Skip to content

Pregabalin, benzodiazepines and other agents

TL;DR — Outside the antidepressants, three agents have real GAD evidence and each carries a distinct problem. Pregabalin is efficacious with a modest effect (Hedges' g=0.37, 95% CI 0.30–0.44 across 8 RCTs and 2,299 patients; g=0.364 in an earlier 7-trial, 1,352-patient analysis) and no excess dropout versus placebo (Generoso 2017, PMID 27643884; Boschen 2011, PMID 21959031); a 2025 meta-analysis of 14 studies and 4,822 patients found it superior on HAM-A at multiple timepoints, with lower discontinuation than comparators (OR 0.80, 0.70–0.91) and favourable cost-effectiveness — and argued for first-line status (Cardoner 2025, PMID 39989902). Benzodiazepines work: 56 studies, 7,556 participants, all BZDs better than placebo with no differences between them and no tolerability difference from placebo except diazepam (RR 1.61, 1.32–1.96) (Fernandes 2025, PMID 40544830) — but the network ranked them poorly tolerated against other agents in the reference GAD network (Slee 2019, PMID 30712879), and dependence is the reason guidelines restrict them (guidelines). Quetiapine has the single largest HAM-A effect of any GAD drug (−3.60, 95% CrI −4.83 to −2.39) and the acceptability problem to match (discontinuation OR 1.44, 1.16–1.80) (Slee 2019, PMID 30712879); the Cochrane review found response OR 2.21 (1.10–4.45) versus placebo alongside more weight gain, sedation, extrapyramidal effects and AE dropout (Depping 2010, PMID 21154392). Buspirone's modern standing rests on 1990s data. This page is where GAD's efficacy-versus-tolerability trade-off is most explicit.

Pregabalin

Pregabalin is an α2δ calcium-channel subunit ligand. Regulatory status varies by jurisdiction and is not used here as an efficacy claim.

Analysis Design Efficacy Tolerability
Boschen 2011 (PMID 21959031) 7 placebo-controlled trials, n=1,352 Overall Hedges' g=0.364; psychic anxiety 0.349; somatic anxiety 0.239 —
Generoso 2017 (PMID 27643884) 8 RCTs, n=2,299 Hedges' g=0.37 (95% CI 0.30–0.44); heterogeneity I²=0%; low publication-bias risk on funnel plot No dropout difference from placebo; lower dropout than benzodiazepines at comparable clinical response
Cardoner 2025 (PMID 39989902) 14 studies, n=4,822; comparators BZDs, SSRIs, SNRIs, placebo HAM-A MD −1.23 (2 wk), −1.12 (4 wk), −2.50 (8 wk), −3.31 (6–12 months); response OR 1.51 (1.31–1.75); CGI-I MD −0.25 (−0.38 to −0.12) Discontinuation OR 0.80 (0.70–0.91); AEs favoured pregabalin over SSRIs/SNRIs and BZDs; better cost-effectiveness (QALY MD 0.02, 0.01–0.03)
Slee 2019 (PMID 30712879) Reference network, 89 trials HAM-A MD −2.79 (95% CrI −3.69 to −1.91) — third-ranked efficacy "Relatively good acceptability"

Two things to note. The 12-week point estimate in Cardoner 2025 is reported as MD 0.99 (95% CI 0.35–1.63) — a sign reversal relative to the other timepoints; this is the value as published and it is not reconciled in the abstract. And the psychic-versus-somatic split (0.349 vs 0.239; Boschen 2011, PMID 21959031) is one of the few analyses in GAD pharmacology that separates worry from arousal.

Pregabalin's clinical position remains contested. Its own review literature has repeatedly framed it as a candidate first-line agent (Baldwin 2015, PMID 26259772; Cardoner 2025, PMID 39989902), while its abuse potential and misuse in populations with substance-use histories are the standing counter-argument (red flags and safety concerns).

