A systematic review of reviews on the prevalence of anxiety disorders in adult populations

Abstract Background A fragmented research field exists on the prevalence of anxiety disorders. Here, we present the results of a systematic review of reviews on this topic. We included the highest quality studies to inform practice and policy on this issue. Method Using PRISMA methodology, extensive electronic and manual citation searches were performed to identify relevant reviews. Screening, data extraction, and quality assessment were undertaken by two reviewers. Inclusion criteria consisted of systematic reviews or meta‐analyses on the prevalence of anxiety disorders that fulfilled at least half of the AMSTAR quality criteria. Results We identified a total of 48 reviews and described the prevalence of anxiety across population subgroups and settings, as reported by these studies. Despite the high heterogeneity of prevalence estimates across primary studies, there was emerging and compelling evidence of substantial prevalence of anxiety disorders generally (3.8–25%), and particularly in women (5.2–8.7%); young adults (2.5–9.1%); people with chronic diseases (1.4–70%); and individuals from Euro/Anglo cultures (3.8–10.4%) versus individuals from Indo/Asian (2.8%), African (4.4%), Central/Eastern European (3.2%), North African/Middle Eastern (4.9%), and Ibero/Latin cultures (6.2%). Conclusions The prevalence of anxiety disorders is high in population subgroups across the globe. Recent research has expanded its focus to Asian countries, an increasingly greater number of physical and psychiatric conditions, and traumatic events associated with anxiety. Further research on illness trajectories and anxiety levels pre‐ and post‐treatment is needed. Few studies have been conducted in developing and under‐developed parts of the world and have little representation in the global literature.


Introduction
Anxiety disordersdefined by excess worry, hyperarousal, and fear that is counterproductive and debilitatingare some of the most common psychiatric conditions in the Western world (Simpson et al. 2010). The prevalence of anxiety disorders in the United States is estimated to be 18% (Kessler et al. 2005), and their annual cost is reported to be $42.3 billion (Greenberg et al. 1999). In the European Union (EU), over 60 million people are affected by anxiety disorders in a given year, making them the most prevalent psychiatric conditions in the EU (Wittchen et al. 2011). The Global Burden of Disease (GBD) study estimated that anxiety disorders contributed to 26.8 million disability adjusted life years in 2010. (Whiteford et al. 2013). While a number of reviews have focused on the burden of depression and its economic, social, and health care policy implications, substantially fewer have assessed anxiety.
The past decade has seen increased research interest into anxiety disorders, in large part because of a greater recognition of their burden and the implications associated with untreated illness. Clinical reviews have shown that the presence of an anxiety disorder is a risk factor for the development of other anxiety and mood disorders and substance abuse. In clinical and populationbased studies, the development of comorbidities makes the treatment of primary and secondary disorders difficult, contributes to low remission rates, poor prognosis and risk of suicide (Nutt and Ballenger 2003;Simpson et al. 2010). Untreated anxiety has been associated with significant personal and societal costs, related to frequent primary and acute care visits, decreased work productivity, unemployment, and impaired social relationships (Simpson et al. 2010).
A number of primary studies on the prevalence of anxiety have been undertaken, but the variability in findings has made generalizability to the wider population difficult. This variability mainly results from differences in study setting (i.e., culture; clinical vs. population-based), age and sex composition of samples, length of follow-up, methods of anxiety assessment, and caseness criteria (i.e., types and number of disorders examined). Systematic reviews on the prevalence of these conditions in highly select, homogeneous population subgroups have been undertaken, but the selective citation of such estimates presents a distorted view of the overall burden of anxiety and limits generalizability.
The aim of this systematic review of reviews was to provide a comprehensive synthesis and description of the prevalence of anxiety disorders in the general population, as well as in clinical outpatient and inpatient groups affected by a range of chronic physical diseases and psychiatric disorders, as reported by individual reviews. Individuals recruited from the community can have different risk factor profiles than those sampled from clinical settings, potentially giving rise to different rates of mental health problems amongst these groups (Nutt and Ballenger 2003;Simpson et al. 2010). As a result, the burden needs to be assessed across different settings and segments of the population. To provide insight into the demographic groups that are most affected, we reported on estimates for men and women and different age groups, if this information was available. Since a number of studies (Walters et al. 2004;Skapinakis et al. 2005;Simpson et al. 2010) have identified the need to better understand the geographical variation of mental health problems, we included reviews that captured studies conducted across the globe at national and subnational levels. To provide insight into the chronicity of anxiety disorders, we provided period (i.e., 12-month) and lifetime prevalence estimates. If the duration criterion was not clearly stated or the "point" or "current" prevalence was indicated, we simply referred to these estimates as "prevalence".
Findings from this systematic review will shed light on the groups that are most affected by anxiety disorders, and can be used to inform targeted screening and treatment efforts. This will be important in the planning of health services and the development of evidence-based policy. Finally, results from this review can be used to identify areas where further research is needed.
This is the first study to provide a comprehensive synthesis of the disparate findings from systematic reviews undertaken on the burden of anxiety across the globe and using a systematic approach.

