OUP user menu

Risk of arrhythmias in 52 755 long-distance cross-country skiers: a cohort study

Kasper Andersen , Bahman Farahmand , Anders Ahlbom , Claes Held , Sverker Ljunghall , Karl Michaëlsson , Johan Sundström
DOI: http://dx.doi.org/10.1093/eurheartj/eht188 First published online: 11 June 2013


Aims We aimed to investigate the association of number of completed races and finishing time with risk of arrhythmias among participants of Vasaloppet, a 90 km cross-country skiing event.

Methods and results All the participants without cardiovascular disease who completed Vasaloppet during 1989–98 were followed through national registries until December 2005. Primary outcome was hospitalization for any arrhythmia and secondary outcomes were atrial fibrillation/flutter (AF), bradyarrhythmias, other supraventricular tachycardias (SVT), and ventricular tachycardia/ventricular fibrillation/cardiac arrest (VT/VF/CA). Among 52 755 participants, 919 experienced arrhythmia during follow-up. Adjusting for age, education, and occupational status, those who completed the highest number of races during the period had higher risk of any arrhythmias [hazard ratio (HR)1.30; 95% CI 1.08–1.58; for ≥5 vs. 1 completed race], AF (HR 1.29; 95% CI 1.04–1.61), and bradyarrhythmias (HR 2.10; 95% CI 1.28–3.47). Those who had the fastest relative finishing time also had higher risk of any arrhythmias (HR 1.30; 95% CI 1.04–1.62; for 100–160% vs. >240% of winning time), AF (1.20; 95% CI 0.93–1.55), and bradyarrhythmias (HR 1.85; 95% CI 0.97–3.54). SVT or VT/VF/CA was not associated with finishing time or number of completed races.

Conclusions Among male participants of a 90 km cross-country skiing event, a faster finishing time and a high number of completed races were associated with higher risk of arrhythmias. This was mainly driven by a higher incidence of AF and bradyarrhythmias. No association with SVT or VT/VF/CA was found.

  • Arrhythmia
  • Atrial fibrillation
  • Bradycardia
  • Exercise training
  • Cross-country skiing


Physical inactivity is a major risk factor for cardiovascular disease, and regular leisure-time exercise reduces cardiovascular risk significantly.14 On the other hand, strenuous physical exercise may induce life-threatening ventricular arrhythmias in patients with pre-existing heart disease; autopsies of 1435 athletes who suffered sudden cardiac death during training have revealed cardiac abnormalities in 97% of the cases.58

Endurance training can lead to a number of physiological structural cardiac changes, including left atrial dilation and left ventricular dilation and hypertrophy.9,10 Further, sinus bradycardia, first-degree atrioventricular block, and second-degree atrioventricular block type I (Wenckebach) are considered normal responses to training.11 These structural changes are at least partly reversible, but an increased incidence of sinus node dysfunction is seen in retired elite cyclists.12,13 Although some of these structural changes and bradyarrhythmias may be physiological, associations with adverse clinical outcomes are unknown. Several small case–control studies and one cohort study have demonstrated that athletes committed to endurance exercise have increased risk of atrial fibrillation and flutter compared with the normal population,1420 but adequately powered cohort studies are lacking.

We hypothesized that a long-term increased workload as a consequence of prolonged endurance training would lead to structural changes in the heart and autonomic disturbances, which could increase arrhythmogenicity. We investigated the association of number of completed races and finishing time in the race, with the risk of hospitalization for arrhythmias in a cohort of 47 477 male and 5278 female participants in a long-distance cross-country ski race, ranging from recreational skiers to elite athletes.


Setting and participants

All Swedish participants in the 90 kilometre skiing event Vasaloppet (www.vasaloppet.se) who completed the race during the period 1989–98 were included. Vasaloppet takes place on the first Sunday in March every year and is a cross-country skiing event from Sälen to Mora in Dalarna, Sweden. Approximately 15 000 participants ranging from recreational to elite skiers complete the 90 kilometres of cross-country skiing every year. The 90 km Vasaloppet has two competitions: (i) the main race on the first Sunday of March where participants start in a large group; (ii) for those who prefer avoiding the stress of the group start, there are two additional race days (‘öppet spår’) where participants can start any time within an hour. The track completed is the same. This study includes both participants starting in the main race and in the ‘öppet spår’ race. During the inclusion period, one race was cancelled (1990, because of thawing).

