Introduction

Work on the epidemiology of common mental disorders is intimately intertwined with research on their nosology. Early epidemiological surveys used simple self-report measures of psychiatric symptoms.1 The publication of operational diagnostic criteria in the third edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM) provided the foundation for a subsequent generation of epidemiological surveys that relied on these criteria to determine caseness. The ECA (Epidemiological Catchment Area) and the NCS (National Comorbidity Survey) studies in the United States, together with similar studies in other countries, provided key information on the prevalence and burden of conditions and demonstrated that anxiety disorders were the most prevalent class of psychiatric disorders.

Ongoing developments in both epidemiology and nosology have contributed to a growing dialogue between these fields. Epidemiological surveys based on DSM criteria have been undertaken in a number of different countries and contexts. The DSM has been revised to reflect advances in our understanding of psychiatric entities and is currently available in its fifth edition; the International Classification of Disease (ICD) is currently under revision and will appear in its 11th edition. In this review, we focus on current intersections between epidemiology and nosology, with a particular emphasis on the anxiety disorders and especially generalized anxiety disorder (GAD).

We address a number of key questions facing the field. The first questions are about prevalence. What is the prevalence of anxiety disorders? Are differences in prevalence estimates across the globe merely artefactual, or do they hold relevant clinical lessons? The second set of questions concerns diagnostic criteria. What are the optimal criteria for anxiety disorders? What are the epidemiological implications of changes in criteria? What are the optimal criteria that differentiate anxiety disorders from normality? The third set is about comorbidity. What are the comorbidities of anxiety disorders? Are high comorbidities of anxiety disorders merely artefactual, or do they have implications for our understanding of these conditions? Finally, the fourth set is about treatment gap and health policy. What is the current treatment gap for anxiety disorders? What are the implications of the current understanding of psychiatric epidemiology and nosology for policy-making relevant to anxiety disorders?

Prevalence of anxiety disorders

The ECA study, the first modern epidemiological survey to use a fully structured research diagnostic interview to operationalize DSM criteria, was carried out in the United States in 1980-1985. The ECA found that certain anxiety disorders were highly prevalent.2,3 The subsequent NCS study, another US survey, but in this case based on a national sample (rather than a series of local samples in the ECA) and including the full range of DSM anxiety disorders (rather than a subset of these disorders in the ECA), indicated that anxiety disorders were the most prevalent class of psychiatric conditions, although this reflects, in part, the fact that this is a relatively large class of conditions.4,5 With the initiation of the World Mental Health Surveys (WMHS), a cross-national expansion of the NCS, this finding was confirmed; in multiple countries, anxiety disorders are more prevalent than mood disorders, substance use disorders, and impulse control disorders.6,7 At the same time, the WMHS has found that prevalence of anxiety disorders differs markedly from country to country. Thus, for example, lifetime prevalence estimates of DSM-IV GAD range from 0.1% in Nigeria to 6.2% in New Zealand. Systematic reviews and meta-regressions have confirmed differences in prevalence of anxiety disorders around the globe, have emphasized that there is considerable uncertainty around estimates, and have suggested a current global prevalence of anxiety disorders of 7.3% (4.8% to 10.9%).8,9

There are several potential explanations for the wide cross-national variation in anxiety disorder prevalence estimates. First, differences across countries in attitudes toward mental illness may express themselves in reluctance to admit to symptoms of psychiatric disorders.10 This issue may be particularly relevant to anxiety disorders, where symptoms may cause a good deal of embarrassment.11 Second, psychiatric disorders may be expressed and experienced in different ways across contexts, and survey questionnaires may not sufficiently address this variation.12 The boundary between (normal) distress and (pathological) disorder may be particularly fuzzy when emotional responses, such as anxiety, are potentially adaptive within a particular context.13 Third, the diagnostic interviews used in epidemiological surveys can contain errors in the ways questions are phrased that introduce biases that vary across countries. Fourth, the field conditions in these surveys (eg, rigor of interviewer training, quality control monitoring, and sampling) can vary across countries in ways that introduce meaningful differences in prevalence estimates. Fifth, there may be real differences in the prevalence of mental disorders across countries, reflecting important differences in risk and resilience factors across geography.