Search strategy
We defined a systematic review in accordance with the Cochrane Collaboration and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement (Moher et al. 2009). (Appendix 1) We included high-quality reviews that reported the prevalence of anxiety disorders in the general population or clinic-based settings. We searched for reviews on young, middle-aged, and older adults with risk behaviors (i.e., drug abuse), chronic or infectious diseases, psychiatric conditions, who are vulnerable, and living in countries across the globe. Reviews on the treatment of anxiety were not included, as we consider this to be a separate review topic that would merit an in-depth analysis.
To identify reviews meeting the inclusion criteria, we searched Medline (inception-May, 2015), PsycInfo (1987-May, 2015, and Embase (inception-May, 2015) using combinations of keywords relating to anxiety and prevalence (Appendix 2). Reference lists were hand-searched for additional reviews. Titles and abstracts of non-English language articles were translated to assess relevance. We excluded unpublished data. The review protocol is registered on PROSPERO (Remes et al. 2014).
reports, letters, and commentaries were excluded. Full-text articles were retrieved for further assessment by OR.

Quality assessment
Quality assessment of the reviews meeting the inclusion criteria was undertaken by OR and RvdL. If reviews met at least five of the criteria stipulated by AMSTAR (Shea et al. 2009), a validated measurement tool for assessing the quality of systematic reviews, they were included. For example, some of the AMSTAR quality criteria assess whether an "a priori" design was established, whether there was duplicate study selection and data extraction, if the literature search was comprehensive, whether the quality of primary studies was examined, etc.

Data extraction and analysis
Data extraction was performed by OR and RvdL using the standardized form capturing device: the dates of publication and literature search; objectives; number of studies reviewed; prevalence of anxiety; sample characteristics; sample size range of primary studies; recommendations for future research, and limitations of primary studies and review. Disagreements were resolved through discussion.
Studies were grouped according to five common themes and prevalence was described in the context of: (1) addiction, (2) other mental and neurological disorders, (3) chronic physical diseases, (4) trauma, and (5) vulnerable population subgroups. If there were fewer than three reviews on a chronic physical disease, it was grouped under: "other chronic physical diseases" or "other chronic physical diseases in end-stage". Vulnerable population subgroups refer to individuals at high risk for poor health, who may experience stigma, marginalization, or health service access barriers.
We did not perform a meta-analysis because of the heterogeneity in study methodology. Quantitative measurement of heterogeneity was not undertaken. Finally, a meta-analysis of primary studies included in 48 systematic reviews would not have been feasible. We described the prevalence of individual and aggregate anxiety disorders, subthreshold disorders, or symptoms of anxiety, as reported by the systematic reviews. If reviews provided clear prevalence estimates for men and women and different age groups, we also included this information.

Results
The search identified 1232 reviews on anxiety. After 338 duplicates were removed, titles and abstracts were screened, and the full text of 198 articles was retrieved. In total, 46 systematic reviews met the inclusion criteria ( Fig. 1). Reference searches identified two additional reviews as relevant, yielding a total of 48 reviews in this systematic review (Appendix 3).
Of the 48 reviews, seven focused on the descriptive epidemiology of anxiety disorders, while five reviewed anxiety in relation to addiction. Four focused on mental and neurological disorders. A total of 19 reviews assessed anxiety in the context of chronic physical diseases: most of these focused on CVD (n = 6) and cancer (n = 7), followed by respiratory disease (n = 3) and diabetes (n = 3); the rest examined end-stage physical disease (n = 4), and conditions that have been less commonly studied in the anxiety field (n = 4). Three reviews examined anxiety in the context of trauma, and ten focused on vulnerable population subgroups. Most of the reviews included international studies.

