Antidepressants versus placebo for generalised anxiety disorder (GAD)

Abstract

Background

Generalised anxiety disorder (GAD) is a mental health condition characterised by excessive anxiety and worry about everyday events. GAD is a common disorder and generally affects women twice as often as men. Treatments include various psychological and pharmacological therapies. Among the pharmacological therapies, antidepressants, in particular, selective serotonin reuptake inhibitors (SSRIs) and serotonin‐noradrenaline reuptake inhibitors (SNRIs), are commonly used for the treatment of GAD and many studies have shown their benefit over placebo. Only one systematic review and meta‐analysis comparing all antidepressants to placebo has been done in the past. Since then, new data on existing antidepressants have emerged and new antidepressants have been introduced. An updated and more comprehensive review is needed to provide a stronger understanding of the efficacy, acceptability, tolerability, and impact on the quality of life of the various types of antidepressants compared to placebo.

Objectives

To assess the effects of antidepressants in GAD in adults, specifically: to determine the efficacy of antidepressants in alleviating symptoms of GAD compared to placebo and to review the acceptability of antidepressants in GAD in terms of adverse effects, including the general prevalence of adverse effects compared to placebo.

Search methods

We searched the Cochrane Common Mental Health Disorders (CCMD) register, CENTRAL, MEDLINE, Embase, PsycINFO, and two trials registers in October 2022.

Selection criteria

We included randomised controlled trials (RCT) or cluster‐RCTs that randomly assigned participants to receive either an antidepressant or placebo for the treatment of GAD. There were no restrictions on dose, frequency, intensity, or duration of treatment. The studies included adults of either sex with a primary diagnosis of GAD and without any serious medical comorbidities. Psychiatric comorbidities were allowed as long as GAD was the primary diagnosis. We excluded studies investigating psychotherapies and those that included participants who had regular use of benzodiazepines. There were no restrictions on setting, country, or language.

Data collection and analysis

Two review authors independently checked eligibility and extracted data following standard Cochrane methodological procedures. We assessed risk of bias using the Cochrane RoB 1 tool. A third review author resolved disagreements between the two primary review authors. We extracted study characteristics, participant characteristics, intervention details, settings, and outcome measures regarding efficacy, acceptability, tolerability, and quality of life. We used GRADE to assess the certainty of the evidence.

Main results

We included 37 unique RCTs with 12,226 participants in the review. The studies included adults with moderate‐severe GAD and without any serious medical comorbidities. Few studies included participants with secondary psychiatric comorbidities. The double‐blind treatment duration ranged from four weeks to 28 weeks.

Antidepressants have a benefit over placebo on rate of treatment response measured as a reduction of at least 50% on the Hamilton Anxiety Rating Scale (HAM‐A) (risk ratio (RR) 1.41, 95% confidence interval (CI) 1.29 to 1.55; 20 studies, 7267 participants; high‐certainty evidence). The magnitude of effect corresponds to a number needed to treat for an additional beneficial outcome (NNTB) of 7 (95% CI 5 to 9). Antidepressants have no difference in acceptability compared to placebo, measured as the number of participants who dropped out during the trial as a proportion of the total number of randomised participants (RR 1.03, 95% CI 0.93 to 1.14; 33 studies, 11,294 participants; high‐certainty evidence). Fewer participants dropped out due to a lack of efficacy in the antidepressant group compared to the placebo group (RR 0.41, 95% CI 0.33 to 0.50; 29 studies, 11,007 participants; high‐certainty evidence) with an NNTB of 27 (95% CI 24 to 32), and more participants dropped out due to adverse effects in the antidepressant group compared to placebo (RR 2.18, 95% CI 1.81 to 2.61; 32 studies, 11,793 participants; high‐certainty evidence) with a number needed to treat for an additional harmful outcome (NNTH) of 17 (95% CI 13 to 112). We observed similar findings when classes of antidepressants were compared with placebo. The certainty of the evidence for the analyses comparing different classes of antidepressants to placebo was high.

Authors' conclusions

This review added to the growing literature on antidepressants in the treatment of GAD. We have high confidence that antidepressants are more effective than placebo at improving treatment response and that antidepressants have similar acceptability to placebo.

Fewer participants dropped out due to a lack of efficacy in the antidepressant group compared to the placebo group and more participants dropped out due to adverse effects in the antidepressant group compared to placebo. We are highly confident in this evidence.

This review identified some important gaps in the literature on antidepressants for GAD and can be used as a tool to guide future research. Future studies may be more transparent with their methodology and outcome reporting. Future reviews may also include people with comorbidities, and explore other sources of heterogeneity.

Plain language summary

Antidepressants for generalised anxiety disorder

Key messages

– Antidepressants are more effective than placebo (pretend tablet) at improving treatment response and have similar acceptability to placebo. Fewer participants dropped out due to a lack of effectiveness in the antidepressant group compared to the placebo group and more participants dropped out due to unwanted effects in the antidepressant group compared to placebo.

– Future studies may be more transparent with their methods and outcome reporting. Future reviews may also include people with co‐occurring medical conditions (comorbidities).

What is generalised anxiety disorder?

Generalised anxiety disorder is a mental health condition characterised by excessive anxiety and ongoing worry about everyday events. Generalised anxiety disorder is common and generally affects women twice as often as men.

How is generalised anxiety disorder treated?

Treatments include various psychological approaches (which work on the mind and a person's behaviour) and medicines. Among the medicines, antidepressants (which are used to treat depression), in particular, two types of antidepressants called selective serotonin reuptake inhibitors and serotonin‐noradrenaline reuptake inhibitors, are commonly used for the treatment of generalised anxiety disorder and many studies have shown their benefit over a sham treatment (known as placebo).

Who will be interested in this review?

People with generalised anxiety disorder, general practitioners, and mental health professionals.

What did we want to find out?

This review aimed to provide an updated summary of all the evidence available on this topic. In particular, we wanted to find out:

– how effective antidepressants are compared to placebo in treating generalised anxiety disorder;

– how acceptable antidepressants are compared to placebo in treating generalised anxiety disorder; and

– how many unintended and harmful effects antidepressants have compared to placebo in people with generalised anxiety disorder.

What did we do?

We searched for studies that compared antidepressants to placebo for treating adults with generalised anxiety disorder but no other serious co‐occurring medical conditions.

What did we find?

We found 37 studies that involved 12,226 adults with generalised anxiety disorder. The studies lasted between four and 28 weeks.

Antidepressants were more effective than placebo in reducing anxiety and there was no difference between antidepressants and placebo in the total number of people leaving the studies early.

Fewer people in the antidepressant group dropped out of the studies early due to the antidepressant being considered ineffective compared to placebo and more people in the antidepressant group dropped out of the studies early because of unwanted effects compared to placebo.

What are the limitations of the evidence?

We are confident in our findings as they apply to people with generalised anxiety disorder but without other co‐occurring medical conditions. However, the evidence is not strong enough to create a clinical guideline for people with other co‐occurring medical conditions as they were excluded from our analyses.

How up to date is this evidence?

The evidence is up to date to October 2022.

Summary of findings

Summary of findings 1. Summary of findings table ‐ All antidepressants compared to placebo for generalised anxiety disorder.

All antidepressants compared to placebo for generalised anxiety disorder
Patient or population: generalised anxiety disorder Setting: inpatient and outpatient Intervention: all antidepressants Comparison: placebo
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with placebo	Risk with all antidepressants
Rate of treatment response measured as a reduction of ≥ 50% on the Hamilton Anxiety Rating Scale (HAM‐A) follow‐up: range 6 weeks to 24 weeks	396 per 1000	559 per 1000 (511 to 614)	RR 1.41 (1.29 to 1.55)	7267 (20 RCTs)	⊕⊕⊕⊕ High	An RR of 1.41 means that treatment with an antidepressant increased the rate of response to treatment by 41% compared to treatment with a placebo. We have high confidence in this evidence.
Acceptability: defined as the total number of dropouts follow‐up: range 4 weeks to 28 weeks	252 per 1000	260 per 1000 (234 to 287)	RR 1.03 (0.93 to 1.14)	11294 (33 RCTs)	⊕⊕⊕⊕ High	An RR close to 1 means that the risk of dropping out of the study early was no different between treatment with an antidepressant and treatment with a placebo. We have high confidence in this evidence
Dropouts due to a lack of efficacy follow‐up: range 6 weeks to 28 weeks	63 per 1000	26 per 1000 (21 to 32)	RR 0.41 (0.33 to 0.50)	11007 (29 RCTs)	⊕⊕⊕⊕ High	An RR of 0.41 means that treatment with an antidepressant decreased the risk of dropping out of the study early because of a lack of efficacy by 59% compared to treatment with a placebo. We have high confidence in this evidence.
Dropouts due to adverse effects follow‐up: range 6 weeks to 28 weeks	50 per 1000	108 per 1000 (90 to 129)	RR 2.18 (1.81 to 2.61)	11793 (32 RCTs)	⊕⊕⊕⊕ High	An RR of 2.18 means that treatment with an antidepressant increased the risk of dropping out of the study early because of adverse effects by 118% compared to treatment with a placebo. We have high confidence in this evidence.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_442518988713638742.

Summary of findings 2. Summary of findings table ‐ Selective serotonin reuptake inhibitors (SSRIs) compared to placebo for generalised anxiety disorder.

Selective serotonin reuptake inhibitors (SSRIs) compared to placebo for generalised anxiety disorder
Patient or population: generalised anxiety disorder Setting: inpatient and outpatient Intervention: selective serotonin reuptake inhibitors (SSRIs) Comparison: placebo
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with placebo	Risk with selective serotonin reuptake inhibitors (SSRIs)
Rate of treatment response measured as a reduction of ≥ 50% on the Hamilton Anxiety Rating Scale (HAM‐A) follow‐up: range 8 weeks to 12 weeks	386 per 1000	583 per 1000 (464 to 734)	RR 1.51 (1.20 to 1.90)	1226 (4 RCTs)	⊕⊕⊕⊕ High	An RR of 1.51 means that treatment with an SSRI increased the rate of response to treatment by 51% compared to treatment with a placebo. We have high confidence in this evidence.
Acceptability: defined as the total number of dropouts follow‐up: range 4 weeks to 12 weeks	214 per 1000	227 per 1000 (203 to 255)	RR 1.06 (0.95 to 1.19)	5031 (16 RCTs)	⊕⊕⊕⊕ High	An RR close to 1 means that the risk of dropping out of the study early was no different between treatment with an SSRI and treatment with a placebo. We have high confidence in this evidence.
Dropouts due to a lack of efficacy follow‐up: range 8 weeks to 12 weeks	42 per 1000	23 per 1000 (16 to 34)	RR 0.55 (0.38 to 0.79)	4832 (14 RCTs)	⊕⊕⊕⊕ High	An RR of 0.55 means that treatment with an SSRI decreased the risk of dropping out of the study early because of a lack of efficacy by 45% compared to treatment with a placebo. We have high confidence in this evidence.
Dropouts due to adverse effects follow‐up: range 8 weeks to 12 weeks	43 per 1000	86 per 1000 (65 to 113)	RR 1.98 (1.51 to 2.61)	5315 (16 RCTs)	⊕⊕⊕⊕ High	An RR of 1.98 means that treatment with an SSRI increased the risk of dropping out of the study early because of adverse effects by 98% compared to treatment with a placebo. We have high confidence in this evidence.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_442534668968434818.

Summary of findings 3. Summary of findings table ‐ Serotonin‐noradrenaline reuptake inhibitors (SNRIs) compared to placebo for generalised anxiety disorder.

Serotonin‐noradrenaline reuptake inhibitors (SNRIs) compared to placebo for generalised anxiety disorder
Patient or population: generalised anxiety disorder Setting: inpatient and outpatient Intervention: serotonin‐noradrenaline reuptake inhibitors (SNRIs) Comparison: placebo
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with placebo	Risk with serotonin‐noradrenaline reuptake inhibitors (SNRIs)
Rate of treatment response measured as a reduction of ≥ 50% on the Hamilton Anxiety Rating Scale (HAM‐A) follow‐up: range 8 weeks to 24 weeks	411 per 1000	551 per 1000 (497 to 604)	RR 1.34 (1.21 to 1.47)	4659 (14 RCTs)	⊕⊕⊕⊕ High	An RR of 1.34 means that treatment with an SNRI increased the rate of response to treatment by 34% compared to treatment with a placebo. We have high confidence in this evidence.
Acceptability: defined as the total number of dropouts follow‐up: range 8 weeks to 28 weeks	314 per 1000	323 per 1000 (273 to 380)	RR 1.03 (0.87 to 1.21)	4863 (15 RCTs)	⊕⊕⊕⊕ High	An RR close to 1 means that the risk of dropping out of the study early was no different between treatment with an SNRI and treatment with a placebo. We have high confidence in this evidence.
Dropouts due to a lack of efficacy follow‐up: range 8 weeks to 28 weeks	87 per 1000	29 per 1000 (22 to 38)	RR 0.33 (0.25 to 0.43)	4775 (13 RCTs)	⊕⊕⊕⊕ High	An RR of 0.33 means that treatment with an SNRI decreased the risk of dropping out of the study early because of a lack of efficacy by 67% compared to treatment with a placebo. We have high confidence in this evidence.
Dropouts due to adverse effects follow‐up: range 8 weeks to 28 weeks	65 per 1000	156 per 1000 (117 to 208)	RR 2.42 (1.81 to 3.22)	5078 (14 RCTs)	⊕⊕⊕⊕ High	An RR of 2.42 means that treatment with an SNRI increased the risk of dropping out of the study early because of adverse effects by 142% compared to treatment with a placebo. We have high confidence in this evidence.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_442534843379916232.

Summary of findings 4. Summary of findings table ‐ Other antidepressants compared to placebo for generalised anxiety disorder.

Other antidepressants compared to placebo for generalised anxiety disorder
Patient or population: generalised anxiety disorder Setting: inpatient and outpatient Intervention: other antidepressants Comparison: placebo
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with placebo	Risk with other antidepressants
Rate of treatment response measured as a reduction of ≥ 50% on the Hamilton Anxiety Rating Scale (HAM‐A) follow‐up: range 8 weeks to 12 weeks	374 per 1000	575 per 1000 (426 to 777)	RR 1.54 (1.14 to 2.08)	1804 (5 RCTs)	⊕⊕⊕⊕ High	An RR of 1.54 means that treatment with an antidepressant from the 'other' category increased the rate of response to treatment by 54% compared to treatment with a placebo. We have high confidence in this evidence.
Acceptability: defined as the total number of dropouts follow‐up: range 8 weeks to 12 weeks	210 per 1000	189 per 1000 (126 to 282)	RR 0.90 (0.60 to 1.34)	1830 (5 RCTs)	⊕⊕⊕⊕ High	An RR close to 1 means that the risk of dropping out of the study early was no different between treatment with an antidepressant from the 'other' category and treatment with a placebo. We have high confidence in this evidence.
Dropouts due to a lack of efficacy follow‐up: range 8 weeks to 12 weeks	72 per 1000	38 per 1000 (20 to 69)	RR 0.52 (0.28 to 0.95)	1826 (5 RCTs)	⊕⊕⊕⊕ High	An RR of 0.52 means that treatment with an antidepressant from the 'other' category decreased the risk of dropping out of the study early because of a lack of efficacy by 48% compared to treatment with a placebo. We have high confidence in this evidence
Dropouts due to adverse effects follow‐up: range 8 weeks to 12 weeks	23 per 1000	53 per 1000 (31 to 94)	RR 2.29 (1.31 to 4.01)	1826 (5 RCTs)	⊕⊕⊕⊕ High	An RR of 2.29 means that treatment with an antidepressant from the 'other' category increased the risk of dropping out of the study early because of adverse effects by 129% compared to treatment with a placebo. We have high confidence in this evidence.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_442534982260661819.

Background

Description of the condition

Generalised anxiety disorder (GAD) is a common psychiatric disorder. The key symptom is disproportionate worry about many, if not most, activities and concerns of daily living (Tyrer 2006). It is associated with somatic symptoms of anxiety, such as trembling, shaking, poor co‐ordination, muscle ache, sweating, nausea, diarrhoea, and an exaggerated startle response (APA 2000). Other symptoms include being easily fatigued and having difficulty concentrating. People with GAD can be irritable and have disturbed sleep (APA 2000). GAD follows a remitting pattern over decades, with periods of disability (Angst 1991; Angst 2009). It has been associated with considerable comorbidity (Wittchen 2002), decreased quality of life, and significant disability (Revicki 2012).

The point prevalence of GAD has been estimated at 1.6%, the 12‐month prevalence at 3.1%, and the lifetime prevalence at 5.1% in the US National Co‐morbidity Survey (Wittchen 2002). The lifetime prevalence in US adolescents in the USA was 3.0% (standard error (SE) 0.6%) for women and 1.5% (SE 0.3%) for men; among the total adolescent population, 0.9% (SE 0.2%) were considered to be severely impaired due to GAD (Merikangas 2010). GAD is more common in women; in those who are from lower socioeconomic groups; or those who are widowed, separated, or divorced (Grant 2005). The female‐to‐male ratio is 3:1 (Lim 2005); men are more likely to have co‐occurring substance abuse than women (odds ratio 3.40, 95% confidence interval (CI) 2.59 to 4.47) (Alegria 2010). However, these rates need to be interpreted with caution. Two‐step designs (a screening test and then a clinical interview) reveal different results. For example, in a survey of primary care patients, 14% of 7900 people met the criteria using a structured screening interview (Romera 2010). However, a clinician who performed the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders (DSM)‐IV (SCID) estimated the point prevalence in primary care at 3.8% (Serrano‐Blanco 2010). One study in Finland among primary care patients found that GAD was present in 4% of people participating in the study (Kujanpää 2014).

A series of treatments has been used for people who have GAD. These include non‐pharmacological treatments, such as applied relaxation and cognitive therapy (Covin 2008; Hofmann 2008; Hunot 2007; Mitte 2005; Otte 2011), and pharmacological treatments including antidepressants (Schmitt 2005), azapirones (Chessick 2006), benzodiazepines, hydroxyzine (Guaiana 2010), the herb kava (Sarris 2011), pregabalin (Samuel 2011), and antipsychotics (Depping 2010; LaLonde 2011; Maher 2011). The herb valerian has been studied in clinical trials, but the results are equivocal (Miyasaka 2006). In general, cognitive therapy is seen as one of the possible treatments (Gale 2011), followed by antidepressants, particularly fluoxetine and sertraline (Baldwin 2011). In older people, psychological and pharmacological interventions have similar effect sizes (Gonçalves 2012).

Description of the intervention

Antidepressants, in particular selective serotonin reuptake inhibitors (SSRIs) and serotonin‐noradrenaline reuptake inhibitors (SNRIs), are being used more frequently in people with GAD (Baldwin 2012a). The main reason antidepressants have become more frequently used is they are unlikely to cause dependence (Rheinold 2011). Recent guidelines consider antidepressants, mainly SSRIs, as the first‐line treatment for GAD due to their more favourable adverse effect profile over benzodiazepines (Baldwin 2014; Katzman 2014). However, antidepressant use also has problems. Adverse effects can occur, such as gastrointestinal (nausea, diarrhoea, dry mouth), agitation, and sexual adverse effects (Schatzberg 2015). Also, antidepressants can be associated with worsening of suicidal ideation, especially in younger people (Termorshuizen 2015).

How the intervention might work

Antidepressants work by augmenting the function of the monoamines serotonin or noradrenaline, or both. Serotonergic antidepressants (SSRIs such as fluoxetine, paroxetine, sertraline, and citalopram) promote the transmission of the neurotransmitter serotonin across brain synapses, most notably in the dorsal raphe nucleus (Briley 1993). They prevent reuptake of serotonin into nerve terminals by inhibiting serotonin transporters, thus allowing more serotonin to be available for neurotransmission. These modifications may remove stress‐induced brain‐derived neurotrophic factor inhibition, and may allow for reconstruction of the pre‐episode neural networks (Doron 2014). Clinically, this could manifest as a slow reduction in anxiety and worry. Antidepressants can also work on forebrain structures, such as medial prefrontal cortex, insula, and amygdala (Graeff 2010). There is some evidence that antagonism at serotonin 5‐HT2C may reduce anxiety (Graeff 2010).

Why it is important to do this review

Antidepressants, in particular SSRIs, continue to be used to treat people with GAD (Baldwin 2012a). Guidelines indicate that antidepressants are first‐line pharmacological treatment for GAD (Baldwin 2014; Katzman 2014). Published RCTs have shown some evidence of efficacy. However, no systematic review on all antidepressants in GADs has been conducted recently. One review on pharmacological treatments in GAD focused on published studies and examined efficacy and tolerability of all treatments in people with GAD (Baldwin 2011). It concluded that fluoxetine is the first drug for response and remission, while sertraline was the first for tolerability. Overall, the study authors concluded that SSRIs are the most effective drug treatment option for GAD. Another relevant meta‐analysis on antidepressants in people with GAD was published in 2005 (Schmitt 2005). The study authors concluded that antidepressants are superior to placebo in GAD, with a number needed to treat for an additional beneficial outcome (NNTB) of 5, and are also well‐tolerated. They recommended that further studies should be conducted to determine which antidepressant should be used for which patient. More studies have been published that may refine the findings of Schmitt 2005 and help expand our knowledge base on antidepressants in GAD. An up‐to‐date review is needed to help prescribers identify the effect size of active treatment compared to placebo in GAD for short‐ and long‐term treatment, in order to be better guided in the choice of the pharmacological agent.

Objectives

To assess the effects of antidepressants in GAD in adults, specifically: to determine the efficacy of antidepressants in alleviating symptoms of GAD compared to placebo and to review the acceptability of antidepressants in GAD in terms of adverse effects, including the general prevalence of adverse effects compared to placebo.

Methods

Criteria for considering studies for this review

Types of studies

All randomised controlled studies (RCTs), including cluster‐randomised studies. We excluded cross‐over RCTs and relapse prevention studies.

Types of participants

Adults (as defined by the study authors) of either sex, with a primary diagnosis of GAD. Studies that adopted any criteria to define participants with GAD were included. More recent studies are likely to have used Diagnostic and Statistical Manual of Mental Disorders. 4th, Text Revision edition (DSM‐IV‐TR) (APA 2000), Diagnostic and Statistical Manual of Mental Disorders. 4th edition (DSM‐IV) (APA 1994), Diagnostic and Statistical Manual of Mental Disorders. 5th edition (DSM‐5) (APA 2013), or International Classification of Diseases ICD‐10 criteria (WHO 1992). Older studies may have used International Classification of Diseases ICD‐9 criteria (ICD‐9) (WHO 1978), Diagnostic and Statistical Manual of Mental Disorders. 3rd edition (DSM‐III) (APA 1980)/Diagnostic and Statistical Manual of Mental Disorders. 3rd edition Revised (DSM‐III‐R) (APA 1987), or other diagnostic systems. We excluded studies that included participants who had a concurrent diagnosis of a medical disorder. However, we included studies that had participants with co‐occurring psychiatric disorders if the primary diagnosis was GAD. We included studies from any setting.

Types of interventions

Experimental

Any trial that compared antidepressants as monotherapy with placebo in the treatment of GAD.

The antidepressants included:

• tricyclic antidepressants (TCAs): amitriptyline, amoxapine, clomipramine, desipramine, dosulepin/dothiepin, doxepin, imipiramine, lofepramine, maprotiline, nortriptyline, proptriptyline, trimipramine;

• SSRIs: fluoxetine, fluvoxamine, sertraline, citalopram, paroxetine, escitalopram;

• monoamine oxidase inhibitors (MAOIs): phenelzine, isocarboxazide, tranylcypromine, moclobemide, brofaromine;

• SNRIs: venlafaxine, desvenlafaxine, duloxetine, milnacipran;

• noradrenergic and specific serotonergic antidepressants (NaSSAs): mirtazapine;

• noradrenergic and dopaminergic reuptake inhibitors (NDRIs): bupropion;

• noradrenergic reuptake inhibitors (NRIs): reboxetine;

• others: agomelatine, trazodone, nefazodone, mianserin, maprotiline, non‐conventional herbal products (e.g. hypericum), vilazodone, vortioxetine.

We included studies in which irregular (i.e. not daily) use of benzodiazepines took place. We excluded studies in which benzodiazepines were regularly administered at a constant dosage for a long time or as part of the study medication. We planned to note possible differences in co‐interventions (such as differential usage of benzodiazepines in antidepressant trials) and planned to examine their influence in sensitivity analyses.

There were no restrictions on dose, frequency, intensity, or duration. We did not include studies with study arms that were outside approved dosing ranges.

We excluded studies that administered psychosocial therapies targeted at GAD. We also excluded studies that had participants with mixed anxiety and depression. We excluded studies where participants had a medical comorbidity.

Control

• Placebo.

Types of outcome measures

Primary outcomes

• Rate of treatment responsemeasured as a reduction of at least 50% on the Hamilton Anxiety Rating Scale (HAM‐A) (Hamilton 1959)

• Acceptability (number of dropouts): number of participants who dropped out during the trial as a proportion of the total number of randomised participants (total dropout rate)

Secondary outcomes

• Rate of treatment response (defined by study authors)

• Remission rate as measured by:

  • the number of participants showing 17 or less on the 14‐item HAM‐A (Hamilton 1959)

  • any other similar cut‐off value on an anxiety scale, depending on the study authors' definition

  • 'not ill or borderline mentally ill' (a score one or two) on Clinical Global Impression (CGI) – Severity (Guy 1976) or

  • according to authors' definition of remitters at follow‐up

• Change in symptom levels. Group mean scores at the end of the trial or changes from baseline on:

  • HAM‐A

  • Any other anxiety scale (example Covi scale) or

  • CGI – Severity scale (Guy 1976) (using standardised mean difference)

• General and specific adverse effects:

  • Total number of participants reporting adverse effects

  • Specific adverse events

    • sleepiness/drowsiness

    • falls

    • hypotension

    • agitation/anxiety

    • suicide wishes/gestures/attempts

    • completed suicide

    • subjective memory impairment

• Average score/change in quality of life/satisfaction

• Death

• Total number of participants experiencing withdrawal symptoms

• Dropouts due to a lack of efficacy

• Dropouts due to adverse effects

Timing of outcome assessment

Whenever possible, we used the end of the double‐blind period as the point of assessment. If there was a period of open‐label follow‐up, this was discounted, including periods of withdrawal of medication (which functionally became single‐blind). Whenever possible, we used intention‐to‐treat analyses. We included data up to the end of the assessment period of the trial.

Hierarchy of outcome measures

As our first primary outcome, we used rate of treatment response using, whenever possible, defined criteria, usually a 50% reduction of HAM‐A or a HAM‐A score of less than 7, or alternatively, response as defined by the study author(s). We preferred to use a more‐defined outcome as this would, in general, decrease bias due to varying instruments. Our other primary outcome was dropout rate, as total number.

We also included, when available, dropout rates due to a lack of effectiveness or due to adverse effects.

We included these outcomes in the summary of findings tables.

Our secondary outcomes were specific scales, particularly HAM‐A change. Other scales, when available, were discussed measures of functional gain and specific adverse effects.

Search methods for identification of studies

Electronic searches

Cochrane Common Mental Disorders register

The Cochrane Common Mental Disorders (CCMD) Group maintained a specialised register of RCTs, the CCMD‐CTR, until 31 March 2023. This register contained over 39,000 reference records (reports of RCTs) for depression, anxiety, and other common mental disorders. A percentage of the reference records were tagged to 12,500 individual, PICO (patient/population, intervention, comparison and outcomes)‐coded study records (with coding based on the EU‐Psi coding manual). Reports of RCTs for inclusion in the register were collated from (weekly) generic searches of MEDLINE, Embase, and PsycINFO; quarterly searches of the Cochrane Central Register of Controlled Trials (CENTRAL); and review‐specific searches of additional databases. Reports of trials were also sourced from international trial registries, drug companies, the handsearching of key journals, conference proceedings, and other (non‐Cochrane) systematic reviews and meta‐analyses.

We searched the CCMD‐CTR‐Studies section of the register using the following controlled vocabulary terms: Diagnosis = "generalized anxiety disorder" and Intervention = placebo*

We searched the CCMD‐CTR‐References register using a more sensitive list of terms for GAD together with antidepressant drug terms to identify additional uncoded study reports (Appendix 1).

We removed duplicates and screened records and, where necessary, retrieved full‐text articles to check for a placebo control arm.

Biomedical database searches

With the relocation of the CCMD Group to the University of York, the CCMD‐CTR is currently out‐of‐date. Therefore, we performed additional searches of the following databases on 22 October 2022.

• CENTRAL (the Cochrane Library; 2022, Issue 10)

• Ovid MEDLINE, Ovid MEDLINE In‐Process & Other Non‐Indexed Citations, Ovid MEDLINE Daily and Ovid OLDMEDLINE (1946 to 20 October 2022)

• Ovid Embase (1974 to 20 October 2022)

• Ovid PsycINFO (1806 to October week 3 2022)

Full search strategies are in Appendix 2.

National and international trial registers

We conducted complementary searches of the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) and ClinicalTrials.gov for unpublished or ongoing trials in October 2022.

Searching other resources

Bibliographies

We planned to examine references and bibliographies from the text of reports of relevant trials for further RCTs not yet identified.

Personal communication

We planned to consult leading researchers and authors of included studies to find out if they knew of any published, unpublished, or ongoing RCTs in the area that were not identified by other means.

We planned to contact the manufacturers of agents of interest to request additional data from unpublished studies.

Data collection and analysis

Selection of studies

Two review authors (KK and CC) independently screened abstracts/titles of the studies identified by the literature searches. We discussed any disagreement with a third review author (GG), documented the decisions, and, where necessary, contacted the study authors to help resolve the issue. Those studies that meet our inclusion criteria constituted the preliminary list and we retrieved their full texts.

Two review authors (KK and CC) independently assessed the full‐text articles in this preliminary list to see if they meet the inclusion criteria. We discussed any disagreement with a third review author (GG), documented the decisions, and, where necessary, contacted the study authors to help resolve the issue. We grouped publications of the same trial as multiple references to one study. We recorded this process in sufficient detail so we could construct a PRISMA diagram.

Data extraction and management

Two review authors (KK and CC) independently extracted data using a specially designed form. We discussed any disagreement with a third review author (GG), documented the decisions, and, where necessary, contacted the study authors to help resolve the issue. When the papers were not in English, we planned to have authors who were fluent in other languages extract data, and if no member of the research team was fluent in that language, we planned to obtain a translation of the paper via CCMD.

Main comparisons

The main comparisons were placebo versus the following comparators:

• all antidepressants (pooled);

• TCAs/heterocyclic antidepressants;

• SSRIs (e.g. citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, sertraline);

• SNRIs (e.g. duloxetine, milnacipran, venlafaxine);

• MAOIs (e.g. moclobemide, phenelzine, tranylcypromine);

• NDRIs (bupropion);

• other antidepressants.

• each type of antidepressant compared to placebo.

Assessment of risk of bias in included studies

Two review authors (KK and CC) independently assessed the risk of bias using the Cochrane RoB 1 tool. We discussed any disagreement with a third review author (GG), documented the decisions and, where necessary, contacted the study authors to help resolve the issue. We assessed the risk of bias following seven domains as per Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

• Method used to generate the allocation sequence should have been described in detail so that we could assess whether it should have produced comparable groups.

• Method used to conceal the allocation sequence should have been described in sufficient detail so that we could assess whether intervention schedules could have been predicted in advance of, or during, recruitment.

• We assessed whether the method by which any measures used to blind (mask) participants adequately limited the chance that any person from any group was able to ascertain which intervention a given participant might have received; this was done separately for different types of outcomes.

• We assessed whether the method by which any measures used to blind (mask) personnel adequately limited the chance that any person from any group was able to ascertain which intervention a given participant might have received; this was done separately for different types of outcomes.

• We assessed whether the method by which any measures used to blind (mask) outcome assessors adequately limited the chance that any person from any group was able to ascertain which intervention a given participant might have received; this was done separately for different types of outcome.

• If studies did not report intention‐to‐treat analyses, we made attempts to obtain missing data by contacting the study authors. We extracted and reported data on attrition and exclusions as well as the numbers involved (compared with total randomised), reasons for attrition/exclusion where reported or obtained from investigators, and any re‐inclusions in re‐analyses that we performed. We reported whether, in our judgement, incomplete data were dealt with adequately (see also Dealing with missing data).

• We attempted to assess the possibility of selective outcome reporting by investigators. We rated these items as 'high', 'low', or 'unclear' using the methods described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

• We considered and reported whether we believed there were any other factors that could put a study at a high risk of bias. We performed a sensitivity analysis and excluded any studies that were at a high risk of bias.

Measures of treatment effect

We analysed dichotomous outcomes by calculating risk ratios (RR) for each trial with the uncertainty in each result expressed using 95% CIs. We described continuous outcomes using mean differences (MD) or, if studies used different scales, using standardised mean differences. We calculated the NNTB or number needed to treat for an additional harmful outcome (NNTH) and its 95% CI using Visual Rx (www.nntonline.net/), while taking into account the event rate in the control group. This was done for the two primary outcomes and for the outcomes 'dropouts due to a lack of efficacy' and 'dropouts due to adverse effects'.

For continuous outcomes, we extracted both change and endpoint data that were available for the same outcome category.

Unit of analysis issues

Cluster‐randomised trials

For included studies that employed a cluster‐randomisation study design (such as randomisation by clinician or practice), we planned to first assess whether clustering had been accounted for by the study authors, in order to reduce the possibility of a unit‐of‐analysis error (Divine 1992). Where clustering was not accounted for in primary studies, we planned to contact the first authors of studies to obtain the intraclass correlation coefficient (ICC) of their clustered data and to adjust for this using accepted methods (Higgins 2022). Where clustering was adequately accounted for, we planned to include the data as if from a parallel‐group randomised study. Where there was no ICC for cluster‐randomised studies, we planned to incorporate these into the analysis using an inflated variance as suggested in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2022).

Cross‐over trials

We excluded cross‐over trials as it was quite likely that there would be carry‐over effects from the first period of treatment after crossing to the second period of treatment.

Studies with multiple treatment groups

If a study had multiple arms and antidepressants were being used as an active control, then we included the arms of the study as pair‐wise comparisons that would meet the criteria for inclusion.

Dealing with missing data

When there were missing data, the first approach was to contact the study authors. We calculated missing standard deviations (SDs) from other data if possible. If these methods were not sufficient, then we considered imputation of missing data.

If within studies the study authors attempted to correct for missing data by using multiple imputation or last observation carried forward (LOCF), then we used those in preference to the completers data, with appropriate care to determine if this made a substantial difference to the results. If no data were available from the text of the study but they could be extracted from a graph or figure, we used GetDataGraph Digitizer version 2.26.0.20 to extract the data.

Assessment of heterogeneity

We examined heterogeneity between comparable trials by visual inspection of the forest plot and by considering the I² statistic value.

We quantified heterogeneity as follows (Higgins 2022):

• 0% to 40%: may not be important;

• 30% to 60%: may represent moderate heterogeneity;

• 50% to 90%: may represent substantial heterogeneity;

• 75% to 100%: may represent considerable heterogeneity.

We explored high levels of heterogeneity (where more than 50% of the variability in outcome between trials could not be explained by sampling variation) through the sensitivity analyses specified below. In addition to the I² statistic (Higgins 2003), we presented the Chi² test for heterogeneity and its P value and considered the direction and magnitude of the treatment effects.

Assessment of reporting biases

We produced a funnel plot (trial effect versus trial size) if more than 10 trials contributed to a meta‐analysis, in an attempt to investigate the likelihood of small‐sample biases. We investigated funnel plot asymmetries (suggesting potential publication bias) (Egger 1997).

Data synthesis

We performed statistical analysis in accordance with the guidelines for statistical analysis in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2022). We entered data into Review Manager (RevMan 2024).

We combined dichotomous data (e.g. response/no response) using a random‐effects model.

We also combined continuous data using a random‐effects model. We conducted a fixed‐effect analysis as part of a sensitivity analysis.

Subgroup analysis and investigation of heterogeneity

Subgroup analyses were planned a priori and were to be performed and interpreted with caution, because multiple analyses can lead to false‐positive conclusions (Oxman 1992). We planned to perform the following subgroup analyses, where possible, for the following a priori reasons.

• Diagnosis type: pre‐DSM‐III, DSM‐III and DSM III‐TR, DSM IV and following. This was needed as there have been considerable changes in the criteria for GAD between revisions of the DSM (Wolk 1996). We analysed studies separately and only combined them if there was no heterogeneity.

