CoChat
Sign Up
Cited by 372Anxiety, Depression, Psychometrics, Treatment, Cognitive ProcessesTreatment of Major DepressionMental Health Treatment and Access

Pharmacological treatments for generalised anxiety disorder: A systematic review and network meta-analysis

Open access
Slee, April, Nazareth, Irwin, Bondaronek, Paulina, Liu, Yifeng, Cheng, Zhihang, Freemantle, Nick·2019·The Lancet
DOI 10.1016/s0140-6736(18)31793-8
Open this paper in CoChat — free accountOpen PDF

Abstract

BACKGROUND Generalised anxiety disorder is a disease that can be associated with substantial dysfunction. Pharmacological treatment is often the first choice for clinicians because of the cost and resource constraints of psychological alternatives, but there is a paucity of comparative information for the multiple available drug choices. METHODS A systematic review and network meta-analysis was performed on randomised trials in adult outpatients with generalised anxiety disorder identified from MEDLINE, Web of Science, Cochrane Library, ClinicalTrials.gov, Chinese National Knowledge Infrastructure (CNKI), Wanfang data, Drugs@FDA and commercial pharmaceutical registries. Placebo and active control trials were included. Data were extracted from all manuscripts and reports. Primary outcomes were efficacy (mean difference [MD] in change in Hamilton Anxiety Scale Score) and acceptability (study discontinuations for any cause). We estimated summary mean treatment differences and odds ratios using network meta-analyses with random effects. This study is registered with PROSPERO, number CRD42018087106. FINDINGS Studies were published between Jan 1, 1994 and Aug 1, 2017, in which 1992 potential studies were screened for inclusion. This analysis is based on 89 trials, which included 25 441 patients randomly assigned to 22 different active drugs or placebo. Duloxetine (MD -3·13, 95% credible interval [CrI] -4·13 to -2·13), pregabalin (MD -2·79, 95% CrI -3·69 to -1·91), venlafaxine (MD -2·69, 95% CrI -3·50 to -1·89), and escitalopram (MD -2·45, 95% CrI -3·27 to -1·63) were more efficacious than placebo with relatively good acceptability. Mirtazapine, sertraline, fluoxetine, buspirone, and agomelatine were also found to be efficacious and well tolerated but these findings were limited by small sample sizes. Quetiapine (MD -3·60 95% CrI -4·83 to -2·39) had the largest effect on HAM-A but it was poorly tolerated (odds ratio 1·44, 95% CrI 1·16-1·80) when compared with placebo. Likewise, paroxetine and benzodiazepines were effective but also poorly tolerated when compared with placebo. Risk of reporting bias was considered low, and when possible all completed studies were included to avoid publication bias. INTERPRETATION To our knowledge, this is the largest contemporary review of pharmacological agents for the treatment of generalised anxiety disorder by use of network analysis. There are several effective treatment choices for generalised anxiety disorder across classes of medication. The failure of initial pharmacological therapy might not be a reason to abandon a pharmacological treatment strategy. FUNDING No funding was received for this research.

AI snapshot

Written from the paper’s full text.

TL;DR

This systematic review and network meta-analysis of 89 randomised trials (25,441 patients) compared 22 active drugs or placebo for generalised anxiety disorder. Duloxetine, pregabalin, venlafaxine, and escitalopram were more efficacious than placebo with relatively good acceptability, while quetiapine had the largest HAM-A reduction but poorer tolerability.

