Psychological Medicine Risk and protective factors for anxiety and

cambridge.org/psm
obsessive-compulsive disorders: an umbrella
review of systematic reviews and meta-analyses
Miquel A. Fullana1,2, *, Miquel Tortella-Feliu3, *, Lorena Fernández de la Cruz4,
Original Article
Jacobo Chamorro5, Ana Pérez-Vigil4, John P. A. Ioannidis6,7,8,9, Aleix Solanes10,11,
*Share first authorship.
Maria Guardiola10, Carmen Almodóvar10, Romina Miranda-Olivos10,
†
Share senior authorship.
Valentina Ramella-Cravaro12,13, Ana Vilar14,15, Abraham Reichenberg16,17,18,19,
Cite this article: Fullana MA et al (2019). Risk
and protective factors for anxiety and David Mataix-Cols4,20, Eduard Vieta21, Paolo Fusar-Poli12,22, Mar Fatjó-Vilas10,23, †
obsessive-compulsive disorders: an umbrella and Joaquim Radua4,10,12,24, †
review of systematic reviews and meta-
analyses. Psychological Medicine 1–16. https:// 1
doi.org/10.1017/S0033291719001247 Institute of Neurosciences, Hospital Clinic, Centro de Investigación Biomédica en Red de Salud Mental
(CIBERSAM), Barcelona, Spain; 2Department of Psychiatry and Forensic Medicine, School of Medicine, Universitat
Received: 8 January 2019 Autònoma de Barcelona, Bellaterra, Spain; 3University Research Institute on Health Sciences (IUNICS), University
Revised: 2 May 2019 of the Balearic Islands, Mallorca, Spain; 4Department of Clinical Neuroscience, Centre for Psychiatry Research,
Accepted: 9 May 2019 Karolinska Institutet, Stockholm, Sweden; 5Anxiety Unit, Institute of Neuropsychiatry and Addictions, Parc de Salut
Mar, Barcelona, Spain; 6Department of Medicine, Stanford Prevention Research Center, Stanford, CA, USA;
Key words: 7
Department of Health Research and Policy, Stanford University School of Medicine, Stanford, CA, USA;
Anxiety disorders; meta-analysis; obsessive- 8
Meta-Research Innovation Center at Stanford, Stanford University, Stanford, CA, USA; 9Department of Statistics,
compulsive disorder; risk factor; umbrella
review Stanford University School of Humanities and Sciences, Stanford, CA, USA; 10FIDMAG Germanes Hospitalaries,
CIBERSAM, Barcelona, Spain; 11Department of Psychiatry and Forensic Medicine, School of Medicine, Autonomous
Author for correspondence: University of Barcelona, Barcelona, Spain; 12Early Psychosis: Interventions and Clinical-detection Lab (EPIC),
Miquel A. Fullana, E-mail: [email protected] Department of Psychosis Studies, Institute of Psychiatry, Psychology & Neuroscience, King’s College London,
London, UK; 13Department of Mental Health, Florence Public Health Center, Florence, Italy; 14Institut de
© Cambridge University Press 2019
Neuropsiquiatria i Addiccions, CSMIJ Sant Martí-La Mina, Parc de Salut Mar, Barcelona, Spain; 15Department of
Experimental and Health Sciences, Pompeu Fabra University (UPF), Barcelona, Spain; 16Institute of Psychiatry,
Psychology & Neuroscience, King’s College London, London, UK; 17Department of Psychiatry, Icahn School of
Medicine at Mount Sinai, New York, NY, USA; 18Department of Preventive Medicine, Icahn School of Medicine at
Mount Sinai, New York, NY, USA; 19Friedman Brain Institute, Icahn School of Medicine at Mount Sinai, New York,
NY, USA; 20Stockholm Health Care Services, Stockholm County Council, Stockholm, Sweden; 21Barcelona Bipolar
Disorders Program, Institute of Neurosciences, Hospital Clinic, University of Barcelona, Institut d’Investigacions
Biomèdiques August Pi i Sunyer, CIBERSAM, Barcelona, Spain; 22Department of Brain and Behavioral Sciences,
University of Pavia, Pavia, Italy; 23Department of Evolutionary Biology, Ecology and Environmental Sciences,
Faculty of Biology, University of Barcelona, Institute of Biomedicine of the University of Barcelona (IBUB),
Barcelona, Spain and 24Institut d’Investigacions Biomèdiques August Pi i Sunyer, CIBERSAM, Barcelona, Spain

Abstract
Background. A multitude of risk/protective factors for anxiety and obsessive-compulsive dis-
orders have been proposed. We conducted an umbrella review to summarize the evidence of
the associations between risk/protective factors and each of the following disorders: specific
phobia, social anxiety disorder, generalized anxiety disorder, panic disorder, and obsessive-
compulsive disorder, and to assess the strength of this evidence whilst controlling for several
biases.
Methods. Publication databases were searched for systematic reviews and meta-analyses
examining associations between potential risk/protective factors and each of the disorders
investigated. The evidence of the association between each factor and disorder was graded
into convincing, highly suggestive, suggestive, weak, or non-significant according to a standar-
dized classification based on: number of cases (>1000), random-effects p-values, 95% predic-
tion intervals, confidence interval of the largest study, heterogeneity between studies, study
effects, and excess of significance.
Results. Nineteen systematic reviews and meta-analyses were included, corresponding to 216
individual studies covering 427 potential risk/protective factors. Only one factor association
(early physical trauma as a risk factor for social anxiety disorder, OR 2.59, 95% CI 2.17–
3.1) met all the criteria for convincing evidence. When excluding the requirement for more
than 1000 cases, five factor associations met the other criteria for convincing evidence and
22 met the remaining criteria for highly suggestive evidence.
Conclusions. Although the amount and quality of the evidence for most risk/protective fac-
tors for anxiety and obsessive-compulsive disorders is limited, a number of factors signifi-
cantly increase the risk for these disorders, may have potential prognostic ability and
inform prevention.
2 Miquel A. Fullana et al.

Introduction hand searched the reference lists of all systematic reviews and
meta-analysis reaching full-text review.
Anxiety disorders are the most common group of mental disor-
Eligibility criteria were: (1) a systematic review or
ders and are associated with enormous societal costs (Kessler
meta-analysis of risk/protective factors for specific phobia, SAD,
et al., 2010; Craske and Stein, 2016). Both ‘genetic’ and ‘non-
GAD, PD, or OCD as defined in any edition of the
genetic’ (i.e. environmental) variables (as well as their interaction)
International Classification of Diseases (ICD) manual or the
have been proposed as potential risk/protective factors for anxiety
DSM; (2) inclusion of a healthy comparison group; and (3) stud-
disorders (Craske et al., 2017), although such a distinction may be
ies reporting sufficient data (or that were retrievable after contact-
somewhat artificial, given that many risk/protective factors
ing the authors) to perform the analyses. We did not apply any
include both genetic and non-genetic components. The evidence
language restrictions in the selection of systematic reviews or
on risk/protective factors for anxiety disorders has been summar-
meta-analyses.
ized in several systematic reviews and meta-analyses. However,
Even though we had hoped to include molecular genetic stud-
findings are conflicting and there have been no previous attempts
ies in the umbrella review, we found that the literature for the dis-
to summarize in a single report the strength of the evidence for
orders investigated is dominated by candidate gene studies, which
the different potential risk/protective factors for each anxiety dis-
are known to have low credibility. Such risk factor assessment
order or to assess possible biases in the literature.
should await thus the publication of large genome-wide studies
We present the results of an umbrella review of risk/protective
(Ioannidis et al., 2008). Moreover, different analytical methods
factors for the most common anxiety disorders. We will focus on
and assessment criteria are required for umbrella reviews of gen-
specific phobia, social anxiety disorder (SAD), generalized anxiety
etic variables (Ioannidis et al., 2008).
disorder (GAD), panic disorder (PD), and obsessive-compulsive
Although in some DSM classifications previous to DSM-5
disorder (OCD) – the latter having been classified as an anxiety
‘panic disorder’ and ‘panic disorder with agoraphobia’ have
disorder until the publication of the 5th edition of the
been classified separately, we included both in our ‘panic disorder’
Diagnostic and Statistical Manual of Mental Disorders (DSM-5;
category. However, we have analyzed them separately where a
American Psychiatric Association, 2013). Umbrella reviews sys-
study reported separate factors for each of these categories. We
tematically collect and assess the existing evidence from individ-
also considered separation anxiety disorder and selective mutism
ual studies included in systematic reviews and/or meta-analyses
(‘anxiety disorders’ in the DSM-5), but they were not included
and have an increasing role in evidence-based health care and
due to the lack of systematic reviews and meta-analyses (online
evidence-based assessments (Ioannidis, 2009; Fusar-Poli and
Supplementary Figs S6 and S7). Posttraumatic stress disorder
Radua, 2018).
(PTSD), grouped as an anxiety disorder until the publication of
In the current absence of valid biomarkers or clear mechanistic
DSM-5, will be covered in a separate manuscript.
explanations for most mental disorders (Kapur et al., 2012), the
Further information about the search strategy and the eligibil-
identification of putative (and, at least for some, modifiable)
ity criteria can be found in the online Supplementary material.
risk/protective factors may lead to the development of more effi-
cient risk prediction models, and may offer clues for prevention
and treatment (Paulus, 2015; Moreno-Peral et al., 2017; Risk/protective factor definition
Fusar-Poli et al., 2018). Our aim was to systematically assess the
We used the following definition of risk factor: ‘that characteristic,
amount of evidence and the robustness of associations between
variable, or hazard preceding the outcome of interest that, if pre-
potential risk/protective factors and each of the aforementioned
sent for a given individual, makes it more likely that this individ-
disorders.
ual, rather than someone selected from the general population,
will develop a given disorder’ (Mrazek and Haggerty, 1994;
Kraemer et al., 1997). Similarly, protective factors are those
Methods where risk is found to be decreased. We assessed both stable fac-
We conducted an umbrella review (Ioannidis, 2009; Fusar-Poli tors (e.g. sex), for which time precedence does not need to be
and Radua, 2018) to assess the relation between potential risk/ established, and factors that are subject to change within-subject.
protective factors and anxiety and obsessive-compulsive disorders. For the latter, we required that the determination of the factor
We followed the Preferred Reporting Items for Systematic Reviews preceded the diagnosis of the outcome (i.e. the disorder) even if
and Meta-Analyses (PRISMA) guidelines (Moher et al., 2009) and the information on the factor and the outcome was collected at
the Meta-analysis of Observational Studies in Epidemiology the same time point (as in the case of cross-sectional studies).
(MOOSE) guidelines (Stroup et al., 2000) (online Supplementary This rule ensured that there would be time precedence for the
Tables S1 and S2). The study protocol was pre-registered with assessments of factors and outcomes, although factors may have
the International Prospective Register of Systematic Reviews existed even before their determination, and disorders may have
(PROSPERO; CRD42017060090). also existed before their diagnosis. Furthermore, when the factor
investigated was related to personality dimensions (e.g. neuroti-
cism), we also required that personality was assessed before the
disorder was diagnosed in order to avoid state-trait influences
Search strategy and eligibility criteria
(Reich et al., 1987). The definitions for each factor were those
We searched PubMed, Web of Science, and Scopus from inception given in the corresponding systematic review or meta-analysis.
to April 30 2018 for systematic reviews and meta-analyses of Following previous work (Radua et al., 2018) we grouped factors
observational studies examining associations between potential into several descriptive categories: sociodemographic, psychopath-
risk/protective factors (see below) separately for each disorder. ology, parental psychopathology, personality dimensions, substance
The search strategy used the keywords ‘systematic review’ or use, life events, perinatal complications, parental rearing styles/
‘meta-analysis’ and each of the disorders of interest. We also attachment, and others.
Psychological Medicine 3

