شیوع بالای اختلالات شخصیت در داوطلبان سالم برای تحقیق: مفاهیم برای تعصب گروه کنترل

Abstract Individuals who volunteer as control subjects for clinical studies are regularly screened for Axis I diagnoses, but seldom screened for Axis II disorders. This study examined the relative rates of Axis II diagnoses among 341 volunteers passing an initial telephone screen for entry into biological research studies. Axis I and II diagnoses by DSM-IV were assigned by best estimate after structured clinical interview, and subjects were categorized into one of three groups based on their diagnostic profiles: (1) volunteers without lifetime Axis I or II diagnoses (“healthy controls”), (2) personality-disordered volunteers without any history of Axis I pathology, and (3) personality-disordered volunteers with past (but not current) Axis I pathology. The results revealed a high prevalence of personality disorders (44.4%) among these volunteers. Several clinically relevant self-report inventories were used to demonstrate important characterological differences between the three comparison groups. Although inventory results demonstrated multiple differences between all three groups, most scales revealed differences between healthy controls and the two personality-disordered groups (with or without lifetime Axis I diagnoses), suggesting that most of the variance was accounted for by the presence or absence of an Axis II disorder, not a past Axis I disorder. These results suggest that personality-disordered volunteers may bias a control group due to the infrequent screening for Axis II disorders among volunteers for medical and psychiatric research. Implications are discussed for routine Axis II screening of volunteers for research with specific diagnostic instruments.

. Introduction One of the defining characteristics of effective psychiatric or medical research is the use of “healthy controls” against which the population of interest may be compared. Although no clear consensus establishes what constitutes a healthy control, control groups are generally designed with two goals in mind. First, control groups may serve to represent the general population as a random sampling. A true randomization ensures that the control group accurately reflects the population from which it is drawn. Second, control group design involves screening to ensure that members lack the defining characteristic of the experimental group or other characteristics that may bias experimental results. Both of these goals are taken into consideration when recruiting volunteers for psychiatric or medical research. Recruiting volunteers from a large non-clinical population supports the goal of random sampling. However, controls are frequently recruited from the community using advertisements offering incentives, such as payment or course credit. The initiative to respond to an advertisement and be willing to participate in research may itself distinguish volunteers from the general population. Similarly, the seeking of the particular incentive may distinguish volunteers from those in the population who do not seek the incentive. Multiple studies have determined that standardized personality rating scales conducted on research volunteers yield means that differ from general population means. For example, Eysenck Personality Questionnaire (EPQ; Eysenck et al., 1985) extraversion and neuroticism have been shown to correlate with willingness to volunteer in college students (Cowles and Davis, 1987), and the Freiburg Personality Inventory dimensions of extraversion, neuroticism, and nervousness correlate with medical student participation in pharmacology studies (Meyer et al., 1995). An Italian publication demonstrated that “healthy volunteers” for clinical trials differ significantly from Italian national means on the three validity scales and nine of 10 clinical scales from the Minnesota Multiphasic Personality Inventory (Berto et al., 1996). In light of their findings, the authors advocate the use of this instrument in evaluating volunteers for Phase I clinical trials. Using an inmate population, Walsh and Nash (1978) found that volunteers for medical research were more impulsive, manipulative, and exhibited more disordered thought processes than inmates who elected not to volunteer. Among a sample of putative healthy control subjects, individuals who chose to participate in a potentially painful procedure (lumbar puncture) were found to be more impulsive than those who declined (Gustavsson et al., 1997). This demonstrates an inherent problem in subject recruitment; namely, the research topic advertised in the recruitment advertisement differentially piques the interest of potential subjects, thus biasing respondents in a way that conforms to particular personality dimensions. The aforementioned studies provide evidence that research volunteers may differ from the general population, and assist in describing these differences. However, they do not address whether volunteers’ personality rating scale differences translate into clinically relevant differences. That is, it is unclear if these personality rating scale differences reflect significant volunteer biases in control groups. Also, these studies do not classify personality rating scale differences into diagnoses based on Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; American Psychiatric Association, 1994) criteria; as such, they provide a limited contribution to the process of screening volunteers for potential exclusion from control groups. The second goal of control group design involves screening to ensure that members lack the defining characteristic of the experimental group or other characteristics that