Psychiatry Research. 26, I3 I-I 39 Elsevier

131

A Family History Method for DSM-III Anxiety and Personality Disorders

James H. Reich

Received February 25, 1988; revised version received June 13, 1988; accepted July 23. 1988.

Abstract. A family history method for DSM-///anxiety and personality disorders (FHPD) was validated by direct interview of 182 relatives. The categories of”any anxiety disorder,” “panic disorder,” and “any personality disorder” achieved sensitivities and specificities equivalent to currently accepted psychiatric family history methods. The family history results for the DSM-III dramatic and anxious personality disorder clusters showed good specificities, but only fair sensitivities. It is possible these cluster sensitivities could be improved in the future by using less stringent criteria. The FHPD is designed so that it can easily be integrated into the Family History-Research Diagnostic Criteria format and scoring.

Key Words. Family history, anxiety, personality.

Many psychiatric disorders, such as depression and schizophrenia, are familial. Although this information is useful to both researchers and clinicians, it is usually not feasible in most clinical and research settings to interview family members individually. As a result, techniques have been developed to estimate familial prevalence by history. Perhaps the best known and validated of these tech- niques is the Family History-Research Diagnostic Criteria (FH-RDC) developed by Andreasen et al. (1977). At present, there is no widely accepted family history method for diagnosing DSM-111 personality and anxiety disorders (American Psychiatric Association, 1980).

There are many articles indicating the familiality of anxiety disorders (Slater and Shields, 1969; Noyes et al., 1978; Crowe et al., 1983; Torgersen, 1983). There is also evidence that personality disorders may have some familial etiology. Several studies have shown a familial relationship between schizophrenia and the schizoid, schizotypal, or paranoid personality disorders (Kety et al., I97 1; Kendler et al., I98 I ; Kendler and Gruenberg, 1982, 1984; Baron et al., 1983). Loranger et al. ( 1982) and Baron et al. (1985) have found evidence for the heritability of borderline personality disorders. Pope et al. (1983) in a 4- to 7-year followup study of borderline personality disorders, found an increased incidence of personality disorders in relatives of borderline personality disorders. Two studies of obsessive and hysterical personality disorders have shown evidence for genetic factors in their etiology

James H. Reich, M.D., M.P.H., is Assistant Professor of Psychiatry, Harvard University, Massachusetts Mental Health Center. (Reprint requests to Dr. J. Reich, Psychiatry Service, I16A, Brockton VA Medical Center, 940 Belmont St., Brockton, MA 02401, USA.)

01651781/88/$03.50 @ 1988 Elsevier Scientific Publishers Ireland Ltd.

132

(Torgersen, 1983, 1984). There are several reports of obsessive-compulsive disorders running in families which appear to extend to the “softer” obsessive personality traits (Brown, 1942; Kringlen, 1965; Lo, 1967).

The development of a family history method for the DSM-III personality dis- orders presents special challenges. One difficulty is that few of the DSM-III

personality disorders have any empirical validation, thus raising the possibility that a family history method might fail due to the proband’s disorder being invalid

(Merikangas and Weissman, 1986). Moreover, many of the available standardized

measures of DSM-IZt personality disorders have little or no agreement with each other (Reich et al., 1987). This raises methodological questions about determining whether the family history instrument is performing adequately. In view of these limitations, even a modest success in developing a family history method would be useful in (1) providing a tool for further family investigations and (2) providing criteria to investigate as a basis for further research.

The present article examines the sensitivity and specificity of a family history method for DSM-III anxiety and personality disorders by comparing the family history method to direct relative interview.

Methods

Subjects. Subjects for this study were drawn from two ongoing studies. The first group was a randomly selected sample of psychiatric outpatients drawn from new intakes to the University of Iowa Outpatient Clinic. Patients with psychotic symptoms, organic brain syndrome, and mental retardation were excluded. Patients were diagnosed on Axis I by a Master’s level interviewer using the Schedule for Affective Disorders and Schizophrenia, Lifetime version (SAD!+L) which uses Research Diagnostic Criteria (RDC) (Endicott and Spitzer, 1978; Spitzer et al., 1978).

