Edited by: Yanhui Liao, Sir Run Run Shaw Hospital, China
Reviewed by: Yu-Tao Xiang, University of Macau, China; Xiao Jun Xiang, Central South University, China; Yanbo Zhang, University of Alberta, Canada
*Correspondence: Yanqing Tang,
This article was submitted to Addictive Disorders, a section of the journal Frontiers in Psychiatry
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Benzodiazepines (BZD) are one of the most frequently prescribed drugs worldwide. However, the cognitive effects of benzodiazepines in the elderly are highly debated. This systematic review and meta-analysis aims to explore the following two questions in the elderly population: (i) Do BZD lead to any impairments in cognitive functions in elderly users? and (ii) Which specific cognitive domains are most affected by BZD use and abuse?
First, we performed a literature search following the PRISMA guidelines. Electronic databases, including PubMed, PsycINFO, EMBASE, Cochrane Library, and Web of Science were searched until May 14th, 2020. After selecting the relevant articles, we integrated the results of the selected studies with a standardized cognitive classification method. Next, we performed meta-analyses with the random-effects model on the cognitive results. Finally, we specifically examined the cognitive impairments of BZD in the abuse subgroup.
Of the included studies, eight of the thirteen had meta-analyzable data. Compared to the controls, elderly BZD users had significantly lower digital symbol test scores (n=253; SMD: -0.61, 95% CI: -0.91 to 0.31, I² = 0%, p < 0.0001). There was no significant difference in Mini-Mental State Examination, Auditory Verbal Learning Test, and Stroop Color and Word Test scores between BZD users and controls. According to the subgroup analyses, BZD abusers performed significantly worse than controls in Mini-Mental State Examination (n=7726; SMD: -0.23, 95% CI: -0.44 to -0.03, I² = 86%, p = 0.02), while there was no significant difference between the regular BZD users and the controls (n=1536; SMD: -0.05, 95% CI: -0.59 to 0.48, I² = 92%, p =0.85).
In the elderly population, the processing speed (digital symbol test scores) was significantly impaired in BZD users; global cognition (Mini-Mental State Examination scores) was significantly impaired in BZD abusers but not in BZD regular users. This study provides insight into the factors that interact with BZD cognitive effects, such as aging, testing tools, and abuse. Clinicians should be cautious when prescribing BZD for the elderly.
PROSPERO, identifier
Benzodiazepines (BZD) are two-ring heterocyclic compounds consisting of a benzene ring fused with a diazepine ring. Since its discovery in the 1950s, BZD’s sedative, hypnotic, anti-anxiety, and anti-convulsive effects have been increasingly accepted, making BZD use highly prevalent among adults (
Since the 1970s, research has found negative effects of BZD on recipients’ cognitive functions (
In recent years, several systematic reviews have found that BZD use was significantly associated with a higher risk of dementia and mild cognitive impairments (MCI). Dementia and MCI introduced a significant growth in mortality and financial burdens worldwide (
Literature in the past 50 years presents contradicting evidence on whether BZD impairs cognitive functioning in the elderly (
This systematic review and meta-analysis aims to explore the following two questions in the elderly: (i) Is BZD use associated with impairment in cognitive functions in the elderly? and (ii) Which cognitive domains have declined functionality associated with BZD use and abuse? The answers could help characterize the specific cognition impairments associated with BZD use, and identify individuals vulnerable to the negative effects of BZD. This meta-analysis may also help identify and monitor the cognitive effects associated with BZD use and abuse to prevent BZD addiction. With the high prevalence of BZD being prescribed to older populations worldwide, it is essential to inform clinicians and patients about the possible cognitive impairments associated with BZD use and abuse. Given the refractory rate and adverse effects of dementia and MCI (
The process of this systematic review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, 2009 (
We searched relevant articles in electronic databases, PubMed, PsycINFO, EMBASE, Cochrane Library, and Web of Science, from their inception to May 14, 2020. Specific search keywords included “benzodiazepines”, “cognition”, and “aged”. The query used for PubMed is ((Benzodiazepines[mh]) OR (Benzodiazepines Compounds[Title/Abstract]) OR (Benzodiazep*[Title/Abstract])) AND (cognit *’ OR memory OR attention OR visual-spatial OR visuospatial OR recall OR recognition OR problem solving OR reaction time OR vigilance OR executive function*’ OR reasoning OR psychomotor OR motor OR processing OR planning OR verbal fluency OR inhibit *’) AND ((Aged[mh]) OR (Elder*[Title/Abstract]) OR (older adults[Title/Abstract])).
