 176.8 ± 6.4 | 175.9 ± 7.0 | 0.1 |
| Body weight (kg) | 71.6 ± 6.9 | 81.7 ± 6.1 | 89.9 ± 6.3 | 98.7 ± 9.7 | <0.0001 |
| Body mass index (kg/m 2 ) | 22.9 ± 0.6 | 26.2 ± 0.8 | 28.7 ± 0.5 | 31.8 ± 2.0 | NA |
| Weight loss from age 25 | 17% | 2% | 0 | 0 | <0.0001 |
| Weight gain ≥10 kg from age 25 years | 9% | 39% | 70% | 88% | <0.0001 |
| Current smoker | 47% | 38% | 41% | 39% | 0.0009 |
| Alcohol consumption | 22% | 22% | 28% | 15% | 0.9 |
| Systolic blood pressure (mm Hg) | 128 ± 17 | 132 ± 18 | 138 ± 19 | 137 ± 18 | <0.0001 |
| Diastolic blood pressure (mm Hg) | 85 ± 10 | 89 ± 10 | 92 ± 10 | 93 ± 11 | <0.0001 |
| Heart rate at rest (beats/min) | 61 ± 10 | 62 ± 10 | 63 ± 10 | 63 ± 9 | 0.0002 |
| Total cholesterol (mmol/L) ⁎ | 6.6 ± 1.2 | 6.8 ± 1.2 | 6.6 ± 1.1 | 6.7 ± 1.4 | 0.04 |
| Physical fitness (kJ/kg) | 2.1 ± 0.8 | 1.7 ± 0.6 | 1.5 ± 0.5 | 1.2 ± 0.4 | <0.0001 |
| Blood glucose (mmol/L) † | 4.4 ± 0.5 | 4.5 ± 0.5 | 4.6 ± 0.7 | 4.8 ± 0.8 | <0.0001 |
⁎ Total cholesterol: mmol/L × 38.7 = mg/dl.
| Variable | Weight Loss (n = 224) | Weight Gain | |||
|---|---|---|---|---|---|
| 0–4.9 kg (n = 778) | 5.0–9.9 kg (n = 517) | ≥10 kg (n = 495) | p Value | ||
| Age (years) | 50.3 ± 5.5 | 49.5 ± 5.5 | 49.5 ± 5.5 | 50.6 ± 5.4 | 0.06 |
| Body height (cm) | 176.2 ± 6.0 | 176.4 ± 6.1 | 176.8 ± 6.0 | 177.4 ± 6.6 | 0.001 |
| Body weight (kg) | 68.4 ± 7.9 | 72.9 ± 7.4 | 78.3 ± 7.5 | 85.2 ± 9.9 | <0.0001 |
| Body mass index (kg/m 2 ) | 22.0 ± 2.0 | 23.4 ± 2.0 | 25.0 ± 1.9 | 27.1 ± 2.7 | <0.0001 |
| Current smoker | 54% | 47% | 40% | 39% | <0.0001 |
| Alcohol consumption | 19% | 24% | 23% | 20% | 0.6 |
| Systolic blood pressure (mm Hg) | 125 ± 16 | 128 ± 17 | 131 ± 18 | 136 ± 19 | <0.0001 |
| Diastolic blood pressure (mm Hg) | 83 ± 10 | 85 ± 10 | 88 ± 10 | 91 ± 11 | <0.0001 |
| Heart rate at rest (beats/min) | 60 ± 10 | 60 ± 9 | 62 ± 10 | 63 ± 10 | <0.0001 |
| Total cholesterol (mmol/L) ⁎ | 6.4 ± 1.2 | 6.6 ± 1.2 | 6.7 ± 1.2 | 6.8 ± 1.2 | <0.0001 |
| Physical fitness (kJ/kg) | 2.2 ± 0.8 | 2.2 ± 0.8 | 1.9 ± 0.7 | 1.5 ± 0.6 | <0.0001 |
| Blood glucose (mmol/L) † | 4.3 ± 0.4 | 4.4 ± 0.4 | 4.4 ± 0.6 | 4.6 ± 0.7 | 0.001 |
⁎ Total cholesterol: mmol/L × 38.7 = mg/dl.