Benzodiazepines

Question Finding Source
Do they work in GAD? All BZDs significantly better than placebo; no significant differences between individual BZDs; high heterogeneity and inconsistency Fernandes 2025, PMID 40544830 (56 studies, 7,556 participants; PROSPERO CRD42022330264)
Are they tolerated? No significant tolerability difference from placebo except diazepam (RR 1.61, 95% CI 1.32–1.96); discontinuation largely undifferentiated Fernandes 2025, PMID 40544830
How do they rank against other drugs? Effective but poorly tolerated relative to duloxetine/pregabalin/venlafaxine/escitalopram Slee 2019, PMID 30712879
Class-level effect SMD −0.40 (95% CrI −0.65 to −0.15) — the smallest of the pharmacological classes that beat placebo in that network Chen 2019, PMID 31494377
Adverse events vs placebo Diazepam is one of only four agents in a 100-trial anxiolytic network with fewer adverse events than placebo (with agomelatine, clobazam and silexan); clobazam had the lowest discontinuation rate of all Müller 2026, PMID 40788541
Certainty CINeMA ratings across 861 comparisons: 814 high, 7 low, 40 very low Fernandes 2025, PMID 40544830

The tension across these rows is real and worth stating plainly: the benzodiazepine-specific network finds them efficacious and well tolerated, while the all-drug networks rank them poorly tolerated. The difference is the comparator set — against placebo, BZDs look benign; against duloxetine and pregabalin, they do not. Neither analysis captures the outcome that drives guideline restriction, which is dependence over months to years rather than dropout over 4–8 weeks (Balon 2020, PMID 32002938).

An empirical trace of that long shadow: in a pooled 735-patient dataset, patients who had recently (<1 month) stopped a benzodiazepine had higher attrition on buspirone, more adverse events, and the smallest buspirone–placebo difference, while those with no or remote prior BZD use responded to buspirone comparably to a benzodiazepine (DeMartinis 2000, PMID 10732655). Prior benzodiazepine exposure changes how the next drug performs.

Quetiapine and other second-generation antipsychotics

Source Design Efficacy Harms
Slee 2019 (PMID 30712879) Reference network HAM-A MD −3.60 (−4.83 to −2.39) — largest of any agent Discontinuation OR 1.44 (1.16–1.80) vs placebo
Depping 2010 (PMID 21154392) Cochrane, 11 RCTs, 4,144 participants, 3 SGAs; 9 studies in GAD Quetiapine vs placebo response OR 2.21 (1.10–4.45), 4 RCTs, n=2,265; no efficacy difference vs antidepressants More AE dropout, weight gain, sedation, extrapyramidal effects than placebo; more AE dropout, weight gain and sedation than antidepressants
Maneeton 2016 (PMID 26834458) 3 RCTs, 2,248 randomised Quetiapine-XR mean change > placebo and comparable to SSRIs; response/remission superior to placebo at 50 and 150 mg/day only Overall and AE discontinuation higher than placebo; comparable to SSRIs only at 50–150 mg
Chen 2019 (PMID 31494377) Class network Second-generation antipsychotics (with serotonin modulators) did not separate from placebo at class level —
Garakani 2024 (PMID 38382649) Umbrella review, 25 SRs/MAs; 24 of 25 rated low quality on AMSTAR-2 "Lack of high-quality studies of antipsychotics in anxiety disorders outside of the use of quetiapine in GAD" Off-label use continues without approval
Altamura 2011 (PMID 21403524) Randomised placebo-controlled augmentation in partial/non-responders Augmentation strategy evidence See treatment-resistant GAD

Quetiapine is a clear example in GAD of an agent whose efficacy estimate and clinical position diverge: it ranked highest for mean HAM-A change in one network, but tolerability and metabolic and extrapyramidal harms restrict its role.