The participants in Vasaloppet generally have higher leisure time physical activity, lower incidence of physical and mental illness, tobacco consumption, fat intake, and higher fibre consumption than average Swedes.21 Further, participants in the race have lower mortality in all major diagnostic groups (cancers and diseases of the circulation system), injuries and poisoning.21

The Vasaloppet office provided information on all Swedish participants who completed any of the races 1989–98 (49 117 men and 5443 women), including year of the race, finishing time, and the unique 10 digit national registration number to The National Board of Health and Welfare, Sweden. The National Board of Health and Welfare linked the data to: (i) the Swedish National In-Patient Register to obtain a list of diagnoses, (ii) censuses 1960, 1970, 1980, and 1990 to receive information on occupation and educational level, and (iii) the Swedish Population Register to obtain information of emigration. Anonymous data were then delivered to the authors. Participants hospitalized with cardiovascular disease (ICD-10 I.00–99 or similar diagnosis in ICD-7/8/9) before the baseline date were excluded (n = 1803). The final study sample comprised 47 477 men and 5278 women without any cardiovascular disease. The study protocol was approved by the Regional Ethical Review Board, Karolinska Institutet, Stockholm, Sweden.

Baseline data

Because improvements in VO2max are directly related to the intensity, duration, and frequency of training,22 we used the number of finished races and finishing time as a measure of the participants' duration and intensity of exposure to physical exercise. We investigated finishing time as a categorical variable, divided into four groups of percentages of the winning time that year [100–160%, 161–200%, 201–240%, and >240% (reference group)]. Only the best relative finishing time for each individual during the 10 year period was considered. Further, we investigated number of races as a categorical variable [1 race (reference group); 2 races; 3–4 races; and ≥5 races]. Number of races and finishing time before 1989 or after 1998 were not considered in the exposure assessment.

The cohort was linked to censuses 1960, 1970, 1980, and 1990 to receive information on occupation and educational level. The latest, most updated information for each person was used in this study. Occupation was grouped into four categories (blue-collar, lower-middle white-collar, high white-collar, and entrepreneur). Highest education level obtained was categorized as low (elementary school only), medium (secondary school), and high (university) levels. There were missing values for education in 3072 participants and for occupational status in 6331 participants.

Follow-up and outcome parameter

Participants were followed from the last participation in the race during the period from 1989–98 (the baseline date) to the date of first diagnosis of the outcome of interest, death, date of emigration, or the end of the follow-up (31st December 2005). Because the national registries were used for follow-up, loss to follow up was negligible.

The primary outcome was for any arrhythmia (all of the diagnoses below, plus ICD-10: I47.9). As secondary endpoint, we used (i) Bradyarrhythmias (ICD-10: I44.1, I44.2, I45.2, I45.3, I45.9, I49.5), (ii) Atrial fibrillation or flutter (AF) (ICD-10: I48.9), (iii) Other supraventricular tachycardias (SVT) (ICD-10: 45.6, I47.1), and (iv) ventricular tachycardia/ventricular fibrillation/cardiac arrest (VT/VF/CA) (ICD-10: I47.0, I47.2, I46.0, I46.1, I46.9, I49.0, R96.0). Corresponding codes for ICD-9 were used; see Supplementary material online, Table S1 for complete list of codes. Both main diagnosis and secondary diagnoses were taken under consideration. Note that use of ICD codes does not allow differentiation between Mobitz-types of A/V-block grade II.

Statistical analysis

Missing values for the variables ‘occupational status’ and ‘education’ were imputed from all other variables using multiple imputations by a chained equation approach (number of imputation cycles 25).23 For each endpoint, person-years of risk were calculated and Nelson-Aalen plots were produced to test the hazard proportionality assumption. No deviation was observed. To investigate associations of the exposures ‘number of races’ and ‘finishing time group’ with risk of arrhythmia outcomes, a Cox proportional hazards regression model was used. We modelled the risk of the primary endpoint (all arrhythmias) and the secondary endpoints (bradyarrhythmias, AF, other SVT, VT/VF/CA) in separate models. Using directed acyclic graphs, a model adjusting for age, occupational status, and education was deemed sufficient to estimate the total effect (evaluating all causal paths) of training level on arrhythmias, also in the absence of important covariates such as blood pressure, diabetes, hyperthyroidism, use of doping, smoking, and alcohol intake (Supplementary material online, Figure S1).24 Hence, two sets of models were investigated: (i) age-adjusted and (ii) adjusted for age, occupational status, and education. Further, to test for a trend by increasing exposure, a model treating ‘numbers of races’ and ‘finishing time group’ as continuous variables were performed. Multiplicative interaction terms between contrasting levels of the variables ‘number of races’ and ‘finishing time group’, and each of those two variables with contrasting age and sex groups, were investigated in the primary endpoint models using likelihood-ratio tests, but did not indicate important interactions. Stata 12 (StataCorp LP, USA) was used for all calculations, and two-tailed 95% confidence intervals (CI) were considered.