may bias experimental results. Unfortunately, it is not always obvious what factors are relevant and have the potential to bias results. Furthermore, the ability to create an appropriate control group hinges on the ability to identify such factors in the first place. For this purpose, psychiatric research often utilizes standardized diagnostic instruments to screen for Axis I disorders and excludes diagnosed volunteers from serving as controls. Personality disorder diagnoses tend to be neglected during this process, although the suggestion that volunteers differ in “personality features” from the general population has been in the literature since the 1950s (Pollin and Perlin, 1958). Screening for Axis II disorders is not standard practice, perhaps due to the additional time, expertise, and expense necessary for diagnosis. This is an unfortunate trend, as no research to date has demonstrated that personality disorders do not introduce clinically relevant biases into control groups. Research has addressed the prevalence of Axis I psychopathology among volunteers recruited to serve as controls in medical and psychiatric research, indicating a higher prevalence of psychopathology as compared to epidemiological population estimates. Prevalence rates of Axis I pathology among research volunteers have been estimated at 16.5–27.7% for current diagnoses and 35.6–41.2% for past (without current) diagnoses according to two studies which administered the Schedule for Affective Disorders and Schizophrenia (SADS) using Research Diagnostic Criteria (Halbreich et al., 1989 and Schechter et al., 1994). These same studies report prevalence rates of 2.2–7% for personality disorder diagnoses, although they did not employ diagnostic tools specifically sensitive to Axis II symptomatology. A study seeking healthy subjects for paid research at a university hospital found that 26.9% of respondents passing a telephone screen were subsequently excluded due to Axis I diagnoses or psychoactive substance use, and 6.7% were excluded due to personality disorders (Shtasel et al., 1991). Diagnoses were made by psychiatric interview employing the Structured Clinical Interview for DSM-III-R: Non-Patient Version, while instruments more specific to personality disorder diagnoses were not utilized. The authors suggested that their telephone screening procedure was more rigid than in other studies, thereby minimizing the number of volunteers with psychopathology that proceeded to interview assessment. A small study of 49 research volunteers that employed the Structured Interview for Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) Personality Disorders (SIDP), as well as the SADS, diagnosed lifetime Axis I disorders at a prevalence of 32.6% and Axis II disorders (in the absence of lifetime Axis I diagnoses) at a prevalence rate of 20.4% (Thaker et al., 1990). By employing a diagnostic tool specific to Axis II disorders, this study more clearly demonstrates the potential for a high proportion of personality disorders among research volunteers. The aforementioned studies imply a lifetime prevalence rate of mental illness as high as 69% (Halbreich et al., 1989 and Schechter et al., 1994) among research volunteers, notably higher than epidemiological study estimates such as the National Comorbidity Survey (Kessler et al., 1994), which indicated a lifetime prevalence of 48% in a sample of 8000 individuals aged 15–54. Similarly, the research utilizing assessment instruments sensitive to personality pathology (Thaker et al., 1990) produced an estimate (20.4%) substantially higher than more recent community-based epidemiological estimates (i.e., 10–13%; de Girolamo and Dotto, 2000). These studies indicate the need to more closely examine the prevalence of diagnosable personality disorders among research volunteers, especially as they relate to current diagnostic criteria. These previous studies do not, however, validate the diagnoses by demonstrating differences between disordered volunteers and healthy volunteers on any particular outcome measure. Therefore, it remains unclear whether personality disordered volunteers have the potential to bias a control group. The current study evaluated paid volunteer research subjects using structured clinical interviews to determine the prevalence of both Axis I and II psychopathology among recruits. Because we were recruiting control subjects to take part in biomedical research procedures both psychiatric and medical health was required. To elucidate the potential bias that might be introduced into a study by the inclusion of personality-disordered volunteers, three distinct groups of volunteers were identified and compared on clinical and personality scales known to correlate with medical and psychiatric outcomes. Based on best-estimate diagnoses, three groups were identified as: (1) volunteers without lifetime Axis I or II diagnoses, i.e., “healthy controls”, (2) personality-disordered volunteers without lifetime Axis I diagnoses, who might normally be missed in a screening that only evaluated Axis I pathology, and (3) personality-disordered volunteers with past (but not current) Axis I diagnoses. In addition, these groups were compared on their responses to a variety of self-report personality questionnaire assessments. To our knowledge, this is the largest study to date to investigate the prevalence of personality disorders among volunteers for medical research using specific diagnostic instruments sensitive to Axis II pathology. It is also the first study of its kind to utilize DSM-IV nomenclature for diagnoses and the first study to validate diagnoses by showing differences on a personality assessment instrument.

