Psychiatric Annals

TREATMENTS REDUCING PERSONALITY DISORDERS 

Clinical Correlates of Stress-Induced Personality Disorder

James Reich, MD, MPH

Abstract

ABSTRACT

Objective: This article starts with the finding that some patients with an Axis I disorder who appear personality disordered when acutely ill may represent a distinct clinical subgroup. It examines the clinical correlates of this "stress induced" personality disorder group.

Methods: A group of male psychiatric outpatients (N = 165) was divided into three groups: The group with life-long personality disorders, "trait personality disorder [trait PD])" group (n = 24); the group with personality symptoms under the stress of an Axis I disorder, "stress-induced PD" or "state PD" group n = 63); and a group that had no personality disorders, "no-PD" group (n = 78). These three groups were compared on personality variables by direct comparison and logistic regression.

Results: Logistic regression showed a reasonable differentiation between the trait and state group. The variables of "reacts criticism," "suicide" and "needs approval" predicted the trait group while the variable "ashamed" predicted the state group. Logistic regression also showed reasonable differentiation between the state and no-PD groups. "Restricted expression of affect" predicted the no-PD group, while the variables of "acts childishly," "suicide," "sensitive to criticism/' "acts emotionally/' "feelings change," and "fearful" predicted the state group. The evidence seems to indicate the previously identified state group can be differentiated from its theoretical near neighbors using clinical criteria.

Abstract

ABSTRACT

Objective: This article starts with the finding that some patients with an Axis I disorder who appear personality disordered when acutely ill may represent a distinct clinical subgroup. It examines the clinical correlates of this "stress induced" personality disorder group.

Methods: A group of male psychiatric outpatients (N = 165) was divided into three groups: The group with life-long personality disorders, "trait personality disorder [trait PD])" group (n = 24); the group with personality symptoms under the stress of an Axis I disorder, "stress-induced PD" or "state PD" group n = 63); and a group that had no personality disorders, "no-PD" group (n = 78). These three groups were compared on personality variables by direct comparison and logistic regression.

Results: Logistic regression showed a reasonable differentiation between the trait and state group. The variables of "reacts criticism," "suicide" and "needs approval" predicted the trait group while the variable "ashamed" predicted the state group. Logistic regression also showed reasonable differentiation between the state and no-PD groups. "Restricted expression of affect" predicted the no-PD group, while the variables of "acts childishly," "suicide," "sensitive to criticism/' "acts emotionally/' "feelings change," and "fearful" predicted the state group. The evidence seems to indicate the previously identified state group can be differentiated from its theoretical near neighbors using clinical criteria.

Personality is generally considered to be stable, or at the very least, something that changes slowly over time. However, as everyone knows, many people respond differently under stress. These responses under stress are not all the same; we don't all act the same way in stressful situations. A person's reaction under stress represents varying degrees of modification of our normal personality styles and sometimes exposes personality vulnerabilities which cause more radical personality style changes.

Personality disorders (PDs) have been conceptualized in different ways by different schools of thought. There is not space in this report to review all the different conceptualizations of personality disorder, but I will mention two of the current major definitions. The current DSM-IV diagnosis of personality disorder is, "An enduring pattern of inner experience and behavior that deviates markedly from the expectations of the individual's culture."1 The pattern is manifested in two or more of the following areas: cognition, affectivity, interpersonal functioning, and impulse control. The pattern is inflexible and pervasive across a broad range of situations, has an early onset, is stable, and leads to significant distress or impairment. Personality disorders, according to the ICD-IO diagnostic guidelines,2 "...comprise deeply ingrained and enduring behaviour patterns, manifesting themselves in inflexible responses to a broad range of personal and social situations. They represent either extreme or significant deviations from the way the average individual in a given culture perceives, thinks, feels, and particularly, relates to others. Such behavior patterns tend to be stable and to encompass multiple domains of behavior and psychological functioning. They are frequently, but not always, associated with various degrees of subjective distress and problems in social functioning and performance."