Key findings

  • In 89 trials with 25,441 patients, duloxetine (MD −3.13; 95% CrI −4.13 to −2.13), pregabalin (MD −2.79; 95% CrI −3.69 to −1.91), venlafaxine (MD −2.69; 95% CrI −3.50 to −1.89), and escitalopram (MD −2.45; 95% CrI −3.27 to −1.63) were more efficacious than placebo with relatively good acceptability.
  • Quetiapine had the largest impact on HAM-A (MD −3.60; 95% CrI −4.83 to −2.39) but was poorly tolerated versus placebo (OR 1.44; 95% CrI 1.16 to 1.80); paroxetine and benzodiazepines were also effective but poorly tolerated versus placebo.
  • Mirtazapine, sertraline, fluoxetine, buspirone, and agomelatine were efficacious and well tolerated, but these findings were limited by small samples. Among active comparisons, quetiapine, duloxetine, and bupropion each reduced anxiety better than tiagabine by more than half a point on HAM-A, and quetiapine improved more than vortioxetine by more than half a point.
  • Subgroup analyses showed no material differences; excluding Chinese trials had almost no impact on results.

Methods

  • Systematic review and Bayesian network meta-analysis of randomised trials in adult outpatients with GAD, registered with PROSPERO (CRD42018087106); searches covered MEDLINE, Web of Science, Cochrane Library, ClinicalTrials.gov, CNKI, Wanfang, Drugs@FDA, and pharmaceutical registries for studies published 1994 to August 1, 2017.
  • Primary outcomes were efficacy (mean difference in change in Hamilton Anxiety Scale score) and acceptability (study discontinuations for any cause); random-effects network meta-analyses used WinBUGS with 100,000 burn-in and 100,000 sample iterations.
  • A common baseline-to-endpoint correlation of r=0.314 was estimated from 22 trials with HAM-A variance data and used to derive missing change variances; subgroup analyses examined median age, proportion of women, baseline anxiety, baseline depression, and non-Chinese trials.

Limitations

  • The paper states that mixed-treatment comparisons rely on the assumption that trials are similar/exchangeable, that trials were conducted across a broad range of settings, that publication bias is always a concern, that the network contains a small number of open loops which may reduce accuracy, and that only pharmacological treatments were compared, not non-pharmacological options such as cognitive behaviour therapy.
  • The paper states that some drugs were limited by small samples (mirtazapine, sertraline, fluoxetine, buspirone, agomelatine), and that maprotiline and mirtazapine were only studied in China, so estimates restricted to other trials are unavailable.
  • Evident from the text: many trials lacked details about randomisation, Chinese trials tended to be lower quality and were mostly open-label or not explicit about blinding, and there were no studies in the review on amitriptyline.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1016/s0140-6736(18)31793-8

Cites

25 of 25 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 12 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    Venlafaxine was more efficacious than placebo in reducing HAM-A (MD -2.69; 95% CrI -3.50, -1.89).

    “Duloxetine (MD: -3•13; 95% CrI: -4•13, -2•13) pregabalin (MD: -2•79; 95% CrI: -3•69, -1•91), venlafaxine (MD: -2•69; 95% CrI: -3•50, -1•89) and escitalopram (MD: -2•45; 95% CrI: -3•27, -1•63) also demonstrated efficacy compared to placebo in large cumulative samples of patients without increased discontinuation.”
    Quote in the paperSubstantial claim
  • Findingnone given

    The findings for mirtazapine, sertraline, fluoxetine, buspirone and agomelatine were limited by small samples.

    “Mirtazapine, sertraline, fluoxetine, buspirone and agomelatine were also found to be efficacious and well tolerated but these findings were limited by small samples.”
    Quote in the paperNot a substantial claim
  • Findingown data

    Duloxetine, pregabalin, venlafaxine and escitalopram had relatively good acceptability.

    “Duloxetine (MD: -3•13; 95% credible interval [CrI]: -4•13, -2•13) pregabalin (MD: -2•79; 95% CrI: -3•69, -1•91), venlafaxine (MD: -2•69; 95% CrI: -3•50, -1•89) and escitalopram (MD: -2•45; 95% CrI: -3•27, -1•63) were more efficacious than placebo with relatively good acceptability.”
    Quote in the paperSubstantial claim
  • Findingown data

    Escitalopram was more efficacious than placebo in reducing HAM-A (MD -2.45; 95% CrI -3.27, -1.63).

