Review Reports - Unresolved Controversies: The Effect of Benzodiazepines on Cognition and Alzheimer’s Disease

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors assessed how benzodiazepines impact cognitive function. They chose 17 articles that explored this topic. Was the measurement of cognition consistent across these studies? Did the patients take only benzodiazepines, or were they also given other medications that could influence cognitive function? Did the patients have any coexisting medical conditions that might impact cognitive function?

Author Response

Was the measurement of cognition consistent across these studies?

Thank you for this important question. The measurement of cognition varied across studies. Among the six studies that assessed cognition as an outcome, three [13,20,21] used composite cognitive scores such as the MMSE or ADAS-Cog, while the remaining three [11,16,19] used selected domain-specific scores, with or without a composite score. The neuropsychological domains assessed included immediate and delayed free recall tests, psychomotor speed test, and executive function. Details of the neuropsychological assessments used in each study are provided in Supplementary Table 1. This heterogeneity is detailed in our Results section (Line 265). We have also elected to include the heterogeneity of cognition measurement as a limitation in Section 4.6 (Line 541), which states:

Line 265: There was also notable heterogeneity in the neuropsychological assessments. Among the six studies that evaluated cognition as an outcome, three [13,20,21] used composite cognitive measures such as the MMSE or ADAS-Cog, whereas the remaining three [11,16,19] employed selected domain-specific tests, with or without a composite score. The neuropsychological domains assessed included immediate and delayed free recall, psychomotor speed, and executive function (Supplementary Table 1).

Line 541: Through detailed comparisons across studies, we identified several methodological and clinical factors contributing to the inconsistencies. These include study design, inclusion of comorbidities and medications, neuropsychological tests applied for assessment of cognitive status, consideration of subgroups, subtypes of BZD drugs, dose and duration.

Did the patients take only benzodiazepines, or were they also given other medications that could influence cognitive function?

Several studies [9,10,12,13] included in our review controlled for concomitant medication use such as antidepressants, whereas others did not differentiate between exclusive benzodiazepine use and polypharmacy. The approach to handling the use of other medications also varied across studies. Some studies excluded participants prescribed other psychotropics, while others reported the effects of concurrent BZD and antidepressant use [13]. We now highlight the potential confounding effects of other medications in results section 3.2 and have revised wording to more clearly indicate which studies addressed polypharmacy in (Line 283).

Line 283: In addition, several studies [9,10,12,13] included in our review controlled for concomitant medication use such as antidepressants, whereas others did not differentiate between exclusive benzodiazepine use and polypharmacy. Approaches to handling the use of other medications also varied across studies. Several studies adjusted for concomitant psychotropic medication use in their analyses [9,10], while others specifically reported the effects of concurrent benzodiazepine and antidepressant use [13]. Therefore, a limitation concerns concomitant medication use, as several studies did not isolate the cognitive effects of benzodiazepines from those of other psychotropic or sedative agents, thereby introducing potential confounding from polypharmacy.

Did the patients have any coexisting medical conditions that might impact cognitive function?

The presence of comorbid medical conditions (such as depression, anxiety, or chronic insomnia) was variably addressed among the included articles. Specifically, seven studies [7,13,14,15,17,18,19] accounted for depression or depressive symptoms while others did not. Two studies [11,12] showed limited inclusion such as demographic characteristics, hypertension, or smoking status. We provided the comorbidities included in each study in Supplementary Table 1. In Section 3.2, we now specifically discuss baseline medical conditions and describe the implications of residual confounding (Line 250). This has also been added to our limitations section (4.6) (Line 541).

Line 250: The presence of comorbid medical conditions (such as depression, anxiety, or chronic insomnia) was variably addressed among the included articles. Specifically, seven studies [7,13,14,15,17,18,19] accounted for depression or depressive symptoms while others did not. Two studies [11,12] showed limited inclusion such as demographic characteristics, hypertension, or smoking status. We provided the comorbidities included in each study in Supplementary Table 1.

