Mortality associated with anxiolytic and hypnotic drugs—A systematic review and meta-analysis

Abstract

Background:

Use of hypnotics or anxiolytic drugs is common and various studies have reported increased mortality with hypnotics or anxiolytic use.

Objective:

To consolidate the evidence on mortality risk associated with hypnotics or anxiolytic use

Methods:

Major databases were searched through April 2014 for studies reporting mortality risk associated with hypnotics or anxiolytics use. A pooled hazard ratio with 95% confidence interval was estimated using random-effects model.

Results:

After screening 2188 articles, 25 studies (24 cohort, 1 case-control) enrolling 2,350,093 patients with 59% females (age 18–102 years) were included in the meta-analysis. Hypnotics or anxiolytic users had 43% higher risk of mortality than non-users (hazard ratio, 1.43; 95% confidence interval, [1.12, 1.84]). Eight studies reported risk estimates for each gender category and pooled results from these studies showed increased risk of mortality among men (hazard ratio = 1.60, 95% confidence interval = [1.29,1.99]) and women (hazard ratio = 1.68, 95% confidence interval = [1.38, 2.04]). Pooled results from 10 studies showed higher mortality among benzodiazepine users compared to non-users (hazard ratio = 1.60, 95% confidence interval = [1.03, 2.49]), while pooled results from five studies showed an increased risk of mortality with Z-drugs use although the effect could not reach statistical significance (hazard ratio = 1.73, 95% confidence interval = [0.95, 3.16]). Significant heterogeneity was observed in the analyses and the quality of included studies was good.

Conclusion:

This meta-analysis suggests that hypnotics or anxiolytics drugs use is associated with increased mortality and hence should be used with caution. Future studies focused on underlying mechanism of increased mortality with hypnotics or anxiolytics use are required.

Keywords

Hypnotic anxiolytic Z-drugs mortality systematic review meta-analysis

Introduction

Benzodiazepines and hypnotics use is common in the general population and is estimated to range between 3.5% and 11.7% (Chong et al., 2013; Mellinger et al., 1985; Morgan et al., 1988; Ohayon and Caulet, 1996; Ohayon et al., 1998; Quera-Salva et al., 1991). Benzodiazepines are mainly used to treat anxiety, insomnia, alcohol withdrawal, seizure, panic disorder, and also used for conscious sedation and as a muscle relaxant. Long-term use of benzodiazepines and hypnotics increases with age, highest among the elderly, ranging from 7% to 43% (Cheng et al., 2008; Donoghue and Lader, 2010; Olfson et al., 2015; Paulose-Ram et al., 2007). Benzodiazepines and hypnotics use may be associated with daytime sleepiness, depressive symptoms, poor physical functioning, car accidents, increased falls and fractures (Cumming and Le Couteur, 2003; Pariente et al., 2008; van Vliet et al., 2009; Vermeeren, 2004), and raises the concern for abuse and dependency. In addition, benzodiazepines may have deleterious effects on memory and cognition (Curran, 1986), and may be associated with an increased risk of Alzheimer type dementia (Billioti de Gage et al., 2014).

In addition to the above negative effects, numerous studies have evaluated the association of hypnotics and anxiolytics (HA) use with mortality risk. However, inconsistent results have been reported (Belleville, 2010; Hausken et al., 2007; Hays et al., 1996; Kojima et al., 2000; Kripke et al., 1998, 2002, 2012; Mallon et al., 2009; Phillips and Mannino, 2005; Rumble and Morgan, 1992). Whereas few studies found no association (Hays et al., 1996; Jaussent et al., 2013; Kojima et al., 2000; Phillips and Mannino, 2005; Rumble and Morgan, 1992), others have reported an increased mortality risk associated with HA use (Belleville, 2010; Hausken et al., 2007; Kripke et al., 1998, 2002, 2012; Mallon et al., 2009). Thus, the association of HA use with mortality risk has remained inconclusive. Therefore, we performed a systematic review and meta-analysis of all available studies evaluating the mortality risk associated with HA use, to quantify the magnitude of this association and appraise the quality of the supporting evidence for this association.

Methods

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement guidelines were used to report this systematic review and meta-analysis (Moher et al., 2009).

Study eligibility

We searched for randomized control trials (RCT), cohort, case control and cross-sectional studies that provide documentation of benzodiazepines and/or anxiolytics and/or hypnotics use, measured all-cause mortality as a predefined outcome of interest and reported risk estimates or frequency data from which we could calculate the mortality risk associated with HA use. Studies were included irrespective of language or publication status. Data were included from most recent comprehensive report in case of overlapping study samples. Studies reporting selective serotonin reuptake inhibitors, serotonin and norepinephrine reuptake inhibitors and tricyclic antidepressants as anxiolytics were excluded.

