One cohort, different modelling questions
Teraoka et al., 2024. All comparisons below use current smokers as the reference. Confidence intervals describe estimate uncertainty, not an individual's range of outcomes.
| Comparison and model | Adjusted HR (95% CI) | What the result supports |
|---|---|---|
| Former smokers at entry; main model | 0.87 (0.83–0.91) | Lower adjusted new-diagnosis rate in that comparison |
| Identified as quitting during follow-up; main model | 0.82 (0.70–0.95) | The source of the relative ‘18% lower’ headline |
| Quitting during follow-up; time-updated sensitivity analysis | 0.79 (0.58–1.07) | An uncertain estimate whose interval includes no difference |
Who was compared—and what was counted?
AF is an irregular heart rhythm, sometimes without noticeable symptoms. Teraoka and colleagues studied 146,772 participants with a smoking history but no recorded AF at baseline. They distinguished former smokers at entry, current smokers, and people identified as quitting during follow-up. Average follow-up was 12.7 years; the outcome was a new recorded AF diagnosis, not day-to-day palpitations, symptom improvement or fewer episodes in people already diagnosed.
The main adjusted hazard ratio (HR) was 0.87 (95% confidence interval 0.83–0.91) for former smokers at entry and 0.82 (0.70–0.95) for those identified as quitting during follow-up, each versus current smokers. An HR compares event rates over time among people not yet diagnosed; 1 means no difference. It is not the probability that a particular person will develop AF, nor an 18-percentage-point reduction.
The two quitting groups were not directly compared with each other. A smaller HR for the follow-up quitting group does not prove that quitting recently works better than having stopped earlier.
Why counting diagnoses before adjustment gives a different impression
Former smokers at entry had a higher crude AF rate than current smokers: 6.46 versus 5.06 cases per 1,000 person-years. A person-year combines the number of people and the time observed; it is not a one-year forecast for each participant. The former-smoker group was older, and age strongly affects AF incidence. The main model's lower adjusted HR therefore does not contradict those raw counts.
The main model used fixed groups from baseline: people later identified as quitting were timed from study entry, so their group also included observation before they actually quit. The updated model kept them in the current-smoking group until their self-reported quitting age, then moved them to the quitting group; diagnoses counted in the group they belonged to at that time. Precision about timing matters too. In the time-updated sensitivity analysis, the HR for quitting during follow-up was 0.79 (0.58–1.07). That interval includes 1, so this analysis did not clearly distinguish lower incidence from no difference. The quitting group was small and post-cessation observation limited. This neither proves quitting has no value nor allows the uncertain result to be advertised as conclusive prevention.
Keep a prevention question separate from a symptom or treatment question
Smoking status was not randomly assigned. Self-report, differences in health and behaviour, variable questionnaire timing and diagnoses missed by routine records limit causal interpretation. The study does not provide an exact ‘after quitting for X days’ threshold, and its UK cohort is not a calibrated prediction for every age or country.
If palpitations persist or worsen, seek clinical assessment rather than labelling them as withdrawal or diagnosing AF yourself. For someone already diagnosed with AF, these new-diagnosis figures cannot decide treatment or changes to prescribed medicines. In England, NHS local stop-smoking services offer a separate support route; ask how it can fit alongside your existing care without submitting symptoms or a smoking history here.
What to keep in mind
Sources
The central claims on this page were checked against the sources below.
- JACC: Clinical Electrophysiology: Teraoka et al. (2024): Smoking Cessation and Incident Atrial Fibrillation in a Longitudinal Cohort; original Methods, Table 2 and time-updated sensitivity analysis
Sources checked: 2026-10-08
- NHS: Atrial fibrillation: definition and clinical-assessment boundary
Sources checked: 2026-10-08
- NHS: Atrial fibrillation
Sources checked: 2026-10-08
- NHS: England: find local stop-smoking support
Sources checked: 2026-10-08
Population-level interpretation of incident AF research only; no diagnosis, personal prediction, symptom classification, treatment or medication selection.