Nine studies were included; three retrospective population-based registry studies (n=91,803) and six prospective cohort studies (n=1,120,411). Problems with these studies included limitations in the accurate certification of cause of death, lack of differentiation between specific benzodiazepines, the methods used to select patients, the definition of users and non-users, and confounding by polydrug use and other factors.
Registry studies: Studies scored between eight and 12 out of 18 points for validity. One study of calls made to a poisons advice centre reported a significantly increased risk of death from benzodiazepine poisoning compared with other outcomes in older (60 years or more) compared with younger (under 60 years) patients (RR 7.1, 95% CI 3.2 to 15.5). One study of driver fatalities in on-road motor vehicle accidents reported no significant association between benzodiazepines and fatalities that were the responsibility of the driver, after adjusting for other drugs. One study reported that age-standardised mortality per million men, with benzodiazepines mentioned, in overdose deaths and poisoning databases, ranged from 7.1 (in 1993) to 6.6 (in 1998).
Cohort studies: Studies scored between six and 10 out of 18 points for validity. The results were mixed with three studies reporting no significant association between benzodiazepine or hypnotic drug use and mortality. One study reported a significant increase in mortality associated with the daily use of benzodiazepines in women aged 40 to 42 years at baseline, after adjusting for confounding factors (RR 1.7, 95% CI 1.1 to 2.6). There was a non significant increase in men. One study reported a significant relationship between the regular use of non-prescribed benzodiazepines and fatal overdose among those who misused drugs, after adjusting for the use of other substances (OR 2.86, 95% CI 1.32 to 6.16). One study reported a significantly increased standardised mortality ratio of 1.41, within six years for women using at least 30 sleeping pills per month after adjusting for 14 variables (p<0.001). There was no significant increase among men. It reported no significant increase in mortality when diazepam and chlordiazepoxide were analysed separately.
The results from a seventh study that did not meet the inclusion criteria were also reported.