Data sources and search/study selection

Major databases including Ovid MEDLINE In-Process & Other Non-Indexed Citations, Ovid MEDLINE, EMBASE, PsycINFO, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews and Scopus were searched through 10 April 2014 by an expert librarian with the input from study team. The complete search strategy is available as online supplement. To further increase the yield of relevant articles, references cited in the potentially eligible articles and conference proceedings of major psychiatry, internal medicine and pharmacology organizations were hand-searched by the study team. This was followed by full article searches of relevant abstracts. All titles and abstracts were screened by two reviewers (A.K.P. and S.S.M.) independently in parallel using a standardized form. The same two reviewers did the full text review of selected articles. Inter-observer agreement on study selection was assessed by Cohen’s κ (Cohen, 1960), and any disagreements were resolved by consensus in the presence of a third reviewer.

Data extraction

Data were extracted independently by two reviewers from selected studies. A predesigned form was used to extract data regarding study characteristics (author, country, study design, study population, number of participants and conclusion), HA exposure and mortality. We collected the most adjusted risk estimates in studies where multiple models for risk estimates were reported. When cohort studies reported both cross-sectional and longitudinal models, we used risk estimates from the longitudinal models.

Quality assessment

Two investigators (A.K.P., S.S.M.) performed the methodological quality assessment of included studies in parallel. Disagreements were resolved by consensus in the presence of a third reviewer. We used the Newcastle-Ottawa Quality Assessment Scale to assess the methodological quality of included studies (Wells et al., n.d.). Each study was assessed for risk of bias in selection, comparability of study groups and ascertainment of the outcome of interest.

Outcome assessment

Our primary outcome was all-cause mortality associated with HA use. The secondary outcomes were sex-stratified mortality risk and mortality associated with benzodiazepines or Z-drugs use only. A priori subgroup analysis based on study location, study population, age, duration of study drug exposure, follow-up, studies adjusted for minimal covariates versus those not adjusted and study quality were planned to explain the potential heterogeneity in the direction and magnitude of the effect estimate among different studies. Studies adjusted for age and medical co-morbidities were considered to be adjusted for minimal covariates.

Data synthesis and analysis

Continuous variables were reported as means with standard deviation (SD) or medians with interquartile range (IQR), and categorical variables were reported as frequency and proportions. Most studies reported risk estimate as an adjusted hazard ratio (HR), while a few reported an odds ratio (OR) and risk ratio (RR) which was converted into HR using a standard formula. The pooled HR for mortality associated with HA use was calculated using DerSimonian and Laird random-effects model (DerSimonian and Laird, 1986). I² statistic and Cochran’s Q test was used to assess the heterogeneity across the included studies (Cochran, 1954). Heterogeneity due to real differences in the study population, methodology and outcome was indicated by an I² value >50% and a small p (<0.10) of the Cochran’s Q test (Higgins et al., 2003). All statistical analyses except for publication bias were performed using Review Manager (RevMan) Version 5.1 (Copenhagen: The Nordic Cochrane Center, The Cochrane Collaboration, 2011). A probability level <0.05 was considered statistically significant (except for publication bias and heterogeneity assessment) and all p were 2-tailed. Publication bias was assessed quantitatively using Egger’s precision test (weighted linear regression) (Egger et al., 1997) and was considered present if p < 0.10. To assess publication bias, we used Comprehensive Meta-Analysis (CMA) Version 2 (Biostat, Englewood, NJ). Quality of evidence was assessed based on Grading of Recommendations Assessment, Development and Evaluation (GRADE) working guidelines (Atkins et al., 2004).

Results

Study selection

A database search identified a total of 2182 articles and six additional articles were found by a references search of relevant articles. An initial screening of titles and abstracts was done by two reviewers in parallel. During the initial screening, 2144 articles were excluded and 44 articles were included for second screening (Figure 1). A full text review of 44 articles was done by the same two reviewers in parallel and 25 studies were included for final data synthesis. Agreement between the two reviewers was high during both steps (kappa > 0.90).

Figure 1.

Flow diagram for the selection of references.

Characteristics of included studies and qualitative summary

A detailed description of included 25 studies is reported in Table 1. Of the 25 included studies, 24 were cohort studies (22 prospective cohorts and two retrospective cohorts), while one study was case control, wherein for mortality analysis, only cases with community-acquired pneumonia were included (Obiora et al., 2013). Studies were conducted in different countries across the world. Total sample size included in this meta-analysis was 2,350,093 (range 500–1,099,830) participants with 59% being females. Age of participants ranged from 18–102 years. Study populations differed across the included studies with detailed description given in Table 1. Duration of follow-up ranged from 1–22 years. A detailed quality assessment of the included studies is reported in eTable 1 (online supplement). All studies scored reasonably highly on the quality assessment tool. Agreement between the reviewers was high on each item of the Newcastle-Ottawa scale (kappa > 0.90).

