Statistics — generalized anxiety disorder¶
Quick-reference tables. Every figure carries its source, year, population and method. Conflicting estimates are shown side by side and never averaged. All PubMed records were retrieved live on 2026-09-02; web sources carry their access date.
The single most important rule for reading this file: a screening-scale figure, an interview-based disorder figure, a claims-code figure and a pooled anxiety-disorder figure are four different quantities. They are separated below.
1. Prevalence of GAD — interview-based, disorder-level¶
| Estimate | Population | Criteria | Instrument | Year(s) of data | Source |
|---|---|---|---|---|---|
| Lifetime 3.7% (SE 0.1); 12-month 1.8% (SE 0.1); 30-day 0.8% | 147,261 adults, 26 countries (WHO WMH) | DSM-5 | CIDI, face-to-face | 2001–2012 | Ruscio 2017, PMID 28297020 |
| Lifetime HIC 5.0% / MIC 2.8% / LIC 1.6% | Same | DSM-5 | CIDI | 2001–2012 | Ruscio 2017, PMID 28297020 |
| Lifetime 4.0% with excessiveness removed (vs 2.6% with it) | 133,614 adults, 28 surveys (WHO WMH) | DSM-5 ± excessiveness | CIDI | — | Ruscio 2024, PMID 39364896 |
| Lifetime, developed countries, by duration rule: 7.5% (1 mo) / 5.2% (3 mo) / 4.1% (6 mo) / 3.0% (12 mo); developing: 2.7 / 1.8 / 1.5 / 1.2% | 85,052 adults, 17 countries | DSM-IV at varying durations | CIDI | — | Lee 2009, PMID 19091158 |
| Lifetime 6.2% (US) | US adults | Review of US estimates | — | — | Szuhany 2022, PMID 36573969 |
| Lifetime ~5%; current 2–3% | US | DSM-IV | Review | — | Weisberg 2009, PMID 19371500 |
| Lifetime 0.9%; 12-month 0.4% | 6,616 adults, Singapore | DSM-IV | CIDI 3.0 | — | Lee 2016, PMID 27969075 |
| Current 0.57% | India NMHS, 12 states | DSM | MINI 6.0.0 | 2016 | Jayasankar 2023, PMID 38298878 |
| Lifetime 3.5% (95% CI 2.9–4.2); women 4.2%, men 2.7%; rural 3.7%, urban 2.8% | 7,270 adults, Bangladesh | DSM-5 | SRQ screen → psychiatrist interview | — | Sarkar 2025, PMID 41061374 |
| Past-year 2.80%; only 0.53% without any Axis I/II comorbidity | 12,312 US adults ≥55 (NESARC wave 2) | DSM-IV | Structured interview | — | Mackenzie 2011, PMID 21427639 |
| Lifetime 2.6% (95% CI 2.4–2.8) | 29,709 children/adolescents 6–18, Iran | K-SADS-PL | Interview | — | Mohammadi 2020, PMID 32470794 |
| ~5% lifetime; up to 10% in women ≥40; ~8% of primary-care attenders | General population and primary care | DSM-III/III-R | Review | — | Wittchen 2001, PMID 11414546 |
Conflict: the range 0.4%–1.8% (12-month) and 0.9%–6.2% (lifetime) across these rows is driven by criteria set, duration rule, instrument and country — not by sampling error. Do not average.
2. Prevalence — other measurement frames¶
| Frame | Estimate | Population | Method | Source |
|---|---|---|---|---|
| ICD-11 algorithm, self-report | 7.1% meeting the ICD-11 GAD algorithm | 2,058 UK-representative adults | International Anxiety Questionnaire (IAQ) | Shevlin 2023, PMID 36215152 |
| Administrative claims | 1-year prevalence 2.1% (2012) → 7.4% (2022); incidence 0.8% → 2.4% | 1,086,618 US insured adults (Optum) | Diagnostic codes | Druet-Cabanac 2025, PMID 40611531 |
| Screening scale | 18.2% any past-2-week anxiety symptoms (up from 15.6% in 2019) | US adults, NHIS 2022 | GAD-7 | Terlizzi 2024, PMID 39591466 |
| Self-report, diagnosed + undiagnosed | 5.3%; only 0.5% reporting a diagnosis | Urban China (NHWS 2012–13) | Self-report survey | Yu 2018, PMID 29524751 |
| Screener, veterans | Probable GAD 7.9% (6.7–9.3); mild anxiety symptoms 22.1% (20.5–23.9) | US veterans, nationally representative | Brief screener | Macdonald-Gagnon 2024, PMID 38325107 |
| Subthreshold GAD | Roughly twice the prevalence of full GAD | 15 high-quality + 3 low-quality studies, 48,214 participants | Systematic review | Haller 2014, PMID 24886240 |
| Subthreshold vs threshold, Canada | Threshold 2.6% (722,000); subthreshold 2.3% (655,000) | 24,785, CCHS-MH 2012 | Structured interview | Gilmour 2016, PMID 27849314 |
| Any 12-month disorder, students | 57.4% screened positive (lifetime 65.2%) | 72,288 first-year students, 18 countries; weighted response 20.8% | Validated screening scales | Mason 2025, PMID 40010072 |
3. Pooled anxiety-disorder figures (not GAD)¶
| Figure | Value | Source |
|---|---|---|
| Global prevalent cases, anxiety disorders, 2021 | 359.2 million | Zhang 2026, PMID 40935255 |