The global distribution of anxiety disorders
Seven reviews focused on the descriptive epidemiology of anxiety disorders, presenting age-, sex-, and time trends. In one international review (Somers et al. 2006), the pooled one-year and lifetime prevalence of total anxiety disorders was estimated to be 10.6% (95% CI: 7.5%, 14.3%) and 16.6% (95% CI: 12.7%, 21.1%), respectively. Given the health care policy and service planning implications of high estimates, a high-quality meta-analysis (Baxter et al. 2014) investigated whether the age-standardized point prevalence of anxiety increased over the last decade. Studies on cultures across the globe were reviewed and findings showed that the prevalence in 1990 (3.8% [95% CI: 3.6%, 4.1%] was very similar to that in 2005 and 2010 (4.0% [95% CI: 3.7%, 4.2%]). A sharp rise in younger people over time was noted, but changing age and population structures were hypothesized to be the drivers of this. Prevalence was found to be lowest in East Asia (2.8% [95% CI: 2.2%, 3.4%]) and highest in North America (7.7% [95% CI: 6.8%, 8.8%]) and the North African/Middle Eastern region (7.7% [95% CI: 6.0%, 10%]) (Baxter et al. 2014). A less rigorous review (Somers et al. 2006) estimated the highest lifetime prevalence of anxiety disorders in Swiss and US populations (23-28.7%), and the lowest in studies on Korea (9.2%). In Pakistan (Mirza and Jenkins 2004), the prevalence of total anxiety ranged from 1.76% to 25%, while a meta-analysis on Germany (Vehling et al. 2012) reported it to be 13.5% (95% CI: 7.1%, 24.3%).
Women are almost twice as likely to be affected as men (female:male ratio of 1.9:1), with sex differences persisting over time and across high and low resource settings (Somers et al. 2006;Baxter et al. 2013;Steel et al. 2014). Irrespective of culture, individuals under the age of 35 years are disproportionately affected by anxiety disorders (Baxter et al. 2013(Baxter et al. , 2014 with the exception of Pakistan, where midlife represents a period of high burden (Mirza and Jenkins 2004).

Addiction
Five reviews focused on anxiety experienced in relation to addictive behaviors, including substance misuse, pathological gambling, and compulsive internet use. A global review on nonmedical prescription opioid use (NMPOU) reported the overall lifetime anxiety prevalence in patients at admissions or in treatment for substance abuse problems to range from 2% to 67% (Fatseas et al. 2010). While the prevalence of anxiety diagnoses is reportedly high at 29% (95% CI: 14%, 44%), that of subthreshold anxiety is higher still, with half of NMPOU populations enrolled in substance abuse treatment in North America reporting symptoms (50% [95% CI: 16%, 84%]) (Goldner et al. 2014). In contrast, general population samples of NMPOU in North America show a substantially lower prevalence of anxiety (16% [95% CI: 1%, 30%]) (Fischer et al. 2012). No significant age or sex-effects were found in NMPOU groups enrolled in substance use treatment (Goldner et al. 2014).
Two other risk behaviors that have received attention in the addiction field include problem and pathological gambling, and more recently, internet addiction. When a global meta-analysis assessed 11 community samples of pathological gamblers, the prevalence of anxiety disorders was reported to be 37. 4% (Lorains et al. 2011). The prevalence of anxiety in the context of internet addiction is lower and comes mostly from studies conducted in Asian countries. A meta-analysis found the prevalence of anxiety to be over two times higher in community samples of people with Internet addiction compared to control subjects (