Data extraction and selection suggestive (class II), suggestive (class III), or weak (class IV)
(Fusar-Poli and Radua, 2018). Convincing evidence required a
We used a systematic approach to extract and select the data. First,
number of cases (n) >1000, a highly significant association ( p <
we identified the factors assessed in each systematic review or
10−6), I 2 < 50%, a statistically significant 95% prediction interval,
meta-analysis. Second, two investigators independently checked
and the absence of small-study effects and excess significance bias.
that each individual article included in the systematic review or
Highly suggestive evidence also required n > 1000, a highly signifi-
meta-analysis met the same eligibility criteria applied to the sys-
cant association ( p < 10−6), and that the largest study had a stat-
tematic review or meta-analysis. Third, two investigators inde-
istically significant effect. Suggestive evidence required n > 1000
pendently extracted the following data (from the systematic
and p < 10−3. Weak evidence required no specific number of
review or meta-analysis or, in most cases, from the individual
cases and p < 0.05. Furthermore, after collecting all the available
studies): first author and year of publication; number of cases
evidence, we noticed that, with few exceptions, there were fewer
and controls; measure and size of the risk and corresponding
than 1000 cases for most factors. Therefore, we also examined
95% confidence intervals (CIs); specific variables depending on
these criteria removing the requirement of n > 1000, so as to
the measure of effect size; and whether the study was a prospect-
obtain a more fine-grained appraisal of the evidence. For associa-
ive cohort. Specific variables depending on the measure of effect
tions with significant evidence (classes I–IV), we also conducted a
size were: number of cases and person-times in exposed and
sensitivity analysis by using only prospective cohort studies.
unexposed for incidence rate ratios (IRR), number of cases and
total number of exposed and unexposed for risk ratios (RR), num-
ber of exposed and unexposed and cases and controls for odds Results
ratios (OR), and means and standard deviations for cases and
We included 19 systematic reviews and meta-analyses (Fig. 1 and
controls for standardized mean differences (SMD). Fourth, two
online Supplementary Figs S1–S5). AMSTAR scores are presented
investigators independently rated the quality of the systematic
in Table 1. All extracted data and results are available at: https://
review or meta-analysis using the Assessment of Multiple
www.umbrellaevidence.com/anxiety/riskfactors/.
Systematic Reviews (AMSTAR; Shea et al., 2007) tool, with sub-
We extracted data for 427 factors from 216 individual studies.
stantial interrater agreement (both weighted Cohen’s kappa and
The number of systematic reviews and meta-analyses, individual
intraclass correlation = 0.71; see online Supplementary material).
studies assessed and included, and factors included are presented
A third investigator reviewed the extracted data to check for
in Table 2. The groups of factors assessed in each systematic
inconsistencies, and disagreements were resolved by consensus.
review or meta-analysis are reported in Table 1 and the specific
For further details on the data extraction, selection, and quality
factors in online Supplementary Table S3. Factors showing con-
assessment, see the online Supplementary material.
vincing, highly suggestive, or suggestive evidence of association
with each disorder are presented in Table 3. All significant factors
(including those showing weak evidence of association) are pre-
Statistical analysis
sented in online Supplementary Table S4.
We performed statistical analyses commonly used in standard Overall, the number of cases was greater than 1000 for 20 fac-
meta-analyses. However, we did not use the statistics provided tors (4.68%). One-hundred eighty-three of the 427 factors
in the included systematic review or meta-analyses because (42.84%) presented a statistically significant effect ( p < 0.05)
there are differences across-studies in the methods employed under the random-effects model, but only 91 (21.31%) had a p
and because some analyses are often not conducted (e.g. the < 0.005 and only 27 (6.32%) reached p < 10−6. Twenty-five factors
test for an excess of significant findings). (36.76%) presented a large estimate of heterogeneity (I 2 > 50%),
We conducted a separate random-effects meta-analysis for while for 29 factors (78.37%) the 95% prediction interval did
each factor and disorder. The outcomes of the meta-analyses not include the null. Additionally, evidence for small-study effects
were the effect sizes with their CIs and p-values, and the statistics and excess significance bias was noted for 2 (5.40%) and 8
required to assess the level of evidence (see below). Depending on (1.87%) factors, respectively (see online Supplementary Table S4).
the factor, we used IRR, RR, OR, or SMD Hedges’ g. For descrip-
tive purposes, we also report OR equivalents (eOR) of IRR, RR,
Results by disorder
and Hedge’s g (see Fusar-Poli and Radua (2018) for additional
details). Specific phobia
We assessed between-study heterogeneity by estimating the No factor showed convincing or highly suggestive evidence as a
95% prediction interval – which evaluates the uncertainty for risk/protective factor for specific phobia using the original
the effect that would be expected in a new study addressing that umbrella review criteria. Removing the n > 1000 criterion, being
same association – and the I 2 metric (Ioannidis et al., 2007). I 2 male showed convincing evidence as protective factor.
> 50% were considered to represent substantial heterogeneity Moreover, neuroticism showed highly suggestive evidence as
(Higgings and Green, 2009). We also assessed whether there risk factor for the disorder, which was maintained after the sen-
was evidence of small-study effects with Egger tests (Egger sitivity analyses (Table 3, Fig. 2, and online Supplementary
et al., 1997), where statistical significance would mean potential Table S4).
reporting or publication bias in the smaller studies. Finally, excess
significance (a relative excess of statistically significant findings) Social anxiety disorder
was assessed with a binomial test that compared the observed v. Early physical and sexual trauma showed, respectively, convincing
the expected number of studies yielding statistically significant and suggestive evidence as risk factors for SAD. Additionally,
results (Radua et al., 2018). when removing the n > 1000 criterion, dysthymia, insecure attach-
The levels of evidence of the associations between each factor ment in childhood, major depression, and neuroticism showed
4 Miquel A. Fullana et al.