| Global prevalence, anxiety disorders | 4.4% of the population; most common of all mental disorders; ~1 in 4 in need (27.6%) receive treatment | WHO — "Anxiety disorders" fact sheet, 8 September 2025, https://www.who.int/news-room/fact-sheets/detail/anxiety-disorders (accessed 2026-09-02) |
| Anxiety + depression as share of mental-disorder burden | 63.1% of mental-health disorder cases; 9.1% of all disease | Zhang 2026, PMID 40935255 |
| Change in age-standardised DALY rate, 1990–2021 | Anxiety +18.2%; depression +13.4% | Zhang 2026, PMID 40935255 |
| Projected anxiety cases, 2040 | >515 million | Zhang 2026, PMID 40935255 |
| Age-standardised DALY rate, 2021 | Anxiety disorders 524.33/100,000 — 2nd of 12 mental disorders after major depression (557.87); largest increase of any subtype | Fan 2025, PMID 40375174 |
| Mental disorders as share of global DALYs | 3.1% (1990) → 4.9% (2019); 80.8M → 125.3M DALYs | GBD 2019 Mental Disorders Collaborators, PMID 35026139 |
| Adolescents/young adults 10–24, 1990–2021 | Anxiety-disorder incidence +52%, sharpest 2019–2021 | Bie 2024, PMID 39691785 |
| Lifetime anxiety disorders | Up to 33.7% | Bandelow 2015, PMID 26487813 |
| 12-month anxiety disorders, US | 18.1% | Kessler 2005, PMID 15939839 |
| Lifetime anxiety disorders, US | 28.8% | Kessler 2005, PMID 15939837 |
There is no GBD estimate for GAD specifically. GBD models anxiety disorders as one cause.
4. Age, sex and onset¶
| Figure | Value | Population/method | Source |
|---|---|---|---|
| Mean age of onset, GAD (with panic, agoraphobia, OCD, PTSD) | 21.1–34.9 years; no sex difference detected | Meta-analysis of 24 general-population studies | Lijster 2017, PMID 27310233 |
| Mean onset, separation anxiety / specific phobia / social phobia | Before age 15 | Same | Lijster 2017, PMID 27310233 |
| Anxiety/fear-related disorders block | Median onset 17 (IQR 9–25); peak 5.5 years; 38.1% before 14, 51.8% before 18 | 192 studies, n=708,561 | Solmi 2022, PMID 34079068 |
| GAD symptom onset, prospective cohort | Mean 15.6 years; 75% before age 20; annual incidence rises considerably between ages 20 and 40 | Zurich cohort, 591 followed 1979–1999 | Angst 2009, PMID 18575915 |
| Female proportion | 66.5% | 1,086,618 US claims | Druet-Cabanac 2025, PMID 40611531 |
| Female excess | OR 1.62 (p=0.012) | Bangladesh national survey | Sarkar 2025, PMID 41061374 |
| Age-related item functioning | Older adults: higher distress/interference, lower fatigue at equal latent severity | 375 clinical participants matched on GAD status/severity | Correa 2019, PMID 31938010 |
5. Comorbidity¶
| Figure | Value | Population | Source |
|---|---|---|---|
| Lifetime any comorbidity | 81.9% (SE 0.7) | WMH, 26 countries | Ruscio 2017, PMID 28297020 |
| Lifetime mood comorbidity | 63.0% (SE 0.9) | Same | Ruscio 2017, PMID 28297020 |
| Lifetime other-anxiety comorbidity | 51.7% (SE 0.9) | Same | Ruscio 2017, PMID 28297020 |
| GAD cases with no other DSM-IV diagnosis | 17.1% of 105 cases; MDD comorbid in 70.4% | Sesto Fiorentino community study | Faravelli 2012, PMID 22578985 |
| Pure GAD / pure MDE / comorbid GAD+MDE, primary care | 3.8% / 4.4% / 1.6% | >20,000 German primary-care patients, one-day survey | Wittchen 2002, PMID 12044105 |
| Confirmed GAD in Spanish primary care | 14% total; GAD+MDD 8%, GAD alone 6% (22% screened positive) | 7,152 patients; HADS-A screen → MINI | Romera 2010, PMID 20541811 |
| Older adults ≥55, GAD without any Axis I/II comorbidity | 0.53% of a 2.80% total | NESARC-2 | Mackenzie 2011, PMID 21427639 |
| Children/adolescents, comorbid anxiety disorder | 57.6% | 29,709 Iranian 6–18 | Mohammadi 2020, PMID 32470794 |
| Pre-existing MDD as GAD risk factor | OR 5.06 (95% CI 5.03–5.08) — largest of all measured | 1.09M US claims | Druet-Cabanac 2025, PMID 40611531 |
| GAD–MDD genetic correlation | +1.00 (women), +0.74 (men) | 37,296 Swedish twins | Kendler 2007, PMID 17121688 |
| GAD heritability | ~30%; meta-analytic 0.32 | Virginia twins (n=1,033 pairs); meta-analysis | Kendler 1992, PMID 1558460; Hettema 2001, PMID 11578982 |
Conflict: 81.9% lifetime comorbidity (WMH) versus pure GAD outnumbering comorbid GAD/MDE 3.8% to 1.6% at a single primary-care consultation (Wittchen 2002). Both are correct; they are lifetime versus point prevalence in different sampling frames.