• Treatment settings (psychiatric inpatients, psychiatric outpatients, primary care).

• Elderly participants (aged 65 years or older) versus other adult participants.

• Trials allowing psychiatric comorbidity versus those excluding such participants.

• Trials lasting over 12 weeks: antidepressants can take weeks to months to work. The longer trials may thus be able to assess the long‐term effect of antidepressants.

We planned to perform subgroup analyses on the primary outcomes only.

Sensitivity analysis

The following sensitivity analyses were planned a priori. We performed the following sensitivity analyses on the primary outcomes only.

• Excluding trials with unclear or high risk of bias in random allocation, unclear or high risk of bias in blinding, or both.

• Excluding trials whose dropout rate was greater than 20%.

• Application of fixed‐effect models were considered to determine if the weighting of small studies might have affected the pooled result.

The following sensitivity analysis was performed on the outcome 'change in symptom levels' only.

• Excluding trials for which the response rates had to be calculated based on the imputation method (Furukawa 2005), and those for which the SD or ICC had to be adopted from other trials (Furukawa 2006).

If subgroups within any of the subgroup or sensitivity analyses differed significantly from one another, we planned to perform a meta‐regression for exploratory analyses of additive or multiplicative influences of the variables in question.

Summary of findings and assessment of the certainty of the evidence

We used GRADEpro GDT to generate a summary of findings table for the primary outcomes, according to the GRADE approach (GRADEpro GDT; Schünemann 2013). When preparing the summary of findings table, we included all populations, all the primary outcome variables, and all time frames. We also included acceptability outcomes, that is, dropouts due to a lack of efficacy and dropouts due to adverse effects. We created one summary of findings table for all antidepressants compared to placebo, and one summary of findings table each for the main categories of antidepressants (SSRIs, SNRIs, and 'other' antidepressants).

Two review authors (KK and CC) independently examined the summary of findings and risk of bias tables to evaluate the certainty of the evidence based on the GRADE approach (Schünemann 2013). We resolved any disagreements by discussion and resolved them with a third review author (GG). We assessed the certainty of evidence for each outcome of interest based on five domains: risk of bias, inconsistency, indirectness, imprecision, and publication (or sponsorship) bias.

In this review, we performed a GRADE assessment for the two primary outcomes, 'rate of treatment response', measured as a reduction of at least 50% on the HAM‐A and 'acceptability', as well as for 'dropouts due to a lack of efficacy' and 'dropouts due to adverse effects' for the analyses comparing all antidepressants to placebo. These outcomes were considered the most clinically relevant. For each outcome, we downgraded the domains by one or two levels depending on the level of concern using the steps and guidelines outlined for each domain in the GRADE Handbook (Schünemann 2013). GRADEpro GDT provided an overall certainty of evidence rating of 'very low', 'low', 'moderate', or 'high' for the outcome depending on how each domain was assessed. The overall certainty of evidence rating is a reflection of how confident we were that the effect estimate we found was close to the true effect (Schünemann 2013).

Results

Description of studies

Results of the search

The search identified 1191 studies. After the removal of duplicates, we screened the titles and abstracts of 1100 studies, and excluded 944 studies that were ineligible. We assessed the remaining 156 studies through full‐text screening. Of those, we excluded 53 for reasons described in Figure 1 and the Characteristics of excluded studies table. We marked 23 studies as awaiting classification, and one study as ongoing. We included the remaining 37 studies in the systematic review and the meta‐analysis.

1.

PRISMA flow diagram to illustrate the flow of studies through the different phases of the review

Included studies

See Characteristics of included studies table.

Study characteristics

We included 37 RCTs (12,226 participants) in this systematic review and meta‐analysis. One study was conducted in China, 15 in the USA, one in the UK, one in Greece, one in Japan, 14 were multinational, and four did not specify the location. In terms of treatment setting, 28 studies enroled outpatients, two studies enroled primary care and psychiatric outpatients, two studies enroled primary care patients, one study enroled volunteers, and four studies did not specify the treatment setting. The total sample sizes ranged from 28 to 781. All included studies were published in English.

Population

The primary diagnosis in all the participants was moderate‐severe GAD, diagnosed using the DSM‐IV, DSM‐IV‐TR, or DSM‐III‐R. Most studies included adults (aged 18 years and older) and only one study enroled elder veterans (aged 60 years and older) (NCT00701675). No studies included people with medical comorbidities. Two studies included people with secondary psychiatric comorbidities (Bose 2008; Gommol 2015).

Interventions and comparators

The Characteristics of included studies table displays the interventions used in each study, along with the dosage and length of the double‐blind treatment period. The antidepressants included agomelatine (Stein 2008; Stein 2014; Stein 2017), duloxetine (Hartford 2007; Koponen 2007; Mahableshwarkar 2014; Nicolini 2009; Rynn 2008; Wen‐Yuan 2011), escitalopram (Baldwin 2006; Bose 2008; Davidson 2004; Goodman 2000; Goodman 2001; NCT00252343; NCT00417118; Stein 2014), imipramine (McLeod 1992), paroxetine (Baldwin 2006; Feltner 2009; NCT00135525; NCT00266747; NCT00658008; PAR 29060 637; Pollack 2001; Rickels 2003), sertraline (Allgulander 2004; Brawman‐Mintzer 2006; NCT00701675), venlafaxine (Allgulander 2001; Bose 2008; Davidson 1999; Gelenberg 2000; Hackett 2003; Hartford 2007; Kasper 2009; Lenox‐Smith 2003; Montgomery 2006; Nicolini 2009; Nimatoudis 2004; Rickels 2000), vilazodone (Gommol 2015), and vortioxetine (Mahableshwarkar 2014).

All studies used placebo as the comparator.

There were six multi‐armed studies comparing two antidepressants to placebo (Baldwin 2006; Bose 2008; Hartford 2007; Mahableshwarkar 2014; Nicolini 2009; Stein 2017). Three of these studies compared different classes of antidepressants to placebo (Bose 2008; Mahableshwarkar 2014; Stein 2014), whereas the other three compared the same class of antidepressants to placebo (Baldwin 2006; Hartford 2007; Nicolini 2009). Of the remaining 31 studies, one compared a TCA to placebo (McLeod 1992), 15 studies compared an SSRI to placebo (Allgulander 2004; Brawman‐Mintzer 2006; Davidson 2004; Feltner 2009; Goodman 2000; Goodman 2001; NCT00135525; NCT00252343; NCT00266747; NCT00417118; NCT00658008; NCT00701675; PAR 29060 637; Pollack 2001; Rickels 2003), and 12 studies compared an SNRI to placebo (Allgulander 2001; Davidson 1999; Gelenberg 2000; Hackett 2003; Kasper 2009; Koponen 2007; Lenox‐Smith 2003; Montgomery 2006; Nimatoudis 2004; Rickels 2000; Rynn 2008; Wen‐Yuan 2011). Three studies compared an antidepressant from the 'other' category to placebo (Gommol 2015; Stein 2008; Stein 2017). There were no studies comparing MAOIs, NaSSAs, NDRIs, or NRIs to placebo.

The minimum double‐blind duration ranged from four weeks to 28 weeks. The studies used both fixed and flexible dosing protocols. Eleven studies compared multiple fixed dosages of the same antidepressant as separate groups, which we combined into one group during data extraction (Allgulander 2001; Baldwin 2006; Brawman‐Mintzer 2006; Davidson 1999; Hackett 2003; Koponen 2007; Mahableshwarkar 2014; Nicolini 2009; Rickels 2000; Rickels 2003; Stein 2017).

Outcomes

Not all the outcomes planned for this review were investigated in each study. Twenty studies reported data on the rate of treatment response measured as a reduction of at least 50% on the HAM‐A (Analysis 1.1), 33 studies reported data on acceptability (Analysis 2.1), 18 studies reported data on the rate of treatment response (defined by study authors) (Analysis 3.1), 17 studies reported remission rates (Analysis 4.1), 34 studies reported change in symptom levels (Analysis 5.1), 23 studies reported the total number of participants reporting adverse effects (Analysis 6.1), 23 studies reported sleepiness/drowsiness (Analysis 7.1), six studies reported agitation/anxiety (Analysis 8.1), three studies reported suicide wishes/gestures/attempts (Analysis 9.1), four studies reported average score/change in quality of life/satisfaction (Analysis 10.1), 29 studies reported dropouts due to lack of efficacy (Analysis 11.1), and 32 studies reported dropouts due to adverse effects (Analysis 12.1). Studies were not included in the analyses if they did not report data on the specified outcome or if the authors did not fully report the specified outcome and could not be contacted for additional information.

1.1. Analysis.

Comparison 1: Rate of treatment response measured as a reduction of at least 50% on the Hamilton Anxiety Scale (HAM‐A), Outcome 1: Comparison 1 All antidepressants versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A

2.1. Analysis.

Comparison 2: Acceptability, Outcome 1: Comparison 1 All antidepressants versus placebo, Outcome 2 Acceptability

3.1. Analysis.

Comparison 3: Rate of treatment response (defined by study authors), Outcome 1: Comparison 1 All antidepressants versus placebo, Outcome 3 Rate of treatment response (defined by study authors)

4.1. Analysis.

Comparison 4: Remission rate, Outcome 1: Comparison 1 All antidepressants versus placebo, Outcome 4 Remission rate

5.1. Analysis.

Comparison 5: Change in symptom levels, Outcome 1: Comparison 1 All antidepressants versus placebo, Outcome 5 Change in symptom levels

6.1. Analysis.

Comparison 6: Total number of participants reporting adverse events, Outcome 1: Comparison 1 All antidepressants versus placebo, Outcome 6 Total number of participants reporting adverse events

7.1. Analysis.

Comparison 7: Sleepiness/drowsiness, Outcome 1: Comparison 1 All antidepressants versus placebo, Outcome 7 Sleepiness/drowsiness

8.1. Analysis.

Comparison 8: Agitation/anxiety, Outcome 1: Comparison 1 All antidepressants versus placebo, Outcome 8 Agitation/anxiety

9.1. Analysis.

Comparison 9: Suicide wishes/gestures/attempts, Outcome 1: Comparison 1 All antidepressants versus placebo, Outcome 9 Suicide wishes/gestures/attempts

10.1. Analysis.

Comparison 10: Average score/change in quality of life/satisfaction, Outcome 1: Comparison 1 All antidepressants versus placebo, Outcome 10 Average score/change in quality of life/satisfaction

11.1. Analysis.

Comparison 11: Dropouts due to a lack of efficacy, Outcome 1: Comparison 1 All antidepressants versus placebo, Outcome 11 Dropouts due to a lack of efficacy

12.1. Analysis.

Comparison 12: Dropouts due to adverse effects, Outcome 1: Comparison 1 All antidepressants versus placebo, Outcome 12 Dropouts due to adverse effects

There were no studies that reported data on falls, hypotension, subjective memory impairment, or the number of participants experiencing withdrawal symptoms. Studies that reported zero events in both treatment arms for an outcome were not included in the analyses for that outcome because they provided no valuable information on the treatment effect. This included eight studies that reported zero deaths in both treatment arms and one study that reported zero completed suicides in both treatment arms.

Of the included studies, we contacted the authors of 18 studies to request additional information (Allgulander 2001; Allgulander 2004; Bose 2008; Brawman‐Mintzer 2006; Davidson 2004; Feltner 2009; Gelenberg 2000; Hackett 2003; Kasper 2009; Koponen 2007; Lenox‐Smith 2003; Mahableshwarkar 2014; Montgomery 2006; Nicolini 2009; Nimatoudis 2004; Pollack 2001; Rickels 2000; Stein 2008). Out of all the authors that we contacted, two authors replied (Rickels 2000; Stein 2008). Rickels 2000 could not provide additional information as they no longer had access to the data due to their retirement. Stein 2008 could not share the data that was requested due to ongoing regulatory activities, but they directed us to the synopsis of the clinical study results which we had already included in our review.

Excluded studies

See Characteristics of excluded studies table and Figure 1.

We excluded 58 references (as 53 studies) for at least one of the following reasons: ineligible study design (23 studies), ineligible outcome (eight studies), ineligible participant population (seven studies), study was cancelled (three studies), ineligible intervention (four studies), concomitant benzodiazepine use (three studies), concomitant medical problem (two studies), and ineligible comparator (three studies). We presented details of 33 studies in the Characteristics of excluded studies table (Alaka 2014; Allgulander 2006; Allgulander 2008; Baldwin 2012b; Baldwin 2012c; Bodkin 2011; Chakhava 2011; Davidson 2005; Davidson 2008; EudraCT 2004‐001500‐13; EudraCT 2004‐002626‐22; EudraCT 2006‐006339‐31; EudraCT 2008‐003421‐17; Goldstein 2002; Hackett 2000; Hong‐Wei 2006; ISRCTN75061806; Kapczinski 2016; Kasper 2014; Kornstein 2009; Lenze 2005; Lenze 2009; Lindsay 1987; Lydiard 1999; Montgomery 2002; Rickels 1993; Rolland 2000; Sheehan 1999; Stein 2011; Stocchi 2003; Westenberg 2004; Wingerson 1992; Wright 2009).

Studies awaiting classification

See Characteristics of studies awaiting classification table.

We classified 23 studies as awaiting classification (Allgulander 2007; Baldwin 2002; Baldwin 2012d; Busnello 1974; Coric 2009; Derivan 1997; Haskins 1998; Hoffman 2009; Huang 2003; Kasper 2002; Meoni 2001; Mychaskiw 2009a; Mychaskiw 2009b; NCT00332891; NCT00390650; NCT00715039; Roberts 2006; Schweizer 1993; Shammas 1977; Smith 1994; Stein 2005; Wurthmann 1998; Xie 2001).

Ongoing studies

See Characteristics of ongoing studies table.

Our search identified one ongoing study comparing venlafaxine to placebo (jRCT2031220156).

Risk of bias in included studies

The overall risk of bias in the included studies was high. See Figure 2; Figure 3; and the Characteristics of included studies table.

2.

Review authors' judgements about each risk of bias item presented as unclear (yellow), low (green), and high (red) across all included studies

3.

Review authors' judgements about each risk of bias item presented as percentages across all included studies

Allocation

Thirteen studies specified how their randomisation sequence was generated and were at low risk of bias, while random sequence generation was unclear for 24 studies. Eight studies specified the methods for allocation concealment and were at low risk of bias, while it was unclear in the remaining 29 studies.

Blinding

In cases where studies only specified they were 'double‐blind', we judged the bias as high, low, or unclear depending on whether there was reasonable evidence for concern that blinding was compromised for participants, personnel, or outcome assessors in any way. In total, 26 studies received a low risk of bias, and nine studies received an unclear risk of bias for both blinding of participants and personnel and blinding of outcome assessment. One study was at low risk of bias for blinding of participants and personnel and unclear risk for assessors. One study was at unclear risk of bias for blinding of participants and personnel and low risk for assessors.

Incomplete outcome data

Although most studies used an intention‐to‐treat analysis, no studies received a low risk of bias for incomplete outcome data. This is because the main approach to dealing with missing data were LOCF. However, LOCF is limited as it carries forward the last observed value and assumes it would not have changed (Higgins 2011). This approach has the potential to lead to bias and, as a result, studies that used LOCF were automatically given an unclear risk of bias (Higgins 2011). We judged studies at high risk of bias if there was insufficient information to make a valid assessment or if they used LOCF and dropout rates differed significantly between groups. In total, 26 studies had an unclear risk of bias, and 11 studies had a high risk of bias in this domain.

Selective reporting

Twenty studies were at low risk of selective outcome reporting bias because they fully reported all outcomes that they specified in their protocol or methods. All other studies were at unclear (four studies) or a high (13 studies) risk of bias if there was insufficient information to make a valid assessment, or if authors failed to report, or only partially reported (i.e. only providing P values or describing the results as 'significant' or 'not significant') outcomes that they specified in their protocol or methods section.

Other potential sources of bias

We rated studies at unclear or high risk of other bias if there was insufficient information to make a valid assessment, or if there was suspicion of the involvement of the funder/sponsor. In this domain, one study was at low risk of bias because it was sponsored by the government, 21 studies had an unclear risk of bias because the sponsor/funding was not specified or the particular involvement of the study sponsor was unclear, and 15 studies at high risk of bias because they were sponsored by a pharmaceutical company and the authors of the study were involved with the pharmaceutical company or had conflicts of interest.

Effects of interventions

See: Table 1; Table 2; Table 3; Table 4

See: Table 1; Table 2; Table 3; Table 4.

Primary outcomes

1. Rate of treatment response measured as a reduction of at least 50% on the Hamilton Anxiety Rating Scale

Twenty studies including 7267 participants provided data on the rate of treatment response measured by a reduction of at least 50% on the HAM‐A.

1.1 All antidepressants versus placebo

Antidepressants have a benefit over placebo in the rate of treatment response measured as a reduction of at least 50% on the HAM‐A scale (RR 1.41, 95% CI 1.29 to 1.55; I² = 63%, P < 0.001; 20 studies, 7267 participants; high‐certainty evidence; Analysis 1.1). There was substantial heterogeneity. The magnitude of effect corresponds to an NNTB of 7 (95% CI 5 to 9), meaning that seven people need to be treated for one more person to experience a response to treatment. Visual inspection of the funnel plot suggested that there may have been publication bias (Figure 4).

4.

Funnel plot created in RevMan to assess publication bias for 1.1 Comparison 1 All antidepressants versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A

One study (682 participants) mentioned it would report the rate of treatment response, measured as a reduction of at least 50% on the HAM‐A, but did not report the LOCF data for the antidepressant groups (escitalopram and paroxetine) compared to placebo at the endpoint (Baldwin 2006). Since there were no data provided, it is unclear how the inclusion of this study may have affected the study findings.

1.2 Each class of antidepressant versus placebo

Analyses among different classes of antidepressants showed that antidepressants have a benefit over placebo (SSRIs: RR 1.51, 95% CI 1.20 to 1.90; I² = 73%; 4 studies, 1226 participants; SNRIs: RR 1.34, 95% CI 1.21 to 1.47; I² = 50%; 14 studies, 4659 participants; 'other' antidepressants: RR 1.54, 95% CI 1.14 to 2.08; I² = 85%; 5 studies, 1804 participants) (all high‐certainty evidence; Analysis 1.2).

1.2. Analysis.

Comparison 1: Rate of treatment response measured as a reduction of at least 50% on the Hamilton Anxiety Scale (HAM‐A), Outcome 2: Comparison 2 Each class of antidepressant versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A

There were no data for TCAs, MAOIs, NaSSAs, NDRIs, or NRIs for this outcome.

1.3 Agomelatine versus placebo

Agomelatine may have a benefit over placebo in increasing the rate of treatment response measured as a reduction of at least 50% on the HAM‐A scale (RR 1.93, 95% CI 1.38 to 2.70; I² = 69%, P = 0.04; 3 studies, 773 participants; Analysis 1.3). There was moderate‐substantial heterogeneity.

1.3. Analysis.

Comparison 1: Rate of treatment response measured as a reduction of at least 50% on the Hamilton Anxiety Scale (HAM‐A), Outcome 3: Comparison 3 Agomelatine versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A

1.4 Duloxetine versus placebo

Duloxetine may have a benefit over placebo in increasing the rate of treatment response measured as a reduction of at least 50% on the HAM‐A scale (RR 1.39, 95% CI 1.22 to 1.58; I² = 42%, P = 0.12; 6 studies, 2070 participants; Analysis 1.4). There was moderate‐substantial heterogeneity.

1.4. Analysis.

Comparison 1: Rate of treatment response measured as a reduction of at least 50% on the Hamilton Anxiety Scale (HAM‐A), Outcome 4: Comparison 4 Duloxetine versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A

1.5 Escitalopram versus placebo

Escitalopram may have a benefit over placebo in increasing the rate of treatment response measured as a reduction of at least 50% on the HAM‐A scale (RR 1.50, 95% CI 1.05 to 2.16; I² = 75%, P = 0.05; 2 studies, 530 participants; Analysis 1.5). There was substantial‐considerable heterogeneity.

1.5. Analysis.

Comparison 1: Rate of treatment response measured as a reduction of at least 50% on the Hamilton Anxiety Scale (HAM‐A), Outcome 5: Comparison 5 Escitalopram versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A

1.6 Sertraline versus placebo

Sertraline may be similar to placebo in rate of treatment response measured as a reduction of at least 50% on the HAM‐A scale (RR 1.52, 95% CI 0.99 to 2.35; I² = 86%, P = 0.008; 2 studies, 696 participants; Analysis 1.6). There was substantial‐considerable heterogeneity.

1.6. Analysis.

Comparison 1: Rate of treatment response measured as a reduction of at least 50% on the Hamilton Anxiety Scale (HAM‐A), Outcome 6: Comparison 6 Sertraline versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A

1.7 Venlafaxine versus placebo

Venlafaxine may have a benefit over placebo in increasing the rate of treatment response measured as a reduction of at least 50% on the HAM‐A scale (RR 1.32, 95% CI 1.18 to 1.49; I² = 50%, P = 0.03; 10 studies, 2913 participants; Analysis 1.7). There was moderate‐substantial heterogeneity.

1.7. Analysis.

Comparison 1: Rate of treatment response measured as a reduction of at least 50% on the Hamilton Anxiety Scale (HAM‐A), Outcome 7: Comparison 7 Venlafaxine versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A

1.8 Vilazodone versus placebo

Vilazodone may have a benefit over placebo in increasing the rate of treatment response measured as a reduction of at least 50% on the HAM‐A scale (RR 1.25, 95% CI 1.01 to 1.55; 1 study, 395 participants; Analysis 1.8).

1.8. Analysis.

Comparison 1: Rate of treatment response measured as a reduction of at least 50% on the Hamilton Anxiety Scale (HAM‐A), Outcome 8: Comparison 8 Vilazodone versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A

1.9 Vortioxetine versus placebo

Vortioxetine may be similar to placebo in rate of treatment response measured as a reduction of at least 50% on the HAM‐A scale (RR 1.04, 95% CI 0.85 to 1.29; 1 study, 610 participants; Analysis 1.9).

1.9. Analysis.

Comparison 1: Rate of treatment response measured as a reduction of at least 50% on the Hamilton Anxiety Scale (HAM‐A), Outcome 9: Comparison 9 Vortioxetine versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A

2. Acceptability

Thirty‐three studies including 11,294 participants provided data on acceptability (total dropout rate).

2.1 All antidepressants versus placebo

There was no difference in acceptability between antidepressants and placebo (RR 1.03, 95% CI 0.93 to 1.14; I² = 52%, P < 0.001; 33 studies, 11,294 participants; high‐certainty evidence; Analysis 2.1). There was moderate‐substantial heterogeneity. Visual inspection of the funnel plot suggested that there may have been publication bias (Figure 5).

5.

Funnel plot created in RevMan to assess publication bias for 2.1 Comparison 1 All antidepressants versus placebo, Outcome 2 Acceptability

2.2 Each class of antidepressant versus placebo

There was no difference in acceptability between SSRIs and placebo (RR 1.06, 95% CI 0.95 to 1.19; I² = 2%; 16 studies, 5031 participants; high‐certainty evidence; Analysis 2.2). There was low heterogeneity.

2.2. Analysis.

Comparison 2: Acceptability, Outcome 2: Comparison 2 Each class of antidepressant versus placebo, Outcome 2 Acceptability

There was no difference in acceptability between SNRIs or 'other' antidepressants and placebo (SNRIs: RR 1.03, 95% CI 0.87 to 1.21; I² = 69%; 15 studies, 4863 participants; 'other' antidepressants: RR 0.90, 95% CI 0.60 to 1.34; I² = 73%; 5 studies, 1830 participants; both high‐certainty evidence; Analysis 2.2). There was substantial heterogeneity.

There were no data for TCAs, MAOIs, NaSSAs, NDRIs, or NRIs for this outcome.

2.3 Agomelatine versus placebo

Fewer total participants may have dropped out of the agomelatine group compared to the placebo group (RR 0.60, 95% CI 0.44 to 0.83; I² = 0%, P = 0.55; 3 studies, 803 participants; Analysis 2.3). There was no heterogeneity.

2.3. Analysis.

Comparison 2: Acceptability, Outcome 3: Comparison 3 Agomelatine versus placebo, Outcome 2 Acceptability

2.4 Duloxetine versus placebo

Duloxetine may have similar acceptability to placebo (RR 1.07, 95% CI 0.82 to 1.39; I² = 73%, P = 0.006; 5 studies, 1749 participants; Analysis 2.4). There was substantial‐considerable heterogeneity.

2.4. Analysis.

Comparison 2: Acceptability, Outcome 4: Comparison 4 Duloxetine versus placebo, Outcome 2 Acceptability

2.5 Escitalopram versus placebo

Escitalopram may have similar acceptability to placebo (RR 0.99, 95% CI 0.83 to 1.17; I² = 0%, P = 0.49; 7 studies, 2197 participants; Analysis 2.5). There was no heterogeneity.

2.5. Analysis.

Comparison 2: Acceptability, Outcome 5: Comparison 5 Escitalopram versus placebo, Outcome 2 Acceptability

2.6 Paroxetine versus placebo

Fewer total participants may have dropped out of the placebo group compared to the paroxetine group (RR 1.25, 95% CI 1.05 to 1.47; I² = 0%, P = 0.56; 7 studies, 2215 participants; Analysis 2.6). There was no heterogeneity.

2.6. Analysis.

Comparison 2: Acceptability, Outcome 6: Comparison 6 Paroxetine versus placebo, Outcome 2 Acceptability

2.7 Sertraline versus placebo

Sertraline may have similar acceptability to placebo (RR 0.95, 95% CI 0.76 to 1.20; I² = 0%, P = 0.42; 3 studies, 758 participants; Analysis 2.7). There was no heterogeneity.

2.7. Analysis.

Comparison 2: Acceptability, Outcome 7: Comparison 7 Sertraline versus placebo, Outcome 2 Acceptability

2.8 Venlafaxine versus placebo

Venlafaxine may have similar acceptability to placebo (RR 0.97, 95% CI 0.81 to 1.16; I² = 66%, P < 0.001; 12 studies, 3693 participants; Analysis 2.8). There was substantial‐considerable heterogeneity.

2.8. Analysis.

Comparison 2: Acceptability, Outcome 8: Comparison 8 Venlafaxine versus placebo, Outcome 2 Acceptability

2.9 Vilazodone versus placebo

Vilazodone may have similar acceptability to placebo (RR 1.43, 95% CI 1.00 to 2.03; 1 study, 402 participants; Analysis 2.9).

2.9. Analysis.

Comparison 2: Acceptability, Outcome 9: Comparison 9 Vilazodone versus placebo, Outcome 2 Acceptability

2.10 Vortioxetine versus placebo

Vortioxetine may have similar acceptability to placebo (RR 1.12, 95% CI 0.81 to 1.55; 1 study, 625 participants; Analysis 2.10).

2.10. Analysis.

Comparison 2: Acceptability, Outcome 10: Comparison 10 Vortioxetine versus placebo, Outcome 2 Acceptability

Secondary outcomes

3. Rate of treatment response (defined by study authors)

Eighteen studies including 6613 participants provided data on the rate of treatment response (defined by study authors).

3.1 All antidepressants versus placebo

One study defined treatment response as a reduction of 40% or more on the HAM‐A (Gelenberg 2000), one study defined treatment response as a score of 1 or 2 on the Clinical Global Impression of Change (CGI‐C) scale (Feltner 2009), while the remaining 16 studies defined it as a score of 1 or 2 on the Clinical Global Impression – Improvement (CGI‐I) scale. The analysis showed a benefit of all antidepressants over placebo in the rate of treatment response as defined by study authors (RR 1.35, 95% CI 1.27 to 1.43; I² = 36%, P = 0.06; 18 studies, 6613 participants; Analysis 3.1). There was low‐moderate heterogeneity. Visual inspection of the funnel plot suggested that there may have been publication bias (Figure 6).

6.

Funnel plot created in RevMan to assess publication bias for 3.1 Comparison 1 All antidepressants versus placebo, Outcome 3 Rate of treatment response (defined by study authors)

Four studies mentioned they would report the rate of treatment response (defined as a CGI‐I score of 1 or 2) but failed to report it (781 participants; Mahableshwarkar 2014), only reported that there was a 'significant difference' but did not specify which group had higher response rates (46 participants; Nimatoudis 2004), or only reported that there was a superior response in the venlafaxine group (349 participants; Rickels 2000) or the agomelatine group (121 participants; Stein 2008) compared to placebo. It was difficult to determine how the inclusion of Mahableshwarkar 2014 and Nimatoudis 2004 may have affected the study findings without further details. The inclusion Rickels 2000 and Stein 2008 may have slightly increased the effect size in favour of antidepressants.

3.2 Each class of antidepressant versus placebo

SSRIs and SNRIs may have a benefit over placebo in rate of treatment response (as defined by study authors) (SSRIs: RR 1.31, 95% CI 1.20 to 1.44; I² = 45%, P = 0.06; 10 studies, 3661 participants; SNRIs: RR 1.41, 95% CI 1.31 to 1.53; I² = 5%, P = 0.39; 8 studies, 2693 participants) (Analysis 3.2). Heterogeneity was moderate among SSRIs and low among SNRIs. 'Other' antidepressants may be similar to placebo in rate of treatment response (as defined by study authors) (RR 1.20, 95% CI 1.00 to 1.43; 1 study, 395 participants; Analysis 3.2).

3.2. Analysis.

Comparison 3: Rate of treatment response (defined by study authors), Outcome 2: Comparison 2 Each class of antidepressant versus placebo, Outcome 3 Rate of treatment response (defined by study authors)

There were no data for TCAs, MAOIs, NaSSAs, NDRIs, or NRIs for this outcome.

3.3 Duloxetine versus placebo

Duloxetine may have a benefit over placebo in rate of treatment response (as defined by study authors) (RR 1.33, 95% CI 1.06 to 1.68; 1 study, 323 participants; Analysis 3.3).

3.3. Analysis.

Comparison 3: Rate of treatment response (defined by study authors), Outcome 3: Comparison 3 Duloxetine versus placebo, Outcome 3 Rate of treatment response (defined by study authors)

3.4 Escitalopram versus placebo

Escitalopram may have a benefit over placebo in rate of treatment response (as defined by study authors) (RR 1.29, 95% CI 1.12 to 1.50; I² = 40%, P = 0.19; 3 studies, 1105 participants; Analysis 3.4). There was low‐moderate heterogeneity.

3.4. Analysis.

Comparison 3: Rate of treatment response (defined by study authors), Outcome 4: Comparison 4 Escitalopram versus placebo, Outcome 3 Rate of treatment response (defined by study authors)

3.5 Paroxetine versus placebo

Paroxetine may have a benefit over placebo in rate of treatment response (as defined by study authors) (RR 1.25, 95% CI 1.11 to 1.40; I² = 46%, P = 0.10; 6 studies, 1998 participants; Analysis 3.5). There was moderate heterogeneity.

3.5. Analysis.

Comparison 3: Rate of treatment response (defined by study authors), Outcome 5: Comparison 5 Paroxetine versus placebo, Outcome 3 Rate of treatment response (defined by study authors)

3.6 Sertraline versus placebo

Sertraline may be similar to placebo in rate of response measured (as defined by study authors) (RR 1.41, 95% CI 1.00 to 2.00; I² = 84%, P = 0.01; 2 studies, 696 participants; Analysis 3.6). There was substantial‐considerable heterogeneity.

3.6. Analysis.

Comparison 3: Rate of treatment response (defined by study authors), Outcome 6: Comparison 6 Sertraline versus placebo, Outcome 3 Rate of treatment response (defined by study authors)

3.7 Venlafaxine versus placebo

Venlafaxine may have a benefit over placebo in rate of treatment response (as defined by study authors) (RR 1.42, 95% CI 1.31 to 1.54; I² = 5%, P = 0.39; 8 studies, 2531 participants; Analysis 3.7). There was low heterogeneity.

3.7. Analysis.

Comparison 3: Rate of treatment response (defined by study authors), Outcome 7: Comparison 7 Venlafaxine versus placebo, Outcome 3 Rate of treatment response (defined by study authors)

3.8 Vilazodone versus placebo

Vilazodone may be similar to placebo in rate of response measured (as defined by study authors) (RR 1.20, 95% CI 1.00 to 1.43; 1 study, 395 participants; Analysis 3.8).

3.8. Analysis.

Comparison 3: Rate of treatment response (defined by study authors), Outcome 8: Comparison 8 Vilazodone versus placebo, Outcome 3 Rate of treatment response (defined by study authors)

4. Remission rate

Seventeen studies including 6286 participants provided data on remission rate. All the studies defined remission as a HAM‐A total score of 7 or less at endpoint except one (PAR 29060 637), which defined remission as a HAM‐A total score of 10 or less at endpoint.

4.1 All antidepressants versus placebo

Antidepressants have a benefit over placebo for remission rates (RR 1.54, 95% CI 1.36 to 1.73; I² = 44%, P = 0.02; 17 studies, 6286 participants; Analysis 4.1). There was moderate heterogeneity. Visual inspection of the funnel plot suggested that there may have been publication bias (Figure 7).

7.

Funnel plot created in RevMan to assess publication bias for 4.1 Comparison 1 All antidepressants versus placebo, Outcome 4 Remission rate

One study (237 participants) only mentioned that escitalopram showed significant improvements compared to placebo in remission rate defined as a score of 1 on the CGI‐I, without giving more details (NCT00417118). It is unlikely that the inclusion of this study would have greatly impacted the results of the study findings.

4.2 Each class of antidepressant versus placebo

SSRIs, SNRIs, and 'other' antidepressants all showed benefit over placebo (SSRIs: RR 1.41, 95% CI 1.22 to 1.64; I² = 33%, P = 0.18; 7 studies, 2827 participants; SNRIs: RR 1.57, 95% CI 1.28 to 1.92; I² = 49%, P = 0.05; 8 studies, 2639 participants; 'other' antidepressants: RR 1.77, 95% CI 1.27 to 2.45; I² = 52%, P = 0.10; 4 studies, 1088 participants) (Analysis 4.2). Heterogeneity was low‐moderate among SSRIs, moderate among SNRIs, and moderate‐substantial among the 'other' antidepressants.

4.2. Analysis.

Comparison 4: Remission rate, Outcome 2: Comparison 2 Each class of antidepressant versus placebo, Outcome 4 Remission rate

There were no data for TCAs, MAOIs, NaSSAs, NDRIs, or NRIs for this outcome.

4.3 Agomelatine versus placebo

Agomelatine may have a benefit over placebo in remission rate (RR 2.06, 95% CI 1.57 to 2.69; I² = 0%, P = 0.54; 3 studies, 799 participants; Analysis 4.3). There was no heterogeneity.

4.3. Analysis.

Comparison 4: Remission rate, Outcome 3: Comparison 3 Agomelatine versus placebo, Outcome 4 Remission rate

4.4 Duloxetine versus placebo

Duloxetine may have a benefit over placebo in remission rate (RR 1.53, 95% CI 1.19 to 1.96; I² = 56%, P = 0.06; 5 studies, 1767 participants; Analysis 4.4). There was moderate‐substantial heterogeneity.

4.4. Analysis.

Comparison 4: Remission rate, Outcome 4: Comparison 4 Duloxetine versus placebo, Outcome 4 Remission rate

4.5 Escitalopram versus placebo

Escitalopram may have a benefit over placebo in remission rate (RR 1.48, 95% CI 1.21 to 1.81; I² = 0%, P = 0.79; 3 studies, 1068 participants; Analysis 4.5). There was no heterogeneity.

4.5. Analysis.

Comparison 4: Remission rate, Outcome 5: Comparison 5 Escitalopram versus placebo, Outcome 4 Remission rate

4.6 Paroxetine versus placebo

Paroxetine may have a benefit over placebo in remission rate (RR 1.31, 95% CI 1.03 to 1.65; I² = 57%, P = 0.07; 4 studies, 1527 participants; Analysis 4.6). There was moderate‐substantial heterogeneity.

4.6. Analysis.

Comparison 4: Remission rate, Outcome 6: Comparison 6 Paroxetine versus placebo, Outcome 4 Remission rate

4.7 Sertraline versus placebo

Sertraline may have a benefit over placebo in remission rate (RR 1.70, 95% CI 1.17 to 2.47; 1 study, 370 participants; Analysis 4.7).