Additionally, we searched in Google Scholar and searched the reference list for relevant articles to ensure that no studies were missed.
We developed the inclusion criteria according to the PICOS guideline. Appropriate papers met the following criteria: 1) Participants were human adults, older than 60 years, mean sample age over 65 years; 2) the treatment groups were BZD users; 3) the studies had placebo or non-users of BZD as controls; 4) the outcomes included performances of cognitive functions measured using any standardized neuropsychological instruments; and 5) any type of published clinical studies except for case reports or conference abstracts were used. We only included articles reported in English.
Endnote was used to delete the duplicated articles. Two reviewers (LL and JL) independently scanned the references by title/abstract to exclude irrelevant articles, then read the full text to identify the appropriate studies based on the above inclusion criteria. Finally, the debated studies were determined through discussions with a third reviewer (YT).
Two reviewers assessed the quality of each included study independently with the modified Newcastle-Ottawa Scale (mNOS) (
After generating a list of the included articles, two reviewers made the data extraction form collectively. The two reviewers then extracted the data independently. The data extraction form contained the following information: (1) author and publication year, (2) study design, (3) setting (country), (4) study design, (5) mean age, (6) gender distribution, (7) education, (8) participants sources, (9) details of BZD use (BZD types, using time, dosage, and the definition of BZD use), and (10) outcomes (cognitive task, cognitive domain, and main findings). The authors were contacted in order to obtain any missing data.
The included studies utilized a variety of psychometric measurements, rendering it challenging to produce generalizable and informative conclusions. In this paper, we organize cognitive domains based on a commonly used framework (
The RevMan software (
The study selection process of this systematic review is summarized in
Flow diagram of the search and study identification process.
The risk bias of the included studies was assessed by the mNOS and detailed in
There were 26033 participants (6374 BZD users) included in this systematic review and 10666 (2318 BZD users) included in the meta-analysis. As shown in
Summary of included articles.
Study | Study design | Country | Samples (BZD users) | Mean Age (BZD users) | Male percentage (BZD users) | Participants sources | Cognitive task | Cognitive domain |
---|---|---|---|---|---|---|---|---|
|
cohort study | U.S. | 3434(1018) | 74.4(74.5) | 40.4(33.1) | population based | CASI | General Cognition |
|
cohort study | U.S. | 2765(400) | range(65-105) | 33.5 | community | SPMSQ; OMC | General Cognition |
|
cohort study | France | 1176(159) | 65(65.3) | 41.6(23.7) | population based | TMT-B; DSS; RAVLT; FTT | General Cognition; Processing speed; Immediate recall: verbal/visual; Gross motor speed |
|
cohort study | Netherlands | 2105(1189) | 69.2 | 47.5 | population based | MMSE; Coding Task; RAVLT; RCPM | General Cognition; Processing speed; Immediate recall: verbal/visual; Reasoning/planning |
|
cohort study | France | 5195(969) | 73.5(74.6) | 40.1(23.1) | community | MMSE; TMT-B; TMT-A; BVRT; IST | General Cognition; Processing speed; Immediate recall: verbal/visual; Verbal fluency |