    “Duloxetine (MD: -3•13; 95% CrI: -4•13, -2•13) pregabalin (MD: -2•79; 95% CrI: -3•69, -1•91), venlafaxine (MD: -2•69; 95% CrI: -3•50, -1•89) and escitalopram (MD: -2•45; 95% CrI: -3•27, -1•63) also demonstrated efficacy compared to placebo in large cumulative samples of patients without increased discontinuation.”
    Quote in the paperSubstantial claim
  • Findingown data

    Quetiapine had the largest impact on HAM-A (MD -3.60; 95% CrI -4.83, -2.39) among drugs with large cumulative sample sizes.

    “Quetiapine had the largest impact on HAM-A (Mean Difference [MD]: -3•60; 95% credible interval [CrI]: -4•83, -2•39,) among the drugs with large cumulative sample sizes”
    Quote in the paperNot a substantial claim
  • May combine two claims. 19
  • Classified as probably not a substantial claim. 16

  • May combine two claims.

    Findingown data

    Paroxetine was effective but poorly tolerated compared to placebo.

    “Likewise, paroxetine and benzodiazepine were effective but also poorly tolerated when compared to placebo.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Benzodiazepine was effective but poorly tolerated compared to placebo.

    “Likewise, paroxetine and benzodiazepine were effective but also poorly tolerated when compared to placebo.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    This analysis is based on 89 trials, which included 25,441 patients randomly assigned to 22 different active drugs or placebo.

    “This analysis is based on 89 trials, which included 25,441 patients randomly assigned to 22 different active drugs or placebo.”
    • Also stated in: § Results · plain
      “In total, 25,441patients were enrolled in these trials.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Active comparison trials were more likely in non-industry settings.

    “While active comparison trials were more likely in non-industry settings”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    Industry trials appear to have adopted modern reporting standards (e.g. CONSORT) sooner.

    “industry trials appear to have adopted modern reporting standards (e.g.CONSORT) sooner.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    For most treatments, efficacy was demonstrably greater than for placebo and acceptability was similar to placebo.

    “For most treatments, efficacy was demonstrably greater than for placebo and acceptability was similar to placebo.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    The trials were conducted across a broad range of settings.

    “First, the trials were conducted across a broad range of settings.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    Pregabalin had a large impact on anxiety symptoms compared to placebo with no clinically important difference in acceptability, which is consistent with recent meta-analyses in GAD.

    “Pregabalin had a large impact on anxiety symptoms compared to placebo with no clinically important difference in acceptability, which is consistent with recent meta-analyses in GAD [24].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [24]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    This network contains a small number of open loops.

    “This network contains a small number of open loops”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: The trials included in the network meta-analysis are similar (exchangeable).

  • May combine two claims.

    Findingown data

    The main difference was that we found convincing effectiveness for venlafaxine, which was previously identified among the poorest for HAM-A response.

    “The main difference was that we found convincing effectiveness for venlafaxine, which was previously identified among the poorest for HAM-A response.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Ocinaplon showed encouraging point estimates but with low statistical certainty due to relatively low number of patients included, resulting in wide credible intervals.

    “Other drugs like ocinaplon showed encouraging point estimates but with low statistical certainty due to the relatively low number of patients included, resulting in wide credible intervals.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Agomelatine was the only drug with established effectiveness among the new pharmacologic options to treat GAD, and showed improved acceptability compared to active drugs including benzodiazepine, quetiapine and paroxetine.

    “Agomelatine (based on our small cumulative samples) was the only drug with established effectiveness among the new pharmacologic options to treat GAD, and showed improved acceptability compared to active drugs including benzodiazepine, quetiapine and paroxetine.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Buspirone and hydroxyzine showed reasonable efficacy.

    “and showed reasonable efficacy.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The trials included in the network meta-analysis are similar (exchangeable).

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Trials from China tend to include active comparators.

    “It is also interesting that the trials from China tend to include active comparators”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    Diagnostics included examining autocorrelation plots for convergence and examination of Markov traces.

