Generalized Anxiety Disorder and Major Depressive Disorder comorbidity in the National Survey of...

9
Research Article GENERALIZED ANXIETY DISORDER AND MAJOR DEPRESSIVE DISORDER COMORBIDITY IN THE NATIONAL SURVEY OF MENTAL HEALTH AND WELL-BEING Caroline Hunt, Ph.D., 1n Tim Slade, Ph.D., 2 and Gavin Andrews, M.D. 2 We report population data on DSM-IV Generalized Anxiety Disorder (GAD) from the Australian National Survey of Mental Health and Well-Being, obtained from a nationwide household survey of adults using a stratified multistage sampling process. A response rate of 78.1% resulted in 10,641 persons being interviewed. Diagnoses were made using the Composite International Diagnostic Interview. The interview was computerised and conducted by trained lay interviewers. We investigated comorbidity between GAD and major depressive disorder (MDD). The results indicate that sociodemographic correlates of GAD, and associated disablement and service use, are influenced by the presence of a comorbid depressive disorder but cannot be fully explained by the presence of that disorder. In addition, GAD was confirmed as significantly disabling, even as a single disorder. We conclude that the results are consistent with the view that GAD has a significant and independent impact on the burden of mental disorders. Depression and Anxiety 20:23–31, 2004. & 2004 Wiley-Liss, Inc. Key words: generalised anxiety disorder; major depression; comorbidity; community sample INTRODUCTION Epidemiological surveys indicate that generalized anxiety disorder (GAD) is one of the more common anxiety disorders [Carter et al., 2001; Hunt et al., 2002; Wittchen et al., 1994]. High comorbidity with other mental disorders is evident and studies of both clinical and community samples have shown that pure cases of GAD are less frequent than cases of GAD with comorbid psychiatric disorders [Bienvenu et al., 1998; Brown and Barlow, 1992; Bruce et al., 2001; Carter et al., 2001; Hunt et al., 2002; Wittchen et al., 1994]. In respect to community samples where help-seeking does not bias sample selection, Judd et al. [1998] reported comorbidity with lifetime unipolar depressive disorder in 67.4% of the National Comorbidity Survey (NCS) sample with a lifetime history of DSM-III-R GAD. Similar patterns of comorbidity have been confirmed using DSM-IV criteria. The German Health Survey (GHS) [Carter at al., 2001] assessed 12-month prevalence and reported that 59.0% of GAD respon- dents had comorbid major depression. In the Austra- lian National Survey of Mental Health and Well-Being 39.3% of individuals with 1-month GAD met criteria for 1-month major depression [Hunt et al., 2002]. High rates of comorbidity and the assumption of low severity and disability have led some to question the diagnostic validity of GAD [e.g., Akiskal, 1998; Barlow and Wincze, 1998; Maser, 1998]. There is growing DEPRESSION AND ANXIETY 20:23–31 (2004) 1 School of Psychology, University of Sydney, Sydney, Australia 2 School of Psychiatry, University of New South Wales, Clinical Research Unit for Anxiety Disorders, New South Wales, Australia n Correspondence to: Dr. Caroline Hunt, School of Psychology (F12), University of Sydney, Sydney, NSW 2006, Australia. E-mail: [email protected] Received for publication 30 August 2003; Revised 6 April 2004; Accepted 1 June 2004 DOI 10.1002/da.20019 Published online 20 August 2004 in Wiley InterScience (www. interscience.wiley.com). & & 2004 WILEY-LISS, INC.

Transcript of Generalized Anxiety Disorder and Major Depressive Disorder comorbidity in the National Survey of...

Page 1: Generalized Anxiety Disorder and Major Depressive Disorder comorbidity in the National Survey of Mental Health and Well-Being

Research Article

GENERALIZED ANXIETY DISORDER ANDMAJOR DEPRESSIVE DISORDER COMORBIDITY

IN THE NATIONAL SURVEY OF MENTAL HEALTH ANDWELL-BEING

Caroline Hunt, Ph.D.,1n Tim Slade, Ph.D.,2 and Gavin Andrews, M.D.2

We report population data on DSM-IV Generalized Anxiety Disorder (GAD)from the Australian National Survey of Mental Health and Well-Being,obtained from a nationwide household survey of adults using a stratifiedmultistage sampling process. A response rate of 78.1% resulted in 10,641persons being interviewed. Diagnoses were made using the CompositeInternational Diagnostic Interview. The interview was computerised andconducted by trained lay interviewers. We investigated comorbidity betweenGAD and major depressive disorder (MDD). The results indicate thatsociodemographic correlates of GAD, and associated disablement and serviceuse, are influenced by the presence of a comorbid depressive disorder but cannotbe fully explained by the presence of that disorder. In addition, GAD wasconfirmed as significantly disabling, even as a single disorder. We conclude thatthe results are consistent with the view that GAD has a significant andindependent impact on the burden of mental disorders. Depression and Anxiety20:23–31, 2004. & 2004 Wiley-Liss, Inc.