Reviewer 2 Report

Comments and Suggestions for Authors

The article by Aidan Butler and colleagues provides a systematic review of The Effect of Benzodiazepines on Cognition. In this article, the authors have mainly evaluated the evidence on benzodiazepine exposure and cognitive outcomes. The review included the studies between 2010 and 2025. The review further showed the findings from 17 studies, and these studies showed inconsistent findings showing the effect of benzodiazepines on cognition with multiple factors such as sex, drug half-life, or dose. Overall, this is a well-written article, and the authors have systematically reviewed the multiple studies and compared the results accordingly. The authors have concluded the need for greater clinical attention to vulnerable populations, including potential sex-specific susceptibility, and high-risk prescribing practices. However, the quality of the review article should be improved, and the following changes can be made.

The authors can include figures in different sections to further improve the quality of the article.

The authors have conducted the search between 2010 and 2025, which limits the findings to only 17 studies. The authors should broaden their search and include further studies to validate findings. This should be included in further discussion about multiple factors. This point needs to be addressed to improve the quality of the review.

Figure 3 is not readable, and the authors should provide high resolution file.

Author Response

The authors can include figures in different sections to further improve the quality of the article.

We thank the reviewer for suggesting the inclusion of additional figures to enhance the presentation of our findings. We carefully considered this recommendation; however, given the substantial heterogeneity among the included studies and the numerous factors contributing to this variability, creating consistent and interpretable figures was not feasible. We believe that the existing tables and narrative synthesis most accurately convey the data while preserving methodological transparency.

The authors have conducted the search between 2010 and 2025, which limits the findings to only 17 studies. The authors should broaden their search and include further studies to validate findings. This should be included in further discussion about multiple factors. This point needs to be addressed to improve the quality of the review.

We appreciate this feedback. The search timeframe (2010–2025) was chosen to reflect the most recent research and ensure high methodological quality, but we acknowledge that this limited the number of included studies. We will expand our discussion (section 4.6: Strengths and Weaknesses) to address this limitation (Line 551).

Line 551: Another potential limitation of this review is the restricted search timeframe (2010-2025). Although this window was selected to capture the most recent evidence, it may have excluded earlier research that could provide additional contextual insights. Future reviews incorporating a broader timeframe may yield a more comprehensive synthesis of the literature.

Figure 3 is not readable, and the authors should provide a high resolution file.

Thank you for raising this point. Table 1 has been resized to prioritize visual clarity.

Reviewer 3 Report

Comments and Suggestions for Authors

Recommendation: Publish after minor revisions noted.

Comments:

The review concludes that the evidence connecting benzodiazepine use to cognitive decline or dementia remains inconclusive due to methodological biases and confounding factors, emphasizing the need for rigorous research and cautious clinical use. The manuscript is interesting, but the authors must explain some points.

How does applying different lag periods (e.g., 2–5 years) to minimize protopathic bias change the observed association between benzodiazepine exposure and incident dementia?
Do drug half-life and cumulative dose jointly modify dementia risk (interaction), and are these effects sex-specific?
Among community-dwelling older adults, does benzodiazepine discontinuation (vs. continuation) alter trajectories of cognitive decline when analyzed with time-varying exposure models?
To what extent do baseline vulnerabilities insomnia, anxiety, depression, SES mediate or confound the relationship between benzodiazepines and cognitive outcomes?
Are domain-specific cognitive effects (e.g., delayed recall, processing speed) detectable in midlife users, and do they predict later-life MCI/dementia independent of BZD exposure?
Do neuroimaging biomarkers (atrophy patterns, white matter integrity) or delirium incidence provide mechanistic links between long-acting benzodiazepines and downstream cognitive impairment?

Author Response

How does applying different lag periods (e.g., 2–5 years) to minimize protopathic bias change the observed association between benzodiazepine exposure and incident dementia?