|
cohort study | U.S. | 5423(405) | 73.0(73.6) | 34.1(30.6) | Alzheimer’s disease center | MMSE; CDR-SB | General Cognition |
|
cohort study | Spain | 64(33) | 73.2(73.5) | 28.13(21.9) | BZD users from hospital; controls are volunteers | MMSE; CPT-II; SDMT; RCFT; CVLT; COWAT FAS; IGT; Tower of London Test; N-Back; SCWT | General Cognition; Vigilance/focus; Processing speed; Immediate recall: verbal/visual; Delayed recall: verbal/visual; Recognition: verbal/visual; Verbal fluency; Reasoning/planning; Working memory; Inhibitory control |
|
cohort study | Spain | 1087(810) | 74.7 | 36.1 | community | MMSE; FCSRT; semantic verbal fluency (animals in one minute); CDT; DST | General Cognition; Immediate recall: verbal/visual; Delayed recall: verbal/visual; Verbal fluency; Reasoning/planning; Processing speed |
|
cross-sectional study | Norway | 241(168) | 78.6(78.1) | 27.8(25.0) | hospital | MMSE; HVLT; SCWT; DVT | General Cognition; Vigilance/focus; Immediate recall: verbal/visual; Delayed recall: verbal/visual; Recognition: verbal/visual; Inhibitory control |
|
cross-sectional study | Canada | 1754(408) | 79.7(79.6) | 38.7 | community | WAIS Block Design; Buschke Free Recall; AVLT Trial 6; Verbal Fluency; Token Test; WAIS Information; WAIS Similarities; WAIS Comprehension | Processing speed; Immediate recall: verbal/visual; Recognition: verbal/visual; Fine motor speed; Verbal Fluency; Semantic processing; Reasoning/planning |
|
cross-sectional study | Netherlands | 2275(702) | range(85-90) | 28.0(20.0) | community | MMSE; LDT; PLT-I; PLT-d; SCWT the Third Chart | General Cognition; Processing speed; Immediate recall: verbal/visual; Delayed recall: verbal/visual; Inhibitory control |
|
cross-sectional study | Finland | 119(64) | 81.6(82.1) | 23.53(15.6) | hospital | MMSE | General Cognition |
|
cross-sectional study | Germany | 395(49) | 78.8(84.0) | 31.9(28.6) | hospital | MMSE | General Cognition |
CASI, Cognitive Abilities Screening Instrument; SPMSQ, Short Portable Mental Status Questionnaire; OMC, Orientation-Memory-Concentration Test; MMSE, Mini Mental State Examination; TMT-B, The Trail Making Test, part B; DSS, The Digit Symbol Substitution Test; RAVLT, Rey Auditory Verbal Learning Test; FTT, Finger Tapping Test; RCPM, Raven’s Coloured Progressive Matrices; TMT-A, The Trail Making Test, part A; BVRT, The Benton Visual Retention Test; IST, Isaacs Set Test; CDR-SB, Clinical Dementia Rating Sum of Boxes; CPT-II, Conners Continuous Performance Test II-Omissions; SDMT, Symbol Digits Modalities Test; RCFT, Rey Complex Figure Test; CVLT, California Verbal Learning Test; COWAT FAS, Controlled Oral Word Association Test; IGT, Iowa Gambling Task; SCWT, Stroop Color and Word Test; FCSRT, total immediate and delayed recall in the Free and Cued Selective Reminding Test; CDT, clock drawing test; DST, digit symbol test; HVLT, Hopkins Verbal Learning Test; DVT, Digit Vigilance Test; LDT, Letter Digit Coding Test; PLT-I, Picture Learning Test-immediately; PLT-d, Picture Learning Test-delay.
Participants’ characteristics were collected according to baseline information.
aThe authors report meta-analyzable data.
bThe BZD abuse studies.
Overview of tasks used to assess cognitive functioning in benzodiazepine users across different cognitive domains.