Key words: generalised anxiety disorder; major depression; comorbidity; communitysample

INTRODUCTIONEpidemiological surveys indicate that generalizedanxiety disorder (GAD) is one of the more commonanxiety disorders [Carter et al., 2001; Hunt et al., 2002;Wittchen et al., 1994]. High comorbidity with othermental disorders is evident and studies of both clinicaland community samples have shown that pure cases ofGAD are less frequent than cases of GAD withcomorbid psychiatric disorders [Bienvenu et al., 1998;Brown and Barlow, 1992; Bruce et al., 2001; Carter etal., 2001; Hunt et al., 2002; Wittchen et al., 1994]. Inrespect to community samples where help-seeking doesnot bias sample selection, Judd et al. [1998] reportedcomorbidity with lifetime unipolar depressive disorderin 67.4% of the National Comorbidity Survey (NCS)sample with a lifetime history of DSM-III-R GAD.Similar patterns of comorbidity have been confirmedusing DSM-IV criteria. The German Health Survey(GHS) [Carter at al., 2001] assessed 12-monthprevalence and reported that 59.0% of GAD respon-dents had comorbid major depression. In the Austra-

lian National Survey of Mental Health and Well-Being39.3% of individuals with 1-month GAD met criteriafor 1-month major depression [Hunt et al., 2002].

High rates of comorbidity and the assumption of lowseverity and disability have led some to question thediagnostic validity of GAD [e.g., Akiskal, 1998; Barlowand Wincze, 1998; Maser, 1998]. There is growing

DEPRESSION AND ANXIETY 20:23–31 (2004)

1School of Psychology, University of Sydney, Sydney,

Australia2School of Psychiatry, University of New South Wales, Clinical

Research Unit for Anxiety Disorders, New South Wales,

Australia

nCorrespondence to: Dr. Caroline Hunt, School of Psychology

(F12), University of Sydney, Sydney, NSW 2006, Australia.

E-mail: [email protected]

Received for publication 30 August 2003; Revised 6 April 2004;

Accepted 1 June 2004

DOI 10.1002/da.20019

Published online 20 August 2004 in Wiley InterScience (www.

interscience.wiley.com).

&& 2004 WILEY-LISS, INC.

Page 2: Generalized Anxiety Disorder and Major Depressive Disorder comorbidity in the National Survey of Mental Health and Well-Being

evidence to support the independent diagnostic statusof GAD as key features of the disorder do not seem tobe affected by the presence or absence of anotherdisorder. For example, comorbidity in GAD does notseem to be correlated with the course or chronicity ofthe disorder [Kessler, 2000; Yonkers et al., 2000] orwith age of onset patterns [Wittchen et al., 2000].Sociodemographic variables (e.g., age, gender, SES) didnot vary in NCS data according to whether GAD wascomorbid with other disorders, or according to whetherthe GAD had an earlier onset than comorbid disorders.Furthermore, the proportion of individuals whoseGAD occurs exclusively during the course of theircomorbid mood disorder is small [Hunt et al., 2002;Wittchen et al., 1994]. In terms of symptom patterns,an examination of the construct validity of diagnosesacross the anxiety and depressive disorders showed thatpatients with a diagnosis of GAD had a score profileacross a number of self-report measures that differ-entiated them significantly from other diagnosticgroups [Zinbarg and Barlow, 1996]. Specifically,patients with GAD differed significantly from patientswith MDD on a discriminant function identified as‘fear of fear.’

Although GAD and MDD may be conceptualised asseparate disorders, a number of authors have arguedthat the relationship between GAD and MDD may bemore meaningful than the relationship between GADand other mental disorders [Andrews et al., 2002;Krueger, 1999; Vollegergh et al., 2001]. For example,Andrews et al. [2002] found that the associationsbetween GAD and depression and dysthymia (oddsratios [OR]¼ 10.2 for depression, 12.6 for dysthymia)were similar in magnitude to the OR within diagnosticgroupings, such as between panic/agoraphobia andsocial phobia (OR¼ 8.6) or between alcohol abuse/dependence and drug abuse/dependence (OR¼ 10.8).This relationship between GAD and depressive dis-order may in part be attributed to the overlap in coresymptoms of these disorders as defined by DSM-IV[Maser, 1998]. For example, problems with concentra-tion, sleep disturbance, and fatigue are all identified assymptoms of both disorders. Alternatively, the closerelationship between GAD and MDD is consistentwith a common etiology. For example, Gorwood[2004] has argued for a genetic pleiotropic explanationfor GAD and MDD comorbidity (i.e., a singlemechanism with dif ferent consequences), primarilyon the basis of evidence of the high and reliablecomorbidity between the two disorders, and their‘‘substantial and shared heritability’’ [Gorwood, 2004].

Longitudinal studies show that the onset of anxietydisorders is more often followed by the onset ofdepressive disorders than depressive disorder is fol-lowed by anxiety disorder [Cole et al., 1998; Wittchenet al., 2000]. In the case of GAD, an increased risk forsecondary depression continues for many years afterthe onset of the temporally primary GAD, but thisincreased risk continues only when GAD is active as

opposed to remitted [Kessler et al., 2001]. Thesefindings are consistent with GAD as a risk factor forthe development of MDD. It has been argued, forexample, that depression secondary to GAD resultsfrom the fatigue and despondency that develops inresponse to persistent anxiety, particularly when thatanxiety becomes sufficiently impairing [Akiskal, 1998].Many individuals with GAD may only recognize theneed for professional intervention once secondarydisorders develop, such that when they do reachspecialist mental health services their presentation islikely to be complex and their disorder difficult to treat[Kessler et al., 2001]. Consistent with this view is thefinding that individuals with GAD report lengthydelays between the onset of symptoms and seekingtreatment, which is likely to be attributed to a delay torecognise such symptoms as a problem requiringprofessional intervention [Thompson et al., in press].