Table 1.

Characteristics of included studies.

AuthoryearCountryStudy type	Total patients(females %)	Age (years)Mean ± SD	Study population	Exposure	Follow-up	Mortality assessment	Conclusion
Ahmad R 2005 UK Longitudinal cohort	1042	75.21	Community-dwelling people aged 65 or older	Hypnotic drugs/ Yes or no	15 years	National health death register	Daily users of HA drugs had higher mortality after adjusting lifestyle and socioeconomic variables.
Belleville G 2010 Canada Longitudinal panel survey	14117 (50.9%)	44.09	Adults between the age of 18–102 years selected from population	HA used in past month	12 years	Death was recorded if subject died during follow-up	Sedative drug use is associated with small but significant increase in mortality risk.
Ekstrom MP 2014 Sweden Prospective cohort	2249 (62.7%)	>45 years	National-wide patients starting long-term oxygen therapy for COPD, aged ⩾45 years	Benzodiazepines current use	1.1 years	Mortality recorded from death registries	Benzodiazepines were associated with increased mortality in patients with chronic obstructive pulmonary disease.
Fukuhara S 2006 12 countries Prospective cohort	7475	NA	National-wide sample of patients with hemodialysis (facilitated from 12 countries)	Benzodiazepine	5 years	Mortality recorded from medical records	Benzodiazepines use in patients with hemodialysis was associated with increased mortality.
Gisev N 2011 Finland Retrospective cohort	2224 (65.6%)	75 ± 7.1	Community-dwelling elderly >65 years in Finland	Benzodiazepine and Benzodiazepine-related drugs	9 years	Dataset of Social Insurance Institute of Finland	Use of benzodiazepines and related drugs was not associated with an increased mortality.
Hartz A 2012 USA Prospective cohort	161748 (100%)	50–79 years	Postmenopausal women between the age of 50–79 years throughout the United States	Hypnotic used five or more times in a week	8 years	Mortality recorded from medical records	Almost daily use of hypnotics was associated with increased mortality risk.
Hausken AM 2007 Norway Prospective cohort	14951 (51.7%)	40–42 years	Men and women aged 40–42 in the two counties of Ostfold and Aust-Agder in Norway were surveyed	HA used daily during the last month	18 years	Mortality recorded from official statistics of the population of Norway	Daily users of HA drugs showed higher mortality compared with non-users.
Hays JC 1996 USA Prospective cohort	3962 (65.1%)	>85 years	Adults 65 years or older living in the community	Sleep medications	4 years	NA	Use of sleeping medications was not associated with increased mortality.
Hublin C 2007 Finland Prospective cohort	19794 (51.9%)	>18 years	Older Finnish twin cohort, who had same sex before 1958 with both co-twins alive in 1975 and selected from the Central Population Registry	Hypnotics and/or tranquilizers used frequently in last one year	22 years	Mortality recorded from population register of Finland	Increased mortality risk was observed with frequent use of hypnotics/tranquilizers.
Jaussent I 2013 France Prospective cohort	6696 (NA)	72.8	Participants aged 65 years or over were randomly selected from electoral rolls between 1999–2001	Hypnotics used intermittently to persistently for <5 years to >20 years	12 years	Mortality recorded from death registries	When controlling for potential confounders, the risk of mortality was not significantly associated with hypnotic use regardless of the type and duration.
Kojima M 1999 Japan Prospective cohort	5322 (54.2%)	47.3 ± 11.6	Residents of Shirakawa town, central Japan, who underwent health examination from 1982–1986 and were available for follow-up in 1996	Sleeping pills used sometime or always	11.9 years	Mortality recorded from death registries	Female users of sleeping pills were at increased risk of mortality.
Kripke DF 1979 USA Prospective cohort	823065 (55.5%)	30 to >90 years	Sample consists of entire range of geographic, economic, racial and ethnic groups within the United States	Sleeping pills used very often	6 years	Mortality recorded from death registries	Participants using sleeping pills often had 1.5 times more mortality compared with those who never used sleeping pills.
Kripke DF 1998 USA Prospective cohort	1099830 (56.9%)	<50 to >70 years	Second Cancer Prevention Study, community-based sample	Hypnotics used >30 times/month for>5 years	6 years	Mortality recorded from death certificate and medical records	Hypnotics use was associated with excess mortality.