6. Burden, impairment and cost¶
| Figure | Value | Population | Source |
|---|---|---|---|
| Severe role impairment, lifetime GAD | 50.6% (SE 1.2) | WMH, 26 countries | Ruscio 2017, PMID 28297020 |
| Impairment of pure GAD vs pure MDD | Similar in magnitude | Large representative samples, review | Hoffman 2008, PMID 17146763 |
| HRQoL | Comparable to depression or panic; GAD+depression significantly worse than either alone | Review of 90 articles | Revicki 2012, PMID 22154706 |
| Median annual medical cost | US $2,375 (GAD) vs $1,448 (primary-care patients without GAD) | North America/Europe review | Revicki 2012, PMID 22154706 |
| Mean annual medical cost vs other anxiety disorders | $2,138 higher (mean $6,475) | Same | Revicki 2012, PMID 22154706 |
| Painful physical symptoms (VAS>30) | 59.0% GAD alone vs 28.3% controls; 78.0% GAD+MDD | 7,152 Spanish primary-care patients | Romera 2010, PMID 20541811 |
| Past-year suicidal ideation, threshold GAD | Men 32.0%, women 21.2% | CCHS-MH 2012, n=24,785 | Gilmour 2016, PMID 27849314 |
| Cardiovascular events, GAD in stable CHD | Annual event rate 9.6% vs 6.6% (p=0.03); adjusted HR 1.62 (1.11–2.37) | 1,015 outpatients, mean 5.6 y follow-up | Martens 2010, PMID 20603456 |
6b. Mortality¶
Source: Wagner 2026, PMID 42136520 — 165 studies, 7,395,722 people with any anxiety or stress-related disorder vs 135,059,023 controls, 27 countries, random-effects meta-analysis with meta-regression.
| Outcome | GAD | Any anxiety/stress-related disorder | Comparators |
|---|---|---|---|
| All-cause mortality | RR 1.48 (1.23–1.78), n=9 | RR 1.54 (1.14–2.08), n=42 | PTSD/stress-related 1.39 (1.15–1.67); not elevated in panic, phobias or mixed |
| Suicide mortality | RR 1.93 (1.17–3.17), n=3 | RR 2.88 (2.13–3.89), n=39 | Panic 3.58 (1.39–9.25); PTSD/stress 3.13 (1.85–5.28); mixed 2.77 (1.89–4.07) |
| Natural-cause mortality | RR 1.55 (1.19–2.02), n=5 | RR 1.25 (1.09–1.44), n=19 | Mixed 1.26 (1.02–1.56); PTSD/stress 1.17 (1.03–1.33); not elevated in panic |
| Suicide attempt | Elevated across all anxiety/stress-related disorders | — | Range RR 6.33 (4.08–9.82) panic to 2.74 (1.72–4.35) phobias |
GAD-specific estimates rest on 3–9 studies each; treat as provisional.
6c. Perceived helpfulness and treatment-seeking¶
Source: Stein 2021, PMID 34372811 — WHO World Mental Health surveys, 23 countries, DSM-5 GAD by CIDI 3.0.
| Metric | Value |
|---|---|
| GAD prevalence in these surveys | 4.5% (2.8% LMIC, 5.3% HIC) |
| Ever obtained treatment for GAD | 34.6% (19.2% LMIC, 38.4% HIC) |
| Of treated, ever considered treatment helpful | 70%, comparable across income bands |
| Modelled probability of obtaining helpful treatment if persisting with up to 10 professionals | Virtually all |
| Estimated proportion who would persist that long | 29.7% |
Correlates of seeking treatment (NESARC-III; Zech 2024, PMID 39047416): comorbid depression, panic disorder and PTSD each independently raised the odds, as did fatigue, panic attacks, reassurance-seeking and interpersonal avoidance.
