METHODS

The subjects were 1763 Japanese men aged 40–74 years who lived in three Japanese communities. They participated in the 2000–2004 annual cardiovascular risk surveys and were recruited for the present sleep study, at a total recruitment rate of 84%. The sensor of the pulse oximeter (PULSOX-3Si, Minolta, Japan) was attached to the index finger during all-night sleep at home. 5 We used a sleep log to exclude waking time from the analysis to minimise potential overestimation of sleep duration. Data from people with total recording time< 4 h, or with artefacts likely due to frequent body movement, inadequate fitting of the probe or excessive pulse pressure (n= 100) were excluded, and data from 1663 men were used for the analyses.