Kripke DF 2012 USA Prospective cohort	34205 (63.1%)	63.3 years	Population from 41 county areas of Pennsylvania served by the Geisinger Health System	Hypnotics used frequently	2.5 years	Mortality recorded from Social Security Death Index	Receiving hypnotic prescription was associated with >threefold increase in death.
Mallon L 2002 Sweden Prospective cohort	1870 (NA)	56 ± 6	A random sample of males and females aged 45–65 years was selected using the population register	Sleeping pills used often and very often	12 years	Mortality recorded from death certificate	Habitual sleeping pill usage was related to cancer deaths in males and deaths from all other causes in females.
Mallon L 2009 Sweden Prospective cohort	3523 (50.3%)	46 ± 10	Participants randomly selected from general population between the age of 30–65 years	Hypnotics used often and very often	20 years	Mortality recorded from death certificate and death registry	Regular hypnotic use was associated with increased risk of all-cause mortality and cause-specific mortality.
Merlo J 1996 Sweden Prospective cohort	500 (0%)	68 years	Men born in 1914 in the city of Malmo, Sweden, designed to investigate determinant factors of cardiovascular disease	HA used in last 2 weeks	10 years	Mortality recorded from death registries	Combined use of HA and analgesics use seems to be associated with increased all-cause and ischemic heart disease mortality in elderly.
Obiora E 2014 UK Case control; however, for mortality analysis-only cases (with community-acquired pneumonia) were included	4964 (53.6%)	<25 to >75 years	Subjects from Health Improvement Network database, which is a large, longitudinal collection of records of patients registered to various primary care facilities all over the United Kingdom	Benzodiazepine Current (within 30 days), recent (31–90 days) and past use	2.8 years	Mortality recorded from database	Benzodiazepines were associated with an increased risk of mortality from community-acquired pneumonia.
Phillips B 2005 USA Prospective cohort	13563 (55.2%)	45–69 years	Participants aged 44 to 66 years from four communities in United States: Forsyth County NC, Jackson MS; Washington County MD and suburbs of Minneapolis, MN	Hypnotics	6.3 Years	Mortality recorded from death certificate	After adjusting for confounders, hypnotics use was not associated with increased mortality.
Rod NH 2011 France Prospective cohort	16989 (26.3%)	45 years	French national gas and electricity company employees aged between 35–50 years at baseline	Sleeping pills	19 years	Mortality recorded from French national death index	People taking sleeping pills were at higher risk of mortality.
Rumble R 1992 UK Prospective cohort	1042 (NA)	>65 years	Subjects were randomly selected from the patient list of general practitioners and demographically representive of British elderly population	Hypnotics used often or all the time	5 years	Mortality recorded from National Health Service central register where all UK deaths are recorded.
Tiihonen J2012FinlandRegistry-based case linkage study	2588(38.0)	37.8 ± 13.7	All residents of Finland with initial hospitalization with a diagnosis of schizophrenia and had not obtained antipsychotic within 6 months before admission	Benzodiazepinescurrent use >28 defined daily uses	4.2 years	Mortality obtained from prescription database of the Social Insurance Institute.	Benzodiazepine use was associated with a marked increase in mortality among patients with schizophrenia.
Vinkers DJ2003NetherlandsProspectiveCohort	599(NA)	⩾ 85 years	85 years or older inhabitants of Leiden, Netherlands	Benzodiazepineused for more than half of the period in 3 months	3 years	Mortality recorded from death registries	Benzodiazepines was not related to all-cause mortality but was associated with increases fracture-related mortality.
Weich S2014UKRetrospective matched cohort	104145(59%)	55.5 years	Patients aged ⩾16 years permanently registered with a practice. Cases: who had received study drugs by incident from January 1998 to December 2001. Controls were age-, sex- and practice- matched subjects with no prescription for any study drugs	HAexposure restricted to receipt of study drugs in the first year after recruitment	7.6 years	Mortality recorded from medical records	HA users were associated with significantly increased risk of mortality.
Winkelmayer WC2007USAProspective cohort	3630(46.3%)	59.4 ± 15.3 years	Patients from the national cohort of Dialysis Morbidity and Mortality study wave 2 conducted in 1996–1997	Benzodiazepines and Zolpidem	9–12 months	Mortality recorded from the US Renal Data System	Benzodiazepine and Zolpidem use is common in incident dialysis patient and may be associated with greater mortality.