    “Diagnostics included examining autocorrelation plots for convergence and examination of Markov traces.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation

    The study is reported as recommended by the PRISMA extension statement for network meta-analysis.

    “It is reported as recommended by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) extension statement for network meta-analysis [13].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [13] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The study was registered with PROSPERO (CRD42018087106).

    “This study is registered with PROSPERO (registration number CRD42018087106).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation

    Risk of bias was assessed using the Cochrane Collaboration's risk of bias assessment tool.

    “Risk of bias was assessed using The Cochrane Collaboration's risk of bias assessment tool [14].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [14] source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    Most of the trials were double-blind, and were conducted by pharmaceutical companies as part of a clinical development program.

    “Most of the trials were double-blind, and were conducted by pharmaceutical companies as part of a clinical development program (Appendix 4).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The results were assessed for coherence compared with pairwise meta-analysis results.

    “We also assessed the results for coherence compared with pairwise meta-analysis results.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductown data

    Chinese trials tended to be lower quality; most were either open-label or not explicit about blinding.

    “Chinese trials tended to be lower quality; most of these were either open-label or not explicit about blinding (Appendix 2).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    This finding is consistent with the work of Maneeton and colleagues, who found that higher doses of quetiapine showed higher rates of response in HAM-A, but also that discontinuation due to adverse events was greater with quetiapine compared to placebo.

    “This finding is consistent with the work of Maneeton and colleagues [23], who found that higher doses of quetiapine showed higher rates of response in HAM-A, but also that discontinuation due to adverse events was greater with quetiapine compared to placebo.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [23]

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Baldwin found that tiagabine had the poorest response rate among all active drugs, which is consistent with our results.

    “Baldwin found that tigabine had the poorest response rate among all active drugs, which is consistent with our results.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [26]

      Not checked in the free run.

  • May combine two claims.

    Study design and conductnone given

    Several of the trials were published prior to the Cochrane collaboration tool, and lacked details about the method of randomisation.

    “Several of the trials were published prior to the Cochrane collaboration tool, and lacked details about the method of randomisation.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductown data

    Investigator-initiated trials were less likely to state a low-risk randomisation strategy and clearly describe an intention-to-treat population.

    “Investigator-initiated trials, although more likely to include active controls, were less likely to state a low-risk randomisation strategy and clearly describe an intention-to-treat population.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premiseown data

    Pregabalin was ranked first for acceptability in keeping with our finding where it was better tolerated than several other active agents.

    “Pregabalin was ranked first for acceptability in keeping with our finding where it was better tolerated than several other active agents.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [26]

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The most comprehensive network meta-analysis for GAD prior to this analysis included 27 randomised trials of 9 active drugs and examined HAM-A response rate.

    “The most comprehensive network meta-analysis for GAD prior to this analysis [26] included 27 randomised trials of 9 active drugs (duloxetine, escitalopram, fluoxetine, lorazepam, paroxetine, pregabalin, sertraline, tiagabine, and venlafaxine) and examined HAM-A response rate (defined as a > 50% reduction from baseline score).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [26]

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Lorazepam had the second-best response rate but the poorest acceptability as measured by trial withdrawal due to adverse events, and our efficacy and acceptability findings for benzodiazepines are consistent with this result.

    “Lorazepam had the second-best response rate but the poorest acceptability as measured by trial withdrawal due to adverse events, and our efficacy and acceptability findings for benzodiazepines are consistent with this result.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [26]

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    The inclusion of previously unpublished studies allows a more complete analysis and comparison of these agents.

    “The inclusion of previously unpublished studies allows a more complete analysis and comparison of these agents.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Interpretationnone given

    Newer sources of clinical trial information such as clinicaltrials.gov and clinical trial reports maintained in pharmaceutical trial registries allow the inclusion of unpublished and earlier-phase trials, which would otherwise be commercial in confidence and thus inaccessible.