The second group consisted of 90 patients drawn from a treatment study of panic disorder (the patients were recruited by advertising). Patients were excluded from this study if they were schizophrenic, mentally retarded, had organic brain syndrome, drug or alcohol abuse in the last year, mania, obsessive-compulsive disorder, or major depression which preceded or dominated their panic disorder symptoms. All patients who met DSM-III-R (American Psychiatric Association, 1987) criteria for panic disorder and who were having at least one panic attack a week were included. Axis I psychiatric disorders were diagnosed by the Structured Clinical Interview for DSM-III Disorders (SCID) (Spitzer and Williams, 1983) given by a board-certified psychiatrist.

Patients were used as probands if they had a Structured Interview for DSM-III Personality Disorders (SIDP) or a Personality Diagnostic Questionnaire (PDQ) diagnosis of avoidant, dependent, borderline, or histrionic personality disorder. There were probands from both the outpatient group and from the panic group.

The patients selected all agreed to have their relatives contacted. Relatives of these probands who could be contacted were asked to participate in the study. One hundred eighty- two agreed to do so and completed an interview either over the telephone or in person. A followup mailing requesting the Millon Clinical Multiaxial Inventory be filled out drew 102 responses.

Instruments. The family history method for DSM-III anxiety and personality disorders (FHPD) was developed by J. Reich, R. Crowe, R. Noyes, and N. Andreasen and required IO-15 min to give per proband when limited to first-degree relatives. (If used for research purposes without “skip-outs”or on larger families, it can take longer.) The instrument has 27 questions rated on a 3-point scale (absent, subthreshold, and present). It is divided into anxiety

133

and personality sections, with each section subdivided by specific disorders. It has a definition for impairment and a scoring system. The kappas for two blind raters on 25 family vignettes are I .OO for any anxiety disorder, 0.88 for individual anxiety disorders (generalized anxiety, panic, agoraphobia, social, and simple phobia), 0.92 for “any” personality disorder, 0.88 for the DSM-Uipersonality disorder clusters, and 0.53 for the individual disorders of narcissictic, borderline, avoidant, dependent, compulsive, and masochistic personality.

The DSM-III divides personality disorders into clusters based on shared characteristics. Cluster A, the schizoid cluster, consists of schizoid, schizotypal, and paranoid disorders. Cluster B, the flamboyant cluster, contains the borderline, narcissistic, histrionic, and antisocial disorders. Cluster C, the anxious cluster, is composed of the dependent, avoidant, passive/aggressive, and compulsive disorders. There is empirical evidence that the clusters are a valid concept (Tyrer and Alexander, 1979; Kass et al., 1985; Reich and Thompson, 1987). The FHPD is divided by DSM-III personality clusters. A patient could meet criteria for Cluster A, B, or C (the broadest criteria) with less difficulty than meeting a specific personality disorder. For Cluster B, if additional criteria are met, the more specific diagnosis of borderline can be made. For Cluster C, if additional criteria are met, the additional diagnoses of avoidant, dependent, compulsive, or masochistic personality disorders can be made. For the anxiety disorders, there are separate sections for generalized anxiety, panic disorder, and the phobias. All sections have screening questions that allow the interviewer to “skip out” of a section if there is no indication for proceeding. Before a disorder is scored positive, there must be some indication that the reported traits or symptoms caused some social or occupational dysfunction. Although the FHPD uses DSM-IIIcriteria for the personality disorders and the FH-RDC uses RDC for its disorders, the two instruments are designed with a similar format so that they can be integrated with each other (we merely expanded the FH-RDC summary score sheet to include our additional disorders). Experienced interviewers require IO to I5 min to give the FHPD to a proband. The present study used only first-degree relatives.