The idea of GAD as an independently disablingdisorder is inconsistent with low levels of specialistservice use and long delays to seek care and maycontribute to a perception that ‘‘pure’’ GAD is not adisabling condition. There is growing evidence thatGAD as a single disorder is disabling. For example,Kessler et al. [2002b] assessed recently the effects ofpure and comorbid major depression and GAD onmeasures of role impairment in the U.S. NationalComorbidity Survey and the Canadian Mental HealthSupplement to the Ontario Health Survey. Comparablelevels of role impairment were independently asso-ciated with both disorders. An aim of this study was tofurther examine the question of whether a diagnosis ofGAD is significant in terms of its public health impactin the absence of comorbid MDD, using data from theAustralian National Survey of Mental Health andWell-Being. We predict that: (1) individuals with bothpure and comorbid GAD will report significant levelsof disability and health service utilisation that iscomparable to other diagnostic groupings; (2) variablesassociated with disability and health service use will besignificant correlates of GAD (in addition to previouslyreported demographic correlates); (3) correlates ofGAD will be independently associated with GAD,rather than mediated by comorbidity with MDD; and(4) MDD will moderate the relationships betweenGAD and demographic, disability, and service usevariables.

MATERIALS AND METHODS

SURVEY DESIGN AND SAMPLE

Data from the National Survey of Mental Health andWellbeing, Australia, a nationwide household survey ofadults conducted in 1997, was used [Andrews et al.,2001]. The survey was conducted by the AustralianBureau of Statistics, a statutory body responsible forconducting such surveys using ethical protocols inconcordance with the Code of Ethics of the World

Hunt et al.24

Page 3: Generalized Anxiety Disorder and Major Depressive Disorder comorbidity in the National Survey of Mental Health and Well-Being

Medical Association (Declaration of Helsinki) thatinclude informed consent. The survey aimed todetermine the prevalence of mental disorders in thecommunity and to describe their associated disabilityand service use. An additional aim was to replicate andextend similar overseas surveys, such as the 1990 U.S.National Comorbidity Survey. A stratified multistagesampling process was used, resulting in a response rateof 78.1% and a final sample of 10,641 persons over theage of 18 years. The sample was weighted to conformto the age and gender distribution of the Australianpopulation and to account for probability of selection.

ASSESSMENT INTERVIEW

Interviews were administered from a laptop compu-ter by experienced lay interviewers from the AustralianBureau of Statistics. Supervisors for each State andTerritory were trained at the World Health Organiza-tion Training and Reference Centre for CIDI inSydney and also attended a course on the training offield staff.

Assessment of diagnosis. Diagnoses of GAD,MDD, and other anxiety, affective, and substance usedisorders were made using the computerised Compo-site International Diagnostic Interview (CIDI, Version2.1). Although inter-rater reliability data is not avail-able for version 2.1, previous versions of the CIDI havebeen shown to be reliable, with kappas greater than 0.9for the anxiety and depressive disorders [Wittchen etal., 1991]. Symptoms were assessed for their presencewithin 12 months before the interview and within 4weeks before the interview to obtain prevalenceestimates for both 12- and 1-month GAD and MDD.Symptoms were assessed against both ICD-10 andDSM-IV criteria for GAD; however, DSM-IV cases arereported in the present analysis. Exclusion criteria werenot applied. Personality diagnoses were made with theuse of screening questions based on the InternationalPersonality Disorder Examination [Loranger et al.,1997].

Assessment of demographic variables. Five demo-graphic variables were examined: age, gender, maritalstatus, education, and employment. Age was examinedin three age groups (18–34 years, 35–54 years, and 55þyears). Marital status was examined in three groups(married/de facto, separated, widowed or divorced, andnever married). Education was examined in two groups(university/diploma/vocational education qualificationand no higher education qualification). Employmentwas examined in three groups (employed, unemployedand not in the labour force).

Assessment of disability. Three measures ofdisability are reported. The Short Form 12 (SF-12)[Ware et al., 1996] measures physical and socialfunctioning and role limitations due to physical andmental health in the 4 weeks before assessment. It hastwo regression-weighted scales; a mental health sum-mary scale and a physical health summary scale (MCS-

12 and PCS-12). The scales are scored such that themean is 50 and the standard deviation is 10. Higherscores indicate less disability. The SF-12 shows goodconcordance with the longer 36-item Short FormHealth Survey (SF36) and good construct validity andtest–retest reliability (e.g., MCS-12 r¼ .76). ‘DisabilityDays’ [Kessler and Frank, 1997] is a summary measureof the number of days in the past 4 weeks an individualhas been unable to preform or has had to cut down ontheir normal activities because of ill health. The BriefDisability Questionnaire (BDQ) assesses the impact ofill health on activities, as well as performance of dailyactivities. It has demonstrated good internal consis-tency (a¼ .88) and good construct validity [von Korffet al., 1996].

Assessment of service use and treatment. Re-spondents were asked a series of questions about theirin-patient and outpatient service use in the 12 monthsbefore interview [Carter, 1998]. For each service use,respondents reported the type of professional seen andwhether the consultation was for a mental healthproblem. Respondents also reported whether or notthey received any of 10 forms of treatment/help for anyproblems they had experienced with their mentalhealth in the 12 months before the interview. The 10forms of treatment/help were information, medicinesor tablets, psychotherapy, cognitive behaviour therapy,counselling, help with house and money problems,help with their ability to work, help with looking afterthemselves, help with meeting people, or any otherkind of help.

The interview also contained the K10, a measure ofnonspecific psychological distress that has shown to bea sensitive screen for DSM-IV disorders in surveyscarried out in the U.S. [Andrews and Slade, 2001]. TheK-10 has strong criterion-related validity, as well asexcellent internal consistency (r¼ .89) [Kessler et al.,2002a]. Finally, the 12 neuroticism items from thewidely used Eysenck Personality Questionnaire-Re-vised [Eysenck et al., 1985] were used. This neuroti-cism scale measures the extent to which people viewthemselves as sensitive or emotional. It has highinternal reliability (r¼ .78).

