Table A2.

Directions for future research and reported limitations

Review details	Directions for future research	Reported limitations	QAa
Global distribution of anxiety disorders
Somers 2006 Search: 2004 # incl. studies 39 Meta‐analysis: yes	Incidence and onset studies needed Research on anxiety risk & protective factors, and social variables as mediators Prev of anxiety in special groups (e.g., medical patients, residents of nursing homes) Clarify epidemiology of anxiety to help with deployment of treatment	Original studies Heterogeneity: diagnosis criteria and instruments used (ex. lower estimates with use of DIS and DSM‐III than CIDI and DSM‐III‐R) Review Heterogeneity: diff countries, response rate, sample size	5
Baxter 2013 Search: 2009 # incl. studies 87 Meta‐analysis: yes	Further research on: Impact of conflict on mental health Aspects of wealth related to anxiety Cultural aspects (ex. psycho‐stressors) related to anxiety Further studies using consistent anxiety definition and methodologies in 1) developing and emerging countries; 2) populations exposed to conflict Interactions of factors associated with prevalence of anxiety	Original studies Limited measurement equivalence across cultures – results should be interpreted with caution Rural study results – should be interpreted with caution Study design differences Review NR	10
Mirza 2004 Search: March 2002 # incl. studies: 20 Meta‐analysis: no	Robust evidence (ex. conduct national, mental health epidemiology surveys) to develop mental health policy with strategic implementation plan for Pakistan More outcome studies, prevention and treatment trials needed	Original studies Most studies from Punjab and Sind Heterogeneity in study design and instruments – limited generalizability Review Publication and selection bias Small number of included studies	5
Vehling 2012 Search: not rep. # incl. studies 89 Meta‐analysis: yes	Representative studies	Original studies Estimate heterogeneity and study quality Limited generalizability Review	7
Baxter 2014 Search: 2009 # incl. studies 91 Meta‐analysis: yes		Original studies Limited or no data from Central Asia, Andean Latin America, Oceania, Central sub‐Saharan Africa, Central Europe, South‐east Asia Possibly biased population samples (ex. conflict region studies may have oversampled those exposed to conflict) Review NR	10
Haller 2014 Search: 2006 # incl. studies: 18 Meta‐analysis: no	Clarify subthreshold GAD vs. nonpathological anxiety – use impairment criterion for this Should treatment strategies used for threshold disorders be used for subthreshold cases?	Original studies Inadequate study response rates Heterogeneous definitions of subthreshold GAD Review Some studies missed Difficult to define search terms for subthreshold GAD Insufficient studies for subpopulations Different study quality	7
Steel 2014 Search: Jan 2014 # incl. studies 174 Meta‐analysis: yes		Original studies Some recall bias with 12‐month estimates Different study age structures contributing to different prev Higher prev with smaller sample sizes Different estimates with the use of different instruments Adaptation of surveys to culture and context & measurement equivalence issues Review Some studies may have been missed Untested search strategies Assessment equivalence across cultures Can only generalize findings to adults	5
Addiction
Fatseas 2010 Search: Jan. 2009 # incl. studies 18 Meta‐analysis: no	Effectiveness of treatment for phobias in opiate‐dependent patients	Original studies Reliability and validity of diagnostic tools (ex. difficult to distinguish substance‐induced anxiety from independent disorders with pre‐DSM‐IV criteria) Heterogeneity in sample characteristics Different time frames for prev of anxiety Review	6
Fischer 2012 Search: Dec. 2011 # incl. studies 9 Meta‐analysis: yes	Longitudinal studies to assess reasons for using NMPOU in individuals with mental health problems	Original studies Heterogeneity: operationalization of anxiety and NMPOU Many screener or epidemiological instruments used (possible overestimation), instead of clinical diagnostic tools All North American studies – limited generalizability Small number of studies Review Between‐study heterogeneity	8
Goldner 2014 Search: April 2012 # incl. studies 11 Meta‐analysis: yes	Relationship between NMPOU and mental illness Retrospective and prospective studies to examine development of mental health problems and NMPOU in those receiving POAs Use standardized and comparable diagnostic instruments Link between chronic pain and mental illness Alternative treatments for and outcomes of patients with both mental health problems and NMPOU	Original studies Cross‐sectional data, thus temporality issues between NMPOU and mental illness Diff instruments used Review Publication bias High between‐study differences Heterogeneity: defining and measuring NMPOU psychiatric problems	8