4.7. Analysis.

Comparison 4: Remission rate, Outcome 7: Comparison 7 Sertraline versus placebo, Outcome 4 Remission rate

4.8 Venlafaxine versus placebo

Venlafaxine may have a benefit over placebo in remission rate (RR 1.74, 95% CI 1.28 to 2.38; I² = 54%, P = 0.07; 5 studies, 1196 participants; Analysis 4.8). There was moderate‐substantial heterogeneity.

4.8. Analysis.

Comparison 4: Remission rate, Outcome 8: Comparison 8 Venlafaxine versus placebo, Outcome 4 Remission rate

4.9 Vortioxetine versus placebo

Vortioxetine may be similar to placebo in terms of remission rate (RR 1.18, 95% CI 0.78 to 1.78; 1 study, 289 participants; Analysis 4.9).

4.9. Analysis.

Comparison 4: Remission rate, Outcome 9: Comparison 8 Vortioxetine versus placebo, Outcome 4 Remission rate

5. Change in symptom levels

Thirty‐four studies including 11,230 participants provided data on changes in symptom levels. All studies reported scores on the HAM‐A scale. A greater reduction in symptoms is indicated by a more negative MD. We extracted the mean change from baseline in HAM‐A total score for 31 studies and the HAM‐A total score at endpoint for four studies (Davidson 1999; Hackett 2003; McLeod 1992; NCT00701675). The imputation method for missing SDs was required for five studies: Allgulander 2001 (imputed SD from Lenox‐Smith 2003), Hackett 2003 (imputed SD from Davidson 1999), Rynn 2008 (imputed SD from Hartford 2007), Koponen 2007 (imputed SD from Hartford 2007), and Rickels 2000 (imputed SD from Kasper 2009). A pooled SD was calculated and imputed for one study (Gelenberg 2000).

5.1 All antidepressants versus placebo

Overall, the analysis showed a benefit for all antidepressants over placebo in reducing GAD symptoms (MD −2.79, 95% CI −3.49 to −2.08; I² = 80%, P < 0.001; 34 studies, 11,230 participants; Analysis 5.1). There was substantial‐considerable heterogeneity. Visual inspection of the funnel plot suggested that there may have been publication bias (Figure 8).

8.

Funnel plot created in RevMan to assess publication bias for 5.1 Comparison 1 All antidepressants versus placebo, Outcome 5 Change in symptom levels

5.2 Each class of antidepressant versus placebo

SSRIs, SNRIs, and 'other antidepressants (excluding TCAs)' showed a benefit for reducing symptoms of GAD over placebo (SSRIs: MD −2.22, 95% CI −2.88 to −1.55; I² = 50%; 15 studies, 4689 participants; SNRIs: MD −3.04, 95% CI −4.12 to −1.97; I² = 81%; 16 studies, 5244 participants; 'other antidepressants (excluding TCAs)': MD −3.81, 95% CI −7.07 to −0.54; I² = 94%; 5 studies, 1691 participants) (Analysis 5.2). There was moderate heterogeneity among SSRIs, substantial‐considerable heterogeneity among SNRIs, and considerable heterogeneity among the 'other antidepressants (excluding TCAs)'.

5.2. Analysis.

Comparison 5: Change in symptom levels, Outcome 2: Comparison 2 Each class of antidepressant versus placebo, Outcome 5 Change in symptom levels

TCAs and placebo may be similar in reducing symptoms levels (MD −3.90, 95% CI −9.49 to 1.69; 1 study, 28 participants).

There were no data for MAOIs, NaSSAs, NDRIs, or NRIs for this outcome.

5.3 Agomelatine versus placebo

Agomelatine may have a benefit over placebo in reducing symptom levels (MD −5.98, 95% CI −9.37 to −2.59; I² = 83%, P = 0.002; 3 studies, 799 participants; Analysis 5.3). There was substantial‐considerable heterogeneity.

5.3. Analysis.

Comparison 5: Change in symptom levels, Outcome 3: Comparison 4 Agomelatine versus placebo, Outcome 4 Change in symptom levels

5.4 Duloxetine versus placebo

Duloxetine may have a benefit over placebo in reducing symptom levels (MD −3.03, 95% CI −3.75 to −2.32; I² = 0%, P = 0.51; 6 studies, 2070 participants; Analysis 5.4). There was no heterogeneity.

5.4. Analysis.

Comparison 5: Change in symptom levels, Outcome 4: Comparison 5 Duloxetine versus placebo, Outcome 4 Change in symptom levels

5.5 Escitalopram versus placebo

Escitalopram may have a benefit over placebo in reducing symptom levels (MD −2.58, 95% CI −3.60 to −1.55; I² = 56%, P = 0.05; 6 studies, 1908 participants; Analysis 5.5). There was moderate‐substantial heterogeneity.

5.5. Analysis.

Comparison 5: Change in symptom levels, Outcome 5: Comparison 6 Escitalopram versus placebo, Outcome 4 Change in symptom levels

5.6 Imipramine versus placebo

Imipramine and placebo may be similar to placebo in reducing symptom levels (MD −3.90, 95% CI −9.49 to 1.69; 1 study, 28 participants; Analysis 5.6).

5.6. Analysis.

Comparison 5: Change in symptom levels, Outcome 6: Comparison 7 Imipramine versus placebo, Outcome 4 Change in symptom levels

5.7 Paroxetine versus placebo

Paroxetine may have a benefit over placebo in reducing symptom levels (MD −1.71, 95% CI −2.69 to −0.74; I² = 47%, P = 0.08; 7 studies, 2183 participants; Analysis 5.7). There was moderate heterogeneity.

5.7. Analysis.

Comparison 5: Change in symptom levels, Outcome 7: Comparison 8 Paroxetine versus placebo, Outcome 4 Change in symptom levels

5.8 Sertraline versus placebo

Sertraline and placebo may be similar in reducing symptom levels (MD −2.00, 95% CI −4.15 to 0.15; I² = 66%, P = 0.05; 3 studies, 736 participants; Analysis 5.8). There was substantial heterogeneity.

5.8. Analysis.

Comparison 5: Change in symptom levels, Outcome 8: Comparison 9 Sertraline versus placebo, Outcome 4 Change in symptom levels

5.9 Venlafaxine versus placebo

Venlafaxine may have a benefit over placebo in reducing symptom levels (MD −3.11, 95% CI −4.60 to −1.63; I² = 85%, P < 0.001; 12 studies, 3498 participants; Analysis 5.9). There was substantial‐considerable heterogeneity.

5.9. Analysis.

Comparison 5: Change in symptom levels, Outcome 9: Comparison 10 Venlafaxine versus placebo, Outcome 4 Change in symptom levels

5.10 Vilazodone versus placebo

Vilazodone may be similar to placebo in reducing symptom levels (MD −1.30, 95% CI −2.77 to 0.17; 1 study, 395 participants; Analysis 5.10).

5.10. Analysis.

Comparison 5: Change in symptom levels, Outcome 10: Comparison 11 Vilazodone versus placebo, Outcome 4 Change in symptom levels

5.11 Vortioxetine versus placebo

Vortioxetine may be similar to placebo in reducing symptom levels (MD −0.50, 95% CI −1.74 to 0.74; 1 study, 497 participants; Analysis 5.11).

5.11. Analysis.

Comparison 5: Change in symptom levels, Outcome 11: Comparison 12 Vortioxetine versus placebo, Outcome 4 Change in symptom levels

6. Total number of participants reporting adverse effects

Twenty‐three studies with 7615 participants provided data on the total number of participants reporting adverse effects.

6.1 All antidepressants versus placebo

The analysis showed a higher number of participants reporting adverse effects among all antidepressants compared to placebo (RR 1.15, 95% CI 1.12 to 1.19; I² = 5%, P = 0.39; 23 studies, 7615 participants; Analysis 6.1). There was low heterogeneity. Visual inspection of the funnel plot suggested that there may have been publication bias (Figure 9).

9.

Funnel plot created in RevMan to assess publication bias for 6.1 Comparison 1 All antidepressants versus placebo, Outcome 6 Total number of patients reporting adverse events

6.2 Each class of antidepressant versus placebo

Analyses among different classes of antidepressants also showed a larger number of participants reporting adverse effects in the SSRI, SNRI, and 'other' antidepressants group compared to placebo (SSRIs: RR 1.16, 95% CI 1.12 to 1.21; I² = 0%, P = 0.93; 14 studies, 4331 participants; SNRI: RR 1.14, 95% CI 1.07 to 1.22; I² = 40%, P = 0.13; 7 studies, 1887 participants; 'other' antidepressants: RR 1.21, 95% CI 1.07 to 1.37; I² = 41%, P = 0.15; 5 studies, 1821 participants) (Analysis 6.2). There was no heterogeneity among SSRIs and low‐moderate heterogeneity for SNRIs and 'other' antidepressants.

6.2. Analysis.

Comparison 6: Total number of participants reporting adverse events, Outcome 2: Comparison 2 Each class of antidepressant versus placebo, Outcome 6 Total number of participants reporting adverse events

There were no data for TCAs, MAOIs, NaSSAs, NDRIs, or NRIs for this outcome.

6.3 Agomelatine versus placebo

A similar number of participants may have reported adverse effects in the agomelatine group compared to the placebo group (RR 1.13, 95% CI 0.94 to 1.37; I² = 0%, P = 0.76; 3 studies, 801 participants; Analysis 6.3). There was no heterogeneity.

6.3. Analysis.

Comparison 6: Total number of participants reporting adverse events, Outcome 3: Comparison 3 Agomelatine versus placebo, Outcome 6 Total number of participants reporting adverse events

6.4 Duloxetine versus placebo

More participants may have reported adverse effects in the duloxetine group compared to the placebo group (RR 1.17, 95% CI 1.10 to 1.25; I² = 0%, P = 0.68; 4 studies, 1169 participants; Analysis 6.4). There was no heterogeneity.

6.4. Analysis.

Comparison 6: Total number of participants reporting adverse events, Outcome 4: Comparison 4 Duloxetine versus placebo, Outcome 6 Total number of participants reporting adverse events

6.5 Escitalopram versus placebo

More participants may have reported adverse effects in the escitalopram group compared to the placebo group (RR 1.13, 95% CI 1.02 to 1.24; I² = 69%, P = 0.002; 8 studies, 2409 participants; Analysis 6.5). There was substantial heterogeneity.

6.5. Analysis.

Comparison 6: Total number of participants reporting adverse events, Outcome 5: Comparison 5 Escitalopram versus placebo, Outcome 6 Total number of participants reporting adverse events

6.6 Paroxetine versus placebo

A similar number of participants may have reported adverse effects in the paroxetine group compared to the placebo group (RR 1.06, 95% CI 0.90 to 1.24; I² = 84%, P < 0.001; 6 studies, 2019 participants; Analysis 6.6). There was substantial‐considerable heterogeneity.

6.6. Analysis.

Comparison 6: Total number of participants reporting adverse events, Outcome 6: Comparison 6 Paroxetine versus placebo, Outcome 6 Total number of participants reporting adverse events

6.7 Sertraline versus placebo

A similar number of participants may have reported adverse effects in the sertraline group compared to the placebo group (RR 1.17, 95% CI 0.35 to 3.84; 1 study, 42 participants; Analysis 6.7).

6.7. Analysis.

Comparison 6: Total number of participants reporting adverse events, Outcome 7: Comparison 7 Sertraline versus placebo, Outcome 6 Total number of participants reporting adverse events

6.8 Venlafaxine versus placebo

More participants may have reported adverse effects in the venlafaxine group compared to the placebo group (RR 1.11, 95% CI 1.01 to 1.23; I² = 54%, P = 0.09; 4 studies, 879 participants; Analysis 6.8). There was substantial heterogeneity.

6.8. Analysis.

Comparison 6: Total number of participants reporting adverse events, Outcome 8: Comparison 8 Venlafaxine versus placebo, Outcome 6 Total number of participants reporting adverse events

6.9 Vilazodone versus placebo

More participants may have reported adverse effects in the vilazodone group compared to the placebo group (RR 1.39, 95% CI 1.22 to 1.59; 1 study, 398 participants; Analysis 6.9).

6.9. Analysis.

Comparison 6: Total number of participants reporting adverse events, Outcome 9: Comparison 9 Vilazodone versus placebo, Outcome 6 Total number of participants reporting adverse events

6.10 Vortioxetine versus placebo

More participants may have reported adverse effects in the vortioxetine group compared to the placebo group (RR 1.13, 95% CI 1.01 to 1.27; 1 study, 622 participants; Analysis 6.10).

6.10. Analysis.

Comparison 6: Total number of participants reporting adverse events, Outcome 10: Comparison 10 Vortioxetine versus placebo, Outcome 6 Total number of participants reporting adverse events

7. Sleepiness/drowsiness

Twenty‐three studies with 8738 participants provided data on the number of participants reporting sleepiness/drowsiness. One study (113 participants) only reported sleepiness/drowsiness for the paroxetine arm but not the placebo arm and was not included in the analysis (Feltner 2009).

7.1 All antidepressants versus placebo

The analysis showed more participants reporting sleepiness/drowsiness among all antidepressants compared to placebo (RR 2.30, 95% CI 1.91 to 2.77; I² = 0%, P = 0.75; 23 studies, 8738 participants; Analysis 7.1). There was no heterogeneity. Visual inspection of the funnel plot suggested that publication bias may not have occurred (Figure 10).

10.

Funnel plot created in RevMan to assess publication bias for 7.1 Comparison 1 All antidepressants versus placebo, Outcome 7 Sleepiness/drowsiness

One study (467 participants) only mentioned that sleepiness/drowsiness was frequently experienced by the participants, but no other details were provided (Hackett 2003). Given the large sample size for this analysis compared to the sample size of the study, it is unlikely that the inclusion of this study would have greatly affected the findings.

7.2 Each class of antidepressant versus placebo

Analyses comparing SSRIs, SNRIs, or 'other' antidepressants to placebo showed more participants reporting sleepiness/drowsiness compared to placebo (SSRIs: RR 2.15, 95% CI 1.70 to 2.71; I² = 0%, P = 0.55; 12 studies, 3933 participants; SNRIs: RR 2.66, 95% CI 1.92 to 3.69; I² = 5%, P = 0.39; 10 studies, 3529 participants; 'other' antidepressants: RR 1.78, 95% CI 1.01 to 3.12; I² = 0%, P = 0.93; 4 studies, 1700 participants) (Analysis 7.2). There was no or low heterogeneity among all analyses.

7.2. Analysis.

Comparison 7: Sleepiness/drowsiness, Outcome 2: Comparison 2 Each class of antidepressant versus placebo, Outcome 7 Sleepiness/drowsiness

There were no data for TCAs, MAOIs, NaSSAs, NDRIs, or NRIs for this outcome.

7.3 Agomelatine versus placebo

A similar number of participants may have reported feeling sleepiness/drowsiness in the agomelatine group as the placebo group (RR 1.94, 95% CI 0.60 to 6.27; I² = 0%, P = 0.60; 2 studies, 680 participants; Analysis 7.3). There was no heterogeneity.

7.3. Analysis.

Comparison 7: Sleepiness/drowsiness, Outcome 3: Comparison 3 Agomelatine versus placebo, Outcome 7 Sleepiness/drowsiness

7.4 Duloxetine versus placebo

More participants may have experienced sleepiness/drowsiness in the duloxetine group compared to the placebo group (RR 4.03, 95% CI 2.47 to 6.58; I² = 0%, P = 0.57; 6 studies, 2094 participants; Analysis 7.4). There was no heterogeneity.

7.4. Analysis.

Comparison 7: Sleepiness/drowsiness, Outcome 4: Comparison 4 Duloxetine versus placebo, Outcome 7 Sleepiness/drowsiness

7.5 Escitalopram versus placebo

More participants may have experienced sleepiness/drowsiness in the escitalopram group compared to the placebo group (RR 1.88, 95% CI 1.32 to 2.67; I² = 0%, P = 0.91; 6 studies, 1933 participants; Analysis 7.5). There was no heterogeneity.

7.5. Analysis.

Comparison 7: Sleepiness/drowsiness, Outcome 5: Comparison 5 Escitalopram versus placebo, Outcome 7 Sleepiness/drowsiness

7.6 Paroxetine versus placebo

More participants may have experienced sleepiness/drowsiness in the paroxetine group compared to the placebo group (RR 2.33, 95% CI 1.71 to 3.18; I² = 7%, P = 0.37; 7 studies, 2294 participants; Analysis 7.6). There was low heterogeneity.

7.6. Analysis.

Comparison 7: Sleepiness/drowsiness, Outcome 6: Comparison 6 Paroxetine versus placebo, Outcome 7 Sleepiness/drowsiness

7.7 Sertraline versus placebo

A similar number of participants may have reported feeling sleepiness/drowsiness in the sertraline group as the placebo group (RR 0.17, 95% CI 0.01 to 3.98; 1 study, 42 participants; Analysis 7.7).

7.7. Analysis.

Comparison 7: Sleepiness/drowsiness, Outcome 7: Comparison 7 Sertraline versus placebo, Outcome 7 Sleepiness/drowsiness

7.8 Venlafaxine versus placebo

More participants may have experienced sleepiness/drowsiness in the venlafaxine group compared to the placebo group (RR 2.20, 95% CI 1.54 to 3.14; I² = 0%, P = 0.82; 6 studies, 1766 participants; Analysis 7.8). There was no heterogeneity.

7.8. Analysis.

Comparison 7: Sleepiness/drowsiness, Outcome 8: Comparison 8 Venlafaxine versus placebo, Outcome 7 Sleepiness/drowsiness

7.9 Vilazodone versus placebo

A similar number of participants may have reported feeling sleepiness/drowsiness in the vilazodone group as the placebo group (RR 1.98, 95% CI 0.76 to 5.17; 1 study, 398 participants; Analysis 7.9).

7.9. Analysis.

Comparison 7: Sleepiness/drowsiness, Outcome 9: Comparison 9 Vilazodone versus placebo, Outcome 7 Sleepiness/drowsiness

7.10 Vortioxetine versus placebo

A similar number of participants may have reported feeling sleepiness/drowsiness in the vortioxetine group as the placebo group (RR 1.55, 95% CI 0.65 to 3.67; 1 study, 622 participants; Analysis 7.10).

7.10. Analysis.

Comparison 7: Sleepiness/drowsiness, Outcome 10: Comparison 10 Vortioxetine versus placebo, Outcome 7 Sleepiness/drowsiness

8. Agitation/anxiety

Six studies with 2026 participants provided data on the number of participants reporting agitation/anxiety. One study (411 participants) only reported this outcome for the treatment arms, but not the placebo arm; therefore, it was not included in the analysis (Stein 2014).

8.1 All antidepressants versus placebo

Overall, the analysis showed that a similar number of participants may have reported experiencing agitation/anxiety among all antidepressants and placebo (RR 1.06, 95% CI 0.74 to 1.53; I² = 0%, P = 0.41; 6 studies, 2026 participants; Analysis 8.1). There was no heterogeneity.

8.2 Each class of antidepressant versus placebo

A similar number of participants may have reported experiencing agitation/anxiety with SSRIs and SNRIs compared to placebo (SSRIs: RR 1.11, 95% CI 0.60 to 2.05; I² = 19%, P = 0.30; 5 studies, 1783 participants; SNRIs: RR 0.99, 95% CI 0.57 to 1.72; 1 study, 243 participants) (Analysis 8.2). There was low heterogeneity among the SSRI group.

8.2. Analysis.

Comparison 8: Agitation/anxiety, Outcome 2: Comparison 2 Each class of antidepressant versus placebo, Outcome 8 Agitation/anxiety

There were no data for TCAs, MAOIs, NaSSAs, NDRIs, NRIs, or 'other' antidepressants for this outcome.

8.3 Escitalopram versus placebo

A similar number of participants may have reported feeling agitation/anxiety in the escitalopram group as the placebo group (RR 0.53, 95% CI 0.06 to 4.79; I² = 72%, P = 0.06; 2 studies, 857 participants; Analysis 8.3). There was substantial heterogeneity.

8.3. Analysis.

Comparison 8: Agitation/anxiety, Outcome 3: Comparison 3 Escitalopram versus placebo, Outcome 8 Agitation/anxiety

8.4 Paroxetine versus placebo

A similar number of participants may have reported feeling agitation/anxiety in the paroxetine group as the placebo group (RR 1.80, 95% CI 0.67 to 4.84; I² = 0%, P = 0.63; 2 studies, 650 participants; Analysis 8.4). There was no heterogeneity.

8.4. Analysis.

Comparison 8: Agitation/anxiety, Outcome 4: Comparison 4 Paroxetine versus placebo, Outcome 8 Agitation/anxiety

8.5 Sertraline versus placebo

A similar number of participants may have reported feeling agitation/anxiety in the sertraline group as the placebo group (RR 1.10, 95% CI 0.60 to 2.00; I² = 0%, P = 0.83; 2 studies, 415 participants; Analysis 8.5). There was no heterogeneity.

8.5. Analysis.

Comparison 8: Agitation/anxiety, Outcome 5: Comparison 5 Sertraline versus placebo, Outcome 8 Agitation/anxiety

8.6 Venlafaxine versus placebo

A similar number of participants may have reported feeling agitation/anxiety in the venlafaxine group as the placebo group (RR 0.99, 95% CI 0.57 to 1.72; 1 study, 243 participants; Analysis 8.6).

8.6. Analysis.

Comparison 8: Agitation/anxiety, Outcome 6: Comparison 6 Venlafaxine versus placebo, Outcome 8 Agitation/anxiety

9. Suicide wishes/gestures/attempts

Three studies with 802 participants provided data on suicide wishes/gestures/attempts.

9.1 All antidepressants versus placebo

Overall, the analysis showed that a similar number of participants may have reported experiencing suicide wishes/gestures/attempts between all antidepressants compared to placebo (RR 0.74, 95% CI 0.40 to 1.36; I² = 0%, P = 0.40; 3 studies, 802 participants; Analysis 9.1). There was no heterogeneity.

9.2 Each class of antidepressant versus placebo

A similar number of participants may have reported experiencing suicide wishes/gestures/attempts between SSRIs, SNRIs and 'other' antidepressants compared to placebo (SSRIs: RR 5.15, 95% CI 0.25 to 105.98; 1 study, 197 participants; SNRIs: RR 0.54, 95% CI 0.16 to 1.79; 1 study, 210 participants; 'other' antidepressants: RR 0.75, 95% CI 0.36 to 1.54; 1 study, 395 participants) (Analysis 9.2).

9.2. Analysis.

Comparison 9: Suicide wishes/gestures/attempts, Outcome 2: Comparison 2 Each class of antidepressant versus placebo, Outcome 9 Suicide wishes/gestures/attempts

There were no data for TCAs, MAOIs, NaSSAs, NDRIs, or NRIs for this outcome.

9.3 Duloxetine versus placebo

A similar number of participants may have reported experiencing suicidal wishes/gestures/attempts in the duloxetine group compared to the placebo group (RR 0.54, 95% CI 0.16 to 1.79; 1 study, 210 participants; Analysis 9.3).

9.3. Analysis.

Comparison 9: Suicide wishes/gestures/attempts, Outcome 3: Comparison 3 Duloxetine versus placebo, Outcome 9 Suicide wishes/gestures/attempts

9.4 Paroxetine versus placebo

A similar number of participants may have reported experiencing suicidal wishes/gestures/attempts in the paroxetine group compared to the placebo group (RR 5.15, 95% CI 0.25 to 105.98; 1 study, 197 participants; Analysis 9.4).

9.4. Analysis.

Comparison 9: Suicide wishes/gestures/attempts, Outcome 4: Comparison 4 Paroxetine versus placebo, Outcome 9 Suicide wishes/gestures/attempts

9.5 Vilazodone versus placebo

A similar number of participants may have reported experiencing suicidal wishes/gestures/attempts in the vilazodone group compared to the placebo group (RR 0.75, 95% CI 0.36 to 1.54; 1 study, 395 participants; Analysis 9.5).

9.5. Analysis.

Comparison 9: Suicide wishes/gestures/attempts, Outcome 5: Comparison 5 Vilazodone versus placebo, Outcome 9 Suicide wishes/gestures/attempts

10. Average score/change in quality of life/satisfaction

Four studies with 1013 participants provided data on the average change/score in quality of life/satisfaction. All studies reported the mean change from baseline on the Quality of Life Enjoyment and Satisfaction Questionnaire (Q‐LES‐Q). Therefore, we reported the MD. A higher score on the Q‐LES‐Q indicates an improved quality of life.

10.1 All antidepressants versus placebo

The analysis showed improvement in average score/change in quality of life/satisfaction with antidepressants over placebo (MD 6.51, 95% CI 4.95 to 8.07; I² = 0%, P = 0.92; 4 studies, 1013 participants; Analysis 10.1). There was no heterogeneity.

Three studies only mentioned that the antidepressant showed improvements over placebo in Q‐LES‐Q scores, but the data for each treatment group were not fully reported and could not be extracted. One study found a significant improvement (NCT00417118: 237 participants), while the other two did not (Bose 2008: 392 participants; Brawman‐Mintzer 2006: 338 participants). One study stated it would report this outcome but did not (Koponen 2007). Given the relatively small sample size for the meta‐analysis of this outcome, the two studies that found non‐significant results may have slightly attenuated the effect estimate.

10.2 Each class of antidepressant versus placebo

Analyses also showed a benefit with SSRIs and SNRIs compared to placebo in the average change/score in quality of life/satisfaction (SSRIs: MD 6.70, 95% CI 5.05 to 8.35; I² = 0%, P = 0.99; 3 studies, 760 participants; SNRIs: MD 5.00, 95% CI 0.29 to 9.71; 1 study, 253 participants) (Analysis 10.2). There was no heterogeneity with SSRIs.

10.2. Analysis.

Comparison 10: Average score/change in quality of life/satisfaction, Outcome 2: Comparison 2 Each class of antidepressant versus placebo, Outcome 10 Average score/change in quality of life/satisfaction

There were no data for TCAs, MAOIs, NaSSAs, NDRIs, NRIs, or 'other' antidepressants for this outcome.

10.3 Escitalopram versus placebo

There may be a benefit with escitalopram compared to placebo in the average change/score in quality of life/satisfaction (MD 6.70, 95% CI 4.46 to 8.94; 1 study, 307 participants; Analysis 10.3).

10.3. Analysis.

Comparison 10: Average score/change in quality of life/satisfaction, Outcome 3: Comparison 3 Escitalopram versus placebo, Outcome 10 Average score/change in quality of life/satisfaction

10.4 Paroxetine versus placebo

There may be a benefit with paroxetine compared to placebo in the average change/score in quality of life/satisfaction (MD 7.20, 95% CI 0.97 to 13.43; 1 study, 113 participants; Analysis 10.4).

10.4. Analysis.

Comparison 10: Average score/change in quality of life/satisfaction, Outcome 4: Comparison 4 Paroxetine versus placebo, Outcome 10 Average score/change in quality of life/satisfaction

10.5 Sertraline versus placebo

There may be a benefit with sertraline compared to placebo in the average change/score in quality of life/satisfaction (MD 6.60, 95% CI 3.96 to 9.24; 1 study, 340 participants; Analysis 10.5).

10.5. Analysis.

Comparison 10: Average score/change in quality of life/satisfaction, Outcome 5: Comparison 5 Sertraline versus placebo, Outcome 10 Average score/change in quality of life/satisfaction

10.6 Venlafaxine versus placebo

There may be a benefit with venlafaxine compared to placebo in the average change/score in quality of life/satisfaction (MD 5.00, 95% CI 0.29 to 9.71; 1 study, 253 participants; Analysis 10.6).

10.6. Analysis.

Comparison 10: Average score/change in quality of life/satisfaction, Outcome 6: Comparison 6 Venlafaxine versus placebo, Outcome 10 Average score/change in quality of life/satisfaction

11. Dropouts due to a lack of efficacy

Twenty‐nine studies with 11,007 participants provided data on dropouts due to a lack of efficacy. Two studies reported dropouts due to lack of efficacy for only one study arm; therefore, due to uncertainty in the data for the other arm, these studies were not included in this analysis (NCT00266747: 246 participants; Nimatoudis 2004: 46 participants).

11.1 All antidepressants versus placebo

Fewer participants dropped out due to a lack of efficacy in the antidepressant group compared to the placebo group (RR 0.41, 95% CI 0.33 to 0.50; I² = 7%, P = 0.35; 29 studies, 11,007 participants; high‐certainty evidence; Analysis 11.1). There was low heterogeneity. The magnitude of effect corresponds to an NNTB of 27 (95% CI 24 to 32). This means that treating 27 people would result in one less person dropping out due to a lack of efficacy. Visual inspection of the funnel plot suggested that publication bias may not have occurred (Figure 11).

11.

Funnel plot created in RevMan to assess publication bias for 11.1 Comparison 1 All antidepressants versus placebo, Outcome 11 Dropouts due to a lack of efficacy

11.2 Each class of antidepressant versus placebo

Fewer participants dropped out due to a lack of efficacy with SSRIs, NRSIs and 'other' antidepressant compared to placebo (SSRIs: RR 0.55, 95% CI 0.38 to 0.79; I² = 6%, P = 0.39; 14 studies, 4832 participants; SNRIs: RR 0.33, 95% CI 0.25 to 0.43; I² = 0%, P = 0.73; 13 studies, 4775 participants; 'other' antidepressants: RR 0.52, 95% CI 0.28 to 0.95; I² = 37%, P = 0.18; 5 studies, 1826 participants) (all high‐certainty evidence; Analysis 11.2). There was no or low heterogeneity.

11.2. Analysis.

Comparison 11: Dropouts due to a lack of efficacy, Outcome 2: Comparison 2 Each class of antidepressant versus placebo, Outcome 11 Dropouts due to a lack of efficacy

No data were available for TCAs, MAOIs, NaSSAs, NDRIs, or NRIs for this outcome.

11.3 Agomelatine versus placebo

Fewer participants may have dropped out due to a lack of efficacy from the agomelatine group compared to the placebo group (RR 0.45, 95% CI 0.21 to 0.96; I² = 54%, P = 0.11; 3 studies, 803 participants; Analysis 11.3). There was substantial heterogeneity.

11.3. Analysis.

Comparison 11: Dropouts due to a lack of efficacy, Outcome 3: Comparison 3 Agomelatine versus placebo, Outcome 11 Dropouts due to a lack of efficacy

11.4 Duloxetine versus placebo

Fewer participants may have dropped out due to a lack of efficacy from the duloxetine group compared to the placebo group (RR 0.24, 95% CI 0.15 to 0.38; I² = 0%, P = 0.93; 6 studies, 2098 participants; Analysis 11.4). There was no heterogeneity.

11.4. Analysis.

Comparison 11: Dropouts due to a lack of efficacy, Outcome 4: Comparison 4 Duloxetine versus placebo, Outcome 11 Dropouts due to a lack of efficacy

11.5 Escitalopram versus placebo

Fewer participants may have dropped out due to a lack of efficacy from the escitalopram group compared to the placebo group(RR 0.48, 95% CI 0.27 to 0.86; I² = 27%, P = 0.22; 7 studies, 2174 participants; Analysis 11.5). There was low heterogeneity.

11.5. Analysis.

Comparison 11: Dropouts due to a lack of efficacy, Outcome 5: Comparison 5 Escitalopram versus placebo, Outcome 11 Dropouts due to a lack of efficacy

11.6 Paroxetine versus placebo

A similar number of participants may have dropped out due to a lack of efficacy from the paroxetine group compared to the placebo group (RR 0.69, 95% CI 0.40 to 1.17; I² = 0%, P = 0.50; 6 studies, 2098 participants; Analysis 11.6). There was no heterogeneity.

11.6. Analysis.

Comparison 11: Dropouts due to a lack of efficacy, Outcome 6: Comparison 6 Paroxetine versus placebo, Outcome 11 Dropouts due to a lack of efficacy

11.7 Sertraline versus placebo

A similar number of participants may have dropped out due to a lack of efficacy from the sertraline group compared to the placebo group (RR 0.57, 95% CI 0.19 to 1.70; I² = 0%, P = 0.55; 2 studies, 699 participants; Analysis 11.7). There was no heterogeneity.

11.7. Analysis.

Comparison 11: Dropouts due to a lack of efficacy, Outcome 7: Comparison 7 Sertraline versus placebo, Outcome 11 Dropouts due to a lack of efficacy

11.8 Venlafaxine versus placebo

Fewer participants may have dropped out due to a lack of efficacy from the venlafaxine group compared to the placebo group (RR 0.39, 95% CI 0.28 to 0.53; I² = 0%, P = 0.74; 9 studies, 3008 participants; Analysis 11.8). There was no heterogeneity.

11.8. Analysis.

Comparison 11: Dropouts due to a lack of efficacy, Outcome 8: Comparison 8 Venlafaxine versus placebo, Outcome 11 Dropouts due to a lack of efficacy

11.9 Vilazodone versus placebo

A similar number of participants may have dropped out due to a lack of efficacy from the vilazodone group compared to the placebo group(RR 1.98, 95% CI 0.18 to 21.66; 1 study, 398 participants; Analysis 11.9).

11.9. Analysis.

Comparison 11: Dropouts due to a lack of efficacy, Outcome 9: Comparison 9 Vilazodone versus placebo, Outcome 11 Dropouts due to a lack of efficacy

11.10 Vortioxetine versus placebo

A similar number of participants may have dropped out due to a lack of efficacy from the vortioxetine group compared to the placebo group (RR 0.59, 95% CI 0.17 to 1.98; 1 study, 625 participants; Analysis 11.10).

11.10. Analysis.

Comparison 11: Dropouts due to a lack of efficacy, Outcome 10: Comparison 10 Vortioxetine versus placebo, Outcome 11 Dropouts due to a lack of efficacy

12. Dropouts due to adverse effects

Thirty‐two studies with 11,793 participants provided data on dropouts due to adverse effects.

12.1 All antidepressants versus placebo

More participants dropped out due to adverse effects with all the antidepressants compared to placebo (RR 2.18, 95% CI 1.81 to 2.61; I² = 27%, P = 0.07; 32 studies, 11,793 participants; high‐certainty evidence; Analysis 12.1). There was low heterogeneity. The magnitude of effect corresponded to an NNTH of 17 (95% CI 13 to 112). This means that treating 17 people would result in one more person dropping out due to an adverse effect. Visual inspection of the funnel plot suggested that there may have been publication bias (Figure 12).

12.

Funnel plot created in RevMan to assess publication bias for 12.1 Comparison 1 All antidepressants versus placebo, Outcome 12 Dropouts due to adverse effects

12.2 Each class of antidepressant versus placebo

More participants dropped out due to adverse effects with SSRIs, SSNIs, and 'other' antidepressants compared to placebo (SSRIs: RR 1.98, 95% CI 1.51 to 2.61; I² = 23%, P = 0.19; 16 studies, 5315 participants; SNRIs: RR 2.42, 95% CI 1.81 to 3.22; I² = 50%, P = 0.02; 14 studies, 5078 participants; 'other' antidepressants: RR 2.29, 95% CI 1.31 to 4.01; I² = 0%, P = 0.56; 5 studies, 1826 participants) (all high‐certainty evidence; Analysis 12.2). SSRI had low heterogeneity, 'other' antidepressants had no heterogeneity, and the SNRI analysis had moderate‐substantial heterogeneity.

12.2. Analysis.

Comparison 12: Dropouts due to adverse effects, Outcome 2: Comparison 2 Each class of antidepressant versus placebo, Outcome 12 Dropouts due to adverse effects

There were no data for TCAs, MAOIs, NaSSAs, NDRIs, or NRIs for this outcome.

12.3 Agomelatine versus placebo

A similar number of participants may have dropped out due to adverse effects in the agomelatine group compared to the placebo group (RR 1.14, 95% CI 0.36 to 3.56; I² = 0%, P = 0.61; 3 studies, 803 participants; Analysis 12.3). There was no heterogeneity.

12.3. Analysis.

Comparison 12: Dropouts due to adverse effects, Outcome 3: Comparison 3 Agomelatine versus placebo, Outcome 12 Dropouts due to adverse effects

12.4 Duloxetine versus placebo

More participants may have dropped out due to adverse effects in the duloxetine group compared to the placebo group (RR 3.13, 95% CI 1.89 to 5.20; I² = 62%, P = 0.02; 7 studies, 2425 participants; Analysis 12.4). There was substantial heterogeneity.

