Sleep duration is the most studied single sleep variable because it is the easiest to ask about. Dozens of prospective cohorts have recorded how long people say they sleep, then followed them for years and counted deaths and diseases. Pooled, those cohorts cover more than five million people.

The result is a U-shaped curve: risk rises on both sides of about seven hours. That shape is repeated for all-cause mortality, cardiovascular disease, stroke and diabetes. The two arms of the U mean different things, and this guide tries to keep them apart.

Short sleep has experimental support as a cause of harm: restricting healthy adults to a few hours a night for days raises blood pressure, impairs glucose handling, increases appetite and hunger hormones, and degrades attention and memory in ways the restricted people do not notice. The chronic version of those changes is a plausible route to hypertension, diabetes and cardiovascular disease.

Long sleep has no equivalent experiment. Sleeping nine or ten hours is associated with depression, inflammation, undiagnosed disease, unemployment and low physical activity, all of which independently predict death. The leading interpretation is that long sleep is a signal of something else being wrong rather than a cause, and the meta-analyses that find it say so.

All-cause mortality. A 2010 meta-analysis of 16 prospective studies with 1.38 million participants found short sleep associated with 12% higher risk of death and long sleep with 30% higher risk. A 2017 dose-response meta-analysis found U-shaped associations for mortality, cardiovascular disease, coronary disease and stroke, with the lowest risk at about seven hours; each hour below that was associated with about 6% higher mortality, and each hour above with about 13%.

Short sleep and disease. A 2017 meta-analysis of 153 studies in over five million people found short sleep associated with higher risk of death, diabetes, hypertension, cardiovascular disease, coronary heart disease and obesity, with a dose-response for mortality below six hours.

Long sleep and disease. The companion 2018 meta-analysis of 137 studies found long sleep associated with higher mortality, diabetes, cardiovascular disease, stroke and coronary disease, with the authors explicitly asking whether the relationship is causal and modifiable, which is the right question.

What a cohort cannot show. All of this is self-reported sleep at one point in time, associated with outcomes years later. The causal direction for long sleep is unresolved. For short sleep, the experiments in the sleep restriction guide supply what the cohorts cannot.

There is no intervention here to be unsafe. The relevant risks are of acting on the curve wrongly: forcing time in bed beyond what a person can sleep worsens insomnia, and the long-sleep arm of the U is not evidence that cutting sleep is protective. Persistent long sleep or persistent unrefreshing sleep is a reason to look for a cause, not something to correct by alarm clock.

Sleep is not a product and has no regulator. Public health bodies in most countries publish recommended sleep ranges by age, derived from the cohort evidence summarized here and from expert consensus, and this site links to those rather than restating them.