Buspirone and the rest

Agent Evidence Standing
Buspirone Pooled analysis of 8 placebo-controlled studies, n=520: significant HAM-A improvement and global improvement vs placebo (p≤0.001); 44–64% of the GAD sample had at least moderate coexisting depressive symptoms, and those patients responded at least as well or better (Gammans 1992, PMID 1454160). Response is attenuated by recent prior benzodiazepine use (DeMartinis 2000, PMID 10732655) Efficacious and well tolerated in the reference network, but "limited by small sample sizes" (Slee 2019, PMID 30712879); azapirones SMD −0.58 (−1.00 to −0.17) at class level (Chen 2019, PMID 31494377)
Agomelatine Relapse prevention 19.5% vs 30.7% over 26 weeks (Stein 2012, PMID 22901350); melatonergic class SMD −0.68 (−1.15 to −0.21) (Chen 2019, PMID 31494377); one of four agents with fewer AEs than placebo (Müller 2026, PMID 40788541) Efficacy supported in this literature; regulatory safety requirements were outside the retrieved source set
Hydroxyzine 3-month double-blind RCT, 334 randomised to hydroxyzine, bromazepam or placebo HAM-A change −12.16 with hydroxyzine vs −9.64 with placebo (p=0.019); response and remission also favoured hydroxyzine (Llorca 2002, PMID 12444816). This establishes short-term efficacy, not comparative long-term safety.
Silexan (lavender oil preparation) HAM-A MD −3.84 (95% CrI −6.31 to −1.34) in a 29-trial herbal network (Zhang 2022, PMID 35378276); described as effective and as acceptable as placebo in a 100-trial anxiolytic network (Müller 2026, PMID 40788541) Evidence is not confined to diagnosed GAD; see other psychological and non-drug therapies
Kava Overall anxiety estimate MD −2.46 (−4.47 to −0.32). The PubMed abstract reports a GAD estimate of MD −0.17 with 95% CrI −2.55 to −1.97, an internally impossible interval because the point estimate lies outside it The GAD-specific magnitude cannot be interpreted without a corrected source value (Zhang 2022, PMID 35378276)
Chamomile Long-term chamomile RCT in GAD Small literature (Mao 2016, PMID 27912875)
Vortioxetine Ranked least efficacious in a 100-trial anxiolytic network (Müller 2026, PMID 40788541); GAD-specific reviews mixed (Pae 2015, PMID 25851751; Meza 2021, PMID 34038400) Not established for GAD
Clomipramine and TCAs Highest efficacy in the 100-trial network but the most study discontinuations (Müller 2026, PMID 40788541) The efficacy–tolerability trade-off in its purest form

Remission rather than mean change: a different ranking

The reference network ranks drugs on mean HAM-A change (Slee 2019, PMID 30712879). Ranking the same literature on remission (HAM-A ≤7) gives a partly different ordering. Across 30 studies comprising 32 double-blind RCTs and 13,338 DSM-IV GAD participants (28 trials moderate risk of bias, 4 low), remission odds ratios versus placebo were: agomelatine 2.70 (95% CI 1.74–4.19), venlafaxine 2.28, escitalopram 2.03 (1.48–2.78), duloxetine 1.88 (1.47–2.40), quetiapine 1.88 (1.39–2.55) and paroxetine 1.74 (1.25–2.42) (Kong 2020, PMID 33343351).

Agomelatine tops the remission ranking, which the mean-change ranking does not show. Two subsequent analyses agree:

Analysis Design Result
Stein 2021 (PMID 33537871) Pooled patient-level data from three placebo-controlled agomelatine 25–50 mg trials, n=669 (340 agomelatine / 329 placebo), 12 weeks HAM-A between-group difference 6.30 ± 2.51 (p=0.012); response 67.1% vs 32.5%; remission 38.8% vs 17.3%; Sheehan Disability Scale difference 5.11 ± 1.81 (p=0.005), functional response 79.1% vs 43.2%, functional remission 55.2% vs 25.4%; all confirmed in the more severely anxious subset
Hood 2025 (PMID 38804215) Network meta-analysis building on Kong 2020; 25 studies identified, 20 analysed for remission and 22 for AE discontinuation Agomelatine significantly better than pregabalin on remission (OR 2.22, 1.19–4.21); all other point estimates favoured agomelatine but were non-significant; probability of having the highest remission rate 67%, of the lowest AE-discontinuation rate 68%
Wang 2020 (PMID 32702221) Meta-analysis of agomelatine in GAD Independent replication

Two caveats. Agomelatine's GAD trial programme is industry-sponsored and small relative to the SSRI/SNRI evidence base. And a 6.30-point HAM-A difference (Stein 2021, PMID 33537871) is roughly double the best estimate in the reference network (Slee 2019, PMID 30712879), which is itself a reason for caution about pooled patient-level industry datasets.