Lorains 2011 Search: Sept. 2010 # incl. studies 11 Meta‐analysis: yes	Health care workers should: Assess for comorbidities Determine whether anxiety developed before gambling problem and should be treated first	Original studies Lifetime estimates may be confounded by age Diff tools (ex. SOGS – satisfactory psychometrics in populations surveys; discordance between NODS and DSM‐IV) Most general population prevalence surveys conducted in the United States and Canada, small sample sizes Review NR	5
Ho 2014 Search: 2012 # incl. studies 8 Meta‐analysis: yes	Genetic transmission of IA Patients with IA should be screened for anxiety and vice versa & integrated treatment recommended Further studies on moderators; other ethnic groups in Europe and North America; older adults Studies on interactions between IA and anxiety (etiology, illness trajectory, treatment outcomes) Consensus on definition of IA Prospective studies Link between anxiety and IA‐specific behaviors (ex. use of social media)	Original Heterogeneity: age of sample, different psychiatric questionnaires, mostly cross‐sectional studies, uncontrolled confounding (ex. environmental stress, parenting) Young patients mainly from Asian countries Review Small number of studies Unable to assess how estimates differ with use of self‐reported questionnaires vs. structured interviews	8
Other mental and neurological disorders
Fajutrao 2009 Search: past 10 years # incl. studies 26 Meta‐analysis: no	Bipolar disorder in Europe	Original studies Anxiety assessment and reporting methods diff Retrospective and nonrepresentative samples Review Focus on electronic databases; language selection criteria	5
Amerio 2014 Search: Mar 2013 # incl. studies: 64 Meta‐analysis: no	Assess history of mood disorders in OCD patients Treatment research (ex. use of mood stabilizers) Studies on hereditary and biological markers, diagnostic validity of BD‐OCD comorbidity and its treatments	Original studies Differences in evaluation, diagnosis, reporting Mostly observational, retrospective studies, lack of control group, small sample size, sampling bias Review NR	5
Swets 2014 Search: Dec 2009 # incl. studies 43 Meta‐analysis: yes	Use random sampling Training needed to assess OCS Diagnostic standardization needed, careful patient selection Detailed assessment of OCD; use SCID OCD def. followed by Y‐BOCS administration Assess OCS in patients with psychosis Shift from descriptive to treatment studies	Original studies Different instruments and criteria used (ex. lower estimates with DSM‐III‐R than later versions; lower prev with DIGS) Sampling variability (different patient characteristics) Possible sampling bias, help‐seeking/patients selection can influence prev rates Limited data on: sub‐Sahara African countries, gender, ethnicity, use of meds (ex. antipsychotics) Review NR	5
Marrie 2015 Search: Nov. 2013 # incl. studies 118 Meta‐analysis: yes	Be consistent: compare psychometric properties of instruments and use same instrument to assess anxiety Standardize estimates to common (world) population	Original Differences in study design: different data sources, populations, definitions of psychiatric disorders Little info on age‐, sex‐, or ethnicity‐specific estimates Review NR	5
Chronic physical diseases
Cardiovascular disease
Janssen 2008 Search: 2007 # incl. studies 39 Meta‐analysis: no	Prospective research that considers view of patients, their families, their physician for symptom management	Original studies Different rates of symptom reporting with different proxies and depending on timing of interview Differences in: methods of reporting; definition of end‐of‐life (ex. different estimates in last week vs. last year of life); patient characteristics; definition and measurement of symptoms Review NR	5
Solano 2006 Search: June 2004 # incl. studies 64 Meta‐analysis: no		Original studies Heterogeneity in definition of symptoms (different criteria), methods to detect cases of symptoms (different questionnaires and screening methods used), study design, sampling, study setting, methods of data collection	5
Tully 2013 Search: May 2011 # incl. studies 12 Meta‐analysis: yes	Further GAD research in CHD Specific anxiety disorders rather than trait/state anxiety “Any anxiety” not clinically informative in cardiac settings	Original studies Heterogeneity: diagnostic criteria for GAD, gender ratio, patient age Review Low rate of publications on GAD	6
Clarke 2009 Search: May 2003 # incl. studies 159 Meta‐analysis: no	Effectiveness of interventions Large prospective studies Anxiety assessed in parallel with chronic conditions	Original studies Different rating tools & diagnostic criteria; low power Review Heterogeneity	7