Fig. 1. Flow chart of the literature search (see online Supplementary material for the flowcharts for each specific disorder).
Psychological Medicine 5

Table 1. Systematic reviews and meta-analyses included in the umbrella review, quality scores, and groups of risk/protective factors examined, by disorder

Systematic review/ AMSTAR score

meta-analysis (0–11)a Groups of risk/protective factors examined Disorders examined

Brander et al. (2016a) 6 Socio-demographic; parental rearing styles/attachment; substance use; life OCD
events; other
Brown et al. (2000) 3 Life events Specific phobia, GAD,
PD
Clarner et al. (2015) 8 Life events PD
Clauss and Blackford 8 Other (behavioral inhibition in childhood) SAD
(2012)
Colonnesi et al. (2011) 8 Parental rearing styles/attachment SAD
Fernandes et al. 8 Life events SAD, GAD, PD
(2015)
Gariepy et al. (2010) 10 Other (obesity) GAD
Guo et al. (2016) 10 Socio-demographic Specific phobia, SAD,
GAD, PD, OCD
Jacobson and 7 Psychopathology Specific phobia, GAD,
Newman (2017) PD
Kedzior and Laeber 7 Substance use Specific phobia, GAD,
(2014) PD, OCD
Kisely et al. (2017) 11 Socio-demographic SAD, GAD, PD
Kossowsky et al. 10 Psychopathology PD
(2013)
Kotov et al. (2010) 5 Personality dimensions Specific phobia, SAD,
GAD, OCD
Micco et al. (2009) 7 Parental psychopathology SAD, GAD, PD, OCD
Moreno-Peral et al. 9 Socio-demographic; psychopathology; parental psychopathology; personality GAD, PD
(2014) dimensions; substance use; life events; perinatal complications; parental rearing
styles/attachment; other
Moylan et al. (2012) 6 Substance use Specific phobia, SAD,
GAD
Osborn et al. (2016) 8 Other (traumatic brain injury) GAD
Tarricone et al. (2012) 8 Socio-demographic GAD
Van Steensel et al. 6 Psychopathology Specific phobia
(2011)
AMSTAR, Measurement tool to assess the methodological quality of systematic reviews; GAD, generalized anxiety disorder; PD, panic disorder; OCD, obsessive-compulsive disorder; SAD,
social anxiety disorder.
Note: Some risk/protective factors were assessed only for some of the disorders included in the corresponding systematic review/meta-analysis.
The specific risk/protective factors assessed in each systematic review or meta-analysis are reported in online Supplementary Table S4.
a
Rounded-up average of two raters.

sensitivity analyses, evidence for both trauma-related factors Panic disorder

became weak, but the rest of factors–except insecure attachment No factor showed convincing or highly suggestive evidence as
in childhood- maintained the same level of evidence (Table 3, risk/protective factor for PD. Removing the n > 1000 criterion,
Fig. 2, and online Supplementary Table S4). being male, separation anxiety in childhood, and early physical
trauma showed convincing evidence as risk/protective factors
Generalized anxiety disorder for PD. The evidence was not maintained, however, after sensitiv-
No factor showed convincing or highly suggestive evidence as a ity analyses. Furthermore, daily cigarette smoking, panic attacks,
risk/protective factor for GAD. Removing the n > 1000 criterion, and major depression showed highly suggestive evidence as risk
being male showed convincing evidence as protective factor for factors for PD, which was maintained after sensitivity analyses
GAD and the following factors showed highly suggestive evidence (Table 3, Fig. 2, and online Supplementary Table S4).
as risk factors for the disorder: psychological malaise at age 33,
borderline personality disorder, parental GAD without comorbid- Obsessive-compulsive disorder
ity, early physical and sexual trauma, and behavioral inhibition No factor showed convincing or highly suggestive evidence as
(assessed as a personality dimension). After sensitivity analyses, risk/protective factor for OCD. Removing the n > 1000 criterion,
all these factors – except both trauma-related variables – main- several parental rearing style variables, neuroticism, and use of
tained the same level of evidence (Table 3, Fig. 2, and online cocaine together with another drug (except marijuana) showed
Supplementary Table S4). highly suggestive evidence as risk/protective factors for OCD.
6 Miquel A. Fullana et al.

Table 2. Number of systematic reviews and meta-analyses included in the umbrella review, individual studies assessed and included, and potential risk/protective
factors included in the umbrella review, by disorder

Number of individual
Number of systematic reviews studies assessed for Number of individual Number of potential risk/
Disorder or meta-analyses included eligibility studies included protective factors included

Specific phobia 6 63 19 13
Social anxiety disorder 10 110 34 20
Generalized anxiety 13 132 57 110
disorder
Panic disorder 11 144 60 78
Obsessive-compulsive 6 160 46 206
disorder

However, the latter was based on a single study reporting one sin- developmental effects (including temporal dynamics and the
gle case in the exposed group. Only neuroticism and use of development of comorbidity over time) are often ignored in cur-
cocaine together with another drug (except marijuana) main- rent nosological systems. The use of longitudinal ‘staging models’ –
tained the same level of evidence after the sensitivity analyses that describe the progression from more simple or ‘pure’ disorders
(Table 3, Fig. 2, and online Supplementary Table S4). to more complex or comorbid disorders – has been proposed to
deal with these issues. Such models could offer a better descrip-
tion of the developmental patterns typical to most mental disor-
Discussion
ders (Beesdo et al., 2009).
This is, to the best of our knowledge, the first umbrella review of Our data suggest that rather than ‘a few’ risk or protective fac-
risk/protective factors for anxiety and obsessive-compulsive disor- tors with large effects, large sets of common ‘variants’ of small
ders. Our study provides a state-of-the-art classification of risk/ effects account for the risk for anxiety and obsessive-compulsive
protective factors based on the robustness of associations between disorders. This idea, which is well established in psychiatry genet-
these factors and five separate disorders, while controlling for sev- ics (Anttila et al., 2018; Sullivan et al., 2018), seems to be also true
eral biases. for ‘non-purely genetic’ factors (Uher and Zwicker, 2017).
Using the original umbrella review criteria, early physical Furthermore, our findings open the door to the potential develop-
trauma was the single most consistent risk factor – class I – for ment of enhanced risk prediction models (Bernardini et al., 2017)
SAD. Early sexual trauma was also associated – class III – with and individual risk prediction scores (see Kessler et al., 2014, and
SAD. Several ‘traditional’ risk/protective factors for anxiety and Shalev et al., 2019 for specific examples in PTSD). In recent years,
obsessive-compulsive disorders were among those that had nom- the use of polygenic risk scores has been validated in disorders
inally statistically significant results (Beesdo et al., 2009; Craske such as schizophrenia (International Schizophrenia Consortium
and Stein, 2016). Although we could not assess exactly the et al., 2009). More recently, the use of ‘poly-environmental scores’
same factors for all disorders, a number of factors showed a simi- has been proposed (Padmanabhan et al., 2017; Uher and Zwicker,
lar association with several of the disorders investigated (Fig. 3). 2017). Given that multiple genetic and non-genetic factors have
For example, being male was associated with decreased risk for much greater explanatory power than considering them one at a
specific phobia, SAD, GAD, and PD; and neuroticism was asso- time in most mental disorders (Uher and Zwicker, 2017), it is
ciated with increased risk for specific phobia, SAD, GAD, and likely that ‘poly-risk’ scores (containing both genetic and non-
OCD. Moreover, early traumatic experiences increased the risk genetic factors) improve the prediction of mental disorders. Our
of all disorders in which they were investigated (SAD, GAD, data may help developing such scores for anxiety and obsessive-
PD, and OCD). Although the evidence for most of these associa- compulsive disorders, although the time of exposure and the
tions was rated as weak, the consistency of these signals across cumulative nature of non-genetic risk will need to be taken into
multiple disorders strengthens the case that they do carry prog- account to improve such prediction abilities (Moffitt et al.,
nostic potential. The fact that the same factors increased the 2005; Sharma et al., 2016). Developmental effects – and their
risk for different disorders may indicate a shared liability within potential interaction with genetic variables- are difficult to study
anxiety and obsessive-compulsive disorders (Blanco et al., using epidemiological data, but they could be investigated using
2014). Moreover, some factors may be shared across mental disor- animal models (Leonardo and Hen, 2008).
ders (i.e. be ‘transdiagnostic’). For example, early traumatic The majority of factors were only classified as having weak evi-
experiences are a significant risk factor for depressive (Köhler dence (class IV). This mainly reflects the methodological limita-
et al., 2018), psychotic (Belbasis et al., 2018; Radua et al., 2018), tions of the data, where less than 5% of the factors included
and bipolar disorders (Bortolato et al., 2017). Importantly, the more than 1000 cases and where the significance of the associa-
results of our umbrella review provide hints not only on the pres- tions for each individual factor was overall low. The (weak)
ence/absence of a particular factor but also on the loading strength of the associations found, together with limitations
(weight) of that factor, which may be still unique (Uher and inherent to the individual study designs employed to date, pre-
Zwicker, 2017). cludes firm causal inferences for any of the significant factors
The non-specificity of the findings for most risk factors inves- identified in our umbrella review (Paulus, 2015). Future work to
tigated here (and probably for most risk factors for mental disor- identify risk/protective factors could focus on large-scale family-
ders in general) may also be partially explained by the fact that based designs, that allow for a more stringent control of
 Psychological Medicine
Table 3. Risk/protective factors showing convincing (class I), highly suggestive (class II), or suggestive (class III) evidence of association with each disorder using the original umbrella review criteria or after removing the n > 1000 cases criterion,
by disorder

Class
Risk/Protective PI ET ESB LS Class (-n > 1000,
Disorder Factor group factor K N Measure ES (95% CI) p sign. I2 sign. sign. sign. eOR Class (-n > 1000) prosp.)