Cognitive domain | Task | Studies | Sensitivitya |
---|---|---|---|
|
|||
|
Conners Continuous Performance Test II-Omissions (CPT-II) | Ros-Cucurull et al. ( |
1/1 |
Digit Vigilance Test (DVT) | Hoiseth et al. ( |
0/1 | |
|
TMT-B | Paterniti et al. ( |
2/2 |
TMT-A | Mura et al. ( |
1/1 | |
The Digit Symbol Substitution (DSS) test | Paterniti et al. ( |
1/1 | |
Coding task | Bierman et al. ( |
0/1 | |
Symbol Digits Modalities Test (SDMT) | Ros-Cucurull et al. ( |
1/1 | |
WAIS Block Design | Helmes and Ostbye ( |
0/1 | |
Letter Digit Coding Test (LDT) | van Vliet et al. ( |
0/1 | |
Digit Symbol Test (DST) | Del Ser et al. ( |
1/1 | |
7/11 | |||
|
|||
|
Rey Auditory Verbal Learning Test (RAVLT) | Paterniti et al. ( |
0/2 |
The Benton Visual Retention Test (BVRT) | Mura et al. ( |
1/1 | |
Rey Complex Figure Test (RCFT) Immediate recall | Ros-Cucurull et al. ( |
1/1 | |
California Verbal Learning Test (CVLT) | Ros-Cucurull et al. ( |
1/1 | |
Hopkins verbal learning test (HVLT) | Hoiseth et al. ( |
0/1 | |
Buschke free recall | Helmes and Ostbye ( |
1/1 | |
Picture Learning Test (PLT-i) | van Vliet et al. ( |
0/1 | |
Free and Cued Selective Reminding Test (FCSRT)-immediate recall | Del Ser et al. ( |
1/1 | |
5/9 | |||
|
Rey Auditory Verbal Learning Test (RAVLT) | Paterniti et al. ( |
1/2 |
Rey Complex Figure Test (RCFT) Delayed recall | Ros-Cucurull et al. ( |
1/1 | |
California Verbal Learning Test (CVLT) | Ros-Cucurull et al. ( |
0/1 | |
Hopkins verbal learning test (HVLT) | Hoiseth et al. ( |
0/1 | |
Picture Learning Test (PLT-d) | van Vliet et al. ( |
0/1 | |
the Orientation- Memory-Concentration Test (OMC) | hanlon1998* ( |
1/1 | |
Buschke free recall | Helmes and Ostbye ( |
1/1 | |
Free and Cued Selective Reminding Test (FCSRT)-delayed recall | Del Ser et al. ( |
0/1 | |
4/9 | |||
|
Rey Auditory Verbal Learning Test (RAVLT) | Paterniti et al. ( |
1/3 |
Rey Complex Figure Test (RCFT) recognition | Ros-Cucurull et al. ( |
0/1 | |
California Verbal Learning Test (CVLT) | Ros-Cucurull et al. ( |
0/1 | |
Hopkins Verbal Learning test (HVLT) | Hoiseth et al. ( |
0/1 | |
1/6 | |||
9/23 | |||
|
|||
|
the Finger Tapping Test (FTT) | Paterniti et al. ( |
1/1 |
|
WAIS Block Design | Helmes and Ostbye ( |
0/1 |
1/2 | |||
|
|||
|
The Isaacs Set Test (IST) | Mura et al. ( |
1/1 |
Controlled Oral Word Association Test (COWAT FAS) | Ros-Cucurull et al. ( |
1/1 | |
verbal fluency | Helmes and Ostbye ( |
0/2 | |
|
Token Test | Helmes and Ostbye ( |
1/1 |
WAIS Information | Helmes and Ostbye ( |
0/1 | |
3/6 | |||
|
|||
|
Raven’s Colored Progressive Matrices (RCPM) | Bierman et al. ( |
1/1 |
Iowa Gambling Task(IGT) | Ros-Cucurull et al. ( |
0/1 | |
Tower of London Test | Ros-Cucurull et al. ( |
1/1 | |
WAIS Similarities | Helmes and Ostbye ( |
0/1 | |
WAIS Comprehension | Helmes and Ostbye ( |
1/1 | |
clock drawing test | Del Ser et al. ( |
0/1 | |
|
N-Back | Ros-Cucurull et al. ( |
1/1 |
|
SCWT | Ros-Cucurull et al. ( |
1/2 |
the third chart of the 40-item SCWT | van Vliet et al. ( |
0/1 | |
5/10 |
* for BZD abuse studies.