6d. Substance-use comorbidity¶
Source: Alegría 2010, PMID 20923623 — NESARC, N=43,093.
| Metric | Value |
|---|---|
| Lifetime GAD with substance use disorder | 2.04% |
| Lifetime GAD without substance use disorder | 2.10% |
| Implication | GAD-SUD is approximately half of all lifetime GAD |
| Treatment-seeking | Equally low in both groups; both equally likely to receive pharmacological treatment for anxiety |
7. Treatment gap¶
| Setting | Measure | Value | Source |
|---|---|---|---|
| 26 countries (WMH) | Any treatment sought, lifetime GAD | 49.2% (59.4% if severe impairment; 59.0% in HICs) | Ruscio 2017, PMID 28297020 |
| Global, all anxiety disorders | Receive any treatment | 27.6% (~1 in 4) | WHO fact sheet, 8 Sep 2025 (accessed 2026-09-02) |
| India (NMHS 2016) | Treatment gap, current GAD | 75.7% | Jayasankar 2023, PMID 38298878 |
| Bangladesh | Sought professional help | 3.9% | Sarkar 2025, PMID 41061374 |
| Urban China | Reported having a diagnosis | 0.5% (against 5.3% self-reported GAD) | Yu 2018, PMID 29524751 |
| US ≥55 | Sought professional help in past year | 18% (no comorbidity) / 28.3% (with) | Mackenzie 2011, PMID 21427639 |
| US primary care | Anxiety-disorder patients with no current treatment | 41% | Kroenke 2007, PMID 17339617 |
| North America/Europe | "Adequately treated" | 20–32% | Revicki 2012, PMID 22154706 |
| German primary care | Correct diagnosis of pure GAD (emotional problem recognised in 72.5%) | 34.4% (vs 64.3% for MDE) | Wittchen 2002, PMID 12044105 |
8. Treatment effect sizes — pharmacological¶
| Comparison | Effect | Certainty / n | Source |
|---|---|---|---|
| Antidepressants vs placebo, response (≥50% HAM-A reduction) | RR 1.41 (1.29–1.55), NNTB 7 (5–9) | High; 20 studies, 7,267 participants | Kopcalic 2025, PMID 39880377 |
| Antidepressants, all-cause dropout | RR 1.03 (0.93–1.14) | High; 33 studies, 11,294 | Kopcalic 2025, PMID 39880377 |
| Antidepressants, dropout for lack of efficacy | RR 0.41 (0.33–0.50), NNTB 27 | High; 29 studies, 11,007 | Kopcalic 2025, PMID 39880377 |
| Antidepressants, AE dropout | RR 2.18 (1.81–2.61), NNTH 17 (13–112) | High; 32 studies, 11,793 | Kopcalic 2025, PMID 39880377 |
| Drug ranking, HAM-A mean difference vs placebo | Quetiapine −3.60 (−4.83 to −2.39); duloxetine −3.13 (−4.13 to −2.13); pregabalin −2.79 (−3.69 to −1.91); venlafaxine −2.69 (−3.50 to −1.89); escitalopram −2.45 (−3.27 to −1.63) | 89 trials, 25,441 patients | Slee 2019, PMID 30712879 |
| Class effects (SMD vs placebo) | NDRI −1.84 (−3.05 to −0.62); NaSSA −0.91; melatonergic −0.68; SSRI −0.67 (−0.90 to −0.43); azapirone −0.58; anticonvulsant −0.56; SNRI −0.54 (−0.79 to −0.30); BZD −0.40 (−0.65 to −0.15) | 91 articles, 14,812 participants, all comorbidities allowed | Chen 2019, PMID 31494377 |
| Pregabalin vs placebo | Hedges' g 0.37 (0.30–0.44), I²=0% | 8 RCTs, n=2,299 | Generoso 2017, PMID 27643884 |
| Pregabalin (earlier) | g 0.364; psychic 0.349, somatic 0.239 | 7 trials, n=1,352 | Boschen 2011, PMID 21959031 |
| Pregabalin vs comparators (HAM-A MD) | −1.23 (2 wk), −1.12 (4 wk), −2.50 (8 wk), −3.31 (6–12 mo); response OR 1.51 (1.31–1.75); discontinuation OR 0.80 (0.70–0.91) | 14 studies, n=4,822 | Cardoner 2025, PMID 39989902 |
| Benzodiazepines | All better than placebo; no differences between individual BZDs; only diazepam differed on tolerability (RR 1.61, 1.32–1.96) | 56 studies, 7,556 participants | Fernandes 2025, PMID 40544830 |
| Quetiapine vs placebo, response | OR 2.21 (1.10–4.45) | 4 RCTs, n=2,265 (Cochrane) | Depping 2010, PMID 21154392 |
| Vilazodone | HAM-A/CGI superior to placebo but NNT 10 (response), NNT 12 (CGI-I), NNH 14; LHH 1.4 (0.48–3.33) | 3 RCTs; 844 drug / 618 placebo | Zareifopoulos 2017, PMID 28483071 |
| Remission ranking (HAM-A ≤7) vs placebo | Agomelatine OR 2.70 (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); paroxetine 1.74 (1.25–2.42) | 30 studies / 32 RCTs, 13,338 participants | Kong 2020, PMID 33343351 |
| Agomelatine 25–50 mg, pooled patient-level | HAM-A between-group difference 6.30 ± 2.51 (p=0.012); response 67.1% vs 32.5%; remission 38.8% vs 17.3%; SDS difference 5.11 ± 1.81 | 3 placebo-controlled trials, n=669 | Stein 2021, PMID 33537871 |
