How do different sleep duration cutoffs change the consistency of reported cardiovascular risk estimates?

The cutoff decides whether short sleep, long sleep, or both look risky

A sleep duration cutoff is simply the boundary used to sort people into groups, such as 'under 7 hours' versus '7 to 9 hours.' Where that boundary falls determines which people are counted as short sleepers and which as long sleepers, and therefore which group carries the elevated risk estimate. In a Swedish twin study of 12,268 adults followed for up to 18 years, using 7 to 9 hours as the reference group produced a U-shaped pattern: both under 7 hours and 10 or more hours were associated with higher cardiovascular disease risk, with hazard ratios of 1.14 and 1.10 respectively [2]. That result depends on treating 7 to 9 hours as the safe middle and placing the long-sleep boundary at 10 hours rather than, say, 9 hours.

A different study using different boundaries can therefore report a different-looking answer without contradicting the first. In 782 older Chinese adults without cardiovascular disease, sleep was split into four groups: under 7, 7 to 8, over 8 to 9, and over 9 hours. Only the over-9-hour group showed a significant positive association with cardiovascular risk compared with the 7-to-8-hour reference, and the short-sleep group did not emerge as elevated in that analysis [5]. Read side by side, the two studies do not necessarily disagree about biology; they placed the reference interval and the long-sleep boundary differently, so the elevated-risk group is drawn differently.

The reference group matters as much as the cutoff itself

Risk estimates are always relative to something, and in sleep research that something is usually a 'normal' sleep band. When the reference band is 7 to 9 hours, as in the twin study, both tails of the distribution can look risky [2]. When the reference band is narrower, such as 7 to 8 hours, the comparison shifts and the long-sleep group may be the one that stands out [5]. This is why two studies can both be internally valid yet appear to point at different sleep durations as the problem.

The cutoff also interacts with how sleep is measured. The twin study relied on self-reported sleep duration at baseline [2], while a UK Biobank analysis of 86,219 participants used seven days of accelerometer-measured sleep and focused on night-to-night irregularity rather than a single duration cutoff [4]. That study found that each one-hour increase in the standard deviation of sleep duration was associated with a 19% higher risk of major cardiovascular events, and the association held across heart attack and stroke subtypes [4]. This is a different exposure — variability rather than average length — so it does not directly test cutoff placement, but it shows that the sleep measure itself changes what the risk estimate describes.

Cutoffs chosen to flag risk factors can point to short sleep, not long sleep

Some studies choose cutoffs not to predict heart attacks but to identify people with cardiovascular risk factors such as obesity, high blood pressure, diabetes, and abnormal cholesterol. In a Korean national survey of 18,677 adults, researchers used statistical cutoffs to find the sleep duration that best flagged these risk factors, and the optimal sleep durations clustered around 6 hours [1]. That is a short-sleep signal, which sits awkwardly beside the long-sleep signal seen in the twin and older-adult studies [2][5].

The difference is partly about the outcome. Flagging who currently has risk factors is not the same as predicting who will later have a cardiovascular event, and a cutoff tuned to one question need not be optimal for the other. The Korean study also found that optimal bedtime ranges differed by sex and age, which reinforces the point that a single universal cutoff is unlikely to perform equally well across groups [1].

Adjusting for other factors can shrink the estimate, and the cutoff cannot fix that

Cutoff choice is not the only reason estimates vary. In the EVasCu study of 296 healthy adults, sleep was grouped into under 6, 6 to 8, and over 8 hours and compared against cardiovascular risk markers including arterial stiffness and skin autofluorescence. Simple analyses showed inverse associations between sleep duration and several markers, but after adjusting for covariates most of those associations lost statistical significance; only skin autofluorescence remained significant, with a difference of 0.223 arbitrary units between the under-6-hour and over-8-hour groups [3]. The authors suggest insufficient sleep may increase cardiovascular risk, but they frame this as a suggestion requiring prospective study rather than a settled finding [3].

This matters for the cutoff question because it shows that a risk estimate can change substantially depending on what else is accounted for, independent of where the sleep boundary is drawn. A cutoff that looks informative in an unadjusted comparison may look weak once age, sex, and health behaviors are included. The practical implication is that reported cardiovascular risk estimates for sleep duration should be read together with the reference group, the outcome being measured, and the adjustment set — not as a single number that transfers cleanly from one study to the next.

Sources used in this answer

Related research questions