Intended for healthcare professionals


Drug treatment for generalised anxiety disorder

BMJ 2011; 342 doi: (Published 11 March 2011) Cite this as: BMJ 2011;342:d1216
  1. Toshi A Furukawa, professor
  1. 1Kyoto University Graduate School of Medicine/School of Public Health, Kyoto, 606-8501 Japan
  1. furukawa{at}

More head to head trials are needed to confirm apparent differences in effectiveness

Generalised anxiety disorder is characterised by excessive worrying over everyday things and is associated with irritability; restlessness; difficulties in concentrating; and somatic symptoms such as muscle tension, fatigue, or sleeplessness. In the linked systematic review (doi:10.1136/bmj.d1199) Baldwin and colleagues assess the relative effectiveness and tolerability of different drugs in the treatments of patients with this disorder.1

Generalised anxiety disorder first appeared in the American diagnostic classification system in 1980 as a residual category after diagnostic criteria for more specific anxiety disorders such as panic disorder, phobias, and obsessive-compulsive disorder had been delineated. Since then, conceptualisations of this disorder have been successively refined, and it is now generally recognised as an independent diagnostic entity. It can be distinguished from other, often coexisting, mental disorders, and its symptoms are uniquely associated with functional impairment and distress. Generalised anxiety disorder is one of the more common mental disorders in the general population, with 12 month prevalence estimates of around 3% and lifetime prevalence estimates of around 6%. It is also a common clinical diagnosis in primary care.2

Various drug and non-drug treatments have been shown to be effective including benzodiazepines, selective serotonin reuptake inhibitors, serotonin noradrenaline reuptake inhibitors, anticonvulsants, and cognitive behavioural therapy. The existence of several treatment options is clinically desirable but raises the difficult question of which is likely to be the most effective and acceptable for an individual patient.

Systematic reviews and meta-analyses of individual drugs can provide partial answers when they make head to head comparisons with active drugs or make comparisons with a common comparator such as placebo. However, evaluating relative superiority among alternative drugs requires larger sample sizes than is needed for comparisons with placebos. In other words, even with comprehensive searches for relevant trials and the use of meta-analytical pooling, the results will still often have wide confidence intervals that may encompass both the point of null effects and the point of minimally important clinical difference. Many reviews then erroneously conclude, or at least imply, that the two treatments have similar effectiveness.

A more refined use of the available evidence is now possible. Multiple treatment meta-analysis, also known as mixed treatment comparison or network meta-analysis, is a relatively new approach to systematic reviews, which combines evidence from both direct head to head comparisons and indirect comparisons via intermediate comparators. It preserves the comparison of randomised treatments within each trial and offers many advantages. Firstly, it can produce tighter confidence intervals than found with other analyses because it uses both direct and indirect estimates. Secondly, it can estimate relative effectiveness between treatments that have never been directly compared. In addition, when conducted within a Bayesian framework, it can rank treatments on the basis of—for example, the probability of each treatment being the best among all the alternatives.

Baldwin and colleagues’ analysis is the first multiple treatment meta-analysis of various drugs for generalised anxiety disorder. The authors found that fluoxetine seemed to be more effective than other treatments and sertraline was better tolerated. They recommend that, among five treatments specifically licensed for generalised anxiety disorder in the United Kingdom, duloxetine, escitalopram, and pregabalin may offer some advantages over venlafaxine and paroxetine.

Although the authors make the best possible use of the available evidence, some caveats are worth consideration. Firstly, of the 46 eligible trials, only half could contribute to the multiple treatment meta-analysis for response, defined as 50% or greater reduction in severity of anxiety, and less than a third to that for remission, defined as achieving minimally symptomatic status. This raises concerns about selective outcome reporting bias.3 Secondly, the available evidence was scanty, especially for direct comparisons between active drugs. For example, fluoxetine was ranked first in effectiveness to bring about response and remission, but data came from 33 patients in one trial only. Thirdly, as the authors argue, few significant differences were seen in terms of response among the active treatments. Yet they base their recommendations on the rankings according to the Bayesian model, which may in itself be a justifiable practice. However, their conclusion that of the five treatments having licensed indications for generalised anxiety disorder in the UK, duloxetine, escitalopram, and pregabalin offer some advantages over venlafaxine and paroxetine may be tenuous at best, particularly when the same rankings show that venlafaxine and paroxetine may be superior to duloxetine, escitalopram, and pregabalin in at least one of their three primary outcomes. Using GRADE criteria, the evidence for these relative rankings should be downgraded two or three levels for publication bias, imprecision, inconsistency, and indirectness.4

With these methodological caveats in mind, evidence suggests that fluoxetine and sertraline have some advantages over others in the short term treatment of generalised anxiety disorder. However, the weaknesses noted above make it difficult to draw firm conclusions. Researchers and the medical community need access to all the outcome data from all of the trials so that more robust multiple treatment meta-analyses can be done. Another unanswered clinical question awaiting rigorous pooling of available evidence is relative effectiveness and tolerability of these drugs in the long term.


Cite this as: BMJ 2011;342:d1216


  • Research, doi:10.1136/bmj.d1199
  • Competing interests: The author has completed the Unified Competing Interest form at and declares: no support from any organisation for the submitted work; he has received honorariums for speaking at CME meetings sponsored by Astellas, Dai-Nippon Sumitomo, Eli Lilly, GlaxoSmithKline, Janssen, Kyorin, MSD, Meiji, Otsuka, Pfizer, Shionogi, and Yoshitomi; he is on advisory boards for Sekisui Chemicals and Takeda Science Foundation; he has received royalties from Igaku-Shoin, Seiwa-Shoten, Nihon Bunka Kagaku-sha, and American Psychiatric Publication.

  • Provenance and peer review: Commissioned; not externally peer reviewed.


View Abstract

Log in

Log in through your institution


* For online subscription