| Silexan 80 mg/d | Baseline severity predicted treatment effect (p<0.001); psychic-domain items improved more than somatic | IPD from all 5 double-blind placebo-controlled trials, n=1,172 | Dold 2024, PMID 38425206 |
| Escitalopram fixed dose | 10 mg and 20 mg both > placebo; 20 mg not superior to 10 mg; 10 mg > paroxetine 20 mg | 681 patients, 12 weeks | Baldwin 2006, PMID 16946363 |
| Escitalopram vs venlafaxine XR vs placebo | Venlafaxine XR separated on primary LOCF (−2.27, p=0.01); escitalopram did not (−1.52, p=0.09); both separated on observed cases | 392 randomised, 8 weeks | Bose 2008, PMID 18050245 |
| Duloxetine, paediatric GAD | PARS-GAD −9.7 vs −7.1 (p≤0.001, d=0.5); response 59% vs 42%; remission 50% vs 34% | 272 youths 7–17, 10 weeks | Strawn 2015, PMID 25791145 |
| Cost-effectiveness (UK NHS/PSS, 2011 prices) | Sertraline lowest cost, highest QALYs, 75% probability most cost-effective at £20,000/QALY | Decision model over 6 first-line drugs | Mavranezouli 2013, PMID 23512146 |
| MM120 (lysergide), single dose | HAM-A LS mean difference −5.0 (−9.6 to −0.4) at 100 µg; −6.0 (−9.8 to −2.0) at 200 µg; 25/50 µg not significant | Phase 2b, 198 randomised, 22 US sites | Robison 2025, PMID 40906494 |
9. Treatment effect sizes — psychological¶
| Comparison | Effect | n | Source |
|---|---|---|---|
| CBT vs treatment as usual | SMD −0.74 (−1.09 to −0.38), moderate certainty | 65 RCTs, 5,048 participants | Papola 2024, PMID 37851421 |
| Third-wave CBT vs TAU | SMD −0.76 (−1.15 to −0.36) | Same | Papola 2024, PMID 37851421 |
| Relaxation therapy vs TAU | SMD −0.59 (−1.07 to −0.11); −0.47 (−1.18 to 0.23), ns, excluding high-RoB studies | Same | Papola 2024, PMID 37851421 |
| Psychotherapy vs control (mostly waitlist) | g 0.84 (0.71–0.97); depression g 0.71 | 41 studies, 2,132 GAD patients | Cuijpers 2014, PMID 24487344 |
| Psychotherapy vs medication, pooled GAD | Psychotherapy g 0.76; medication g 0.38 — not directly comparable (different control types) | 79 RCTs, 11,002 GAD participants | Carl 2020, PMID 30760112 |
| CBT vs placebo (pooled anxiety disorders) | g 0.56 overall; GAD in the large-effect stratum; response OR 2.97 | 41 studies, N=2,843 | Carpenter 2018, PMID 29451967 |
| CBT vs controls at ≥12 months, GAD | Hedges' g 0.22 (k=10); 0.07–0.40 at earlier follow-ups | 69 RCTs, 4,118 outpatients | van Dis 2020, PMID 31758858 |
| Individual vs remote CBT | SMD 0.96 (0.13–1.79) favouring individual | 52 trials, 4,361 patients | Liu 2025, PMID 40506439 |
| Metacognitive therapy vs CBT | Recovery 65% vs 38% post-treatment; 57% vs 38% at 9 years; GAD re-diagnosis 9.5% vs 23.1% | 81 randomised (32/28/21); 39 of 60 at 9 y | Nordahl 2018, PMID 30294448; Solem 2021, PMID 34520637 |
| MBSR vs escitalopram (anxiety disorders) | CGI-S difference −0.07 (−0.38 to 0.23), non-inferior (margin −0.495); AEs 78.6% vs 15.4% | 276 randomised / 208 per-protocol | Hoge 2023, PMID 36350591 |
| Yoga vs stress education | Response 54.2% vs 33.0%, OR 2.46 (1.12–5.42), NNT 4.59; CBT 70.8%, OR 5.00, NNT 2.62; yoga not non-inferior to CBT | 226 with primary GAD | Simon 2021, PMID 32805013 |
| Internet-delivered treatment, GAD | Anxiety g 0.79; worry g 0.75 | 20 RCTs | Eilert 2021, PMID 33225589 |
| Digital interventions, GAD subgroup | g 0.62 (overall anxiety 0.80); guided 0.84 vs unguided 0.64 (ns); digital vs face-to-face 0.14 (−0.01 to 0.30) | 47 RCTs, 4,958 participants | Pauley 2023, PMID 34047264 |
| Digital CBT vs active psychoeducation | GAD-7 d 1.09 (0.81–1.37) at 10 wk; remission 71.0% vs 34.6% (OR 4.63, 2.85–7.54) | 351 randomised | Parsons 2025, PMID 41396602 |
| Psychological treatment effect on intolerance of uncertainty | Within-group g 0.88 (post), 1.05 (follow-up); between-group g 1.35; CBT-IU > general CBT at post | 26 studies, 1,199 GAD patients | Wilson 2023, PMID 37271039 |
| CBT in older adults vs working age | g 0.55 (0.22–0.88) vs 0.94 (0.52–1.36); no older-adult study used ITT | 15 studies, 22 comparisons, 770 patients | Kishita 2017, PMID 28119196 |