HA: hypnotics or anxiolytic; COPD: chronic obstructive pulmonary disease; NC: North Carolina; MS: Mississippi; MD: Maryland; MN: Minnesota.

Meta-analysis

The pooled adjusted HR for mortality among HA users was significantly higher as compared to non-users (HR = 1.43; 95% confidence interval [CI] = [1.12,1.84]). Overall, HA users had 43% higher risk of mortality compared with non-users (low quality). However, significant heterogeneity was observed between the studies (Cochran’s Q test p < 0.001, I² = 99%) (Figure 2). After removing six studies causing heterogeneity, HA users had 21% higher risk of mortality compared with non-users and results were homogenous (Figure 3) (Ahmad and Bath, 2005; Kripke et al., 2012; Mallon et al., 2002, 2009; Vinkers et al., 2003; Weich et al., 2014). Eight studies had reported a risk estimate for each sex category and pooled results from subset analysis of these studies showed similar increase in the risk of mortality among men (HR = 1.60, 95% CI = [1.29,1.99]) (Figure 4) and women (HR = 1.68, 95% CI = [1.38, 2.04]) (Figure 5). On stratification, 10 studies reported mortality risk associated with only benzodiazepines use and pooled analysis showed 60% higher mortality among benzodiazepines users as compared to non-users (HR = 1.60, 95% CI = [1.03, 2.49]) (Figure 6). Only five studies reported association of mortality with Z-drugs use and pooled effect showed a 73% increased risk of mortality with Z-drugs use, although the effect did not reach statistical significance (HR = 1.73, 95% CI = [0.95, 3.16]) (Figure 7).

Figure 2.

Mortality risk associated with hypnotics and anxiolytics use.

Figure 3.

Mortality risk associated with hypnotics and anxiolytics use after excluding six studies causing heterogeneity.

Figure 4.

Mortality risk associated with hypnotics and anxiolytics use among males.

Figure 5.

Mortality risk associated with hypnotics and anxiolytics use among females.

Figure 6.

Mortality risk associated with benzodiazepines (only) use.

Figure 7.

Mortality risk associated with Z-drugs (only) use.

Twelve studies provided duration of HA use, while remaining 13 studies reported HA use as ‘yes or no’ without giving details about the duration of use. Subgroups analysis showed 1.25 times higher risk of mortality among HA users reporting use as ‘yes’ compared with non-users reporting use as ‘no’. Subgroup analysis as per study location, study population, age, duration of HA use, duration of follow-up and quality of included studies, has been reported in Table 2. Of the different variables, only population characteristics—age was able to explain the study heterogeneity (p = 0.01, Table 2). Studies with mixed-age subjects had the highest risk of mortality (OR 1.74, 95% CI = [1.25, 2.44]). To assess whether any individual study or group of studies had a dominant effect on the meta-analytic HR, each study was excluded and its effect on the main summary estimate and Cochran’s Q test p for heterogeneity was evaluated. No individual study markedly affected the summary estimate or p value for heterogeneity but after removing six studies from the pool(Ahmad and Bath, 2005; Kripke et al., 2012; Mallon et al., 2002, 2009; Vinkers et al., 2003; Weich et al., 2014), risk estimate decreased but heterogeneity in the pooled HR was resolved.

Table 2.

Subgroup analysis.

Subgroups	Number of studies	Risk ratio	95% confidence interval	p for difference between the subgroups
a) Study location
US	7	1.46	[1.08, 1.97]	0.93
Non-US	18	1.43	[1.03, 1.98]
b) Duration of anxiolytics/hypnotics
=1 month	5	1.30	[1.05, 1.60]	0.53
>1 month	7	1.54	[0.95, 2.51]
c) Follow-up duration
⩽10 years	16	1.38	[1.01, 1.90]	0.97
>10 years	9	1.37	[1.15, 1.64]
d) Quality of studies using Newcastle-Ottawa quality assessment scale (NOS)
NOS score ⩽7	5	1.86	[1.36, 2.55]	0.11
Score >7	20	1.33	[1.00, 1.75]
e) Study population
Population-based	11	1.18	[1.02, 1.36]	0.15
Specific cohort	14	1.54	[1.11, 2.15]
f) Covariates adjustment
Adjusted for minimal covariates	14	1.50	[1.06, 2.13]	0.34
Not adjusted for minimal covariates	11	1.26	[1.14, 1.40]
g) Population characteristics age, overall p = 0.01*
1. Age <50 years	2	1.30	[1.00, 1.69]	0.17 (1 vs 2)
2. Age >50 years	8	1.07	[0.98, 1.16]	0.006* (3 vs 2)
3. Mixed age	14	1.74	[1.25, 2.44]	0.18 (3 vs 1)

p < 0.10 implies that difference between the two subgroups is significant.

Publication bias

Quantitatively, there was no evidence of publication bias (Egger’s regression test: t = 0.66, df = 23, p = 0.51).

Mortality causes

Most studies reported all-cause mortality, but 12 studies reported cause-specific mortality with suicides, cardiovascular and cancer-related deaths among the common causes (Table 2).