12.4. Analysis.

Comparison 12: Dropouts due to adverse effects, Outcome 4: Comparison 4 Duloxetine versus placebo, Outcome 12 Dropouts due to adverse effects

12.5 Escitalopram versus placebo

More participants may have dropped out due to adverse effects in the escitalopram group compared to the placebo group (RR 2.02, 95% CI 1.41 to 2.90; I² = 0%, P = 0.79; 8 studies, 2411 participants; Analysis 12.5). There was no heterogeneity.

12.5. Analysis.

Comparison 12: Dropouts due to adverse effects, Outcome 5: Comparison 5 Escitalopram versus placebo, Outcome 12 Dropouts due to adverse effects

12.6 Paroxetine versus placebo

More participants may have dropped out due to adverse effects in the paroxetine group compared to the placebo group (RR 2.40, 95% CI 1.60 to 3.61; I² = 24%, P = 0.25; 7 studies, 2344 participants; Analysis 12.6). There was low heterogeneity.

12.6. Analysis.

Comparison 12: Dropouts due to adverse effects, Outcome 6: Comparison 6 Paroxetine versus placebo, Outcome 12 Dropouts due to adverse effects

12.7 Sertraline versus placebo

A similar number of participants may have dropped out due to adverse effects in the sertraline group compared to the placebo group (RR 1.37, 95% CI 0.39 to 4.76; I² = 68%, P = 0.08; 2 studies, 699 participants; Analysis 12.7). There was substantial heterogeneity.

12.7. Analysis.

Comparison 12: Dropouts due to adverse effects, Outcome 7: Comparison 7 Sertraline versus placebo, Outcome 12 Dropouts due to adverse effects

12.8 Venlafaxine versus placebo

More participants may have dropped out due to adverse effects in the venlafaxine group compared to the placebo group (RR 2.05, 95% CI 1.54 to 2.72; I² = 35%, P = 0.13; 10 studies, 3311 participants; Analysis 12.8). There was low to moderate heterogeneity.

12.8. Analysis.

Comparison 12: Dropouts due to adverse effects, Outcome 8: Comparison 8 Venlafaxine versus placebo, Outcome 12 Dropouts due to adverse effects

12.9 Vilazodone versus placebo

More participants may have dropped out due to adverse effects in the vilazodone group compared to the placebo group (RR 3.11, 95% CI 1.36 to 7.12; 1 study, 398 participants; Analysis 12.9).

12.9. Analysis.

Comparison 12: Dropouts due to adverse effects, Outcome 9: Comparison 9 Vilazodone versus placebo, Outcome 12 Dropouts due to adverse effects

12.10 Vortioxetine versus placebo

A similar number of participants may have dropped out due to adverse effects in the vortioxetine group compared to the placebo group (RR 2.52, 95% CI 0.90 to 7.03; 1 study, 625 participants; Analysis 12.10).

12.10. Analysis.

Comparison 12: Dropouts due to adverse effects, Outcome 10: Comparison 10 Vortioxetine versus placebo, Outcome 12 Dropouts due to adverse effects

Subgroup analyses

Subgroup analyses were not performed as covariates were too unevenly distributed to provide meaningful results.

Sensitivity analyses

Excluding trials with unclear or high risk of bias in random allocation, unclear or high risk of bias in blinding, or both

We performed this sensitivity analysis for the two primary outcomes to investigate whether removing studies with an unclear or high risk of bias rating in the domains of random sequence generation and blinding of participants and personnel had an effect on the study findings.

Removing 10 studies with an unclear or high risk of bias did not substantially change the rate of treatment response measured as a reduction of at least 50% on the HAM‐A (RR 1.45, 95% CI 1.28 to 1.64; I² = 67%; 10 studies, 4745 participants) compared to the original analysis (RR 1.41, 95% CI 1.29 to 1.55; I² = 63%; 20 studies, 7267 participants).

Removing 21 studies with an unclear or high risk of bias did not substantially change acceptability (RR 0.90, 95% CI 0.77 to 1.06; I² = 62%; 12 studies, 5250 participants) compared to the original analysis (RR 1.03, 95% CI 0.93 to 1.14; I² = 52%; 33 studies, 11,294 participants).

Excluding trials whose dropout rate was greater than 20%

We performed this sensitivity analysis for the two primary outcomes.

Removing 15 studies with more than 20% dropout rate did not substantially change the rate of treatment response (RR 1.59, 95% CI 1.22 to 2.06; I² = 79%; 5 studies, 1632 participants) compared to the original analysis (RR 1.41, 95% CI 1.29 to 1.55; I² = 63%; 20 studies, 7267 participants).

Removing 23 studies with more than 20% dropout rate did not substantially change acceptability (RR 0.97, 95% CI 0.77 to 1.23; I² = 52%; 10 studies, 3636 participants) compared to the original analysis (RR 1.03, 95% CI 0.93 to 1.14; I² = 52%; 33 studies, 11,294 participants).

Excluding trials for which the response rates had to be calculated based on the imputation method (Furukawa 2005), and those for which the standard deviation or intraclass correlation coefficient had to be borrowed from other trials (Furukawa 2006)

We performed this sensitivity analysis for the outcome change in symptom levels.

Removing six studies with imputed or pooled SDs did not substantially change the analysis (MD −2.74, 95% CI −3.55 to −1.92; I² = 82%; 28 studies, 8823 participants) compared to the original (MD −2.79, 95% CI −3.49 to −2.08; I² = 80%; 34 studies, 11,230 participants).

Application of fixed‐effect models to determine if the weighting of small studies might have affected the pooled result

We performed this sensitivity analysis using fixed‐effect models to the two primary outcomes.

Fixed‐effect analysis did not substantially change the rate of treatment response (RR 1.39, 95% CI 1.32 to 1.47; I² = 63%; 20 studies, 7267 participants) compared to the original analysis (RR 1.41, 95% CI 1.29 to 1.55; I² = 63%; 20 studies, 7267 participants).

Fixed‐effect analysis did not substantially change acceptability (RR 1.01, 95% CI 0.95 to 1.08; I² = 52%; 33 studies, 11,294 participants) compared to the original analysis (RR 1.03, 95% CI 0.93 to 1.14; I² = 52%; 33 studies, 11,294 participants).

Discussion

Summary of main results

This review included 37 studies (12,226 participants). For the two primary outcomes, there was evidence showing that antidepressants have a benefit over placebo in rate of treatment response measured as a 50% or more reduction on the HAM‐A scale (high‐certainty evidence). Analyses of different classes of antidepressants also showed that SSRIs, SNRIs, and 'other' antidepressants have a benefit over placebo in rate of treatment response measured as a reduction of at least 50% or more on the HAM‐A. There was no difference in acceptability between antidepressants and placebo. There were similar results among different classes of antidepressants (SSRIs, SNRIs, and 'others') (high‐certainty evidence).

Secondary outcomes showed that all antidepressants may have a benefit over placebo for the rate of treatment response (defined by study authors). There were similar results for SSRIs and SNRIs. 'Other' antidepressants may be similar to placebo in rate of treatment response (as defined by study authors). All antidepressants and different classes of antidepressants (SSRIs, SNRIs, and 'others') showed benefit over placebo for remission rates. There was a benefit for all antidepressants over placebo in reducing GAD symptoms, although one study of TCAs (imipramine) showed TCAs and placebo may be similar in reducing symptoms levels.

Limited evidence showed that SSRIs, SNRIs, and all antidepressants may have a benefit over placebo in improving quality of life. Quality of life measures allow researchers to understand the impact of the treatments on people's lives, such as their overall functioning, well‐being, and satisfaction, from the perspective of the patient (Addington‐Hall 2001; Wyrwich 2011). In this review, the Q‐LES‐Q was the primary participant‐reported quality of life measure. The Q‐LES‐Q rates the person's overall satisfaction in subjective domains such as family and social relationships, physical health, work, economic status, and daily functioning. The Q‐LES‐Q has previously shown good psychometric properties in GAD as well as good reliability, validity, and stability in other psychiatric disorders (Wyrwich 2011). Other measures of quality of life that were not used in our review tend to be more disorder‐specific and less generalisable compared to the Q‐LES‐Q (Wyrwich 2011).

In general, more participants may have reported adverse effects in the all antidepressants group compared to the placebo group. This was consistent among all classes of antidepressants (SSRIs, SNRIs, and 'others'). Regarding specific adverse effects, more participants reported sleepiness/drowsiness with all antidepressants compared to placebo and with each class of antidepressant (SSRIs, SNRIs, and 'others') compared to placebo. Limited evidence also suggested there may be no differences between all antidepressants and placebo and classes of antidepressants and placebo in number of participants reporting suicide wishes/gestures/attempts and in participants reporting agitation/anxiety.

Fewer people dropped out due to a lack of efficacy in the antidepressant group and in the SSRIs, NRSIs and 'other' antidepressant groups compared to placebo (high‐certainty evidence). More people dropped out due to adverse effects among all antidepressants, SSRIs, SSNIs, and 'other' antidepressants compared to the placebo group (high‐certainty evidence).

This systematic review and meta‐analysis also investigated the efficacy and acceptability of various antidepressants compared to placebo across multiple outcome measures. Regarding primary outcomes, agomelatine, duloxetine, escitalopram, and venlafaxine showed potential benefits over placebo in increasing the rate of treatment response, measured as a reduction of at least 50% on the HAM‐A scale, with varying levels of heterogeneity observed among the studies. However, sertraline and vortioxetine appeared to have similar rates of treatment response compared to placebo.

In terms of acceptability, agomelatine demonstrated fewer total dropouts compared to placebo, suggesting better tolerability. Duloxetine, escitalopram, sertraline, venlafaxine, vilazodone, and vortioxetine exhibited similar acceptability to placebo, while paroxetine may exhibit higher dropout rates due to adverse effects compared to placebo.

Studies also evaluated secondary outcomes such as remission rate, change in symptom levels, adverse effects, sleepiness/drowsiness, and agitation/anxiety. Overall, the findings suggest that while some antidepressants may have efficacy over placebo in reducing symptoms and improving remission rates, they also exhibited varying levels of adverse effects. Further research and analyses are needed to better understand the comparative effectiveness and tolerability of these antidepressants in treating GAD.

Overall completeness and applicability of evidence

The comprehensive search identified numerous published and unpublished studies for inclusion in this review. These studies reported on many of the predefined outcomes in our protocol (Guaiana 2018), and predominantly compared SSRIs, SNRIs, and 'other' antidepressants to placebo. The SSRIs that were investigated were escitalopram, paroxetine, and sertraline. The SNRIs that were investigated were duloxetine and venlafaxine, while imipramine was the only TCA that was included. 'Other' antidepressants that studies investigated were agomelatine, vilazodone, and vortioxetine. This allowed for a considerable number of participants to be included, and useful information to be derived, for these classes of antidepressants on several predefined efficacy and acceptability measures.

Despite the comprehensive search, we found no studies comparing MAOIs, NaSSAs, NDRIs, or NRIs to placebo. There were also no studies that compared the following SSRIs to placebo: fluoxetine, fluvoxamine, and citalopram; the following SNRIs to placebo: desvenlafaxine and milnacipran; and the following TCAs to placebo: amitriptyline, amoxapine, clomipramine, desipramine, dosulepin/dothiepin, doxepin, lofepramine, maprotiline, nortriptyline, proptriptyline, and trimipramine. 'Other' antidepressants for which there were no studies were: trazodone, nefazodone, mianserin, maprotiline, and non‐conventional herbal products.

Furthermore, we found no data that could be synthesised for the following outcomes: falls, hypotension, completed suicide, subjective memory impairment, deaths, and total number of participants experiencing withdrawal. This precluded any conclusions from being made about these outcomes.

The inclusion and exclusion criteria may have limited the applicability of the results. For example, we excluded people with other medical comorbidities from this review. Although other psychiatric comorbidities were allowed, only two studies included people with other psychiatric comorbidities. Given that many people with GAD also experience other psychiatric and medical comorbidities, the population in this review may not be representative of the typical GAD population. For example, people with GAD often experience comorbid pain syndromes, hypertension, and cardiovascular and gastric conditions (Katzman 2014). Some psychiatric conditions that have been found to be commonly comorbid among people with GAD are major depressive disorder, social phobia, specific phobia, bipolar disorder, alcohol abuse disorder, and panic disorder (Noyes 2001; Simon 2009). Future reviews and studies may also consider including participants with such comorbid conditions.

Studies included in this review also inherently had their own inclusion criteria that further limited the applicability of the results. Studies often excluded people taking other medications to prevent interference with the study results. The high frequency of comorbidities among people with GAD often leads to the use of additional medications. For example, it is common in clinical practice for those initiating treatment with an SSRI or SNRI to also take a benzodiazepine as adjunctive treatment to provide faster relief of symptoms while waiting for the effects of the antidepressants to begin (Rapaport 2006). Conversely, people with comorbid GAD and bipolar disorder may also take additional mood stabilisers (Coplan 2015). However, studies included in this review largely excluded concomitant use of psychotropic drugs to prevent interference with the study drug, and ultimately, the study findings.

Quality of the evidence

The outcomes chosen to assess the certainty of the evidence were: 1. rate of treatment response measured as a reduction of at least 50% on the HAM‐A, 2. acceptability, 3. dropouts due to a lack of efficacy, and 4. dropouts due to adverse effects. The certainty of evidence was high based on GRADE methodology. This means that we have high confidence that the effect estimates for the two primary outcomes and for dropouts due to a lack of efficacy and adverse effects, are close to the true effect.

We did not downgrade the evidence for risk of bias because sensitivity analyses suggested that excluding studies with high risk of bias did not impact the results.

We evaluated certainty of evidence for inconsistency following guidelines described in Guyatt 2011a. The analyses showed strong overlap between CIs and given the analyses pooled studies of different drug dosages, length of follow‐up, diagnostic criteria, settings, and risk of bias, heterogeneity was to be expected.

Participants with other serious medical comorbidities were excluded from this review and there were limited studies that included participants with secondary psychiatric comorbidities. Although many people with GAD also experience other medical and psychiatric comorbidities, we did not downgrade the certainty of the evidence for indirectness because this was not the population of interest for this review. However, future reviews and even RCTs, should investigate the efficacy of antidepressants without excluding participants with comorbidities discussed above. This would increase the applicability of the results.

Imprecision refers to the precision of the effect estimates (Schünemann 2013). For dichotomous outcomes, the GRADE Handbook suggests using a relative risk reduction of 25% when calculating the optimal information size (OIS) (Schünemann 2013). Using Figure 4 in Guyatt 2011b and assuming a relative risk reduction of 25%, all four outcomes met the minimum total sample size (OIS) criteria. Furthermore, the GRADE Handbook suggests using a threshold of appreciable benefit and harm of 25% or more when evaluating imprecision (Schünemann 2013). The effect estimates for the rate of treatment response, dropouts due to a lack of efficacy, and dropouts due to adverse effects excluded the null value and their 95% CIs also excluded the threshold of appreciable benefit and harm of 25%. Therefore, these outcomes were not downgraded for imprecision. Although the effect estimate for acceptability did include the null value, the 95% CIs did not include the threshold for appreciable benefit or harm of 25% and was not downgraded for imprecision.

Finally, we did not downgrade the certainty of the evidence for sponsorship bias as there were no substantial concerns.

Potential biases in the review process

Although we conducted a thorough search of electronic databases, we did not search bibliographies or reach out to experts in the field. There was some asymmetry in the funnel plots suggesting that small studies may have been missed during the search process. Small studies with non‐significant results or serious adverse effects will sometimes be rejected by the journal or not be submitted for publication entirely, which could lead to an overestimation of the benefits and an underestimation of harms (Mlinarić 2017; Siddaway 2019; Turner 2013).

Agreements and disagreements with other studies or reviews

Overall, the results of this meta‐analysis generally agree with the results of other systematic reviews and meta‐analyses comparing antidepressants to placebo for the treatment of GAD. This review adds to the overall understanding of the efficacy and acceptability of antidepressants in the treatment of GAD and provides more detailed information on their tolerability profiles.

The review by Schmitt 2005 investigated imipramine, paroxetine, venlafaxine, and sertraline. They found a benefit of antidepressants over placebo in the risk of non‐response to treatment, which is in line with our finding of higher treatment response with antidepressants. They defined treatment response as a score of 1 or 2 on the CGI, which was also the common definition of treatment response found in this review.

Schmitt 2005 defined acceptability as the total number of people dropping out during the trials and post‐randomisation exclusions, and specific adverse effects. They found no differences in acceptability between all antidepressants and for each type of antidepressant compared to placebo. This result concurs with this review as we also found no differences between antidepressants and placebo in acceptability measured as the total number of dropouts. In terms of adverse effects, they investigated common adverse effects only for venlafaxine, as it was the only antidepressant for which there was more than one study. They found that people taking venlafaxine were more likely to report nausea, dry mouth, insomnia, constipation, somnolence, anorexia, sexual dysfunction, and flatulence. Although our review did not consider all these potential adverse effects and did not specifically examine individual antidepressants, there was a higher risk of people reporting somnolence with SNRIs compared to placebo.

It is evident that Schmitt 2005 was limited in that it only included eight studies investigating imipramine, paroxetine, venlafaxine, and sertraline. Our review was able to retrieve many more studies and with more antidepressants, thus it was a more comprehensive analysis.

Gomez 2018 investigated the efficacy of SSRIs and SNRIs compared to placebo in reducing anxiety symptoms as measured by the HAM‐A. They found that SSRIs and SNRIs were more effective than placebo and were similar in magnitude. These findings are consistent with the current review. Gomez 2018 did not evaluate the safety and tolerability of antidepressants and, therefore, could not be compared to the results of the current review. The current review also investigated other classes of antidepressants, while Gomez 2018 was restricted to SSRIs and SNRIs.

Several other reviews have directly compared antidepressants to placebo in the treatment of GAD. Some network meta‐analyses on pharmacological treatments have included antidepressants as a part of their analyses and have generally found that antidepressants have better efficacy and variable tolerability compared to placebo (He 2019; Slee 2019). Slee 2019 for example, included agomelatine, bupropion, citalopram, duloxetine, escitalopram, fluoxetine, imipramine, maprotiline, mirtazapine, paroxetine, sertraline, venlafaxine, vilazodone, and vortioxetine as a part of their analysis. They compared each individual antidepressant to placebo and generally found them to have better efficacy at reducing GAD symptoms. The exceptions were imipramine, maprotiline, vilazodone, and vortioxetine whose 95% CIs included the potential for no difference. Acceptability, measured as the odds of not completing the study, was mostly no different from placebo, with the exception of paroxetine and vilazodone, which showed worse tolerability compared to placebo.

He 2019 also conducted a network meta‐analysis on the efficacy and acceptability of first‐line treatments in GAD. In particular, their review included studies on duloxetine, escitalopram, fluoxetine, paroxetine, sertraline, venlafaxine, vilazodone, and vortioxetine. Similarly to Slee 2019, He 2019 did not pool the results of their included studies but conducted individual comparisons for each drug–placebo pair. They found greater improvement in symptoms and response with antidepressants compared to placebo, except with fluoxetine and vortioxetine, which was no different from placebo. There were also no differences between antidepressants and placebo in acceptability, except with vilazodone, which had worse acceptability than placebo. Tolerability, measured as the number of participants dropping out due to adverse effects, was higher with antidepressants, except with fluoxetine, sertraline, and vortioxetine, which showed no difference.

Network meta‐analyses are advantageous in that they can compare the relative effectiveness of several interventions both directly and indirectly. The two network meta‐analyses by He 2019 and Slee 2019 discussed above were limited in the variety of outcomes that they investigated as they did not consider improvements in functioning (i.e. quality of life) and specific adverse effects. Moreover, a network meta‐analysis comparing the relative efficacy of antidepressants only could also provide valuable information to clinicians who have patients who do not respond well to other treatments. Future research should consider conducting a network meta‐analysis that compares antidepressants to other antidepressants (and potentially other pharmacological and non‐pharmacological treatments) on a wide range of outcomes. However, network meta‐analyses have limitations. For example, the validity of indirect comparisons in network meta‐analyses strongly depends on the transitivity assumption. This assumption requires trials that are used for indirect comparisons to be similar with respect to their effect modifiers (Cipriani 2013; Higgins 2022). Otherwise, indirect comparisons should not be made. Attempts to improve transitivity can also be challenging if there are a few studies included in each comparison (Cipriani 2013). Nevertheless, a network meta‐analysis would allow clinicians and researchers to gain a more comprehensive understanding of how various pharmacotherapies compare to each other and which results in the best outcomes for people with GAD.

This was the first systematic review and meta‐analysis since Schmitt 2005 to attempt to compare all antidepressants (and only antidepressants) to placebo in the treatment of GAD. Other studies, as described above, have either restricted their analysis to first‐line treatments, or the main purpose was to perform a network meta‐analysis of all pharmacotherapies.

Authors' conclusions

Implications for practice.

The results from the meta‐analyses investigating rate of treatment response, remission rates, and reduction in symptom levels suggest that antidepressants are more effective than placebo at treating generalised anxiety disorder (GAD), with a high degree of confidence. Antidepressants were also comparable to placebo in the total number of dropouts (acceptability) and had fewer dropouts due to lack of efficacy. However, antidepressants were also less tolerable than placebo, as they had greater dropouts due to adverse effects.

Applying a clinically meaningful interpretation of these results is difficult as it depends on the perspective of the person who is considering it and what their goals are (Keefe 2013). Clinicians, patients, policymakers, and health economists for example, may place different emphasis on factors such as the availability of other interventions, the condition of the patients, the risk‐to‐benefit ratio, and the cost of treatment (Keefe 2013). Nonetheless, the results of this review add to the growing literature on antidepressants in the treatment of GAD.

Implications for research.

This review helped to identify some important limitations that currently exist in the literature and some implications for future studies. Briefly, future research should be conducted with higher methodological standards and with more transparency to reduce the risk of bias. No studies had an overall low risk of bias and the distribution of risk of bias among the domains differed between studies. Many studies failed to describe random sequence generation and allocation concealment, resulting in an unclear risk of bias. This highlights the need for more rigorous reporting criteria. Furthermore, there were some concerns with incomplete outcome data as many studies had a high dropout rate and selective outcome reporting.

Future studies may also consider other potential sources of heterogeneity. Although random‐effects meta‐analysis accounts for some unexplained heterogeneity, this should still be investigated to determine whether the effects differ between samples with heterogeneous characteristics. In the current review, none of the sensitivity analyses substantially affected the results compared to the original analysis. Further investigations of sources of heterogeneity could provide stronger conclusions about why it may be occurring and whether effects differ among different subgroups. A network meta‐analysis could be performed given that the quality of the data and the characteristics of the included studies meet the transitivity and coherence requirements needed for valid results to be derived. A network meta‐analysis would allow researchers to investigate head‐to‐head comparisons among antidepressants and establish how they compare to each other.

The subgroup analyses were also limited in the number of included studies and thus could not be performed due to an uneven distribution of covariates. This limits the ability to investigate whether the effects of treatment differed between clinical groups. For example, the double‐blind period in most of the studies that were eligible for this review had a duration of 12 weeks or less. Given that antidepressants can take about four to eight weeks to provide relief of symptoms and up to 12 weeks to experience a full response to treatment, trials of longer duration are needed to better understand the long‐term effects of treatment and how long treatment should be continued (Katzman 2014).

Psychotherapy has also been considered for the treatment of GAD. Although there is strong evidence supporting the use of psychotherapy over placebo in the treatment of GAD, evidence directly comparing psychotherapy to pharmacotherapy is lacking (Katzman 2014). The few studies that have compared psychotherapy to pharmacotherapy, however, have shown that their efficacy is similar in magnitude (Katzman 2014). Furthermore, only a few studies have investigated the effects of combined psycho‐ and pharmacotherapy in the treatment of GAD and have found conflicting results (Katzman 2014). Given this evidence, current guidelines generally recommend that psychotherapy can be used if pharmacotherapies are ineffective and do not recommend combined therapy in the treatment of GAD (Katzman 2014). Future studies should directly compare psychotherapy to pharmacotherapies to clearly establish which is more effective and investigate whether their use in combination is more effective than pharmacotherapy or psychotherapy alone. There is also a lack of evidence investigating different variables that could affect the efficacy of psychotherapy (such as population, setting, treatment duration, frequency, etc.) and future studies should take these factors into consideration.

Furthermore, outcomes such as agitation/anxiety, suicide wishes/gestures/attempts, and average score/change in quality of life/satisfaction were not as commonly reported as some other outcomes. Considering GAD is a very debilitating disorder, with high rates of suicide, these outcomes may be considered important to patients when choosing treatment. More research should investigate these outcomes with antidepressants.

Moreover, GAD is also a chronic and persistent mental disorder as illustrated by one study that found the probability of achieving recovery among people with GAD was 58%, and the probability of recurrence among those who had recovered was 45% after 12 years of follow‐up (Bruce 2005). Relapse prevention studies are generally longer in duration and provide insight into the long‐term efficacy of antidepressants. In particular, studies on duloxetine (Davidson 2008), escitalopram (Allgulander 2006), vortioxetine (Baldwin 2012c), paroxetine (Stocchi 2003), and venlafaxine (Hackett 2000; Rickels 2010) have all found lower relapse rates in those who continued the antidepressant for at least six months after an open‐label period compared to those who switched to placebo. In other words, the risk of relapse was reduced in those who continued taking the antidepressants after initially responding to treatment during the open‐label period. These results suggest that continuing to take the medication even after achieving response/remission may be beneficial in preventing relapse and maintaining efficacy. More studies are needed to understand the benefits and drawbacks of long‐term treatment with antidepressants.

Our search also did not identify many studies investigating the effects of antidepressants compared to placebo in older adults. The search identified only one unpublished trial among veterans aged over 60 years (NCT00701675). One systematic review and meta‐analysis investigating pharmacotherapies among older adults in GAD found that antidepressants have significantly more responders compared to placebo (Gonçalves 2012). However, the authors only identified five studies comparing antidepressants to placebo in this population. Older adults tend to have more cognitive decline and medical comorbidities that require them to take additional medications and that may affect their response to treatment and induce drug–drug interactions. As such, future research on this population to identify the best treatment options would be useful.

What's new

Date	Event	Description
11 February 2025	Amended	Amended to change typographical errors in the Plain Language Summary, Effects of interventions, and Discussion.

History

Protocol first published: Issue 2, 2018
Review first published: Issue 1, 2025

Appendices

Appendix 1. Search strategy: CCMD‐CTR‐references

We searched the CCMD‐CTR‐references register using the following terms:

#1. (general* NEAR2 anxi* or GAD)

#2. antidepress* or anti‐depress* or "anti depress*" or MAOI* or RIMA* or “monoamine oxidase inhibit*” or ((serotonin or norepinephrine or noradrenaline or neurotransmitter* or dopamin*) NEAR (uptake or reuptake or re‐uptake or "re uptake")) or SSRI* or SNRI* or NARI* or SARI* or NDRI* or TCA* or tricyclic* or tetracyclic*

#3. Agomelatine or Alaproclate or Amoxapine or Amineptine or Amitriptylin* or Amitriptylinoxide or Atomoxetine or Befloxatone or Benactyzine or Binospirone or Brofaromine or (Buproprion or Amfebutamone) or Butriptyline or Caroxazone or Cianopramine or Cilobamine or Cimoxatone or Citalopram or (Chlorimipramin* or Clomipramin* or Chlomipramin* or Clomipramine) or Clorgyline or Clovoxamine or (CX157 or Tyrima) or Demexiptiline or Deprenyl or (Desipramine* or Pertofrane) or Desvenlafaxine or Dibenzepin or Diclofensine or Dimetacrin* or Dosulepin or Dothiepin or Doxepin or Duloxetine or Desvenlafaxine or DVS‐233

#4. Escitalopram or Etoperidone or Femoxetine or Fluotracen or Fluoxetine or Fluvoxamine or (Hyperforin or Hypericum or “St John*”) or Imipramin* or Iprindole or Iproniazid* or Ipsapirone or Isocarboxazid* or Levomilnacipran or Lofepramine* or (“Lu AA21004” or Vortioxetine) or "Lu AA24530" or (LY2216684 or Edivoxetine) or Maprotiline or Melitracen or Metapramine or Mianserin or Milnacipran or Minaprine or Mirtazapine or Moclobemide or Nefazodone or Nialamide or Nitroxazepine or Nomifensine or Norfenfluramine or Nortriptylin* or Noxiptilin*

#5. Opipramol or Oxaflozane or Paroxetine or Phenelzine or Pheniprazine or Pipofezine or Pirlindole or Pivagabine or Pizotyline or Propizepine or Protriptylin* or Quinupramine or Reboxetine or Rolipram or Scopolamine or Selegiline or Sertraline or Setiptiline or Teciptiline or Thozalinone or Tianeptin* or Toloxatone or Tranylcypromin* or Trazodone or Trimipramine or Venlafaxine or Viloxazine or Vilazodone or Viqualine or Zalospirone

#6. (#1 and (#2 or #3 or #4 or #5))

Appendix 2. New search – other database search strategies

MEDLINE ALL
(includes: Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, Ovid MEDLINE Daily and Ovid MEDLINE)
via Ovid http://ovidsp.ovid.com/
Date range: 1946 to October 20, 2022
Date searched: 22nd October 2022
Records retrieved: 687

1 *Anxiety Disorders/ (24914)
2 Anxiety Disorders/dt [Drug Therapy] (5405)
3 ((general* adj2 anxi*) or GAD).ti,ab,kf. (21059)
4 or/1‐3 (42562)
5 exp Antidepressive Agents/ (158209)
6 exp Neurotransmitter Uptake Inhibitors/ (154672)
7 exp Monoamine Oxidase Inhibitors/ (22528)
8 (antidepress* or anti depress* or MAOI* or monoamine oxidase inhibit* or noradrenerg* or antiadrenergic or anti adrenergic or SSRI* or SNRI* or TCA* or tricyclic* or tetracyclic* or heterocyclic* or psychotropic*).mp. (236833)
9 (serotonin or norepinephrine or noradrenaline or nor epinephrine or nor adrenaline or neurotransmitt* or dopamine*).mp. (475050)
10 (uptake or reuptake or re‐uptake).mp. (449062)
11 9 and 10 (61850)
12 (Agomelatine or Alaproclate or Amoxapine or Amineptine or Amitriptylin* or Amitriptylinoxide or Atomoxetine or Befloxatone or Benactyzine or Binospirone or Brofaromine or Bupropion or Amfebutamone or Butriptyline or Caroxazone or Cianopramine or Cilobamine or Cimoxatone or Citalopram or Chlorimipramin* or Clomipramin* or Chlomipramin* or Clomipramine or Clorgyline or Clovoxamine or CX157 or Tyrima or Demexiptiline or Deprenyl or Desipramine* or Pertofrane or Desvenlafaxine or Dibenzepin or Diclofensine or Dimetacrin* or Dosulepin* or Dothiepin or Doxepin* or Duloxetine or Desvenlafaxine or DVS‐233 or Escitalopram or Etoperidone or Femoxetine or Fluotracen or Fluoxetine or Fluvoxamine or Hyperforin or Hypericum or St John* or Imipramin* or Iprindole or Iproniazid* or Ipsapirone or Isocarboxazid* or Levomilnacipran or Lofepramine* or Lu AA21004 or Vortioxetine or Lu AA24530 or LY2216684 or Edivoxetine or Maprotiline or Melitracen or Metapramine or Mianserin or Milnacipran or Minaprine or Mirtazapine or Moclobemide or Nefazodone or Nialamide or Nitroxazepine or Nomifensine or Norfenfluramine or Nortriptylin* or Noxiptilin* or Opipramol or Oxaflozane or Paroxetine or Phenelzine or Pheniprazine or Pipofezine or Pirlindole or Pivagabine or Pizotyline or Propizepine or Protriptylin* or Quinupramine or Reboxetine or Rolipram or Scopolamine or Selegiline or Sertraline or Setiptiline or Teciptiline or Thozalinone or Tianeptin* or Toloxatone or Tranylcypromin* or Trazodone or Trimipramine or Venlafaxine or Viloxazine or Vilazodone or Viqualine or Zalospirone).mp. (112144)
13 5 or 6 or 7 or 8 or 11 or 12 (451731)
14 4 and 13 (5110)
15 randomized controlled trial.pt. (579185)
16 randomi#ed.ti,ab,kf. (751604)
17 controlled clinical trial.pt. (95077)
18 Double‐Blind Method/ (173344)
19 clinical trials as topic.sh. (200471)
20 randomly.ab. (393763)
21 (RCT or at random or (random* adj (assign* or allocat* or divid* or division or number))).ti,ab,kf. (269931)
22 trial.ti,kf. (291288)
23 (animals not (humans and animals)).sh. (5023551)
24 or/15‐22 (1531971)
25 24 not 23 (1416231)
26 (placebo* or dummy or sugar pill).mp. (259201)
27 14 and 25 and 26 (687)

Embase
via Ovid http://ovidsp.ovid.com/
Date range: 1974 to 2022 October 20
Date searched: 22nd October 2022
Records retrieved: 935

1 *anxiety disorder/ (24794)
2 anxiety disorder/dt [Drug Therapy] (12695)
3 generalized anxiety disorder/ (13938)
4 ((general* adj2 anxi*) or GAD).ti,ab,kw. (30051)
5 or/1‐4 (64151)
6 exp antidepressant agent/ (542965)
7 exp serotonin uptake inhibitor/ (302315)
8 exp serotonin noradrenalin reuptake inhibitor/ (202288)
9 exp noradrenalin uptake inhibitor/ (251434)
10 (Agomelatine or Alaproclate or Amoxapine or Amineptine or Amitriptylin* or Amitriptylinoxide or Atomoxetine or Befloxatone or Benactyzine or Binospirone or Brofaromine or Bupropion or Amfebutamone or Butriptyline or Caroxazone or Cianopramine or Cilobamine or Cimoxatone or Citalopram or Chlorimipramin* or Clomipramin* or Chlomipramin* or Clomipramine or Clorgyline or Clovoxamine or CX157 or Tyrima or Demexiptiline or Deprenyl or Desipramine* or Pertofrane or Desvenlafaxine or Dibenzepin or Diclofensine or Dimetacrin* or Dosulepin* or Dothiepin or Doxepin* or Duloxetine or Desvenlafaxine or DVS‐233 or Escitalopram or Etoperidone or Femoxetine or Fluotracen or Fluoxetine or Fluvoxamine or Hyperforin or Hypericum or St John* or Imipramin* or Iprindole or Iproniazid* or Ipsapirone or Isocarboxazid* or Levomilnacipran or Lofepramine* or Lu AA21004 or Vortioxetine or Lu AA24530 or LY2216684 or Edivoxetine or Maprotiline or Melitracen or Metapramine or Mianserin or Milnacipran or Minaprine or Mirtazapine or Moclobemide or Nefazodone or Nialamide or Nitroxazepine or Nomifensine or Norfenfluramine or Nortriptylin* or Noxiptilin* or Opipramol or Oxaflozane or Paroxetine or Phenelzine or Pheniprazine or Pipofezine or Pirlindole or Pivagabine or Pizotyline or Propizepine or Protriptylin* or Quinupramine or Reboxetine or Rolipram or Scopolamine or Selegiline or Sertraline or Setiptiline or Teciptiline or Thozalinone or Tianeptin* or Toloxatone or Tranylcypromin* or Trazodone or Trimipramine or Venlafaxine or Viloxazine or Vilazodone or Viqualine or Zalospirone).mp. (260282)
11 (antidepress* or anti depress* or MAOI* or monoamine oxidase inhibit* or ((serotonin or norepinephrine or noradrenaline or nor epinephrine or nor adrenaline or neurotransmitt* or dopamine*) and (uptake or reuptake or re‐uptake)) or noradrenerg* or antiadrenergic or anti adrenergic or SSRI* or SNRI* or TCA* or tricyclic* or tetracyclic* or heterocyclic* or psychotropic*).mp. (374928)
12 or/6‐11 (734512)
13 major clinical study/ (4639494)
14 Randomized controlled trial/ (733077)
15 Controlled clinical study/ (467336)
16 double blind procedure/ (199840)
17 randomization/ (95355)
18 (RCT or randomi#ed).ti,ab,kw. (1086039)
19 ((at random or random*) adj2 (allocat* or assign* or divide* or division or number)).ti,ab,kw. (329369)
20 ((double or single or doubly or singly) adj (blind or blinded or blindly)).ti,ab,kw. (261848)
21 or/13‐20 (5843011)
22 ((animal or nonhuman) not (human and (animal or nonhuman))).de. (6224878)
23 21 not 22 (5686058)
24 5 and 12 and 23 (4596)
25 (placebo* or dummy or sugar pill).mp. (509246)
26 24 and 25 (1170)
27 elsevier.cr. (29206743)
28 26 and 27 (1119)
29 (random* adj sampl* adj7 ("cross section*" or questionnaire*1 or survey* or database*1)).ti,ab. not (comparative study/ or controlled study/ or randomi?ed controlled.ti,ab. or randomly assigned.ti,ab.) (9159)
30 Cross‐sectional study/ not (randomized controlled trial/ or controlled clinical study/ or controlled study/ or randomi?ed controlled.ti,ab. or control group*1.ti,ab.) (324230)
31 (((case adj control*) and random*) not randomi?ed controlled).ti,ab. (20366)
32 (Systematic review not (trial or study)).ti. (225598)
33 (review.ab. and review.pt.) not trial.ti. (1033830)
34 or/29‐33 (1514911)
35 28 not 34 (935)

