10.1192/bjp.135.2.163Access the most recent version at doi: 1979 135: 163-167 The British Journal of Psychiatry

P Tyrer and J Alexander

Classification of personality disorder

References http://bjp.rcpsych.org/cgi/content/abstract/135/2/163#otherarticles

Article cited in:

permissionsReprints/

permissions@rcpsych.ac.ukto To obtain reprints or permission to reproduce material from this paper, please write

to this article atYou can respond http://bjp.rcpsych.org/cgi/eletter-submit/135/2/163

serviceEmail alerting

click heretop right corner of the article or Receive free email alerts when new articles cite this article - sign up in the box at the

fromDownloaded

The Royal College of PsychiatristsPublished by on February 1, 2011 bjp.rcpsych.org

http://bjp.rcpsych.org/subscriptions/ go to: The British Journal of PsychiatryTo subscribe to

Brit. J. Psychiat.(1979), 135, 163—67

Classification of Personality DisorderBy PETER TYRER and JOHN ALEXANDER

SUMMARY An interview schedule was used to record the personalitytraits of 130 psychiatric patients, 65 with a primary clinical diagnosisof personality disorder and 65 with other diagnoses. The results wereanalysed by factor analysis and three types of cluster analysis. Factoranalysis showed a similar structure of personality variables in bothgroups of patients, supporting the notion that personality disordersdiffer only in degree from the personalities of other psychiatricpatients. Cluster analysis revealed five discrete categories; sociopathic,passive-dependent, anankastic, schizoid and a non-personality-disordered group. Of all the personality-disordered patients 63 per centfell into the passive-dependent or sociopathic category. The resultssuggest that the current classification of personality disorder could besimplified.

Despite many criticisms the concept ofpersonality disorder remains a useful one forpsychiatrists (Lewis, 1974; Shepherd and Saxtorius, 1974) and is included in formal classifications of illness (World Health Organization,1965; American Psychiatric Association, 1968).Unfortunately it has not achieved the samediagnostic status as other psychiatric disordersbecause of major difficulties in establishing avalid and reliable classification (Walton et al,1970; Walton and Presly, 1973). Currentclassifications of personality disorders are largelyunsupported by measurement and have notbeen independently confirmed. Using a structured interview schedule for assessing disordered personality we have examined thepersonality characteristics of 130 psychiatricpatients, half of whom were clinically assessed tohave personality disorders.

MethodPersonality disorderscheduleUsing a structured interview schedule (Tyrer

et al, 1979) 24 personality attributes (Table I)were rated on 9 point scales for all patients. Theschedule can be used with either subjects orinformants although it is more suitable for thelatter.

PatientsandprocedureOver an 18 month period all patients

attending PT's out-patient clinic at the Department of Psychiatry, Southampton, were assessedprovided that they satisfied one of the followingcriteria: (i) a relative or close friend of thepatient who had known the patient for at least10 years could be interviewed, or (ii) theassessor had seen the patient at least threetimes, one of which was at a time when thepatient had no formal psychiatric disorder.These criteria were introduced because currentpsychiatric status may influence personalityassessment (Wittenborn and Maurer, 1977). Alldiagnostic conditions, with the exception ofmental handicap, were included. The patientswere classified according to the InternationalClassification of Disease (lCD) (World HealthOrganization, 1965). After 18 months 65patients with diagnoses other than personalitydisorder but only 37 with personality disordershad been assessed. Patients with an lCDclassification of personality disorder thereforecontinued to be seen until 65 assessments werealso obtained so that adequate numbers wereavailable for comparison. The diagnosticfeatures of the 130 patients is shown (Table II).

163

AggressionIrritabilityAloofnessLabilityAnxiousness

CallousnessOptimismPessimismChildishnessResourcelessnessConscientiousness

DependenceRigiditySensitivityEccentricityShynessHypochondriasis

ImpulsivenessSubmissivenessSuspiciousnessIntrospectionVulnerabilityIrresponsibilityWorthlessness

Cluster 7- Non-Personality Disorders

AIRI Al ISlW@e@ V biChi @IIr @ICo E Sus In P MxS@ Res L kn Ca 0

FIG—Cluster profiles of personality disorder. The first two or three letters of the personality attributes listed in Table Iare shown on the abscissae, with the exception of Sn (Sensitivity) and Ip (Irresponsibility). The personality attributes

are given in the same order for all seven clusters.

