OUP user menu

★ editor's choice ★

Big men and atrial fibrillation: effects of body size and weight gain on risk of atrial fibrillation in men

Annika Rosengren, Paul J. Hauptman, Georg Lappas, Lars Olsson, Lars Wilhelmsen, Karl Swedberg
DOI: http://dx.doi.org/10.1093/eurheartj/ehp076 1113-1120 First published online: 20 March 2009

Abstract

Aims Obesity is a recognized risk factor for atrial fibrillation (AF), partly because of the association between body mass index (BMI) and atrial volume. We aimed to determine whether other factors relating to body size were related to AF.

Methods and results Data were derived from a random population sample of 6903 men (mean age 51.5 years) who underwent a single midlife evaluation as part of the multifactor Swedish Primary Prevention Study. A total of 1253 men (18.2%) had a subsequent hospital discharge diagnosis (principal or secondary) of AF during a maximum follow-up of 34.3 years. Body surface area (BSA) at age 20 (calculated from recalled weight and measured height) was strongly related to subsequent AF (P < 0.0001), as were midlife BMI and weight gain from age 20 to midlife (P < 0.0001). In a Cox regression model which adjusted for midlife BMI, weight gain and other risk factors, hazard ratios (HR) [95% confidence intervals (CI)] for AF for the second, third, and fourth quartile of BSA at age 20, compared with the lowest quartile, were 1.47 (95% CI, 1.22–1.76), 1.66 (95% CI, 1.38–2.00), and 2.22 (95% CI, 1.82–2.70) (P for trend <0.0001).

Conclusion Large body size in youth, in an era when obesity was rare, as well as weight gain from age 20 to midlife, were both independently related to the development of AF. Given the current trends not only for obesity but also for height, a substantial increase in the incidence of AF is likely.

Keywords
  • Atrial fibrillation
  • Obesity
  • Body size
  • Stature

Introduction

Atrial fibrillation (AF) often accompanies other cardiovascular conditions and is a major cause of morbidity and mortality. In addition to coronary disease, heart failure, and valvular disease,1 prior studies have established that key risk factors for the development of AF include increasing age, hypertension,24 and pulse pressure.5 More recently, increased body mass index (BMI)3,4,69 and weight gain4 have been implicated. As a result, secular trends with respect to obesity may contribute to the well-documented increase in prevalence and incidence of AF.1015

The association between obesity and AF may be partly mediated through a complex relationship with atrial volume,6 which is determined not only by the degree of obesity but also by other factors related to body size. In a recent cross-sectional study from a large patient population with impaired left ventricular function, stature measured by height as well as by body surface area (BSA) was strongly correlated with AF.16 Several studies have found a correlation between height and AF,4,7 raising the possibility that a large body size in itself might be a determinant of AF. In adulthood, a large body size will usually be owing to obesity, and only to a lesser extent to other structures, such as muscle mass and skeletal frame. In the present study, we sought to separate the influences on body size by studying self-reported weight in youth and subsequent weight gain on the risk of future AF between midlife and old age.

Participants and methods

Data were derived from men participating in the intervention group of the multifactor Primary Prevention Study (PPS), which began in Göteborg, Sweden in 1970.17 All men in the city born between 1915 and 1925, except those born in 1923, were randomized into three equally large groups. The men in one of the groups (intervention group; n = 10 004) were offered a medical examination to identify and treat risk factors. The screening examination included 7495 men (participation rate 75%) in the intervention group, and took place between January 1970 and March 1973 at which time the men ranged in age from 47 to 56 years. The intervention was essentially a high-risk strategy directed towards men with pronounced hypercholesterolaemia, severe hypertension, or heavy smoking habits, according to predefined criteria, with treatment offered at specialist clinics. During the first 12-year follow-up, there were no significant differences in outcomes with respect to cardiovascular disease or all-cause mortality between the intervention and control groups.17 All participants gave their informed consent to participate in the study, which was approved by the Ethics Committee for Medical Research at Göteborg University, complying with the Declaration of Helsinki.