    “First and foremost, newer sources of clinical trial information such as clinicaltrials.govand clinical trial reports maintained in pharmaceutical trial registries allow the inclusion of unpublished and earlier-phase trials, which would otherwise be commercial in confidence and thus inaccessible.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning

    The motivation for these studies is different from that of the rest of the world.

    “which suggests that the motivation for these studies is different from that of the rest of the world.”
    Quote in the paperNot a substantial claim
31 of 31

  • The authors state this as an assumption.

    The included trials are similar enough across comparisons to justify the network meta-analysis (similarity/transitivity assumption).

    “Trials of refractory GAD, comorbid conditions and discontinuation designs were excluded to preserve the similarity assumption of the mixed treatment comparison.”

  • The paper relies on this without support.

    There is little incentive for pharmaceutical companies to conduct active comparison trials for GAD, and the cost of clinical trials can be prohibitive for academic researchers, governments and other health care payers.

    “There is little incentive for pharmaceutical companies to conduct active comparison trials for GAD, and the cost of clinical trials can be prohibitive for academic researchers, governments and other health care payers.”

  • The authors state this as an assumption.

    The trials included in the network meta-analysis are similar (exchangeable).

    “A limitation of the mixed treatment method is the reliance on the assumption that the trials are similar (exchangeable).”

  • The paper relies on this without support.

    Newer agents for GAD are unlikely to be tested against older agents such as benzodiazepine, buspirone and hydroxyzine.

    “Newer agents for GAD have few studies against active agents, and are unlikely to be tested against older agents such as benzodiazepine, buspirone and hydroxyzine.”

  • The paper relies on this without support.

    The choice of what and where to publish may be influenced by corporate development strategy, especially prior to regulatory approval.

    “The choice of what and where to publish may be influenced by corporate development strategy, especially prior to regulatory approval.”
5 of 5

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Related

Authors
Slee, A., Nazareth, I., Bondaronek, P., Liu, Y., Cheng, Z., & Freemantle, N.
Affiliations
  • University College London (GB)
  • University Hospital of Wales (GB)
Corresponding author
Freemantle, Nick
ORCID
Venue
The Lancet
Publisher
Elsevier BV
Published
2019-02
Volume
393
Issue
10173
Pages
768-777
ISSN
  • 0099-5355 (ISSN-L)
  • 0140-6736
DOI
10.1016/s0140-6736(18)31793-8
OpenAlex
W2914055690
Semantic Scholar
3f0a45f8ce52efa54335736e710fccf94487afc7
Cited by
372
References
25
Citations by year
2026: 37 · 2025: 46 · 2024: 80 · 2023: 50 · 2022: 41 · 2021: 67 · 2020: 35 · 2019: 16
FWCI
5.93
Type
journal article
Language
en
Open access
Licence
www.elsevier.com/tdm/userlicense/1.0/ (tdm)
Locations
  • The Lancet (publishedVersion)
  • PubMed (publishedVersion)
  • UCL Discovery (University College London) (submittedVersion)
Funders
  • National Institute for Health and Care Research (NF-SI-0515-10041)
  • Medical Research Council
Topics
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Anxiety, Depression, Psychometrics, Treatment, Cognitive Processes
  • Health Sciences › Medicine › Pharmacology › Treatment of Major Depression
  • Social Sciences › Psychology › Social Psychology › Mental Health Treatment and Access
Keywords
generalized anxiety disorder, pharmacotherapy, antidepressants, anxiolytics, duloxetine, pregabalin, venlafaxine, escitalopram, mirtazapine, sertraline, fluoxetine, buspirone, agomelatine, quetiapine, paroxetine, treatment efficacy, treatment acceptability, Hamilton Anxiety Rating Scale, comparative drug effectiveness
MeSH terms
Antidepressive Agents, Anxiety Disorders, Clinical Trials as Topic, Double-Blind Method, Female, Humans, Hypnotics and Sedatives, Male, Research Design, Anti-Anxiety Agents, Randomized Controlled Trials as Topic, Treatment Outcome

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-09.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-09.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review