| Collaborative care (CALM) vs usual care | NNT 5.27 response, 5.50 remission at 12 months; response 63.66% vs 44.68% | 1,004 patients, 17 clinics | Roy-Byrne 2010, PMID 20483968 |
| Acupuncture vs sham | HAMA MD −2.71 (−4.17 to −1.25); GAD-7 −2.99 (−5.52 to −0.45); certainty low to very low | 14 RCTs, 968 participants | Jiang 2025, PMID 41312341 |
| Low-intensity CBT (self-help, guided self-help, psychoeducational groups) | Anxiety g=−0.63; depression −0.48; worry −0.64 | 12 RCTs; n=1,201 / 1,164 / 908 | Powell 2024, PMID 38166836 |
| Remote CBT, GAD-specific | Within-group pre-post g=1.30 (1.03–1.58); between-group g=0.76 (0.47–1.06); low- vs high-intensity not different | 10 studies, 1,071 participants | Trenoska Basile 2022, PMID 35403706 |
| Digital interventions, interview-confirmed generalized anxiety | g=0.80 (vs depression 0.62, PTSD 0.57, OCD 0.68, panic 1.05, specific phobia 1.18) | 168 studies, 22,144 patients | Harrer 2025, PMID 40703853 |
| Conversational agents, generalized anxiety symptoms | g=0.29 (0.21–0.36) | 32 RCTs, 6,089 participants | He 2023, PMID 37115595 |
| MI-augmented CBT vs CBT alone, severe GAD | No post-treatment difference; steeper follow-up decline in worry (γ=−0.13, p=0.03) and distress (γ=−0.12, p=0.01); ~5× odds of no longer meeting GAD criteria at 12 months; dropout 10% vs 23% | 85 randomised, allegiance-controlled | Westra 2016, PMID 26985729 |
| Acceptance-/mindfulness-based interventions, DSM-5 anxiety disorders | ACT and MBCT comparable to CBT; MBSR significantly lower; no differences from TAU or CBT at 6–12 months | 23 RCTs, 1,815 adults | Haller 2021, PMID 34650179 |
| TMS in GAD (cross-disorder analysis) | SMD −1.8 (−2.6 to −1.0), without significant heterogeneity | 208 RCTs across disorders | Hyde 2022, PMID 35365806 |
| NIBS modality comparison in GAD | cTBS and rTMS > tDCS, which showed no significant effect; more headaches in intervention groups | 7 RCTs | Qi 2024, PMID 39208534 |
| rTMS vs sham | SMD −1.857 (−2.219 to −1.494) | 6 studies, 152 patients | Parikh 2022, PMID 34791241 |
10. Course, relapse and prognosis¶
| Figure | Value | Population | Source |
|---|---|---|---|
| Five-year remission probability | 0.38 | 167 GAD patients, HARP clinic cohort | Yonkers 2000, PMID 10974960 |
| 12-year course | Chronic for the majority; recovered patients have high recurrence probability | HARP, prospective multicentre | Bruce 2005, PMID 15930067 |
| 20-year community outcome | 52% symptom-free, 29% subthreshold, 16% re-diagnosed GAD, 7% persistent | 75 of 105 DSM-III GAD cases, Zurich cohort | Angst 2009, PMID 18575915 |
| 2-year recovery, naturalistic cohort | 54.6%; ML prediction AUC 0.67 (sens 62.0%, spec 62.8%) from 569 predictors | 887 anxiety-disorder patients (NESDA) | Bokma 2022, PMID 32524918 |
| Personality disorder and remission | 30% lower likelihood (avoidant 34%, dependent 14%), independent of MDD | 514 patients, HARP | Massion 2002, PMID 11982447 |
| Relapse on placebo vs escitalopram 20 mg | 56% vs 19%; risk 4.04× higher on placebo | 375 randomised responders, 24–76 weeks | Allgulander 2006, PMID 16316482 |
| Relapse on placebo vs duloxetine | 41.8% vs 13.7% | 405 randomised responders, 26 weeks | Davidson 2008, PMID 18559291 |
| Relapse on placebo vs agomelatine | 30.7% vs 19.5% (log-rank p=0.046) | 227 randomised responders, 26 weeks | Stein 2012, PMID 22901350 |
10b. Cognition, electrophysiology and inflammation¶
| Measure | Finding | n | Source |
|---|---|---|---|
| Executive function | Poorer cognitive flexibility and working memory, not inhibitory control; absent in older adults; both slower RT and poorer accuracy | 32 studies, 13,084 participants | Nguyen 2025, PMID 40513710 |
| Longitudinal cognition | No consistent evidence of sustained cognitive decline | 13 studies | Li 2026, PMID 42191132 |
| ERP | P3 amplitude g=−0.54 (−0.70 to −0.38, I²=20%); ERN amplitude g=−0.42 (−0.72 to −0.12); delayed P3 (0.43), N2 (0.36) and MMN (0.63) latencies | 37 studies, 1,086 patients vs 1,315 controls | Xie 2025, PMID 39675130 |