There were three personality measures given to the relatives. (In some cases not all information is complete on every subject.) The three measures were the Structured Interview for DSM-III Personality Disorder (SIDP) (Pfohl et al., 1982) the Personality Diagnostic Questionnaire (PDQ) (Hyler et al., 1983), and the Millon Clinical Multiaxial Inventory (MCMI) (Millon, 1982).

The SIDP is a 160-item semistructured interview organized into sections relevant to personality styles; it measures all I1 DSM-III, Axis II diagnoses. Joint kappas for inpatient samples were 0.85 for borderline, 0.75 for histrionic, 0.62 for schizotypal, 0.45 for avoidant and 0.90 for dependent personalities (Stangl et al., 1985). As recommended by the instru- ment’s developers, significant others were also interviewed before the instrument was scored. Our raters were trained in part by one of the instrument’s developers, D. Stangl. In this study we followed the suggestion of the instrument’s developers and used informants as well as patient information to arrive at the patient’s SIDP diagnosis.

The PDQ is a 152-item self-administered true/false instrument measuring all I I DSM-III Axis II personality disorders. Test-retest reliability for psychiatric outpatients at I month was a kappa 2 0.56 for paranoid, schizotypal, antisocial, borderline, avoidant, and compulsive personality disorders (Hurt et al., 1984).

The MCMI is a 175-item true/false self-report instrument with scales whose personality diagnoses, in at least one study, are highly correlated with DSM-III Axis 11 diagnoses (Reich, 1987). Test-retest reliability of personality diagnoses at 4-6 weeks ranged from 0.77 to 0.85 in a group of psychiatric patients (Millon, 1982). The scores of Millon’s personality scales run from 0 to 100. For this study, scores > 85, the level designated prominent personality traits by Millon, were used to define a personality disorder.

The Brief Symptom Inventory (BSI) is a 50-item self-report scale of acute symptomatology (Derogatis and Spencer, 1982). Since it has been shown that high levels of anxiety and/or depression can distort personality measurements (Reich et al., 1986, 1987) this test was given to all relatives to ascertain their anxiety and depression levels during the prior 7 days. Two- week test-retest reliabilities were 0.78 and 0.84 for anxiety and depression subscales, respec- tively (Derogatis and Spencer, 1982).

134

Procedures. All probands were administered the FHPD when they were acutely ill and untreated pharmacologically for at least I week. Relatives who agreed were interviewed using the SADS-L, SIDP, PDQ, and MCMI, and all also filled out the BSI. Master’s level interviewers had at least 3 months’experience with the instruments involved. All interviewers were unaware of the diagnosis of the proband whose relative they were interviewing.

Although the anxiety diagnoses in the FHPD are designed to be compatible with DSM-III criteria, RDC were used as the “gold standard" for anxiety disorders for two reasons: (I) The FHPD is designed to be compatible with the FH-RDC, which uses RDC diagnoses. (2) The major anxiety diagnosis categories examined (any anxiety disorder, panic disorder, and agoraphobia) are fairly similarly defined in the two diagnostic systems.

Analysis. For the FHPD the diagnoses of “any” anxiety disorder, panic disorder, agoraphobia, simple phobia, and social phobia and the RDC diagnoses of the SADS-L in relatives over age I8 were compared to similar interview diagnoses. The FH PD diagnoses of “any” personality disorder, Cluster B personality disorder, Cluster C personality disorder, and the individual personality disorders which the FHPD measures were compared to SIDP, PDQ, and MCMI diagnoses in relatives. Since the proband group excluded schizophrenia and psychoses, it was felt to be an inappropriate sample to examine Cluster A personality disorders. Sensitivities, specificities, predictive power positive, and predictive power negative were calculated. For the MCMI, scaled scores > 85 were considered personality disorders. For the DSM-III personality disorder clusters, a relative was considered to have that cluster if he/she had any of the individual personality diagnoses in that cluster. A relative was placed in the “any” personality diagnosis cluster if he/she was positive for any personality disorder.