The DSM-IV does not allow for the possibility of a stress-induced personality disorder. The ICD10 allows for a personality disorder to be created by stress. An enduring personality change is defined as: a disorder of adult personality and behaviour that has developed following catastrophic or excessive prolonged stress, or following a severe psychiatric illness, in an individual with no previous personality disorder. There is a definite and enduring change in the individual's pattern of perceiving, relating to, or thinking about the environment and the self. The personality change is associated with inflexible and maladaptive behavior that was not present before the pathogenic experience and is not a manifestation of another mental disorder or a residual symptom of any antecedent mental disorder.2,3 The 1CD-10 definition does not allow for a stressinduced personality disorder to reverse itself.

The two definitions of personality cited above emphasize the concept of personality functioning being stable over time. However, there is no question that measures of personality characteristics can be elevated if measured when the patient is acutely ill with an Axis I disorder. These measures then return to baseline after resolution of the Axis I disorder.4"10 One must consider the possibility that this is a measurement artifact. If this were the case, traits distorted by the presence of an Axis I disorder would have no clinical value (ie, they would just be noise confusing the clinical picture.) Taking this approach some DSM personality disorder test developers have worked to reduce this effect in their instruments. For example Loranger et al.11 have taken that approach and eliminated much of this noise from their test results. They have shown that it is possible to separate a group of trait personality disorders that do not change under stress.

The separation out of a trait personality disorder group does not eliminate the possibility of a state personality disorder (PD) group; it would merely form one boundary of the group with the other boundary being the no-PD group, personalities that are also relatively immune from stress, but are stably healthy from a personality perspective. A further argument for the value of categorizing a state personality group would be clinical significance. We know that personality measures taken during an acute Axis I illness predict outcome of treatment of an Axis I disorder.12-13 Ibis strengthens the argument that this may be a clinically relevant subgroup.

Other researchers have speculated about the possibility of stress-induced personality disorders. As far back as 1968 Leonhard theorized about this concept.14 Mischel15 examined the issue and found that although high levels of personality traits could predict behavioral response much of the time, as the personality trait was present at lower levels there was more variability in response. Some of this variability was presumed to be environmental. This conceptualization would fit the concept of stress-induced personality disorders. The high trait (personality disorders) would be more predictable in their dysfunctional responses, the very low trait (normals) would have the greatest adaptive flexibility and the stress-induced group would be intermediate. After a review of the literature on personality and the anxiety and depressive disorders, Bronisch and Klerman16 concluded that a stress-induced personality disorder was a reasonable concept. They referred to the concept as "personality change." They postulated five different areas where fluxuations in personality might occur: mood and affect, impulse control, attitudes toward self, attitudes toward the world, and social and interpersonal behavior.

Other researchers have approached the subject of personality from dimensional and genetic perspectives. Livesley et al.17 examined the heritabiiity of personality traits in twin pairs. They found personality traits had varying levels of heritability, some high and some low. They did not see discrete categories of personality, but rather personality traits behaving as dimensionally distributed attributes in the population. For most personality dimensions the best fitting model specified additive genetic and unique environmental effects. In this model the state PD group would be the middle rank in those who responded to environmental stress. They would be between those who responded maladaptively to minor environmental stress (trait PD) and those who were relatively resilient to environmental stress (no-PD.) Although this model would not give clear categorical boundaries, the state PD group would still be of clinical interest.

Tyrer et al.18"19 in the United Kingdom developed an empirically based personality disorder system measured by an instrument called the Personality Assessment Schedule. This system categorizes no-PD, sub-threshold PD, complex PD, and severe PD. In this system the state PD group might be considered somewhere near the border of sub-threshold and the simple PD (Tyrer, personal communication).

A previous report separated three groups by demographic, family history and functioning variables.20 These three groups were patients whose personalities may have been changed temporarily by the stress of an Axis I disorder (state or stress-induced PDs); patients whose personalities appear permanently disordered (trait PDs); and those patients whose personalities did not change under the influence of an Axis I disorder (no-PDs). That report indicated the possibility that a stress-induced group could be separated from a trait PD group and a no-PD group.