The WMHS undertook several methodological approaches to address these issues.14 The Composite International Diagnostic Interview (CIDI), the diagnostic interview used in the WMHS, was carefully designed to maximize accuracy of respondent answers. Features of the CIDI that aim to improve the reliability and validity of the data obtained from lay-administered interview include efforts to increase respondent understanding, motivation, and ability to provide accurate survey responses. The DSM has been revised over time in an evidence-based attempt to be applicable to a range of different contexts,15,16 and updated versions of the CIDI have incorporated these changes. Clinical reappraisal studies have supported the validity of the CIDI, with good concordance for a CIDI diagnosis of any anxiety disorder with clinical diagnosis (area under the curve, 0.88).17 Nevertheless, there are limited data on concordance from low- and middle-income countries in the WMHS, where prevalence estimates of anxiety disorders have sometimes been implausibly low and require cautious interpretation.

Despite differences in prevalence of anxiety disorders across countries, several other WMHS findings suggest that there are important universal features of these conditions. Thus, for example, across WMHS, (i) certain anxiety disorders are more prevalent (eg, specific phobia), whereas others are less prevalent (eg, agoraphobia without a history of panic disorder); (ii) anxiety disorders have an earlier age of onset than do mood disorders, substance use disorders, and impulse control disorders (with such onset ranging from modal onset in childhood in specific phobia to modal onset in adolescence or early adulthood in social phobia, with somewhat later and more widely dispersed distributions of onset in panic disorder, agoraphobia, and GAD), and the course is often chronic-recurrent (with relatively higher ratios of 12-month to lifetime prevalence than for other common psychiatric disorders); (iii) anxiety disorders have typical sociodemographic correlates across the globe (eg, female sex); (iv) anxiety disorders are highly comorbid with one another and with other mental disorders; and (v) first treatment of anxiety disorders usually does not occur until at least a decade after onset, even in well-resourced countries.7

It is also notable that epidemiological data have drawn attention to diagnostic entities that have been relatively overlooked by clinicians. A key exemplar of this is separation anxiety disorder. Although clinical reports have long appeared in the literature, these primarily focused on separation anxiety in children, with less attention paid to adults.18 WMHS data convincingly demonstrated that this condition is found in adults and that it is prevalent around the globe.19 Around one third of childhood-onset cases persisted into adulthood, whereas the majority of adult respondents in WMHS reported adult onset of symptoms. These data provide solid support for the decision to include separation anxiety disorder as a new entity in the DSM-5 chapter on anxiety disorders.

Our view is that it is problematic to draw conclusive lessons from apparent differences in prevalence estimates across countries. Despite rigorous revisions of both diagnostic criteria and survey methodology, regional differences in willingness to endorse mental symptoms in general and to answer probes for anxiety symptoms in particular, probably affect calculated prevalence estimates. Cross-national differences in field conditions in the WMHS have probably also played a part in cross-national differences in prevalence estimates, as indicated by the existence of a significant cross-national association between survey response rate and disorder prevalence estimates. In light of these observations, it might be more prudent to focus on the more cross-nationally stable patterns of within-survey associations to draw lessons about universal features of anxiety disorders (eg, sociodemographic correlates, course, and comorbidity).7

Optimizing diagnostic criteria

Efforts have been made to ensure revisions to the DSM are evidence based where possible; any proposed change must reflect the relevant data on diagnostic validity and clinical utility.15,16 However, for many conditions, there is often a relative paucity of research contrasting and comparing alternative diagnostic criteria sets.20,21 Clinical field trials provide one opportunity to undertake such work, but relatively few DSM-5 field trials used this sort of design.22,23 In addition, analysis of data from epidemiological surveys may be useful in contributing to the evaluation of different diagnostic criteria sets; respondents who do and do not meet such criteria sets can be compared to determine whether or not they have different risk profiles or clinical characteristics.

An analysis of data from the NCS-R (National Comorbidity Survey Replication) assessed the impact of removing the excessive worry requirement; nonexcessive worriers meeting all other DSM-IV criteria for GAD were compared with respondents who met full GAD criteria, as well as with other survey respondents.24 The estimated lifetime prevalence of GAD increased by approximately 40% when the excessive worry requirement was removed. Compared with nonexcessive-worry GAD, excessive-worry GAD begins earlier in life, has a more chronic course, and is associated with greater symptom severity and psychiatric comorbidity. However, compared with respondents without GAD, nonexcessive cases have substantial persistence and impairment of GAD, high rates of treatment seeking, and significantly elevated comorbidity. Furthermore, nonexcessive cases and excessive cases have comparable sociodemographic characteristics and family history of GAD. Taken together, these findings challenge the validity of the DSM-IV excessiveness requirement.