Other mental and neurological disorders
In Europe, approximately 13-28% of people with bipolar disorder recruited from clinical and community settings have comorbid anxiety, with GAD and panic disorder being frequently experienced by this population (Fajutrao et al. 2009). In the US and Italian samples with bipolar disorder (Amerio et al. 2014), OCD is also common. The prevalence of this anxiety disorder in those who are bipolar has been shown to range from 11.1% to 21% in population-based studies, and 1.8% to 35.1% in clinical samples.
OCD is also highly comorbid with schizophrenia. A global review (Swets et al. 2014) estimated the prevalence of this disorder in people diagnosed with schizophrenia to be 12.3% (95% CI: 9.7%, 15.4%). The prevalence of obsessive compulsive symptoms (OCS) not meeting full caseness criteria was over twice that of OCD (30.7% [95% CI: 23%, 39.6%]). Lower anxiety prevalence was linked to sub-Saharan African origin. Age and sex did not influence OCD or OCS rates (Swets et al. 2014). These estimates were mainly based on groups from clinical settings.
One of the highest prevalence figures of psychopathology was found by a review on multiple sclerosis (MS) (Marrie et al. 2015), which reported that almost 32% of people with MS have an anxiety disorder and over half experience symptoms. Some of the primary studies included in this review were based on participants recruited from the general population, suggesting that men and women with MS are at high risk for psychopathology. Health anxiety may be an important issue in this population subgroup, given that 26.4% of those with MS are affected. Study methodology made a significant contribution to the figures reported. Estimates of anxiety prevalence were substantially higher if they were derived through self-reported questionnaires (

Cardiovascular disease
Six reviews reported the prevalence of anxiety in the context of cardiovascular disease (CVD). Approximately a tenth of patients with cardiovascular disease and living in Western countries are affected by GAD (10.94% [95% CI: 7.8%, 14.0%] (Tully and Cosh 2013), with women showing higher anxiety levels than men (Clarke and Currie 2009). Anxiety symptom prevalence among patients with congestive heart failure is 2- 49% (Janssen et al. 2008), and in end-stage patients suffering from heart disease, it is 49% (Solano et al. 2006). Further, panic disorder is a common diagnosis in patients with coronary artery disease, with the prevalence ranging from 10% to 50% in this subgroup (Clarke and Currie 2009).
Individuals with noncardiac or nonspecific chest pain presenting to emergency departments, particularly women and those who are younger, appear to be disproportionately affected by anxiety. Compared to those with a determined cause of chest pain, anxiety prevalence was found to be higher in those with unknown etiology (21-53.5% of noncardiac chest pain patients have probable anxiety) (Webster et al. 2012).
A high-quality, global meta-analysis of population-, hospital-, and rehabilitation-based studies found the prevalence of anxiety disorders in stroke patients to vary between 18% (95% CI: 8%, 29%) and 25% (95% CI: 21%, 28%) when measured by clinical interview and rating scales, respectively (Campbell Burton et al. 2013). Age and sex did not influence the probability of having anxiety after stroke in most of the included studies. GAD and phobic disorders were the commonest anxiety disorders post-stroke.

Cancer
Seven reviews assessed anxiety among individuals diagnosed with or receiving treatment for cancer and in spouses of cancer patients. The prevalence of anxiety among cancer patients varies between 15% and 23%, with symptoms rising to 69-79% in the later stages of disease. There was no reported evidence with respect to age and sex (Solano et al. 2006;Clarke and Currie 2009 prevalence of anxiety was comparatively lower than in Chinese patients, ranging from 28.0% to 33.0% (Vehling et al. 2012).
Randomized controlled trials (RCT) and non-RCT studies conducted across the globe showed that approximately a fourth to over half of individuals undergoing or who had undergone breast cancer treatment experienced anxiety (Lim et al. 2011). Lower levels of anxiety were observed in patients undergoing radiotherapy rather than chemotherapy. During chemotherapy, young age and high trait anxiety measured before infusions were correlated with the intensity of anxiety experienced (Lim et al. 2011). Among ovarian cancer patients, younger age groups were also disproportionately affected by anxiety. Following treatment for ovarian cancer, psychopathology tended to persist, with almost half (47%) of individuals experiencing anxiety symptoms at three months posttreatment (Arden-Close et al. 2008).
Long-term cancer survivors and their spouses also experience elevated levels of anxiety. In a global metaanalysis of outpatient clinic, hospital, and populationbased samples (Mitchell et al. 2013), the prevalence of anxiety in individuals who had been diagnosed with cancer at least 2 years previously was found to be much higher than in healthy controls (17.9% [95% CI: 12.8%, 23.6%] and 13.9% [95% CI: 9.8%, 18.5%], respectively). Further, almost half (40.1% [95% CI: 25.4%, 55.9%]) of spouses of long-term cancer survivors developed anxiety. No age or sex effects were reported.