Discussion

To the best of our knowledge, this is the first meta-analysis that examines the cumulative effects of HA on mortality risk. Our pooled results from 25 observational studies including 2,350,093 subjects showed that HA users had 43% higher risk of mortality during a follow-up period of 1–22 years. Similarly, higher mortality was found among population using benzodiazepine and Z-drugs, even though the effect did not reach statistical significance for Z-drugs likely due to limited number of studies. Mortality risk was significantly higher among men and women with no significant interaction of sex with the mortality. Based on limited evidence, higher deaths due to suicides, cancers and cardiovascular-related causes were observed among HA users; however, no casual association was reported. Increased suicides may be secondary to increased depression. Patients with depression have higher sleep disturbances, which may lead to higher sedatives and hypnotics use, but only limited studies controlled for depression (Jaussent et al., 2013; Mallon et al., 2002, 2009; Phillips and Mannino, 2005; Weich et al., 2014). Furthermore, HA may impair judgment and promote risky behavior including suicides (Drummer, 2002; Neutel and Patten, 1997). Depression may increase the mortality risk in subjects with underlying cardiovascular disease (Shah et al., 2014). Therefore, it is possible that subjects using HA had higher depression leading to increased cardiovascular deaths given that limited studies controlled for depression. HA may increase the risk of cancer by suppressing immune function or promoting viral infection (Kripke, 2008). These findings are supported by carcinogenic effects of hypnotics in rodents, by causing chromosomal damage (Kripke, 2008). However, it is possible that patients receiving hypnotics get more medical care and thus higher cancer detection than non-users.

A significant heterogeneity was observed in the outcome analysis. Heterogeneity could be due to several methodological issues including study design, differing study populations, follow-up duration, sample size, type and duration of HA use and differing adjusted confounders. Heterogeneity in our study could not be explained by the origin of study, study population, sex, duration of HA use, covariates adjustment, follow-up duration or quality of studies. On subgroup analysis by age, significant subgroup difference was seen in the outcome, suggesting that heterogeneity was possibly secondary to differing age of the study population. We also assessed whether any individual study or group of studies had a dominant effect on the meta-analytic HR. After removing six studies, mortality risk decreased to 21% and results were homogenous (Ahmad and Bath, 2005; Kripke et al., 2012; Mallon et al., 2002, 2009; Vinkers et al., 2003; Weich et al., 2014). Ahmad et al. analyzed data on community elderly (>65 years old) enrolled in longitudinal study of activity and aging, to identify risk factors for 15-years mortality (Ahmad and Bath, 2005). They used Cox Regression with genetic algorithms model for analysis in which they adjusted for age, stomach ache, chest pain, joint pain, handgrip strength, ability to raise GBP200 during emergency, self-rated activities, time since last visit to social worker and dentist, but did not adjust for actual medical illnesses. Kripke et al. (2012) reported mortality risk associated with currently popular hypnotics use only in a matched cohort study adjusted for age, gender, smoking, body mass index, ethnicity, marital status and alcohol use. They collected hypnotic information from electronic medical record which provided details about medication orders but not about dispensing or use by the subjects, which might have overestimated the risk. Mallon et al. (2002) studied association between sleep complaints and coronary artery disease-related mortality in a middle-age population. They reported increased mortality among sleeping pills users when adjusted for age; however, it was unknown if sleeping pills included sedatives and hypnotics only. It may be possible that sleeping pills included other sleeping promoting medications like antipsychotics or antidepressants. In another study on population-based cohort, authors reported increased mortality among sleeping pills users; however, the survey was done in 1983 without any descriptions of sleeping pills (Mallon et al., 2009). Mortality data were collected for the period of 1983–2003 with no follow-up on sleeping pills usage which might have changed significantly over the long duration of follow-up. Welch et al. (2014) found increased mortality risk among HA users over 7 years; however, they used primary care prescription database, which did not include psychiatrists’ prescription. Since they used prescription data, it was unknown if patient took the medications or not. Vinkers et al. (2003) found no association between benzodiazepine use and mortality risk; however, they enrolled only community-dwelling elderly individuals aged 85 years or older and used pharmacy registries for benzodiazepine use, which did not confirm if patient actually took the medications. Quality of included studies was reasonably good and ranged from 7–9 on Newcastle-Ottawa Quality Assessment Scale. However, due to significant heterogeneity and limited covariate adjustments among the studies, the overall quality of evidence according to the GRADE guidelines for association of HA is low to moderate. Therefore, the results need to be interpreted with caution.

The strengths of our meta-analysis include the detailed literature search, thorough screening of studies and quality assessment in duplicate, subgroup analysis and the good methodological quality of included studies. There are certain limitations of this meta-analysis which include analysis performed with study-level data rather than using an individual patient data, which would be the preferred methodology to account for patient-level characteristics. Different studies evaluated the mortality risk associated with different groups of HA, and usage varied in duration across the included studies. Another limitation is that adjustment for confounders varied across the studies. However, on subgroup analysis, there was no significant difference among studies who adjusted for major covariates versus who did not. Thus, it could not explain the heterogeneity.