Of the 7495 men in the intervention group, 457 had no record of recalled weight at age 20, 39 had prior AF (by ECG or record), and 96 prior myocardial infarction, or stroke, leaving 6903 men as the study population. Variables collected at the time of the midlife evaluation included recalled weight at age 20, smoking history, physical activity during leisure time (categorized as sedentary, moderate activity, or regular strenuous activity), treatment for hypertension, diabetes mellitus, occupation,18 and alcohol history (dichotomized by the presence or absence of abuse determined by registration with the Swedish Board of Social Welfare for medical or legal problems attributed to alcohol).

The screening examinations were performed in the afternoon. Weight was measured to the nearest 0.10 kg and height to the nearest 0.01 m. Body mass index (BMI) was calculated as measured weight in kilograms divided by measured height in meters squared. For calculation of BMI at age 20, recalled weight was used and height measured at the investigation. Weight change from age 20 to midlife was calculated as the ratio between measured weight and recalled weight at age 20. BSA was calculated according to the Mosteller formula19 as the square root of [height (cm) × weight (kg)/3600], using measured weight for midlife BSA and self-stated weight at age 20 for BSA in youth. Systolic and diastolic blood pressures (BP) were measured to the nearest 2 mmHg after 5 min of rest with the participant seated.

All participants were followed-up from the date of their baseline examination until 31 December 2004, using their unique personal identification number. A computer file of the study cohort was run against the Swedish national register on cause of death and the Swedish hospital discharge register. This procedure was reviewed and approved by the Ethics Committee.

The hospital discharge register has operated on a nationwide basis since 1987, but all discharges from Göteborg hospitals have been entered in the national register since 1970 (except 1976 owing to a legislative change for that year). For the purpose of the current study, AF was defined as a discharge from hospital with a primary or secondary diagnosis code of 427.92 [International Classification of Diseases, Eighth revision (ICD 8)], 427D [International Classification of Diseases, Ninth revision (ICD 9)], or I48 [International Classification of Diseases, 10th revision (ICD 10)]. Heart failure was defined as any discharge code of 427.00 or 427.10 (ICD 8), 428A, 428B, or 428X (ICD 9) or I50 (ICD 10). Non-fatal myocardial infarction was defined as discharge codes of 410 (ICD 8 and 9) or I21 (ICD 10).

Statistics

Differences in distribution across BSA and weight change categories were examined using Mantel–Haenzsel tests or trend tests, as appropriate. Prospective analyses were accomplished using Cox proportional hazards regression models to identify factors related to a hospital discharge with a diagnosis of AF. Time at risk was calculated to first hospitalization with a diagnosis of AF, to death, or to 31 December 2004.

Univariate regression analyses were used to evaluate potential confounders of the body size–AF relationship. We created dichotomous predictor variables using clinically relevant categories of BMI and weight change; six for the former (<20, 20–22.5, 22.5–25.0, 25.0–27.5, 27.5–30.0, and >30.0 kg/m2), and five for the latter (loss of more than 4%, stable weight defined as ±4% change, increase by 5–15%, 15–35%, and more than 35% increase). Variables for height, BMI at age 20, weight, and BSA at age 20 and at baseline were created using quartiles of the distribution as cut-off levels. Covariates selected for adjustment included age, systolic BP, use of antihypertensive therapy, diabetes at baseline, smoking category, alcohol abuse, and occupational class. In the multivariable analyses, age, body size variables, and systolic BP were entered as continuous variables, unless stated otherwise, and all others as categorical variables. We created variables for a discharge diagnosis of heart failure or AMI occurring before a diagnosis of AF and, using the admission dates for these hospitalizations, entered these into the multivariable models as time-dependent variables. We checked the assumption of proportional hazards for BSA in youth and for weight gain with the methods of cumulative sums.20 There was no evidence that the assumption of proportional hazards was violated in these models. To assess the linearity of any of the indicators of body size variables in the models we introduced an additional binary term with a value of 1 for values lower than the mean of the variable in question, and zero for the rest. A significant coefficient for this additional term would indicate a departure from the linear effect of the corresponding variable because the effect would be different along the range of possible measurements (values). The linearity assumption was further tested with the method of the fractional polynomials21 The analysis showed that the continuous variables in our models: BSAY (body surface when young), WKVOT (weight ratio, current/young), and BMI, all support the linearity assumption on the log hazard scale. This means that no other functional form of these variables (from a broad range of possible forms, the so called fractional polynomials) gives us a significantly better model fit than the linear form. All analyses were performed with SAS version 9.1 (SAS Institute, Cary, NC, USA). A two-sided P < 0.05 was considered statistically significant.