Results

One hundred eighty-two relatives were interviewed on the SADS-L and the SIDP, and filled out the PDQ. There were 80 (44%) males and 102 (56%) females; their mean age was 41.2 (15.9) years. A somewhat smaller group (n = 102) returned their MCMIs. (The MCMI was sent out in a second mailing at a later date.) This group consisted of 47 (46%) males and 55 (54%) females; their mean age was 42.1 (16.2) years.

The frequency distribution, sensitivites, specificities, predictive power positive, and predictive power negative are shown in Table 1. As can be seen, specificity is good in every case. For the broad categories of any anxiety disorder and any personality disorder and for panic disorder, the sensitivities are as good as reported for the same analysis of affective disorder by the FH-RDC (Andreasen et al., 1977, p. 1233, Table 6). Sensitivities for the DSM-III clusters are lower, however. The MCMI, although having the lowest sensitivities, had the highest predictive power positive. This, except in Cluster B, appears to have been achieved by lower predictive power negative values. In general, predictive value positives are higher than their corresponding sensitivities and are indicative of possible screening utility for research.

There were too few anxiety disorders diagnosed in the categories agoraphobia, simple phobia, and social phobia for adequate analysis. Similarly, there were too few individual personality disorders identified by the FHPD to allow for adequate analysis.

The mean BSI depression and anxiety scores for the entire relative group were 0.56 (0.63) and 0.44 (0.48) respectively. These scores represent the 62nd and 57th percentiles for a normal population and are not unduly elevated.

135

Table 1. Frequency distribution, sensitivity, and specificity of the Family History of Personality and Anxiety Disorders (FHPD)’

Disorder and Relative Predictive Predictive diagnostic frequency power power criteria n W) Sensitivity Specificity positive negative

Anxiety by RDC

Any anxiety disorder 20 11.0 0.500 0.913 0.416 0.936

Panic disorder 12 6.7 0.417 0.976 0.556 0.959

Personality by SIDP

Any personality disorder 33 18.2 0.364 0.865 0.375 0.860

Cluster B 10 5.5 0.300 0.924 0.188 0.958

Cluster C 28 15.4 0.250 0.928 0.389 0.872

Personality by PDQ

Any personality disorder 37 20.4 0.351 0.868 0.406 0.840

Cluster B 4 2.2 0.286 0.928 0.250 0.939

Cluster C 32 17.7 0.219 0.926 0.389 0.847

Personality by MCMI

Any personality disorder 56 54.9 0.161 0.826 0.529 0.447

Cluster B 17 16.7 0.117 0.964 0.400 0.845

Cluster C 37 36.2 0.156 0.892 0.416 0.644

Abbreviations. RDC = Research Diagnostic Criteria. SIDP = Structured Interview for DSM-III Personality Disorder. PDQ = Personality Diagnostic Questionnaire. MCMI = Millon Clinical Multiaxial Inventory.

1. RDC diagnoses were made on the basis of the Schedule for Affective Disorders and Schizophrenia - Lifetime version.

Discussion

Although the specific criteria for anxiety disorders in RDC are fairly well delineated and validated, this is not so for the DSM-III personality disorders. One study has found three jointly given measures of DSM-IIIpersonality disorders (MCMI, PDQ, and SIDP) had virtually no agreement with each other (Reich et al., 1987). It is possible that some of the differences could have been due to criteria slightly differing from DSM-II/. For although there is some empirical evidence of agreement between MCMI and DSM-I/I categories (Reich, 1987) a content analysis by Widiger et al. (1985) revealed differences in some categories. However, poor agreement was present even in categories where content analysis would not indicate difficulties. It is possible that differences between self-report and interview techniques could lower agreement. These findings are consistent with the review of Merikangas and Weissman (1986), who found little empirical validation work on most DSM-III personality disorders.