An analysis of data from the NCS-R also assessed the impact of relaxing the DSM-IV requirements for GAD of 6-month duration, excessive worry, and three associated symptoms.25 Relaxing all three criteria led to a more than doubling of the prevalence estimates of GAD. Nevertheless, broadly defined GAD significantly predicts the subsequent onset of a wide range of temporally defined secondary disorders. The odds of secondary disorders are somewhat smaller for broadly defined GAD than for DSM-IV GAD, but few of these differences are statistically significant. Once again, then, these data suggest that from an epidemiological perspective, broadening the DSM-IV diagnostic criteria may increase the validity of GAD diagnosis.

Analysis of WMHS data allowed an investigation of such diagnostic decisions across the globe.26 Lifetime prevalence estimates for GAD lasting 1 month, 3 months, 6 months, and 12 months were higher in developed than in developing countries, but in both country groups, prevalence decreased with increasing duration (7.5%, 5.2%, 4.1%, and 3% for developed countries and 2.7%, 1.8%, 1.5%, and 1.2% for developing countries, respectively). There was little difference between GAD of 6 months' duration and GAD of shorter durations (1-2 months, 3-5 months) in key characteristics (age of onset, symptom severity or persistence, comorbidity, impairment). These analyses suggested that the clinical profile of GAD is similar across the globe, regardless of duration. Furthermore, the DSM-IV 6-month duration criterion excludes many individuals who present with shorter episodes, which may nevertheless be recurrent and impairing.

At the same time, as discussed in more detail in the next section, epidemiological data have been important in emphasizing the validity and the independence of the GAD diagnosis.27 Taken together, the epidemiological data suggest that although GAD is a valid diagnostic entity, there is potential for further refining several aspects of the relevant diagnostic criteria. It is crucially important to remember that the DSM provides only one perspective on complex psychiatric disorders and that a range of alternative diagnostic sets may be useful for clinicians to be aware of.20 There is, for example, some evidence that mental and somatic symptoms of “tension” may be particularly worth foregrounding in the operationalization of GAD.28 At the same time, thresholds for change in DSM-5 were understandably conservative,16 and GAD remained largely unchanged in this edition of the nosology.

Epidemiological data have also shed light on the implications of alternative decisions about where to draw thresholds, for example, between a disorder and normality, or between subtypes of disorders. Mild disorders and subthreshold cases may, for example, be associated with considerable impairment, and thus important to recognize.25,29 It is noteworthy that in general medicine, measures such as cholesterol level fall on a continuum, and thresholds for defining hypercholesterolemia are influenced by data on cost-efficiency of treatment (with the introduction of statins and then of generic statins leading to progressive lowering of thresholds).30

In anxiety disorders such as GAD, community data also suggest that disorder and subthreshold disorder fall on a continuum.25,31 In DSM-IV, social anxiety disorder (SAD) is classified into generalized and nongeneralized subtypes. WMHS data found, however, that in those with SAD, there is a dose-response relationship between number of fears and persistence, severity, comorbidity, and treatment. Thus, there was no clear evidence of a distinction between generalized and nongeneralized SAD. Whereas DSM-5 has included a performance subtype of SAD, WMHS found no evidence to support subtyping on the basis of the number of performance fears versus the number of interactional fears.32 For those with threshold anxiety disorder, there is increasing evidence of the cost-efficiency of treatment.33

Comorbidity and impairment

The NCS, the NCS-R, the WMHS, and a range of clinical surveys have provided a great deal of epidemiological data on the comorbidity of mental disorders, including anxiety disorders. Key findings from this epidemiological research are that anxiety disorders very often precede the onset of other psychiatric disorders,7 that anxiety symptoms may be a predictor of worse outcome (eg, suicidality in patients with depression),34 that anxiety disorders are associated with substantial individual impairment (including reduced educational attainment, more unstable marriage, lower occupational status) and staggering direct and indirect economic costs,2,4 and that certain anxiety disorders (eg, specific phobia) are important in predicting onset or course of other conditions.35 The Global Burden of Disease study found that in 2010, anxiety disorders were the sixth leading cause of disability in terms of years of life lived with disability in both high-income and in low- and middle-income countries, accounting for 390 disability-adjusted life years per 100 000 persons (95% uncertainty interval, 191-371), with highest burden in women and in those aged 15 to 34 years, but with no change over time and no identifiable differences in burdens across regions.36 These epidemiological findings have immediate clinical relevance insofar as they underscore the potential value of early diagnosis and intervention for anxiety symptoms and disorders.37