APA PsycINFO
via Ovid http://ovidsp.ovid.com/
Date range: 1806 to October Week 3 2022
Date searched: 22nd October 2022
Records retrieved: 353

1 generalized anxiety disorder/ (3442)
2 ((general* adj2 anxi*) or GAD).ti,ab,id. (14609)
3 *anxiety disorders/ (16026)
4 1 or 2 or 3 (27484)
5 exp antidepressant drugs/ (41005)
6 neurotransmitter uptake inhibitors/ or exp serotonin norepinephrine reuptake inhibitors/ or exp serotonin reuptake inhibitors/ (14356)
7 exp monoamine oxidase inhibitors/ (2299)
8 exp tricyclic antidepressant drugs/ (6493)
9 (antidepress* or anti depress* or MAOI* or monoamine oxidase inhibit* or ((serotonin or norepinephrine or noradrenaline or nor epinephrine or nor adrenaline or neurotransmitt* or dopamine*) and (uptake or reuptake or re‐uptake)) or noradrenerg* or antiadrenergic or anti adrenergic or SSRI* or SNRI* or TCA* or tricyclic* or tetracyclic*or heterocyclic* or psychotropic*).ti,ab,id,hw. (75055)
10 (Agomelatine or Alaproclate or Amoxapine or Amineptine or Amitriptylin* or Amitriptylinoxide or Atomoxetine or Befloxatone or Benactyzine or Binospirone or Brofaromine or Bupropion or Amfebutamone or Butriptyline or Caroxazone or Cianopramine or Cilobamine or Cimoxatone or Citalopram or Chlorimipramin* or Clomipramin* or Chlomipramin* or Clomipramine or Clorgyline or Clovoxamine or CX157 or Tyrima or Demexiptiline or Deprenyl or Desipramine* or Pertofrane or Desvenlafaxine or Dibenzepin or Diclofensine or Dimetacrin* or Dosulepin* or Dothiepin or Doxepin* or Duloxetine or Desvenlafaxine or DVS‐233 or Escitalopram or Etoperidone or Femoxetine or Fluotracen or Fluoxetine or Fluvoxamine or Hyperforin or Hypericum or St John* or Imipramin* or Iprindole or Iproniazid* or Ipsapirone or Isocarboxazid* or Levomilnacipran or Lofepramine* or Lu AA21004 or Vortioxetine or Lu AA24530 or LY2216684 or Edivoxetine or Maprotiline or Melitracen or Metapramine or Mianserin or Milnacipran or Minaprine or Mirtazapine or Moclobemide or Nefazodone or Nialamide or Nitroxazepine or Nomifensine or Norfenfluramine or Nortriptylin* or Noxiptilin* or Opipramol or Oxaflozane or Paroxetine or Phenelzine or Pheniprazine or Pipofezine or Pirlindole or Pivagabine or Pizotyline or Propizepine or Protriptylin* or Quinupramine or Reboxetine or Rolipram or Scopolamine or Selegiline or Sertraline or Setiptiline or Teciptiline or Thozalinone or Tianeptin* or Toloxatone or Tranylcypromin* or Trazodone or Trimipramine or Venlafaxine or Viloxazine or Vilazodone or Viqualine or Zalospirone).ti,ab,id,hw. (38290)
11 or/5‐10 (95598)
12 4 and 11 (2306)
13 (RCT or at random or (random* adj (assign* or allocat* or divid* or division or number))).ti,ab,id. (58796)
14 trial.ti,id. (41651)
15 randomi#ed.ti,ab,id. (100128)
16 ((singl* or doubl* or trebl* or tripl*) adj3 (blind* or mask* or dummy)).ti,ab,id. (28612)
17 or/13‐16 (165038)
18 (placebo* or dummy or sugar pill).ti,ab,id,hw. (44515)
19 12 and 17 and 18 (353)

Cochrane Central Register of Controlled Trials (CENTRAL)
via The Cochrane Library, Wiley http://www.cochranelibrary.com/
Issue 10 of 12, October 2022
Date searched: 22nd October 2022
Records retrieved: 949

#1 (generalised or generalized) near anxiety:ti,ab,kw 3894
#2 GAD:ab 2537
#3 #1 or #2 4786
#4 (Agomelatine or Alaproclate or Amoxapine or Amineptine or Amitriptylin* or Amitriptylinoxide or Atomoxetine or Befloxatone or Benactyzine or Binospirone or Brofaromine or Bupropion or Amfebutamone or Butriptyline or Caroxazone or Cianopramine or Cilobamine or Cimoxatone or Citalopram or Chlorimipramin* or Clomipramin* or Chlomipramin* or Clomipramine or Clorgyline or Clovoxamine or CX157 or Tyrima or Demexiptiline or Deprenyl or Desipramine* or Pertofrane or Desvenlafaxine or Dibenzepin or Diclofensine or Dimetacrin* or Dosulepin* or Dothiepin or Doxepin* or Duloxetine or Desvenlafaxine or DVS‐233 or Escitalopram or Etoperidone or Femoxetine or Fluotracen or Fluoxetine or Fluvoxamine or Hyperforin or Hypericum or (St next John*) or Imipramin* or Iprindole or Iproniazid* or Ipsapirone or Isocarboxazid* or Levomilnacipran or Lofepramine* or "Lu AA21004" or Vortioxetine or "Lu AA24530" or LY2216684 or Edivoxetine or Maprotiline or Melitracen or Metapramine or Mianserin or Milnacipran or Minaprine or Mirtazapine or Moclobemide or Nefazodone or Nialamide or Nitroxazepine or Nomifensine or Norfenfluramine or Nortriptylin* or Noxiptilin* or Opipramol or Oxaflozane or Paroxetine or Phenelzine or Pheniprazine or Pipofezine or Pirlindole or Pivagabine or Pizotyline or Propizepine or Protriptylin* or Quinupramine or Reboxetine or Rolipram or Scopolamine or Selegiline or Sertraline or Setiptiline or Teciptiline or Thozalinone or Tianeptin* or Toloxatone or Tranylcypromin* or Trazodone or Trimipramine or Venlafaxine or Viloxazine or Vilazodone or Viqualine or Zalospirone) 28414
#5 antidepress* or anti‐depress* or MAOI* or monoamine next oxidase next inhibit* or ((serotonin or norepinephrine or noradrenaline or nor next epinephrine or nor next adrenaline or neurotransmitt* or dopamine*) and (uptake or reuptake or re‐uptake)) or noradrenerg* or antiadrenergic or anti next adrenergic or SSRI* or SNRI* or TCA* or tricyclic* or tetracyclic* or heterocyclic* or psychotropic* 28369
#6 #4 or #5 43961
#7 #3 and #6 in Trials 949

Data and analyses

Comparison 1. Rate of treatment response measured as a reduction of at least 50% on the Hamilton Anxiety Scale (HAM‐A).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Comparison 1 All antidepressants versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A	20	7267	Risk Ratio (IV, Random, 95% CI)	1.41 [1.29, 1.55]
1.2 Comparison 2 Each class of antidepressant versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A	20		Risk Ratio (IV, Random, 95% CI)	Subtotals only
1.2.1 SSRIs	4	1226	Risk Ratio (IV, Random, 95% CI)	1.51 [1.20, 1.90]
1.2.2 SNRIs	14	4659	Risk Ratio (IV, Random, 95% CI)	1.34 [1.21, 1.47]
1.2.3 Other	5	1804	Risk Ratio (IV, Random, 95% CI)	1.54 [1.14, 2.08]
1.3 Comparison 3 Agomelatine versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A	3	733	Risk Ratio (IV, Random, 95% CI)	1.93 [1.38, 2.70]
1.4 Comparison 4 Duloxetine versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A	6	2070	Risk Ratio (IV, Random, 95% CI)	1.39 [1.22, 1.58]
1.5 Comparison 5 Escitalopram versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A	2	530	Risk Ratio (IV, Random, 95% CI)	1.50 [1.05, 2.16]
1.6 Comparison 6 Sertraline versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A	2	696	Risk Ratio (IV, Random, 95% CI)	1.52 [0.99, 2.35]
1.7 Comparison 7 Venlafaxine versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A	10	2913	Risk Ratio (IV, Random, 95% CI)	1.32 [1.18, 1.49]
1.8 Comparison 8 Vilazodone versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A	1	395	Risk Ratio (IV, Random, 95% CI)	1.25 [1.01, 1.55]
1.9 Comparison 9 Vortioxetine versus placebo, Outcome 1 Rate of treatment response measured as a reduction of at least 50% on the HAM‐A	1	610	Risk Ratio (IV, Random, 95% CI)	1.04 [0.85, 1.29]

Comparison 2. Acceptability.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Comparison 1 All antidepressants versus placebo, Outcome 2 Acceptability	33	11294	Risk Ratio (IV, Random, 95% CI)	1.03 [0.93, 1.14]
2.2 Comparison 2 Each class of antidepressant versus placebo, Outcome 2 Acceptability	33		Risk Ratio (IV, Random, 95% CI)	Subtotals only
2.2.1 SSRIs	16	5031	Risk Ratio (IV, Random, 95% CI)	1.06 [0.95, 1.19]
2.2.2 SNRIs	15	4863	Risk Ratio (IV, Random, 95% CI)	1.03 [0.87, 1.21]
2.2.3 Other	5	1830	Risk Ratio (IV, Random, 95% CI)	0.90 [0.60, 1.34]
2.3 Comparison 3 Agomelatine versus placebo, Outcome 2 Acceptability	3	803	Risk Ratio (IV, Random, 95% CI)	0.60 [0.44, 0.83]
2.4 Comparison 4 Duloxetine versus placebo, Outcome 2 Acceptability	5	1749	Risk Ratio (IV, Random, 95% CI)	1.07 [0.82, 1.39]
2.5 Comparison 5 Escitalopram versus placebo, Outcome 2 Acceptability	7	2197	Risk Ratio (IV, Random, 95% CI)	0.99 [0.83, 1.17]
2.6 Comparison 6 Paroxetine versus placebo, Outcome 2 Acceptability	7	2215	Risk Ratio (IV, Random, 95% CI)	1.25 [1.05, 1.47]
2.7 Comparison 7 Sertraline versus placebo, Outcome 2 Acceptability	3	758	Risk Ratio (IV, Random, 95% CI)	0.95 [0.76, 1.20]
2.8 Comparison 8 Venlafaxine versus placebo, Outcome 2 Acceptability	12	3693	Risk Ratio (IV, Random, 95% CI)	0.97 [0.81, 1.16]
2.9 Comparison 9 Vilazodone versus placebo, Outcome 2 Acceptability	1	402	Risk Ratio (IV, Random, 95% CI)	1.42 [1.00, 2.03]
2.10 Comparison 10 Vortioxetine versus placebo, Outcome 2 Acceptability	1	625	Risk Ratio (IV, Random, 95% CI)	1.12 [0.81, 1.55]

Comparison 3. Rate of treatment response (defined by study authors).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Comparison 1 All antidepressants versus placebo, Outcome 3 Rate of treatment response (defined by study authors)	18	6613	Risk Ratio (IV, Random, 95% CI)	1.35 [1.27, 1.43]
3.2 Comparison 2 Each class of antidepressant versus placebo, Outcome 3 Rate of treatment response (defined by study authors)	18		Risk Ratio (IV, Random, 95% CI)	Subtotals only
3.2.1 SSRIs	10	3661	Risk Ratio (IV, Random, 95% CI)	1.31 [1.20, 1.44]
3.2.2 SNRIs	8	2693	Risk Ratio (IV, Random, 95% CI)	1.41 [1.31, 1.53]
3.2.3 Other	1	395	Risk Ratio (IV, Random, 95% CI)	1.20 [1.00, 1.43]
3.3 Comparison 3 Duloxetine versus placebo, Outcome 3 Rate of treatment response (defined by study authors)	1	323	Risk Ratio (IV, Random, 95% CI)	1.33 [1.06, 1.68]
3.4 Comparison 4 Escitalopram versus placebo, Outcome 3 Rate of treatment response (defined by study authors)	3	1105	Risk Ratio (IV, Random, 95% CI)	1.29 [1.12, 1.50]
3.5 Comparison 5 Paroxetine versus placebo, Outcome 3 Rate of treatment response (defined by study authors)	6	1998	Risk Ratio (IV, Random, 95% CI)	1.25 [1.11, 1.40]
3.6 Comparison 6 Sertraline versus placebo, Outcome 3 Rate of treatment response (defined by study authors)	2	696	Risk Ratio (IV, Random, 95% CI)	1.41 [1.00, 2.00]
3.7 Comparison 7 Venlafaxine versus placebo, Outcome 3 Rate of treatment response (defined by study authors)	8	2531	Risk Ratio (IV, Random, 95% CI)	1.42 [1.31, 1.54]
3.8 Comparison 8 Vilazodone versus placebo, Outcome 3 Rate of treatment response (defined by study authors)	1	395	Risk Ratio (IV, Random, 95% CI)	1.20 [1.00, 1.43]

Comparison 4. Remission rate.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 Comparison 1 All antidepressants versus placebo, Outcome 4 Remission rate	17	6286	Risk Ratio (IV, Random, 95% CI)	1.54 [1.36, 1.73]
4.2 Comparison 2 Each class of antidepressant versus placebo, Outcome 4 Remission rate	17		Risk Ratio (IV, Random, 95% CI)	Subtotals only
4.2.1 SSRIs	7	2827	Risk Ratio (IV, Random, 95% CI)	1.41 [1.22, 1.64]
4.2.2 SNRIs	8	2639	Risk Ratio (IV, Random, 95% CI)	1.57 [1.28, 1.92]
4.2.3 Other	4	1088	Risk Ratio (IV, Random, 95% CI)	1.77 [1.27, 2.45]
4.3 Comparison 3 Agomelatine versus placebo, Outcome 4 Remission rate	3	799	Risk Ratio (IV, Random, 95% CI)	2.06 [1.57, 2.69]
4.4 Comparison 4 Duloxetine versus placebo, Outcome 4 Remission rate	5	1767	Risk Ratio (IV, Random, 95% CI)	1.53 [1.19, 1.96]
4.5 Comparison 5 Escitalopram versus placebo, Outcome 4 Remission rate	3	1068	Risk Ratio (IV, Random, 95% CI)	1.48 [1.21, 1.81]
4.6 Comparison 6 Paroxetine versus placebo, Outcome 4 Remission rate	4	1527	Risk Ratio (IV, Random, 95% CI)	1.31 [1.03, 1.65]
4.7 Comparison 7 Sertraline versus placebo, Outcome 4 Remission rate	1	370	Risk Ratio (IV, Random, 95% CI)	1.70 [1.17, 2.47]
4.8 Comparison 8 Venlafaxine versus placebo, Outcome 4 Remission rate	5	1196	Risk Ratio (IV, Random, 95% CI)	1.74 [1.28, 2.38]
4.9 Comparison 8 Vortioxetine versus placebo, Outcome 4 Remission rate	1	289	Risk Ratio (IV, Random, 95% CI)	1.18 [0.78, 1.78]

Comparison 5. Change in symptom levels.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
5.1 Comparison 1 All antidepressants versus placebo, Outcome 5 Change in symptom levels	34	11230	Mean Difference (IV, Random, 95% CI)	‐2.79 [‐3.49, ‐2.08]
5.2 Comparison 2 Each class of antidepressant versus placebo, Outcome 5 Change in symptom levels	34		Mean Difference (IV, Random, 95% CI)	Subtotals only
5.2.1 TCAs	1	28	Mean Difference (IV, Random, 95% CI)	‐3.90 [‐9.49, 1.69]
5.2.2 SSRIs	15	4689	Mean Difference (IV, Random, 95% CI)	‐2.22 [‐2.88, ‐1.55]
5.2.3 SNRIs	16	5244	Mean Difference (IV, Random, 95% CI)	‐3.04 [‐4.12, ‐1.97]
5.2.4 Other	5	1691	Mean Difference (IV, Random, 95% CI)	‐3.81 [‐7.07, ‐0.54]
5.3 Comparison 4 Agomelatine versus placebo, Outcome 4 Change in symptom levels	3	799	Mean Difference (IV, Random, 95% CI)	‐5.98 [‐9.37, ‐2.59]
5.4 Comparison 5 Duloxetine versus placebo, Outcome 4 Change in symptom levels	6	2070	Mean Difference (IV, Random, 95% CI)	‐3.03 [‐3.75, ‐2.32]
5.5 Comparison 6 Escitalopram versus placebo, Outcome 4 Change in symptom levels	6	1908	Mean Difference (IV, Random, 95% CI)	‐2.58 [‐3.60, ‐1.55]
5.6 Comparison 7 Imipramine versus placebo, Outcome 4 Change in symptom levels	1	28	Mean Difference (IV, Random, 95% CI)	‐3.90 [‐9.49, 1.69]
5.7 Comparison 8 Paroxetine versus placebo, Outcome 4 Change in symptom levels	7	2183	Mean Difference (IV, Random, 95% CI)	‐1.71 [‐2.69, ‐0.74]
5.8 Comparison 9 Sertraline versus placebo, Outcome 4 Change in symptom levels	3	736	Mean Difference (IV, Random, 95% CI)	‐2.00 [‐4.15, 0.15]
5.9 Comparison 10 Venlafaxine versus placebo, Outcome 4 Change in symptom levels	12	3498	Mean Difference (IV, Random, 95% CI)	‐3.11 [‐4.60, ‐1.63]
5.10 Comparison 11 Vilazodone versus placebo, Outcome 4 Change in symptom levels	1	395	Mean Difference (IV, Random, 95% CI)	‐1.30 [‐2.77, 0.17]
5.11 Comparison 12 Vortioxetine versus placebo, Outcome 4 Change in symptom levels	1	497	Mean Difference (IV, Random, 95% CI)	‐0.50 [‐1.74, 0.74]

Comparison 6. Total number of participants reporting adverse events.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
6.1 Comparison 1 All antidepressants versus placebo, Outcome 6 Total number of participants reporting adverse events	23	7615	Risk Ratio (IV, Random, 95% CI)	1.15 [1.12, 1.19]
6.2 Comparison 2 Each class of antidepressant versus placebo, Outcome 6 Total number of participants reporting adverse events	23		Risk Ratio (IV, Random, 95% CI)	Subtotals only
6.2.1 SSRIs	14	4331	Risk Ratio (IV, Random, 95% CI)	1.16 [1.12, 1.21]
6.2.2 SNRIs	7	1887	Risk Ratio (IV, Random, 95% CI)	1.14 [1.07, 1.22]
6.2.3 Other	5	1821	Risk Ratio (IV, Random, 95% CI)	1.21 [1.07, 1.37]
6.3 Comparison 3 Agomelatine versus placebo, Outcome 6 Total number of participants reporting adverse events	3	801	Risk Ratio (IV, Random, 95% CI)	1.13 [0.94, 1.37]
6.4 Comparison 4 Duloxetine versus placebo, Outcome 6 Total number of participants reporting adverse events	4	1169	Risk Ratio (IV, Random, 95% CI)	1.17 [1.10, 1.25]
6.5 Comparison 5 Escitalopram versus placebo, Outcome 6 Total number of participants reporting adverse events	8	2409	Risk Ratio (IV, Random, 95% CI)	1.13 [1.02, 1.24]
6.6 Comparison 6 Paroxetine versus placebo, Outcome 6 Total number of participants reporting adverse events	6	2019	Risk Ratio (IV, Random, 95% CI)	1.06 [0.90, 1.24]
6.7 Comparison 7 Sertraline versus placebo, Outcome 6 Total number of participants reporting adverse events	1	42	Risk Ratio (IV, Random, 95% CI)	1.17 [0.35, 3.84]
6.8 Comparison 8 Venlafaxine versus placebo, Outcome 6 Total number of participants reporting adverse events	4	879	Risk Ratio (IV, Random, 95% CI)	1.11 [1.01, 1.23]
6.9 Comparison 9 Vilazodone versus placebo, Outcome 6 Total number of participants reporting adverse events	1	398	Risk Ratio (IV, Random, 95% CI)	1.39 [1.22, 1.59]
6.10 Comparison 10 Vortioxetine versus placebo, Outcome 6 Total number of participants reporting adverse events	1	622	Risk Ratio (IV, Random, 95% CI)	1.13 [1.01, 1.27]

Comparison 7. Sleepiness/drowsiness.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
7.1 Comparison 1 All antidepressants versus placebo, Outcome 7 Sleepiness/drowsiness	23	8738	Risk Ratio (IV, Random, 95% CI)	2.30 [1.91, 2.77]
7.2 Comparison 2 Each class of antidepressant versus placebo, Outcome 7 Sleepiness/drowsiness	23		Risk Ratio (IV, Random, 95% CI)	Subtotals only
7.2.1 SSRIs	12	3933	Risk Ratio (IV, Random, 95% CI)	2.15 [1.70, 2.71]
7.2.2 SNRIs	10	3529	Risk Ratio (IV, Random, 95% CI)	2.66 [1.92, 3.69]
7.2.3 Other	4	1700	Risk Ratio (IV, Random, 95% CI)	1.78 [1.01, 3.12]
7.3 Comparison 3 Agomelatine versus placebo, Outcome 7 Sleepiness/drowsiness	2	680	Risk Ratio (IV, Random, 95% CI)	1.94 [0.60, 6.27]
7.4 Comparison 4 Duloxetine versus placebo, Outcome 7 Sleepiness/drowsiness	6	2094	Risk Ratio (IV, Random, 95% CI)	4.03 [2.47, 6.58]
7.5 Comparison 5 Escitalopram versus placebo, Outcome 7 Sleepiness/drowsiness	6	1933	Risk Ratio (IV, Random, 95% CI)	1.88 [1.32, 2.67]
7.6 Comparison 6 Paroxetine versus placebo, Outcome 7 Sleepiness/drowsiness	7	2294	Risk Ratio (IV, Random, 95% CI)	2.33 [1.71, 3.18]
7.7 Comparison 7 Sertraline versus placebo, Outcome 7 Sleepiness/drowsiness	1	42	Risk Ratio (IV, Random, 95% CI)	0.17 [0.01, 3.98]
7.8 Comparison 8 Venlafaxine versus placebo, Outcome 7 Sleepiness/drowsiness	6	1766	Risk Ratio (IV, Random, 95% CI)	2.20 [1.54, 3.14]
7.9 Comparison 9 Vilazodone versus placebo, Outcome 7 Sleepiness/drowsiness	1	398	Risk Ratio (IV, Random, 95% CI)	1.98 [0.76, 5.17]
7.10 Comparison 10 Vortioxetine versus placebo, Outcome 7 Sleepiness/drowsiness	1	622	Risk Ratio (IV, Random, 95% CI)	1.55 [0.65, 3.67]

Comparison 8. Agitation/anxiety.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
8.1 Comparison 1 All antidepressants versus placebo, Outcome 8 Agitation/anxiety	6	2026	Risk Ratio (IV, Random, 95% CI)	1.06 [0.74, 1.53]
8.2 Comparison 2 Each class of antidepressant versus placebo, Outcome 8 Agitation/anxiety	6		Risk Ratio (IV, Random, 95% CI)	Subtotals only
8.2.1 SSRIs	5	1783	Risk Ratio (IV, Random, 95% CI)	1.11 [0.60, 2.05]
8.2.2 SNRIs	1	243	Risk Ratio (IV, Random, 95% CI)	0.99 [0.57, 1.72]
8.3 Comparison 3 Escitalopram versus placebo, Outcome 8 Agitation/anxiety	2	857	Risk Ratio (IV, Random, 95% CI)	0.53 [0.06, 4.79]
8.4 Comparison 4 Paroxetine versus placebo, Outcome 8 Agitation/anxiety	2	650	Risk Ratio (IV, Random, 95% CI)	1.80 [0.67, 4.84]
8.5 Comparison 5 Sertraline versus placebo, Outcome 8 Agitation/anxiety	2	415	Risk Ratio (IV, Random, 95% CI)	1.10 [0.60, 2.00]
8.6 Comparison 6 Venlafaxine versus placebo, Outcome 8 Agitation/anxiety	1	243	Risk Ratio (IV, Random, 95% CI)	0.99 [0.57, 1.72]

Comparison 9. Suicide wishes/gestures/attempts.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
9.1 Comparison 1 All antidepressants versus placebo, Outcome 9 Suicide wishes/gestures/attempts	3	802	Risk Ratio (IV, Random, 95% CI)	0.74 [0.40, 1.36]
9.2 Comparison 2 Each class of antidepressant versus placebo, Outcome 9 Suicide wishes/gestures/attempts	3		Risk Ratio (IV, Random, 95% CI)	Subtotals only
9.2.1 SSRIs	1	197	Risk Ratio (IV, Random, 95% CI)	5.15 [0.25, 105.98]
9.2.2 SNRIs	1	210	Risk Ratio (IV, Random, 95% CI)	0.54 [0.16, 1.79]
9.2.3 Other	1	395	Risk Ratio (IV, Random, 95% CI)	0.75 [0.36, 1.54]
9.3 Comparison 3 Duloxetine versus placebo, Outcome 9 Suicide wishes/gestures/attempts	1	210	Risk Ratio (IV, Random, 95% CI)	0.54 [0.16, 1.79]
9.4 Comparison 4 Paroxetine versus placebo, Outcome 9 Suicide wishes/gestures/attempts	1	197	Risk Ratio (IV, Random, 95% CI)	5.15 [0.25, 105.98]
9.5 Comparison 5 Vilazodone versus placebo, Outcome 9 Suicide wishes/gestures/attempts	1	395	Risk Ratio (IV, Random, 95% CI)	0.75 [0.36, 1.54]

Comparison 10. Average score/change in quality of life/satisfaction.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
10.1 Comparison 1 All antidepressants versus placebo, Outcome 10 Average score/change in quality of life/satisfaction	4	1013	Mean Difference (IV, Random, 95% CI)	6.51 [4.95, 8.07]
10.2 Comparison 2 Each class of antidepressant versus placebo, Outcome 10 Average score/change in quality of life/satisfaction	4		Mean Difference (IV, Random, 95% CI)	Subtotals only
10.2.1 SSRIs	3	760	Mean Difference (IV, Random, 95% CI)	6.70 [5.05, 8.35]
10.2.2 SNRIs	1	253	Mean Difference (IV, Random, 95% CI)	5.00 [0.29, 9.71]
10.3 Comparison 3 Escitalopram versus placebo, Outcome 10 Average score/change in quality of life/satisfaction	1	307	Mean Difference (IV, Random, 95% CI)	6.70 [4.46, 8.94]
10.4 Comparison 4 Paroxetine versus placebo, Outcome 10 Average score/change in quality of life/satisfaction	1	113	Mean Difference (IV, Random, 95% CI)	7.20 [0.97, 13.43]
10.5 Comparison 5 Sertraline versus placebo, Outcome 10 Average score/change in quality of life/satisfaction	1	340	Mean Difference (IV, Random, 95% CI)	6.60 [3.96, 9.24]
10.6 Comparison 6 Venlafaxine versus placebo, Outcome 10 Average score/change in quality of life/satisfaction	1	253	Mean Difference (IV, Random, 95% CI)	5.00 [0.29, 9.71]

Comparison 11. Dropouts due to a lack of efficacy.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
11.1 Comparison 1 All antidepressants versus placebo, Outcome 11 Dropouts due to a lack of efficacy	29	11007	Risk Ratio (IV, Random, 95% CI)	0.41 [0.33, 0.50]
11.2 Comparison 2 Each class of antidepressant versus placebo, Outcome 11 Dropouts due to a lack of efficacy	29		Risk Ratio (IV, Random, 95% CI)	Subtotals only
11.2.1 SSRIs	14	4832	Risk Ratio (IV, Random, 95% CI)	0.55 [0.38, 0.79]
11.2.2 SNRIs	13	4775	Risk Ratio (IV, Random, 95% CI)	0.33 [0.25, 0.43]
11.2.3 Other	5	1826	Risk Ratio (IV, Random, 95% CI)	0.52 [0.28, 0.95]
11.3 Comparison 3 Agomelatine versus placebo, Outcome 11 Dropouts due to a lack of efficacy	3	803	Risk Ratio (IV, Random, 95% CI)	0.45 [0.21, 0.96]
11.4 Comparison 4 Duloxetine versus placebo, Outcome 11 Dropouts due to a lack of efficacy	6	2098	Risk Ratio (IV, Random, 95% CI)	0.24 [0.15, 0.38]
11.5 Comparison 5 Escitalopram versus placebo, Outcome 11 Dropouts due to a lack of efficacy	7	2174	Risk Ratio (IV, Random, 95% CI)	0.48 [0.27, 0.86]
11.6 Comparison 6 Paroxetine versus placebo, Outcome 11 Dropouts due to a lack of efficacy	6	2098	Risk Ratio (IV, Random, 95% CI)	0.69 [0.40, 1.17]
11.7 Comparison 7 Sertraline versus placebo, Outcome 11 Dropouts due to a lack of efficacy	2	699	Risk Ratio (IV, Random, 95% CI)	0.57 [0.19, 1.70]
11.8 Comparison 8 Venlafaxine versus placebo, Outcome 11 Dropouts due to a lack of efficacy	9	3008	Risk Ratio (IV, Random, 95% CI)	0.39 [0.28, 0.53]
11.9 Comparison 9 Vilazodone versus placebo, Outcome 11 Dropouts due to a lack of efficacy	1	398	Risk Ratio (IV, Random, 95% CI)	1.98 [0.18, 21.66]
11.10 Comparison 10 Vortioxetine versus placebo, Outcome 11 Dropouts due to a lack of efficacy	1	625	Risk Ratio (IV, Random, 95% CI)	0.59 [0.17, 1.98]

Comparison 12. Dropouts due to adverse effects.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
12.1 Comparison 1 All antidepressants versus placebo, Outcome 12 Dropouts due to adverse effects	32	11793	Risk Ratio (IV, Random, 95% CI)	2.18 [1.81, 2.61]
12.2 Comparison 2 Each class of antidepressant versus placebo, Outcome 12 Dropouts due to adverse effects	32		Risk Ratio (IV, Random, 95% CI)	Subtotals only
12.2.1 SSRIs	16	5315	Risk Ratio (IV, Random, 95% CI)	1.98 [1.51, 2.61]
12.2.2 SNRIs	14	5078	Risk Ratio (IV, Random, 95% CI)	2.42 [1.81, 3.22]
12.2.3 Other	5	1826	Risk Ratio (IV, Random, 95% CI)	2.29 [1.31, 4.01]
12.3 Comparison 3 Agomelatine versus placebo, Outcome 12 Dropouts due to adverse effects	3	803	Risk Ratio (IV, Random, 95% CI)	1.14 [0.36, 3.56]
12.4 Comparison 4 Duloxetine versus placebo, Outcome 12 Dropouts due to adverse effects	7	2425	Risk Ratio (IV, Random, 95% CI)	3.13 [1.89, 5.20]
12.5 Comparison 5 Escitalopram versus placebo, Outcome 12 Dropouts due to adverse effects	8	2411	Risk Ratio (IV, Random, 95% CI)	2.02 [1.41, 2.90]
12.6 Comparison 6 Paroxetine versus placebo, Outcome 12 Dropouts due to adverse effects	7	2344	Risk Ratio (IV, Random, 95% CI)	2.40 [1.60, 3.61]
12.7 Comparison 7 Sertraline versus placebo, Outcome 12 Dropouts due to adverse effects	2	699	Risk Ratio (IV, Random, 95% CI)	1.37 [0.39, 4.76]
12.8 Comparison 8 Venlafaxine versus placebo, Outcome 12 Dropouts due to adverse effects	10	3311	Risk Ratio (IV, Random, 95% CI)	2.05 [1.54, 2.72]
12.9 Comparison 9 Vilazodone versus placebo, Outcome 12 Dropouts due to adverse effects	1	398	Risk Ratio (IV, Random, 95% CI)	3.11 [1.36, 7.12]
12.10 Comparison 10 Vortioxetine versus placebo, Outcome 12 Dropouts due to adverse effects	1	625	Risk Ratio (IV, Random, 95% CI)	2.52 [0.90, 7.03]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Allgulander 2001.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): venlafaxine 44.8 (SE = NA) years; placebo 46.1 (SE = NA) years Sex (women): venlafaxine 61.3%; placebo 58% Location: Belgium, Finland, France, Sweden, UK Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups. 1. Venlafaxine (n = 411) Duration: up to 24 weeks Treatment protocol: fixed dosage (37.5 mg/day, 75 mg/day, or 150 mg/day) 2. Placebo (n = 130) Duration: 24 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: baseline, weeks 1, 2, 3, 4, 6, 8, 10, 12, 16, 20, 24, 25 Primary outcomes Rate of treatment response: defined as a reduction of ≥ 50% on the HAM‐A Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: not specified Funding source: Wyeth‐Ayerst Research Declarations of interest among primary researchers: DH and Es are employees of Wyeth‐Ayerst. CS is an employee of the Karolinska Institutet, Stockholm, and was an investigator for 1 study centre Author of this study, Christer Allgulander (christer.allgulander@neurotec.ki.se) contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "A randomization schedule in blocks of four was generated for packaging and labelling by the biostatistics section of Wyeth‐Ayerst Research."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quote: "ITT [intention‐to‐treat] population and LOCF [last observation carried forward] analysis were used. Reasons for withdrawal were reported as well as numbers. Quote: "Number of patients who discontinued did not differ significantly among the treatment groups, but there were relatively fewer discontinuations among patients receiving the higher doses of venlafaxine ER. Patients who received placebo were significantly more likely to discontinue because of lack of efficacy than patients receiving venlafaxine ER (P<0.001) … There was no differences between the treatment groups in the percentage of patients who discontinued and gave adverse events as either a primary or secondary reason." Comment: unclear if rates of discontinuations differed between groups among other reasons for discontinuations.
Selective reporting (reporting bias)	Low risk	Quote: "All prespecified outcomes were reported."
Other bias	High risk	Quote: "This study was sponsored by Wyeth‐Ayerst Research, of which D.H. and E.S. are employees."

Allgulander 2004.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): sertraline 40.3 (SE 11.1) years; placebo 42.4 (SE 11.5) years Sex (women): sertraline 59%; placebo 51% Location: Australia, Canada, Denmark, Norway, Sweden Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Sertraline (n = 188) Duration: 12 weeks Treatment protocol: flexible dosage (50–150 mg/day). Sertraline treatment was initiated at 25 mg/day for the first week, followed by 3 weeks of treatment with 50 mg/day. From week 5 to week 6, flexible dosing was permitted in the range of 50–100 mg/day, increasing from week 7 to week 12 to a range of 50–150 mg/day. 2. Placebo (n = 190) Duration: 12 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: weeks 1, 2, 4, 6, 8, 12 Primary outcomes Rate of treatment response: defined as a reduction of ≥ 50% on the HAM‐A Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Agitation/anxiety Average score/change in quality of life/satisfaction: measured as the mean change from baseline on the Quality of Life Enjoyment and Satisfaction Questionnaire Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: not specified Funding source: Pfizer Declarations of interest among primary researchers: not specified Author of this study, Christer Allgulander (christer.allgulander@neurotec.ki.se) contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quotes: "ITT [intention‐to‐treat] population and LOCF [last observation carried over] analysis were used. Reasons and numbers of dropouts in each group were reported and were balanced between groups." "A nonsignificantly higher proportion of patients assigned to sertraline than placebo completed all 12 weeks of study treatment."
Selective reporting (reporting bias)	Low risk	Quote: "All prespecified outcomes were reported."
Other bias	Unclear risk	Quote: "The clinical trial was sponsored by Pfizer, Inc. The authors thank Hana Kosar, B.S.N., and Michelle Hosten of Pfizer and the other principal investigators for their contribution to the generation and collection of data." Comment: declarations of interest and sponsor involvement in the study were not clearly outlined.