164

Analysis of dataFactor analysis using the Varimax rotation

was applied to the ratings of both the normaland personality disordered groups separatelyand together to find out if similar factorsloaded in both groups. A cluster analysis of thetotal data was also carried out to determine thebest classification of groups, particularly withinthe personality disorders. This represented a testof the lCD classification of personality disorders,a classification which has never receivedindependent validation. Both hierarchical andnon-hierarchical methods of cluster analysiswere employed. Because of evidence thatsimilarity of cluster profiles may be moreimportant than numerical differencesbetweenthem (Strauss et al, 1973) both distance andcorrelational measures of similarity were em

CLASSIFICATION OF PERSONALITY DISORDER

ployed for the hierarchical analysis. Nearestneighbour, centroid and furthest neighbour Imethods of continuing clusters were explored.The non-hierarchical analysis chosen employed

TABLE IPersonal it, traits assessed in interview schedule

60

40

20

60Ckjsterl-E plo

40

20

--@

Cluster4 - Hsstrion@c

60@ç,y,pr\,@ /@Clusters-Jst@40

0)

c 20

01

o 60

40

20

60Cluster3 - Paranoid Ag@essive Ckister6- Schizoid

iiIAtAz Ist@IwlSeI v ó o@i'@@Co E Sus lo P Anx&tRes L ka Ca 0 Co E Sos@ P Ana S&@Res L ka Ca 0

C

a0

Personality disordergroupNon-personality disordergroupAnankastic

8Affectivepsychosis9Hysterical11Depressiveneurosis30Asthenic11Anxietyneurosis8Hypochondriaóal3Phobic anxietystate6Schizoid

6Alcoholism3Antisocial7Schizophrenia4Explosive

7Paranoidpsychosis3Paranoid7+ morbidjealousyAffective

3ObsessionalOthers2neurosis2Total

65Total65

Non-personality disordergroupPersonality disorder groupAllpatientsVarianceVarianceVarianceMain

trait(% ofMain trait(% ofMain trait(%ofFactorloadingstotal)loadingstotal)loadingstotal)SociopathyAggression

IrritabilityImpulsivenessCallousnessIrresponsibilityLability21.0Callousness

AggressionIrritabilityImpulsiveness19.3Aggression

CallousnessIrritabilityImpulsiveness

Irresponsibility16.6PassiveVulnerability13.9Dependence23.9Dependence21.1dependenceSubmissiveness

SensitivityDependenceResourcelessness

SubmissivenessChildishnessVulnerabilityAnxiousnessResourcelessness

ChildishnessIrresponsibilityVulnerability

LabilityAnankasticNot

representedRigidityConscientiousnessImpulsiveness*7.8Rigidity

Conscientiousness4.4SchizoidAloofness

Suspiciousness5.9Aloofness EccentricityLability3.0Aloofness

Eccentricity5.0DysthymicAnxiousness4.2Not

representedIntrospectionSensitivityPessimismAnxiousness9:6

165PETER TYRER AND JOHN ALEXANDER

T4@zI1 IIClassification of patients according to the International

Classificationof Disease

the ratio of the between groups to withingroups sum of squares as a criterion to bemaximized using an iterative process. Theseanalyses were carried out using the GENSTATcomputer package.

Results

In all but 25 of the 130 patients the personality schedule was completed with aninformant. The results ofthe factor analysis weresimilar in both groups of patients. In bothpersonality and non-personality disorder groupsthe two main factors, termed sociopathic andpassive-dependent, accounted for most of thevariance. A dysthymic factor was also detectedin the non-personality disorder group (TableIII).

TABLE IIIFactorstructureof personalityvariables.Onlypersonalitytraits with afactor co-efilcientof 0.5 or overare included

* Significant negative loadings for this variable.