Specific Socio-demographic Male gender 9 689 OR 0.43 (0.36–0..51) <0.000001 Yes 0% No No Yes 0.43 IV I NA
phobia

Personality Neuroticism 1 79 g 0.81 (0.57–1.05) <0.000001 NA NA NA No Yes 4.35 IV II II

dimensions

SAD Psychopathology Dysthymia 1 52 IRR 14.81 (6.7–32.73) <0.000001 NA NA NA No Yes 14.81 IV II II

Major depression 1 52 IRR 9.35 (4.71–18.54) <0.000001 NA NA NA No Yes 9.35 IV II II

Personality Neuroticism 1 89 g 0.89 (0.67–1.12) <0.000001 NA NA NA No Yes 5.02 IV II II

dimensions

Life events Early emotional 3 720 OR 2.8 (1.84–4.24) 0.000001 No 36% No No Yes 2.80 IV III III
trauma

Early physical 4 1191 OR 2.59 (2.17–3.1) <0.000001 Yes 0 No No Yes 2.59 I I IV

trauma

Early sexual 5 1239 OR 3.18 (1.73–5.86) 0.00019 No 85% No No Yes 3.18 III III IV
trauma

Parental rearing Insecure 1 76 g 1.26 (0.91–1.61) <0.000001 NA NA NA No Yes 9.83 IV II NA

styles/attachment attachment in
childhood

Other Behavioral 7 257 OR 7.52 (3.04–18.61) 0.000013 No 78% No Yes Yes 7.52 IV III III
inhibition in
childhooda

GAD Socio-demographic Age (30 to 54) 1 390 OR 2.92 (1.78–4.78) 0.000022 NA NA NA No Yes 2.92 IV III III

Male gender 15 999 OR 0.5 (0.41–0.59) <0.000001 Yes 0% No No Yes 0.5 IV I NA

Psychopathology Bipolar I disorder 1 390 OR 2.58 (1.48–4.49) 0.00081 NA NA NA No Yes 2.58 IV III III

Borderline 1 390 OR 4.71 (2.93–7.57) <0.000001 NA NA NA No Yes 4.71 IV II II

personality
disorder

History of one 1 288 OR 1.7 (1.27–2.26) 0.00029 NA NA NA No Yes 1.70 IV III III
psychological
disorder

Internalizing 1 288 OR 2.01 (1.34–3) 0.00065 NA NA NA No Yes 2.01 IV III III

disorder at age 16

Internalizing 1 288 OR 1.91 (1.31–2.79) 0.00081 NA NA NA No Yes 1.91 IV III III

disorder at age 7

Narcissistic 1 390 OR 2.31 (1.49–3.6) 0.00019 NA NA NA No Yes 2.31 IV III III

personality
disorder
(Continued )

7
.
 8
Table 3. (Continued.)

Class
Risk/Protective PI ET ESB LS Class (-n > 1000,
Disorder Factor group factor K N Measure ES (95% CI) p sign. I2 sign. sign. sign. eOR Class (-n > 1000) prosp.)

Psychological 1 288 OR 4.73 (3.43–6.52) <0.000001 NA NA NA No Yes 4.73 IV II II

malaise at age 33

Schizotypal 1 390 OR 2.6 (1.52–4.44) 0.00045 NA NA NA No Yes 2.60 IV III III

personality
disorder

Subsyndromal 1 563 OR 2.25 (1.56–3.24) 0.000014 NA NA NA No Yes 2.25 IV III III

depression no
distress

Parental Anxiety 8 254 OR 3.45 (1.97–6.02) 0.000013 Yes 0% No No Yes 3.45 IV III NA
psychopathology

GAD without 1 106 HR 3.77 (2.27–6.26) <0.000001 NA NA NA No Yes 3.77 IV II II

comorbidity

Major depression 1 65 OR 3.7 (2.01–6.79) 0.000024 NA NA NA No Yes 3.70 IV III III

in both parents

Major depression 1 79 OR 2.51 (1.47–4.29) 0.00074 NA NA NA No Yes 2.51 IV III III

in one parent

Personality Behavioral 1 106 HR 1.97 (1.66–2.33) <0.000001 NA NA NA No Yes 1.97 IV II II

dimensions inhibitiona

Harm avoidance 1 106 HR 1.69 (1.37–2.09) 0.000001 NA NA NA No Yes 1.69 IV III III

Substance use Cannabis use 1 83 OR 2.79 (1.55–5.02) 0.00059 NA NA NA No Yes 2.79 IV III III

Life events Early physical 1 350 OR 2.39 (1.92–2.98) <0.000001 NA NA NA No Yes 2.39 IV II NA
trauma

Early sexual 1 350 OR 3.28 (2.6–4.14) <0.000001 NA NA NA No Yes 3.28 IV II NA

trauma

Physical abuse in 1 165 OR 1.82 (1.33–2.48) 0.00017 NA NA NA No Yes 1.82 IV III IV

childhood

Separation events 1 106 HR 2.44 (1.54–3.85) 0.00013 NA NA NA No Yes 2.44 IV III IV

in childhood

Other Received mental 1 52 OR 6.15 (2.81–13.45) 0.000005 NA NA NA No Yes 6.15 IV III III
health treatment
from 20 to 32

Received 1 52 OR 5.19 (1.98–13.55) 0.00078 NA NA NA No Yes 5.19 IV III III

Miquel A. Fullana et al.

psychiatric
medication from
20 to 32

PD Socio-demographic Male gender 11 439 OR 0.5 (0.39–0.64) <0.000001 Yes 0% No No Yes 0.5 IV I NA

Psychopathology Major depression 2 771 OR 2.03 (1.66–2.49) <0.000001 NA 0% NA No Yes 2.03 IV II II

Panic attacks 1 811 OR 2.73 (1.93–3.88) <0.000001 NA NA NA No Yes 2.73 IV II II

.
 Post-traumatic 1 224 OR 2.59 (1.5–4.47) 0.00062 NA NA NA No yYes 2.59 IV III III

Psychological Medicine
stress disorder

Separation anxiety 10 880 OR 6.11 (4.31–8.66) <0.000001 Yes 5% No No Yes 6.11 IV I NA

in childhood

Parental Panic attacks (for 1 54 OR 3.93 (1.91–8.07) 0.00019 NA NA NA No Yes 3.93 IV III III
psychopathology PDA)

Substance use Cigarette smoking 2 201 HR 3.46 (2.21–5.41) <0.000001 NA 21% NA No Yes 3.46 IV II II
(daily)

Cigarette smoking 1 51 HR 14.46 (4.81.–43.5) 0.000002 NA NA NA No Yes 14.46 IV III III

(persistence in
daily smokers)

Cigarette smoking 1 149 HR 3.18 (1.99–5.1) 0.000001 NA NA NA No Yes 3.18 IV III III
(persistence in
prior daily
smokers)

Life events Early emotional 1 123 OR 2.71 (1.57–4.68) 0.00035 NA NA NA No Yes 2.71 IV III NA
trauma

Early trauma 2 194 OR 3.56 (1.86–6.8) 0.00012 NA 0 NA No Yes 3.56 IV III NA

Early physical 4 449 OR 2.46 (1.95–3.11) <0.000001 Yes 0% No No Yes 2.46 IV I III
trauma

Early sexual 5 518 OR 2.91 (1.67–5.08) 0.00017 No 73% No No Yes 2.91 IV III III
trauma

Other Joint 1 14 RR 22.34 (5.3–94.29) 0.000023 NA NA NA No Yes 22.34 IV III III

hypermobility
syndrome

OCD Socio-demographic Paternal age >35 1 122 OR 5.34 (2.15–13.27) 0.00030 NA NA NA No Yes 5.34 IV III NA

Perinatal Ear infection 1 68 OR 57.81 (7.59–440.61) 0.00009 NA NA NA No Yes 57.81 IV III NA

complications

Early 1 13 OR 11.53 (2.84–46.78) 0.00062 NA NA NA No Yes 11.53 IV III NA

developmental
problems

Excess weight gain 1 68 OR 9.31 (2.62–33.1) 0.00057 NA NA NA No Yes 9.31 IV III NA

in pregnancy

Hyperemesis 1 68 OR 8 (3.21–19.97) 0.000008 NA NA NA No Yes 8.00 IV III NA

Medication during 1 68 OR 5.45 (2.6–11.45) 0.000007 NA NA NA No Yes 5.45 IV III NA

pregnancy

Mumps 1 68 OR 11.41 (3.23–40.28) 0.00015 NA NA NA No Yes 11.41 IV III NA

Other postnatal 1 68 OR 5.81 (2.04–16.52) 0.00096 NA NA NA No Yes 5.81 IV III NA

problems

Other problems in 1 68 OR 12.18 (3.46–42.9) 0.0001 NA NA NA No Yes 12.18 IV III NA

pregnancy

Throat infection 1 68 OR 4.7 (2.28–9.7) 0.000028 NA NA NA No Yes 4.70 IV III NA

(Continued )

9
.
 Table 3. (Continued.)