In the present study, 270 men (13%) developed AF during ≤35 years of follow-up (median 30). The mean age at the first documented AF was 71 ± 9 years. The men who subsequently developed AF had a greater baseline BMI and systolic and diastolic blood pressure than the men who maintained sinus rhythm ( Table 1 ). Of the 270 participants with AF, 17 had no information about the time of diagnosis and were excluded from the survival analyses. Age, blood pressure at rest and at maximum exercise, maximum heart rate, BMI, and physical fitness were univariate predictors of AF. Age, blood pressure at rest, and BMI remained significant predictors on multivariate analyses, with age as the strongest predictor ( Table 4 ).
| Variable | Univariate Analysis | Multivariate Analysis ⁎ | ||||||
|---|---|---|---|---|---|---|---|---|
| HR | 95% CI | p Value | Chi-Square | HR | 95% CI | p Value | Chi-Square | |
| Age at baseline (10 years) | 1.94 | 1.53–2.43 | <0.001 | 31.6 | 1.77 | 1.40–2.24 | <0.001 | 22.6 |
| Systolic blood pressure (1 SD) | 1.34 | 1.19–1.50 | <0.001 | 21.9 | 1.22 | 1.08–1.37 | <0.005 | 9.5 |
| Diastolic blood pressure (1 SD) | 1.32 | 1.17–1.48 | <0.001 | 19.5 | 1.24 | 1.10–1.40 | <0.001 | 11.5 |
| Physical fitness (1 SD) | 0.78 | 0.68–0.89 | <0.001 | 13.7 | 0.75 | 0.20–2.70 | 0.67 | 0.2 |
| Exercise maximal heart rate (1 SD) | 0.79 | 0.70–0.90 | <0.001 | 12.4 | 0.40 | 0.13–1.25 | 0.11 | 2.5 |
| Pulse pressure (1 SD) | 1.25 | 1.10–1.41 | <0.001 | 11.5 | 1.12 | 0.99–1.27 | 0.08 | 3.2 |
| Body mass index (1 SD) | 1.22 | 1.08–1.37 | <0.005 | 10.0 | 1.16 | 1.02–1.32 | <0.05 | 5.4 |
| Exercise maximal systolic blood pressure (1 SD) | 1.18 | 1.04–1.34 | <0.05 | 6.5 | 1.08 | 0.95–1.23 | 0.22 | 1.5 |
| Total cholesterol (1 SD) | 1.11 | 0.98–1.26 | NS | 2.8 | ||||
| Body height (1 SD) | 1.07 | 0.94–1.20 | NS | 1.1 | ||||
| PR interval (1 SD) | 1.07 | 0.94–1.20 | NS | 1.1 | ||||
| Heart rate at rest (1 SD) | 0.98 | 0.86–1.11 | NS | 0.1 | ||||
| Blood glucose (1 SD) | 1.02 | 0.89–1.16 | NS | 0.1 | ||||
| Smoking (yes/no) | 1.04 | 0.80–1.32 | NS | 0.1 | ||||
⁎ Blood pressure components analyzed separately in multivariate analysis; multivariate analysis of other than blood pressure variables analyzed with systolic blood pressure in the full model.
The dichotomy of the population into men with age-adjusted physical fitness above and below the median demonstrated that the fittest men had an overall slightly reduced AF risk (118 vs 135 AF events; hazard ratio [HR] 0.77, 95% CI 0.60 to 0.99, p = 0.04) adjusted for age and systolic blood pressure. However, the 1% of the men with the absolutely greatest age-adjusted physical fitness at baseline had an increased crude AF incidence of 24% (6 AF events among 25 men; none reported participation in competition sports). The next 100 men with the greatest age-adjusted physical fitness had a crude AF incidence of 9% (p = 0.04 between these 2 groups).
BMI as a continuous variable predicted incident AF in the full multivariate model with a HR of 1.16 (95% CI 1.02 to 1.32) per SD (2.8 kg) increase. The categories of BMI related to AF risk are presented in Figure 1 and Table 5 . Men with a BMI of ≥28 kg/m 2 had a significantly increased multivariate adjusted risk of incident AF compared to the men in the lower BMI categories (HR 1.68, 95% CI 1.14 to 2.40, comparing BMI ≥28 and <28 kg/m 2 ). The influence of physical fitness was further explored in fitness groups above and below the median ( Table 5 ). Men with the greatest physical fitness and BMI of ≥28 kg/m 2 experienced few AF events, and no statistically significant differences in the risk of AF were found between the BMI categories; however, the HR of 1.52 compared to a BMI <25 kg/m 2 does not exclude a possible association. When assessing men with low fitness, a BMI of ≥28 kg/m 2 was associated with a 1.59-fold (95% CI 1.05 to 2.36) risk of AF compared to all men with a BMI <28 kg/m 2 . Before the occurrence of AF, 115 men were diagnosed with diabetes mellitus and/or myocardial infarction, coronary artery bypass grafting, aortic valve replacement, and heart failure. Adjustments for the presence of these disorders in the multivariate model tended to strengthen the previously established risk association between AF and a BMI of ≥28 kg/m 2 versus a BMI of <28 kg/m 2 (HR 1.91, 95% CI 1.10 to 3.15 for the entire cohort, after dichotomy; HR 1.29, 95% CI 0.31 to 3.55 for above and HR 2.00, 95% CI 1.08 to 3.52 for below the median physical fitness; data not shown in the tables).