Implications for clinical practice

Based on pooled results from 25 observational studies conducted across the world, there is low to moderate quality evidence (based on GRADE working group guidelines (Atkins et al., 2004)) suggesting that HA use is associated with increased mortality although the evidence supporting cause-specific mortality was limited. Therefore, physicians should carefully consider the increased mortality risk while prescribing HA drugs in addition to other risk factors and contraindications. Serious adverse effects and contraindications of sedatives and hypnotics include benzodiazepine abuse and dependence; concomitant alcohol, opioid and barbiturate abuse due to the risk of life-threatening interactions; pregnancy due to potential for harm in the unborn; myasthenia gravis, sleep apnea, bronchitis and chronic obstructive pulmonary disease (COPD) because of their muscle relaxant action, which may cause respiratory depression, and severe depression which may precipitate suicidal tendencies. Sedatives and hypnotics should be used with cautions in older age due to increased risk of dependence, and elderly are more sensitive to the adverse effects such as memory problems, daytime sedation, impaired motor coordination and falls. However, that does not mean that all patients should be treated identically or not be started on HA. Patients should be informed about the increased association of mortality with HA use. This further reiterates the importance of involving patients in decision making and explaining the merits of alternatives including non-pharmacological management for sleep and anxiety disorder so that patients can make well-informed decisions.

In conclusion, evidence warranting moderate confidence suggests an increased mortality with HA drugs use including benzodiazepine. Further well-designed studies adjusted for potential confounders are warranted to evaluate the temporal relationship between mortality risk and HA. Future studies should also focus on evaluating the underlying mechanism for increased mortality associated with HA use.

Footnotes

Acknowledgements

Ajay K. Parsaik has full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Contents of this manuscript are solely the responsibility of the authors and do not represent the official view of any institutions. These data have been presented at the APA 168th Annual Meeting in Toronto, Canada, 16–20 May 2015.

Declaration of interest

The author(s) declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Ahmad

Bath

(2005) Identification of risk factors for 15-year mortality among community-dwelling older people using Cox regression and a genetic algorithm. Journal of Gerontology, Series A: Biological Sciences and Medical Sciences 60: 1052–1058.

Atkins

Best

Briss

. (2004) Grading quality of evidence and strength of recommendations. British Medical Journal 328: 1490.

Belleville

(2010) Mortality hazard associated with anxiolytic and hypnotic drug use in the National Population Health Survey. Canadian Journal of Psychiatry / Revue canadienne de psychiatrie 55: 558–567.

Billioti de Gage

Moride

Ducruet

. (2014) Benzodiazepine use and risk of Alzheimer’s disease: Case-control study. British Medical Journal 349: g5205.

Cheng

Huang

Lin

. (2008) Characteristics associated with benzodiazepine usage in elderly outpatients in Taiwan. International Journal of Geriatric Psychiatry 23: 618–624.

Chong

Fryer

(2013) Prescription sleep aid use among adults: United States, 2005–2010. NCHS Data Brief 127: 1–8.

Cochran

(1954) The combination of estimates from different experiments. Biometrics 10: 101–129.

Cohen

(1960) Coefficient of agreement for nominal scales educational and psychological measurement. Educational and Psychological Measurement 20: 37–46.

Cumming

Le Couteur

(2003) Benzodiazepines and risk of hip fractures in older people: A review of the evidence. CNS Drugs 17: 825–837.

10.

Curran

(1986) Tranquillising memories: A review of the effects of benzodiazepines on human memory. Biological Psychology 23: 179–213.

11.

DerSimonian

Laird

(1986) Meta-analysis in clinical trials. Controlled Clinical Trials 7: 177–188.

12.

Donoghue

Lader

(2010) Usage of benzodiazepines: A review. International Journal of Psychiatry in Clinical Practice 14: 78–87.

13.

Drummer

(2002) Benzodiazepines—Effects on human performance and behavior. Forensic Science Review 14: 1–14.

14.

Egger

Davey Smith

Schneider

. (1997) Bias in meta-analysis detected by a simple, graphical test. British Medical Journal 315: 629–634.

15.

Hausken

Skurtveit

Tverdal

(2007) Use of anxiolytic or hypnotic drugs and total mortality in a general middle-aged population. Pharmacoepidemiology and Drug Safety 16: 913–918.

16.

Hays

Blazer

Foley

(1996) Risk of napping: Excessive daytime sleepiness and mortality in an older community population. Journal of the American Geriatrics Society 44: 693–698.