This report examines the clinical characteristics of stress-induced personality disorders. Three different groups (trait PD, stress-induced or state PD, and no-PD) are compared on clinical characteristics. It is hypothesized that the stress-induced PD has a distinct set of personality characteristics that will separate it from trait PDs and no-PDs.

METHODS

Population

Subjects for this study were drawn from a freestanding Veterans Administration outpatient clinic in a city in the northeast United States with a population of approximately 300 000. The population was 100% male. Subjects were a random sample of non-psychotic psychiatric outpatients. After complete description of the study to the subjects, written informed consent was obtained.

Instruments

The information used in this report was gathered by direct interview. The interview consisted of an established measure of Axis I disorders, the structured clinical interview for DSM-III-R diagnoses (SCID-I),21 an established measure of DSMIH-R personality disorders, the personality disorder examination, version 2 (PDE).22'23 The PDE has been designed to be resistant to the effects of state.11 The personality diagnostic questionnairerevised (PDQ-R), a 152-item self-report by patient or informant(s) using DSM-IH-R criteria was the second personality instrument administered.2425 This is a personality instrument that appears to be affected by state effects.6'8,9

Procedures

Patients in the psychiatric clinic were approached, either by mail or in person at the time of their visit, to take part in the study. Approximately 65% of those subjects eligible to be in the study performed the interviews. No differences at the demographic or diagnostic differences between the responders and non-responders were found.

Research assistants who had undergone extensive training on both of the instruments involved performed the interviews. This training included reading, videotapes, didactic sessions, and supervised practice interviews. The Biometrie Unit of the New York State Psychiatric Institute supplied training materials and consultation. Interviews were performed in person. Research assistants were blind to the purpose of the study and did not know in which group a given subject would be placed. The developer of the PDE participated in the training of some of the research assistants on the PDE.

Table

TABLE 1Personality Disorder Examination Items Used In the Analysis

TABLE 1

Personality Disorder Examination Items Used In the Analysis

Patients were divided into three groups based on the results of their personality testing. The trait PD group was formed by using any subject who scored in the DSM-IU-R cluster B grouping on the PDE. (Cluster A [schizoid cluster] includes the schizoid, schizoptypal, and paranoid personality disorders. Cluster B [impulsive cluster] includes borderline, histrionic, antisocial and narcissistic PDs. Cluster C [anxious cluster] consists of the avoidant, dependent, and compulsive PDs.) Of those remaining in the pool, those with a DSM-IlI-R cluster B grouping on the PDQ-R were placed in the stress-induced PD or state group. The remaining subjects without a personality disorder diagnosis on either the PDE or PDQ-R) were placed in the no-PD group.

Individual items from the PDQ-R and PDE were selected by the author. These .selections were made to select a small group of items that appeared to have face validity for the stressinduced PD concept. These were the items used in the analyses.

Statistical Analyses

Statistical analyses were made using SAS version 8.1 for personal computers.26 Individual comparisons used Fisher's Exact Test where the variables were categorical and analysis of variance where the variables were continuous. Odds ratios were calculated using the Proc Legist in the SAS program. A modeling procedure was done using Proc Logist to detennine combinations of variables that best differentiated the three clinical groups. A "forward" modeling process was used. If a variable met 0.1 level of significance, it was added to the model. After a new variable was added, all variables in the model that did not reach a 0.1 level of significance were dropped. This process continued until there was a final model.

RESULTS

As reported in a previous publication20 there were 78 patients in the no-PD group, 63 in the state or stress-induced PD group, and 24 in the trait PD group. (In the previous manuscript the group currently labeled state personality or "stress induced," is now labeled "state PD" or "stress induced." The group previously labeled "state PD" or "severe PD" is now labeled "trait PD." This is done to bring the terms in this article closer to other state /trait literature conventions.) There was a significant difference in the ages of the groups (61.3, 54.8, and 47.2 years respectively, F =14.5, df =2.0, ? =0.0001). There were no significant differences in Axis I diagnoses: schizophrenia; bipolar disorders (1.7%, 2.2%, and 14%); major depression (44.3%, 53.2%, and 66.7%); alcohol dependence (23.0%, 42.6%, and 46.7%); panic disorder (11.5%, 14.9%, and 26.7%); and posttraumatic stress disorder (27.9%, 46.7%, and 40.0%). In measures of functioning the rough order was that the trait group had the lowest functioning and highest level of Axis I symptoms, followed by the state PD group with the next highest level of symptoms and a higher level of functioning. The no-PD group had the lowest level of Axis I symptoms and had the highest functioning.