Silexan: the best-evidenced non-prescription option

Analysis Design Result
Zhang 2022 (PMID 35378276) Bayesian network meta-analysis, 29 trials, 12 herbs, HAM-A endpoint Silexan MD −3.84 (95% CrI −6.31 to −1.34) — the largest herbal effect and comparable to licensed drugs
Müller 2026 (PMID 40788541) 100-trial, 28,637-participant anxiolytic network Silexan "both highly effective and as acceptable as a placebo"; one of only four agents with fewer adverse events than placebo (with diazepam, agomelatine and clobazam)
Dold 2024 (PMID 38425206) Individual-patient-data analysis of all five double-blind placebo-controlled Silexan trials in anxiety disorders; 1,172 patients (587 Silexan 80 mg/d, 585 placebo), 10 weeks Baseline severity significantly predicted treatment effect (p<0.001); psychic-domain HAM-A items improved more than somatic items, particularly in mild-to-moderate cases; CGI global improvement favoured Silexan at all severity strata
Donelli 2019 (PMID 31655395) Systematic review and meta-analysis of lavender for anxiety Broader lavender literature
Kasper 2016 (PMID 26718792) RCT in mixed anxiety-depression Extends beyond GAD
Bartova 2023 (PMID 35262795) Randomized placebo-controlled data on co-occurring depressive symptoms in subthreshold anxiety and anxiety disorders The subthreshold population that prescription trials exclude

Silexan's profile is unusual in this condition: an effect size in the same range as licensed drugs, an adverse-event rate below placebo, and demonstrated efficacy in subthreshold and mild anxiety — the group that meets no diagnostic threshold and is therefore invisible to the rest of the treatment literature (the diagnostic boundary). It is not a GAD-only evidence base: the pooled IPD covers anxiety disorders (Dold 2024, PMID 38425206).

Cannabinoids

Source Design Result
Black 2019 (PMID 31672337) Systematic review and meta-analysis, 83 eligible studies (40 RCTs), Jan 1980–April 2018, across depression, anxiety, ADHD, Tourette, PTSD and psychosis; GRADE-assessed The reference synthesis; anxiety findings largely derive from studies of anxiety secondary to other medical conditions
Gundugurti 2024 (PMID 38797087) Phase 3, 178 randomised (89/89), multicentre in India, 15 weeks, nanodispersible CBD oral solution vs placebo, mild-to-moderate anxiety GAD-7 difference −7.02 (95% CI −7.52 to −6.52, p<0.0001); HAM-A difference −11.9 (−12.6 to −11.3, p<0.0001); no serious adverse events
Hsu 2026 (PMID 41296368) JAMA review of therapeutic cannabis and cannabinoids Current overview
Chen 2025 (PMID 39352792) Harvard South Shore late-life algorithm Places cannabidiol among low-priority agents with some supportive evidence

The Gundugurti effect sizes are extraordinary — a −7.02-point GAD-7 difference exceeds anything in the SSRI literature and the confidence interval is implausibly narrow for n=178 — and the population is mild-to-moderate anxiety rather than diagnosed GAD. This knowledge base treats cannabidiol in GAD as unestablished pending independent replication.

The trade-off, summarised

Agent Efficacy rank (HAM-A MD, Slee 2019) Tolerability Practical limiter
Quetiapine 1st (−3.60) Poor Weight gain, sedation, EPS; off-label
Duloxetine 2nd (−3.13) Good AE-related dropout NNTH 17 for antidepressants as a class (Kopcalic 2025, PMID 39880377)
Pregabalin 3rd (−2.79) Good Misuse potential
Venlafaxine 4th (−2.69) Good Discontinuation symptoms (Henssler 2024, PMID 38851198)
Escitalopram 5th (−2.45) Good —
Benzodiazepines Effective Poor in cross-drug networks, good in within-class networks Dependence over months–years, not captured by trial endpoints

Open questions

  • Should pregabalin be first-line? One meta-analysis argues yes on efficacy, tolerability and cost-effectiveness (Cardoner 2025, PMID 39989902), while guideline placement varies (guidelines). A targeted PubMed search rerun on 2026-09-02 located no pregabalin-versus-SSRI GAD trial explicitly powered for non-inferiority; this is a dated comparative gap.
  • Why do benzodiazepines look well tolerated within-class (Fernandes 2025, PMID 40544830) and poorly tolerated across-class (Slee 2019, PMID 30712879)? The answer determines whether guideline restriction rests on trial evidence or on dependence data that trials do not measure.
  • Does quetiapine's largest-in-class efficacy (Slee 2019, PMID 30712879) survive at doses whose tolerability is acceptable? Maneeton 2016 (PMID 26834458) suggests only 50–150 mg/day, and only for some outcomes.
  • Is buspirone under-used? Its evidence base is 1990s-era with a positive signal specifically in GAD with depressive symptoms (Gammans 1992, PMID 1454160), a group the modern trials exclude.
  • Does silexan's effect (Zhang 2022, PMID 35378276; Müller 2026, PMID 40788541) replicate in a regulatory-grade GAD trial programme?

References

  1. Generoso MB, et al. Pregabalin for generalized anxiety disorder: an updated systematic review and meta-analysis. Int Clin Psychopharmacol. 2017;32:49-55. PMID 27643884
  2. Boschen MJ. A meta-analysis of the efficacy of pregabalin in the treatment of generalized anxiety disorder. Can J Psychiatry. 2011;56:558-66. PMID 21959031
  3. Cardoner N, et al. Does pregabalin offer potential as a first-line therapy for generalized anxiety disorder? A meta-analysis of efficacy, safety, and cost-effectiveness. Front Pharmacol. 2025;16:1483770. PMID 39989902
  4. Baldwin DS, et al. Efficacy and safety of pregabalin in generalised anxiety disorder: A critical review of the literature. J Psychopharmacol. 2015;29:1047-60. PMID 26259772
  5. Fernandes H, et al. Comparative Efficacy and Safety of Benzodiazepines in the Treatment of Patients with Generalized Anxiety Disorder: A Systematic Review and Network Meta-Analysis. Psychother Psychosom. 2025;94:373-388. PMID 40544830
  6. Balon R, Starcevic V. Role of Benzodiazepines in Anxiety Disorders. Adv Exp Med Biol. 2020;1191:367-388. PMID 32002938
  7. DeMartinis N, et al. Prior benzodiazepine use and buspirone response in the treatment of generalized anxiety disorder. J Clin Psychiatry. 2000;61:91-4. PMID 10732655
  8. Slee A, et al. Pharmacological treatments for generalised anxiety disorder: a systematic review and network meta-analysis. Lancet. 2019;393:768-777. PMID 30712879
  9. Chen TR, et al. Pharmacological and psychological interventions for generalized anxiety disorder in adults: A network meta-analysis. J Psychiatr Res. 2019;118:73-83. PMID 31494377
  10. Müller TJ, et al. Comparative efficacy and acceptability of anxiolytic drugs for the treatment of anxiety disorders: a systematic review and network meta-analysis. Eur Arch Psychiatry Clin Neurosci. 2026;276:1879-1894. PMID 40788541
  11. Depping AM, et al. Second-generation antipsychotics for anxiety disorders. Cochrane Database Syst Rev. 2010:CD008120. PMID 21154392
  12. Maneeton N, et al. Quetiapine monotherapy in acute treatment of generalized anxiety disorder: a systematic review and meta-analysis of randomized controlled trials. Drug Des Devel Ther. 2016;10:259-76. PMID 26834458
  13. Garakani A, et al. Antipsychotic agents in anxiety disorders: An umbrella review. Acta Psychiatr Scand. 2024;149:295-312. PMID 38382649
  14. Altamura AC, et al. Augmentative quetiapine in partial/nonresponders with generalized anxiety disorder: a randomized, placebo-controlled study. Int Clin Psychopharmacol. 2011;26:201-5. PMID 21403524