Results

The average recalled weight at age 20 was 69.1 (8.0) kg, with an estimated BMI of 22.4 (2.2) kg/m2, and mean change in weight from youth to midlife of +14.7 (14.3)%. Baseline characteristics by categories of BSA at age 20 and of weight change from age 20 to midlife are shown in Table 1. The relative and absolute increase in weight was greater in men with low BSA. As expected, cardiovascular risk factor profile deteriorated with increasing weight gain, except for midlife smoking which was less common in men who gained weight, and diabetes which was not related to weight change from age 20.

View this table:
Table 1

Anthropometric indices and cardiovascular risk factors

(A) BSA at age 20 (m2)
<1.75 (n = 1716)1.75–1.82 (n = 1744)1.83–1.91 (n = 1714)>1.91 (n = 1729)P for trend
 Age (years)51.6 ± 2.351.6 ± 2.351.5 ± 2.351.5 ± 2.20.008
 Weight at age 20 (kg)59.7 ± 4.166.4 ± 2.271.0 ± 2.179.1 ± 5.6<0.0001
 BMI at age 20 (kg/m2)a20.7 ± 1.922.0 ± 1.722.7 ± 1.724.1 ± 2.1<0.0001
 Height (cm)170 ± 5174 ± 4177 ± 4181 ± 5<0.0001
 Midlife weight (kg)70.7 ± 8.976.6 ± 8.680.5 ± 9.087.8 ± 11.0<0.0001
 Midlife BMI (kg/m2)24.5 ± 3.025.3 ± 3.025.7 ± 3.026.7 ± 3.4<0.0001
 Midlife BSA (m2)1.68 ± 0.061.92 ± 0.111.99 ± 0.122.10 ± 0.14<0.0001
 Systolic BP (mmHg)149 ± 22148 ± 22149 ± 22149 ± 210.18
 Diastolic BP (mmHg)94 ± 1394 ± 1395 ± 1396 ± 13<0.0001
 Serum cholesterol (mmol/L)6.46 ± 1.156.47 ± 1.186.46 ± 1.136.44 ± 1.150.36
 Regular smokers54.450.646.348.4<0.0001
 Treatment for hypertension5.05.35.05.90.28
 Diabetes2.02.12.01.60.39
 Alcohol problems8.96.66.25.5<0.0001
 Physically active12.516.616.618.9<0.0001
 Non-manual occupation23.027.929.332.3<0.0001
(B) Weight change from age 20
<4% (n = 1414)4–15% (n = 2455)15–35% (n = 2470)>35% (n = 564)
 Age (years)51.7 ± 2.351.4 ± 2.351.5 ± 2.351.7 ± 2.30.82
 Weight at age 20 (kg)71.9 ± 8.669.8 ± 7.568.0 ± 7.363.9 ± 8.5<0.0001
 BMI at age 20 (kg/m2)a23.4 ± 2.322.6 ± 1.922.0 ± 2.020.5 ± 2.4<0.0001
 BSA at age 20 (m2)1.87 ± 0.131.84 ± 0.121.82 ± 0.121.77 ± 0.14<0.0001
 Height (cm)175 ± 6176 ± 6176 ± 6176 ± 6<0.0001
 Midlife weight (kg)71.9 ± 8.676.5 ± 8.483.4 ± 9.492.8 ± 12.6<0.0001
 Midlife BMI (kg/m2)22.7 ± 2.324.8 ± 2.227.0 ± 2.629.8 ± 3.7<0.0001
 Midlife BSA (m2)1.84 ± 0.131.93 ± 0.132.02 ± 0.142.13 ± 0.16<0.0001
 Systolic BP (mmHg)143 ± 21147 ± 22152 ± 22158 ± 23<0.0001
 Diastolic BP (mmHg)90 ± 1293 ± 1297 ± 13101 ± 13<0.0001
 Serum cholesterol (mmol/L)6.21 ± 1.136.42 ± 1.126.58 ± 1.166.71 ± 1.23<0.0001
 Regular smokers62.750.642.547.2<0.0001
 Treatment for hypertension2.23.87.012.1<0.0001
 Diabetes2.21.51.93.60.18
 Alcohol problems8.15.86.210.60.62
 Physically active18.918.114.29.0<0.0001
 Non-manual occupation26.229.028.726.60.44
  • Values are means ± SD or percentages.