Baldwin 2006.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age (mean): escitalopram 41.2 (SE 12.3) years; paroxetine 41.7 (SE 12.0) years; placebo 41.8 (SE 11.6) years Sex (women): escitalopram 64.8%; paroxetine 60%; placebo 67% Location: not specified Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Escitalopram (n = 403) Duration: 12 weeks Treatment protocol: fixed dose (5 mg/day, 10 mg/day, or 20 mg/day orally) 2. Paroxetine (n = 139) Duration: 12 weeks Treatment protocol: fixed dose (20 mg/day orally) 3. Placebo (n = 139) Duration: 12 weeks Treatment protocol: fixed dose, orally
Outcomes	Time points of assessment: baseline and after 1, 2, 4, 6, 8, 10, 12, 13, 14 weeks Primary outcomes Acceptability Secondary outcomes Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Agitation/anxiety Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: not specified Funding source: H. Lundbeck Declarations of interest among primary researchers: the University of Southampton has received support from Lundbeck and GlaxoSmithKline. DSB has received consultancy honoraria from Lundbeck and GlaxoSmithKline. AKTH and EM are employees of H. Lundbeck A/S, who sponsored this study.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Patients who met selection criteria at the screening and baseline visits were assigned to 12 and baseline visits were assigned to 12 weeks of double‐blind treatment in a 1:1:1:1:1 ratio of 5 mg escitalopram to 10 mg escitalopram to 20 mg escitalopram to 20 mg paroxetine to placebo according to a computer‐generated randomisation list drawn up by H. Lundbeck A/S … At each study centre, sequentially enrolled patients were assigned the lowest randomisation number available in blocks of ten."
Allocation concealment (selection bias)	Low risk	Quote: "The details of the randomisation series were unknown to randomisation series were unknown to any of the investigators and were contained in a set of sealed opaque envelopes."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: study was double‐blind and (quotes) "All study personnel and participants were masked to treatment assignment for the duration of the study." … "Study medications were capsules for oral administration, of identical appearance, taste and smell." Study was double‐blind and "All study personnel and participants were masked to treatment assignment for the duration of the study." … "Study medications were capsules for oral administration, of identical appearance, taste and smell."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: study was double‐blind. Quote: "All study personnel and participants were masked to treatment assignment for the duration of the study … Study medications were capsules for oral administration, of identical appearance, taste and smell."
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: intention‐to‐treat population and last observation carried forward. All reasons for withdrawal were reported with numbers as well. Quote: "The proportion of patients that withdrew because of adverse events was relatively low (<11% in any treatment group and <7% overall). Compared with the placebo group, significantly more patients (chi‐square test, P<0.05) in the escitalopram 20 mg and paroxetine 20 mg groups withdrew because of adverse events. Withdrawal rates due to lack of efficacy in the escitalopram 5 and 20 mg, paroxetine 20 mg and placebo groups were comparable. Compared with placebo, significantly fewer patients in the escitalopram 10 mg group withdrew because of lack of efficacy." Comment: the number of dropouts was relatively low for each group compared to total number of people in each group so this may not have affected the outcomes very much.
Selective reporting (reporting bias)	High risk	Quote: "Results not fully reported for response rate measured as a reduction of at least 50% on the HAM‐A: only significant comparisons provided."
Other bias	High risk	Quote: "The University of Southampton has received support from Lundbeck and GlaxoSmithKline. D.S.B. has received consultancy honoraria from Lundbeck and GlaxoSmithKline. A.K.T.H. and E.M. are employees of H.Lundbeck A/S, who sponsored this study."

Bose 2008.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): escitalopram 38.2 (SE 11.5) years; venlafaxine 37.1 (SE 10.8) years; placebo 37.6 (SE 12.3) years Sex (women): escitalopram 64.6%; paroxetine 59.7%; placebo 62.5% Location: USA Comorbidities: protocol: people were excluded from the study if they met DSM‐IV criteria for primary diagnoses for any axis I disorder other than GAD Results: 14 (10.3%) placebo, 14 (11.0%) escitalopram, and 9 (7.0%) venlafaxine XR participants had ongoing secondary psychiatric disorders, the most prevalent of which were social phobia and depression
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Escitalopram (n = 131) Duration: 8 weeks Treatment protocol: flexible dosage (10–20 mg/day). Participants received 10 mg/day for the first week, after which the dosage could be increased to 20 mg/day if clinically indicated. 2. Venlafaxine (n = 133) Duration: 8 weeks Treatment protocol: flexible dosage (75–225 mg/day). Participants received 75 mg/day for the first week, after which the dosage could be increased at the discretion of the investigator to a maximum of 150 mg/day for week 2 and 225 mg/day for weeks 3–8. 3. Placebo (n = 140) Duration: 8 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: baseline and after 1, 2, 4, 6, 8 weeks Primary outcomes Rate of treatment response: defined as a reduction of ≥ 50% on the HAM‐A Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2003–2004 Funding source: not specified Declarations of interest among primary researchers: not specified Author of this study, Anjana Bose (anjana.bose@frx.com) contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Incomplete outcome data (attrition bias) All outcomes	High risk	Quotes: "ITT [intention‐to‐treat] population and LOCF [last observation carried forward] analysis were used. Reasons and numbers of drops outs for each group were reported but there was imbalance between groups." "There were no differences between groups in specific reasons for premature discontinuation, with the following exceptions: more venlafaxine XR‐treated patients withdrew due to AEs than placebo‐treated patients (13.2 versus 5.1%, P =.031); more placebo‐treated patients withdrew due to insufficient therapeutic response than venlafaxine XR‐treated patients (4.4 versus 0%, P = 030)."
Selective reporting (reporting bias)	Low risk	Quote: "All prespecified outcomes were reported."
Other bias	Unclear risk	Quote: "Did not specify funding source."

Brawman‐Mintzer 2006.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): sertraline 40.1 (SE 13.2) years; placebo 40.8 (SE 12.3) years Sex (women): sertraline 59.8%; placebo 56.8% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Sertraline (n = 168) Duration: 10 weeks Treatment protocol: flexible dosage (50–200 mg/day). Participants who entered the double‐blind phase received sertraline 25 mg/day for 1 week followed thereafter by a flexible daily dosing of sertraline at 50–200 mg/day (once daily) based on clinical response and tolerability. Dose titration included medication increases at 2, 3, 4, and 7 weeks by 50‐mg increments up to a maximum of 200 mg/day at the investigator's discretion. Dosage reduction was permitted at any time during the study, and only 1 subsequent increase was allowed thereafter. 2. Placebo (n = 170) Duration: 10 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: baseline, and at the end of study weeks 1, 2, 3, 4, 6, 8, 10 Primary outcomes Rate of treatment response: defined as a reduction of ≥ 50% on the HAM‐A Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2000–2002 Funding source: Pfizer Declarations of interest among primary researchers: Dr Brawman‐Mintzer has served as a consultant to Pfizer, UCB Pharma, Cephalon, Janssen, and AstraZeneca; has received grant/research support from Pfizer, Forest, AstraZeneca, Janssen, Biovail, UCB Pharma, and Eli Lilly; and has served on the speakers bureau of AstraZeneca, Forest, Pfizer, Eli Lilly, and Janssen. Dr Rynn has served as a consultant to Eli Lilly, Pfizer, Wyeth, and AstraZeneca and has received grant/research support from Eli Lilly, Organon, AstraZeneca, Pfizer, and the National Institutes of Health. Dr Rickels has received honoraria from and served as a consultant to or on the advisory boards of Cephalon, DOV, Eli Lilly, MediciNova, Merck, Novartis, Pfizer, Pharmacia, Pherin, Predix, Sanofi‐Synthelabo, and Wyeth and has received research grants from AstraZeneca, Bristol‐Myers Squibb, Cephalon, Fabre‐Kramer, Merck, National Institute of Mental Health, Pfizer, Predix, Somerset, and Wyeth. Dr Knapp and Dr Carter report no other significant commercial relationships relevant to the study. Author of this study, Olga Brawman‐Mintzer (mintzero@musc.edu) contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "A computerized randomization list, stratified by site, was generated by the Coordinating Center at the Medical University of South Carolina. Equal allocation of the 2 treatment groups was used with a block size of either 4 or 6 to ensure that masking was not compromised. The study drug was packaged and labeled accordingly and delivered to the sites by a central pharmacy. The list of Participant Study Numbers and corresponding Study Drug Code Numbers (randomization numbers) was maintained at each site. Subjects were assigned the next available unique Study Drug Code Number in consecutive order."
Allocation concealment (selection bias)	Low risk	Quote: "Equal allocation of the 2 treatment groups was used with a block size of either 4 or 6 to ensure that masking was not compromised. The study drug was packaged and labeled accordingly and delivered to the sites by a central pharmacy. The list of Participant Study Numbers and corresponding Study Drug Code Numbers (randomization numbers) was maintained at each site. Subjects were assigned the next available unique Study Drug Code Number in consecutive order."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Study was double‐blind and no reason to believe blinding was compromised."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quotes: "ITT [intention‐to‐treat] population and mixed models analysis were used. Reasons for dropouts were reported." "Study completers did not differ significantly from dropouts with respect to baseline demographic and clinical characteristics. There were no statistically significant differences in reasons for dropping out between the sertraline and placebo groups (χ2 = 4.3, df = 4, p = .4)." "LOCF [last observation carried forward] analysis was done as a secondary analysis "to evaluate sensitivity of results to choice of statistical method." Comment: these results were only reported as treatment vs. placebo effects.
Selective reporting (reporting bias)	Low risk	Quote: "All prespecified outcomes were reported."
Other bias	High risk	Quote: "This study was conducted with the support of an unrestricted investigator‐initiated grant from Pfizer Inc … Dr. Brawman‐Mintzer has served as a consultant to Pfizer; has received grant/research support from Pfizer, Forest, AstraZeneca, Janssen, Biovail, UCB Pharma, and Eli Lilly; and has served on the speakers bureau of AstraZeneca, Forest, Pfizer, Eli Lille, and Janssen. Dr. Rynn has served as a consultant to Eli Lilly, Pfizer, Wyeth, and AstraZeneca, Pfizer, and the National Institutes of Health. Dr. Rickels has received honoraria from and served as a consultant to or on the advisory board of Cephalon, DOV, Eli Lilly, MediciNova, Merck, Novartis, Pfizer, Pharmacia, Pherin, Predic, Sanofi‐Synthelabo, and Wyeth and has received research grants from AstraZeneca, Bristol‐Myers Squibb, Cephalon, Fabre‐Kramer, Merck, National Institute of Mental Health, Pfizer, Predix, Somerset, and Wyeth. Drs Knapp and Carter report no other significant commercial relationships relevant to the study."

Davidson 1999.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): venlafaxine 37.5 (SE 10.5) years; buspirone 37.0 (SE 10.0) years; placebo 39.0 (SE 11.0) years Sex (women): venlafaxine 64.4%; buspirone 51%; placebo 62.2% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Venlafaxine (n = 174) Duration: 8 weeks Treatment protocol: fixed dosage (75 mg/day or 150 mg/day) 2. Buspirone (n = 93) Duration: 8 weeks Treatment protocol: divided dosage (30 mg/day). Buspirone was administered daily in 3 divided doses and was titrated according to the following schedule: 15 mg/day on study days 1 and 2, 20 mg/day on days 3 and 4, 25 mg/day on days 5–7, and 30 mg/day on days 8–56 3. Placebo (n = 98) Duration: 8 weeks Treatment protocol: matching placebo
Outcomes	Time points of assessment: baseline and at days 8, 15, 22, 29, 43, 57 during the double‐blind phase Primary outcomes Rate of treatment response: defined as a reduction of ≥ 50% on the HAM‐A Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Change in symptom levels: measured as the score at endpoint on the HAM‐A total score Dropouts due to adverse effects
Identification
Notes	Date of study: not specified Funding source: Wyeth‐Ayerst Research Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quotes: "Study was double‐blind." "Study medications were provided in blister packs, with double‐dummy techniques for blinding purposes. Venlafaxine XR and matching placebo were supplied in identical peach‐colored capsules and buspirone and matching placebo in identical gray capsules."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Incomplete outcome data (attrition bias) All outcomes	High risk	Quote: "ITT [intention‐to‐treat] population and LOCF [last observation carried forward] analysis were used. No information on whether attrition rates differed between groups and no information on reasons for dropouts in each group."
Selective reporting (reporting bias)	High risk	Quotes: "Did not fully report adverse events." "Venlafaxine XR and buspirone were both well tolerated, and adverse events were consistent with those reported in the literature for these agents."
Other bias	Unclear risk	Quote: "Supported by Wyeth‐Ayerst Research." Comment: declarations of interest and involvement in the study were not clearly reported.

Davidson 2004.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): escitalopram 39.5 (SE 12.1) years; placebo 39.5 (SE 13.1) years Sex (women): escitalopram 52.5%; placebo 52.9% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Escitalopram (n = 158) Duration: 8 weeks Treatment protocol: flexible dosage (10–20 mg/day). For the first 4 weeks, participants received escitalopram 10 mg/day. At week 4 or 6, if the investigator judged the therapeutic response to be insufficient, participants were dispensed double‐blind medication from a separate bottle containing either escitalopram 20 mg tablets and were told to continue taking 1 tablet per day. Participants could be returned to the starting dose thereafter if necessary for tolerability reasons. 2. Placebo (n = 157) Duration: 8 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: baseline and weeks 1, 2, 4, 6, 8 Primary outcome Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Agitation/anxiety Average score/change in quality of life/satisfaction: measured as the mean change from baseline on the Quality of Life Enjoyment and Satisfaction Questionnaire Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: not specified Funding source: Forest Laboratories Declarations of interest among primary researchers: not specified Author of this study, Jonathan Davidson (tolme@acpub.duke.edu) was contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: intention‐to‐treat population and last observation carried forward analysis used. No information on reasons or rates of dropouts in each group and whether discontinuation rates differed between groups. Except for adverse events. Quote: "the rate of discontinuation due to adverse events was not significantly different between the escitalopram and placebo groups."
Selective reporting (reporting bias)	Unclear risk	Quote: "All prespecified outcomes were reported. Adverse events not fully reported."
Other bias	Unclear risk	Quote: "Contract grant sponsor: Forest Laboratories, Inc." Comment: declarations of interest and involvement in the study were not clearly reported.

Feltner 2009.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): paroxetine 35.0 (SE 12.7) years; lorazepam 38.3 (SE 12.0) years; placebo 35.0 (SE 10.4) years Sex (women): paroxetine 55.4%; lorazepam 64.3%; placebo 54.4% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Paroxetine (n = 56) Duration: 4 weeks Treatment protocol: fixed dosage (20 mg/day orally) 2. Lorazepam (n = 56) Duration: 4 weeks Treatment protocol: fixed dosage (4.5 mg/day in 3 doses orally). Lorazepam was titrated (1 mg in 3 doses for the first 3 days) for tolerability 3. Placebo (n = 57) Duration: 4 weeks Treatment protocol: fixed dosage, orally
Outcomes	Time points of assessment: baseline and weeks 1, 2, 4 Primary outcomes Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐C score Change in symptom levels: measured as the mean change from baseline on the Quality of Life Enjoyment and Satisfaction Questionnaire Average score/change in quality of life/satisfaction: measured using the mean change from baseline on the Quality of Life Employment Satisfaction Questionnaire
Identification
Notes	Date of study: not specified Funding source: Pfizer Declarations of interest among primary researchers: Dr Feltner, Dr Capelleri, Dr Morlock, Ms Harness, and Ms Brock were Pfizer employees at the time of this work. Dr Sambunaris has received research funding from Pfizer. Author of this study, Douglas Feltner (douglas.feltner@pfizer.com) was contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Study was double‐blind and no reason to believe it was compromised."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Incomplete outcome data (attrition bias) All outcomes	High risk	Quote: "LOCF population and ITT analysis were used. Dropout rates were not clearly reported."
Selective reporting (reporting bias)	Unclear risk	Quote: "A lot of missing data; failed to report certain data that should have been reported such as dropouts due to inefficacy, dropouts due to side effects, total number of patients reporting side effects."
Other bias	High risk	Quote: "This study was funded by Pfizer, Inc. Dr. Feltner, Dr. Capelleri, Dr. Morlock, Ms. Harness, and Ms. Brock were Pfizer employees at the time this work was done. Dr. Sambunaris has received research funding from Pfizer."

Gelenberg 2000.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): venlafaxine 41.0 (SE 12.0) years; placebo 38.0 (SE 11.0) years Sex (women): venlafaxine 59%; placebo 59% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Venlafaxine (n = 124) Duration: 28 weeks Treatment protocol: flexible dosage (75–225 mg/day orally). For the first week of treatment, participants received venlafaxine XR 75 mg/day. Beginning on day 8 for participants with suboptimal responses, investigators were allowed to increase the dose of venlafaxine XR to 150 mg/day, and beginning on day 15, to 225 mg/day. No additional dosage increases beyond 225 mg/day were allowed. Dosage reduction to improve tolerability for the study drug was allowed to a minimum dosage of 75 mg/day. 2. Placebo (n = 127) Duration: 28 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: baseline and weeks 1, 2, 3, 4, 6, 8, 12, 16, 20, 24, 28 Primary outcomes Acceptability Secondary outcomes Rate of treatment response: defined as a reduction of ≥ 40% on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 1996–1997 Funding source: Wyeth‐Ayerst Research Declarations of interest among primary researchers: Dr Gelenberg has received grants or research support from Bristol‐Myers Squibb, Organon, Pfizer, Lilly Research Laboratories, Janssen, Merck Sharp & Dohme, SmithKline Beecham, Wyeth Ayerst, Hoechst Marion Roussel, and Forest Laboratories; is on the speakers' bureaus of Abbott, Bristol Myers Squibb, Pfizer, and SmithKline Beecham; is a stockholder or has other ownership interest in Pfizer, Warner Lambert, and Eli Lilly; and is a consultant for Merck‐Medco, Eli Lilly, Scios Inc, Glaxo Wellcome, and Parke‐Davis. Dr Lydiard has received grants or research support from Bristol‐Myers Squibb, Pfizer, SmithKline Beecham, Wyeth‐Ayerst, Forest Laboratories, Glaxo Wellcome, Parke‐Davis, Eli Lilly, Interneuron, Roche, Solvay, Organon, and UpjohnPharmacia and is a consultant for Bristol‐Myers Squibb, Pfizer, Eli Lilly, SmithKline Beecham, Wyeth‐Ayerst, Forest Laboratories, Glaxo Wellcome, Parke‐Davis, Roche, Novartis, and Organon. He has made presentations sponsored by Bristol‐Myers Squibb, Eli Lilly, SmithKline Beecham, Wyeth‐Ayerst, Forest Laboratories, Glaxo Wellcome, Parke‐Davis, Upjohn‐Pharmacia, Ravizza Farmaceutici, and Hoechst and has been a consultant for Bristol‐Myers Squibb, Pfizer, Eli Lilly, SmithKline Beecham, Wyeth‐Ayerst, Forest Laboratories, Glaxo Wellcome, Parke‐Davis, Roche, Dupont, Organon, and Zeneca. Drs Rudolph, Aguiar, Haskins, and Salinas are employees of Wyeth‐Ayerst Research, a division of American Home Products Corporation (AHP). They own stock in AHP, which manufactures and sells venlafaxine XR Author of this study, Alan Gelenberg (alang@u.arizona.edu) was contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "In this study, a double‐blind, parallel group design randomized patients in blocks of 4 using a table of random numbers."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: study says "double‐blind" and that "study medications were identical‐appearing capsules containing 75mg venlafaxine XR or placebo that were administered orally in the morning."
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Quote: "Not enough information on outcome assessors."
Incomplete outcome data (attrition bias) All outcomes	High risk	Quote: "LOCF [last observation carried forward] analysis was used. Study did not use ITT [intention‐to‐treat] analysis (used "evaluable‐patient basis"). Significant differences between the two groups in dropout rates. Quote: "Primary reasons for premature withdrawal included unsatisfactory response, with significantly less attrition in the venlafaxine XR group than in the placebo group (10 [8%] vs 28 [22%]; P = .002); adverse reaction, with the placebo group showing significantly less attrition than the venlafaxine XR group (18 [14%] vs 30 [24%]; P = .05); and failure to return (any reason) with significantly less attrition in the venlafaxine XR group (10 [8%] vs 22 [17%]; P = .04)." It is unclear if the people who dropped out differed from the people who remained in the study.
Selective reporting (reporting bias)	High risk	Comment: not all results were reported. Authors seemed to choose significant results to report without and did report full information. For example: no tables or values were reported for any secondary outcome measures, only vague statements such as "significantly reduced" or only P values are reported. Authors mentioned that they will perform both analyses on OC and LOCF but did not report any results for OC analysis, just LOCF and make vague statements such as "there was a high degree of concordance between the results obtained using LOCF analyses vs OC analyses." Authors reported differences between groups for certain dropout reasons but not for adverse events.
Other bias	High risk	Quote: "This study was supported by Wyeth‐Ayerst Research." Dr Gelenberg has received grants from Wyeth‐Ayerst. Dr Lydiard has received grants or research support from Wyeth‐Ayerst, is a consultant for Wyeth‐Ayerst, and has made presentations sponsored by Wyeth‐Ayerst. "Drs Rudolph, Aguiar, Haskins, and Salinas are employees of Wyeth‐Ayerst Research, a division of American Home Products Corporation (AHP). They own stock in AHP, which manufactures and sells venlafaxine XR."

Gommol 2015.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age (mean): vilazodone 40.5 (SE 13.2) years; placebo 40.1 (SE 13.0) years Sex (women): vilazodone 72.5%; placebo 66.2% Location: USA Comorbidities: protocol: people were excluded if they had a DSM‐IV‐TR‐based Axis I diagnosis other than generalised anxiety disorder within 6 months; secondary diagnoses of comorbid social anxiety disorder or specific phobias (or both) were allowed Results: mood disorders, which were the most frequently reported psychiatric comorbidities, were reported in 25% of participants overall and at similar rates in the 2 treatment groups.
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Vilazodone (n = 201) Duration: 8 weeks Treatment protocol: flexible dosage (20–40 mg/day with food). Participants received vilazodone 10 mg/day during week 1 and 20 mg/day during week 2. At the end of weeks 2 and 4, participants with inadequate response and no significant tolerability issues could have their dose increased to 40 mg/day; participants with adequate response continued taking 20 mg/day. 2. Placebo (n = 201) Duration: 8 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: baseline and weeks 1, 2, 4, 6, 8 Primary outcomes Rate of treatment response: defined as a reduction of at least 50% on the HAM‐A Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of patients reporting adverse effects Sleepiness/drowsiness Suicide wishes/gestures/attempts Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2013–2014 Funding source: Forest Laboratories Declarations of interest among primary researchers: Angelo Sambunaris, MD, has received consultant and speaking fees from Actavis Inc. and Takeda Pharmaceuticals. He has also received research support from Alkermes, Forest Pharmaceuticals, an affiliate of Actavis Inc., Otsuka, Pfizer, Sunovion, Shire, and Takeda. He was a principal investigator in both vilazodone phase III studies. Carl Gommoll, Xiongwen Tang, and Suresh Durgam acknowledge a potential conflict of interest as employees of Forest Research Institute, an affiliate of Actavis Inc.; Giovanna Forero, Maju Mathews, and Rene Nunez acknowledge a potential conflict of interest as employees of Forest Research Institute, an affiliate of Actavis Inc., at the time of the study.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Patients were randomized by computer‐generated numbers …"
Allocation concealment (selection bias)	Low risk	Quote: "The blind was maintained through a secured randomization code list and was broken only in case of emergency."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Investigators and patients were blinded to the allocation of study drug throughout treatment and down‐taper … Removing blind for any reason disqualified a patient from further participation."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Investigators and patients were blinded to the allocation of study drug throughout treatment and down‐taper. The blind was maintained through a secured randomization code list and was broken only in the case of emergency. Removing the blind for any reason disqualified a patient from further participation."
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quote: "ITT [intention‐to‐treat] population and MMRM [mixed model for repeated measures] analysis were used. LOCF [last observation carried forward] analysis was used for sensitivity analysis. Reasons for attrition/exclusions were reported."
Selective reporting (reporting bias)	Low risk	Quote: "All prespecified outcomes were reported."
Other bias	High risk	Quote: "Forest Laboratories, LLC was involved in the study design, collection (through contracted clinical investigator sites), analysis and interpretation of data, and the decision to present these results."

Goodman 2000.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): escitalopram 39.6 (SE 13.4) years; placebo 40.9 (SE 14.0) years Sex (women): escitalopram 59.5%; placebo 62.5% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 if 2 groups 1. Escitalopram (n = 129) Duration: 8 weeks Treatment protocol: flexible dosage (10–20 mg/day orally) 2. Placebo (n = 128) Duration: 8 weeks Treatment protocol: flexible dosage, orally
Outcomes	Time points of assessment: baseline and week 8 Primary outcomes Acceptability Secondary outcomes Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2000–2001 Funding source: not specified Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Quote: "Not enough information."
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Quote: "This study was a synopsis. It was double‐blind but not enough details are given to make a full assessment."
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quote: "LOCF [last observation carried forward] population and ITT [intention‐to‐treat] analysis were used. Reasons and rates of withdrawal were reported by group but unclear if the rates differ between groups."
Selective reporting (reporting bias)	Low risk	Quote: "All prespecified outcomes were reported."
Other bias	Unclear risk	Quote: "Did not specify funding source."

Goodman 2001.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): escitalopram 36.8 (SE 12.2) years; placebo 38.6 (SE 12.5) years Sex (women): escitalopram 61.4%; placebo 48.6% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Escitalopram (n = 149) Duration: 8 weeks Treatment protocol: flexible dosage (10–20 mg/day, orally) 2. Placebo (n = 145) Duration: 8 weeks Treatment protocol: flexible dosage, orally
Outcomes	Time points of assessment: baseline and week 8 Primary outcomes Acceptability Secondary outcomes Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2001–2002 Funding source: not specified Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Quote: "Not enough information."
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Quote: "This study was a synopsis. It was double‐blind but not enough details are given to make a full assessment."
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quote: "ITT [intention‐to‐treat] population and LOCF [last observation carried forward] analysis were used. Numbers and rates of dropouts in each group reported but not clear if these rates differed between the groups."
Selective reporting (reporting bias)	Low risk	Quote: "All prespecified outcomes were reported."
Other bias	Unclear risk	Quote: "Did not specify funding source."

Hackett 2003.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): venlafaxine 44.5 (SE NA) years; diazepam 44 (SE NA) years; placebo 43 (SE NA) years Sex (women): venlafaxine 66.5%; diazepam 64%; placebo 64% Location: not specified Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Venlafaxine (n = 370) Duration: 8 weeks Treatment protocol: fixed dosage (75 mg/day or 150 mg/day) 2. Diazepam (n = 89) Duration: 8 weeks Treatment protocol: fixed dosage (15 mg/day) 3. Placebo (n = 97) Duration: 8 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: baseline and days 7, 14, 21, 28, 42, 56 Primary outcomes Rate of treatment response: defined as a reduction of at least 50% on the HAM‐A Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Change in symptom levels: measured as the score at endpoint on the HAM‐A total score Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: not specified Funding source: not specified Declarations of interest among primary researchers: not specified Author of this study, David Hackett (hackettd@war.wyeth.com) was contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: intention‐to‐treat population and last observation carried forward analysis were used. Reasons for dropouts and numbers in each group were reported. Quote: "A similar proportion of patients in each of the four treatment groups discontinued from the study, the primary reasons being adverse reaction, unsatisfactory response and patient request (Table 2). Adverse reactions led to the withdrawal of a significantly higher percentage of venlafaxine XR‐treated patients than placebo or diazepam‐treated patients (P < 0.05)." Comment: due to high dropouts in venlafaxine 75 mg/day group and venlafaxine 150 mg/day group, there could have been bias in the results.
Selective reporting (reporting bias)	High risk	Comment: did not fully report adverse events. Quote: "The most frequently reported treatment‐emergent adverse events (TEAEs) were nausea, headache, asthenia, somnolence, dry mouth and dizziness. Nausea, reported most commonly in the venlafaxine XR groups, was mild to moderate in severity and tended to occur early in the course of treatment, subsiding with continued therapy. Conversely, asthenia, the most common adverse event in the diazepam group, persisted throughout the course of the study."
Other bias	Unclear risk	Quote: "Did not specify funding source."

Hartford 2007.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): venlafaxine 40.1 (SE 13.2) years; duloxetine 40.4 (SE 13.6) years; placebo 41.9 (SE 14.2) years Sex (women): venlafaxine 62.2%; duloxetine 64.2%; placebo 61.5% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Venlafaxine (n = 164) Duration: 10 weeks Treatment protocol: flexible dosage (75–225 mg/day). The venlafaxine XR group began treatment with 37.5 mg/day for the first week and then was increased to 75 mg/day for the second week. Participants could then have their dose increased to 150 mg/day (for at least 1 week) and then to 225 mg/day. Dose increases were allowed on the basis of investigator judgement to maximise efficacy irrespective of the CGI‐I scale score; however, the dosage had to be increased if the CGI‐I scale score was ≥ 3 or (minimal improvement, no change, or worse) at visit 4, visit 5, or visit 6 unless the participant was unable to tolerate an increased dosage. A total of 2 downward dose adjustments for tolerability reasons were allowed as long as the dose did not fall below venlafaxine XR 75 mg/day. 2. Duloxetine (n = 162) Duration: 10 weeks Treatment protocol: flexible dosage (60–120 mg/day). The duloxetine group started at 30 mg/day for 1 week followed by an increase to 60 mg/day. After titration to 60 mg/day, flexible dosing was allowed in weekly increments of 30 mg/day up to a maximum dose of 120 mg/day. A total of 2 downward dose adjustments for tolerability reasons were allowed as long as the dose did not fall below duloxetine 60 mg/day. 3. Placebo (n = 161) Duration: 10 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: baseline and weeks 1, 2, 4, 7, 10 Primary outcomes Rate of treatment response: defined as a reduction of at least 50% on the HAM‐A Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Remission rate: defined as a score of ≤ 7 on the Hamilton Anxiety Rating Scale Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: not specified Funding source: Eli Lilly and Company and Boehringer Ingelheim Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: study used block randomisation. Quote: "Patients were randomly assigned in a 1:1:1 ratio."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: study stated double‐blind but no other information. No reason to believe that the blinding was broken.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: double‐blind. No reason to believe blinding was compromised.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: intention‐to‐treat population and last observation carried forward analysis used. Reasons for discontinuation were reported. Quote: "Treatment groups did not differ in their reason for discontinuation with exception of adverse events … these events, however, occurred in less than 2% of the entire study sample, and there was no statistically significant difference between the active treatment groups and the placebo group with respect to any specific adverse event leading to discontinuation."
Selective reporting (reporting bias)	Low risk	Quote: "All prespecified outcomes were reported."
Other bias	Unclear risk	Quote: "This work was sponsored by Eli Lilly and Company and Boehringer Ingelheim." Comment: declarations of interest and involvement in the study were not clearly reported.

Kasper 2009.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age (mean): venlafaxine 42.6 (SE 11.8) years; pregabalin 39.5 (SE 11.9) years; placebo 40.2 (SE 12.1) years Sex (women): venlafaxine 58%; pregabalin 64%; placebo 61% Location: Belgium, Canada, France, Ireland, Italy, Netherlands, Spain, Sweden Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Venlafaxine (n = 125) Duration: 8 weeks Treatment protocol: flexible dosage (75–225 mg/day). Venlafaxine treatment started at 75 mg/day for first week. Thereafter, dosing was flexible at 75–225 mg/day, administered in the morning. 2. Pregabalin (n = 121) Duration: 8 weeks Treatment protocol: flexible dosage (150–600 mg/day). Pregabalin treatment started at 150 mg twice daily for first week; thereafter, dosing was flexible based on clinical response and tolerability at 300–600 mg/day, administered twice daily. 3. Placebo (n = 128) Duration: 8 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: baseline, week 1, 2, 3, 4, 6, 8 Primary outcomes Rate of treatment response: defined as a reduction of at least 50% on the HAM‐A Acceptability Secondary outcomes Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Sleepiness/drowsiness Average score/change in quality of life/satisfaction: measured as the mean change from baseline on the Quality of Life Enjoyment and Satisfaction Questionnaire Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: not specified Funding source: Pfizer Declarations of interest among primary researchers: Dr Kasper has received grants/research support, consulting fees, and honoraria within the last 3 years from AstraZeneca, Bristol‐Myers Squibb, CSC, Eli Lilly, GlaxoSmithKline, Janssen Pharmaceutical, Lundbeck, MSD, Novartis, Organon, Pierre Fabre, Pfizer, Schwabe, Sepracor, Servier, Wyeth. Dr Herman is a full‐time employee of Pfizer Inc. Dr Nivoli has no disclosures to declare. Dr Van Ameringen has received grant/research support form AstraZeneca, Cephalon, GlaxoSmithKline, Janssen‐Ortho Inc., National Institute of Health, Novartis, Pfizer, and Wyeth‐Ayerst; has served as a consultant for Biovail, Cephalon, GlaxoSmithKline, Janssen‐Ortho Inc., Novartis, Pfizer, and Wyeth‐Ayerst; and has served on the speakers bureaus of GlaxoSmithKline, Janssen‐Ortho Inc., Pfizer, and Wyeth‐Ayerst. Dr Petralia has no disclosures to declare Dr Mandel is a full‐time employee of Pfizer Inc, Dr Baldinetti was a full‐time employee of Pfizer Inc. at the time this study was conducted, and at the time the initial draft manuscript was prepared. She is currently affiliated with IRCCS S. Lucia Foundation‐Neurology Department, University of Rome, Rome, Italy. Dr Bandelow has received consulting fees and honoraria within the last 3 years from AstraZeneca, Bristol‐Myers‐Squibb, Cephalon, Dainippon‐Sumitomo, Glaxo, Janssen, Jazz, Lilly, Lundbeck, Pfizer, Roche, Servier, Solvay, and Wyeth. Author of this study, Siegfried Kasper (sekretariat.bandelow@med.uni‐goettingen.de) was contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: study stated it was double‐blind but no other information provided. Some assessments were done by telephone. Results did not suggest blinding was compromised.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: double‐blind. No reason to believe blinding was compromised.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: intention‐to‐treat population and last observation carried forward analysis were used. Unclear if there was difference between groups in dropouts rates or if those who dropped out differed from those who remained in the study. Only information on dropouts was (quote) "rate of attrition because of adverse events was numerically higher on VXR [venlafaxine] compared with placebo (17.6 vs 5.5%) and again was intermediate (12.4% on PGB [placebo])."
Selective reporting (reporting bias)	Low risk	Quote: "All prespecified outcomes were reported."
Other bias	High risk	Quote: "The study was funded by Pfizer Inc. Paid editorial support was provided by Edward Schweizer, MD, and funded by Pfizer Inc. The authors acknowledge the contribution of the many individual investigators for their participation in this trial. Dr Kasper has received grants/research support, consulting fees, and honoraria within the last 3 years from AstraZeneca, Bristol‐Myers Squibb, CSC, Eli Lilly, GlaxoSmithKline, Janssen Pharmaceutical, Lundbeck, MSD, Novartis, Organon, Pierre Fabre, Pfizer, Schwabe, Sepracor, Servier, Wyeth. Dr Herman is a full‐time employee of Pfizer Inc. Dr Nivoli has no disclosures to declare. Dr Van Ameringen has received grant/research support form AstraZeneca, Cephalon, GlaxoSmithKline, Janssen‐Ortho Inc., National Institute of Health, Novartis, Pfizer, and Wyeth‐Ayerst; has served as a consultant for Biovail, Cephalon, GlaxoSmithKline, Janssen‐Ortho Inc., Novartis, Pfizer, and Wyeth‐Ayerst; and has served on the speakers bureaus of GlaxoSmithKline, Janssen‐Ortho Inc., Pfizer, and Wyeth‐Ayerst. Dr Petralia has no disclosures to declare Dr Mandel is a full‐time employee of Pfizer Inc. Dr Baldinetti was a full‐time employee of Pfizer Inc. at the time this study was conducted, and at the time the initial draft manuscript was prepared. She currently is affiliated with IRCCS S. Lucia Foundation‐Neurology Department, University of Rome, Rome, Italy. Dr Bandelow has received consulting fees and honoraria within the last 3 years from AstraZeneca, Bristol‐Myers‐Squibb, Cephalon, DainipponSumitomo, Glaxo, Janssen, Jazz, Lilly, Lundbeck, Pfizer, Roche, Servier, Solvay, and Wyeth."