DiagnosisCluster1 234567TotalAnankastic11518Hysterical1

4611Asthenic154111Hypochondria1113Schizoid156Antisocial2417Explosive2417Paranoid21317Affective1113Others112No

personalitydisorder35115565Total8

9 1418111357130

166 CLASSIFICATION OF PERSONALITY DISORDER

Seven clusters were defined in both thehierarchical and the non-hierarchical clusteranalysis using the distance measure of similarity.One cluster was largely composed of patientswith no personality disorder and the other sixclusters were mainly subsets of the disorderedgroup. The profiles for the seven clusters areillustrated (Fig), and labelled according to thepredominant pattern of abnormal personality.The profiles of the â€̃¿ explosive'and â€̃¿ paranoidaggressive' clusters, and the â€̃¿ asthenic' andâ€̃¿ histrionic' clusters were similar and thedifferenceswere mainly in ratingsof severity.The correlational cluster analysis showed fiveclearly identifiable clusters which were largelyreproduced by the three methods used. Thepatients in the explosive and paranoid aggressive clusters, and the asthenic and histrionicclusters were merged early in the hierarchicalprocedure reflecting the similar profiles of thesetwo pairs of clusters. The final group of fourpersonality disorder clusters was very similar tothat derived from factor analysis, with passivedependent, sociopathic, schizoid and anankasticcategories.

The relationshipbetween clustergroup andsub-category of personality disorder accordingto the International Classification of Disease isshown in Table IV. Patients in the hysterical

TABLE IV

Relationship between cluster groups and lCD diagnosis ofpersonality disorder

and asthenic categories are equally distributedbetween clusters 2 and 4 and those in theexplosive and antisocial categories betweenclusters 1 and 3. The anankastic and schizoidpersonality disorders largely correspond withtheir cluster counterparts.

DiscussionThe results of the factor analysis reveal that

the underlying structure of variables is similarin both those with and those without primarypersonality disorder and hence supports theconcept of personality disorders as being at theextreme ofa multidimensional continuum.

The cluster analysis showed that the personality-disordered patients could be classifiedinto at least four and possibly six categories. Themerging of six to four categories using thecorrelational cluster analysis and the closesimilarity between two of the seven profiles(Fig) suggests that only four distinct groupsof abnormal personality could be identifiedin the patients studied. The term passivedependent personality is preferred to eitherasthenic or histrionic personality, as the mainfeatures of the cluster were resourcelessness,dependence, childishness and vulnerability.The higher scoring of the asthenic sub-group(cluster 2) compared with the histrionic group(cluster 4) suggests that histrionic behaviour is astage along the road to asthenia rather than aseparate entity. The sociopathic cluster cornbines explosive and antisocial sub-groups for thesame reason. The anankastic and schizoidclusters include most ofthe patients diagnosed asanankastic and schizoid personality disordersusing the lCD (Table IV) and so the samedescription is retained.

Several of the classical descriptions ofabnormal personality types are not identifiablein the clusters. Kretschmer's well-known descriptions of cycloid (Kretschmer, 1922) and sensitivepersonality (Kretschmer, 1918) are reflected inthe separation of affective(lCD 301.1) andparanoid (lCD 301.0) categories in the International Classification of Disease but they didnot cluster separately in our analysis. This mayindicate that such abnormal personality typesare rare, but could also be due to the relativeabsence of social and personal difficulties in such

167PETER TYRER AND JOHN ALEXANDER

personalities. A statistically abnormal personality need not be a personality disorder, as astable adjustment may be made to the abnormality (Slater and Roth, 1969). The same mayapply to the dysthymic personality defined bythe factor analysis (Table III), as this groupwas found only in the patients without personality disorders. Hypochondriacal personalityalso does not appear in the clusters, but ashypochondriasis is statistically a heterogeneousentity (Bianchi, 1973) this result is not surprising. Some support for the majordivision ofpersonality disorders into sociopathic andpassive-dependent categories comes from Preslyand Walton (1973). Using a principal components analysis of ratings made on 140patients they found that components of socialdeviance and submissiveness accounted fornearly half the total variance. These twocomponents correspond closely with our twomain clusters, which contain 63 per cent of allthe patients originally diagnosed as personalitydisorders. Concepts of personality disorder heldby psychiatrists also suggest that its classification can be simplified. Plutchik and Platman(1977) studied psychiatrists' views of theattributes ascribed to the seven personalitytypes described in the diagnostic manual of theAmerican Psychiatric Association (1968) andfound only two types, compulsive and sociopathic personalities, showed clear different@tion. Histrionic and cyclothymic types, ands@iizoid, paranoid and passive-aggressive (aterm not used in the lCD) types shared manycharacteristics. These two composite groupsshow similarity with the passive-dependent andschizoid categories identified in our analysis.Because of the absence of an alternative