Baseline characteristics of the participants are shown in Table 1. For the primary outcome any arrhythmia, the median follow-up time was 9.7 years (minimum 0.01 years; maximum 16.8 years). Mean age at inclusion was 38.5 (SD 12.2) years, and mean age of first diagnosis was 56.8 (SD 13.5) years. Tables with separate results for men and women are shown in Supplementary material online, Table S2.

View this table:
Table 1

Baseline characteristics and incidence of any arrhythmias

Any arrhythmias

During a total of 513 496 person-years at risk, 919 cases of arrhythmia were reported (17.9; 95% CI 16.8–19.1/10 000 person-years at risk). Cumulative incidence of arrhythmias by number of completed races and finishing time is shown in Figure 1. Adjusting for age, education, and occupational status, we observed higher incidence of arrhythmias with increasing number of races (HR 1.30; 95% CI 1.08–1.58; for ≥5 vs. 1 completed races) and by faster finishing time (HR 1.30; 95% CI 1.04–1.62; for 100–160% vs. >240% of winning time) (Figure 2). Treating exposure as a continuous variable resulted in a HR of 1.06; 95% CI 0.99–1.13 by each step of ‘finishing time group’ and a HR of 1.10; 95% CI 1.03–1.16 by each step of ‘number of races’ group. A model only adjusting for age was also tested showing similar results (data not shown). Noteworthy, because of the natural loss of performance with increasing age, the number of elderly athletes in the fastest finishing time groups was limited.

Figure 1

Cumulative incidence of any arrhythmia by finishing time group in per cent of winning time and number of completed races during the period 1989–98.

Figure 2

Hazard ratios of any arrhythmia with 95% confidence intervals (log scale) by finishing time group in per cent of winning time and number of previous races. Model adjusted for age, occupation, and education level.


During follow-up, 119 participants were diagnosed with a bradyarrhythmia (2.3; 95% CI 1.9–2.8/10 000 person-years at risk), mainly grade II and III atrioventricular blocks and sick sinus syndromes (Table 2). When adjusting for age, education, and occupational status, higher risk of bradyarrhythmias was observed with increasing number of races (HR 2.10; 95% CI 1.28–3.47; for ≥5 vs. 1 completed races) and with faster finishing times (HR 1.85; 95% CI 0.97–3.54; for 100–160% vs. 240% of winning time) (Tables 3 and 4). Treating exposure as a continuous variable resulted in a HR of 1.29; 95% CI 1.10–1.52 by each step of ‘number of races’ group and a HR of 1.16; 95% CI 0.95–1.40 by each step of ‘finishing time group’. As a sensitivity analysis, we excluded potentially non-pathological bradyarrhythmias (atrioventricular blocks II and bi- and tri-fascicular blocks) from the outcome, with comparable results (Supplementary material online, Table S2).

View this table:
Table 2

List of bradyarrhythmia diagnoses

View this table:
Table 3

Risk of secondary outcomes by finishing time group

View this table:
Table 4

Risk of secondary outcomes by number of completed races

Atrial fibrillation and flutter

The most frequent arrhythmia was atrial fibrillation, which occurred in 681 skiers (13.2; 95% CI 12.3–14.3/10 000 person-years at risk). In a model adjusted for age, educational, and occupational status, we observed higher incidence of atrial fibrillation with higher number of races (HR 1.29; 95% CI 1.04–1.61 for ≥5 vs. 1 completed races) and a tendency to higher incidence of atrial fibrillation with faster finishing times (HR 1.20; CI 0.93–1.55; for 100–160% vs. 240% of winning time) (Tables 3 and 4). Treating exposure as a continuous variable resulted in a HR of 1.09; 95% CI 1.02–1.17 by each step of ‘number of races’ group and a HR of 1.04; 95% CI 0.96–1.13 by each step of ‘finishing time group’.

Other arrhythmias

The secondary endpoints of other SVT (n = 105) and VT/VF/CA (n = 90) were analysed in the same way. No associations of number of completed races or finishing time group with risk of SVT or VT/VF/CA were found (Tables 3 and 4).


