REF-1577 · Newman et al., 2021 · British Journal of Sports Medicine · Paper · PubMed

Do athletes have higher odds of atrial fibrillation?

Pooled case-control and cohort data: OR 2.46 (1.73-3.51) vs non-athletes, higher in mixed than in endurance sport; no absolute risk or dose read. Association only. Palpitations, chest pain, fainting: a doctor.

Main caveat: Abstract only. No full text is held; see the Protocol list for what is missing.
limited evidenceMeasured on a mixed group

The search covered PubMed, MEDLINE, Science Direct, SPORTDiscus and Cochrane, for articles published after 1990 and before 2 December 2020 that reported AF case numbers in athletes with non-athlete controls and allowed an OR. Sport mode and age were covariates set in advance.

The authors call for future studies by exercise dose, including training and competition history, because this pool could not address it. Our reading: this establishes that a signal exists in the published comparisons, not its size for any one athlete.

Protocol

Eligibility: case-control or cohort; athlete vs non-athlete control groups; enough data to compute an OR.
Analysis: meta-analysis with heterogeneity analysis, then meta-regression.
Not in the abstract: study count and participant numbers, definition of athlete and of AF ascertainment, whether ORs are adjusted, risk-of-bias method, heterogeneity value, sex split.
Funding and interests: not in what was read.

Key Results

ComparisonResultP
Athletes vs non-athlete controlsOR 2.46 (95% CI 1.73-3.51), Z = 4.97<0.001
Sport mode and AF, meta-regressionB = 0.12590.0193
Mixed vs endurance sportB = -0.54760.0204
Younger (<55 years) vs older (≥55 years)B = -0.02293<0.001
Reading the coefficients. The abstract gives their coding nowhere, so the size of the sport and age effects cannot be turned into odds here. It also calls a regression coefficient a moderate correlation.
OR is not risk. An odds ratio of 2.46 says nothing about the absolute rate, which the abstract does not give.
The claim about younger athletes. The abstract says younger athletes were more likely to develop AF than older ones, yet this is study-level meta-regression with no absolute rates, so it may reflect a larger contrast with same-age controls rather than a higher rate in a younger athlete (inference).

Mechanisms

Measured: none.
Authors' interpretation: none offered in the abstract beyond calling for dose-specific data.
Inference: athletes may be monitored and may notice symptoms more, which would raise detected AF. Case-control designs add recall and selection bias. Nothing here tests either way, and no mechanism is shown.

Practical Application

Do athletes have higher odds of atrial fibrillation? Pooled case-control and cohort data: OR 2.46 (1.73-3.51) vs non-athletes, higher in mixed than in endurance sport; no absolute risk or dose read. Association only. Palpitations, chest pain, fainting: a doctor.
Good for: taking symptoms seriously in a fit person instead of explaining them away by fitness (Inference).
Not good for: telling an athlete to cut training. No dose, threshold or reversal is shown.
Not good for: quoting 2.46 as an individual's risk, or as a causal effect.
Not from this paper; the base's own practice: palpitations, unexplained fatigue or a sudden loss of exercise capacity go to a doctor and an ECG rather than to a lighter week (SYNTH-35). This entry is not a diagnosis tool.

Limitations

Abstract only. No full text is held; see the Protocol list for what is missing.
Case-control and cohort pooled. Case-control studies of a disease are prone to selection and recall bias.
Controls could be active, which would shrink the contrast, or inactive, which would not; their split is not given.
Detection bias is possible in monitored athletes (inference).
Meta-regression on study-level age and sport can mislead about individuals (ecological bias, inference).
No dose, no absolute rates, no outcomes such as stroke or death in the abstract.
How this entry was checked and changed
Why this level

(from the abstract) — a meta-analysis of observational studies is capped at B+, not A. One step down because only the abstract was read, case-control studies are mixed in, and the number of studies, participants, heterogeneity and adjustment are all absent