17.

Higgins

Thompson

Deeks

. (2003) Measuring inconsistency in meta-analyses. British Medical Journal 327: 557–560.

18.

Jaussent

Ancelin

Berr

. (2013) Hypnotics and mortality in an elderly general population: A 12-year prospective study. BMC Medicine 11: 212.

19.

Kojima

Wakai

Kawamura

. (2000) Sleep patterns and total mortality: A 12-year follow-up study in Japan. Journal of Epidemiology 10: 87–93.

20.

Kripke

(2008) Possibility that certain hypnotics might cause cancer in skin. Journal of Sleep Research 17: 245–250.

21.

Kripke

Garfinkel

Wingard

. (2002) Mortality associated with sleep duration and insomnia. Archives of General Psychiatry 59: 131–136.

22.

Kripke

Klauber

Wingard

. (1998) Mortality hazard associated with prescription hypnotics. Biological Psychiatry 43: 687–693.

23.

Kripke

Langer

Kline

(2012) Hypnotics’ association with mortality or cancer: A matched cohort study. BMJ Open 2: e000850.

24.

Mallon

Broman

Hetta

(2002) Sleep complaints predict coronary artery disease mortality in males: A 12-year follow-up study of a middle-aged Swedish population. Journal of Internal Medicine 251: 207–216.

25.

Mallon

Broman

Hetta

(2009) Is usage of hypnotics associated with mortality? Sleep Medicine 10: 279–286.

26.

Mellinger

Balter

Uhlenhuth

(1985) Insomnia and its treatment: Prevalence and correlates. Archives of General Psychiatry 42: 225–232.

27.

Moher

Liberati

Tetzlaff

. (2009) Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. British Medical Journal 339: b2535.

28.

Morgan

Dallosso

Ebrahim

. (1988) Prevalence, frequency, and duration of hypnotic drug use among the elderly living at home. British Medical Journal 296: 601–602.

29.

Neutel

Patten

(1997) Risk of suicide attempts after benzodiazepine and/or antidepressant use. Annals of Epidemiology 7: 568–574.

30.

Obiora

Hubbard

Sanders

. (2013) The impact of benzodiazepines on occurrence of pneumonia and mortality from pneumonia: A nested case-control and survival analysis in a population-based cohort. Thorax 68: 163–170.

31.

Ohayon

Caulet

(1996) Psychotropic medication and insomnia complaints in two epidemiological studies. Canadian Journal of Psychiatry / Revue canadienne de psychiatrie 41: 457–464.

32.

Ohayon

Caulet

Priest

. (1998) Psychotropic medication consumption patterns in the UK general population. Journal of Clinical Epidemiology 51: 273–283.

33.

Olfson

King

Schoenbaum

(2015) Benzodiazepine use in the United States. JAMA Psychiatry 72: 136–142.

34.

Pariente

Dartigues

Benichou

. (2008) Benzodiazepines and injurious falls in community dwelling elders. Drugs & Aging 25: 61–70.

35.

Paulose-Ram

Safran

Jonas

. (2007) Trends in psychotropic medication use among U.S. adults. Pharmacoepidemiology and Drug Safety 16: 560–570.

36.

Phillips

Mannino

(2005) Does insomnia kill? Sleep 28: 965–971.

37.

Quera-Salva

Orluc

Goldenberg

. (1991) Insomnia and use of hypnotics: Study of a French population. Sleep 14: 386–391.

38.

Rumble

Morgan

(1992) Hypnotics, sleep, and mortality in elderly people. Journal of the American Geriatrics Society 40: 787–791.

39.

Shah

Ghasemzadeh

Zaragoza-Macias

. (2014) Sex and age differences in the association of depression with obstructive coronary artery disease and adverse cardiovascular events. Journal of the American Heart Association 3: e000741.

40.

van Vliet

van

der

Mast

van

den

Broek

. (2009) Use of benzodiazepines, depressive symptoms and cognitive function in old age. International Journal of Geriatric Psychiatry 24: 500–508.

41.

Vermeeren

(2004) Residual effects of hypnotics: Epidemiology and clinical implications. CNS Drugs 18: 297–328.

42.

Vinkers

Gussekloo

van der Mast

. (2003) Benzodiazepine use and risk of mortality in individuals aged 85 years or older. Journal of the American Medical Association 290: 2942–2943.

43.

Weich

Pearce

Croft

. (2014) Effect of anxiolytic and hypnotic drug prescriptions on mortality hazards: Retrospective cohort study. British Medical Journal 348: g1996.

44.

Wells

Shea

O’Connell

. (n.d.) The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomized studies in meta-analyses. Available at: www.ohri.ca/programs/clinical_epidemiology/oxford.asp (accessed 19 February 2013).