By design the no-PD group had no personality disorders in Cluster B as measured by the PDQ-R and the PDE. The state PD group had no cluster B disorders by PDE, but had cluster B disorders by the PDQ-R. The trait PD group had personality disorders on Cluster B by the PDE. These were: antisocial, 37.5%; narcissistic 8.3%; histrionic, 12.5%; and borderline 62.5%. (These diagnoses are not mutually exclusive.)

Family history of personality disorders was significantly different for the three groups in all three clusters. The results for the Cluster A were 16.5%, 8%, and 5.5% for the trait PD, state PD and no-PD groups respectively (?2 = 24.3, df = 2.0, ? = .0001). For Cluster B the results were 45.0%, 27.2%, and 21.6% (?2=37.2, df =2, p=.0001) and for Cluster C 32.0%, 15.6%, and 11.3% (?2 =42.9, df = 2.0, V =.0001).

The candidate variables are shown on Table 1 (page 584) and Table 2.

Table 3 (page 586) compares the frequencies of the candidate variables across the three groups, (trait, state, and no-PD.) The table shows that all of the variables are significantly different across groups. Most variables are highest in the trait PD group, followed by the state PD group and the no-PD group. However, for some PDQ-R variables, the state group is highest, followed by the trait group and then the no-PD group. These variables are "sensitive to criticism," "easily offended," "ashamed," and "feelings change." The variable of "acts emotionally" is about equal for the trait and state groups, both of which are higher than the no-PD group.

Table

TABLE 2Personality Diagnostic Questionnaire Items Used in the Analysts

TABLE 2

Personality Diagnostic Questionnaire Items Used in the Analysts

Tables 4 and 5 (page 587) examine whether logistic regression can successfully differentiate the trait and state groups using the candidate variables. Table 4 examines the PDE and PDQ-R variables separately and Table 5 combines them. Table 4 shows that the PDQ-R variables, by themselves, have difficulty making this distinction, with a maximum Rp 2 value of only 0.15. The PDE variables perform this differentiation better with a maximum Rp 2 value of 0.42. The analysis shown on Table 5, which uses both PDE and PDQ-R variables, demonstrates increased ability to separate the groups. The maximum Rp 2 is now 0.63. The variables of "suicide" (PDE) (OR = 41.7), "reacts criticism" (PDE) (OR - 37.0), and "needs approval" (OR = 37.0) predict the trait group while "ashamed" (PDQ-R) (OR = 0.006) predicts the state group.

Tables 6 and 7 (page 587) examine whether logistic regression can successfully differentiate the state and no-PD groups using the candidate variables. Table 6 examines the PDE and PDQ-R variables separately, while Table 7 combines them. The PDE variables, by themselves, do not show a great deal of ability to distinguish these two groups (maximum Rp 2 = 0.09). The PDQ-R shows better ability to differentiate these groups (maximum Rp 2 0.52). The analysis on Table 6 using both PDE and PDQ-R variables shows improved ability to distinguish between the two groups. The maximum Rp 2 value is 0.71. The most powerful predictor variables appear to be "restricted expression of affect" (PDE) (OR= 0.06) (predicting the no-PD group) and "acts childishly" (PDQ-R) (OR =166.6) and "feelings change" (PDQ-R) (OR =10.6) (predicting the state group).

