Reading the evidence
Prospective cohort studies follow people over years and record who develops disease, which allows the exposure to precede the outcome. This temporal order rules out the simplest form of reverse causation, where disease causes short sleep. Adjustment for standard risk factors addresses confounding. What remains is an association that is consistent across populations, shows a dose-response gradient, and is supported by a plausible physiological mechanism — the three criteria most often used to judge whether an association is likely causal.
Shift work as a natural experiment
Shift workers rotate or work nights, so their sleep is both shortened and misaligned with the circadian clock. Prospective studies report higher rates of hypertension, heart attack, and stroke in this group, and risk increases with the number of years spent on shift work. The dose-response with exposure duration is important: if the association were driven purely by who chooses shift work, risk would not rise steadily with years on the job.
The U-shaped mortality curve
Plotting sleep duration against all-cause mortality produces a U-shaped curve, with the lowest risk near seven hours. Short sleep shows a consistent elevation. Long sleep also shows elevated risk, but this arm is harder to interpret because long sleep is often a symptom of existing illness rather than a cause. Treat the long-sleep arm as a marker, not a recommendation to sleep less.