Sleep-disordered breathing is much more prevalent in elderly people than in middle-aged or young populations, but its clinical significance in this age group is unclear. This study investigated retrospectively the rates of all-cause mortality in elderly people (>or= 65 years) with a laboratory diagnosis of sleep apnoea, and compared their rates of mortality with that of age-, gender- and ethnicity-matched national mortality data. Survival of 611 elderly people was ascertained after a follow-up of 5.17 +/- 1.13 years. Their age was 70.4 +/- 4.8 years, body mass index 30.4 +/- 5.9 kg m(-2) and respiratory disturbance index (RDI) 28.9 +/- 20.1 events h(-1). Seventy-five (12.27%) patients died during the follow-up period. In comparison with the demographically matched cohort from the general population, the standardized mortality rate of the sleep laboratory cohort was 0.67 [95% confidence interval (CI): 0.53-0.88; chi(2) = 11.69, P < 0.0006]. When calculated separately for patients with RDI < 20 (no/mild apnoea), RDI 20-40 (moderate apnoea) and RDI > 40 events h(-1) (severe apnoea) there was a significant survival advantage for the moderate group with a standardized mortality rate of 0.42 (P < 0.0002), while elderly people with no/mild apnoea tended to have lower mortality and those with severe sleep apnoea had the same mortality as the matched population cohorts. Cox regression analysis revealed that sleep latency and comorbidities but not sleep apnoea severity were associated independently with mortality. The survival advantage of elderly people with moderate sleep apnoea, combined with recent findings on the potential cardioprotective effects of chronic intermittent hypoxia, raise the possibility that apnoeas during sleep may activate adaptive pathways in the elderly.