Respiratory disease
Three reviews focusing on anxiety in the context of respiratory disease indicated that the prevalence of anxiety was high among adults with COPD (32-57%) (Janssen et al. 2008), and higher still among those with far-advanced, end-stage respiratory disease (51-75%) (Solano et al. 2006). Among acute lung injury/acute respiratory distress syndrome (ALI/ARDS) survivors discharged from intensive care units in the United States and Germany, anxiety levels ranged from 23% to 48% (Davydow et al. 2008). No age or sex effects were reported.

Diabetes
Three systematic reviews assessed anxiety in adults with diabetes. One high-quality global review of mostly North American and European studies (Smith et al. 2013) showed that the prevalence is significantly elevated in those with diabetes compared to other groups, but is also dependant on caseness criteria. Approximately 15% to 73% of people with diabetes have anxiety symptoms not meeting threshold criteria (vs. 19.9% to 43.1% of nondiabetic individuals), while 1.4% to 15.6% of people with diabetes meet threshold criteria for an anxiety disorder (vs. 1.6% to 8.8% of nondiabetic individuals). In another review capturing studies predominantly conducted in primary care or clinical settings, women with diabetes were found to have an almost two-fold higher prevalence of anxiety than men with diabetes (55.3% and 32.9%) (Grigsby et al. 2002). Age effects were not reported. The anxiety disorders that are most common in the context of diabetes are anxiety not otherwise specified, specific phobia, GAD, and social phobia (Grigsby et al. 2002;Clarke and Currie 2009).

Other chronic physical diseases
Four reviews assessed anxiety in population subgroups with polycystic ovary syndrome (PCOS), benign joint hypermobility syndrome, musculoskeletal pain, and agerelated macular degeneration. Clinical, mostly Western samples of women with polycystic ovary syndrome (PCOS) had a much higher prevalence of generalized anxiety symptoms than control groups (20.4% and 3.9%, respectively) (Dokras et al. 2012). There is some evidence that social phobia and OCD are comorbid with PCOS. Differences in anxiety levels according to age were not found (Dokras et al. 2012).
Widely varying anxiety prevalence figures have been reported for Mediterranean populations with benign joint hypermobility syndrome (BJHS) (5-68%) (Smith et al. 2014), as well as for Western populations with musculoskeletal pain (0-20.9%) (Andersen et al. 2014). In relation to the latter group, the link between fibromyalgia and anxiety appears to be particularly strong. In people with BJHS, commonly occurring comorbidities are agoraphobia and panic disorder (Smith et al. 2014). The only chronic condition that has failed to show a link with anxiety is age-related macular degeneration; while this review recruited patients from clinics, it was largely based on US studies (Dawson et al. 2014).

Other chronic physical diseases in end-stage
Four reviews assessed anxiety in end-stage conditions. A global meta-analysis of mostly Western studies (Mitchell et al. 2011) estimated the pooled prevalence of anxiety disorders in palliative cancer patients to be 9.8% (95% CI: 6.8%, 13.2%). Estimates appear to vary widely by condition. Among patients with chronic renal failure, the prevalence of anxiety symptoms was found to be 25% in the terminal stage (Janssen et al. 2008), whereas another review found a prevalence of 38% in patients with endstage renal disease (Murtagh et al. 2007). Although patients suffering from end-stage AIDS showed a high symptom prevalence of 8-34%, the highest estimates were found for end-stage COPD (51-75%) and cancer patients (13-79%) (Solano et al. 2006). No associations between age or sex and anxiety were found in palliative-care settings (Mitchell et al. 2011).

Trauma
Three reviews tackled the issue of anxiety in the context of trauma. The first was primarily based on findings from UK and US studies and focused on traumatic limb amputees, and included veterans that had served in Vietnam, Iraq and Afghanistan (Mckechnie and John 2014). Very high prevalence figures were found, with anxiety affecting a fourth of traumatic limb amputees in some studies to over half in others. The second review was global in scope and assessed the frequency of lifetime anxiety among individuals with a history of sexual abuse (Chen et al. 2010). Widely varying anxiety estimates were reported by this review, ranging from 2% to 82%. Finally, a third review focused on GAD in refugees resident in high-income Western countries; over half of the refugees were from southeast Asia. This meta-analysis estimated that 4% of refugees experience GAD (Fazel et al. 2005). No age or sex effects in relation to anxiety disorders were reported.