| Metacognition | Negative beliefs about uncontrollability and danger of thoughts most prevalent in GAD specifically | 47 studies, 3,772 patients vs 3,376 controls | Sun 2017, PMID 28763680 |
| Inflammation | CRP Cohen's d=0.38 (0.06–0.69), I²=75%; CRP, IFN-γ and TNF-α raised in ≥2 studies; 5 of 14 studies found no difference in ≥1 cytokine | 14 studies, 1,188 patients vs 10,623 controls | Costello 2019, PMID 31326932 |
| Network bridge symptoms (PHQ/GAD-7) | Most central: sad mood, uncontrollable worry, worrying too much. Most frequent bridges: sad mood, restlessness, motor disturbance. Most robust edge: sleep–fatigue | 33 studies, 78,721 participants | Cai 2024, PMID 38238548 |
11. Measurement performance¶
| Instrument | Threshold | Sensitivity | Specificity | AUC | Source |
|---|---|---|---|---|---|
| GAD-7 (development) | ≥10 | 0.89 | 0.82 | — | Spitzer 2006, PMID 16717171 |
| GAD-7 (Cochrane pooled, GAD) | ≥10 | 0.64 (0.56–0.72) | 0.91 (0.87–0.93) | 0.86 (0.84–0.88) | Aktürk 2025, PMID 40130828 |
| GAD-7 (Cochrane pooled, any anxiety disorder) | ≥10 | 0.48 (0.40–0.57) | 0.91 (0.89–0.93) | 0.80 (0.78–0.82) | Aktürk 2025, PMID 40130828 |
| GAD-2 (Cochrane pooled, GAD) | ≥3 | 0.68 (0.59–0.75) | 0.86 (0.82–0.89) | 0.82 (0.81–0.86) | Aktürk 2025, PMID 40130828 |
| GAD-7 (single-country validation) | ≥10 | 0.87 | 0.78 | — | Muñoz-Navarro 2017, PMID 28666201 |
| GAD-7 (Rational Clinical Examination) | — | LR+ 5.1 (4.3–6.0); LR− 0.13 (0.07–0.25) | — | — | Herr 2014, PMID 25058220 |
| HADS-A (Cochrane pooled, GAD) | ≥8 | 0.82 (0.76–0.87) | 0.74 (0.70–0.77) | 0.82 (0.80–0.84) | Fomenko 2025, PMID 40600405 |
| GAD-7, perinatal (vs psychiatrist DSM-IV diagnosis) | ≥13 (optimal) | 0.613 | 0.727 | — | Simpson 2014, PMID 25161068 |
| GAD-7 MCID | — | 4 points | — | — | Toussaint 2020, PMID 32090765 |
| HAM-A severity bands | None/minimal ≤7; mild 8–14; moderate 15–23; severe ≥24 | — | — | — | Matza 2010, PMID 20718076 |
| GAD lifetime diagnosis reliability | CIDI test–retest κ 0.53; κ 0.78 without excessiveness; CIDI–SCID concordance κ 0.35 (0.66 without excessiveness) | — | — | — | Wittchen 1995, PMID 7666382 |
| DSM-IV vs ICD-10-DCR GAD agreement | κ 0.86 | — | — | — | Starcevic 1999, PMID 9885394 |
| Diagnostic reliability by method | Audio re-rating mean κ 0.80 vs independent re-interview mean κ 0.47, same clinic | — | — | — | Chmielewski 2015, PMID 26098046 |
12. Trial-methodology statistics¶
| Figure | Value | Source |
|---|---|---|
| Placebo pre-post effect, GAD | d_av 1.23 (1.06–1.41) — 2nd of 9 disorders, behind MDD 1.40 | Bschor 2024, PMID 38809560 |
| Placebo pre-post, comparison disorders | Panic/ADHD/PTSD/social phobia/mania 0.68–0.92; OCD 0.65; schizophrenia 0.59 | Bschor 2024, PMID 38809560 |
| SSRI vs SNRI improvement trajectory (57 trials, N=16,056) | SSRI linear, SNRI logarithmic across 12 weeks; higher SSRI (not SNRI) doses give greater benefit; higher doses of both increase AE dropout | Jakubovski 2019, PMID 30479005 |
| App uptake / adherence / attrition (79 trials) | Uptake 92.4%; adherence 61.8%; post-test attrition 18.6%; follow-up attrition 28.4% | Liu 2026, PMID 41259035 |
| Guidance and completion of digital programmes (22 studies) | Amount completed g 0.29 (0.18–0.40); full completion +12 percentage points | Musiat 2022, PMID 34802474 |
| GAD trial recruitment failure | Three-quarters of eligible patients declined, most refusing randomisation to medication; only ~12% of eligible patients identified; only 5% of those with GAD-7 ≥10 would have been eligible | Kalpakidou 2019, PMID 31126337 |
13. Safety statistics¶
| Harm | Value | Source |
|---|---|---|
| Antidepressant AE dropout, GAD | NNTH 17 (13–112) | Kopcalic 2025, PMID 39880377 |
| Falls, older adults (adjusted OR) | SSRIs 2.02 (1.85–2.20); antidepressants 1.57; antipsychotics 1.54; benzodiazepines 1.42; long-acting BZD 1.81; TCAs 1.41 | Seppala 2018, PMID 29402652 |
| Hyponatraemia | Antidepressant exposure OR 3.160 (1.911–5.225); SNRI 7.44% > SSRI 5.59% > TCA 2.66% > mirtazapine 1.02% | Gheysens 2024, PMID 38403888 |
| Bleeding under SSRIs | Adjusted OR 1.41 (1.27–1.57); 42 observational studies | Laporte 2017, PMID 27521835 |