Given these findings, the formulation and validation of family history criteria becomes a bootstrapping operation. The chances of initial success are reduced since the initial formulations of the disorders are not well validated. However, should the disorder be familial and moderate success be achieved, the family history or family study method can provide a method of validation and a technique for refinement of the criteria of these disorders. Under these circumstances, it is certainly justified to use three different measures of DSM-III personality disorder as “gold standards” in relatives.

136

One method of ascertaining the relative usefulness of an instrument is to com- pare pertinent criteria to an established instrument of a similar type. The relevant parameters in this case would be sensitivity and specificity. One established family history method in psychiatry is the FH-RDC. For that instrument when only the proband is used as an informant, the reported sensitivity for affective disorder is 4 1% and the reported specificity is 89% (Andreasen et al., 1977). By this measure, our findings for any anxiety disorder (sensitivity = 0.50, specificity = 0.913) and panic disorder (sensitivity = 0.42, specificity = 0.98) are perfectly adequate. Any personality disorder as measured by the PDQ (sensitivity = 0.35, specificity = 0.87) or by the SIDP (sensitivity = 0.37, specificity = 0.87) comes close to meeting this criterion. I did not have multiple informants per relative in this study and therefore cannot report empirical data. However, if the effect of multiple informants on this method is similar to that reported by Andreasen’s et al. (1977) for major depression, I would expect sensitivity to increase approximately 33% while specificity remained roughly the same.

The results for Cluster B (Table 1) for validation for the PDQ and SIDP achieve sensitivities of 0.30 and 0.29, respectively, with corresponding sensitivities for Cluster C of 0.25 and 0.22. These sensitivities, although somewhat low, are potentially of use when combined with their high associated specificities. They could detect large differences in prevalence between populations, be used as a screening device to select personality disordered probands, or serve as a starting point for further research to improve their sensitivity. The exceptionally high specificities involved (range 0.924 to 0.928) raise the question of whether the criteria are merely too stringent and whether a loosening of them might result in higher sensitivities with only a modest loss of specificity. This last question can only be answered by empirical research.

For the MCMI, although specificities were high, sensitivities were low for every category. There are several possible reasons for this. One would be that the criterion of a scaled score of > 85 was too stringent to allow for adequate sensitivities. However, the high relative frequency of personality disorders found (54%) makes that unlikely. Alternatively, the instrument could be tapping into different personality aspects than the FHPD measures. This would be consistent with previously cited findings on content analysis (Widiger et al., 1985). It also should be remembered that less relative information was available for the MCMI subject (n = 102) than was available for the SIDP and PDQ (n = 182). Both the smaller sample size and potential biases in the personality-disordered responses to a followup mailing may have confounded the MCMI results. (Examples of potential biases include fewer paranoid and schizoid subjects returning the forms and a greater percentage of compulsive subjects returning the forms.)

Although I have referred to sensitivites and specificities throughout this report as “good” or “fair,” it must be understood that this is a value judgment. The actual utility of any given set of specificities and sensitivities is tied to the base rate of the disorder. In general, a given set will become more useful and efficient at higher base rates and less so at lower ones. Values that are excellent for high base rates may be inadequate for rare disorders.

As with any other study, this one has strengths and limitations. Its strengths are its selection of an appropriate population for study, its use of the best available

137

standardized instruments as “gold standards,” its ascertainment that state anxiety and depression were not confounding variables, and its double-blind status. Its weaknesses include failure to test for Cluster A due to an inappropriate sample and a smaller sample size for the MCMI comparison. An area also to be considered is the fact that, a trait’s being familial does not necessarily equate to its being genetic. Familiality of a trait could reflect a common familial environment, a genetic factor, or a socialization process. Future studies will have to tease these variables apart.

