Exercise linked to 22% lower odds of atrial fibrillation, with higher odds only at very high training loads

Exercise goes with better heart health on most of the measures that have been studied. One exception has puzzled cardiologists for years: endurance athletes seem to get more atrial fibrillation, a common rhythm disorder, than people who do no sport at all. A new analysis has tried to find where on the scale of exercise the benefit gives way to that extra risk.

Pooling 37 studies of nearly 1.3 million people, it found that ordinary physical activity went with 22% lower odds of atrial fibrillation. The lowest odds came at a moderate dose. Higher odds appeared only beyond a level the authors equate to 6 to 7 hours of endurance training a week. That threshold comes from studies that watched people, not from experiments, and the studies differed widely from one another.

Why exercise and atrial fibrillation is debated

Atrial fibrillation is an irregular and often fast heartbeat that starts in the upper chambers of the heart. One research group that pooled earlier studies described it this way: atrial fibrillation (AF) is the most common arrhythmia worldwide and is associated with increased morbidity and mortality and reduced quality of life. An arrhythmia is any disturbance of the heart’s rhythm. Morbidity means illness.

For most heart conditions, more exercise goes with less disease. For this one the record has been mixed. A 2021 analysis in Heart Rhythm opened by saying that the place of physical activity in atrial fibrillation (AF) risk is less well established than its place in preventing other heart disease.

The new paper frames the puzzle directly. Regular activity may lower the risk in the general population, it says, but individuals engaging in the highest levels of PA may exhibit a paradoxical rise in AF risk. PA is the authors’ shorthand for physical activity. In an earlier conference version of the same work, the team noted that this pattern has a name, termed the extreme exercise hypothesis.

What nobody had pinned down was the turning point. The point at which the dose-response curve reverses remains unclear, the paper says. A dose-response curve is a line showing how an outcome changes as the amount of something, here exercise, goes up.

One reason was the way earlier work had been organized. Previous meta-analyses have typically examined PA within the general population and athletes separately, the conference abstract observed. A meta-analysis is a study that pools the results of many earlier studies. With recreational exercisers in one pile and athletes in another, there was no single scale on which to find a crossing point.

How the atrial fibrillation analysis worked

The analysis was carried out by a team with C. L. Davies as first author and G. P. McCann as last author, and published on October 9 in the European Journal of Preventive Cardiology.

Four databases were searched (from January 1990 to June 2025). To be included, a study had to have reported PA exposure compared against a sedentary control group and assessed AF incidence or prevalence. Sedentary means inactive. Incidence counts new cases over time, and prevalence counts everyone who has a condition at one moment.

In the end 37 studies were included, covering 1,298,291 people. Twenty-two were studies of physical activity in the general population. Fifteen compared athletes with non-athletes.

To put a weekend walker and a marathon runner on one scale, the team converted every study’s measure of activity into a single unit: MET-hours per week. MET stands for metabolic equivalent of task. One MET is the energy a person uses sitting quietly, so an activity rated at 8 METs burns eight times that. An hour of it counts as 8 MET-hours. The conference abstract records that the conversion was done using the 2024 Compendium of Physical Activities, a standard reference table.

Two kinds of result came out. The first was a simple comparison of active with inactive people, and of athletes with non-athletes, expressed as odds. Odds are the chance of something happening set against the chance of it not happening. The second was the dose-response curve itself, fitted across the 26 studies, of 1,022,790 people, that gave enough detail on dose.

What the analysis found about activity and atrial fibrillation

In the general population, physical activity went with 22% lower odds of atrial fibrillation than being sedentary. The range around that estimate ran from 11% lower to 32% lower.

Among athletes the direction flipped. Athlete status was associated with an increase in the odds of AF. Athletes had 2.74 times the odds of non-athletes, with a range from 1.72 to 4.39. That range is wide, a sign that the athlete studies disagreed with one another or were small.

The dose-response curve joined the two results into one shape. A weekly total of 26 MET-hours was associated with the greatest reduction in odds of AF. Past that point the benefit shrank. Beyond 78 MET-hours a week the odds were higher than in sedentary people, which the authors call the equivalent of 6-7 hours per week of endurance training.

The dose with the lowest odds is one third of the dose beyond which odds exceeded those of sedentary people. The authors summarize the pattern as U-shaped, with odds high at both ends and lowest in the middle. Their conclusion is hedged. Beyond the threshold, they write, the protective effect of exercise may reverse.

How firm is the number 78? The conference version of the analysis, published in June, had put the threshold at 73 MET-hours and the best range at 20 to 37. It described the threshold then as the equivalent of 6-7 hours per week of high-intensity sport. The published figures moved a little. A curve fitted to several dozen studies gives a region where the line turns, and small changes in method shift it.