We appreciate this insightful question. Bietry et al., in particular, examined the association between benzodiazepine use and dementia risk both with and without adjustment for prodromal symptoms. They found that the initially observed association became non-significant after this adjustment. These findings suggest that accounting for lag time or prodromal symptom onset can substantially attenuate the observed association between benzodiazepine exposure and dementia risk. We also discussed the implications of incorporating a lag period in the main text, which states:

Line 218: Specifically, Bietry et al. examined the association between benzodiazepine use and dementia risk both with and without adjustment for prodromal symptoms. They found that the initially observed association became non-significant after this adjustment. These findings suggest that accounting for lag time or prodromal symptom onset can substantially attenuate the observed association between benzodiazepine exposure and dementia risk.

Do drug half-life and cumulative dose jointly modify dementia risk (interaction), and are these effects sex-specific?

Several included studies [15,22] suggest that drug half-life and cumulative dose jointly influence dementia risk, with some evidence for sex-specific effects (e.g., higher risk among women at high doses). We enhanced Section 4.4 (Pharmacology) to discuss interactions between drug pharmacokinetics, cumulative exposure, and sex, providing relevant references and quantitative data (HR, OR) from the analyzed studies, which states:

Line 495: Interaction analyses in several studies [15,22] indicate that dementia risk depends jointly on benzodiazepine half-life and cumulative dose, with long-acting, high-exposure regimens producing the strongest associations. These effects were more pronounced among women, suggesting potential sex-specific pharmacokinetic vulnerability.

Among community-dwelling older adults, does benzodiazepine discontinuation (vs. continuation) alter trajectories of cognitive decline when analyzed with time-varying exposure models?

Thank you for the insightful question. Two studies [17,18] included time-varying measures of BZD use over the study period. However, none of the studies examined alteration of trajectories following discontinuation of BZD use. Specifically, while Teverovsky et al. found that BZD use is associated with incident MCI, there is a lack of evidence regarding whether discontinuation attenuates this risk. We included this point in our discussion section which states:

Line 512: Of 17 articles, two studies, two studies [17,18] included time-varying measures of BZD use over the study period. However, none of the studies examined alteration of trajectories following discontinuation of BZD use. Specifically, while Teverovsky et al. found that BZD use is associated with incident MCI, there is a lack of evidence regarding whether discontinuation attenuates this risk. This gap warrants further investigation in future studies.

To what extent do baseline vulnerabilities insomnia, anxiety, depression, SES mediate or confound the relationship between benzodiazepines and cognitive outcomes?

We again appreciate the question. We agree that insomnia, anxiety, depression, and socioeconomic status (SES) can serve as potential confounding factors of the relationship between benzodiazepines and cognition. One study [15] progressively fitted Cox proportional hazards models, first adjusting for age and sex at Stage I; then for age, sex, body mass index, living status, education, self-perceived health, alcohol consumption, smoking, diabetes, history of hypertension, cardiovascular or cerebrovascular disease, hypercholesterolemia, cranial trauma, and baseline cognitive status at Stage II; and finally for all Stage I and II variables plus depression, anxiety, and insomnia at Stage III. As a result, the hazard ratio for benzodiazepine (BZD) use and incident dementia gradually decreased (Stage I: HR = 1.23, 95% CI 1.02-1.48; Stage II: HR = 1.14, 95% CI 0.94–1.38; Stage III: HR = 1.10, 95% CI 0.90–1.34). We therefore note a noticeable attenuation in the association between BZD use and incident dementia after adjusting for insomnia, anxiety, and depression. We therefore included this aspect as a limitation of our study in Discussion, which states: .

Line 473: One study [15] examined the extent to which baseline vulnerabilities insomnia, anxiety, depression, SES explain the relationship between benzodiazepines and cognitive outcomes. Shash et al. progressively fitted Cox proportional hazards models, first adjusting for age and sex at Stage I; then for age, sex, body mass index, living status, education, self-perceived health, alcohol consumption, smoking, diabetes, history of hypertension, cardiovascular or cerebrovascular disease, hypercholesterolemia, cranial trauma, and baseline cognitive status at Stage II; and finally for all Stage I and II variables plus depression, anxiety, and insomnia at Stage III. As a result, the hazard ratio for benzodiazepine (BZD) use and incident dementia gradually decreased (Stage II: HR = 1.14, 95% CI 0.94–1.38; Stage III: HR = 1.10, 95% CI 0.90–1.34). We therefore note a noticeable attenuation in the association between BZD use and incident dementia after adjusting for insomnia, anxiety, and depression.