10
Class
Risk/Protective PI ET ESB LS Class (-n > 1000,
2
Disorder Factor group factor K N Measure ES (95% CI) p sign. I sign. sign. sign. eOR Class (-n > 1000) prosp.)

Substance use Alcohol use 1 105 RR 2.41 (1.6–3.62) 0.000024 NA NA NA No Yes 2.41 IV III III
disorder

Use of cocaine and 1 105 RR 5.92 (4.97–7.05) <0.000001 NA NA NA No Yes 5.92 IV II II

others (no
marijuana)

Life events Emotional neglect 1 74 g 0.75 (0.32–1.18) 0.00066 NA NA NA No Yes 3.90 IV III NA
in childhood

History of verbal 1 33 OR 4.36 (1.88–10.11) 0.00061 NA NA NA No Yes 4.36 IV III NA

abuse in family

Sexual assault in 2 32 RR 4.03 (1.83–8.87) 0.00052 NA 0 NA No Yes 4.03 IV III NA

childhood

Personality Neuroticism 1 62 g 1.23 (0.96–1.5) <0.000001 NA NA NA No Yes 9.31 IV II II

dimensions

Parental rearing Interference from 1 94 g 0.85 (0.55–1.14) <0.000001 NA NA NA No Yes 4.67 IV II NA

styles/attachment father

Overprotection 6 716 g 0.44 (0.21–0.68) 0.00017 No 65% No No Yes 2.24 III III NA
from father

Punishment from 1 94 g 0.71 (0.42–1) 0.000001 NA NA NA No Yes 3.62 IV III NA

father

Refusal from father 1 94 g 1.28 (0.98–1.59) <0.000001 NA NA NA No Yes 10.19 IV II NA

Warmth from 3 248 g −0.64 (−0.87 to <0.000001 No 23% No No Yes 0.31 IV II NA

father −0.42)

Other Postpartum 1 29 OR 12.05 (3.5–41.52) 0.000081 NA NA NA No Yes 12.05 IV III NA

Class, class of evidence, Class (−n > 1000)- class of evidence after removing the n > 1000 cases criterion, Class (−n > 1000, prosp.)– class of evidence after removing the n > 1000 cases criterion and after sensitivity analyses (including only prospective
studies); CI, confidence interval; ES, effect size; ET, Egger test; eOR, equivalent odds ratio; ESB, excess significance bias; g, Hedge’s g; GAD, generalized anxiety disorder; HR, hazard ratio; I 2, heterogeneity; IRR, incidence rate ratio; K, number of studies
for each factor; LS, largest study with significant effect; N, number of cases; NA, not assessable; ns, not significant; OCD, obsessive-compulsive disorder; OR, odds ratio; PD, panic disorder; PDA, panic disorder with agoraphobia; PI, prediction interval;
SAD, social anxiety disorder; sign., significant; RR, relative risk.
a
‘Behavioral inhibition’ referred to ‘the chronic tendency to respond to novel persons, places, and objects with wariness or avoidant behaviours’ in one meta-analysis (Clauss and Blackford, 2012) and to a personality/character dimension referring to
‘consistent restraint in response to social and non social situations’ in one systematic review (Moreno-Peral et al., 2014).

Miquel A. Fullana et al.

.
Psychological Medicine 11

Fig. 2. Forest plots of risk (in red) and protective (in green)
factors showing convincing (class I) or highly suggestive
(class II) evidence of association with each disorder, after
removing the n > 1000 cases criterion.
12 Miquel A. Fullana et al.

Fig. 3. Forest plots of risk/protective factors assessed

in at least four of the disorders under study and show-
ing convincing (class I) or highly suggestive (class II)
evidence of association with at least one of the disor-
ders, after removing the n > 1000 cases criterion.
Psychological Medicine 13

unmeasured familial confounders (D’Onofrio et al., 2013) and included in the systematic reviews and meta-analyses (because
should improve the confidence in the identification of ‘non-purely this is beyond the scope of an umbrella review). Moreover,
genetic’ risk/protective factors that are in the causal pathway for there may be differences across-studies in the exact definitions
anxiety and obsessive-compulsive disorders. For example, recent and methods of assessment for each factor. Finally, the almost
population-based work in OCD has confirmed that a range of ubiquitously limited amount of evidence made us explore also
perinatal complications are robustly associated with the disorder, what would happen if we removed the need for >1000 cases to
even after strict control of unmeasured genetic and environmental have highly suggestive evidence. Nevertheless, great caution is
confounders, and that the number of perinatal complications needed in trusting associations, no matter how strong and consist-
cumulatively contribute to risk for the disorder (Brander et al., ent, where data are sparse.
2016b). Similarly, as the field of psychiatric genetics is clearly In summary, we found a number of nominally statistically sig-
shifting away from the candidate gene approach into the less arbi- nificant risk and protective factors for anxiety and obsessive-
trary genome-wide association studies (GWAS) approach, the compulsive disorders, although very few were supported by robust
identification of genetic variants implicated in these disorders evidence. The limited amount of evidence was the main restrict-
should increase dramatically in the next few years, as exemplified ing factor, and this means that there is plenty of room to improve
by the recent formation of an anxiety disorders group within the the standards of evidence in this field. Our findings may help
psychiatric genetics consortium (Sullivan et al., 2018). optimize current prediction models and may provide hints for
We also note that we identified very few protective factors that testing prevention strategies.
were not reciprocal to risk factors. This indicates that most
research so far has focused on adverse/negative factors, and high- Author ORCIDs. Miquel A. Fullana, 0000-0003-3863-5223.
lights another important aspect that will need to be addressed in Supplementary material. The supplementary material for this article can
future studies. be found at https://doi.org/10.1017/S0033291719001247.
Our results may also offer opportunities for prevention.
Current prevention programs for anxiety disorders have shown Acknowledgements. Dr Fernández de la Cruz is supported by a Junior
modest benefits (Moreno-Peral et al., 2017) and there is a need Researcher grant from the Swedish Research Council for Health, Working
for new strategies. Our findings lend support to identifying Life and Welfare (FORTE grant number 2015-00569). Ms. Pérez-Vigil is sup-
ported by a grant from the Alicia Koplowitz Foundation. Drs. Vieta, Radua,
those individuals with several risk factors for inclusion in preven-
and Fatjó-Vilas have received support from the Instituto de Salud Carlos III,
tion programs (Blanco et al., 2014). Large sets of risk factors of Ministry of Economy and Competitiveness of Spain (PI 12/00912; CP14/
small effects seem to account for the risk for anxiety and 00041; CD16/00264), integrated into the Plan Nacional de I+D+I and
obsessive-compulsive disorders, and therefore interventions that cofounded by ISCIII- Subdirección General de Evaluación and Fondo
try to modulate several of them concurrently should be devised. Europeo de Desarrollo Regional (FEDER) and Centro para la Investigación
For example, parental psychopathology and parental rearing styles Biomédica en Red de Salud Mental (CIBERSAM). Dr Vieta has also received
were significant risk factors for several of the disorders investi- support from Secretaria d’Universitats i Recerca del Departament d’Economia
gated here and could be a combined target for prevention efforts. i Coneixement (2014_SGR_398), Seventh European Framework Programme
Recent data on moderators and mediators of prevention strategies (ENBREC), and the Stanley Medical Research Institute. Dr Fatjó-Vilas has
also received support from Comissionat per a Universitats i Recerca del
should help optimise such efforts (Ginsburg et al., 2015). Our
DIUE, of the Generalitat de Catalunya regional authorities (2017_SGR_1271).
results also support focusing on those modifiable risk factors
with the largest effects (e.g. trauma), and whose reduction Conflict of interest. Dr Fernández de la Cruz and Prof. Mataix-Cols
would have a greater prospective impact (Li et al., 2016). receive royalties for contributing articles to UpToDate, Wolters Kluwer
Claims of success should await the results of randomized trials, Health. Dr Vieta has received grants and honoraria from AB-Biotics,
since observational associations may not necessarily represent Allergan, Angelini, AstraZeneca, Ferrer, Forest Research Institute, Gedeon
causal effects. Richter, GlaxoSmithKline, Janssen, Lundbeck, Medscape, Otsuka, Pfizer,
Our study has several strengths. We used systematic search Sanofi-Aventis, Sunovion, and Takeda as well as from the CIBERSAM,
Grups Consolidats de Recerca 2014 (SGR 398), the Seventh European
methods and both the study selection and data extraction were
Framework Programme (ENBREC), Horizon 2020 (R-LINK) and the
conducted by independent raters. Moreover, we assessed that
Stanley Medical Research Institute. The rest of authors report no competing
each individual study included in the systematic review or interests.
meta-analysis fulfilled our inclusion criteria and used standard
approaches to assess the methodological quality of the systematic Ethical standards. The authors assert that all procedures contributing to
reviews or meta-analysis (Fusar-Poli and Radua, 2018). We offer this work comply with the ethical standards of the relevant national and insti-
as online Supplementary material all data collected in our tutional committees on human experimentation and with the Helsinki
umbrella review. Beyond encouraging open science, this databank Declaration of 1975, as revised in 2008.
may contribute to the creation of a database of risk/protective fac-
tors for anxiety and obsessive-compulsive disorders that can be References
updated in the future. We also note several limitations. First, we
assessed each of disorders separately and did not use a mixed American Psychiatric Association (2013). Diagnostic and Statistical Manual
‘anxiety disorders’ category as an outcome. There have been of Mental Disorders, 5th Edn. Washington, DC: American Psychiatric
changes in the specific disorders included under the ‘anxiety dis- Association.
Anttila V, Bulik-Sullivan B, Finucane HK, Walters RK, Bras J, Duncan L,
orders’ category, complicating the interpretation of such analyses.
Escott-Price V, Falcone GJ, Gormley P, Malik R, Patsopoulos NA,
Second, we collected only information about factors assessed in Ripke S, Wei Z, Yu D, Lee PH, Turley P, Grenier-Boley B,
systematic reviews and meta-analyses, and studies not included Chouraki V, Kamatani Y, Berr C, Letenneur L, Hannequin D,
in this type of publication were not eligible for inclusion. Amouyel P, Boland A, Deleuze J-F, Duron E, Vardarajan BN, Reitz C,
Moreover, not all factors were evaluated for all the disorders. Goate AM, Huentelman MJ, Kamboh MI, Larson EB, Rogaeva E, St
Third, we did not assess the quality of the individual studies George-Hyslop P, Hakonarson H, Kukull WA, Farrer LA, Barnes LL,
14 Miquel A. Fullana et al.