  • aCalculated from recalled weight at age 20, and measured midlife height.

Estimated body surface area at age 20, weight gain, and midlife body mass index as univariable predictors

During a maximum of 34.3 years of follow-up, 1253 (18.2%) men were discharged from hospital with a principal or secondary diagnosis of AF. The median time between the midlife examination and discharge for AF was 25.0 (interquartile range 20.1–28.7) years.

Univariate predictors of AF are shown in Tables 2 and 3. There was a progressive increase in risk of AF with increasing BMI. Similar graded associations were noted for all measures of body size, with the exception of estimated BMI at age 20. Subjects in the highest as opposed to the lowest quartile of body height and BSA, in youth or in midlife (not shown), had an approximate doubling in risk. The probability of developing AF over time increased across categories of BSA at age 20 and weight increase (Figures 1 and 2).

Figure 1

Cumulative incidence of atrial fibrillation by body surface area at age 20 [gray = quartile 1 (<1.75 m2), stippled = quartile 2 (1.75–1.91 m2) to 3, black = quartile 4 (>1.91 m2)].

Figure 2

Cumulative incidence of atrial fibrillation by weight gain (gray = <5%, stippled = gain of 5–35%, black = gain in excess of 35%).

View this table:
Table 2

Hazard ratios of atrial fibrillation (primary or secondary hospital diagnosis) by anthropometric variables

Risk factorNumber at riskCases per 100 000 observation years (number)Age-adjusted hazard ratio (95% CI)
Midlife BMI (kg/m2)
 <20.00191484 (20)0.98 (0.61–1.57)
 20.00–22.49883567 (123)Referent
 22.50–24.992104659 (345)1.14 (0.93–1.40)
 25.00–27.492053806 (408)1.40 (1.15–1.72)
 27.50–29.991101901 (235)1.67 (1.34–2.08)
 ≥30.00571978 (122)2.03 (1.58–2.60)
 Per kg/m21.07 (1.05–1.09)
P for trend<0.0001
Measured midlife height (cm)
 <1721737564 (235)Referent
 172–1751692714 (295)1.26 (1.06–1.49)
 176–1791651796 (319)1.46 (1.23–1.73)
 >1791823911 (404)1.68 (1.43–1.97)
 Per cm1.04 (1.03 to 1.04)
P for trend<0.0001
BMI at age 20 (kg/m2)a
 <211772672 (285)Referent
 21–22.21657718 (295)1.02 (0.87–1.20)
 22.3–23.71807747 (332)1.06 (0.91–1.25)
 >23.71667863 (341)1.28 (1.10–1.50)
 Per kg/m21.05 (1.02–1.08)
P for trend0.0004
Body surface area at age 20 (m2)a
 <1.751716521 (215)Referent
 1.75–1.831744699 (296)1.36 (1.14–1.62)
 1.83–1.911714781 (331)1.48 (1.25–1.76)
 >1.911729992 (411)1.97 (1.67–2.33)
P for trend<0.0001
Weight change from age 20 to midlife (%)
 Loss of more than 4%491659 (74)1.09 (0.83–1.44)
 No change (−4 to 4%)923656 (150)Referent
 Gain of 5–15%2455700 (427)1.11 (0.92–1.33)
 Gain of 16–35%2470821 (492)1.34 (1.12–1.61)
 Gain of >35%564886 (110)1.61 (1.26–2.06)
P for trend<0.0001
  • aCalculated from recalled weight at age 20, and measured midlife height.