. Results 3.1. Prevalence of Axis I/II diagnoses Subjects with any lifetime history of Axis I diagnosis comprised 44.9% (153/341) of the initial subject sample. The most prevalent Axis I disorders were the substance use disorders (22.3% lifetime alcohol-related disorder, 11.7% lifetime drug-related disorder), depressive disorders (19.1% lifetime depressive disorder, 11.4% lifetime major depressive disorder most prevalent), and anxiety disorders (8.5% lifetime anxiety disorder, 2.6% specific phobia most prevalent). Subjects with any lifetime history of Axis II diagnosis comprised 48.4% (165/341) of the initial subject sample. The frequencies of the various Axis I/II combinations (i.e., presence/absence) are depicted in Table 1 for the initial 341 subjects. Subjects with no history of Axis I or II pathology were designated as healthy controls. The prevalence of the specific personality disorder diagnoses is presented in Table 2. Twenty-six subjects were diagnosed with multiple personality disorders. Subjects diagnosed as PD-NOS met DSM-IV Criteria for General Personality Disorder and had only modestly higher GAF scores compared to those subjects diagnosed with a specific personality disorder (64.0 ± 7.7 vs. 60.5 ± 7.8, p < 0.05 by Tukey HSD) and significantly lower GAF scores compared with subjects with Axis I Disorders only (64.0 ± 7.7 vs. 76.5 ± 15.6, p < 0.001 by Tukey HSD) or without any Axis I/II disorder (64.0 ± 7.7 vs. 84.0 ± 5.1, p < 0.001 by Tukey HSD). All personality disorder diagnoses were represented in the sample. Volunteers diagnosed with Axis II disorders comprised 44.4% (110/248) of those who passed the structured interviews for current Axis I disorders, and 19% (47/248) of those who passed screenings for both current and past Axis I diagnoses. (As stated previously, the final sample size of 248 used in the rating scale analyses resulted from the exclusion of all subjects with a current Axis I diagnosis, non-PD subjects with a past history of Axis I diagnosis, and subjects with deferred diagnoses; see Methods/Subject Groups section for additional details.) The most frequent diagnosis was PD-NOS (24.9%), followed by obsessive-compulsive personality disorder (7%); the least frequent diagnosis was dependent personality disorder (0.3%). Table 1. Frequency of Axis I/II diagnostic combinations in initial sample (n = 341) Axis I Current diagnosis Full remission No history Diagnosis deferred Axis II Current diagnosis 55/341 (16.1%) 63a /341⁎(18.5%) 47b /341 (13.8%) 0 No history 7/341 (2.1%) 25/341 (7.3%) 138c /341 (40.5%) 0 Diagnosis deferred 1/341 (0.3%) 2/341 (0.6%) 2/341 (0.6%) 1/341 (0.3%) a Group designated as PDV + 1. b Group designated as PDV. c Group designated as Healthy Controls (HC). ⁎ Boldface denotes groups that were included in the final rating scale analyses (n = 248). Table options Table 2. Prevalence of personality disorders among volunteers Number (n = 341) % of Total volunteers Any personality disorder 165 48.4 Personality disorder NOS 85 24.9 Obsessive-compulsive PD 24 7.0 Narcissistic PD 21 6.2 Antisocial PD 19 5.6 Paranoid PD 20 5.9 Histrionic PD 8 2.3 Schizoid PD 10 2.9 Avoidant PD 9 2.6 Schizotypal PD 3 0.9 Borderline PD 2 0.6 Dependent PD 1 0.3 Table options 3.2. Demographic data Demographic data for the three comparison groups are presented in Table 3. Univariate analysis revealed a significant effect for age (F(2,245) = 5.50, p = 0.005). Post hoc analysis revealed that PDV + 1 subjects were older than both HC (p = 0.002) and PDV subjects (p = 0.012). No difference was found between the HC and PDV groups. This pattern suggests a potential interaction between age and a lifetime diagnosis of Axis I disorder. It is plausible that greater age confers greater risk of past Axis I pathology by time effects alone. In contrast, Axis II disorders begin, by definition, in adolescence or early adulthood. Table 3. Demographic characteristics of subject groups and total sample following initial screen for current Axis I psychopathology HC (n = 138) PDV no Axis I (n = 47) PDV + Axis I (n = 63) Full sample (n = 248) Mean age in years ± SD 28.4 ± 7.4 28.3 ± 6.4 31.2 ± 8.1 29.3 ± 7.5 Race Caucasian 84 (60.9%) 18 (38.3%) 33 (52.4%) 135 (54.4%) African–American 43 (31.2%) 27 (57.4%) 26 (41.3%) 96 (38.7%) Asian 7 (5.1%) 0 3 (4.8%) 10 (4.0%) Hispanic 2 (1.4%) 0 1 (1.6%) 3 (1.2%) Other 2 (1.4%) 2 (4.3%) 0 4 (1.6%) Gender Male 93 (67.4%) 36 (76.6%) 50 (79.4%) 179 (72.2%) Female 45 (32.6%) 11 (23.4%) 13 (20.6%) 69 (27.8%) Marital statusa Marriedb 21 (15.2%) 4 (8.5%) 5 (8.0%) 30 (12.1%) Never married 111 (80.4%) 39 (83.0%) 48 (76.2%) 198 (79.8%) Divorced 5 (3.6%) 4 (8.5%) 10 (15.9%) 19 (7.7%) SES Class II 22 (15.9%) 3 (6.4%) 4 (6.3%) 29 (11.7%) Class III 49 (35.5%) 18 (38.3%) 21 (33.3%) 88 (35.5%) Class IV 37 (26.8%) 11 (23.4%) 22 (34.9%) 70 (28.2%) Class V 30 (21.7%) 15 (31.9%) 16 (25.4%) 61 (24.6%) a One subject did not indicate marital status, thus slightly underestimating 100% of the total sample. b Includes separated (4), remarried (2), and married by common law (2). Table options Due to the small number of non-African American minority participants, analyses for race were limited to Caucasian and African–American participants. A chi-square revealed a significant effect of race within conditions (χ2 = 9.56; p = 0.008). Analysis of the cell distribution revealed that African–American participants were more likely to receive a diagnosis of PDV than HC (27/43) relative to Caucasians (18/84). However, there were no differences in the percentage of African–American versus Caucasians diagnosed as PDV + 1. Further chi-square analyses showed no significant effect of socioeconomic status on diagnostic condition (χ2 = 7.36; p = 0.289) or race (χ2 = 5.71; p = 0.127), and thus does not account for the race difference between the HC and PDV groups. No effects were found for gender (p = 0.16). The reason for the cultural differences in PDV diagnosis is unclear, although a number of studies have found discrepancies in the prevalence of specific personality disorders among various ethnic groups ( Canino et al., 1987, Chavira et al., 2003, Karno et al., 1987, Nestadt et al., 1990, Robins et al., 1984 and Swartz et al., 1990). Explanations for these discrepancies range from the tendency to apply Western views to non-Western belief systems, to the use of culturally biased assessment instruments. To our knowledge, no epidemiological studies have examined the prevalence of overall personality disorders across ethnic groups. Thus, it is difficult to speculate as to whether the discrepancies between African–Americans and Caucasians in this study were the result of genuine differences or artifact of the assessment and/or specific subject sample. Further investigation is warranted to that end. 3.3. Self-report personality questionnaire data Group means and standard deviations for all questionnaire data are presented in Table 4. Multivariate analysis revealed group differences on all five of the