This study confirms findings from earlier studies of higher incidence of arrhythmias with higher training level in athletes.1420 We observed a higher rate of AF and bradyarrhythmias among skiers with higher number of completed races and with faster relative finishing times, over a 10 year follow-up period. We did not observe associations between the number of completed races or finishing time group and the risk of other SVT or the more dangerous VT/VF/CA. This study extends current knowledge by providing cohort study data on persons in the higher end of the distribution of physical activity.


Abnormal ECG patterns are very common among elite athletes. Distinctly or mildly abnormal ECGs have been observed in as much as 50–60% of athletes committed to endurance sports.11 Sinus bradycardia, grade I atrioventricular block, and grade II Mobitz type I block should be considered as normal findings11,12,2527 but even the incidence of AV-block grade III is reported more frequently among athletes than that in the general population.26 Further, one earlier study reports higher rate of subclinical sinus node dysfunction among retired elite cyclists.13 The present study observed an approximately two-fold higher risk of hospitalization for bradyarrhythmias in athletes who completed five or more races compared with those who only completed one race. Of note, in order to capture clinically relevant bradyarrhythmias (i.e. potentially requiring a pacemaker), sinus bradycardia and grade I atrioventricular block were not included in this outcome. The use of ICD codes did not give opportunity to differ between atrioventricular blocks II Mobitz type I and II. For this reason, it is possible that some normal variants are included as pathological findings. However, grade II atrioventricular block only represents 23% of the bradyarrhythmia cases, and the results were consistent after excluding potentially non-pathological bradyarrhythmias (atrioventricular blocks II and bi- and tri-fascicular blocks) from the outcome. Interestingly, there may exist a link between bradycardia and increased risk of atrial fibrillation since atrial size, long PQ time, and bradycardia seem to be risk factors for atrial fibrillation among endurance trained athletes.19

Atrial fibrillation

Earlier case–control studies report higher incidence of atrial fibrillation in endurance sports-trained athletes and present odds ratios between 1.9 and 8.8 compared with non-athletes.1518 The effects of physical training in the present study may seem modest compared with earlier findings, but there are several methodological differences between the present study and the earlier studies, and the risks of bias in case–control studies are higher than in cohort studies. It is important to stress that this study only compares the risk of arrhythmias in an active population, and there may be a threshold training amount, below the level of most Vasaloppet participants. The Physicians Health study among 1572 healthy physicians showed increased risk of atrial fibrillation by increasing days per week of exercise vigorous enough to break a sweat.14 Although neither that study nor the present study had direct measurements of fitness level, the present study likely investigates individuals at a higher fitness level and hence extends the dose–response curve. It is possible that associations of physical training level with atrial fibrillation would have been even greater had the cohort included also less physically active people.

Incidence rates of atrial fibrillation have been sparsely reported for young people. The Rotterdam and Framingham studies report incidence rates of atrial fibrillation among men aged 55–64 representing the general population at 22 and 31 per 10 000 person-years at risk, respectively.28,29 In the present cohort, we find an incidence rate of atrial fibrillation and flutter of 49 (95% CI 43–51) per 10 000 person-years at risk among men aged 55–64. Although there are several methodological differences between the studies, this could indicate a higher rate in the skiers of this study than in the general population.

Other arrhythmias and death

It is well established that young athletes with pre-existing heart disease are at higher risk of sudden cardiac death.5 The heart disease is often unknown to the athlete, and this may also pertain to Vasaloppet participants. We did not observe higher incidence of sudden cardiac death with higher number of completed races or finishing time, but again it must be stressed that this study does not compare with the normal population. We have previously shown that participants in Vasaloppet have lower mortality than the general population and that mortality decreases with increasing number of races.21 Similar associations were observed in a Dutch long-distance skating event.30 We hence believe that it is generally safe to prepare for and participate in the race.

The Swedish National Board of Health and Welfare recommends, in accordance with the recommendations of the European Society of Cardiology,31 that all athletes competing at elite level have a physical evaluation including personal and family medical history, physical examination, and resting ECG. Only if these investigations reveal abnormal findings, further examinations (like echocardiography or stress testing) are performed. Which athletes are regarded as elite is defined by the specific sports associations (for skiers, the Swedish Ski Association). The elite group in Vasaloppet consists of a few hundred skiers of different nationality (this study only includes Swedish participants). We hence believe that the potential bias of screening is low, and for all practical purposes, this cohort should be regarded as non-screened.