View this table:
Table 3

Hazard ratios of atrial fibrillation (primary or secondary hospital diagnosis) by cardiovascular risk factors

Risk factorNumber at riskCases per 100 000 observation years (number)Age-adjusted hazard ratio (95% CI)
Systolic blood pressure (mmHg)a
 <1331733605 (270)Referent
 133–1451645692 (283)1.15 (0.98–1.36)
 146–1611784805 (345)1.40 (1.19–1.64)
 >1611730912 (354)1.73 (1.48–2.03)
P for trend<0.0001
Treated hypertension
 No6538725 (1158)Referent
 Yes3651234 (95)2.05 (1.67–2.53)
Serum cholesterol (mmol/L)b
 <5.701664779 (317)Referent
 5.70–6.301775751 (331)0.98 (0.84–1.15)
 6.40–7.101805759 (334)1.02 (0.88–1.19)
 >7.101594692 (258)0.96 (0.82–1.14)
P for trend0.80
Smokingc
 Never smoker2054765 (409)Referent
 Former smoker1404780 (279)1.07 (0.92–1.25)
 1–14 g/day2047698 (332)1.04 (0.90–1.21)
 >14 g/day1376767 (231)1.30 (1.11–1.53)
Occupational class
 Not classifiable365722 (59)0.97 (0.73–1.29)
 Unskilled workers1522747 (264)Referent
 Semi-skilled and skilled workers1775760 (326)0.96 (0.82–1.13)
 Lower officials1299679 (213)0.87 (0.73–1.04)
 Intermediate non-manual1183801 (239)0.96 (0.81–1.15)
 High officials, professionals759769 (152)0.88 (0.72–1.07)
Leisure time physical activityd
 Low1729721 (284)Referent
 Moderate3991733 (717)0.96 (0.84–1.10)
 Regular exercise1101859 (243)1.09 (0.92–1.29)
Alcohol problems
 No6433749 (1183)Referent
 Yes470747 (70)1.28 (1.01–1.63)
Diabetes
 No6769749 (1236)Referent
 Yes134707 (17)1.49 (0.92–2.41)
  • aMissing data for blood pressure in 11 men.

  • bMissing data for serum cholesterol for 65 men.

  • cSmoking habits were defined using four categories: never smokers, former smoker of more than 1 month's duration, and current daily smoking of 1–14 g, and 15 g or more of tobacco. One cigarette was considered to contain 1 g of tobacco; one cigarillo, 2 g; and one cigar, 5 g. Missing data for smoking in 22 men.

  • dMissing data on physical activity for 82 men.

Multivariable predictors of atrial fibrillation

Results of the multivariable Cox proportional hazards regressions for BMI, height, BSA at age 20, and weight gain are shown in Table 4. After multivariable adjustment, including interim MI and heart failure, the risk of AF in obese men compared with men with low normal BMI remained significantly elevated [hazard ratio (HR) 1.56 (95% CI 1.20–2.02)], with an estimated 4% increase in risk per one unit increase in BMI. Similarly, height was independently associated with risk of AF. There was a significantly increased risk for AF in the second, third, and fourth quartile of BSA at age 20, compared with the lowest [HR for the highest quartile 1.95 (1.65–2.31)] and for increase in weight from age 20 to midlife [HR for 35% or more 1.31 (1.02–1.68) compared with stable weight]. In a multivariable model which mutually adjusted for BSA, weight change, and BMI (continuous), BMI was not independently associated with AF (P = 0.63), whereas the association with both BSA and weight change became stronger [HR for the highest BSA quartile 2.22 (1.82–2.70) and for the highest weight gain category 1.90 (1.37–2.64)]. The exclusion or inclusion of BMI in the final model made little difference to the HR:s for BSA or weight change.