Beach TG, Demirci FY, Head E, Hulette CM, Jicha GA, Kauwe JSK, MacIntyre DJ, Porteous D, Homuth G, Rivera M, Grove J,
Kaye JA, Leverenz JB, Levey AI, Lieberman AP, Pankratz VS, Middeldorp CM, Hickie I, Pergadia M, Mehta D, Smit JH, Jansen R, de
Poon WW, Quinn JF, Saykin AJ, Schneider LS, Smith AG, Sonnen JA, Geus E, Dunn E, Li QS, Nauck M, Schoevers RA, Beekman ATF,
Stern RA, Van Deerlin VM, Van Eldik LJ, Harold D, Russo G, Knowles JA, Viktorin A, Arnold P, Barr CL, Bedoya-Berrio G, Joseph
Rubinsztein DC, Bayer A, Tsolaki M, Proitsi P, Fox NC, Hampel H, Bienvenu O, Brentani H, Burton C, Camarena B, Cappi C, Cath D,
Owen MJ, Mead S, Passmore P, Morgan K, Nöthen MM, Schott JM, Cavallini M, Cusi D, Darrow S, Denys D, Derks EM, Dietrich A,
Rossor M, Lupton MK, Hoffmann P, Kornhuber J, Lawlor B, Fernandez T, Figee M, Freimer N, Gerber G, Grados M, Greenberg E,
McQuillin A, Al-Chalabi A, Bis JC, Ruiz A, Boada M, Seshadri S, Hanna GL, Hartmann A, Hirschtritt ME, Hoekstra PJ, Huang A,
Beiser A, Rice K, van der Lee SJ, De Jager PL, Geschwind DH, Huyser C, Illmann C, Jenike M, Kuperman S, Leventhal B, Lochner C,
Riemenschneider M, Riedel-Heller S, Rotter JI, Ransmayr G, Lyon GJ, Macciardi F, Madruga-Garrido M, Malaty IA, Maras A,
Hyman BT, Cruchaga C, Alegret M, Winsvold B, Palta P, Farh K-H, McGrath L, Miguel EC, Mir P, Nestadt G, Nicolini H, Okun MS,
Cuenca-Leon E, Furlotte N, Kurth T, Ligthart L, Terwindt GM, Pakstis A, Paschou P, Piacentini J, Pittenger C, Plessen K, Ramensky V,
Freilinger T, Ran C, Gordon SD, Borck G, Adams HHH, Lehtimäki T, Ramos EM, Reus V, Richter MA, Riddle MA, Robertson MM,
Wedenoja J, Buring JE, Schürks M, Hrafnsdottir M, Hottenga J-J, Roessner V, Rosário M, Samuels JF, Sandor P, Stein DJ, Tsetsos F, Van
Penninx B, Artto V, Kaunisto M, Vepsäläinen S, Martin NG, Nieuwerburgh F, Weatherall S, Wendland JR, Wolanczyk T, Worbe Y,
Montgomery GW, Kurki MI, Hämäläinen E, Huang H, Huang J, Zai G, Goes FS, McLaughlin N, Nestadt PS, Grabe H-J, Depienne C,
Sandor C, Webber C, Muller-Myhsok B, Schreiber S, Salomaa V, Konkashbaev A, Lanzagorta N, Valencia-Duarte A, Bramon E,
Loehrer E, Göbel H, Macaya A, Pozo-Rosich P, Hansen T, Werge T, Buccola N, Cahn W, Cairns M, Chong SA, Cohen D, Crespo-Facorro B,
Kaprio J, Metspalu A, Kubisch C, Ferrari MD, Belin AC, van den Crowley J, Davidson M, DeLisi L, Dinan T, Donohoe G, Drapeau E,
Maagdenberg AMJM, Zwart J-A, Boomsma D, Eriksson N, Olesen J, Duan J, Haan L, Hougaard D, Karachanak-Yankova S, Khrunin A,
Chasman DI, Nyholt DR, Anney R, Avbersek A, Baum L, Berkovic S, Klovins J, Kučinskas V, Chee Keong JL, Limborska S, Loughland C,
Bradfield J, Buono R, Catarino CB, Cossette P, De Jonghe P, Lönnqvist J, Maher B, Mattheisen M, McDonald C, Murphy KC,
Depondt C, Dlugos D, Ferraro TN, French J, Hjalgrim H, Jamnadas- Murray R, Nenadic I, van Os J, Pantelis C, Pato M, Petryshen T,
Khoda J, Kälviäinen R, Kunz WS, Lerche H, Leu C, Lindhout D, Quested D, Roussos P, Sanders AR, Schall U, Schwab SG, Sim K, So H-
Lo W, Lowenstein D, McCormack M, Møller RS, Molloy A, Ng P-W, C, Stögmann E, Subramaniam M, Toncheva D, Waddington J,
Oliver K, Privitera M, Radtke R, Ruppert A-K, Sander T, Schachter S, Walters J, Weiser M, Cheng W, Cloninger R, Curtis D, Gejman PV,
Schankin C, Scheffer I, Schoch S, Sisodiya SM, Smith P, Sperling M, Henskens F, Mattingsdal M, Oh S-Y, Scott R, Webb B, Breen G,
Striano P, Surges R, Thomas GN, Visscher F, Whelan CD, Zara F, Churchhouse C, Bulik CM, Daly M, Dichgans M, Faraone SV,
Heinzen EL, Marson A, Becker F, Stroink H, Zimprich F, Gasser T, Guerreiro R, Holmans P, Kendler KS, Koeleman B, Mathews CA,
Gibbs R, Heutink P, Martinez M, Morris HR, Sharma M, Ryten M, Price A, Scharf J, Sklar P, Williams J, Wood NW, Cotsapas C, Palotie A,
Mok KY, Pulit S, Bevan S, Holliday E, Attia J, Battey T, Smoller JW, Sullivan P, Rosand J, Corvin A and Neale BM (2018)
Boncoraglio G, Thijs V, Chen W-M, Mitchell B, Rothwell P, Analysis of shared heritability in common disorders of the brain. Science
Sharma P, Sudlow C, Vicente A, Markus H, Kourkoulis C, Pera J, 360, eaap8757.
Raffeld M, Silliman S, Boraska Perica V, Thornton LM, Huckins LM, Beesdo K, Knappe S and Pine DS (2009) Anxiety and anxiety disorders in
William Rayner N, Lewis CM, Gratacos M, Rybakowski F, Keski- children and adolescents: developmental issues and implications for
Rahkonen A, Raevuori A, Hudson JI, Reichborn-Kjennerud T, DSM-V. Psychiatric Clinics of North America 32, 483–524.
Monteleone P, Karwautz A, Mannik K, Baker JH, O’Toole JK, Belbasis L, Köhler CA, Stefanis N, Stubbs B, van Os J, Vieta E, Seeman MV,
Trace SE, Davis OSP, Helder SG, Ehrlich S, Herpertz-Dahlmann B, Arango C, Carvalho AF and Evangelou E (2018) Risk factors and periph-
Danner UN, van Elburg AA, Clementi M, Forzan M, Docampo E, eral biomarkers for schizophrenia spectrum disorders: an umbrella review
Lissowska J, Hauser J, Tortorella A, Maj M, Gonidakis F, Tziouvas K, of meta-analyses. Acta Psychiatrica Scandinavica 137, 88–97.
Papezova H, Yilmaz Z, Wagner G, Cohen-Woods S, Herms S, Julià A, Bernardini F, Attademo L, Cleary SD, Luther C, Shim RS, Quartesan R and
Rabionet R, Dick DM, Ripatti S, Andreassen OA, Espeseth T, Compton MT (2017) Risk prediction models in psychiatry. The Journal of
Lundervold AJ, Steen VM, Pinto D, Scherer SW, Aschauer H, Clinical Psychiatry 78, 572–583.
Schosser A, Alfredsson L, Padyukov L, Halmi KA, Mitchell J, Blanco C, Rubio J, Wall M, Wang S, Jiu CJ and Kendler KS (2014) Risk fac-
Strober M, Bergen AW, Kaye W, Szatkiewicz JP, Cormand B, Ramos- tors for anxiety disorders: common and specific effects in a national sample.
Quiroga JA, Sánchez-Mora C, Ribasés M, Casas M, Hervas A, Depression and Anxiety 31, 756–764.
Arranz MJ, Haavik J, Zayats T, Johansson S, Williams N, Elia J, Bortolato B, Köhler CA, Evangelou E, León-Caballero J, Solmi M, Stubbs B,
Dempfle A, Rothenberger A, Kuntsi J, Oades RD, Banaschewski T, Belbasis L, Pacchiarotti I, Kessing LV, Berk M, Vieta E and Carvalho AF
Franke B, Buitelaar JK, Arias Vasquez A, Doyle AE, Reif A, Lesch K-P, (2017) Systematic assessment of environmental risk factors for bipolar dis-
Freitag C, Rivero O, Palmason H, Romanos M, Langley K, Rietschel M, order: an umbrella review of systematic reviews and meta-analyses. Bipolar
Witt SH, Dalsgaard S, Børglum AD, Waldman I, Wilmot B, Molly N, Disorders 19, 84–96.
Bau CHD, Crosbie J, Schachar R, Loo SK, McGough JJ, Grevet EH, Brander G, Pérez-Vigil A, Larsson H and Mataix-Cols D (2016a) Systematic
Medland SE, Robinson E, Weiss LA, Bacchelli E, Bailey A, Bal V, review of environmental risk factors for obsessive-compulsive disorder: a
Battaglia A, Betancur C, Bolton P, Cantor R, Celestino-Soper P, proposed roadmap from association to causation. Neuroscience &
Dawson G, De Rubeis S, Duque F, Green A, Klauck SM, Leboyer M, Biobehavioral Reviews 65, 36–62.
Levitt P, Maestrini E, Mane S, De-Luca DM, Parr J, Regan R, Brander G, Rydell M, Kuja-Halkola R, Fernández de la Cruz LF,
Reichenberg A, Sandin S, Vorstman J, Wassink T, Wijsman E, Cook E, Lichtenstein P, Serlachius E, Rk C, Almqvist C, D’Onofrio BM,
Santangelo S, Delorme R, Rogé B, Magalhaes T, Arking D, Schulze TG, Larsson H and Mataix-Cols D (2016b) Association of perinatal risk factors
Thompson RC, Strohmaier J, Matthews K, Melle I, Morris D, with obsessive-compulsive disorder a population-based birth cohort, sibling
Blackwood D, McIntosh A, Bergen SE, Schalling M, Jamain S, Maaser A, control study. JAMA Psychiatry 73, 1135–1144.
Fischer SB, Reinbold CS, Fullerton JM, Grigoroiu-Serbanescu M, Brown ES, Fulton MK, Wilkeson A and Petty F (2000) The psychiatric
Guzman-Parra J, Mayoral F, Schofield PR, Cichon S, Mühleisen TW, sequelae of civilian trauma. Comprehensive Psychiatry 41, 19–23.
Degenhardt F, Schumacher J, Bauer M, Mitchell PB, Gershon ES, Rice J, Clarner A, Graessel E, Scholz J, Niedermeier A, Uter W and Drexler H
Potash JB, Zandi PP, Craddock N, Ferrier IN, Alda M, Rouleau GA, (2015) Work-related posttraumatic stress disorder (PTSD) and other emo-
Turecki G, Ophoff R, Pato C, Anjorin A, Stahl E, Leber M, Czerski PM, tional diseases as consequence of traumatic events in public transportation:
Edenberg HJ, Cruceanu C, Jones IR, Posthuma D, Andlauer TFM, a systematic review. International Archives of Occupational and
Forstner AJ, Streit F, Baune BT, Air T, Sinnamon G, Wray NR, Environmental Health 88, 549–564.
Psychological Medicine 15