ANALYSIS

A number of dichotomous variables were computedfor statistical analysis due to the high positive skew inthe distribution of the predictor variables. The vari-ables were dichotomised at arbitrary yet meaningfulcut-off points. The three disability variables were: (1)having a score of o40 (vs. a score of Z40) on the SF-12 Mental Component Summary scale; (2) having ascore of o2 (vs. a score of Z2) on the role functioningscale of the BDQ; and (3) having one or more days (vs.no days) recorded on the disability days scale. A cut-offof 40 was chosen on the MCS-12 to correspond withone standard deviation below the general populationmean. The cut-off score of two or more on the role

Research Article: Comorbidity of GAD 25

Page 4: Generalized Anxiety Disorder and Major Depressive Disorder comorbidity in the National Survey of Mental Health and Well-Being

functioning scale of the BDQ was chosen to corre-spond with previous studies that have used this measure[Ormel et al., 1999]. The remaining predictor variableswere dichotomised to achieve a median split in the fullsurvey sample. For example, the median number ofdisability days was zero in the full survey sample, andtherefore ‘greater than zero’ was chosen as the cut-offfor this variable. Two health service use variables wereused: (1) consultations with any health professional fora mental health problem (no consultation vs. 1 or moreconsultations); and (2) consultations with a specialistmental health professional for a mental health problem(no consultation vs. 1 or more consultations). Specialistmental health professionals were considered to bepsychiatrists, psychologists or a mental health team.Lastly, a treatment variable was computed based onwhether the individual had received any kind oftreatment for a mental health problem versus notreatment for a mental health problem. The healthservice use and treatment questions were based on theprevious 12 months. The neuroticism variable was alsosplit at the median into high neuroticism (neuroticismscore 42) versus low neuroticism (neuroticism score�2).

Logistic regression models were used to determinethe significance of univariate and multivariate associa-tions. The sample was weighted to conform to the ageand gender distribution of the Australian populationand to account for probability of selection. TheSUDAAN software package designed specifically foruse with complex survey samples was used for allanalyses [Shah et al., 1997]. The logistic regressionanalyses tested three models, with the outcome variablein the regression analysis being the sample ofrespondents with current GAD (n¼ 335) versus therest of the survey sample (n¼ 10306). The first modeltested the univariate association between current GADand demographic, disability and service use variableand neuroticism. This analysis aimed to confirm

disability and service use as significant correlates of adiagnosis of GAD. As demographic variables wereincluded, this was in part a replication of the analysisreported in Hunt et al. [2002]. The second modelincluded the same set of predictor variables, but addedMDD as a covariate to determine whether MDDmediated the relationship between GAD and itscorrelates. The third model tested the interactionbetween MDD and the predictor variables to deter-mine whether MDD moderated the relationshipbetween GAD and its correlates.

RESULTSDisability and service use in ‘‘pure’’ and comor-

bid diagnostic groups. Disability associated withvarious comorbidity groupings for GAD, as well asthe total survey sample, is reported in Table 1.Disability, as measured by the SF-12 Mental HealthScale, was significantly greater when GAD co-occurredwith an affective disorder, than if GAD co-occurredwith another anxiety disorder (t¼ 3.7, Po.001). Inrelation to mean disability days, no difference wasapparent between GAD accompanied by an affectivedisorder and GAD accompanied by an anxiety disorder(t¼�1.4, P¼ .19). No differences were found acrossthe comorbidity groups reported in Table 1 in regardto mean service utilisations. For instance, there were nosignificant differences in mean consultations betweenGAD with comorbid MDD compared to GAD withcomorbid anxiety disorder (t¼ 0.9, P¼ .39).

The data in Table 1 also indicate that pure GAD is insome respects at least as disabling as pure MDD orpure panic disorder. In terms of the mean number ofdisability days, pure GAD was not significantlydifferent from pure MDD (t¼ 1.8, P¼ .07), butshowed significantly greater number of disability daysthan pure panic disorder (t¼�2.8, Po.01). In terms ofdisability measured by the SF-12 mental health scale,

TABLE 1. Disability and service use for persons with 1-month DSM-IV generalised anxiety disorder with and withoutcomorbid disorders, major depression, and panic disorder

nDisabilitydaysn

SF-12 MentalHealth Scale

Serviceutilisations

Pure GADa 100 6.2 (1.1) 38.8 (1.5) 1.4 (0.2)Pure major depressiona 142 9.1 (1.0) 34.9 (0.9) 1.9 (0.2)Pure panic disordera 25 2.5 (1.0) 44.7 (5.0) 1.3 (0.5)No mental disorder 9,254 2.5 (0.1) 53.4 (0.1) 0.8 (0.0)GAD with comorbid disordersAny affective disorder 158 11.4 (1.0) 29.7 (1.2) 1.2 (0.2)

Major depression 139 10.9 (1.1) 29.0 (1.2) 1.2 (0.2)Dysthymia 60 12.8 (1.6) 31.3 (1.9) 1.2 (0.3)

Any anxiety disorderb 128 11.2 (1.4) 31.0 (0.9) 1.5 (0.2)Any other disorder 235 9.9 (0.8) 31.5 (0.8) 1.1 (0.1)

nTaken from the Service Utilisation Days out of Role questions (SUDOR). Disability days questions [Kessler and Frank, 1997]. All values are mean (se).aPersons with specified disorder and no other DSM-IV disorder.bIncludes all anxiety disorders other than GAD.