How earlier research on activity and atrial fibrillation compares

The lower-odds half of the curve matches earlier pooled work. The 2021 Heart Rhythm analysis combined 15 studies of 1,464,539 people. People who met the guideline level of activity, which it defined as 450 MET-minutes a week, had a 6% lower risk, with a range from 3% to 10%. That level is 7.5 MET-hours a week.

Its dose-response result stopped short of the new one. PA levels up to 1900 MET-minutes per week were associated with a lower risk of AF, with less certainty beyond that level. That is about 32 MET-hours a week, close to where the new analysis puts the greatest benefit. Above it the authors could say only that at 2000 MET-minutes per week and beyond, the benefit is less clear. The new analysis added studies of athletes to extend the curve.

On these figures, the level beyond which odds were raised in the new analysis is about ten times the guideline level.

Individual cohorts have traced the same bend. In the Tromsø Study in Norway, which followed 20,484 adults for 20 years, moderately active individuals had a 19% lower risk of any AF than the least active. The most vigorous exercisers did not share the benefit. The authors concluded that leisure time physical activity was associated with AF in a J-shaped pattern, and that higher activity levels attenuated the benefits of moderate activity. Attenuated means weakened.

Sex may matter. A pooled analysis of 23 cohorts, with 1,930,725 participants and 45,839 cases, found no overall difference between the most and least active people. Split by sex, it found an increased risk in men and a decreased risk in women. In UK Biobank, a study of 402,406 people aged 40-69 years, with over 2.8 million person-years of follow-up, the lower risk in active people was also more marked in women. Person-years are the years of follow-up added up across all participants.

The athlete result has precedent too. A 2021 analysis in the British Journal of Sports Medicine set out to investigate the incidence of atrial fibrillation (AF) among athletes compared with non-athlete controls and found 2.46 times the odds, with a range from 1.73 to 3.51. It called for future studies to measure risk according to specific exercise dose parameters, including training and competition history. The new analysis is an attempt at that.

A large Swedish study adds perspective on what the extra cases mean. It followed 208,654 people who had completed the Vasaloppet, a long cross-country ski race, and 527,448 matched non-skiers. Female skiers in Vasaloppet had a lower incidence of atrial fibrillation than did female nonskiers. Male skiers had a similar incidence to that of nonskiers. Within the skiers, those with the highest number of races or fastest finishing times had the highest incidence.

The same study tracked strokes, the main danger of atrial fibrillation. Skiers of either sex had a lower incidence of stroke than did nonskiers, by about a third. Skiers who developed atrial fibrillation had more strokes than skiers who did not.

What may explain the link between heavy training and atrial fibrillation

The new paper is about dose, not cause. Others have proposed explanations, with caution. A 2018 review began its account with the words although the mechanisms are not well understood. It pointed to exercise-induced changes in autonomic tone alongside the development of an arrhythmogenic atrial substrate. Autonomic tone is the balance of the nerves that speed and slow the heart without conscious control. An arrhythmogenic substrate means heart tissue that has changed in a way that makes rhythm problems easier to start.

The 23-cohort analysis gave a longer list of suggested mechanisms: increased vagal tone, bradycardia, increased volume load and stretch, atrial fibrosis and inflammation. The vagus nerve slows the heart, bradycardia is a slow heart rate, and fibrosis is scarring. The idea is that years of heavy training load and stretch the upper chambers and slow the resting heart rate. These are proposals, and none was tested in the new analysis.

What the atrial fibrillation analysis cannot show

Association, not cause. Every study in the analysis observed people as they were. Active people differ from sedentary people in weight, blood pressure, smoking and much else. The analysis cannot separate the effect of exercise from those differences.

Uneven studies. The conference version of the analysis was plain about this. Heterogeneity was high, and athlete studies were of poorer quality than general population PA studies. Heterogeneity means the studies disagreed with each other more than chance would explain. The authors wrote then that higher quality studies, particularly in athletes, are required to confirm this finding.

Self-reported exercise. Most large studies in this field ask people how much they exercise. The 23-cohort analysis noted of its own data that all studies assessed physical activity through self-reported questionnaires. Converting such answers into MET-hours adds further approximation.

A threshold from a fitted curve. The figure of 78 MET-hours is where a statistical curve crosses a line. It moved between the conference version and the published paper.

Prevalence as well as incidence. The analysis accepted studies that counted existing cases at one moment as well as studies of new cases. The first kind cannot show whether the exercise came before the rhythm problem.

No absolute risks. The pooled results are odds relative to sedentary people. They do not say how many athletes in every thousand develop atrial fibrillation, or how that compares with the number who avoid other heart disease through training.

Whether a given training load is a concern for a particular person’s heart rhythm is assessed by their doctor.

Across 26 studies of about 1 million people, the odds of atrial fibrillation were lowest in the moderately active and higher only at very high training loads.