Are domain-specific cognitive effects (e.g., delayed recall, processing speed) detectable in midlife users, and do they predict later-life MCI/dementia independent of BZD exposure?

One study [32] demonstrated that domain-specific cognitive effects are detectable in midlife and can predict later-life MCI or dementia. However, this does not address whether domain-specific cognitive deficits observed in midlife among benzodiazepine (BZD) users may lead to later-life MCI or dementia independently of BZD exposure. As noted earlier in this review, the answer likely depends on whether discontinuation of BZD use alters the long-term cognitive trajectory. Since this potential effect warrants further investigation, the question remains unresolved. We included this discussion in the main text, which states:

Line 516: One study [19] included in this review found that long-term BZD use was associated with impaired delayed free recall among midlife women. Another study [32] demonstrated that domain-specific cognitive effects are detectable in midlife and can predict later-life MCI or dementia. Based on these findings, an important question arises: do domain-specific cognitive changes observed in midlife BZD users predict later-life MCI or dementia independently of BZD exposure in later life? As discussed earlier, the answer likely depends on whether discontinuation of BZD use restores the long-term cognitive trajectory. This potential effect remains unresolved and warrants further investigation.

Do neuroimaging biomarkers (atrophy patterns, white matter integrity) or delirium incidence provide mechanistic links between long-acting benzodiazepines and downstream cognitive impairment?

Limited studies directly addressed neuroimaging or delirium as a mechanistic link between benzodiazepine use and cognitive outcomes. Some reported neuroanatomical changes with high cumulative anxiolytic use, but the evidence is preliminary. We expanded Section 4.4 (Pharmacology) to highlight emerging neuroimaging and delirium findings and call for more research in this area. We included this discussion in the main text, which states:

Line 526: While none of the studies included in this review directly examined potential mediation pathways between BZD exposure and cognition, Hofe et al. reported an association between long-term benzodiazepine use and both structural and functional brain changes, including subtle neuroanatomical alterations linked to high cumulative anxiolytic doses. In addition, Dyer et al. identified an increased incidence of delirium among benzodiazepine-exposed patients with Alzheimer’s disease. Together, these findings suggest potential mechanistic pathways that warrant further investigation through neuroimaging and neuropathological studies.

Reviewer 4 Report

Comments and Suggestions for Authors

The manuscript is well-structured and addresses a relevant, timely, and somewhat controversial topic. However, a few minor revisions should be made before it can be accepted for publication:

In the abstract, you mentioned only two databases used for the selection process. Why was Scopus not included?
How did you choose the databases? Why were specific databases containing clinical studies not considered?
Line 177: Please explain the abbreviation MCI. The explanation can be added below the first mention.
Line 207: Please provide an explanation for MMSE.
A synthetic table summarizing the main findings would be helpful.
It would be beneficial if more of the narrative interpretations were supported by quantitative data reported in the analyzed studies (e.g., OR, HR, etc.).
Lines 286–298: This paragraph requires appropriate bibliographic references.
Line 273: Does this refer to Table 1 from the Supplementary Material section? Please verify this information.
Figure 3: In its current format, the figure is difficult to read. The authors should find a clearer way to present it (for example, using a landscape layout).
A synthesis of the BZDs reported in the included studies should be added to Section 4.5. A table might be useful for this purpose.
Section 4.6 should be expanded to include other limitations identified by the authors throughout the manuscript.

Author Response

1. Abstract: In the abstract, you mentioned only two databases used for the selection process. Why was Scopus not included?

Thank you for highlighting this aspect. During our initial search strategy development, we prioritized PubMed and Google Scholar for their broad coverage and ease of access to clinical and epidemiological. We included this limitation in Discussion, which states:

Line 546: A limitation of this review is inclusion of only two databases, which are PubMed and Google Scholar. As a result, some relevant studies indexed exclusively in Scopus, Embase, or PsycINFO may not have been captured although the inclusion of PubMed and Google Scholar provided broad coverage of clinical and epidemiological literature, partially mitigating this limitation.