multi-scaled questionnaires (ANOVA found the same for the STAI scale). Subsequent post-hoc analysis revealed that the HC group differed from both PDV and PDV + 1 groups on most subscales. Specifically, the HC group differed from both PDV and PDV + 1 groups on 12 of the 18 (66.6%) scales examined. These included: EPQ Psychoticism, BDHI Hostility, BIS Non-Planning and Attentional Impulsivity, STAI Trait Anxiety, all ALS scales, and CAS Social Anhedonia. PDV subjects scored similarly to HC subjects on only two scales where group differences were noted (EPQ Neuroticism and BDHI Aggression) and PDV subjects scored differently from PDV + 1 subjects on only four scales where group differences were noted (EPQ Neuroticism and Psychoticism, BDHI Aggression, and STAI Trait Anxiety). Table 4. Comparative Rating Scale Data between healthy controls, personality-disordered volunteers without Axis I diagnoses, and personality-disordered volunteers with Axis I diagnoses Scale Healthy controls PDV no Axis I PDV + Axis 1 df F p Post-hoc n M (SD) n M (SD) n M (SD) EPQ 126 45 59 6,450 8.16 0.000 Neuroticism 5.8 (3.27) 6.3 (3.84) 10.02 (5.26) 2,227 23.30 0.000 A = B < C Psychoticism 2.8 (1.58) 3.4 (1.88) 4.1 (2.22) 2,223 11.1 0.000 A < B < C Extroversion 14.4 (4.27) 14.3 (4.50) 14.3(4.31) 2,227 0.04 0.965 BDHI 128 43 60 4,454 13.4 0.000 Aggressive acts 14.1 (5.90) 15.5 (6.58) 20.4 (6.46) 2,228 21.5 0.000 A = B < C Hostile feelings 2.9 (2.75) 4.7 (3.59) 5.63 (3.76) 2,228 16.2 0.000 A < B = C BIS 127 45 59 6,452 3.4 0.003 Motor 18.9 (3.70) 19.2 (4.08) 20.4 (4.10) 2,228 3.11 0.047 A < C NP 25.6 (4.85) 27.2 (4.57) 28.4 (4.97) 2,228 7.4 0.001 A < B = C Attentional 15.2 (3.16) 16.4 (3.10) 16.2 (3.25) 2,228 3.3 0.04 A < B = C Trait anxiety 125 30.9 (7.00) 43 34.6 (9.28) 59 39.4 (9.41) 2,224 22.1 0.000 A < B < C ALS 125 45 59 12,442 4.00 0.000 Depression 18.1 (5.30) 21.2 (5.76) 22.6 (5.96) 2,226 14.8 0.000 A < B = C Hypomania 20.8 (6.64) 25.3 (7.91) 25.2 (7.01) 2,226 11.7 0.000 A < B = C Biphasic 13.1 (3.87) 16.1 (5.10) 17.1 (5.46) 2,226 18.9 0.000 A < B = C Anxiety 9.2 (2.71) 11.3 (3.60) 11.9 (4.05) 2,226 16.1 0.000 A < B = C Anger 9.2 (3.03) 10.9 (3.73) 11.7 (4.39) 2,226 11.4 0.000 A < B = C Anxiety/depression 10.1 (3.10) 12.7 (4.58) 14.0 (5.46) 2,226 19.0 0.000 A < B = C CAS 77 39 44 6,310 2.26 0.037 Social anhedonia 6.3 (4.26) 8.5 (4.95) 8.1 (5.34) 2,157 3.6 0.029 A < B = C Physical anhedonia 10.7 (5.42) 12.5 (5.91) 11.0 (5.15) 2,157 1.4 0.249 Perceptual aberration 3.0 (3.81) 3.7 (3.87) 4.7 (3.53) 2,157 3.4 0.037 A < C Note: F-statistics were generated using multivariate analysis of variance. Scale abbreviations: EPQ = Eysenck Personality Questionnaire, BDHI = Buss-Durkee Hostility Index, BIS = Barratt Impulsiveness Scale, NP= Non-Planning subscale, ALS = Affective Lability Scale, CAS = Chapman Anhedonia Scale. Table options Given that many, if not most, epidemiological studies of personality disorder do not include a PD-NOS category, the rating scale analyses were also run while excluding individuals with a diagnosis of PD-NOS in order to ensure that results were not confounded by the addition of this more ill defined diagnostic category. The resulting sample of n = 185 produced the same general patterns of categorical differences as the first set of analyses, except for some changes in significance due to reduced statistical power. Every single rating scale continued to demonstrate significant differences between personality disordered and non-personality disordered groups.