There are several possible future directions that can be taken in working with the FHPD. The categories of any anxiety disorder, panic disorder, and any DSM-III personality disorder are probably appropriate for clinical or research use. The specific anxiety disorders in the FHPD should be tested at some future time. Using DSM-III clusters could serve as an initial screening device to identify relatives with

specific personality disorder clusters. The clusters also represent an opportunity for

further research, both in determining the appropriate stringency of criteria as well as the specific appropriate criteria themselves. Cluster A will, of course, have to be tested on an appropriate sample. Another important area to investigate would be the effect of acute illness state on familial reporting by probands. The current instrument is available to clinicians or researchers who would like to use it in their work and/ or help develop it further.

Acknowledgment. This research was supported by NIMH small grant MSMB IR03 MH- 41055-01 and also by a grant from the Upjohn Corporation. Thanks are due Russell Noyes, M.D., and Ray Crowe, M.D., for their helpful comments at an early stage of this project and to Nancy Andreasen, M.D., for her comments at various stages of the project.

References

American Psychiatric Association. DSM-III: Diagnostic and Statistical Manual of Mental Disorders. 3rd ed. APA, Washington, DC (1980).

American Psychiatric Association. DSM-III-R: Diagnostic and Statistical Manual of Mental Disorders. 3rd ed. revised. APA, Washington, DC (1987).

Andreasen, N.C., Endicott, J., Spitzer, R.L., and Winokur, G. The family history method using diagnostic criteria: Reliability and validity. Archives of General Psychiatry, 34, 1229 (1977).

Baron, M., Gruen, R., Asnis, L., and Kane, J. Familial relatedness of schizophrenia and schizotypal status. American Journal of Psychiatry, 140, I437 (I 983).

Baron, M., Gruen, R., Asnis, L., and Lord, S. Familial transmission of schizotypal and borderline personality disorders. American Journal of Psychiatry, 142,927 (1985).

Brown, F.W. Heredity in the psychoneuroses. Proceedings of the Royal Society of Medicine, 35, 785 ( 1942).

Crowe, R.R., Noyes, R., Pauls, D.L., and Slymen, D. A family study of panic disorder. Archives of General Psychiatry, 40, 1065 (1983).

Derogatis, L., and Spencer, P.M. Brief Symptom Inventory (BSI). Administration, Scoring and Procedures Manual. Clinical Psychometric Research, John Hopkins University, Baltimore (1982).

Endicott, J., and Spitzer, R.L. A diagnostic interview: The Schedule for Affective Disorders and Schizophrenia. Archives of General Psychiatry, 35,837 (I 978).

Hurt, S.W., Hyler, SE., Frances, A., Clarkin, J.F., and Brent, R. Assessing borderline personality disorder with self-report, clinical interview or semistructured interview. American Journal of Psychiatry, 141, I228 (1984).

138

Hyler, S., Rieder, R.O., and Spitzer, R.L. Personality Diagnostic Questionnaire (PDQ). New York State Psychiatric Institute, New York (1983).

Kass, F., Skodol, A.E., Charles, M.A., Spitzer, R.L., and Williams, J.B. Scaled ratings for DSM-III personality disorders. American Journal of Psychiatry, 142,627 (1985).

Kendler, K.S., and Gruenberg, A.M. Genetic relationship between paranoid personality and the “schizophrenia spectrum” disorders. American Journal of Psychiatry, 139, 1185 (1982).

Kendler, K.S., and Gruenberg, A.M. An independent analysis of the Danish adoption study of schizophrenia: VI. The relationship between psychiatric disorders as defined by DSM-IIIin relatives and adoptees. Archives of General Psychiatry, 41, 555 (1984).

Kendler, KS., Gruenberg, A.M., and Strauss, J.S. An independent analysis of the Copenhagen sample of Danish adoption study of schizophrenia: II. Relationship between schizotypal personality disorder and schizophrenia. Archives of General Psychiatry, 38, 892 (1981).

Kety, S.S., Rosenthal, D., and Wender, P.H. Mental illness in the biologic and adoptive families of adopted schizophrenics. American Journal of Psychiatry, 128, 302 (1971).

