AHA Scientific Sessions 2026 Abstract · Nov 2, 2026
Association of Low-Sleep Nights with Imminent Myocardial Infarction and Stroke: Evidence from the All of Us Program Wearable Data
By Sadek M, Sadek C, Sui Y et al.
Background: Chronic sleep deprivation and sleep irregularity are associated with increased risk of major adverse cardiovascular events (MACEs). However, it remains unclear whether sleep patterns change during the acute period preceding MACEs. Research question: Do individuals exhibit changes in wearable-derived sleep measures during the acute period preceding myocardial infarction (MI) or stroke? Methods: A retrospective analysis of the All of Us Research Program was conducted. Participants with >30 nights of wearable-derived sleep data and a first-time incident MI/stroke occurring during the wearable monitoring period were included. Nightly sleep duration was restricted to values between 120-720 min. For each participant, sleep metrics were analyzed during pre-event windows of 3, 5, 7, 14, and 30 days and compared with duration-matched historical windows. The sleep measures analyzed were average sleep duration, sleep variability (mean absolute successive difference), and the proportion of low-sleep nights (defined as sleep duration >1.5 SD below individual’s mean). Paired t-tests and logistic regression were used to assess differences between pre-event and historical windows. Results: 257 participants with incident MI (n=134) and stroke (n=128) contributed 168,481 pre-event nights of wearable data. Participants had a mean sleep duration of 7.19 ± 1.09 hours. Mean sleep duration and variability did not differ between pre-event and historical windows for both MI and stroke across all window durations (p>0.05). However, low-sleep nights comprised a greater proportion of pre-event windows relative to historical windows for both MI and stroke. This was most prominent for the 5-day pre-MI window, where low-sleep nights accounted for 11.2% vs 6.3% of historical window nights (p=0.019), and the 14-day pre-stroke window, where low-sleep nights accounted for 11.4% vs 6.8% of historical window nights (p=0.014). In matched analysis, low-sleep nights (z < -1.5) were more likely than normal-sleep nights (-0.5 < z < 0.5) to occur in pre-event windows for both MI and stroke, with the strongest association within 5 days for MI (OR 1.91; 95% CI, 1.11-3.28) and 3 days for stroke (OR 2.34; 95% CI, 1.07-5.11), progressively attenuating with increasing window duration (Figure). Conclusion: MI and stroke are preceded by increased occurrence of low-sleep nights, which suggests that acute sleep reduction may be associated with imminent MACEs. Hence, wearable-derived sleep data has the potential to identify those at heightened risk for impending MACEs.