  15. Gammans RE, et al. Use of buspirone in patients with generalized anxiety disorder and coexisting depressive symptoms. A meta-analysis of eight randomized, controlled studies. Neuropsychobiology. 1992;25:193-201. PMID 1454160
  16. Stein DJ, et al. Agomelatine prevents relapse in generalized anxiety disorder: a 6-month randomized, double-blind, placebo-controlled discontinuation study. J Clin Psychiatry. 2012;73:1002-8. PMID 22901350
  17. Zhang W, et al. Medicinal herbs for the treatment of anxiety: A systematic review and network meta-analysis. Pharmacol Res. 2022;179:106204. PMID 35378276
  18. Mao JJ, et al. Long-term chamomile (Matricaria chamomilla L.) treatment for generalized anxiety disorder: A randomized clinical trial. Phytomedicine. 2016;23:1735-1742. PMID 27912875
  19. Pae CU, et al. Vortioxetine, a multimodal antidepressant for generalized anxiety disorder: a systematic review and meta-analysis. J Psychiatr Res. 2015;64:88-98. PMID 25851751
  20. Meza N, et al. Vortioxetine for generalised anxiety disorder in adults. Medwave. 2021;21:e8172. PMID 34038400
  21. Kopcalic K, et al. Antidepressants versus placebo for generalised anxiety disorder (GAD). Cochrane Database Syst Rev. 2025;1:CD012942. PMID 39880377
  22. Henssler J, et al. Incidence of antidepressant discontinuation symptoms: a systematic review and meta-analysis. Lancet Psychiatry. 2024;11:526-535. PMID 38851198
  23. Kong W, et al. Comparative Remission Rates and Tolerability of Drugs for Generalised Anxiety Disorder: A Systematic Review and Network Meta-analysis of Double-Blind Randomized Controlled Trials. Front Pharmacol. 2020;11:580858. PMID 33343351
  24. Stein DJ, et al. Efficacy of Agomelatine 25-50 mg for the Treatment of Anxious Symptoms and Functional Impairment in Generalized Anxiety Disorder: A Meta-Analysis of Three Placebo-Controlled Studies. Adv Ther. 2021;38:1567-1583. PMID 33537871
  25. Hood SD, et al. Systematic review and network meta-analysis of agomelatine for the treatment of generalized anxiety disorder in adult patients. Int Clin Psychopharmacol. 2025;40:62-74. PMID 38804215
  26. Wang SM, et al. Agomelatine for the Treatment of Generalized Anxiety Disorder: A Meta-Analysis. Clin Psychopharmacol Neurosci. 2020;18:423-433. PMID 32702221
  27. Dold M, et al. Baseline symptom severity and efficacy of Silexan in patients with anxiety disorders: A symptom-based, patient-level analysis of randomized, placebo-controlled trials. Eur Psychiatry. 2024;67:e23. PMID 38425206
  28. Donelli D, et al. Effects of lavender on anxiety: A systematic review and meta-analysis. Phytomedicine. 2019;65:153099. PMID 31655395
  29. Kasper S, et al. Efficacy of Silexan in mixed anxiety-depression - A randomized, placebo-controlled trial. Eur Neuropsychopharmacol. 2016;26:331-340. PMID 26718792
  30. Bartova L, et al. Beneficial effects of Silexan on co-occurring depressive symptoms in patients with subthreshold anxiety and anxiety disorders. Eur Arch Psychiatry Clin Neurosci. 2023;273:51-63. PMID 35262795
  31. Black N, et al. Cannabinoids for the treatment of mental disorders and symptoms of mental disorders: a systematic review and meta-analysis. Lancet Psychiatry. 2019;6:995-1010. PMID 31672337
  32. Gundugurti PR, et al. Evaluation of the efficacy, safety, and pharmacokinetics of nanodispersible cannabidiol oral solution (150 mg/mL) versus placebo in mild to moderate anxiety subjects. Asian J Psychiatr. 2024;97:104073. PMID 38797087
  33. Hsu M, et al. Therapeutic Use of Cannabis and Cannabinoids: A Review. JAMA. 2026;335:345-359. PMID 41296368
  34. Chen A, et al. A Proposed Algorithm for the Pharmacological Treatment of Generalized Anxiety Disorder in the Older Patient. J Geriatr Psychiatry Neurol. 2025;38:155-171. PMID 39352792
  35. Llorca PM, et al. Efficacy and safety of hydroxyzine in the treatment of generalized anxiety disorder: a 3-month double-blind study. J Clin Psychiatry. 2002;63:1020-1027. PMID 12444816