Older people and their caregivers
Five reviews assessed anxiety in older people and their caregivers. The prevalence of anxiety disorders in old age varies widely in community (1.2-14%) and clinical (1-28%) studies conducted mostly in European and North American settings. Estimates are even higher when anxiety symptoms are accounted for. GAD is the commonest anxiety disorder in old age, with the prevalence ranging from 1.3% to 4.7% (Bryant et al. 2008). A random-effects model (Volkert et al. 2013) showed that specific phobia also occurs frequently in older samples living in the community, while agoraphobia is the rarest anxiety disorder (Bryant et al. 2008). Women are at higher risk for psychopathology than men (Bryant et al. 2008).
Older population subgroups with cognitive dysfunction and their caregivers are disproportionately affected by anxiety (Monastero et al. 2009). In older people with mild cognitive impairment (MCI), the prevalence of anxiety symptoms varies from 11% to 75% (Monastero et al. 2009;Yates et al. 2013). Caregivers of older people with cognitive impairment are also affected by anxiety (prevalence estimates of 3.7-76.5%), with women and younger caregivers showing elevated levels (Cooper et al. 2007;Bryant et al. 2008).

Pregnant women
Three reviews focused on pregnant women. A meta-analysis of international studies (Russell et al. 2013) reported higher OCD prevalence in pregnant (2.07%, [95% CI: 1.26%, 3.37%]) and postpartum (up to 12 months) (2.43%, [95% CI: 1.46%, 4.00%]) women compared to the general population (1.08%, [95%: 0.80%, 1.46%]). Asia and Europe had the lowest prevalence of OCD across conditions, while the Middle East and Africa had the highest. In Ethiopian and Nigerian samples recruited from health clinics and the community (Sawyer et al. 2010), the prevalence of anxiety was found to be high during both the pre-and post-natal periods (14.8% [95% CI: 12.3%, 17.4%] and 14.0% [95% CI: 12.9%, 15.2%], respectively), with younger women showing elevated anxiety compared to older women (Sawyer et al. 2010). There is also some evidence from UK and US studies that a high BMI may contribute to anxiety symptoms during pregnancy (Molyneaux et al. 2014).

Individuals identifying as lesbian, gay or bisexual, and self-harm patients
Two reviews focused on (1) predominantly Western individuals living in the community and identifying as lesbian, gay or bisexual (LGB), and (2) self-harm patients presenting to general hospitals in countries across the globe. In LBG men, anxiety prevalence was estimated to be 3-20%, while LGB women showed somewhat higher estimates, at 3-39% (King et al. 2008). In a global metaanalysis of self-harm patients presenting to hospitals, the prevalence of anxiety disorders was found to be 35% (95% CI: 21.9%, 48.6%). Age-and sex-based differences were small, while rates of anxiety were highest in young and old age groups of self-harm adults (Hawton et al. 2013). All non-Western studies of self-harm patients were based in Asia, while most of the Western studies were conducted in the United Kingdom.