9lassification, the subdivisions of personalitydisorder given in the eighth revision of theInternational Classification of Disease have beenretained in the ninth revision (Shepherd andSartorius, 1974). Although personality types aremany, personality disorders are few, and ourresults suggest that the number of subdivisions

should be reduced and a category including thepassive-dependent concept introduced. If thiswere done the currently low reliability of thelCD categories of personality disorder (Waltonand Presly, 1973) might be improved.

AcknowledgementsWe thank Drs Carol Trotter, John Lambourn and

CoHnGodber for their helpwith ratings.

ReferencesAMERICAN PSYCHIATRIC ASSOCIATION (1968) Diagnostic and

Statsstical Manual for Mental Disorders, 2nd edition.Washington.

BIANcIn, G. N. (1973) Patterns of hypochondriasis:a principal components analysis. British Journal ofP@ychiat@y,122,541-8.

Kaarscsnszit, E. (1918) Der SensitiveBeziehungswahn.Berlin: Springer.

—¿ (1922) Körperbau und Charakter, pp. 98-102. Berlin:

Springer.Lzwis, A. (1974) Psychopathic personality: a most

elusivecategory.P@hologicalMedicine,4,133-40.PLUTCHIK, R. & Pi.@AN, S. R. (1977) Personality

connotations of psychiatric diagnosis: implicationsfor a similarity model. Journal of Xervosuand MentalDisease, 165,418—22.

PRESLY, A. S. & WAIr0N, H. J. (1973) Dimensions ofabnormal personality. Bntith Journal of P@ychiatty, 122,269—76.

Siixpsrnw, M. & SARroiuus, N. (1974) Personalitydisorder and the International Classification ofDiseases. Psychological Medicine, 4,141-6.

SLATER,E. & Rom, M. (1969) Clinical Psychiatrj, 3rdedition.London: Bailliêre,Tindall and Cassell.

S@ra@uss,J., BARTRo, J. J. & CARPENTER, W. T. (1973)The use of clustering techniques for the classificationof psychiatric patients. British Journal of Psychiatry,122,531-40.

TvRER, P., ALEXANDER,M., CICCHETrI, D., Coimti, M. &REMINGTON, M. (1979) Reliability of a schedule forrating personality disorders. British Journal of Psych.iatiy, 135, 168—74.

WALTON, H. J., Fouu)s, G. A., Lrrriw@i, S. K. & PRESLY,A. S. (1970) Abnormal personality. British Journal ofPsychiatry,116,497—510.

—¿ & PRESLY, A. S. (1973) Use of a category system in

the diagnosis of abnormal personality. British Journalof Psychiatry, 122,259-68.

WITTENBORN, J. R. & MAURER, H. S. (1977) Persistingpersonalities among depressed women. Archives ofGeneralPsychiatry,34,968-71.

Woiu.r HEALTH ORGANIZATION (1965) InternationalClassification of Disease. (Eighth revision). Geneva.

Peter Tyrer, M.D..M.R.C.P.M.R.C.PSYCIL.Senior Lecturer in Psychiatry, Universi@' of Southampton, Royal SouthHants Hospital, Graham Road, Southampton S09 4PE

* John Alexander, B.Sc@,M.Sc. (Social Studies). Lecturer in Medical Statistics, University of Southampton* Present address: Department of Computer Studies and Mathematics, The Polytechnic, Queensgate, Huddersfleld,

W. Yorkshire.(Received 9 Xovember 1978; revised 17 January 1979)