Clauss JA and Blackford JU (2012) Behavioral inhibition and risk for devel- Kovess-Masfety V, Lee S, Medina-Mora ME, Murphy SD,
oping social anxiety disorder: a meta-analytic study. Journal of the Navarro-Mateu F, Piazza M, Posada-Villa J, Scott K, Torres Y and
American Academy of Child and Adolescent Psychiatry 51, 1066–1075. Viana MC (2014) How well can post-traumatic stress disorder be predicted
Colonnesi C, Draijer EM, Stams GJJM, van der Bruggen CO, Bögels SM from pre-trauma risk factors? An exploratory study in the WHO World
and Noom MJ (2011) The relation between insecure attachment and Mental Health Surveys. World Psychiatry 13, 265–274.
child anxiety: a meta-analytic review. Journal of Clinical Child and Kisely S, Alichniewicz KK, Black EB, Siskind D, Spurling G and Toombs M
Adolescent Psychology 40, 630–645. (2017) The prevalence of depression and anxiety disorders in indigenous
Craske MG and Stein MB (2016) Anxiety. The Lancet 388, 3048–3059. people of the Americas: a systematic review and meta-analysis. Journal of
Craske MG, Stein MB, Eley TC, Milad MR, Holmes A, Rapee RM and Psychiatric Research 84, 137–152.
Wittchen H-U (2017) Anxiety disorders. Nature Reviews Disease Primers Köhler CA, Evangelou E, Stubbs B, Solmi M, Veronese N, Belbasis L,
3, 17024. Bortolato B, Melo MCA, Coelho CA, Fernandes BS, Olfson M,
D’Onofrio BM, Lahey BB, Turkheimer E and Lichtenstein P (2013) Critical Ioannidis JPA and Carvalho AF (2018) Mapping risk factors for depres-
need for family-based, quasi-experimental designs in integrating genetic sion across the lifespan: an umbrella review of evidence from meta-analyses
and social science research. American Journal of Public Health 103, S46–S55. and Mendelian randomization studies. Journal of Psychiatric Research 103,
Egger M, Davey Smith G, Schneider M and Minder C (1997) Bias in 189–207.
meta-analysis detected by a simple, graphical test. British Medical Journal Kossowsky J, Pfaltz MC, Schneider S, Taeymans J, Locher C and Gaab J
315, 629–634. (2013) The separation anxiety hypothesis of panic disorder revisited: a
Fernandes V, Osório FL, Osó rio F, Preto R and Paulo S (2015) Are there meta-analysis. American Journal of Psychiatry 170, 768–781.
associations between early emotional trauma and anxiety disorders? Kotov R, Gamez W, Schmidt F and Watson D (2010) Linking ‘ Big ’ person-
Evidence from a systematic literature review and meta-analysis. European ality traits to anxiety, depressive, and substance use disorders: a
Psychiatry 30, 756–764. meta-analysis. Psychological Bulletin 136, 768–821.
Fusar-Poli P, Hijazi Z, Stahl D and Steyerberg EW (2018) The science of Kraemer HC, Kazdin AE, Offord DR, Kessler RC, Jensen PS and Kupfer DJ
prognosis in psychiatry. JAMA Psychiatry 75, 1289–1297. (1997) Coming to terms with the terms of risk. Archives of General Psychiatry
Fusar-Poli P and Radua J (2018) Ten simple rules for conducting umbrella 54, 337–343.
reviews. Evidence Based Mental Health 21, 95–100. Leonardo ED and Hen R (2008) Anxiety as a developmental disorder.
Gariepy G, Nitka D and Schmitz N (2010) The association between obesity Neuropsychopharmacology 33, 134–140.
and anxiety disorders in the population: a systematic review and Li M, D’Arcy C and Meng X (2016) Maltreatment in childhood substantially
meta-analysis. International Journal of Obesity (2005) 34, 407–419. increases the risk of adult depression and anxiety in prospective cohort
Ginsburg GS, Drake KL, Tein J-Y, Teetsel R and Riddle MA (2015) studies: systematic review, meta-analysis, and proportional attributable frac-
Preventing onset of anxiety disorders in offspring of anxious parents: a ran- tions. Psychological Medicine 46, 717–730.
domized controlled trial of a family-based intervention HHS public access. Micco JA, Henin A, Mick E, Kim S, Hopkins CA, Biederman J and
American Journal of Psychiatry 172, 1207–1214. Hirshfeld-Becker DR (2009) Anxiety and depressive disorders in offspring
Guo X, Meng Z, Huang G, Fan J, Zhou W, Ling W, Jiang J, Long J and Su L at high risk for anxiety: a meta-analysis. Journal of Anxiety Disorders 23,
(2016) Meta-analysis of the prevalence of anxiety disorders in mainland 1158–1164.
China from 2000 to 2015. Scientific Reports 6, 28033. Moffitt TE, Caspi A and Rutter M (2005) Strategy for investigating interac-
Higgings JPT and Green S (2009). Cochrane {H}andbook for {S}ystematic {R} tions between measured genes and measured environments. Archives of
eviews of {I}nterventions. Version 5.1.0. General Psychiatry 62, 473–481.
International Schizophrenia Consortium IS, Purcell SM, Wray NR, Moher D, Liberati A, Tetzlaff J and Altman DG (2009) Preferred reporting
Stone JL, Visscher PM, O’Donovan MC, Sullivan PF and Sklar P items for systematic reviews and meta-analyses: the PRISMA statement.
(2009). Common polygenic variation contributes to risk of schizophrenia BMJ 339, b2535–b2535.
and bipolar disorder. Nature 460, 748–752. Moreno-Peral P, Conejo-Cerón S, Motrico E, Rodríguez-Morejón A,
Ioannidis JPA (2009) Integration of evidence from multiple meta-analyses: a Fernández A, García-Campayo J, Roca M, Serrano-Blanco A,
primer on umbrella reviews, treatment networks and multiple treatments Rubio-Valera M and Bellón JÁ (2014) Risk factors for the onset of panic
meta-analyses. Canadian Medical Association Journal 181, 488–493. and generalised anxiety disorders in the general adult population: a systematic
Ioannidis JPA, Patsopoulos NA and Evangelou E (2007) Uncertainty in het- review of cohort studies. Journal of Affective Disorders 168, 337–348.
erogeneity estimates in meta-analyses. BMJ 335, 914–916. Moreno-Peral P, Conejo-Cerón S, Rubio-Valera M, Fernández A,
Ioannidis JPA, Boffetta P, Little J, O’Brien TR, Uitterlinden AG, Vineis P, Navas-Campaña D, Rodríguez-Morejón A, Motrico E, Rigabert A, De
Balding DJ, Chokkalingam A, Dolan SM, Flanders WD, Higgins JPT, Dios Luna J, Martín-Pérez C, Rodríguez-Bayón A, Ballesta-Rodríguez MI,
Mccarthy MI, McDermott DH, Page GP, Rebbeck TR, Seminara D and Luciano JV and Bellón JÁ (2017) Effectiveness of psychological and/or educa-
Khoury MJ (2008) Assessment of cumulative evidence on genetic associa- tional interventions in the prevention of anxiety: a systematic review,
tions: interim guidelines. International Journal of Epidemiology 37, 120–132. meta-analysis, and meta-regression. JAMA Psychiatry 74, 1021–1029.
Jacobson NC and Newman MG (2017) Anxiety and depression as bidirec- Moylan S, Jacka FN, Pasco JA and Berk M (2012) Cigarette smoking, nico-
tional risk factors for one another: a meta-analysis of longitudinal studies. tine dependence and anxiety disorders: a systematic review of population-
Psychological Bulletin 143, 1155–1200. based, epidemiological studies. BMC Medicine 10, 123.
Kapur S, Phillips AG and Insel TR (2012) Why has it taken so long for bio- Mrazek PJ and Haggerty RJ (1994) Reducing Risks for Mental Disorders:
logical psychiatry to develop clinical tests and what to do about it? Frontiers for Preventive Intervention Research. vol. 636. Washington, DC:
Molecular Psychiatry 17, 1174–1179. National Academies Press.
Kedzior KK and Laeber LT (2014) A positive association between anxiety dis- Osborn A, Mathias J and Fairweather-Schmidt A (2016) Prevalence of anx-
orders and cannabis use or cannabis use disorders in the general iety following adult traumatic brain injury: a meta-analysis comparing mea-
population--a meta-analysis of 31 studies. BMC Psychiatry 14, 136. sures, samples and postinjury intervals. Neuropsychology 30, 247–261.
Kessler RC, Ruscio AM, Shear K and Wittchen H-U (2010) Epidemiology of Padmanabhan JL, Shah JL, Tandon N and Keshavan MS (2017) The “poly-
anxiety disorders. Current Topics in Behavioral Neurosciences 2, 21–35. enviromic risk score”: aggregating environmental risk factors predicts con-
Kessler RC, Rose S, Koenen KC, Karam EG, Stang PE, Stein DJ, version to psychosis in familial high-risk subjects. Schizophrenia Research
Heeringa SG, Hill ED, Liberzon I, McLaughlin KA, McLean SA, 181, 17–22.
Pennell BE, Petukhova M, Rosellini AJ, Ruscio AM, Shahly V, Paulus MP (2015) Pragmatism instead of mechanism: a call for impactful bio-
Shalev AY, Silove D, Zaslavsky AM, Angermeyer MC, Bromet EJ, De logical psychiatry. JAMA Psychiatry 72, 631–632.
Almeida JMC, De Girolamo G, De Jonge P, Demyttenaere K, Radua J, Ramella-Cravaro V, Ioannidis JPA, Reichenberg A,
Florescu SE, Gureje O, Haro JM, Hinkov H, Kawakami N, Phiphopthatsanee N, Amir T, Yenn Thoo H, Oliver D, Davies C,
16 Miquel A. Fullana et al.