Discussion
We have synthesized 48 reviews on prevalence studies conducted across the globe. This is the first review to undertake a comprehensive synthesis of the systematic reviews conducted to date on the prevalence of anxiety disorders. It provides a comprehensive, up-to-date summary of the state of knowledge in this area.
Compared to healthy populations or control groups, prevalence was higher in individuals with chronic physical diseases (Mitchell et al. 2013;Yang et al. 2013), and the burden was particularly high in the end stage (Solano et al. 2006;Murtagh et al. 2007;Mitchell et al. 2011). Anxiety symptoms tended to persist post-disease if present before disease onset (Sawyer et al. 2010), reflecting a chronic, unremitting pattern of psychopathology. Individuals exposed to trauma or who were vulnerable and at risk for stigma (Cooper et al. 2007;Bryant et al. 2008;King et al. 2008;Monastero et al. 2009;Sawyer et al. 2010;Hawton et al. 2013;Russell et al. 2013;Volkert et al. 2013;Yates et al. 2013;Molyneaux et al. 2014), such as older people with cognitive impairment (Yates et al. 2013), were also more likely to experience anxiety. Prevalence figures were heterogeneous, and this made comparison between studies difficult. Heterogeneity was driven by differences in caseness criteria and sampling methods. For example, a metaregression (Swets et al. 2014) that assessed the influence of instrument differences on OCD prevalence in the context of schizophrenia showed that the prevalence was higher with the use of the Yale-Brown Obsessive Compulsive Scale (YBOCS)/Obsessive Compulsive Inventory (OCI) (Goodman et al. 1989;Foa et al. 1998) compared to other instruments. Also, the lower the threshold of the YBOCS, the higher the estimated prevalence. A range of methods was used to measure anxiety, such as, standardized, structured interviews administered by trained professionals, clinician diagnoses, symptom checklists, and self-report. Some reviews attempted to handle the assessment of anxiety in alternative ways. For example, one review (Baxter et al. 2013) mapped estimates onto ICD or DSM diagnostic criteria and conducted a meta-analysis to provide an aggregate measure of anxiety. Other reviews either did not attempt a meta-analysis, or because of very large differences in sampling methods within primary studies, reported disaggregated estimates and ranges found in primary studies. Across reviews, higher prevalence figures were found when subthreshold disorders or symptoms were assessed (Bryant et al. 2008;Goldner et al. 2014;Haller et al. 2014;Swets et al. 2014;Marrie et al. 2015) and when lifetime rather than past-year or current prevalence was estimated (Somers et al. 2006;Volkert et al. 2013).
With the exception of one review (Monastero et al. 2009), authors did not account for the use of psychoactive prescription medicines, such as anxiolytics, which could influence the reporting of anxiety symptoms.
Reviews produced inflated prevalence estimates with the use of less robust methodologies. Within reviews, low and variable response rates across primary studies were identified as another limitation (Somers et al. 2006;Haller et al. 2014). In one review, response rates across studies ranged from 45.9% to 99.5% (Steel et al. 2014).
The areas that received the most attention in the anxiety field include addiction and chronic physical diseases (mainly cancer, CVD, and respiratory diseases), while anxiety disorders other than PTSD in the context of (1) trauma and (2) psychiatric or neurological conditions, such as internet addiction and multiple sclerosis, are relatively new and underresearched areas. Surprisingly, only one review (King et al. 2008) examined LGB groups, despite this population being at high risk for poor health (Fredriksen-Goldsen et al. 2013). Authors of this review (King et al. 2008) called for further research to produce more refined and consistent definitions of LGB and the recruitment of more representative samples.
Although most of the reviews included in this systematic review were conducted in the last few years, the field of anxiety is rapidly gaining research interest. Some differences in findings and methodologies between older and more recent reviews were noted. For example, recent reviews are increasingly recognizing that early adulthood is the period with the highest peak in anxiety (Arden-Close et al. 2008;Sawyer et al. 2010;Lim et al. 2011;Webster et al. 2012;Hawton et al. 2013;Yates et al. 2013;Baxter et al. 2014;Haller et al. 2014;Ho et al. 2014;Baxter et al. 2013), and the contexts within which psychopathology is assessed are expanding to a greater number of physical diseases and newly emergent disorders [e.g. internet addiction (Ho et al. 2014)]. Also, newer research is starting to expand its scope to Asian countries (Yang et al. 2013;Ho et al. 2014), a previously identified limitation. More recent reviews are of higher quality, and have started considering instrument differences and their effects on prevalence estimates (Lorains et al. 2011;Swets et al. 2014), another previously identified limitation.

Recommendations for future research and clinical practice
Recommendations for future research were made by review authors, such as the use of longitudinal designs to address temporality issues (Murtagh et al. 2007;Arden-Close et al. 2008;Bryant et al. 2008;Janssen et al. 2008;King et al. 2008;Clarke and Currie 2009;Sawyer et al. 2010;Dokras et al. 2012;Fischer et al. 2012;Webster et al. 2012;Russell et al. 2013;Smith et al. 2013;Goldner et al. 2014;Ho et al. 2014; Mckechnie and John 2014); population-based research that is less susceptible to the help-seeking/self-selection bias often present in clinical studies (Grigsby et al. 2002;Murtagh et al. 2007); and the use of valid and reliable instruments and consistent approaches to examine anxiety levels pre-and post-disease (Davydow et al. 2008;Monastero et al. 2009;Sawyer et al. 2010;Webster et al. 2012;Campbell Burton et al. 2013;Smith et al. 2013;Volkert et al. 2013;Goldner et al. 2014;Molyneaux et al. 2014;Swets et al. 2014;Marrie et al. 2015). The measure of "total" or "any anxiety" is not clinically meaningful and is discouraged in favor of the assessment of individual disorders (Smith et al. 2013;Tully and Cosh 2013). Consensus on definitions used to define study samples (e.g., sexual orientation) (King et al. 2008;Fischer et al. 2012;Ho et al. 2014) and diagnostic standardization with respect to the measurement of psychiatric disorders were also emphasized (Monastero et al. 2009;Goldner et al. 2014;Swets et al. 2014), as well as research into the risk factors, illness trajectory, hereditary, and biological markers of anxiety (Somers et al. 2006;Davydow et al. 2008;Monastero et al. 2009;Chen et al. 2010;Dokras et al. 2012;Russell et al. 2013;Amerio et al. 2014;Ho et al. 2014;Smith et al. 2014), and the appropriateness of anxiety screening measures in the context of physical diseases and cultures around the world (who may express distress differently) (Fazel et al. 2005;Bryant et al. 2008;Sawyer et al. 2010;Baxter et al. 2013;Hawton et al. 2013;Steel et al. 2014). Research questions should be structured around theories (Arden-Close et al. 2008;Webster et al. 2012). Recommendations were made for the inclusion of appropriate control subjects in studies to determine whether prevalence differs between exposed and comparison groups (Yang et al. 2013;Dawson et al. 2014). Finally, further treatment or intervention studies are needed to alleviate anxiety (Mirza and Jenkins 2004;Murtagh et al. 2007;Arden-Close et al. 2008;Clarke and Currie 2009;Fatseas et al. 2010;Lim et al. 2011;Amerio et al. 2014;Goldner et al. 2014;Haller et al. 2014;Ho et al. 2014;Smith et al. 2014;Swets et al. 2014).
Clinical recommendations included the administration of targeted anxiety screening and, if necessary, treatment. For example, suggestions were made for the screening of substance users at treatment entry (Fatseas et al. 2010) or patients with noncardiac chest pain presenting to acute care (Webster et al. 2012). It was also shown that certain anxiety disorders were more common in certain groups, such as OCD in schizophrenia (Swets et al. 2014), PD andGAD in CVD (Campbell Burton et al. 2013), and SP in diabetes (Grigsby et al. 2002). Additional research on individual anxiety disorders is needed to confirm these findings, but once this is underway, further impetus will be provided for the targeted screening of high-risk groups in relation to individual anxiety disorders.
This review has some limitations. Despite extensive database searches, it is possible that some reviews have been missed. Also, the high heterogeneity in anxiety assessment methods and sampling frameworks within primary studies contributed to large differences in prevalence estimates within and across reviews, making it difficult to draw conclusions about the burden of anxiety. Also, a number of the reviews were based on English-language studies conducted in predominantly Western settings, making generalizability to other parts of the world difficult.

Conclusions
Anxiety disorders are increasingly being recognized as important determinants of poor health and major contributors to health service use across the globe (Nutt and Ballenger 2003;Simpson et al. 2010). Despite epidemiologic advances in this field, important areas of research remain under-or unexplored. There is a need for further studies on the prevalence of anxiety disorders in the context of: personality disorders; Indigenous cultures in Canada, the United States, New Zealand, and Australia; African, Middle Eastern, Eastern European, Asian and South American countries; and marginalized populations, such as injection drug users, street youth, and sex workers. These recommendations can serve to guide the research agenda, and most importantly, help develop tailored and timely interventions. PsycInfo 1 exp Meta Analysis/ 2 meta analy$.tw. 3 metaanaly$.tw. 4 (systematic adj -n -(review$1 or overview$1)).tw. 5 exp "Literature Review"/ 6 or/1-5 7 cochrane.ab. 8 embase.ab. 9 (psychlit or psyclit).ab. 10 (cinahl or cinhal