Koponen 2007.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): duloxetine 43.6 (SE 12.7) years; placebo 44.1 (SE 13.4) years Sex (women): duloxetine 68.3%; placebo 66.9% Location: Finland, France, Germany, South Africa, Spain, Sweden, USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Duloxetine (n = 338) Duration: 9 weeks Treatment protocol: fixed dosage (60 mg/day or 120 mg/day). Participants in both duloxetine treatment groups were started with duloxetine 60 mg/day. If there were tolerability concerns, the dose could be lowered initially to 30 mg/day, but all participants were gradually increased to their randomly assigned dose over a 2‐week period. 2. Placebo (n = 175) Duration: 9 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: baseline, and weeks 1, 2, 4, 6, 9 Primary outcome Rate of treatment response: defined as a reduction of ≥ 50% on the HAM‐A Secondary outcomes Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2004–2005 Funding source: Eli Lilly and Company and Boehringer Ingelheim Declarations of interest among primary researchers: Dr Koponen has received honoraria from or participated in speakers bureaus for AstraZeneca, Bristol‐Myers Squibb, Eli Lilly, GlaxoSmithKline, H. Lundbeck, Janssen‐Cilag, and Organon. Dr Allgulander has served as a consultant or speaker for Eli Lilly, H. Lundbeck, Pfizer, and Wyeth. Drs Erickson, Detke, Ball, and Russell are employees or shareholders (or both) of Eli Lilly. Drs Dunayevich and Pritchett were previously employees and shareholders of Eli Lilly. Author of this study, Hannu Koponen (hannujuhani.koponen@uku.fi) was contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: stratified randomisation used. Quote: "Treatment randomization was stratified by HAM‐A total score at the randomization visit in order to ensure that severity did not differ between groups."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Study says it was double‐blind. No reason to think blinding was compromised."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quote: "Significant differences between groups in dropouts rates due to adverse events and lack of efficacy. ITT [intention‐to‐treat] population and LOCF [last observation carried forward] analyses were used but LOCF analysis was not reported."
Selective reporting (reporting bias)	High risk	Quote: "Authors mention that both MMRM [mixed model for repeated measures] and LOCF [last observation carried forward] analysis but it seems they only reported MMRM results because Table 2 is MMRM and no LOCF analysis were reported. Otherwise all prespecified outcomes were reported."
Other bias	High risk	Quote: "The research was supported by Eli Lilly and Co. and Boehringer Ingelheim. Dr. Koponen has received honoraria from or participated in speakers bureaus for AstraZeneca, Bristol‐Myers Squibb, Eli Lilly, GlaxoSmithKline, H. Lundbeck, Janssen‐Cilag, and Organon. Dr. Allgulander has served as a consultant or speaker for Eli Lilly, H. Lundbeck, Pfizer, and Wyeth. Drs. Erikson, Detke, Ball, and Russell are employees and/or shareholders of Eli Lilly. Drs. Dunayevick and Pritchett were previously employees and shareholders of Eli Lilly."

Lenox‐Smith 2003.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): venlafaxine 48 (SE NA) years; placebo 46 (SE NA) years Sex (women): duloxetine 61.5%; placebo 56.6% Location: UK Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Venlafaxine (n = 122) Duration: 24 weeks Treatment protocol: flexible dosage (75–150 mg/day). Initially venlafaxine XL 75 mg/day, although this could be increased any time after 2 weeks to 150 mg/day, and then reduced again if there were any tolerability issues. 2. Placebo (n = 122) Duration: 24 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: baseline, and weeks 2, 4, 8, 16, 24 Primary outcomes Acceptability Secondary outcomes Rate of treatment response: defined as a reduction of ≥ 50% on the HAM‐A Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Agitation/anxiety
Identification
Notes	Date of study: not specified Funding source: Wyeth Ayerst Declarations of interest among primary researchers: not specified Author of this study, Alan Lenox‐Smith (lenoxa@wyeth.com) was contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Randomisation was by randomly permuted blocks that had been generated centrally."
Allocation concealment (selection bias)	Low risk	Quote: "Randomisation was by randomly permuted blocks that had been generated centrally."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Study was double‐blind with matching placebos and active medications."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "The study was double‐blind with matching placebos and active medication."
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quote: "ITT [intention‐to‐treat] population and LOCF [last observation carried forward] analysis were used. No information or analysis on dropouts nor were number of dropouts in each group reported."
Selective reporting (reporting bias)	High risk	Quote: "Efficacy based on CGI scores were not reported. Rates for dropouts and reasons were not reported."
Other bias	Unclear risk	Quote: "This study was fully funded by Wyeth." Comment: Drs Lenox‐Smith and Reynolds are senior medical advisors and scientific affairs managers for Wyeth Pharmaceuticals. Declarations of interest and involvement in the study were not clearly reported.

Mahableshwarkar 2014.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age (mean): duloxetine 39.5 (SE 12.3) years; vortioxetine 38.9 (SE 12.1) years; placebo 36.8 (SE 12.1) years Sex (women): duloxetine 72.4%; vortioxetine 67.1%; placebo 65.0% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Duloxetine (n = 156) Duration: 8 weeks Treatment protocol: fixed dosage (60 mg/day) 2. Vortioxetine (n = 468) Duration: 8 weeks Treatment protocol: fixed dosage (2.5 mg/day, 5 mg/day, or 10 mg/day) 3. Placebo (n = 157) Duration: 8 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: baseline, and weeks 1, 2, 4, 6, 8 Primary outcomes Rate of treatment response: defined as a reduction of ≥ 50% on the HAM‐A Acceptability Secondary outcomes Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2008–2009 Funding source: Takeda Pharmaceutical Company and H. Lundbeck Declarations of interest among primary researchers: Dr Atul R Mahableshwarkar, Dr Yinzhong Chen and Ms Paula L Jacobsen are employees of Takeda Development Center Americas; Dr Jeffrey S Simon has received research grant/research support from Takeda Pharmaceutical Company, Ltd Author of this study, Atul Mahableshwarkar (atul@tgrd.com), was contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Eligible patients were randomised (1:1:1:1) to receive placebo, vortioxetine 2.5mg, vortioxetine 5mg, vortioxetine 10mg or duloxetine 60mg once daily during the 8 week double‐blind treatment period using an interactive voice response system (IVRS)."
Allocation concealment (selection bias)	Low risk	Quote: "Vortioxetine tablets (2.5, 5 or 10 mg) and duloxetine 60 mg (for treatment) or 30 mg (for tapering dose) capsules were enclosed in brownish‐orange capsules; identical capsules containing lactose monohydrate/magnesium stearate filler were used for placebo. Patients were dispensed their double‐blind study drug supplies (using an interactive voice response system) …"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Study was double‐blind and "capsules were enclosed in brownish‐orange capsules; identical capsules containing lactose monohydrate/magnesium stearate filler were used for placebo. Patients were dispended their double‐blind study drug supplied (using IVRS) …"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: study was double‐blind. Quote: "capsules were enclosed in brownish‐orange capsules; identical capsules containing lactose monohydrate/magnesium stearate filler were used for placebo. Patients were dispended their double‐blind study drug supplied (using Interactive voice response system) …"
Incomplete outcome data (attrition bias) All outcomes	High risk	Quote: "MMRM [mixed model for repeated measures] analysis was used for primary efficacy endpoint. Secondary efficacy analysis of primary endpoint (sensitivity analysis) used LOCF and Observed Cases. Reasons and number of dropouts in each group were reported but they do no report whether there were significant differences between the groups in terms of dropout rates."
Selective reporting (reporting bias)	High risk	Comment: specified response rate should be measured but there was no mention of it in the results; remission was not fully reported. Quotes: "There was no separation from placebo for any of the vortioxetine treatment groups on the HAM‐A remission rate (logistic regression on LOCF [last observation carried forward])." Comment: suicide was not fully reported. Quote: "There were no differences between treatment groups in any of the C‐SSRS [Columbia‐Suicide Severity Rating Scale] items during the study. The incidence of suicidal ideation was low and similar for all treatment groups; none of the patients reported suicide attempts."
Other bias	High risk	Quote: "Drs. Atul R. Mahableshwarkar, Yinzhong Chen and Ms. Paula L. Jacobsen are employees of Takeda Development Center Americas. At the time of this study, Michael Serenko was an employee of Takeda Development Center Americas. This study was supported by the Takeda Pharmaceutical Company, Ltd and H. Lundbeck A/S. Assistance with writing and manuscript preparation was provided by Ann C. Sherwood, PhD and Philip Sjostedt, BPharm of The Medicine Group and funded by the Takeda Pharmaceutical Company, Ltd. and H. Lundbeck A/S."

McLeod 1992.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐III‐R Age (mean): imipramine 41.8 (SE 8.2) years; alprazolam 41.4 (SE 9.8) years; placebo 40.3 (SE 7.9) years Sex (women): imipramine 64%; alprazolam 64%; placebo 64% Location: not specified Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 11 Imipramine (n = 14) Duration: 6 weeks Treatment protocol: flexible dosage (25–200 mg/day). Participants took 1 capsule 3 times per day unless they developed excessive adverse effects, in which case the dose was decreased. Across the weeks of treatment, the dose was adjusted according to clinical response and ranged from a minimum of 1 capsule to a maximum of 12 capsules per day. 2. Alprazolam (n = 14) Duration: 6 weeks Treatment protocol: flexible dosage (0.5–5.5 mg/day). Participants took 1 capsule 3 times per day unless they developed excessive adverse effects, in which case the dose was decreased. Across the weeks of treatment, the dose was adjusted according to clinical response and ranged from a minimum of 1 capsule to a maximum of 12 capsules per day. 3. Placebo (n = 14) Duration: 6 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: baseline, and week 6 Secondary outcome Change in symptom levels: measured as score at endpoint on the HAM‐A total score
Identification
Notes	Date of study: not specified Funding source: supported in part by NIMH grants MH35435 and MH42579, and by a grant from the Upjohn Company, Kalamazoo, Michigan, USA Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: study was double‐blind. Quote: "Medications were prepared in identical capsules that contained either placebo, 0.5 mg alprazolam, or 25 mg imipramine."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: study was double‐blind. Quote: "Medications were prepared in identical capsules that contained either placebo, 0.5 mg alprazolam, or 25 mg imipramine."
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: no information on dropouts and no information on whether intention‐to‐treat population or last observation carried forward analysis were used.
Selective reporting (reporting bias)	Low risk	Quote: "All prespecified outcomes were reported."
Other bias	Low risk	Quote: "This study was supported in part by NIMH grants MH35435 and MH42579, and by a grant from the Upjohn Company, Kalamazoo, Michigan."

Montgomery 2006.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): venlafaxine 46.0 (SE 12.0) years; pregabalin 43.4 (SE 12.1) years; placebo 43.0 (SE 12.0) years Sex (women): venlafaxine 65%; pregabalin 62.1%; placebo 58% Location: Austria, Belgium, Germany, Netherlands, UK Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Venlafaxine (n = 113) Duration: 6 weeks Treatment protocol: fixed dosage (75 mg/day twice daily). Participants assigned to venlafaxine began treatment at the full 37.5 mg twice daily dosage. 2. Pregabalin (n = 207) Duration: 6 weeks Treatment protocol: fixed dosage (400 mg/day or 600 mg/day). Participants assigned to pregabalin 400 mg/day received 100 mg/day for 2 days, then 200 mg/day for 2 days, before receiving the full dosage of 400 mg/day on day 5. Participants assigned to pregabalin 600 mg/day received 150 mg/day for 2 days, 300 mg/day for 2 days, and 450 mg/day for 2 days, before receiving the full dosage of 600 mg/day after their day 7 visit. 3. Placebo (n = 101) Duration: 6 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: baseline, and weeks 1, 2, 3, 4, 6 Primary outcomes Rate of treatment response: defined as a reduction of ≥ 50% on the HAM‐A Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 1999–2001 Funding source: Pfizer Declarations of interest among primary researchers: Dr Montgomery has been a consultant for, received honoraria from, and served on the speakers or advisory boards for Wyeth, Lundbeck, and GlaxoSmithKline. Dr Tobias is an employee of Pfizer. Dr Zornberg has been an employee of Pfizer. Dr Kasper has received grant/research support from Eli Lilly, Lundbeck, Bristol‐Myers Squibb, GlaxoSmithKline, Organon, and Servier; has been a consultant or served on the advisory boards for AstraZeneca, Bristol‐Myers Squibb, GlaxoSmithKline, Eli Lilly, Lundbeck, Pfizer, Organon, Janssen, and Novartis; and has served on the speakers bureau for AstraZeneca, Eli Lilly, Lundbeck, and Janssen. Dr Pande is an employee of and a major stock shareholder in Pfizer. Author of this study, Stuart Montgomery (stuart@samontgomery.co.uk) was contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: no information other than "double‐blind" and there are no reasons to think the blinding was compromised.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: study says double‐blind and there are no reasons to believe that blinding was compromised.
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: intention‐to‐treat population and last observation carried forward analysis were used. Observed cases was used for early onset of efficacy evaluation. Reasons for discontinuations were reported. Quote: "significantly more patients treated with pregabalin 40 mg/day completed the study than did patients with venlafaxine. There were, however, no notable differences in demographic or clinical variables between the group of patients who dropped out and those who completed the study … The attrition rate due to discontinuations associated with adverse events in the venlafaxine 75mg/day group was significantly greater than that in the pregabalin 400mg/day group." Comment: unclear whether dropouts due to other reasons differed between groups. Since there were more dropouts in the venlafaxine groups it is possible this affected the outcomes.
Selective reporting (reporting bias)	Low risk	Comment: all prespecified outcomes were reported.
Other bias	High risk	Quote: "This study was funded by Pfizer Inc, New York, N.Y. Dr. Montgomery has been a consultant for, received honoraria from, and served on the speakers or advisory boards for Wyeth, Lundbeck, and GlaxoSmithKline. Dr. Tobias is an employee of Pfizer. Dr. Zornberg has been an employee of Pfizer. Dr. Kasper has received grant/research support from Eli Lilly, Lundbeck, Bristol‐Myers Squibb, GlaxoSmithKline, Organon, and Servier; has been a consultant or served on the advisory boards for AstraZeneca, Bristol‐Myers Squibb, GlaxoSmithKline, Eli Lilly, Lundbeck, Pfizer, Organon, Janssen, and Novartis; and has served on the speakers bureau for AstraZeneca, Eli Lilly, Lundbeck, and Janssen. Dr. Pande is an employee of and a major stock shareholder in Pfizer."

NCT00135525.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): paroxetine 39.5 (SE 12.2) years; placebo 40.6 (SE 12.7) years Sex (women): paroxetine 60.5%; placebo 59.1% Location: Japan Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Paroxetine (n = 170) Duration: 8 weeks Treatment protocol: fixed dosage (20 mg/day orally). Participants began treatment with 10 mg/day for 1 week followed by forced titration to 20 mg/day for 7 weeks. 2. Placebo (n = 170) Duration: 8 weeks Treatment protocol: fixed dosage, orally
Outcomes	Time points of assessment: baseline and weeks 1, 2, 4, 6, 8 Primary outcomes Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Change in symptom levels: measured as the mean change from baseline on the Hamilton Anxiety Rating Scale total score Total number of participants reporting adverse effects Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2003–2005 Funding source: not specified Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Double‐blind. No reason to believe blinding was compromised."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: double‐blind. No reason to believe blinding was compromised.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: study used a full analysis set (FAS‐8W) and LOCF analysis. Not a full list of reasons for dropouts was reported and unclear if the rates between groups differed and why people dropped out exactly.
Selective reporting (reporting bias)	Low risk	Comment: all prespecified outcomes were reported.
Other bias	Unclear risk	Comment: did not specify funding source.

NCT00252343.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age (mean): total 40.3 (SE 13.9) years Sex (women): total 61.9% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Escitalopram (n = 122) Duration: 8 weeks Treatment protocol: fixed dosage (10 mg/day once daily) 2. Amibegron (n = 118) Duration: 8 weeks Treatment protocol: fixed dosage (700 mg/day in 2 doses) 3. Placebo (n = 119) Duration: 8 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: baseline, day 56 Secondary outcomes Total number of participants reporting adverse effects Dropouts due to adverse effects
Identification
Notes	Date of study: 2005–2007 Funding source: Sanofi‐Aventis Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Comment: this study was a synopsis. It was double‐blind but not enough details are given to make a full assessment.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Comment: this study was a synopsis. It was double‐blind but not enough details are given to make a full assessment.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: this study was a synopsis. Not enough information was given to make a full assessment. Intention to treat and mixed model for repeated measures used for primary efficacy analysis. Mixed model for repeated measures and last observation carried forward used for secondary efficacy endpoints. Not enough information on dropouts in each groups and number of dropouts in each group.
Selective reporting (reporting bias)	High risk	Judgement Comment: "Results were not fully reported for change in symptom levels. Quote: "The comparison between placebo and escitalopram, which was chosen as reference treatment, confirmed the validity of the study design and the conduct of the study."
Other bias	Unclear risk	Judgement Comment: "Study was sponsored by Aventis Sanofi but given this study was a synopsis a full assessment could not be made"

NCT00266747.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age (mean): total 40.8 (SE 12.3) years Sex (women): total 66.9% Location: NA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Paroxetine (n = 124) Duration: 8 weeks Treatment protocol: fixed dosage (20 mg/day) 2. Amibegron (n = 120) Duration: 8 weeks Treatment protocol: fixed dosage (700 mg/day) 3. Placebo (n = 122) Duration: 8 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: baseline, and day 56 Secondary outcomes Total number of participants reporting adverse events Dropouts due to adverse effects
Identification
Notes	Date of study: 2005–2007 Funding source: Sanofi‐Aventis Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Comment: this study was a synopsis. It was double‐blind but not enough details are given to make a full assessment.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Comment: this study was a synopsis. It was double‐blind but not enough details are given to make a full assessment.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: this study was a synopsis. Not enough information was given to make a full assessment. Intention to treat and mixed model for repeated measures analyses were done on the primary efficacy measure. Last observation carried forward analysis done on the secondary efficacy measures.
Selective reporting (reporting bias)	Unclear risk	Comment: this study was a synopsis and only reported adverse event data.
Other bias	Unclear risk	Comment: drug company sponsored the trial but given this study was a synopsis a full assessment could not be made.

NCT00417118.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age (mean): total 41.6 (SE NA) years Sex (women): total 66.0% Location: Belgium, Canada, Finland, France, Italy, Sweden, Turkey Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Escitalopram (n = 113) Duration: 8 weeks Treatment protocol: fixed dosage (10 mg/day orally) 2. Saredutant (n = 124) Duration: 8 weeks Treatment protocol: fixed dosage (100 mg/day orally) 3. Placebo (n = 124) Duration: 8 weeks Treatment protocol: fixed dosage, orally
Outcomes	Time points of assessment: baseline, day 56 Primary outcomes Acceptability Secondary outcomes Total number of participants reporting adverse events Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2006–2008 Funding source: not specified Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Comment: this study was a synopsis. It was double‐blind but not enough details are given to make a full assessment.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Comment: this study was a synopsis. It was double‐blind but not enough details are given to make a full assessment.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: this study was a synopsis. Not enough information was given to make a valid assessment. Analysis was based on mixed model for repeated measures. Last observation carried forward was used as supportive analysis. No information on whether intention to treat was used.
Selective reporting (reporting bias)	High risk	Comment: results not fully reported for response, remission, and quality of life. Quotes: "Significant differences between the saredutant 100 mg and placebo groups were not observed with any of the other secondary endpoints. Escitalopram showed significant improvements for a number of the other secondary endpoints including HAM‐A factor (psychic anxiety) and item (anxious mood and tension) scores, CGI‐I scores, MADRS [Montgomery–Åsberg Depression Rating Scale] scores, Q‐LES‐Q‐SF [Quality of Life Enjoyment and Satisfaction Questionnaire Short Form] scores, EWPS [Endicott Work Productivity Scale] scores, and clinical remission based on CGI‐I scores." "A statistically significant greater reduction was observed in the escitalopram 10 mg group compared to placebo, confirming the assay sensitivity of the trial. Supportive analysis based on ANCOVA‐LOCF [analysis of covariance last observation carried forward] did not show statistically significant differences for either the saredutant 100 mg or escitalopram 10 mg groups versus placebo."
Other bias	Unclear risk	Comment: study was sponsored by Aventis Sanofi but given this study was a synopsis a full assessment could not be made.

NCT00658008.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): paroxetine 43.5 (SE 13.5) years; imagabalin 40.9 (SE 12.8) years; placebo 42.1 (SE 12.5) years Sex (women): paroxetine 63.9%; imagabalin 61.2%; placebo 65% Location: Hungary, Italy, Korea, US Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Paroxetine (n = 97) Duration: 8 weeks Treatment protocol: fixed dosage (20 mg/day once daily). Participants randomised to paroxetine received treatment with 20 mg once daily on day 1 and 20 mg every morning for the remainder of the 8‐week treatment phase. 2. Imagabalin (PD 0332334) (n = 295) Duration: 8 weeks Treatment protocol: fixed dosage (150 mg/day, 350 mg/day, or 450 mg/day in 2 doses). The study drugs consisted of blinded oral capsules containing imagabalin 25 mg or imagabalin 100 mg. Study drug was administered to each participants in a twice daily regimen for imagabalin. Participants randomised to receive imagabalin 175 mg BID or imagabalin 225 mg twice daily began dosing with 125 mg at bedtime (HS) on study day 1 (day of randomisation to double‐blind treatment). Participants were titrated up to their full treatment dose over the first week of treatment. Following titration, participants in both imagabalin treatment groups received the full treatment dose for the remainder of the 8‐week double‐blind treatment phase. 3. Placebo (n = 101) Duration: 8 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: baseline, and week 8 Primary outcome Acceptability Secondary outcomes Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Suicide wishes/gestures/attempts Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2008–2009 Funding source: not specified Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: study was double‐blind and (quote) "study drugs consisted of blinded oral capsules." No other reasons to believe that blinding was compromised.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: study was double‐blind (quote) "Study drugs consisted of blinded oral capsules." There are no other reasons to believe that blinding was compromised.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: full analysis set and last observation carried forward analysis were used. Reasons and number of dropouts in each group were reported and were similar between groups. Quote: "Subject discontinuations from day 1 to week 10 overall were similar between the paroxetine 20 mg, PD 0332334 75 mg BID [twice daily], PD 0332334 175 mg BID, and PD 0332334 225 mg BID treatment groups; the lowest overall number of discontinuations occurred in the placebo group."
Selective reporting (reporting bias)	Low risk	Comment: all prespecified outcomes were reported.
Other bias	Unclear risk	Comment: did not specify funding source.

NCT00701675.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): sertraline NA; placebo NA Sex (women): sertraline 3.6%; placebo 0% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Sertraline (n = 28) Duration: 11 weeks Treatment protocol: fixed dosages (50 mg/day or 100 mg/day) 2. Placebo (n = 14) Duration: 11 weeks Treatment protocol: fixed dosages
Outcomes	Time points of assessment: baseline and after 11 weeks Primary outcomes Acceptability Secondary outcomes Change in symptom levels: measured as the score at endpoint on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Agitation/anxiety
Identification
Notes	Date of study: 2005–2009 Funding source: not specified Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Comment: study stated it was "triple blind" but not enough information to properly assess.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Comment: study stated it was triple‐blind but there not enough information to make a full assessment.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: this is an unpublished trial. Not enough information.
Selective reporting (reporting bias)	Unclear risk	Comment: this was an unpublished trial. Not enough information.
Other bias	Unclear risk	Comment: study was sponsored by Veterans Affairs.

Nicolini 2009.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): total 42.8 (SE NA) years Sex (women): total 57.1% Location: Australia, Argentina, Belgium, Canada, Mexico, Russia, Taiwan, UK Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Duloxetine (n = 242) Duration: 10 weeks Treatment protocol: fixed and flexible dosage (20 mg/day or 60–120 mg/day). A duloxetine 20 mg once daily dose was included in this study to explore whether doses of duloxetine < 60 mg/day were effective in the treatment of generalised anxiety disorder. Participants in this group started with a 20‐mg dose that remained fixed during the study. For participants in the duloxetine 60–120 mg/day group, treatment was initiated at 30 mg/day for 1 week and then increased to 60 mg/day. Flexible dosing was allowed in increments of duloxetine 30 mg/day up to a maximum dose of duloxetine 120 mg/day, based on the investigator's judgement. However, a dose increase was required if the CGI‐I scale score was ≥ 3 (minimal improvement, no change, or worse) after 3 weeks of treatment. Dose of study medication could be decreased a total of 2 times for tolerability reasons, provided the participant maintained a minimal dose of duloxetine 60 mg/day. Doses were stabilised after 6 weeks of treatment. 2. Venlafaxine (n = 169) Duration: 10 weeks Treatment protocol: flexible dosage (75–225 mg/day). For the venlafaxine XR 75–225 mg/day group, treatment began with 37.5 mg/day for 1 week and then increased to 75 mg/day. Flexible dosing was allowed in increments of venlafaxine XR 75 mg/day up to a maximum dose of venlafaxine XR 225 mg/day, based on the investigator's judgement. However, a dose increase was required if the CGI‐I scale score was ≥ 3 (minimal improvement, no change, or worse) after 3 weeks of treatment. Dose of study medication could be decreased a total of 2 times for tolerability reasons, provided the participant maintained a minimal dose of venlafaxine XR 75 mg/day. Doses were stabilised after 6 weeks of treatment. 3. Placebo (n = 170) Duration: 10 weeks Treatment protocol: matched placebo
Outcomes	Time points of assessment: baseline (week 0), weeks 1, 2, 4, 7, 10 Primary outcomes Rate of treatment response: defined as a reduction of ≥ 50% on the HAM‐A Acceptability Secondary outcomes Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2005–2007 Funding source: not specified Declarations of interest among primary researchers: Dr Nicolini has received research funding from Eli Lilly and Company, Wyeth Pharmaceuticals, and the National Institute of Mental Health. Dr Bakish has received grant support and has served as a consultant and a member of the speaker's bureau for Eli Lilly and Company. Drs Duenas, Spann, Ball, Sagman and Russell, and C Hallberg are employees and shareholders of Eli Lilly and Company. Dr Erickson was a former employee of Eli Lilly and Company. Author of this study, David Bakish (dbakish@opctrials.com) was contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Assignment to therapy was determined by a computer‐generated random sequence using an Interactive Voice Response System (IVRS)."
Allocation concealment (selection bias)	Low risk	Quote: "assignment to therapy was determined by a computer‐generated random sequence using an Interactive Voice Response System (IVRS) … The IVRS automatically dispensed medication for all treatment arms based on physician assessment."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: study stated "double‐blind" and there is no reason to believe that blinding of participants and personnel was broken.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: double‐blind. No reason to believe blinding was compromised.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: intention to treat population and last observation carried forward analysis were used. Reasons and numbers of dropouts were reported. There is imbalance in discontinuation rates between groups only due to lack of efficacy. Quote: "Significantly more placebo‐treated patients discontinued due to lack of efficacy compared with each of the three active treatment groups (duloxetine 20 mg/day, p<0.05; duloxetine 60–120 mg/day and venlafaxine XR 75–225 mg/day, p < 0.001 for both comparisons)."
Selective reporting (reporting bias)	Low risk	Comment: all prespecified outcomes were reported.
Other bias	Unclear risk	Comment: did not specify funding source.

Nimatoudis 2004.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): venlafaxine 41.0 (SE 14.0) years; placebo 44.0 (SE 12.0) years Sex (women): venlafaxine 66.7%; placebo 68.2% Location: Greece Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Venlafaxine (n = 24) Duration: 8 weeks Treatment protocol: 75–150 mg/day. Participants with < 30% decrease in their HAM‐A total score at the end of 2 weeks, compared to the end of the prestudy period, doubled their dose (venlafaxine XR 150 mg/day) for the rest of the treatment period (6 weeks) 2. Placebo (n = 24) Duration: 8 weeks Treatment protocol: not specified
Outcomes	Time points of assessment: baseline, and week 8 Primary outcomes Rate of treatment response: defined as a reduction of ≥ 50% on the HAM‐A Acceptability Secondary outcomes Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects
Identification
Notes	Date of study: not specified Funding source: not specified Declarations of interest among primary researchers: not specified Author of this study, Ioannis Nimatoudis (zissisn@wyeth.com) was contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: study was double‐blind. Quote: "study medications were provided in blister packs, with double‐dummy techniques for blinding purposes. Venlafaxine XR and matching placebo were supplied in identical colored capsules.""
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: double‐blind. No reason to believe blinding was compromised.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: intention‐to‐treat population and last observation carried forward analysis were used. Little information on dropouts; little information on reasons for dropouts in each group and no statistics were provided to determine if there was a difference between the 2 groups.
Selective reporting (reporting bias)	High risk	Comment: did not fully report results for response rate (defined by study authors) and adverse events. Quote: "Responses based on CGI‐I scores of 1 (very much improved) or 2 (much improved) did not indicate statistically significant differences until the last visit (day 57) … Adverse events were consistent with those reported in the literature."
Other bias	Unclear risk	Comment: did not specify funding source.

PAR 29060 637.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): paroxetine 46.5 (SE 14.9) years; placebo 45.4 (SE 15.0) years Sex (women): paroxetine 74.6%; placebo 66.5% Location: UK, France, Ireland, Germany, Austria, Italy Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Paroxetine (n = 186) Duration: 8 weeks Treatment protocol: flexible dosage (20–50 mg/day). All participants randomised to paroxetine received 20 mg/day for the first 2 weeks of the active treatment phase. The dose could then be uptitrated in 10 mg/day increments at intervals no more frequently than every 7 days at the discretion of the investigator, according to clinical response and tolerability. Participants receiving ≥ 30 mg/day, were allowed a dose reduction to the next lower level consequent to an adverse event. If the adverse events abated, the dose could be returned to the original level. Participants requiring > 1 dose reduction were withdrawn from the study. 2. Placebo (n = 188) Duration: 8 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: baseline and week 8 Primary outcomes Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Remission rate: defined as the proportion of subjects with a score of 10 or less on the HAM‐A at endpoint Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Agitation/anxiety Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 1998–1999 Funding source: not specified Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Comment: this study was a synopsis. It was double‐blind but not enough details are given to make a full assessment.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Comment: this study was a synopsis. It was double‐blind but not enough details are given to make a full assessment.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: intention‐to‐treat population and last observation carried forward analysis were used. Number and reasons for dropouts in each group were reported but it is unclear if these rates differed between the groups.
Selective reporting (reporting bias)	Low risk	Comment: all prespecified outcomes were reported
Other bias	Unclear risk	Comment: did not specify funding source.

Pollack 2001.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): paroxetine 39.7 (range 19–69) years; placebo 41.3 (range 19–80) years Sex (women): paroxetine 60.9%; placebo 66.3% Location: USA and Canada Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Paroxetine (n = 161) Duration: 8 weeks Treatment protocol: flexible dosage (20–50 mg/day). Participants randomly assigned to paroxetine started treatment at 10 mg/day for the first week and received 20 mg/day during the second week. Participants who could not tolerate the study medication during the first 2 weeks were removed from the study. After week 2, the paroxetine dosage could be increased every 7 days by 10 mg/day up to 50 mg/day. During the study treatment, only a single dosage reduction (because of physical illness or an adverse event) was allowed in patients taking ≥ 30 mg/day. 2. Placebo (n = 163) Duration: 8 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: baseline, and weeks 1, 2, 3, 4, 5, 6, 8 Primary outcomes Acceptability Secondary outcomes Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: not specified Funding source: GlaxoSmithKline Declarations of interest among primary researchers: financial disclosure: Dr Pollack has served as a consultant for Pfizer, SmithKline Beecham, and Wyeth; has received grant/research support from Forest, Glaxo, Lilly, Pfizer, SmithKline Beecham, and Wyeth; has received honoraria from Forest, Lilly, Pfizer, SmithKline Beecham, and Wyeth; and has served on the speakers or advisory boards for Forest, Lilly, Pfizer, SmithKline Beecham, and Wyeth. Mr McCafferty is an employee of and a major stock shareholder in GlaxoSmithKline. Drs Zaninelli, Burnham, and Iyengar and Mr Bellew are employees of GlaxoSmithKline. Author of this study, Mark Pollack (mpollack@partners.org) was contacted for additional information but could not be reached.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Quote: study was double‐bind but authors did not specify details on how blinding was maintained when doses were adjusted. Quote: "During the study treatment, only a single dosage reduction (because of physical illness or an adverse event) was allowed in patients taking at least 30 mg/day (or placebo equivalent)."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: double‐blind. No reason to believe blinding was compromised.
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: intention‐to‐treat population and last observation carried forward analysis were used. Study did not report dropouts nor reasons for dropouts.
Selective reporting (reporting bias)	Low risk	Comment: all prespecified outcomes were reported.
Other bias	High risk	Quote: "Support by GlaxoSmithKline. Dr. Pollack has served as a consultant for Pfizer, SmithKline Beecham, and Wyeth; has received grant/research support from Forest, Glaxo, Lilly, Pfizer, SmithKline Beecham, and Wyeth; has received honoraria from Forest, Lilly, Pfizer, SmithKline Beecham, and Wyeth; and has served on the speakers or advisory boards for Forest, Lilly, Pfizer, SmithKline Beecham, and Wyeth. Mr. McCafferty is an employee of and a majority stock shareholder in GlaxoSmithKline. Drs. Zaninelli, Burnham, and Iyengar and Mr. Bellew are employees of GlaxoSmithKline."

Rickels 2000.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): venlafaxine 40.8 (SE 12.4) years; placebo 40.9 (SE 11.3) years Sex (women): venlafaxine 55.3%; placebo 57% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Venlafaxine (n = 253) Duration: 8 weeks Treatment protocol: fixed dosage (75 mg/day, 150 mg/day, or 225 mg/day). During the first week of treatment all participants who were taking venlafaxine XR received 75 mg/day; on days 8–14 participants randomly assigned to 150 mg/day or 225 mg/day received 150 mg/day; beginning on day 15 participants in the high‐dose group received 225 mg/day. After day 15, all participants were maintained at their respective venlafaxine XR doses for the remainder of the study. 2. Placebo (n = 96) Duration: 8 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: baseline and days 8, 15, 22, 29, 43, and 57 Primary outcomes Acceptability Secondary outcomes Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification	Date of study: not specified Funding source: Wyeth‐Ayerst Laboratories Declarations of interest among primary researchers: not specified Author of this study, Karl Rickels (krickels@pennmedicine.upenn.edu) was contacted for additional information but could not provide additional information due to retirement.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: study stated double‐blind. Results do not suggest blinding was compromised.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: double‐blind. No reason to believe blinding was compromised.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: intention‐to‐treat population and last observation carried forward analysis were used. There was significantly less attrition due to adverse events in the placebo group compared to the other groups but no significant difference between groups for other sources of attrition (table 2). Unclear if those who dropped out differed from those who did not.
Selective reporting (reporting bias)	High risk	Comment: response rate (defined by study authors) was not fully reported. Quote: "The results of the analyses of secondary efficacy variables support the findings of a superior response with venlafaxine XR."
Other bias	Unclear risk	Quote: "Supported by Wyeth‐Ayerst Laboratories." Comment: declarations of interest and involvement in the study were not clearly reported.