View this table:
Table 4

Hazard ratios of atrial fibrillation (primary or secondary hospital diagnosis) by anthropometric variables in adjusted multivariable models

Risk factorMultiple-adjusted hazard ratios (HR, 95% CI)aFinal model of multiple-adjusted hazard ratios (95% CI) for BSA at age 20 and weight gainb
Midlife BMI (kg/m2)
 <20.001.11 (0.67–1.78)
 20.00–22.49Referent
 22.50–24.991.15 (0.94–1.41)
 25.00–27.491.32 (1.10–1.66)
 27.50–29.991.50 (1.21–1.88)
 ≥30.001.56 (1.20–2.02)
 Per kg/m21.04 (1.02–1.06)
P for trend (continuous)<0.0001
Measured midlife height (cm)
 <172Referent
 172–1751.32 (1.11–1.57)
 176–1791.51 (1.27–1.79)
 >1791.81 (1.53–2.13)
 Per cm1.04 (1.03–1.05)
P for trend<0.0001
Body surface area at age 20
 <1.75ReferentReferent
 1.75–1.831.39 (1.16–1.66)1.47 (1.22–1.76)
 1.83–1.911.52 (1.28–1.81)1.66 (1.38–2.00)
 >1.911.95 (1.65–2.31)2.22 (1.82–2.70)
P for trend<0.0001<0.0001
Weight change from age 20 to midlife, %
 Loss of more than 4%1.11 (0.84–1.47)1.00 (0.72–1.27)
 No change (−4 to 4%)ReferentReferent
 Gain of 5–15%1.08 (0.90–1.30)1.16 (0.96–1.41)
 Gain of 16–35%1.22 (1.02–1.47)1.46 (1.17–1.81)
 Gain of >35%1.31 (1.02–1.68)1.90 (1.37–2.64)
P for trend<0.0001<0.0001
  • aAdjusted for age at baseline, intercurrent heart failure (time-dependent), intercurrent myocardial infarction (time-dependent), systolic blood pressure, treatment for hypertension, smoking, diabetes, alcohol problems, occupational class.

  • bSame as above + BMI (continuous) at baseline. Both body surface area (BSA) at age 20 and weight change in the model. Final models for trend include all three variables as continuous.

Discussion

In this long-term prospective study of men, we found that large body size in youth, as well as weight gain from age 20 predicted the subsequent development of AF. Associations were graded, and to be in the highest, as opposed to the lowest, quartile of body height or BSA, in youth or in midlife, was associated with an approximate doubling in risk of developing AF.

Obesity and weight gain as risk factors

Several recent studies have documented an increased risk of AF with increasing BMI. In the Danish Cancer, Diet, and Health study,7 an increase of one unit of BMI corresponded to an increase in risk of AF of 1.08, which is close to the age-adjusted estimate of 1.07 in the present study. A recent meta-analysis found that obesity increased the risk of developing AF by 49% in the general population, and the risk escalated in parallel with increased BMI.22

Recalled weight was validated against measured weight 28 years earlier in the Charleston heart study;23 the overall correlation was 0.82 but slim men tended to overestimate their past weight, whereas overweight men underestimated past weight. This phenomenon, if applied to our population, is, however, unlikely to be of a magnitude that would meaningfully alter our results. In a secondary analysis we assumed a range of errors between −4 to 4 kg in weight self-reporting to generate nine normal random variables as a proxy for the real weight at 20. The final estimates for the main effects associated with the variables of interest varied very little, regardless of the size of the assumed error. Our findings underline the importance of weight gain for the later development of AF.

Body size

Several studies have noted a relationship between body height and AF.2,4,7,8,16 In a recent large cross-sectional study of patients with reduced left ventricular ejection fraction, the effect of height on AF risk persisted after adjusting for clinical and demographic variables, with an estimated 3% increase in risk per cm increase in height.16 Mean left atrial diameter was found to be significantly larger in patients with height above the population median. In the present study we were able to extend this observation to a healthy population.

