This secondary analysis of the Systolic Blood Pressure Intervention Trial (SPRINT) examined whether changes in sleep disturbance over time predict incident atrial fibrillation (AF) and, conversely, whether AF predicts subsequent clinically significant sleep disturbance within the same study population. The analysis focused on adults with hypertension and increased cardiovascular risk who did not have diabetes at enrollment.
Two prospective analytic cohorts were defined for reciprocal analyses. For the analysis of sleep disturbance predicting incident AF, 7,232 participants without AF at baseline were included. For the reciprocal analysis examining AF predicting incident clinically significant sleep disturbance, 6,079 participants without clinically significant sleep disturbance at baseline were included. The source is a secondary analysis of SPRINT; details about additional inclusion or exclusion beyond these numbers were not reported in the source.
Sleep disturbance was assessed repeatedly during follow-up using a patient-reported symptom score ranging from 0 to 3. Clinically significant sleep disturbance was prespecified as a score of ≥2. AF was ascertained from protocol 12-lead electrocardiograms collected as part of SPRINT. The source did not report additional objective sleep measures (for example, polysomnography or actigraphy) or symptom item wording beyond the 0–3 score system.
Time-dependent Cox proportional hazards models were used to evaluate both directions of association. Models treated sleep disturbance and AF as time-updated exposures: time-updated sleep disturbance in relation to incident AF, and time-updated AF in relation to incident clinically significant sleep disturbance. Models adjusted for demographic, lifestyle, and cardiovascular risk factors; the specific covariates included in adjustment were not itemized in the source text.
During a median follow-up of 3.3 years, 189 incident AF events occurred among the cohort without baseline AF. Time-updated clinically significant sleep disturbance (score ≥2) was associated with a 78% higher adjusted risk of incident AF (hazard ratio [HR] 1.78; 95% confidence interval [CI], 1.27–2.51).
When analyzed per point of the sleep score, each 1-point increase in the time-updated sleep score was associated with a 26% higher risk of AF (HR 1.26; 95% CI, 1.09–1.46). Compared with a sleep score of 0, the adjusted HRs for incident AF were reported as:
These findings indicate a graded association between higher time-updated sleep symptom scores and risk of developing AF in this hypertensive population.
In the reciprocal analysis, during a median follow-up of 3.2 years there were 1,422 incident clinically significant sleep-disturbance events among participants without clinically significant sleep disturbance at baseline. Time-updated AF was associated with a 59% higher adjusted risk of incident clinically significant sleep disturbance (HR 1.59; 95% CI, 1.16–2.16).
This result indicates that the onset or presence of AF during follow-up was associated with subsequent increases in clinically meaningful sleep symptoms in the same cohort population.
The authors reported that associations in both temporal directions were consistent across prespecified subgroups, and no significant effect modification was identified. The source did not provide subgroup-specific point estimates or list the prespecified subgroup definitions in detail.
Among adults with hypertension and elevated cardiovascular risk (without diabetes), longitudinally assessed sleep disturbance and AF were prospectively associated in both directions. The data support a bidirectional relationship: worsening or clinically significant sleep symptoms predicted higher risk of incident AF, and AF predicted higher risk of subsequent clinically significant sleep disturbance.
These findings highlight the potential clinical relevance of routine assessment of sleep health in patients with or at risk for AF and suggest that monitoring and addressing sleep symptoms may be important in comprehensive cardiovascular care. The source notes these relations were observed after adjustment for multiple covariates, although the exact covariate list was not reported in the article text provided.
Strengths of this analysis include use of repeated, time-updated measures of sleep disturbance and AF ascertainment via protocol 12-lead ECGs within a well-characterized randomized trial cohort. Limitations inherent to the report include reliance on a patient-reported 0–3 sleep symptom score rather than objective sleep testing, lack of detail in the source about the full set of adjusted covariates and subgroup definitions, and that this is a secondary analysis of SPRINT rather than a primary trial designed to test these hypotheses. The authors stated that SPRINT data used for the analysis are publicly available through the NHLBI BioLINCC repository. The investigators declared no competing interests.