Hunt et al.26

Page 5: Generalized Anxiety Disorder and Major Depressive Disorder comorbidity in the National Survey of Mental Health and Well-Being

pure GAD was reported to be less disabling than pureMDD (t¼�2.1, Po.05) but not significantly differentto pure panic disorder (t¼ 1.1, P¼ .28). Furthermore,there was no dif ference in service utilisation betweenthe samples of pure GAD and pure MDD (t¼ 1.4,P¼ .19) and between pure GAD and pure panicdisorder (t¼�0.2, P¼ .83). Comparisons between thepure GAD sample and the sample of survey respon-dents with no diagnosed mental disorder showedsignificant dif ferences on disability days (t¼�3.4,Po.01), the SF-12 mental health scale (t¼�2.3,Po.05) and service utilisation (t¼ 9.8, Po.001).

Model 1: Associations between GAD and dis-ability, service use and demographic variables.Univariate associations between current GAD and thedemographic, disability, and service use variables andneuroticism, represented by Wald w2 statistics, arereported in Table 2. The results concerning GAD andthe five demographic variables (age, gender, maritalstatus, educational attainment, and employment) are adirect replication of those reported in Hunt et al.[2002]. Using data from the Australian survey, Hunt etal. [2002] reported that a 1-month diagnosis of GADwas positively and significantly associated with beingseparated, divorced or widowed, not holding tertiaryqualification, or being unemployed or not in the labourforce. The data also showed that the prevalence ofGAD declined over the age of 55 years, and wasrelatively lower for younger males, with these age-related differences possibly due to a tendency for olderadults and young males to be less likely to reportpsychological symptoms. Gender was not significantlyassociated with GAD. All additional ‘‘impairment’’

variables (disability, neuroticism, health service use,and treatment) were significantly associated withcurrent GAD at Po.001.

Model 2: Does MDD mediate the relationshipbetween GAD and predictor variables? Table 2reports the associations between the presence orabsence of GAD and the demographic and impairmentvariables with the presence or absence of current MDDentered as a covariate. These results indicate the extentto which the presence of MDD was responsible for theassociation between these variables and the presence orabsence of GAD. The results show that once MDDwas entered into the model, the associations betweenthe variables and GAD were weaker but, in all but onecase, still significant. The relationship between thepresence and absence of GAD and marital status failedto reach significance once MDD was taken intoaccount.

Model 3: Does MDD moderate the relationshipbetween GAD and predictor variables? Although thepresence of MDD did not fully account for therelationship between GAD and the demographic andimpairment variables, it was important to furtherconsider the role of MDD in moderating theserelationships. In other words, does the presence orabsence of MDD significantly modify the associationbetween GAD and the predictor variables? The resultslisted in Table 2 indicate that, with one exception (theBDQ variable), all the impairment variables were foundto interact with MDD in predicting the presence orabsence of GAD. On closer inspection of theseinteraction effects, all variables showed a similarpattern. Taking disability on the SF-12 as an example,

TABLE 2. Logistic regression results for three models assessing the relationship between current GAD and predictorvariablesw

Model 1: Significance ofpredictor variable

Model 2: Significance ofpredictor variable with major

depression as confound

Model 3: Significance ofinteraction between predictorvariable and major depression

Demographic variablesAge 22.15nn 15.22n 4.34Gender 1.94 0.13 3.91n

Marital status 8.98n 4.18 2.02Education 11.38nn 4.57n 0.13Employment 12.69nn 7.17n 3.76

Disability variablesDisabled on SF-12 302.94nn 203.50nn 18.98nn

Disabled on BDQ 91.69nn 36.99nn 2.64Disabled on disability days 534.37nn 175.91nn 17.94nn

Neuroticism 113.20nn 87.67nn 4.70n

Health service useAny health professional 486.78nn 143.82nn 27.76nn

Specialist mental health professional 187.66nn 26.75nn 16.16nn

Treatment receivedAny treatment 500.18nn 120.27nn 32.86nn

wValues are Wald w2.nPo.05, nnPo.001.

Research Article: Comorbidity of GAD 27

Page 6: Generalized Anxiety Disorder and Major Depressive Disorder comorbidity in the National Survey of Mental Health and Well-Being

the results indicated that the association between GADand disability was significantly stronger for thoseindividuals who did not have current MDD, comparedto those who did have current MDD. In other words,the presence of disability (or any other of thesignificant variables) increases the likelihood of alsohaving GAD to a greater degree in those withoutdepression than in those with depression. Gender wasthe only demographic variable that significantly inter-acted with MDD in predicting GAD, although thisinteraction effect was only just significant at P¼ .048,using a¼ .05. Again, the results indicated that theassociation between gender and GAD was stronger forthose who were not currently depressed compared tothose who were depressed.

DISCUSSIONThe aim of this study was to investigate the impact of

comorbidity between GAD and MDD on demographicand impairment profiles using data from the AustralianNational Survey of Mental Health and Well-Being.First, we considered the association between thepresence of comorbidity and level of disability experi-enced, and the extent of service use for thoseindividuals with a diagnosis of GAD. Second, we wereinterested in the extent to which the presence of acomorbid MDD might account for findings in relationto significant correlates of GAD, as well as associateddisability and service use.

Pure GAD (i.e., in the absence of a comorbiddisorder) was associated with disability comparable tothat associated with pure major depression or purepanic disorder. In functional terms, the sample ofpersons with pure GAD had been unable to engage intheir usual activities on an average of 6 days in theprevious month, and their mean disability score on theSF-12 mental health scale fell more than one standardbelow the population average. Given recent argumentsabout the level of impairment associated with GAD[Kessler et al., 2000; Olfson, 2000] the data in theAustralian survey are consistent with the notion thatGAD, even as a single disorder, is significantlydisabling.