Rickels 2003.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): paroxetine 40.4 (SE 12.7) years; placebo 40.8 (SE 12.6) years Sex (women): paroxetine 55%; placebo 56% Location: USA and Canada Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Paroxetine (n = 385) Duration: 8 weeks Treatment protocol: fixed dosage (20 mg/day or 40 mg/day). All participants receiving paroxetine started at 10 mg/day for the first week of treatment. Thereafter, the paroxetine dose was increased by 10 mg/week until the fixed‐dose was reached (at the beginning of week 2 and week 4 for the 20‐mg and 40‐mg arms, respectively) where they remained for the duration of the study. Dosage adjustments or interruptions were not permitted during the study for any reason. 2. Placebo (n = 188) Duration: 8 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: weeks 1, 2, 3, 4, 6, and 8 Primary outcomes Acceptability Secondary outcomes: Rate of treatment response: defined as a score of 1 or 2 on the CGI‐I scale Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: not specified Funding source: GlaxoSmithKline Declarations of interest among primary researchers: not specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Comment: study stated "double‐blind" but did not mention who was blinded. Study 170 (which is the same as this one) mentions that (quote) "PBO [placebo] subjects received study medication identical in appearance to the active medication."
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Quote: "Not enough information on outcome assessors."
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: intention‐to‐treat population and last observation carried forward analysis were used. Authors reported reasons for dropout. Quote: "Completion rates did not differ substantially among patients given placebo (77.8%), 20mg of paroxetine (76.1%), or 40mg of paroxetine (72.6%) (chi‐squared = 1.42, df = 2, p <0.05), and there were no significant differences in the demographic or clinical characteristics between the patients who dropped out and those who completed the study … The number of patient withdrawals from the study because of lack of efficacy, protocol violations, or being lost to follow‐up was similar across the three treatments." However, it is unclear whether dropouts due to adverse events was different between groups."
Selective reporting (reporting bias)	Low risk	Comment: all prespecified outcomes were reported.
Other bias	Unclear risk	Quote: "Supported by GlaxoSmithKline. Dr. Rickels assumes full responsibility for the scientific content of this article. He was involved to a significant degree in the protocol development, data collection, data analysis, and manuscript preparation phases of this project. Dr. Rickels had full access to all data collected and major input into the data analysis plan."

Rynn 2008.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): duloxetine 42.2 (SE 13.9) years; placebo 41.0 (SE 14.2) years Sex (women): duloxetine 61.3%; placebo 62.3% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Duloxetine (n = 168) Duration: 10 weeks Treatment protocol: flexible dosage (60–120 mg/day). The starting dose was 60 mg/day, but a dose decrease to 30 mg was possible during the first 2 weeks to allow participants to acclimate to the medication. By week 2, participants in the duloxetine group were required to take a minimum dose of 60 mg/day. Participants' doses were progressively titrated at each subsequent visit; that is, they were required to increase their dose (duloxetine 30 mg/day) if their CGI‐I rating was ≥ 3 (minimal improvement, no change, or worsening) and they were able to tolerate a dose increase. Participants could be increased to a maximum dose of 120 mg/day. 2. Placebo (n = 159) Duration: 10 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: baseline, and week 1, 2, 4, 7, 10 Primary outcomes Rate of treatment response: defined as a reduction of at least 50% on the HAM‐A Acceptability Secondary outcomes Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: not specified Funding source: Eli Lilly and Company, Boehringer Ingelheim Declarations of interest among primary researchers: Karl Rickels, MD, has received funding or served as a speaker for DOV Pharmaceuticals, Eli Lilly and Company, Medinova, Merck, Novartis, Pfizer, Pherin Pharmaceuticals, PreDix Pharmaceuticals, Sanofi‐Synthelabo Research, and Wyeth Laboratories. Dr Rynn has served as a consultant, and speaker, or received research funding from Astra Zeneca, Eli Lilly and Company, Pfizer, Predix Pharmaceuticals, Forest Pharmaceuticals, Abbott Laboratories, Wyeth Pharmaceuticals, and NIMH.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Comment: study stated it was "double‐blind" but did not specify who was double‐blinded (other than participants); i.e. whether it was the principal investigators, subinvestigators, or raters that were also blinded.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Comment: not enough information on outcome assessors.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: intention‐to‐treat population and last observation carried forward and mixed model for repeated measures analyses were used. Reasons for attrition or exclusions (or both) were reported. There was no difference between the treatment groups for reasons for dropout (except due to adverse events). Quote: "Treatment groups did not differ in their reason for discontinuation with the exception of adverse events for the duloxetine‐treated group (P = .002)."
Selective reporting (reporting bias)	Low risk	Comment: all prespecified outcomes were reported.
Other bias	High risk	Quote: "Contract grant sponsor: Eli Lilly and Company; Contract grant sponsor: Boehringer Ingelheim." Comment: Drs Russell, Erikson, Detke, Ball, and Dinkel are all employees or shareholders (or both) of Eli Lilly and Company.

Stein 2008.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age (mean): total 41.7 (SE 12.2) years Sex (women): total 68.6% Location: Finland, South Africa Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 or 2 groups 1. Agomelatine (n = 63) Duration: 12 weeks Treatment protocol: 25–50 mg/day. The dosage of agomelatine could be increased from 25 mg/day to 50 mg/day based on insufficient improvement from 2 weeks onward in a blinded manner 2. Placebo (n = 58) Duration: 12 weeks Treatment protocol: not specified
Outcomes	Time points of assessment: baseline, and weeks 2, 4, 6, 8, 12 Primary outcomes Rate of treatment response: defined as a reduction of at least 50% on the HAM‐A Acceptability Secondary outcomes: Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2005–2006 Funding source: Servier Declarations of interest among primary researchers: Drs Stein and Ahokas have received consultancy honoraria from Servier. Dr Stein has also received research grants or consultancy honoraria (or both) from AstraZeneca, Eli Lilly, GlaxoSmithKline, Johnson & Johnson, H. Lundbeck, Orion, Pfizer, Pharmacia, Roche, Solvay, Sumitomo, Tikvah, and Wyeth. Dr de Bodinat is a full‐time employee of Servier. Author of this study, Dan Stein (dan.stein@uct.ac.za) was contacted for additional information but referred us to the synopsis of the clinical study results, which we were already aware of and had included in the review.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Randomization was balanced and stratified on the factors center and previous intake of psychotropic treatment (benzodiazepines and/or antidepressants) and was done using an Interactive Voice Response System."
Allocation concealment (selection bias)	Low risk	Quote from study 141: "The dose increase and the treatment allocation were made centrally in a double‐blind procedure, so that both patients and investigators were blind to the dose increase."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: study was double‐blind and treatments were identically labeled to maintain blinding. Quote: "Dosage of agomelatine could be increased from 25 to 50 mg daily based on insufficient improvement from 2 weeks onward in a blinded fashion according to a predefined dose adjustment algorithm that was shared neither with investigators nor subjects."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: study was double‐blind. Quote: "treatments were identically labeled so as to maintain blinding … dosage of agomelatine could be increased from 25 to 50 mg daily based on insufficient improvement from 2 weeks onward in a blinded fashion according to a predefined dose adjustment algorithm that was shared neither with investigators nor subjects."
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: intention‐to‐treat population was used. Last observation carried forward was not used (last value analysis was used). Reasons and rates of withdrawal were reported for each group. Quote: "there were no differences in rates of withdrawal between agomelatine and placebo." Comment: sample size was small and this could have affected the results.
Selective reporting (reporting bias)	High risk	Comment: response rate (defined by study authors) not fully reported. Quote: "Similar results were obtained when response was defined in a CGI improvement scale of 2 or less." Comment: adverse events not fully reported.
Other bias	Unclear risk	Quote: "Supported by IRIS. Professors Stein and Seedat are supported by the Medical Research Council of South Africa. Dr. de Bodinat was a full‐time employee of Servier at the time of the study." Comment: declarations of interest and involvement in the study were not clearly reported.

Stein 2014.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age (mean): agomelatine 43.6 (SE 12.5) years; escitalopram 41.2 (SE 12.5) years; placebo 43.0 (SE 12.2) years Sex (women): agomelatine 74.8%; escitalopram 68.3%; placebo 71.8% Location: Finland, Russia, Poland, Czech Republic, Slovakia, Argentina, South Korea Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Agomelatine (n = 139) Duration: 12 weeks Treatment protocol: 10–20 mg/day. On the basis of insufficient improvement, daily dosage of agomelatine could be increased at week 4 (agomelatine: from 25 mg to 50 mg). 2. Escitalopram (n = 142) Duration: 12 weeks Treatment protocol: 25–50 mg/day. On the basis of insufficient improvement, daily dosage of escitalopram could be increased at week 4 (escitalopram: from 10 mg to 20 mg). 3. Placebo (n = 131) Duration: 12 weeks Treatment protocol: not specified
Outcomes	Time points of assessment: baseline, and weeks 2, 4, 8, 12 Primary outcomes Rate of treatment response: defined as a reduction of ≥ 50% on the HAM‐A Acceptability Secondary outcomes Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: not specified Funding source: Servier Declarations of interest among primary researchers: potential conflicts of interest: Dr Stein has received research grants or consultancy honoraria (or both) from Abbott, AstraZeneca, Eli Lilly, GlaxoSmithKline, Jazz Pharmaceuticals, Johnson & Johnson, Lundbeck, Orion, Pfizer, Pharmacia, Roche, Servier, Solvay, Sumitomo, Takeda, Tikvah, and Wyeth. Dr Ahokas has received research grants or consultancy honoraria (or both) from AstraZeneca, Bristol‐Myers Squibb, Eli Lilly, GlaxoSmithKline, Lundbeck, Orion, Otsuka, Sanofi‐Aventis, Servier, and Wyeth. Dr Márquez has received research grants or consultancy honoraria (or both) from Pfizer, Servier, Bagó, Eli Lilly, Gador, Temis Lostaló, Forester, Otsuka, Shering, Sanofi‐Aventis, and Wyeth; and has served on speakers or advisory boards for Janssen‐Cilag, GlaxoSmithKline, and AstraZeneca. Dr Höschl has received research support from Servier, has received honoraria from Lundbeck Institute, and has served on speakers or advisory boards for Eli Lilly. Dr Oh has been a consultant to Servier and has received grant/research support from Servier, Lilly, GlaxoSmithKline, and Otsuka. Dr Jarema has been a consultant to Bristol‐Myers Squibb, Eli Lilly, GlaxoSmithKline, Janssen, Servier, and Roche; and has received honoraria from and served on speakers or advisory boards for Bristol‐Myers Squibb, Eli Lilly, GlaxoSmithKline, Janssen, Medagro, and Servier. Dr Avedisova has received honoraria from AstraZeneca, Bristol‐Myers Squibb, Eli Lilly, GlaxoSmithKline, Lundbeck, Sanofi‐Aventis, Servier; and has served on speakers or advisory boards for Servier and AstraZeneca. Drs Albarran and Olivier are employees of Servier.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Randomization was balanced and stratified by center and was done using an interactive response system."
Allocation concealment (selection bias)	Low risk	Quote: "Randomization was balanced and stratified by center and was done using an interactive response system. Treatments were identically labelled."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "Treatments were identically labelled, daily dosage of agomelatine or escitalopram could be increased at week 4 in blinded fashion according to a predefined dose adjustment algorithm (blind for investigators and patients)."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: double‐blind. No reason to believe blinding was compromised.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: full analysis set and unclear which analysis was used. All attritions and exclusions were reported along with reasons.
Selective reporting (reporting bias)	Low risk	Comment: all prespecified outcomes were reported.
Other bias	High risk	Quote: "This study was sponsored by Servier. The sponsored played a key role in leading the design and conduct of the study, including data collection, management, and analysis." Comment: Drs Albarran and Olivier are employees of Servier.

Stein 2017.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age (mean): agomelatine 43.9 (SE 14.3) years; placebo 44.1 (SE 13.1) years Sex (women): agomelatine 70%; placebo 63.4% Location: Finland, Russia, Poland, Slovakia, Ukraine Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Agomelatine (n = 270) Duration: 12 weeks Treatment protocol: fixed dosage (10 mg/day or 25 mg/day) 2. Placebo (n = 142) Duration: 12 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: baseline, and weeks 2, 4, 8, 12 Primary outcomes Rate of treatment response: defined as a reduction of at least 50% on the HAM‐A Acceptability Secondary outcomes Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2013–2015 Funding source: Servier Declarations of interest among primary researchers: Dr Stein has received research grants or consultancy honoraria (or both) from Abbott, AstraZeneca, Eli Lilly, GlaxoSmithKline, Jazz Pharmaceuticals, Johnson & Johnson, Lundbeck, Orion, Pfizer, Pharmacia, Roche, Servier, Solvay, Sumitomo, Takeda, Tikvah, and Wyeth. A Ahokas has received research grants or consultancy honoraria (or both) from AstraZeneca, Bristol‐Myers Squibb, Eli Lilly, GlaxoSmithKline, Lundbeck, Orion, Otsuka, Sanofi‐Aventis, Servier and Wyeth. M Jarema has received research grants or consultancy honoraria (or both) from BMS, Eli‐Lilly, GlaxoSmithKline, Janssen, Medagro, Roche, Servier. As Avedisova has received research grants or consultancy honoraria (or both) from AstraZeneca, Bristol‐Myers Squibb, Eli Lilly, GlaxoSmithKline, Lundbeck, Sanofi‐Aventis, Servier. L Vavrusova has no disclosures. O Chaban has received research grants or consultancy honoraria (or both) from Alkermes, AstraZeneca, Eli Lilly, Janssen, Lundbeck, Novartis, MSD, Otsuka, Pfizer and Servier. C Gruget, V Olivier, F Picarel‐Blanchot, and C de Bodinat are employees at Servier.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "The treatments were assigned at the inclusion visit by a balanced (non‐adaptive and noncentralized) randomization with stratification by centre."
Allocation concealment (selection bias)	Unclear risk	Quote: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: study was double‐blind and treatments were "identically labeled." No other reasons to believe blinding was compromised.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: study was double‐blind. Quote: "Treatments were identically labeled." Comment: no other reasons to believe blinding was compromised.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quote: "Full Analysis Set (FAS) and LOCF [last observation carried forward] analysis were used. Mixed‐effects model for repeated measures (MMRM) was used to test robustness of results. Reasons and rates of withdrawal for each treatment group were provided but it is unclear of these rates differed statistically between groups."
Selective reporting (reporting bias)	High risk	Comment: all prespecified outcomes were reported. Secondary outcome measures were performed post‐hoc making results available to investigators before analysis was done. Results could have been altered based on knowledge of assignment to intervention. Quote: "Post‐hoc analyses using a two‐way analysis of covariance model on treatment and centre …"
Other bias	High risk	Quote: "This study was funded by Servier. Servier employees were involved in the collection and analysis of data."

Wen‐Yuan 2011.

Study characteristics
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): duloxetine 37.3 (SE 11.9) years; placebo 38.0 (SE 12.0) years Sex (women): duloxetine 46.3%; placebo 54.9% Location: China Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Duloxetine (n = 108) Duration: 15 weeks Treatment protocol: 60–120 mg/day orally. Participants who were randomly assigned to duloxetine received 60 mg orally for 7 weeks; at that point, participants who had no response had their daily duloxetine dose increased to 120 mg for the remaining 8. Participants who did not tolerate the increase to the 120‐mg dose were discontinued from the study. 2. Placebo (n = 102) Duration: 15 weeks Treatment protocol: not specified
Outcomes	Time points of assessment: baseline, and weeks 1, 3, 7, 11, 15 Primary outcomes Rate of treatment response: defined as a reduction of ≥ 50% on the Hamilton Anxiety Rating Scale Acceptability Secondary outcomes Remission rate: defined as a score of ≤ 7 on the HAM‐A Change in symptom levels: measured as the mean change from baseline on the HAM‐A total score Total number of participants reporting adverse effects Sleepiness/drowsiness Suicide wishes/gestures/attempts Dropouts due to a lack of efficacy Dropouts due to adverse effects
Identification
Notes	Date of study: 2008–2010 Funding source: Eli Lilly and Company and Boehringer Ingelheim Declarations of interest among primary researchers: Drs WU Wen‐yuan and WANG Gang served as investigators in this study and received funding from Eli Lilly and Company, Indianapolis, IN, USA. Dr ANG Qui‐qing is now working in GlaxoSmithKline (China), Shanghai, China. Drs Susan G Ball and Durisala Desaiah are employees of Eli Lilly and Company and hold company stocks.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "No information."
Allocation concealment (selection bias)	Unclear risk	Comment: "No information."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: double‐blind. No reason to believe blinding was compromised.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: double‐blind. No reason to believe blinding was compromised.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: intention‐to‐treat population and last observation carried forward analysis were used. All exclusions and attritions were reported, as well as reasons.
Selective reporting (reporting bias)	Low risk	Comment: all prespecified outcomes were reported.
Other bias	High risk	Quote: "Drs. WU Wen‐yuan and WANG Gang served as investigators in this study and received funding from Eli Lilly and Company, Indianapolis, IN, USA. Dr. ANG Qui‐qing is now working in GlaxoSmithKline (China), Shanghai, China. Drs. Susan G Ball and Durisala Desaiah are employees of Eli Lilly and Company and hold company stocks."

CGI‐C: Clinical Global Impression of Change; CGI‐I: Clinical Global Impression – Improvement; DSM‐IV: Diagnostic and Statistical Manual of Mental Disorders. 4th edition; DSM‐IV‐TR: Diagnostic and Statistical Manual of Mental Disorders. 4th, Text Revision edition; n: number of participants; HAM‐A: Hamilton Anxiety Rating Scale; NA: not available; SE: standard error.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Alaka 2014	Concomitant medical problem
Allgulander 2006	Ineligible study design
Allgulander 2008	Ineligible study design
Baldwin 2012b	Ineligible study design
Baldwin 2012c	Ineligible study design
Bodkin 2011	Ineligible outcomes
Chakhava 2011	Ineligible study design
Davidson 2005	Ineligible study design
Davidson 2008	Ineligible study design
EudraCT 2004‐001500‐13	Study was cancelled
EudraCT 2004‐002626‐22	Study was cancelled
EudraCT 2006‐006339‐31	Ineligible intervention
EudraCT 2008‐003421‐17	Study was cancelled
Goldstein 2002	Ineligible participant population
Hackett 2000	Ineligible study design
Hong‐Wei 2006	Ineligible study design
ISRCTN75061806	Ineligible comparator
Kapczinski 2016	Ineligible study design
Kasper 2014	Ineligible participant population
Kornstein 2009	Ineligible study design
Lenze 2005	Concomitant benzodiazepine use
Lenze 2009	Concomitant benzodiazepine use
Lindsay 1987	Ineligible intervention
Lydiard 1999	Ineligible study design
Montgomery 2002	Ineligible study design
Rickels 1993	Concomitant medical problem
Rolland 2000	Concomitant medical problem
Sheehan 1999	Ineligible study design
Stein 2011	Ineligible study design
Stocchi 2003	Ineligible study design
Westenberg 2004	Ineligible participant population
Wingerson 1992	Ineligible study design
Wright 2009	Ineligible study design

Characteristics of studies awaiting classification [ordered by study ID]

Allgulander 2007.

Methods	Study design: parallel, randomised, double‐blind, flexible‐dose, multinational, clinical trial designed to evaluate the clinical efficacy of escitalopram in relapse prevention in GAD
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age (mean): escitalopram 42.1 (SE 11.3); placebo 40.7 (SE 11.9) Sex (women): escitalopram 62.5%; placebo 59.3% Location: France, Canada, Germany, Sweden Comorbidities: none stated
Interventions	Interventions 1. Escitalopram (n = 187) Duration: 24–76 weeks Treatment Protocol: fixed dosage (20 mg/day, orally) 2. Placebo (n = 188) Duration: 24–76 weeks Treatment protocol: fixed dosage, orally
Outcomes	Time points of assessment: not specified Health‐related quality of life: SF‐36 scale Worked productivity: Work Limitation Questionnaire and the work efficacy and work satisfaction visual analogue scale
Notes	Study did not report results that were relevant to our review

Baldwin 2002.

Methods	Study design: 3 randomised, double‐blind, placebo‐controlled trials
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age: not specified Sex: not specified Location: not specified Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of the following Study 1 1. Paroxetine (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage (20 mg/day or 40 mg/day) 2. Placebo (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage Studies 2 and 3 1. Paroxetine (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage (20–50 mg/day) 2. Placebo (n = NA) Duration: 8 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: not specified Efficacy: mean change from baseline in HAM‐A total score Disability: Sheehan Disability Scale
Notes	Only reported Sheehan Disability Scale scores

Baldwin 2012d.

Methods	Study design: pooled analysis of 4 double‐blind, placebo‐controlled clinical trials
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): pregabalin 39.7 (SE NA) years; alprazolam 40.3 (SE NA) years; venlafaxine XR 42.6 (SE NA) years; placebo 41.2 (SE NA) years Sex (women): pregabalin: 61.7%; alprazolam: 65.9%; venlafaxine: 57.6%; placebo: 58.6% Location: USA, Austria, Belgium, Germany, Netherlands, UK, Canada, France, Ireland, Italy, Spain, and Sweden Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of the following 1. Pregabalin (n = 752) Duration: 4–8 weeks Treatment protocol: initial doses of pregabalin 250–300 mg/day, and were titrated by week 1 to 1 of the following fixed doses: 200 mg (n = 78), 300 mg (n = 91), 400 mg (n = 186), 450 mg (n = 178), 600 mg (n = 199), or flexible dosage (300–600 mg/day) 2. Alprazolam (n = 88) Duration: 4–8 weeks Treatment protocol: fixed dosage (1.5 mg/day) 3. Venlafaxine (n = 120) Duration: 4–8 weeks Treatment protocol: flexible dosage (75–225 mg/day) 4. Placebo (n = 394) Duration: 4–8 weeks Treatment protocol: not specified
Outcomes	Time points of assessment: not specified Change in HAM‐A total score
Notes	This study was looking at whether early improvement could predict endpoint response. Also looked at the specificity and sensitivity of various measures of early improvement. Authors did not report results that were relevant to our study.

Busnello 1974.

Methods	NA
Participants	NA
Interventions	NA
Outcomes	NA
Notes	Could not find full‐text

Coric 2009.

Methods	Study design: the Sheehan‐STS was incorporated into a multicenter, randomised, double‐blind, active comparator (escitalopram), and placebo‐controlled, proof‐of‐concept study
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age: not specified Sex: not specified Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of the following 1. Pexacerfont (n = 24) Duration: 8 weeks Treatment protocol: fixed dosage (100 mg/day) 2. Escitalopram (n = 12) Duration: 8 weeks Treatment protocol: fixed dosage (10 mg/day) 3. Placebo (n = 25) Duration: 8 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: baseline, weeks 2, 4, 8 Mean change from baseline in HAM‐A total score and HAM‐D‐17 total score Sheehan Disability Scale was designated an exploratory outcome measure
Notes	This study was evaluating efficacy of the Sheehan Disability Scale tool. Study did not report results that were relevant to our study.

Derivan 1997.

Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age: not specified Sex: not specified Location: not specified Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Venlafaxine (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage (75 mg/day or 150 mg/day). The venlafaxine group had their dose titrated during the first week of therapy 2. Buspirone (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage (30 mg/day). The buspirone group had their dose titrated during the first week of therapy 3. Placebo (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: weeks 1, 2, 3, 4, 6, 8 HAM‐A total score HAM‐A psychic anxiety factor Clinical Global Impressions scale The anxiety subscale from the Hospital Anxiety and Depression Scale HAM‐A anxious mood item Safety
Notes	Study was not a full text, did not report results that could be extracted.

Haskins 1998.

Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age: not specified Sex: not specified Location: not specified Comorbidities: none stated
Interventions	Participants were randomly assigned to either 1 of 2 groups (total n = 377): 1. Venlafaxine Duration: 8 weeks Treatment protocol: fixed dosage (75 mg/day, 150 mg/day, 225 mg/day). Participants began treatment with venlafaxine 75 mg daily. At week 2, the venlafaxine middle‐dose and high‐dose groups were increased to 150 mg/day; at week 3 the venlafaxine high‐dose group was increased to 225 mg/day. 2. Placebo Duration: 8 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: 1, 2, 3, 4, 6, 8 weeks Improvement: HAM‐A total score, the HAM‐A psychic anxiety factor, the Clinical Global Impressions scale, the anxiety subscale of the Hospital Anxiety and Depression Scale, the HAM‐A somatic anxiety factor, the Clinical Anxiety Scale, and the anxious mood and tension items of the HAM‐A Safety
Notes	Study was not a full text, did not report results that could be extracted.

Hoffman 2009.

Methods	NA
Participants	NA
Interventions	NA
Outcomes	NA
Notes	Could not find full‐text

Huang 2003.

Methods	NA
Participants	NA
Interventions	NA
Outcomes	NA
Notes	Could not find full‐text

Kasper 2002.

Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: not specified Age (mean): total 44 years Sex (women): total 62% Location: not specified Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of 3 groups 1. Pregabalin (n = 209) Duration: 6 weeks Treatment protocol: fixed dosage (400 mg/day or 600 mg/day in 2 doses) 2. Venlafaxine (n = 114) Duration: 6 weeks Treatment protocol: fixed dosage (75 mg/day in 2 doses) 3. Placebo (n = 103) Duration: 6 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: not specified Efficacy: mean change from baseline to endpoint in the HAM‐A total score Safety was assessed through clinical and laboratory observations
Notes	Study was not a full text, did not report results that could be extracted.

Meoni 2001.

Methods	Study design: pooled analysis of 5 randomised, placebo‐controlled, double‐blind studies
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age: not specified Sex: not specified Location: not specified Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of the following Study 1 1. Venlafaxine (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage (75 mg/day, 150 mg/day, 225 mg/day) 2. Placebo (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage Study 2 1. Venlafaxine (n = NA) Duration: 8 weeks Treatment protocol: flexible dosage (75 mg/day or 150 mg/day) 2. Buspirone (n = NA) Duration: 8 weeks Treatment protocol: flexible dosage (30 mg/day) 3. Placebo (n = NA) Duration: 8 weeks Treatment protocol: flexible dosage Study 3 1. Venlafaxine (n = NA) Duration: up to 6 months Treatment protocol: flexible dosage (75–225 mg/day) 2. Placebo (n = NA) Duration: up to 6 months Treatment protocol: flexible dosage Study 4 1. Diazepam (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage (15 mg/day) 2. Venlafaxine (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage (75 mg/day or 150 mg/day) 3. Placebo (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage Study 5 1. Venlafaxine (n = NA) Duration: up to 6 months Treatment protocol: fixed dosage (37.5 mg/day, 75 mg/day, 150 mg/day) 2. Placebo (n = NA) Duration: up to 6 months Treatment protocol: fixed dosage
Outcomes	Time points of assessment: not specified Change in the total HAM‐A score between baseline and the end of therapy Body surface area evaluations at baseline and during the course of the studies
Notes	No data relevant to our review could be extracted.

Mychaskiw 2009a.

Methods	NA
Participants	NA
Interventions	NA
Outcomes	NA
Notes	Could not find full‐text

Mychaskiw 2009b.

Methods	NA
Participants	NA
Interventions	NA
Outcomes	NA
Notes	Could not find full‐text

NCT00332891.

Methods	Study design: double‐blind, randomised, parallel‐group, placebo‐controlled study
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age: not specified Sex: not specified Location: Bulgaria, Croatia, Czech Republic, Poland, Serbia Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of the following Study 1 1. Amibegron (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage (175 mg and 350 mg every 12 hours) 2. Paroxetine (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage (20 mg/day every 12 hours) 3. Placebo (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage, orally
Outcomes	Time points of assessment: baseline, day 56 Change from baseline in 14‐item HAM‐A total score Changes from baseline in the Clinical Global Impression – Severity of Illness Score Change from baseline in the percentage of participants with HAM‐A treatment response Change from baseline in the HAM‐A somatic and psychic anxiety factor scores
Notes	No results posted

NCT00390650.

Methods	Study design: randomised, parallel‐group, double‐blind study
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV‐TR Age: not specified Sex: not specified Location: US, Canada Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of the following Study 1 1. Sareduatant (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage (100 mg) 2. Placebo (n = NA) Duration: 8 weeks Treatment protocol: fixed dosage
Outcomes	Time points of assessment: baseline, day 56 Change from baseline in the HAM‐A total score Change from baseline in the Clinical Global Impression – Severity of Illness score
Notes	No results posted

NCT00715039.

Methods	NA
Participants	NA
Interventions	NA
Outcomes	NA
Notes	No data found

Roberts 2006.

Methods	NA
Participants	NA
Interventions	NA
Outcomes	NA
Notes	Could not find reference

Schweizer 1993.

Methods	Study design: 7 studies from a systematic literature review and 1 double‐blind, placebo‐controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: pre‐DSM‐III for the systematic review, unclear for the trial Age: not specified Sex: not specified Location: not specified Comorbidities: likely had comorbid depression or panic disorder
Interventions	Participants were randomly assigned to 1 of the following Study 1 1. Imipramine (n = NA) Duration: 8 weeks Treatment protocol: mean dose 143 mg 2. Trazodone (n = NA) Duration: 8 weeks Treatment protocol: mean dose 255 mg 3. Diazepam (n = NA) Duration: 8 weeks Treatment protocol: mean dose 26 mg
Outcomes	Unclear
Notes	This was a summary of a systematic literature review and 1 placebo‐controlled trial. Did not report results that could be extracted.

Shammas 1977.

Methods	NA
Participants	NA
Interventions	NA
Outcomes	NA
Notes	Could not find full‐text

Smith 1994.

Methods	NA
Participants	NA
Interventions	NA
Outcomes	NA
Notes	Could not find full‐text

Stein 2005.

Methods	Study design: 3 multicentre, randomised, 8‐week, double‐blind, placebo‐controlled studies
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age (mean): escitalopram 38.6 (SD 12.6) years; placebo 39.6 (SD 13.2) years Sex (women): escitalopram: 58%; placebo: 54% Location: USA Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of the following Study 1 1. escitalopram (n = 429) Duration: 8 weeks Treatment protocol: flexible dosage (10–20 mg/day). Each trial was initiated with a 1‐week single‐blind placebo lead‐in. Participants were then randomised double‐blind to escitalopram 10 mg/day or placebo. Investigators had the option of increasing the medication dose to escitalopram 20 mg/day at weeks 4 and 6. Participants unable to tolerate escitalopram 20 mg/day could be returned to the starting dose for the remainder of the study. 2. Placebo (n = 427) Duration: 8 weeks Treatment protocol: flexible dosage
Outcomes	Time points of assessment: weeks 1, 2, 4, 6, 8 Primary efficacy measure was the HAM‐A HAMA psychic anxiety subscale (items 1–6, 14) Covi anxiety scale Hospital Anxiety and Depression scale (divided into anxiety and depression subscales) HAM‐D‐17 (total, depression (items 1, 2, 7, 8, 10, 13) and anxiety (items 10, 11, 12, 13, 15, 17) subscales) Raskin Depression scale Quality of life (Quality of Life, Enjoyment, and Satisfaction Questionnaire)
Notes	Study reported results were not provided in full. Could not extract any information.

Wurthmann 1998.

Methods	Study design: randomised, placebo‐controlled, double‐blind, single subject experiment
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐IV Age: not specified Sex: not specified Location: not specified Comorbidities: none stated
Interventions	Participants were randomly assigned to 1 of the following Study 1 1. Amitriptyline (n = NA) Duration: unclear Treatment protocol: fixed dosage (30 mg/day) 2. Flupentixole (n = NA) Duration: unclear Treatment protocol: fixed dosage (1.5 mg/day) 3. Clotiazepam (n = NA) Duration: unclear Treatment protocol: fixed dosage (15 mg/day) 4. Placebo (n = NA) Duration: unclear Treatment protocol: fixed dosage
Outcomes	Not specified
Notes	Study was not a full text, only U tests and meta‐analysis of the results were reported.

Xie 2001.

Methods	NA
Participants	NA
Interventions	NA
Outcomes	NA
Notes	Did not have access to full text

DSM‐IV‐TR: Diagnostic and Statistical Manual of Mental Disorders. 4th, Text Revision edition; DSM‐IV: Diagnostic and Statistical Manual of Mental Disorders. 4th edition; GAD: generalised anxiety disorder; HAM‐A: Hamilton Anxiety Rating Scale; HAM‐D‐17: Hamilton Rating Scale for Depression – 17 item; n: number of participants; NA: not available; SD: standard deviation; SE: standard error; SF‐36: 36‐item Short Form.

Characteristics of ongoing studies [ordered by study ID]

jRCT2031220156.

Study name	A placebo‐controlled, randomized, double‐blind, multicenter study to evaluate the efficacy and safety of venlafaxine in Japanese outpatients with generalized anxiety disorder
Methods	Study design: randomised controlled trial
Participants	Diagnosis: generalised anxiety disorder Method of diagnosis: DSM‐5 Age: still recruiting (minimum age 18 years) Sex: still recruiting Location: Japan Comorbidities: NA
Interventions	Participants were randomly assigned to 1 of 2 groups 1. Venlafaxine Duration: 8 weeks Treatment protocol: flexible dosage (37.5–225 mg/day) 2. Placebo Duration: 8 weeks Treatment protocol: flexible dosage
Outcomes	Primary endpoint: change in the HAM‐A total score from baseline to treatment completion (week 8) or discontinuation.
Starting date	Not specified
Contact information	Madoka Kitano: Viatris_GAD_CL@iqvia.com
Notes

DSM‐5: Diagnostic and Statistical Manual of Mental Disorders 5th edition; HAM‐A: Hamilton Anxiety Rating Scale.

Differences between protocol and review

We made the following changes from the review protocol (Guaiana 2018).

• Relapse prevention studies were excluded to avoid carry‐over effects.

• Vortioxetine and vilazodone were not specified in the protocol but were included because they are antidepressants and we found studies relevant to our review.

• For clarity, we changed the name of the outcomes 'response rate', 'total number of people experiencing side effects', 'dropouts due to inefficacy', and 'dropouts due to side effects' to 'rate of treatment response', 'total number of people experiencing adverse effects', 'dropouts due to a lack of efficacy', and 'dropouts due to adverse effects', respectively.

• We did not search bibliographies or contact experts in the field because we were satisfied with the results of the search.

• We did not report non‐congruence in the selection of trials as percentage disagreement.

• All papers were in English, therefore no translations were required.

• We added the comparison 'other antidepressants versus placebo' for all outcomes as this comparison was deemed clinically important.

• We added analyses of each type of antidepressant compared to placebo.

• We calculated the number needed to treat for an additional beneficial outcome (NNTB) or harmful (NNTH) outcome and their 95% CI using Visual Rx (www.nntonline.net/) while taking into account the event rate in the control group. This was done for the two primary outcomes and for the outcomes 'dropouts due to a lack of efficacy' and 'dropouts due to adverse effects'.

• We extracted change from baseline and endpoint data for continuous outcomes instead of just endpoint data because they can be combined into a meta‐analysis.

• We did not assess the skewness of continuous data because we included change‐from‐baseline measures and the Cochrane Handbook for Systematic Reviews of Interventions suggests that the check for skewness may not be appropriate for change‐from‐baseline measures (Higgins 2022).

• We used the mean difference as the effect measure for continuous outcomes instead of the standardised mean difference because the measurement tool was consistent between the studies.

• Unit‐of‐analysis issues for cluster‐randomised trials were not required as this review included no cluster‐randomised trials.

• We used GetDataGraph Digitizer to extract data from figures where needed.

• Subgroup analyses were not conducted due to an uneven distribution of covariates.

• We did not perform a sensitivity analysis that investigated the influence of possible differences in co‐interventions (such as differential usage of benzodiazepines in antidepressant trials) because studies generally excluded concomitant use of other medications.

• We added sensitivity analysis for 'change in symptom levels' where we removed studies with imputed standard deviations to investigate whether this affected the results. The protocol specified this sensitivity analysis but did not specify the outcome for which it was planned.

• Meta‐regression was not performed due to a lack of studies.

• In the protocol (Background section, 'Why it is important to do this review' subsection) we wrote: "This review is part of a series of Cochrane reviews that assess the pharmacological and psychological treatments in people with GAD (Chessick 2006; Gale 2012; Guaiana 2010). A network meta‐analysis is also planned. This suite of Cochrane reviews will provide current and comprehensive information on pharmacological treatments of GAD." We deleted this sentence from the section and we are no longer planning to prepare a Cochrane network meta‐analysis.

Contributions of authors

KK helped with extraction and was responsible for conducting the analysis and writing of the review.

JA helped with creating Characteristics of excluded studies table.

AP provided suggestions and helped modify the protocol.

SA provided advice on the review methodology in the early stages.

CB provided suggestions and helped modify the protocol.

CC helped with data extraction and quality assessment.

JM led and supervised the project.

GG conceived the idea and wrote the first draft of the protocol; and led and supervised the project.

Sources of support

Internal sources

• Stipend, Other

  Authors were volunteers, paid by their hospital or university, or students.

External sources

• No external source of support, Other

  No external source of support

Declarations of interest

KK: none.

JA: none.

AP: none.

SA: none.

CB: none. CB is a Cochrane Editor but was not involved in the editorial process for this review.

CC: none.

JM: none.

GG: is a Diplomate of the Academy of Cognitive Therapy. GG is a Cochrane Editor but was not involved in the editorial process for this review.

Edited (no change to conclusions)