Morgan C, McGuire P, Murray RM and Fusar-Poli P (2018) What causes Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D,
psychosis? An umbrella review of risk and protective factors. World Moher D, Becker BJ, Sipe TA and Thacker SB (2000) Meta-analysis of
Psychiatry 17, 49–66. observational studies in epidemiology: a proposal for reporting. Journal of
Reich J, Noyes R, Hirschfeld R, Coryell W and O’Gorman T (1987) State and the American Medical Association 283, 2008–2012.
personality in depressed and panic patients. American Journal of Psychiatry Sullivan PF, Agrawal A, Bulik CM, Andreassen OA, Børglum AD, Breen G,
144, 181–187. Cichon S, Edenberg HJ, Faraone SV, Gelernter J, Mathews CA,
Shalev AY, Gevonden M, Ratanatharathorn A, Laska E, van der Mei WF, Qi W, Nievergelt CM, Smoller JW and O’Donovan MC (2018) Psychiatric genom-
Lowe S, Lai BS, Bryant RA, Delahanty D, Matsuoka YJ, Olff M, Schnyder U, ics: an update and an agenda. American Journal of Psychiatry 175, 15–27.
Seedat S, DeRoon-Cassini TA, Kessler RC and Koenen KC (2019) Estimating Tarricone I, Stivanello E, Poggi F, Castorini V, Marseglia MV, Fantini MP
the risk of PTSD in recent trauma survivors: results of the International and Berardi D (2012) Ethnic variation in the prevalence of depression and
Consortium to Predict PTSD (ICPP). World Psychiatry 18, 77–87. anxiety in primary care: a systematic review and meta-analysis. Psychiatry
Sharma S, Powers A, Bradley B and Ressler KJ (2016) Gene × environment Research 195, 91–106.
determinants of stress- and anxiety-related disorders. Annual Review of Uher R and Zwicker A (2017) Etiology in psychiatry: embracing the reality of
Psychology 67, 239–261. poly-gene-environmental causation of mental illness. World Psychiatry 16,
Shea BJ, Grimshaw JM, Wells GA, Boers M, Andersson N, Hamel C, 121–129.
Porter AC, Tugwell P, Moher D and Bouter LM (2007) Development of van Steensel FJA, Bögels SM and Perrin S (2011) Anxiety disorders in chil-
AMSTAR: a measurement tool to assess the methodological quality of sys- dren and adolescents with autistic spectrum disorders: a meta-analysis.
tematic reviews. BMC Medical Research Methodology 7, 10. Clinical Child and Family Psychology Review 14, 302–317.