Eleven out of 14 experiments on global cognition in the 13 studies showed no relationship between BZD use and decreased performance in global cognition. However, two experimental results (
Seven studies tested the processing speed of BZD users. Four studies concluded that BZD use in the elderly population may result in significant impairment (
An overview of the eight studies that tested memory and learning ability is presented in
Results of the tasks for motor, language ability, and executive functions are controversial (
We conducted subgroup analyses of BZD abusers (
Subgroup analysis of the effect of BZD abuse and regular use on mini mental state examination in the elderly.
Due to the discrepancy in the type of cognitive tasks used in the included studies, we could not obtain meta-analyzable data to examine the effects of BZD abuse on specific cognitive functions. Nevertheless, the results of the experiments in these studies were summarized in
This meta-analysis and systematic review included 6374 BZD users and 19,659 controls to comprehensively investigate the affected cognition domains by BZD use and abuse in elders. A total of 13 papers met the inclusion criteria and were included in the literature review. Eight out of the 13 papers had appropriate tests and sufficient information for the meta-analysis.
Consistent with previous systematic reviews, our meta-analysis suggests no impairment in global cognition among elderly BZD users (
The results consistently showed impairment in elderly BZD users’ processing speed, but not inhibitory control. Processing speed, defined by the time it takes for a person to complete a mental task, has been found to be associated with caudate activity in neuroimaging studies (
The BZD abuse subgroup in this review included 1673 BZD abusers and 6053 controls. The subgroup analysis showed that BZD abusers received significantly lower MMSE scores than the controls, while the BZD regular users’ scores were not significantly different from controls. These results demonstrated that impairment in global cognition occurs after the BZD user develops into abuse. However, the high heterogeneity in the results cannot be ignored. According to the results of the subgroup analysis, the confidence intervals of the two subgroups overlapped, and the difference between the two groups was not statistically significant (
After searching the databases, we did not find a meta-analysis investigating the cognitive effects of BZD abuse. The most relevant studies are two systematic reviews and meta-analyses papers on dementia risk in the elderly with long-duration and high dosage BZD users (
This study is, to our knowledge, the first systematic review and meta-analysis of the cognitive effects of BZD use and abuse in elders. We attempted to consolidate results from studies testing different cognitive functions and data from a variety of cognitive tasks by utilizing a mature cognitive domain classification catalogue. The subgroup analysis of BZD abusers allowed us to preliminarily compare the effects of BZD use and abuse. This analysis encourages further studies to examine the qualitative difference of BZD use and abuse. These results can help identify and monitor the cognitive effects of BZD use and abuse, shedding light on awareness and prevention of BZD addiction at early stages, providing evidence for clinical decision-making, and improving the life quality of the elderly.