Similar to other clinical and community studies,comorbidity with MDD was substantial in the Aus-tralian survey, with 39.3% (SE¼ 3.5) of persons with 1-month GAD also experiencing MDD [Hunt et al.,2002]. The data from the current study suggests thatcomorbidity with MDD has significant implicationsfor disability relative to when GAD comorbidity occurswith any other anxiety disorder, substance use disorder,or a personality disorder. This finding is consistentwith previous reports that the presence of anycomorbid disorder with GAD is associated with greaterlevels of disability [e.g., Hunt et al., 2002; Wittchenet al., 1994]. That comorbid MDD seems to beassociated with greater disability than comorbiditywith other disorder groupings is not surprising given

that affective disorder per se is more disabling thatanxiety disorders per se [e.g., Andrews et al., 2001].Surprisingly, the higher levels of disability exhibited bythe sample with GAD and MDD comorbidity was notreflected in levels of service use. It is possible that in thecase of individuals with GAD, comorbidity per se is thecritical factor in regard to service use because thepresence of any additional disorder increases problemrecognition, itself a significant factor in the decision toseek treatment [Thompson et al., in press].

The disabling nature of the disorder is also reflectedin a level of service utilisation that is comparable toother major mental disorders and almost twice the levelof individuals without a mental disorder. This findingadds to previous reports of treatment seeking, whichsuggest that all three disorders are associated withsignificant mental health service use [e.g., Katz et al.,1997] but also surprising given that many reports showaffective disorders to have higher levels of serviceutilisation than anxiety disorders [e.g., Andrews et al.,2001]. It is important to note that as the current samplewas limited to 12-month (as opposed to lifetime)prevalence rates, the sample is less likely to includeindividuals who might have experienced mental dis-orders that have since remitted. Thus, the currentsample may be biased toward having a greaterproportion of individuals with a chronic disorder thanin the general population. Perhaps once disordersbecome chronic, dif ferences between service utilizationacross different diagnostic groups disappear. It is alsoworth noting that in the current sample of persons withpure GAD, all contact with health professionalsoccurred at the primary care level [Hunt et al., 2002].

The results show that in addition to the socio-demographic variables reported elsewhere (age, maritalstatus, education and employment) [Hunt et al., 2002],a diagnosis of GAD was significantly associated with anumber of variables that reflect level of impairment(Model 1). These variables included disability, neuroti-cism, health service use, and having previous treatment.The data also show that although comorbidity withMDD does, to some extent, moderate the relationshipbetween the impairment variables and GAD (Model 3),it cannot fully account for the significant associationsthat were found (Model 2). GAD seems to beindependently associated with age, education, andemployment, but the association with marital statusseems to be mediated by the presence or absence of acomorbid MDD. Arguably the symptoms of depres-sion, as opposed to anxiety, will have a more significantimpact on interpersonal relationships. Alternatively,poor relationships or relationship breakdown are morelikely to contribute to emotions associated with loss(depression) than threat (anxiety).

The presence of a comorbid MDD was found tosignificantly moderate the relationship between GADand a number of variables. It is of interest that althoughgender was not a significant correlate of GAD acrossthe total sample, there was a significant interaction

Hunt et al.28

Page 7: Generalized Anxiety Disorder and Major Depressive Disorder comorbidity in the National Survey of Mental Health and Well-Being

between MDD and gender in predicting the presenceof GAD. Whereas the interaction just reached statis-tical significance, it does seem that there is a strongerrelationship between gender and GAD for thoseindividuals who did not have a comorbid MDD,relative to those who had a comorbid MDD. Aprevious study suggested that that there might havebeen a gender difference in younger age groups, withyounger males being less likely to report symptomsconsistent with a diagnosis of GAD [Hunt et al., 2002].We might speculate that the presence of MDD,particularly in older cohorts, had obscured a realgender dif ference associated with GAD in the sampleas a whole.

All impairment variables interacted significantly withMDD in predicting the presence of GAD, with theexception of the role functioning scale of the BDQ.Disability as measured by the SF-12, disability days,service use, and previous treatment had weakerassociations with GAD for individuals who had MDDrelative to those who did not. In the case of individualswithout comorbid MDD, it seemed that having or nothaving GAD has a significantly greater impact ondisability, service use, and treatment seeking. Again,this finding is expected as MDD, as a significantlydisabling disorder, will reduce the impact of thepresence or absence of GAD on these factors.Neuroticism scores were also more closely associatedwith GAD for the not depressed as opposed to thedepressed sample. One might speculate that, in theabsence of MDD, neuroticism is a significant vulner-ability factor for GAD. Neuroticism might become lessimportant for GAD in the presence of MDD, as MDDitself will present an already significant association withthis vulnerability factor.

CONCLUSIONSThis study provides further insight into the nature of

the relationship between the commonly co-occurringdisorders of GAD and MDD. In particular it considerswhether GAD is itself associated with significantdisability, or whether disability is associated withMDD comorbidity. When interpreting the results, itis important to keep in mind the problems associatedwith reliance on retrospective reporting and the cross-sectional design of the study. For example, the lack of aprospective design means that the causal direction ofthe significant relationships is not known. Further-more, the focus on samples with current and 12-monthdiagnoses may have produced a sample with increasedlevels of chronicity relative to the population ofindividuals with GAD. Notwithstanding these limita-tions, the report is strengthened by the use of a non-clinical community sample derived from a nationwidestratified epidemiological survey, and the use ofinstruments with robust psychometric properties.