| Discontinuation symptoms | ≥1 symptom 31% vs 17% placebo (RCT difference 8%); severe 2.8% vs 0.6% | Henssler 2024, PMID 38851198 |
| Discontinuation, symptom nature | DESS week 1 SMD 0.31 (≈1 extra symptom); dizziness OR 5.52 (3.81–8.01); mood worsening not associated | Kalfas 2025, PMID 40632531 |
| Paediatric antidepressant suicidality | RR 1.95 (1.28–2.98) all indications; risk difference 0.02; no completed suicides in 4,582 patients | Hammad 2006, PMID 16520440 |
| Pheochromocytoma screening false positives | 10–22% of urine normetanephrine results above the reference limit; only 0.7% of 12,572 patients compatible with likely PPGL | Kline 2020, PMID 31978379 |
| Caffeine challenge, panic disorder | Panic attacks in 51.1% of 237 patients after caffeine, 0% after placebo; patients 53.9% vs controls 1.7% (log RR 3.47) | Klevebrant 2022, PMID 34871964 |
| MM120 perceptual adverse events | Visual perceptual changes in 46.2% of dosed participants | Robison 2025, PMID 40906494 |
Known conflicts and caveats¶
- Prevalence is a criterion artefact. 2.6% vs 4.0% depending on the excessiveness rule (Ruscio 2024); 7.5% vs 3.0% depending on the duration rule (Lee 2009). Neither variation is measurement error.
- Four different quantities are routinely conflated: interview-diagnosed GAD (1.8% 12-month), screening-scale symptoms (18.2% past two weeks), administrative-code prevalence (7.4%), and pooled anxiety-disorder prevalence (4.4% globally). Every published "anxiety affects X%" claim belongs to one of these frames and to only one.
- Comorbidity rates depend on sampling frame and time horizon. 81.9% lifetime (WMH) and 3.8% pure vs 1.6% comorbid at a single consultation (Wittchen 2002) are both correct.
- Prognosis depends on where the cohort was recruited. 0.38 five-year remission in a clinic cohort (Yonkers 2000) versus 52% symptom-free in a community cohort (Angst 2009).
- Responder and continuous analyses of the same drug literature tell different stories. NNTB 7 for response (Kopcalic 2025) alongside a best-drug HAM-A difference of 3.6 points (Slee 2019), against a placebo arm improving by d_av 1.23 (Bschor 2024). Nobody has reconciled these.
- Psychotherapy and drug effect sizes are not comparable across control types. Carl 2020 reports psychotherapy g 0.76 vs medication 0.38 and explicitly warns against direct comparison; Bandelow 2015 reports the reverse ordering on pre-post effects and warns equally.
- Two evidence bases disagree about digital delivery. Digital ≈ face-to-face (g 0.14, ns; Pauley 2023) versus individual CBT > remote CBT (SMD 0.96; Liu 2025).
- Two evidence bases disagree about benzodiazepines. Efficacious with favourable safety within class (Fernandes 2025) versus poorly tolerated across drugs (Slee 2019). Neither measures dependence over months to years.
- Neurostimulation effect sizes are not credible as point estimates. rTMS SMD −1.86 from 152 patients (Parikh 2022); high-frequency rTMS response OR 291.40 with a 95% CI of 13.08–6490.21 (Duan 2025, PMID 40203547). The intervals are the finding.
- GBD does not model GAD. Every burden figure in section 3 is for anxiety disorders pooled.
- Guideline-concordant care has not been shown to improve outcomes. 39% concordance in a 721-patient Dutch primary-care cohort; concordant patients were more severe at baseline and improved equally at 12 months, at higher cost (Prins 2010, PMID 20049547; Prins 2011, PMID 22099636; Prins 2011, PMID 20586845).
- In paediatric anxiety, treatment type does not predict long-term remission — acute response does. 46.5% remission at ~6 years and 21.7% stable remission across 4 follow-up years, with treatment assignment non-predictive in both (Ginsburg 2014, PMID 24477837; Ginsburg 2018, PMID 29960692).
- The GAD-7's cultural bias runs in the counter-intuitive direction. Black/African American participants with high symptoms scored lower than others at equivalent symptom levels (Parkerson 2015, PMID 25725310), which would deflate, not inflate, measured prevalence in that group.
Last curated: 2026-09-02.