NPC Archive Item: Meta-analysis supports SSRIs as initial drug treatment for anxiety

NOTE – This is an archive post from the NPC and has not been updated since first publication. Therefore, some hyperlinks may no longer be working.
MeReC Rapid Review NPC Logo

4th May 2011

This meta-analysis of 27 randomised controlled trials (RCTs) of nine drug treatments for generalised anxiety disorder (GAD) ranked sertraline first for tolerability and fluoxetine first for response and remission. Whilst there are limitations to the data, this study supports the drug treatment recommendations within the NICE guidance on generalised anxiety disorder.

Level of evidence
Level 2 (limited quality patient-oriented evidence) according to the SORT criteria.

Health professionals should continue to follow NICE guidance on generalised anxiety disorder and panic disorder, which was updated in January 2011. With regard to GAD, NICE recommends non-drug management at steps 1 and 2 of treatment. At step 3, i.e. where patients have GAD with marked functional impairment or GAD that has not improved after step 2 interventions (low-intensity psychosocial interventions), drug treatment is an alternative to individual high-intensity psychological intervention (cognitive behavioural therapy [CBT] or applied relaxation). If drug treatment is chosen at step 3, an SSRI, usually sertraline, should be offered first.

Benzodiazepines should not be offered for treatment of GAD in primary or secondary care except as a short-term measure during crises. Neither should antipsychotics be offered for treating GAD in primary care. When drug treatment is chosen, informed consent should be obtained and documented and the person monitored carefully for adverse reactions. A previous MeReC Rapid Review discusses the updated NICE anxiety guideline in more detail.

What is the background to this?
Published systematic reviews and meta-analyses have compared various psychotropic drugs used in GAD. However, only selected treatments (for example, benzodiazepines) have been compared, rather than all drug treatments available. This systematic review of 27 RCTs compared nine drug treatments for GAD (duloxetine▼, escitalopram, fluoxetine, lorazepam, paroxetine, pregabalin, sertraline, tiagabine, and venlafaxine) and ranked them in terms of response, remission and tolerability, using Bayesian probabilistic mixed treatment meta-analyses.

What does this study claim?
Fluoxetine was ranked highest for response (probability 62.9%) and remission (probability 60.6%). Sertraline was ranked highest in terms of tolerability (probability 49.3%); i.e. it was associated with the lowest percentage of withdrawals due to adverse effects. Of the five drugs licensed for GAD in the UK (duloxetine, escitalopram, paroxetine, pregabalin and venlafaxine), duloxetine was ranked first for response (but it only had a probability of 2.7% for being the most effective treatment), escitalopram was ranked first for remission (probability 26.7%) and pregabalin was ranked first for tolerability (probability 7.7%).

So what?
While this study represents some of the best evidence we have to guide our practice in this area, the results of this meta-analysis are based on very limited data. For example, even though fluoxetine was ranked highest for response and remission, only one RCT of fluoxetine contributed to the analysis. As outlined in an accompanying editorial, this amounted to only 33 patients. The editorial also expresses concerns about selective reporting bias because only half of the 46 trials deemed eligible could be included in the multiple treatments meta-analysis for response and less than a third were included for remission. In addition, only 6 out of the 27 RCTs included, were direct comparisons of active treatment. Most compared active treatment only with placebo, which gives highly variable response rates in GAD. Although GAD is considered to be a long-term condition, only the initial 6 to 8 weeks of treatment was assessed, so this does not provide information on treatment over longer periods. The authors point out that, in contrast to clinical studies, the results relating to differing measures of response and remission do not correlate well. This might be a result of comparing outcomes in the meta-analysis across a broad range of clinical settings, where patient symptom severities and treatments differed. They also indicate that all studies were sponsored by pharmaceutical companies, which might be a source of bias. Therefore, when considering the importance of this study, it is worth noting the conclusions of the editorial, that “the evidence for these relative rankings should be downgraded…for publication bias, imprecision, inconsistency, and indirectness”.

