AHA Scientific Sessions 2026 Abstract · Nov 2, 2026
Wearable Trends Predict Major Adverse Cardiovascular Events: Analysis of the National Institutes of Health All of Us Cohort
By Polsani D, Mata Quinonez LR, Dasi L
Introduction: Physiological and behavioral metrics, such as heart rate, physical activity, and sleep architectures, collected passively from consumer wearable devices may be associated with major adverse cardiovascular events (MACE). Whether consumer wearables add incremental value to the current standards for MACE risk association, like hypertension and diabetes, is unknown. Hypothesis: We assessed the hypothesis that wearable-derived metrics including sleep architecture, heart rate, and physical activity are independently associated with incident MACE after adjusting for covariates including hypertension, diabetes, and beta-blockers. Methods: The cohort included 16,079 participants from the NIH All of Us controlled tier (C2024Q3R9), who had Fitbit sleep, heart rate, and activity data characteristics. MACE had to occur after at least 12 months of Fitbit wear and was defined as incident ICD-10 codes I20-I22 and I25. The metrics were calculated from a 12-month pre-event window for cases and prior to the median Fitbit wear date for controls. Minute-level heart rate metrics data was captured on days with fewer than 5,000 steps per the Tudor-Locke sedentary lifestyle definition (2013). Sleep metrics were derived from main sleep episodes, defined as a boolean by Fitbit. Through complete case analysis 3,588 participants were excluded due to insufficient wearable data. Cox regression with L2 penalization (λ=0.1) was adjusted for age, sex, hypertension, diabetes, and beta-blockers. Results: Among 16,079 participants (mean age 52.2 years, 27.7% male, 14.7% hypertension, 5.0% diabetes), 402 patients developed MACE over a median follow-up of 1,510 days. In adjusted Cox regression (C-index 0.808, clinical only C-index 0.819, wearable only C-index 0.721), wearable-derived REM sleep duration (HR 0.941, 95% CI [0.900, 0.984]) and deep sleep duration (HR 0.951, 95% CI [0.910, 0.994]) were independently associated with MACE. Participants who developed MACE averaged 9.6 fewer minutes of nightly REM sleep and 7.1 fewer minutes of deep sleep compared to controls (71.2 vs 80.8 and 52.9 vs 60.0 minutes). Heart rate metrics and physical activity metrics were not independently associated with MACE. Conclusion: REM and deep sleep are independently associated with incident MACE in a diverse cohort while adjusting for established cardiovascular risk factors, age, and sex. Thus, consumer wearable may be able to complement traditional clinical assessment through capturing sleep-related cardiovascular risk. Notably, a wearable-only model (C-index 0.721) approached the performance of established cardiovascular risk tools without clinical measurements. Findings should be interpreted in the context of consumer-grade sleep staging accuracy and the observational study design.