Table

TABLE 3Frequency of Items By Group

TABLE 3

Frequency of Items By Group

COMMENTS

What we now have some evidence for is an intermediate group, state PDs, that can be separated from its nearest neighbors (trait PDs and no-PDs) by demographic variables, functioning variables, and clinical personality symptoms. This subgroup differs from trait PD (which constantly reacts in maladaptive ways to low-level stress) and from no-PD (which functions adaptively in spite of relatively high levels of stress). State PDs function well until exposed to a moderate amount of stress and then function in a maladaptive way for only the period of time that the stress is present.

There are two possible explanations of this phenomenon. One is that we are looking at a dimensional phenomenon, and the second is that we are examining a categorical diagnosis. The dimensional approach would tell us that personality maladaptive responses are relatively regularly distributed throughout the population, with extremes in either the high or low end because of random distribution of the trait. In this case we wouldn't be examining a syndrome with discrete boundaries but one that merged imperceptibly with the next category. This would then be similar to the concept of the personality trait neuroticism.

Table

TABLE 4Logisitic Regression Results Comparing State and Trait Groups

TABLE 4

Logisitic Regression Results Comparing State and Trait Groups

Even if the category is not a discrete syndrome, this does not mean that it would not have clinical significance. The diagnoses of social phobia and avoidant personality disorder might be an example of how the concept would be clinically significant. In that case, the relevant traits range from nondisabling and normal to a source of clinical concern (social phobia), to severe and generalized to the point that many life functions are colored by it (avoidant personality disorder).27 This dimensional aspect of social phobia does not affect its conceptual and clinical relevance and it also would not affect the value of the concept and clinical importance of state PDs. State PD's clinical relevance is increased by its association to the variables of suicide20 and response of an Axis I disorder to treatment.12"13

Table

TABLE 5Logisitic Regression Results Comparing Traft and State Groups

TABLE 5

Logisitic Regression Results Comparing Traft and State Groups

The next possibility is that we are examining a discrete syndrome. In this case the limits of the syndrome still would have to be determined. Is it a more general phenomenon existing in a multitude of different stress situations as postulated by Bronisch and Klerman16 or does it just apply to more circumscribed situations? For example, in this report, shame appears to be one of the individual variables that most characterize the state PD clinical correlates. We do not know if this is an artifact of our data set, reflects a larger class of difficult interpersonal affects or conflicts that might be related to state PD, or whether shame has some specific association for this concept. But it is possible that some specific subset of stressful circumstances, such as shame, could be more specific of state personality than stress in general.

Table

TABLE 6Logisitic Regression Results Comparing State and No-PD Groups

TABLE 6

Logisitic Regression Results Comparing State and No-PD Groups

Table

TABLE 7Logisitic Regression Results Comparing State and No-PD Groups

TABLE 7

Logisitic Regression Results Comparing State and No-PD Groups

Limitations to this report include using only a male population, a somewhat small sample size, and, in one instance, the use of a somewhat higher number of factors in the logistic regression than would be used in a nonexploratory study. These limitations are appropriate given the exploratory nature of the study in a clinically relevant area where little is published.

Future research would first need to replicate finding a state personality group in different clinical or community populations and to refine the key descriptive and underlying variables involved. This would probably have to be done by longitudinal studies to identify the different groups. Once variables are identified that characterize state personality, a measurement instrument to identify it on a crosssectional basis would need to be developed. This would be followed by studies designed to determine its natural course, risk factors, associations to other mental disorders, and morbidity.

REFERENCES

1. Diagnostic and Statistical Manual of Mental Disorders. 4th ed, Washington, DC: American Psychiatric Association; 1994.

2. The ICD-W Classification of Mental and Behavioral Disorders: Clinical Descriptions and Diagnostic Guidelines. Geneva: World Health Organization; 1992.

3. The International Statistical Classification of Diseases and Related Health Problems, 10th rev. Vol. 1. Geneva: World Health Organization; 1992.

4. Kerr TA, Schapira K, Roth M, Garside RF. The relationship between the Maudsley Personality Inventory and the course of affective disorders. Br J Psychiatry. 1970;116:11-19.