Earlier cross-sectional studies report similar prevalence of SVT among athletes and the normal population, and this study supports independence of training level and SVT, at least in the setting of recreational skiers to elite athletes.26 This is not surprising since these arrhythmias (e.g. AV nodal reentrant tachycardia and Wolff-Parkinson-White) often rely on congenital or developmental changes in the heart's conduction system.

Strength and limitations

Strengths of the present study include the very large sample size of more than 52 000 athletes without known pre-existing heart disease, the cohort design, and the use of official registries for determination of socioeconomic measures and identification of the outcomes, which also minimizes loss-to-follow-up.

Some limitations of the study are worth noting. The incidence of atrial fibrillation increases dramatically with age, and because this cohort is quite young, possibly important associations at older age remain unknown. The exposure may be imprecisely measured, as the exposure is not limited to the participation in the race but may also include training before the race. Furthermore, many participants are likely committed to other sports activities during the summer. Indeed, in an earlier study of this population, 79.1% participants, compared with 29.5% in the normal population, reported strenuous exercise over the last year.21 Hence, our exposure measure may reflect all-year activity, which on the other hand may be viewed as a strength. Since we do not have any information on participation in races before the index date, it is possible that some of the participants have completed other races before the index date and only one afterwards. Likewise, follow-up was longer than the inclusion time, and participants could continue to race. Thus, it is likely that we may underestimate this exposure. This misclassification is not suspected to affect exposure groups differentially and should hence not bias the associations, but may affect the interpretation of the absolute numbers of races. It is possible that some participants did not complete the race because they developed arrhythmia during the race, and such participants are not included in the cohort. We do not have any information on non-fatal arrhythmias during the race, but we know that only one participant died in the track during the inclusion period.8

It is possible that some of the athletes have suffered from ischaemia-induced arrhythmias or arrhythmias of other specific aetiology, but we excluded persons with previous cardiovascular disease, and the incidence of subsequent cardiovascular disease is low in this population of physically active persons.21 Residual protopathic bias by pre-existing cardiovascular disease is possible (which would tend to bias results towards the null) but is unlikely a major explanation, as Vasaloppet is a strenuous race.

The study design limits the possibility of adjusting for several potentially important confounders, including smoking, alcohol use, blood pressures, and diabetes, but these were not identified as crucial covariates in order to minimize bias using the directed acyclic graphs approach, although residual confounding by these and other unmeasured factors is possible. This study is also limited by the lack of comparison of the sport active persons with inactive persons. The use of data from the Swedish National In-patient Register likely identifies outcomes with some random misclassification, which would tend to bias us towards the null and also precludes separation of atrial flutter and fibrillation or grade II atrioventricular blocks type I or II. On the other hand, this study uses diagnoses that can be read from standard ECG also by non-experienced doctors. Furthermore, it is possible that physically active persons have a lower threshold for seeking hospital service for arrhythmia symptoms than less physically active persons. We believe this potential confounding to be limited because the present cohort consists of persons with a training level needed for 90 km strenuous cross-country skiing; hence, most of the participants will likely experience symptoms and seek hospital services for arrhythmias regardless of race result or number of completed races. The generalizability to other ethnic groups is unknown. We did not find any interactions between sexes, but the limited number of women did not allow us to draw any firm conclusions regarding any sex-specific effect of exercise and risk of arrhythmias in women. This study cannot assess causality, but other studies suggest that longstanding endurance training could lead to vagal predominance32 and increased risk of extrasystoles and increases left atrial size33 and the amount of myocardial fibrosis,34,35 which may act as arrhythmia triggers and substrates.


In this study of 52 755 participants of a 90 km cross-country skiing event, a fast finishing time and a high number of completed races were associated with higher risk of arrhythmias. This was mainly driven by associations with risk of atrial fibrillation and bradyarrhythmias. No associations of number of completed races or finishing time with other SVT or dangerous arrhythmias such as VT/VF/CA were observed, which parallels previous observations in this cohort of lower standardized mortality with higher number of completed races.21


J.S. was funded by the Swedish Heart-Lung Foundation (grant 20041151) and the Swedish Research Council (grants 2007-5942 and 2010-1078). K.A. received a grant from the Geriatric Fund, Sweden.

Conflict of interest: J.S. is on the scientific advisory board for Itrim, a weight-loss company. The other authors did not report any conflicts of interest.


The authors thank Monica Eriksson at Vasaloppet's Office for providing information on study participants.


View Abstract