Webster 2012 Search: Nov. 2010 # incl. studies 12 Meta‐analysis: no	Theory‐driven research to examine link between patients’ perceptions (ex. chest pain) and mental health Does providing explanations for patients’ chest pain reduce their anxiety? Longitudinal design to assess mental health trajectory in NCCP Use reliable and valid measures for mental disorders with recommended cut‐offs	Original studies Different caseness cut‐offs Risk factor research used cross‐sectional designs Few studies on correlates of poor mental health in NCCP Review Possible publication bias Large heterogeneity in study settings	5
Campbell Burton 2013 Search: March 2011 # incl. studies 44 Meta‐analysis: yes	Mood assessment tools appropriate for stroke patients Guidance on best time to screen for anxiety What is the impact of anxiety and its economic burden in the context of stroke?	Original studies Different cut‐off scores used Most studies cross‐sectional, so difficult to determine whether pre‐stroke anxiety is linked to post‐stroke anxiety Few studies differentiated btw. “first‐ever” and “current anxiety” Some scales were not validated in stroke populations Review Potential publication bias & heterogeneity Some studies not included in review	10
Cancer
Clarke 2009 – previously described
Solano 2006 – previously described
Yang 2013 Search: Sep. 2012 # incl. studies 17 Meta‐analysis: yes	Use control groups with diseases other than cancer	Original studies Anxiety assessed using different instruments Studies were cross‐sectional so cannot determine temporality between anxiety and cancer development Review Few studies & lacking international literature Potential publication bias	9
Vehling 2012 – previously described
Lim 2011 Search: 2010 # incl. studies 10 Meta‐analysis: no	Studies in different settings assessing effect of cancer treatment on anxiety Interventions for anxiety in women with breast cancer Ways to decrease state anxiety and help women cope with chemotherapy, despite their level of trait anxiety	Original studies Small sample sizes Review Difference in treatment, tools & timing of measurement	6
Arden‐Close 2008 Search: May 2007 # incl. studies 18 Meta‐analysis: no	Longitudinal studies and RCTs needed to clarify directionality between immunity and mental illness Prospective research needed to test trajectories of change in mental illness following cancer diagnosis and treatment Interventions targeting distress (ex. coping) Attention to sample size and validation of questionnaires Theory‐driven research needed Authors should state limitations/directions for future research	Original studies Certain correlates of mental illness tested in too few studies Lack of validation of assessment tools Small sample sizes Residual confounding Limited generalisability (US) Review Published studies	6
Mitchell 2013 Search: March 2013 # incl. studies 43 Meta‐analysis: yes	Link between health‐related quality of life and anxiety Studies on anxiety in palliative settings or in patients with advanced cancer More reliable estimates by use of interview methods	Original studies Differences in: quality of matching with healthy controls, study quality, study design, case ascertainment Possible uncontrolled factors Heterogeneity in healthy controls (review authors had limited info on recruitment of healthy controls in studies) Review – NR	11
Respiratory disease
Janssen 2008 – previously described
Solano 2006 – previously described
Davydow 2008 Search: April 2007 # incl. studies 10 Meta‐analysis: no	Risk factors for psychopathology More rigorous assessment of psychopathology Anxiety in ICU as risk factor for post‐ALI/ARDS psychopathology To what extent are risk factors for ALI/ARDS related to development of mental illness in those without ALI/ARDS	Original studies Mostly psychiatric questionnaires used with diff. sensitivities, ex. screening instruments or measures of symptom severity (not necessarily validated for ARDS survivors) Small sample sizes Review Small number of studies	5
Diabetes
Smith 2013 Search: July 2012 # incl. studies 12 Meta‐analysis: yes	Individual anxiety disorders associated with diabetes Relevant confounders should be included Studies on diabetes and anxiety using accurate measurements Prospective studies to clarify directionality between anxiety and diabetes	Original studies Different time frames resulting in different likelihood of capturing symptoms Measurement differences Cross‐sectional data Temporality between diabetes and anxiety Review Publication bias, language biases	10
Grigsby 2002 Search: 2001 # incl. studies 18 Meta‐analysis: yes	Longitudinal studies to identify behavioral and physiological mechanisms related to anxiety in diabetes More community‐based studies to estimate anxiety prev in diabetes Assess potential moderators Studies on causal mechanisms	Original studies Small sample sizes Lacking data on race/ethnicity influence on anxiety prev Differences in scales used to measure anxiety and in aggregation/reporting of results (ex. assessment of 1 anxiety disorder vs. aggregate of several anxiety disorders) Lack of data on prev of anxiety by diabetes type Review Small number of studies Few studies included nondiabetic comparison group	6