2. How did you choose the databases? Why were specific databases containing clinical studies not considered?

We sincerely appreciate your observation regarding this limitation. This point has been added to the study’s limitations section, as described above.

3. Line 177: Please explain the abbreviation MCI. The explanation can be added below the first mention.

Mild Cognitive Impairment, a revision was made according to the suggestion in the main text at line 116.

4. Line 207: Please provide an explanation for MMSE.

The full name of MMSE (Mini-mental state of examination) has been included in the main text at line 267, as suggested.

5. A synthetic table summarizing the main findings would be helpful.

We have summarized main findings in Table 1 of the main text.

6. It would be beneficial if more of the narrative interpretations were supported by quantitative data reported in the analyzed studies (e.g., OR, HR, etc.).

Line 333-335: Shash et al. reported a 62% increased risk of dementia for long-acting BZDs (HR = 1.62, CI = 1.11–2.37) compared with short half-life users. Sun et al. similarly found 2.86-fold higher odds of dementia for clonazepam (OR = 2.86) and 2.60-fold higher odds for diazepam (OR = 2.60) [7] compared to controls, representing the two highest odds among the BZDs investigated.

Line 342-344: Hofe et al. reported 33% increased dementia risk with higher cumulative anxiolytic dose compared with never use (HR = 1.33).

Line 345-350: A 21% increased risk of dementia was observed among individuals with exposure of 91–180 defined daily doses (DDDs) (HR = 1.21), and a 28% increased risk of dementia among those with exposure 180 DDDs or more (HR = 1.28), both compared with individuals exposed to fewer than 90 DDD.

Line 352-355: Interaction analyses indicated that the risk of dementia may depend jointly on BZD half-life and cumulative dose, rather than either factor alone. Torres-Bondia et al. also reported that long half-life BZDs were associated with a 21% higher dementia risk only at the highest exposure level (>180 DDDs; HR = 1.21), and women demonstrated stronger dose–response effects than men across all exposure groups.

7. Lines 286–298: This paragraph requires appropriate bibliographic references.

We included references for the paragraph in line 315 in the main text.

8. Line 273: Does this refer to Table 1 from the Supplementary Material section? Please verify this information.

We identified an inaccurate summary in the main text for the lines the reviewer noted and have made the necessary corrections. This revision has also been applied to both Table 1 and Supplementary Table 1. The corrected summary now reads as follows:

Line 303: In addition, one study reported that many long-term BZD users most frequently showed impaired performance in processing speeds (32.6%) and sustained attention (27.2%), with women demonstrating poorer performance than men [11].

9. Figure 3: In its current format, the figure is difficult to read. The authors should find a clearer way to present it (for example, using a landscape layout).

We appreciate this feedback. Table 1 has been resized to prioritize visual clarity.

10. A synthesis of the BZDs reported in the included studies should be added to Section 4.5. A table might be useful for this purpose.

As suggested, we have included a supplementary table (Supplementary Table 2) listing the BZDs reported in the studies reviewed.

11. Section 4.6 should be expanded to include other limitations identified by the authors throughout the manuscript.

We expanded the limitations of the study, which now states:

Line 546: A limitation of this review is inclusion of only two databases, which are PubMed and Google Scholar. As a result, some relevant studies indexed exclusively in other databases such as Scopus or Embase may not have been captured. Another potential limitation of this review is the restricted search timeframe (2010-2025). Although this window was selected to capture the most recent evidence, it may have excluded earlier research that could provide additional contextual insights. Future reviews incorporating a broader timeframe may yield a more comprehensive synthesis of the literature. Moreover, our review does not capture the holistic impact of BZD use on a person’s life including effects on physical functionality, basic and instrumental activities of daily living. Furthermore, most of the studies included were based on cognitively healthy individuals at baseline and examined either cognitive change or incident dementia. Given that persistent prescription—often at high BZD doses—has been reported among patients with moderate-to-severe dementia, further research in this population is warranted.