5. Ingham JG. Changes in MPI scores in neurotic patients: a three year follow-up. Br J Psychiatry. 1966;112:931-939.

6. Reich J, Noyes, R Coryell W, O'Gorman T. The effect of state anxiety on personality measurement. Am J Psychiatry. 1986;143:760-763.

7. Reich J, Noyes R, Hirschfeld RP, Coryell W, O'Gorman T. State effects on personality measures in depressed and panic patients. Am J Psychiatry. 1987;144:181-187.

8. Reich J, Noyes R. DSM-U1 personality disorders in unrecovered panic and depressed patients. / Anxiety Disord. 1987;1:123-131.

9. Reich J, Troughton, E. A comparison of DSM-HI personality disorders in recovered depressed and panic disorder patients. / Nerv Ment Dis. 1988;176:300-304.

10. Hirschfeld RA, Klerman GL, Clayton PJ, Keller MB, McDonald-Scott P, Larkin BH. Assessing personality effects of the depressive state on trait measurement. Am } Psychiatry. 1983;140:695-699.

11. Loranger A, Lenzenweger MF, Gartner AF, et al. Traitstate artifacts and the diagnosis of personality disorders. Arch Gen Psychiatr. 1991;48:720-728.

12. Reich JH, Green AI. Effect of personality disorders on outcome of treatment. / New Ment Dis. 1991;179:74-83.

13. Reich J, Vasile R. The effect of personality disorders on the treatment outcome of axis I conditions: an update. J Nero Ment Dis. 1993;181:475-484.

14. Leonhard K. Akzentuierte Personlichenkeiten. Berlin: Verlag Volk und Gesundheit; 1968.

15. Mischel W. Introduction to Personality: A New Look. 4th ed. New York: Holt Rinehart and Winston; 1986.

16. Bronisch T, Klerman G. Personality functioning; change and stability in relationship to symptoms and psychopathology. / Personal Disord. 1991;5:307-317.

17. Livesley J, Jang KJ, Jackson DN, Vernon PA. Genetic and environmental contributions to dimensions of personality disorder. Am J Psychiatry. 1993;150:1826-1831.

18. Tyrer P, Johnson T. Establishing the severity of personality disorder. Am J Psychiatry. 1996;153:1593-1597.

19. Tyrer P, Ferguson "B. Current classification of personality disorder. In: Tyrer P, ed. Personality Disorders: Diagnosis, Management and Course. 2nd ed. London: Arnold; 2000.

20. Reich J. Empirical evidence for "stress induced" personality disorders. Psychiatric Annals. 1999;29:701-706.

21. Spitzer RL, Williams JB, Gibbon M, First MB. Structured Clinical Interview for DSM-IU-R-Patient Version (SCID-P). New York: Biometrics Research Department, New York State Psychiatric Institute; 1988.

22. Loranger AW, Susman VL, Oldham JM, Russakoff LM. The Personality Disorder Examination: a preliminary report. / Personal Disord. 1987;1:1-1.

23. Loranger AW. Personality Disorder Exam (PDE) Manual. Yonkers, NY: DV Communications; 1988.

24. Reich JH. Instruments measuring DSM-HI and DSM-LTI-R personality disorders. J Personal Disord. 1987;1:220-240,

25. Reich JH. Update on instruments to measure DSM-III and DSM-ITJ-R personality disorders. / Neri; Ment Dis. 1989;177:366-370.

26. SAS [computer program]. Version 6.11. Cary, NC: SAS Institute; 2000.

27. Reich J. The relationship of social phobia to the personality disorders. Eur Psychiatry. 2000;15:151-159.

TABLE 1

Personality Disorder Examination Items Used In the Analysis

TABLE 2

Personality Diagnostic Questionnaire Items Used in the Analysts

TABLE 3

Frequency of Items By Group

TABLE 4

Logisitic Regression Results Comparing State and Trait Groups

TABLE 5

Logisitic Regression Results Comparing Traft and State Groups

TABLE 6

Logisitic Regression Results Comparing State and No-PD Groups

TABLE 7

Logisitic Regression Results Comparing State and No-PD Groups

10.3928/0048-5713-20021001-06

Sign up to receive

Journal E-contents