Nevertheless, these findings are consistent with NICE guidance which recommends that if drug treatment is chosen (at step 3 – see Action above), an SSRI should be offered first. NICE also recommends considering offering sertraline first because it is the most cost-effective drug, but noting that it does not currently have a licence for GAD. This NICE recommendation was based on an economic analysis, informed by multiple treatment network meta-analyses conducted for NICE. The NICE meta-analyses, which looked at fewer drugs, also found sertraline to have the lowest probability of discontinuation due to side effects (which is consistent with the meta-analysis considered in this Rapid Review) as well as the second highest probability of response (duloxetine first) in people who did not discontinue treatment due to side effects. In addition, sertraline had the lowest relative risk versus placebo in terms of non-response in the classical NICE meta-analysis. On the other extreme, lorazepam was ranked worst for tolerability in the meta-analysis considered in this Rapid Review, which confirms the NICE recommendation not to offer benzodiazepines for treating GAD in primary or secondary care except as a short-term measure during crises.

Study details
Baldwin D, Woods R, Lawson R, et al. Efficacy of drug treatments for generalised anxiety disorder: systematic review and meta-analysis. BMJ 2011;342:d1199

Systematic review of 27 RCTs. In the primary analyses, treatments were ranked for effectiveness for each outcome measure using primary Bayesian probabilistic mixed treatment meta-analyses and given a percentage probability of being the most effective treatment. Secondary analyses (frequentist mixed treatment meta-analyses conducted with random effects model) were designed to support the validity of any findings identified in the primary analyses.

Adults aged >18 years receiving any drug treatment for GAD. Intervention and comparison
Nine drugs for GAD (duloxetine, escitalopram, fluoxetine, lorazepam, paroxetine, pregabalin, sertraline, tiagabine, and venlafaxine) were compared with one another and placebo. Most RCTs included in the meta-analysis compared active treatment with placebo; only 6 RCTs were direct comparisons.

Outcomes and results
The three primary outcomes were response, remission and adverse events. Response was the proportion of patients who experienced a reduction of at least 50% from their baseline score on the Hamilton Anxiety Scale (HAM-A). Remission was the proportion of patients with a final score <7. Tolerability was defined as the percentage of patients withdrawing from the trial because of adverse events.

In the primary probabilistic analysis, fluoxetine was ranked first for response (probability 62.9%) and remission (probability 60.6%). In terms of tolerability, sertraline was ranked first (probability 49.3%), whereas lorazepam was ranked last. In a sub analysis of UK licensed treatments, duloxetine was ranked first for response (but it only had a probability of 2.7% for being the most effective treatment), escitalopram was ranked first for remission (probability 26.7%) and pregabalin was ranked first for tolerability (probability 7.7%). Table 1 gives the ranking of all nine drugs in terms of these outcomes. The secondary frequentist mixed treatment meta-analyses showed that most active treatments were significantly favoured over placebo in terms of response and remission but placebo was favoured over most active treatments in terms of tolerability. Few significant differences were seen between active treatments in response and remission, but this might be expected because many of the studies were only powered to show superiority to placebo and non-inferiority to active treatment.

Table 1 Ranking of treatments by outcome measure according to probabilistic analysis, based on currently available evidence for GAD

Ranking Response Remission Withdrawal
1 Fluoxetine* Fluoxetine* Sertraline*
2 Lorazepam* Escitalopram Pregabalin
3 Duloxetine Venlafaxine Fluoxetine*
4 Sertraline* Paroxetine Paroxetine
5 Paroxetine Sertraline* Tiagabine*
6 Pregabalin Duloxetine Venlafaxine
7 Venlafaxine Tiagabine* Escitalopram
8 Escitalopram NA Duloxetine
9 Tiagabine* NA Lorazepam*

NA = not available (not all studies reported on this outcome)
* Not licensed for GAD in the UK

This study was funded by Lundbeck

More information on anxiety can be found within the anxiety NPC eLearning materials.

Please comment on this rapid review in the NPC discussion rooms, or using our feedback form.

Make sure you are signed up to NPC Email updates — the free email alerting system that keeps you up to date with the NPC news and outputs relevant to you