The findings of the current study are consistent withprevious reports that disorder variables and socio-

demographic correlates do not vary whether comor-bidity is present, and thus consistent with the idea thatGAD and MDD are diagnostically independentdisorders. On this basis there is good support for theretention of GAD as a valid diagnostic category. Thepresence of a comorbid disorder is not necessary forGAD to be a disabling and chronic condition, whichalso lends support to the significance of the disordereven in its pure form. Furthermore, the increaseddisability associated with MDD comorbidity cannot beattributed wholly to the comorbid disorder. It istherefore argued that effective treatment of GAD as asingle disorder should be a priority for health services.This is particularly important given evidence for theadverse impact of comorbidity on the course of GAD,as well as the increased level of disablement. A largestudy following the course of disorder in individualsrecruited from specialist anxiety clinics showed that thepresence of a comorbid diagnosis of MDD or panicdisorder decreased the probability that an individualwould remit from their GAD [Bruce et al., 2001].

Unfortunately, although GAD as a single disorder isassociated with significant disability, it seems that thepresence of a comorbid disorder is required to bring anindividual into contact with specialist mental healthservices. Using data from a specialist anxiety disordersclinic, we have found that the average time between theonset of symptoms and seeking treatment for indivi-duals with a primary diagnosis of GAD was 9.4 years[Thompson et al., in press]. We have argued that therecognition of symptoms as resulting from emotionaldisorder will play a role in the decision to seektreatment. The results of the current study areconsistent with a view that the onset of a secondarycomorbid disorder leads to increase problem recogni-tion in individuals with GAD, which in turn leads toincreased use of specialised health services. Given thatthe majority of individuals with GAD as a singledisorder remain in primary health care services, it willbe important to investigate whether problem recogni-tion on the part of primary health care providers is alsoa barrier to effective treatment. The education ofhealth care providers and the improvement of mentalhealth literacy in the community may have a role inimproving outcome for individuals with GAD.

Investigations of comorbidity are important to ourunderstanding of the diagnostic independence ofpsychiatric disorders, and might also inform publicmental health initiatives. There is still much to belearned about the true nature of the relationshipbetween GAD and MDD, particularly in regard toaetiological and prognostic factors given the dif ficul-ties inherent in determining such factors in twocomplex and multifactorial disorders [Gorwood,2004]. For example, the current data is compatiblewith a role for neuroticism as a shared risk factor forboth GAD and MDD, yet the specific geneticcontribution to these disorders remains in some dispute[Kessler, 2000]. Alternatively, future studies using

Research Article: Comorbidity of GAD 29

Page 8: Generalized Anxiety Disorder and Major Depressive Disorder comorbidity in the National Survey of Mental Health and Well-Being

prospective designs and clinical populations cancomplement findings from epidemiological surveyssuch as those reported here. For example, earlyintervention approaches that track the developmentof secondary depression after treatment may add toexisting evidence for GAD as a risk factor in thedevelopment of MDD. We argue that improvedrecognition and effective treatment of GAD, which isa diagnostically independent and significantly disablingdisorder, should made a priority in health servicedelivery.

Acknowledgments. This study was supported by acontract from the Australian Department of Healthand Aged Services to the WHO Collaborating Centrefor Evidence in Mental Health Policy, Sydney, tosupport a survey data analysis consortium (G. Andrews,V. Carr, G. Carter, R. Crino, W. Hall, A. Henderson, I.Hickie, C. Hunt, L. Lampe, J. McGrath, A. McFar-lane, P. Mitchell, L. Peters, M. Teesson, K. Wilhelm).The Australian Bureau of Statistics, who does notnecessarily endorse the view expressed in this study,conducted the survey.

REFERENCESAkiskal HS. 1998. Toward a definition of generalized anxiety disorder

as an anxious temperament type. Acta Psychiatr Scand98(Suppl):66–73.

Andrews G, Henderson S, Hall W. 2001. Prevalence, comorbidity,disability and service utilisation. Overview of the AustralianNational Mental health Survey. Br J Psychiatry 178:145–153.

Andrews G, Slade T. 2001. Interpreting scores on the KesslerPsychological Distress Scale (K10). Aust NZ J Publ Heal 25:494–497.

Andrews G, Slade T, Issakidis C. 2002. Deconstructing currentcomorbidity: Data from the Australian National Survey of MentalHealth and Well-Being. Br J Psychiatry 181:306–314.

Barlow DH, Wincze J. 1998. DSM-IV and beyond: What isgeneralized anxiety disorder? Acta Psychiatr Scand 98(Suppl):23–29.

Bienvenue OJ, Nestadt G, Eaton WW. 1998. Characterizinggeneralized anxiety: Temporal and symptomatic thresholds. J NervMent Dis 186:51–56.

Brown TA, Barlow DH. 1992. Comorbidity among anxiety disorders:Implications for treatment and DSM-IV. J Consult Clin Psych60:835–844.

Bruce SE, Machan JT, Dyck I, Keller MB. 2001. Infrequency of‘‘pure’’ GAD: Impact of psychiatric comorbidity on clinical course.Depress Anxiety 14:219–225.

Carter G. 1998. Service utilization instrument development for theAustralian National Survey of Mental Health and Well-Being.New South Wales: University of Newcastle, NSW, Australia.

Carter RM, Wittchen HU, Pfister H, Kessler RC. 2001. One-yearprevalence of subthreshold and threshold DSM-IV generalizedanxiety disorder in a nationally representative sample. DepressAnxiety 13:78–88.

Cole DA, Peeke LG, Martin JM, Truglio R, Seroczynski AD. 1998.A longitudinal look at the relation between depression andanxiety in children and adolescents. J Consult Clin Psychol66:451–460.

Eysenck SBG, Eysenck KHJ, Barrett P. 1985. A revised version of thepsychoticism scale. Pers Individ Diff 6:21–29.

Gorwood P. 2004. Generalized anxiety disorder and major depressivedisorder comorbidity: An example of genetic pleiotrophy? EurPsychiatry 19:27–33.

Hunt C, Issakidis C, Andrews G. 2002. DSM-IV generalized anxietydisorder in the Australian National Survey of Mental Health andWell-Being. Psychol Med 32:649–659.

Judd LL, Kessler RC, Paulus MP, Zeller PV, Wittchen HU, KunovacJL. 1998. Comorbidity as a fundamental feature of generalizedanxiety disorders: Results from the National Comorbidity Study(NCS). Acta Psychiatr Scand 98(Suppl):6–11.

Katz SJ, Kessler RC, Frank RG, Leaf P, Lin E, Edlund M. 1997. Theuse of outpatient mental health services in the United States andOntario: The impact of mental morbidity and perceived need forcare. Am J Public Health 87:1136–1143.

Kessler RC. 2000. The epidemiology of pure and comorbidgeneralized anxiety disorder: A review and evaluation of recentresearch. Acta Psychiatr Scand 406:7–13.

Kessler RC, Andrews G, Colpe LJ, Hiripi E, Mroczek DK, NormandSLT, Walters EE, Zaslavsky AM. 2002a. Short screening scales tomonitor population prevalences and trends in non-specificpsychological distress. Psychol Med 32:959–976.

Kessler RC, Berglund PA, Dewit DJ, Ustun TB, Wang PS, WittchenHU. 2002b. Distinguishing generalized anxiety disorder frommajor depression: Prevalence and impairment from current pureand comorbid disorders in the US and Ontario. Int J MethodsPsychiatr Res 11:99–111.

Kessler RC, Dupont R, Wittchen H-U, Berglund P. 2000. Impair-ment in generalized anxiety disorder [letter]. Am J Psychiatry157:2061.

Kessler RC, Frank G. 1997. The impact of psychiatric disorders onwork loss days. Psychol Med 27:861–873.

Kessler RC, Keller MB, Wittchen H-U. 2001. The epidemiologyof generalized anxiety disorder. Psychiatr Clin North Am 24:19–39.

Krueger R. 1999. The structure of common mental disorders. ArchGen Psychiatry 56:921–926.

Loranger AW, Janca A, Sartorius N. 1997. Assessment and diagnosisof personality disorders: The ICD-10 international personalitydisorder examination (IPDE). Cambridge: Cambridge UniversityPress.

Maser JD. 1998. Generalized anxiety disorder and its comorbidities:Disputes at the boundaries. Acta Psychiatr Scand 98(Suppl):12–22.

Olfson M. 2000. Impairment in generalized anxiety disorder [letter].Am J Psychiatry 157:2060–2061.

Ormel J, von Korff M, Oldehinkel AJ, Simon GE, Tiemens BJ,Ustun T. 1999. Onset of disability in depressed and non-depressedprimary care patients. Psychol Med 29:847–853.

Shah BV, Barnwell BG, Beigler GS. 1997. SUDAAN user’smanual. Research Triangle Park, NC: Research Triangle Institute.

Thompson AE, Hunt C, Issakidis C. Why wait? Reasons for delayand prompts to seek help for mental health problems inan Australian clinical sample. Soc Psychiatry Psychiatr Epidem(in press).

Vollegergh WAM, Iedema J, Bijl RV, de Graaf R, Smit F, Ormel J.2001. The structure and stability of common mental disorders.Arch Gen Psychiatry 58:597–603.

Von Kroff M, Ustun TB, Ormel J, Kaplan I, Simon GE. 1996. Self-report disability in an international primary care study ofpsychological illness. J Clin Epidemiol 49:297–303.

Ware JE, Kosinski M, Keller SD. 1996. A 12-item short-form healthsurvey. Construction of scales and preliminary tests of reliabilityand validity. Med Care 34:220–233.

Hunt et al.30

Page 9: Generalized Anxiety Disorder and Major Depressive Disorder comorbidity in the National Survey of Mental Health and Well-Being

Wittchen HU, Kessler RC, Pfister H, Lieb M. 2000. Why do peoplewith anxiety disorders become depressed? A prospective-long-itudinal community study. Acta Psychiatr Scand 406:14–23.

Wittchen HU, Robins LN, Cottler LB, Sartorius N, Burke JD,Regier D, and participants in the multicentre WHO/ADAMHAfield trials. 1991. Cross-cultural feasibility, reliability and sourcesof variance of the Composite International Diagnostic Interview(CIDI). Br J Psychiatry 159:645–653.

Wittchen HU, Zhao Z, Kessler RC, Eaton WW. 1994. DSM-III-Rgeneralized anxiety disorder in the National Comorbidity Survey.Arch Gen Psychiatry 51:355–364.

Yonkers KA, Dyck IR, Warshaw M, Keller MB. 2000. Factorspredicting the clinical course of generalised anxiety disorder. Br JPsychiatry 176:544–549.

Zinbarg RE, Barlow DH. 1996. Structure of anxiety and the anxietydisorders: A hierarchical model. J Abnorm Psychol 105:181–193.

Research Article: Comorbidity of GAD 31