A cohort study published online November 2, 2026 in Circulation: Population Health and Outcomes followed 10,362 US adults in the National Institutes of Health All of Us Research Program whose Fitbit data had been linked to their electronic health records. Over a median of 3.7 years, how much a person's sleep length swung from one night to the next was associated with new cardiovascular disease. The average number of hours they slept was not.

The researchers built 11 sleep metrics from a 180-day baseline period that required at least 90 valid nights of wearable data, covering average duration, night-to-night variability in duration, and variability in bedtime, wake time and the midpoint of sleep. They then counted diagnoses recorded after a further 180-day washout, so that illness already under way was less likely to be driving the sleep pattern. The cohort had a median age of 53 and was 70 percent female.

What the numbers showed

For every one standard deviation more night-to-night variability in sleep duration, about 29 minutes, the hazard of a composite of heart attack, heart failure, cardiomyopathy or ischemic stroke was 24 percent higher (hazard ratio 1.24, 95 percent confidence interval 1.09 to 1.41). There were 230 such events. Variability in bedtime (1.23), wake time (1.20) and sleep midpoint (1.21) carried similar associations. Mean sleep duration did not: its hazard ratio was 0.96, with an interval that crossed one. Duration variability was also associated with heart failure (1.32), hypertension (1.16) and type 2 diabetes (1.20).

In an exploratory analysis of 7,199 participants and just over one million nights of sleep-stage data, a higher count of wearable-estimated sleep cycles per night was associated with lower risk of atrial arrhythmia (0.72, and 0.58 after adjusting for duration, regularity and timing). The authors describe that finding as new and in need of confirmation. The study is observational, so it shows association rather than cause, and people who wear a Fitbit and join a research program are not a cross-section of the country. The abstract available to us did not state a funding source.

A second cohort, same direction

A paper in Frontiers in Medicine, published September 29, 2026, asked a similar question of 7,604 US adults in the 2011 to 2014 National Health and Nutrition Examination Survey who wore a wrist accelerometer for four to seven days and were then followed through the National Death Index for a median of 81 months. There were 714 deaths, 229 of them cardiovascular. For each additional hour in the night-to-night standard deviation of the sleep midpoint, all-cause mortality was 22 percent higher (1.22, 1.12 to 1.33). A 10-point lower score on the sleep regularity index was associated with 29 percent higher all-cause mortality. Adding each person's average sleep duration to the model changed those estimates little, and average duration itself was not linearly associated with either outcome. The authors reported no funding and name reverse causation, residual confounding from shift work, sleep apnea and depression, the short recording window, and the small number of cardiovascular deaths as limits.

Hours still count

Regularity has not displaced duration in the evidence. A UK Biobank cohort study published November 1, 2026 in the journal Sleep followed 69,153 to 76,811 adults aged 40 to 69 for an average of eight years after a seven-day accelerometer recording. Risk followed an inverse J-shaped curve for most outcomes. Compared with a five-hour reference, a device-measured sleep period of seven hours carried a 30 percent lower hazard of death from any cause (0.70, 0.61 to 0.79). The lowest-risk duration was about 7.2 hours by self-report and about 7.7 hours by device, and the authors conclude that seven to nine hours of nightly sleep remains the range to aim for whichever way sleep is measured. The work was funded by UK Biobank and an Australian National Health and Medical Research Council grant.

Where the guideline stands

The American Academy of Sleep Medicine and the Sleep Research Society issued their joint consensus statement in the journal Sleep in June 2015. A panel of 15 sleep experts reviewed 5,314 articles across nine health areas using a modified RAND appropriateness method over 12 months. The recommendation is that adults aged 18 to 60 sleep seven or more hours a night on a regular basis. The statement links habitual sleep under seven hours with weight gain, diabetes, hypertension, heart disease, stroke, depression and higher mortality, and says it is uncertain whether regularly sleeping more than nine hours carries risk for most adults. Notably, the regularity wording is already in the recommendation.

Read together, the three 2026 cohorts and the 2015 statement do not contradict each other. Two found that an irregular schedule tracked with worse outcomes after accounting for hours; one found that hours tracked with mortality on their own. None was a trial, so none shows that fixing a schedule or adding an hour changes what happens next. For a reader whose wearable reports a sleep regularity score or a night-to-night range, the useful next step is to bring that number, and the week-to-week pattern behind it, to a clinician rather than to read it alone.

This article reports research and expert guidance. It is not medical advice. Talk to your clinician before changing medication, treatment or a medical diet.

Sources