Several factors may contribute to the high comorbidity of mental disorders, including anxiety disorders. First, these conditions may share important risk factors, so that patients with one condition are more likely to also develop a second. Second, having a psychiatric disorder may itself be a risk factor for a second disorder; for example, individuals with anxiety disorders may self-medicate with alcohol and eventually develop alcohol dependence. Third, comorbidity may be artefactual38; patients with major depression, for example, may have significant anxiety and so may be inappropriately conceptualized as having a comorbid anxiety disorder. Epidemiological data on comorbidity have made a number of important contributions to teasing out these possibilities and to understanding the nosology of anxiety disorders.

As alluded to above, an important early nosological debate was whether GAD deserved recognition as an independent diagnostic entity. Critics emphasized that GAD was rarely seen in psychiatric practice, that its symptoms overlapped with those of depression, and that GAD patients invariably had comorbidity. However, epidemiological data demonstrated that GAD does not have a higher comorbidity than most other mood or anxiety disorders, that the symptoms of GAD form a cluster separate from the symptoms of depression, that the sociodemographic predictors of GAD differ from those of major depression, that the clinical course of GAD is less consistently related to comorbidity than is the course of depression and other anxiety disorders, and that the impairments associated with GAD are no less than those associated with other severely impairing mental and chronic physical disorders.27

Epidemiological data on comorbidity have also influenced nosological conceptualizations of panic disorder and agoraphobia. An influential clinical perspective has emphasized that many patients with panic attacks go on to develop panic disorder and then subsequently develop agoraphobia.39,40 Epidemiological data have, however, demonstrated that panic attacks in the absence of panic disorder are prevalent and impairing and that agoraphobia often exists as an independent disorder.39,40 These findings have provided important support for the decisions in DSM-5 and ICD-11 to provide separate codes for panic attacks (which may be diagnosed in a range of different mental disorders) and for agoraphobia.

As noted earlier, specific phobia is a predictor of the course of GAD.35 Data from the International Consortium on Psychiatric Epidemiology combined data from Brazil, Canada, the Netherlands, and the United States.35 Six disorders predict first onset of GAD in all four surveys: agoraphobia, panic disorder, simple phobia, dysthymia, major depression, and mania. Respondents with active disorders have elevated risk of GAD, but in the case of specific phobia, respondents with a history of remitted disorder also have elevated risk of GAD. Specific phobia is also the only disorder that predicts the persistence of GAD. Taken together, these data indicate that specific phobia is a risk marker for GAD, a hypothesis that is consistent with biological data on individual differences in fear conditioning and risk for GAD.41

Factor-analytic studies have suggested that anxiety and mood disorders have high factor loadings on an “internalizing” dimension, with some studies also indicating that phobias and panic fall onto a secondary “fear” dimension whereas GAD and depression fall onto a secondary “distress” dimension. WMHS data have extended this work by focusing on the role of latent variables in the development of comorbidity.42 A two-factor internalizing-externalizing structure was found, but with no evidence for a distinction between fear and distress disorders (a finding supportive of the DSM-5 and ICD-11 decision not to group GAD and depression together). Associations between different disorders were stronger and more consistent within rather than between the internalizing and externalizing domains, and the vast majority of associations were explained by a model that assumed the existence of mediating latent internalizing and externalizing variables. Specific phobia stood out as the most robust predictor of internalizing disorders, consistent with the view that it may be a useful risk marker.

Epidemiological research has also emphasized the importance of comorbidity between anxiety disorders and physical disorders, with some inconsistent evidence of an association between anxiety disorders and increased mortality.43,44 Data from the WMHS indicates associations of anxiety and mood disorders with a range of subsequent chronic physical disorders.45,46 Further WMHS analyses have indicated that childhood adversities and early onset mental disorders have independent, broad-spectrum effects that increase the risk of such conditions later in life.47 Associations between anxiety disorders and subsequent coronary heart disease seem particularly robust,48,49 and there is also emerging evidence of associations of anxiety with stroke and diabetes.50 A range of psychobiological and behavioral research is attempting to address the precise mechanisms that underlie such associations, and there is also ongoing work directed at developing clinically useful assessment measures and at undertaking rigorous clinical trials in this area.51,52

Treatment gap and health policy

Epidemiological surveys have made a significant contribution to increasing awareness of the burden of mental disorders.53 The Global Burden of Disease consortium has further emphasized the burden of mental disorders compared with infectious disorders and other noncommunicable diseases.54 As emphasized in the previous section, the anxiety disorders are associated with a broad range of profound negative sequelae. The treatment gap for common mental disorders, including the anxiety disorders, is particularly wide in low- and middle-income countries.55 In these regions of the world, there are significant attitudinal and structural barriers to treatment, including somatization, low mental health literacy, and a relative lack of mental health clinicians.10,56

Nevertheless, evidence of the treatment gap for common mental disorders has not necessarily led to appropriate mental health policies. For example, on the African continent, the number of countries with a formally adopted mental health policy remains relatively few.55 There are several potential reasons for the policy and financing gap. Not all policy-makers are convinced of the credibility of the treatment gap, or of the possibility that addressing it will save monies rather than further increase the strain on health budgets. In addition, the prevalence of serious mental illness is relatively low. However, in the case of anxiety and depressive disorders, given associated impairments, there is good evidence of the cost-efficiency of treatment.33

On the one hand, in the case of mild anxiety disorders, data demonstrating cost-efficiency remain relatively limited. Although mild anxiety disorders are certainly impairing, anxiety responses are often adaptive, and supportive interventions are hypothetically more cost-efficient than intensive management. On the other hand, anxiety disorders have an early age of onset and are associated with significant subsequent comorbidity and morbidity; early and robust intervention may therefore be a cost-efficient option. Ultimately, the decision of where to optimally draw thresholds for diagnosis and intervention is one that requires empirical validation.30 Similarly, the study of early interventions to treat anxiety disorders deserves attention in order to clarify whether this has a positive effect on the onset, persistence, or severity of secondary disorders.7

In the interim, several points about the epidemiology and nosology of anxiety disorders should be emphasized to health policy-makers. First, anxiety disorders are often more impairing than physical disorders, yet they are often less likely to be diagnosed and treated.57 Second, anxiety disorders have an early age of onset with a great deal of subsequent morbidity and comorbidity; thus, policies that focus on early intervention may be particularly cost-efficient.58 Third, anxiety disorders are associated not only with other mental disorders, but also with physical disorders; evaluation and treatment in integrative or collaborative care settings is therefore important.59 Fourth, there continue to be high levels of structural and attitudinal barriers to appropriate care for individuals with anxiety disorders; consumer advocacy and government policies addressing mental health literacy and removing such barriers are therefore key.60,61

Conclusion

This review has focused on the relatively narrow set of DSM-5 anxiety disorders, rather than also addressing closely related conditions such as obsessive-compulsive disorder and posttraumatic stress disorder.62 Nevertheless, it is important to emphasize that when taken together, the prevalence, comorbidity, and morbidity of the anxiety and related disorders is particularly high. Furthermore, there have been a series of productive conversations at the intersection of the epidemiology and nosology of obsessive-compulsive spectrum disorders and posttraumatic stress disorders.63-66 Although it is beyond the scope of the current review to address such findings in detail, here too, there has been interactive and iterative nosological and epidemiological progress.67

Our review has emphasized that: (i) rather than overly focusing on cross-national comparisons of prevalence in anxiety disorders, it is more productive to pool survey data and to explore cross-national themes that emerge (eg, about the early onset and chronic course of these conditions); (ii) optimizing diagnostic criteria for anxiety disorders is an iterative process to which epidemiological data can make a crucial contribution; (iii) high comorbidity in the anxiety disorders is not merely artefactual, but rather provides researchers key opportunities to explore pathways to mental disorders and provides clinicians key opportunities to intervene accordingly; and (iv) work on the epidemiology and nosology of anxiety disorders has provided a number of important targets for mental health policy to address.

Psychiatric nosology has come in for a great deal of criticism in recent years; with an influential view being that a neuroscientifically based approach will ultimately lead to improved assessment and treatment. From this perspective, it can be argued that anxiety and its disorders are “low-hanging fruit”; as animal models are potentially productive and as there has already been significant translation between bench and bedside in this area.68 At the same time, it should also be emphasized how important nosology and epidemiology have been in this area, as well as the need for translation between clinical findings and community implementation.69 Indeed, over the next few years of research, integration of nosological, epidemiological, and psychobiological research can be expected, as methods to include genomic data, physiological markers, and experiential sampling, become more widely incorporated into a future generation of community and clinical survey research.