There are some limitations to this meta-analysis and systematic review. First, the number of studies and cognitive domains examined in this meta-analysis was limited. Five of the thirteen studies did not report cognitive tests data, and the domains of motor and language did not have meta-analysis results. Second, although the n is quite large, the meta-analysis results of the digital symbol test, AVLT, and SCWT were drawn from two small sample size studies, which is difficult to justify or interpret, considering the biased nature of the original studies. In the subgroup analysis of MMSE, the sample size of studies in the BZD regular users’ subgroup varied from hundreds to 2500, which are much smaller than the sample size of the abusers’ subgroup (>7000). Although the random effects model was used to reduce the impact caused by the difference between the sample size of the two subgroups, there was no analyzable data for further analysis. Third, due to the limitation of meta-analysis methodology, we could not directly compare the cognitive differences between BZD users and abusers. Fourth, other variabilities existed in the 13 studies in terms of the participants’ source (hospital or community), sample size, sex ratio, and cognitive measurements. These factors could also bias the findings to an uncertain extent. Fifth, users and controls were not well matched in the analysis. Participants in the experimental groups of the studies had different preexisting conditions such as anxiety, depression, insomnia, and APOE e4 status. One study did not age match the intervention group with the control group or as a covariable, which means the study could not completely distinguish the cognition impaired by BZD and by natural aging. Other factors related to cognition, such as the use of other psychotropic medications, certain physical diseases, and educational level, were not examined in most studies. Finally, due to the limited number of studies, sources of heterogeneity were not examined.
Our meta-analysis confirmed the negative effects of BZD on elderly users’ processing speed. Therefore, doctors should be cautious when prescribing BZD drugs to elderly patients, especially those with family histories of dementia, Alzheimer’s disease, and other aging-related cognitive deficits. Additionally, although global cognition was not impaired in BZD regular users, BZD abusers had significantly worse performance in global cognition. This research can inform more individualized prescription decisions. For example, elderly patients whose daily activities require higher cognitive processing should be informed of BZD’s potential side effects on their cognitive processing speed. Patients with a history of addiction should prioritize alternative treatments to BZD therapy to prevent BZD dependence and abuse.
Another important finding in this study is that the results of cognitive performance are highly dependent on the type of cognitive measurements in the study. For example, as previously reported, BZD users had significantly lower processing speeds when tested with the TMT. However, studies measuring processing speed with the coding task or block design task did not reveal any significant findings. Therefore, clinical practitioners should be mindful when selecting cognitive tests. It might be reasonable to use tests with higher sensitivity to reduce missed diagnoses.
In addition, through our exploration in the literature on the cognitive effects of BZD use, few studies paid attention to BZD abuse and addiction in participants (
In conclusion, this meta-analysis indicated no significant global cognition deficit (MMSE scores) in BZD users, but did reveal deficits in elders with BZD abuse behaviors. BZD users performed significantly worse in the cognition domain of processing speed (digit symbol test scores) than the controls, but not in memory and learning (AVLT scores) or inhibitory control (SCWT scores). Studies that tested the other cognitive domains, however, showed conflicting results. Unfortunately, these cognitive domains’ measurements varied across studies, rendering it unavailable to be merged into meta-analysis. Clinicians should be cautious when prescribing BZD for the elderly, especially to patients with a family history of age-related cognitive deficits. Moreover, the majority of the included studies did not clearly distinguish between the use and abuse of BZD, making it challenging to evaluate the effects of BZD abuse. Future well-designed studies are needed in order to verify the cognitive effects of BZD use and abuse.
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation, to any qualified researcher.
LL and YT: conceptualization. LL and LJ: data curation and analysis. FW and YT: project administration. LL, YZ, JZ, and PJ: supervision, writing—review, and editing. LL and PJ: writing—original draft.
National Key Research and Development Program of China (2018YFC1311604 and 2016YFC1306900 to YT), National Science Fund for Distinguished Young Scholars (81725005 to Fei Wang), Liaoning Education Foundation (Pandeng Scholar to Fei Wang), Innovation Team Support Plan of Higher Education of Liaoning Province (LT2017007 to Fei Wang), Major Special Construction Plan of China Medical University (3110117059 and 3110118055 to Fei Wang), Joint fund of National Natural Science Foundation of China (U1808204 to FW), Natural Science Foundation of Liaoning Province (2019-MS-05 to FW).
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The authors thank Xiang YT and Guo WB for providing theoretic guidance.
The Supplementary Material for this article can be found online at: