BACKGROUND: Hypnotics are widely used by the elderly, and their impact on mortality remains controversial. The inconsistent findings could be due to methodological limitations, notably the lack of control for underlying sleep symptoms or illness associated with hypnotic use, for example, insomnia symptoms and excessive daytime sleepiness, depression and anxiety. Our objective was to examine the association between the use of hypnotics and mortality risk in a large cohort of community-dwelling elderly, taking into account a wide range of potential competing risks including sociodemographic characteristics, lifestyle, and chronic disorders as well as underlying psychiatric disorders and sleep complaints. METHODS: Analyses were carried out on 6,696 participants aged 65 years or older randomly recruited from three French cities and free of dementia at baseline. Adjusted Cox proportional hazards models with delayed entry, and age of the participants as the time scale, were used to determine the association between hypnotic use and 12-year survival. RESULTS: At baseline, 21.7% of the participants regularly used at least one hypnotic. During follow-up, 1,307 persons died, 480 from cancer and 344 from cardiovascular disease. Analyses adjusted for study center, age and gender showed a significantly greater risk of all-cause and cardiovascular-related mortality with hypnotics, particularly benzodiazepines, and this increased with the number of hypnotics used. None of these associations were significant in models adjusting for sociodemographic and lifestyle characteristics, chronic disorders including cardiovascular pathologies, sleep and psychiatric disorders. Results remained unchanged when duration of past hypnotic intake or persistent versus intermittent use during follow-up were taken into account. CONCLUSIONS: When controlling for a large range of potential confounders, the risk of mortality was not significantly associated with hypnotic use regardless of the type and duration. Underlying psychiatric disorders appear to be the principal confounders of the observed association.