Clarke 2009 – previously described
Other chronic physical diseases
Dokras 2012 Search: April 2011 # incl. studies 9 Meta‐analysis: yes	Effect of clinical or biochemical factors in relation to hyperandrogenism and anxiety in PCOS Link between PCOS‐specific characteristics and anxiety Larger sample sizes Longitudinal studies for insight into etiology and trajectory of anxiety in PCOS	Original studies Few studies on prev on anxiety in PCOS using validated anxiety screening tools Mostly cross‐sectional studies Review Small sample sizes, possible publication bias	5
Smith 2014 Search: January 2013 # incl. studies 14 Meta‐analysis: yes	Degree of BJHS related to mental illness Biological link between BJHS and anxiety (ex. abnormal reactive autonomic nervous system) Influence of nonpharmacologic treatment on alleviating anxiety in those with BJHS Anxiety in BJHS in other cultures	Original studies Limited generalizability (mainly Mediterranean adult populations), mostly cross‐sectional designs Possible cross‐cultural differences in expression of anxiety Review	7
Andersen 2014 Search: Sept. 2012 # incl. studies 24 Meta‐analysis: no		Original studies Different recruitment methods, study inclusion criteria Most study patients were women, thus, possible overestimation of significance of results Different measurement methods: questionnaires, clinical evaluations, structured interviews (some methods not validated for pain patients) Review Search strategy	6
Dawson 2014 Search: Feb 2012 # incl. studies 16 Meta‐analysis: no	Does anxiety come before onset of AMD? Link between length of time since AMD diagnosis and AMD treatments in relation to patient's mental health Include control group to compare prev of anxiety between AMD and non‐AMD populations Use tools with clear cut‐off for clinical anxiety	Original studies Is anxiety different in different forms of AMD? Different definition and measurement of anxiety Comparison group may not be representative Review Small number of studies	5
Other chronic physical diseases in end‐stage
Mitchell 2011 Search: Nov. 2010 # incl. studies 94 Meta‐analysis: yes		Original studies No consensus about optimum psychiatric diagnostic approach in cancer settings Studies of variable quality, mostly cross‐sectional designs, some used convenience sampling, different anxiety measurement methods Could not determine correlates of anxiety Few studies with defined period of prevalence Review Possible publication bias	8
Janssen 2008 – previously described
Murtagh 2007 Search: April 2005 # incl. studies 60 Meta‐analysis: No	Studies on incidence and prevalence of symptoms in ESRD, their causes, and interventions Population‐based, longitudinal studies More information on generalizability of available studies How do symptoms vary between those managed without dialysis and those withdrawing from dialysis? Symptom burden in ESRD Symptoms experienced at end of life Identify what is common and different between those dying from ESRD and other palliative populations	Original studies Heterogeneity: symptom definition, who defines a symptom (reporting), different periods over which prevalence is measured, different tools used No population‐based studies Review Search strategy	6
Solano 2006 – previously described
Trauma
Mckechnie 2014 Search: June 2013 # incl. studies 13 Meta‐analysis: no	Prospective studies assessing long‐term levels of anxiety in post‐traumatic amputees, and whether rehab programmes are successful and mental health issues continue after the programme ends	Original No info on how prev changes with time since amputation (anxiety assessed at fixed time point) Different scoring systems in different populations at various follow‐up times Selected specialist samples not representative of all traumatic amputees Sampling – possible selection bias Attrition during follow‐up Review Some studies may have been missed	8
Chen 2010 Search: Dec. 2008 # incl. studies 37 studies Meta‐analysis: yes	Interplay between stressful life events, vulnerability genes, and development of psychiatric disorders (gene‐environment interactions)	Original studies Self‐report (recall bias), abuse underreport Anxiety affected by unmeasured forms of abuse? Review	8
Fazel 2005 Search: Dec. 2002 # incl. studies 20 Meta‐analysis: yes		Original studies Measurement equivalence issues: differences in sampling methods, diagnostic instruments Insufficient data on refugees in developing countries, asylum seekers, people internally displaced in their own countries Updated info on recently displaced refugees Review NR	5
Vulnerable population subgroups
Older people and their caregivers
Bryant 2008 Search: 2007 # incl. studies 49 Meta‐analysis: no	Hypothesis‐driven research with late‐life anxiety as primary focus Longitudinal designs Studies on anxiety in old age Prevention and early treatment should target old people in poor health and who are at risk for anxiety	Original studies Differences in definition and measurement of anxiety Measurement equivalence issues in elderly – is anxiety experienced differently in elderly? (case definition) Difficult to disentangle physical symptoms & anxiety in elderly Possible selection bias Older people may underreport anxiety Mostly cross‐sectional studies Review	5
Volkert 2013 Search date: Dec. 2011 # incl. studies 25 Meta‐analysis: yes	Studies on anxiety in elderly using improved methodology and accounting for changes in old age (adapted instruments)	Differences in instruments and diagnostic criteria Difficult to disentangle anxiety from physical diseases, somatoform disorders, and depression in elderly Instruments not designed for elderly – what constitutes anxiety in elderly? Heterogeneity: studies of different geographic and cultural regions and using different case definitions and case identification methods Difficult to recruit elderly for studies Review Studies in English and German – limited generalizability No missing data analysis	8
Monastero 2009 Search: Aug. 2008 # incl. studies 27 Meta‐analysis: no	Health care worker to distinguish primary behavioral changes from cognitive impairment Large, cohort studies using standardized instruments to assess NPS as prognostic factors in MCI Optimum ways to assess NPS in those with MCI Genetic and biological markers linking NPS to MCI and dementia	Original studies Possible selection bias Differences in age and sex distributions within studies Differences in instruments used/methods of reporting symptoms Review NR	5
Yates 2013 Search: Nov. 2012 # incl. studies 18 Meta‐analysis: yes	Anxiety and depression should both be considered Classification systems for MCI should consider anxiety Clarify directionality between anxiety and MCI	Original studies Heterogeneity: sampling differences, small samples (may not be representative), different ways of assessing mood/NPS Lacking info on link between anxiety and MCI subtypes Review Possible publication bias, English articles	5
Cooper 2007 Search:2005 # incl. studies 33 Meta‐analysis: no	Cohort studies Research on coping in relation to anxiety (this could be intervention target)	Original studies Lack of info on determinants of anxiety caseness in caregivers Review NR	5
Pregnant women
Russell 2013 Search: August 2012 # incl. studies 17 Meta‐analysis: yes	Prospective studies examining OCD during pregnancy and postpartum period Incidence studies needed Course of OCD across reproductive events Influence of biological determinants on OCD exacerbation throughout reproductive period	Original studies Small samples Difficult to match control studies on various factors Possible overestimation of OCD prev in some control studies OCD evaluated at different pregnancy time points, making comparisons difficult Review Published studies	8
Molyneaux 2014 Search: Jan 2013 # incl. studies 62 Meta‐analysis: insufficient studies for meta‐a for anxiety	Validation of anxiety scales for specific populations needed, ex. women in early pregnancy	Original Heterogeneity: different screening measures and cut‐offs Review English language papers only Published studies eligible Few studies carried out in low and middle‐income countries	6
Sawyer 2010 Search: January 2009 # incl. studies 35 Meta‐analysis: yes	Longitudinal studies to determine anxiety prev at different time points during and after pregnancy Develop cross‐cultural measures of mental health	Original studies Small number of studies Measurement issues, timing of mental health assessment varied (thus, anxiety trajectory over time is unclear) Few studies on antenatal mental health and associated risk factors in African women Insufficient info on how maternal psychological problems impact children Review – NR	6
LGB and self‐harm patients
King 2008 Search: 2005 # incl. studies 25 Meta‐analysis: yes	Prospective studies to determine risk factors of mental disorders Refine definition of sexual orientation	Original studies Difficult to recruit and define LGB group Study design heterogeneity Heterogeneity in definitions of exposure and outcome Review Heterogeneity: study designs and LGB definition Small number of studies included	7
Hawton 2013 Search: Nov. 2011 # incl. studies 50 Meta‐analysis: yes	Studies on mental disorders in those who repeat self‐harm	Original studies Measurement equivalence issues Heterogeneity: methods used to recruit participants, different diagnostic measures used, differences in study participant gender ratios cross‐sectional studies Review English language studies	6

^aprev, prevalence; anx, anxiety; NR, not reported; QA, quality assessment based on AMSTAR criteria.