Kringlen, E. Obsessional neurotics: A long term follow-up. British Journal of Psychiatry, 111,709 (1965).

Lo, W.H. A follow-up study of obsessional neurotics in Hong Kong Chinese. British Journal of Psychiatry, 113,823 (1967).

Loranger, A.W., Oldham, J.M., and Tulis, E.H. Familial transmission of DSM-If/ borderline personality disorder. Archives of General Psychiatry, 39,795 (I 982).

Merikangas, K.R., and Weissman, M.M. Epidemiology of DSM-IfI, Axis 11 personality disorders. In: Frances, A., and Hales, R.E., eds. American Psychiatric Association Annual Review. Vol. 5. American Psychiatric Press, Washington, DC (1986).

Millon, T. Millon Clinical Multiaxial Inventory. 2nd ed. Interpretive Scoring Systems, Minneapolis, MN (1982).

Noyes, R., Clancy, J., and Crowe, R.R. The familial prevalence of anxiety neuroses. Archives of General Psychiatry, 35, 1057 (1978).

Pfohl, B., Stangl, D., and Zimmerman, M. The Structured Interview for DSM-III Personality Disorders (SIDP). University of Iowa Hospitals and Clinics, Iowa City, IA (1982).

Pope, H.G., Jonas, J., Hudson, J., Cohen, B.M., and Gunderson, J.G. The validity of DSM-III borderline personality disorder. Archives of General Psychiatry, 40,23 (1983).

Reich, J.H. Instruments measuring DSM-III and DSM-III-R personality disorders. Journal of Personality Disorders, 1,220 ( 1987).

Reich, J.H., Noyes, R., Coryell, W. and O’Gorman, T. The effect of state anxiety on personality measurement. American Journal of Psychiatry, 143,760 (1986).

Reich, J.H., Noyes, R. Hirschfeld, R., Coryell, W., and O’Gorman, T. State effects on personality measures in depressed and panic patients. American Journal of Psychiatry, 144, 181 (1987).

139

Reich, J.H., Noyes, R., and Troughton, E. Comparison of instruments to measure DSM- III, Axis II. In: Millon, T., ed. Proceedings oj the Millon Clinical Multiaxial Inventory Conjerence f986. National Computer Systems, Minetonka, MN ( 1987).

Reich, J.H., and Thompson, W.D. Differential assortment of DSM-III personality disorder clusters in three populations. British Journal of Psychiatry, 150,47 I ( 1987).

Slater, E., and Shields, J. General aspects of anxiety. In: Lader, M.H., ed. “Studies of Anxiety.” British Journal of Psychiatry. 3,62 ( 1969).

Spitzer, R.L., Endicott, J., and Robins, E. Research Diagnostic Criteria: Rationale and reliability. Archives of General Psychiatry, 35, 773 ( 1978).

Spitzer, R.L., and Williams, J.B. Structured Clinical Interview jbr DSM-III Disorders (SCID). New York State Psychiatric Institute, New York (1983).

Stangl, D., Pfohl, B., Zimmerman, M., Bowers, W., and Corenthal, C. A structured interview for the DSM-III personality disorders. Archives I$ General Psychiatry. 42, 591 (1985).

Torgersen, S. Genetics of neuroses: The effect of sampling variation upon the twin concordance ratio. British Journal of Psychiatry, 142, I26 (1983).

Torgersen, S. Genetic and nosologic aspects of schizotypal and borderline personality disorders: A twin study. Archives of General Psychiatry, 41,546 (1984).

Tyrer, P., and Alexander, J. Classification of personality disorder. British Journal o/ Psychiatry, 135, 238 (I 979).

Widiger, T.A., Williams, J.B.W., and Spitzer, R.L. The MCMI as a measure of DSM-III. Journal of Personality Assessments, 49, 366 ( 1985).