Body size in an adult population is the net sum of body size when young and accumulation of weight during adult life. Accordingly, an elevated midlife BSA is usually to a large extent because of obesity. The self-stated weight of men in this study when young dates from the period 1935 to 1945, when obesity was rare. Only 25 men (0.4%) had an estimated BMI of 30 kg/m2 or more. Overweight and obesity are less prevalent in Sweden than in many other European countries and in the USA, but the current rate of increase is almost high.24 It is also noteworthy that middle-aged men in Sweden born in 1913 and in 1943 display a difference in height of 3 cm, and in body weight of almost 7 kg.25 Given the current trends in both weight and height in several Western countries,26 the impact on the incidence of AF may be significant.

The association between body size, whether expressed as height, weight, BMI, or BSA, and AF is biologically plausible, given the association between body size and left atrial size.6,16,27 In the Cardiovascular Health Study, left atrial size was strongly and independently associated with the incidence of AF during 3 years of follow-up.2 Wang et al.6 found that the association between BMI and AF risk became insignificant after adjustment for LA size and concluded that LA enlargement accounted for the entire observed association between BMI and AF risk. We have no echocardiographic data, but it is plausible that the relationship between body size and AF we have reported is mediated through effects on atrial size.

Limitations

There are several important limitations to our study. First, the development of AF was defined by admission to the hospital. Many episodes of AF are undetected, especially if asymptomatic. In the Cardiovascular Health Study,2 which documented 302 AF cases over a 3-year period from self-report, ECG, and hospital discharge records, two out of three were identified by hospital records. With our extended follow-up, it is likely that a higher proportion were identified. In addition, although the Framingham study28 suggested a one in four lifetime risk of developing AF for those greater than 40 years of age, the cumulative 30-year risk of AF in 55-year-old men was estimated to be 20% in the Rotterdam study,29 similar to the 18% documented in our study. Accordingly, it is probable that a comparatively large proportion of clinically important cases were detected. However, the fact that the men took part in a prevention trial could theoretically mean that they may have become more health-conscious, which could underestimate the burden of AF.

Secondly, our AF cases were not formally validated. However, register-based data for heart failure and AMI diagnoses in Sweden according to the hospital discharge register have been shown to have good validity.30,31 Similar hospital discharge diagnoses for AF from Denmark have been validated; in 174 retrieved records, evidence for AF was found in 99%.32

Thirdly, some of the increase in risk could reflect greater co-morbidity in obese men, but estimates for the HRs for AF did not change when we included intercurrent AMI or heart failure as potential confounding factors. Body size in youth was not related to AMI or heart failure, meaning that any misclassification with respect to AF is probably non-differential. In addition, we were unable to distinguish between AF and atrial flutter, because these two conditions have the same diagnostic codes. Similarly, we cannot distinguish paroxysmal from persistent AF as a cause of or contributor to hospitalization; accordingly some participants may only have had a single episode of paroxysmal AF. However, an index AF event is associated with a high rate of recurrence and of conversion to persistent AF.33 Other limitations include that lifestyle factors, such as smoking, alcohol consumption or physical activity, but also biological markers such as blood pressure are likely to change during follow-up. Thus, there is a potential for residual confounding not only for factors not assessed in the study but also with respect to covariates included in the analysis. In addition, only men were investigated, and hence, results may not be generalizable to women.

In summary, using longitudinal data over an extended time frame, this study confirms the known association between obesity and risk of AF, while adding that large body stature in youth confers an independent and substantial increase in risk. Given the current trends not only for obesity but also for height in many Western countries, we may be facing a substantial increase in the prevalence and incidence of AF in the future.

Funding

The Swedish Research Council, the Swedish Council for Working Life and Social Research, and the Swedish Heart and Lung Foundation.

Conflict of interest: none declared.

References

View Abstract