This study examined associations between hypnotic intake and risk of excess mortality (all-causes and specific causes) over a 12-year period in a large elderly cohort, taking into account a wide range of potential confounding factors. As in several previous studies we observed significant associations between hypnotic use, notably BZDs, and mortality; however, these associations became non-significant after adjustment for all potential confounding factors, notably psychiatric disorder. These findings persisted even after taking into account up to 20 years duration of past hypnotic intake or persistent versus intermittent use.
Previous studies have been inconsistent, with some studies observing significant relationships between hypnotic prescriptions and mortality
[16, 20–25, 37] and others not
[16–19]. Our findings suggest that these differences are probably largely due to failure to take into account confounding associations, notably common affective symptoms and sleep complaints, although other factors such as study design, participant age, and class of hypnotics probably also influence study outcome.
Insomnia symptoms often lead to the use of hypnotics, a condition frequently associated with EDS, anxiety, and mood disorders. Depression and anxiety are also risk factors for mortality
[14, 15]. Depressive symptomatology and insomnia are both common in the elderly and in France there are no official guidelines for management, so antidepressants are often used to treat sleep disorder and hypnotics to treat depression, especially where sleep disturbance is one of the presenting symptoms
. EDS is also of multifactorial origin, and commonly associated with depression
, cognitive decline
, physical illness (particularly CVD), and mortality in older adults
[5, 6, 13]. Thus, all these conditions may increase the risk of mortality in elderly patients through pathways independently of hypnotics. However, few previous studies have controlled for psychological status
[17, 18, 22, 23] and in studies where depressive symptoms have been considered, antidepressant use has not been necessarily taken into account. This is important because antidepressant use may relieve depressive symptomatology, but the underlying biological risk factors associated with increased mortality may still be operating. No previous studies have controlled for anxiety or simultaneously for insomnia and EDS symptoms as potential independent confounding factors. To our knowledge, our study is the first one controlling for such a large range of potential confounding factors, especially the underlying diseases associated with hypnotic use, such as anxiety and depressive symptomatology and antidepressant use, as well as EDS and insomnia complaints. Our finding that psychiatric disorder could be a principal determinant driving the association between hypnotics and mortality risk explains previous inconsistencies.
Chronic use of hypnotic drugs, particularly BZD, may be associated with the risk of addiction and insomnia-rebound after withdrawal, psychomotor impairment and cognitive problems, OSAS, EDS, and car accidents
[12, 39, 40]. In our sample, only one participant died from a car accident, and this person did not use hypnotics. We did not find any interaction between individuals clinically at risk for OSAS, hypnotics intake, and mortality, suggesting that if hypnotics trigger or aggravate OSAS they do not impact on mortality risk. The use of BZD may also favor falls and hip fractures and thus increase the risk for disability and death especially in the elderly
[41, 42]. However some studies have suggested that nighttime sleep problems may also be significant risk factors for falls in the elderly, independently of hypnotic use
[43–45]. In our study, the associations between hypnotic use and all-causes or CVD-related death became non-significant after adjustment for health behavior and status variables, plus EDS and insomnia complaints.
An increased incidence risk for cancer was also reported in individuals using hypnotics in some studies
[21, 22, 24], even in infrequent hypnotic users
. Our study did not report any association between hypnotic use and cancer-related death. Again, differences in adjustment of underlying co-morbid conditions frequently associated with the chronic use of hypnotics appear to explain previous findings.
The present study has some limitations. Unfortunately, data related to hypnotic dose were not available. Bias could have been introduced by the low participation rate at baseline and the non-random exclusion of participants with missing data at baseline – these participants were older, were more commonly hypnotics users, and more often had psychiatric and other chronic disorders that may limit the generalizability of our findings. Although unlikely, the possibility of overadjusment can not be excluded: potential confounding variables should be intermediate variables in the causal pathway between hypnotics intake and mortality. Finally, the absence of significant association between the use of hypnotics and mortality (all-causes, and CVD) after adjustments for covariates should be interpreted with caution regarding the small number of events per predictor variable.
Our prospective study based on a large community sample has several strengths, including the duration of the follow-up and adjustment for a wide range of possible confounding factors including sociodemographic and lifestyle factors, chronic disorders, and sleep complaints as well as depression and anxiety disorders that were found as key confounding factors in this study. Prescriptions and medications themselves were checked by the interviewer and the causes of death were established by an independent committee. Finally, excluding participants who died during the first two years of follow-up did not modify the main results, suggesting a modest confounding effect of severe undiagnosed conditions in relation to hypnotic use and death.