Algorithms for pharmacological treatment of personality dimensions: 

Symptom-specific treatments for cognitive-perceptual, affective, and impulsive-behavioral dysregulation

by Paul H. Soloff, MD

This article is based on a presentation at the 19th annual Menninger Winter Psychiatry Conference, held March 9-14, 1997, at Park City, Utah. This article is supported by NIMH Grant # MH 48463. Dr. Soloff is professor of psychiatry, Department of Psychiatry, University of Pittsburgh School of Medicine. Correspondence may be sent to Dr. Soloff at Western Psychiatric Institute and Clinic, 3811 O'Hara Street, Pittsburgh, PA 15213.

A pharmacological approach to treating patients with personality disorders (PD) is based on evidence that some dimensions of personality are mediated by variations in neurotransmitter physiology and are responsive to medication effects. Target symptoms for pharmacotherapy in the PD patient are derived from expressions of cognitive-perceptual, affective, and impulsive-behavioral dysregulation of central neurotransmitter functions. Pharmacotherapy is directed at state symptoms during periods of acute decompensation and at trait vulnerabilities, which represent the diathesis to future episodes. A basic assumption of this approach is that neurotransmitter biology transcends Axis I and Axis II definitions and that closely related symptoms may share a common pathophysiology, independent of categorical definition. A common pathophysiology implies the possibility of shared responsiveness to medication. Using a dimensional definition of symptom domains, the author has developed treatment algorithms for cognitive-perceptual symptoms, affective dysregulation, and impulsive-behavioral dyscontrol in personality disorder patients. (Bulletin of the Menninger Clinic, 62[2], 195-214)

Personality diagnoses are syndromes of related and interacting dimensions, each of which may have unique origins in psychosocial develop-merit or biological temperament. Popular empirical models define three (Eysenck & Eysenck, 1976), five (Costa & McCrae, 1992), or seven (Cloninger, Svrakic, &; Przybeck, 1993) basic dimensions of personality that can be used to define recognizable clinical types. Personality dimensions may be genetically mediated or acquired in the course of development. While it is customary to discuss the biologically mediated dimensions of personality (e.g., temperament) apart from the learned characteristics (e.g., character), this distinction is artificial and obscures the extensive interaction and overlap between neurobiology and the psychosocial origins of personality. The neurobiology of perception, cognition, information processing, and regulation of affect and impulse strongly conditions the expression of one's needs and the range of one's behavioral responses, providing a matrix in which interpersonal relatedness develops. A healthy mother-infant bond depends on the conditioning of each to the other's temperament. Life experiences, especially early life trauma, may profoundly alter both biology and psychology of personality development. A personality disorder (PD) is said to be present when dimensions of personality lead to patterns of experience and behavior that are pervasive and inflexible and that cause distress or impairment (American Psychiatric Association, 1994).

A pharmacological approach to the treatment of personality disorders is based on evidence that some dimensions of personality appear to be mediated by variations in neurotransmitter physiology and are responsive to medication effects. Specifically, the regulation of cognition, perception, affect, and impulse appears to be dependent on the actions of specific neurotransmitter systems. Symptoms arising from dysregulation of these systems are the true targets of pharmacotherapy in the personality disorder patient. Pharmacotherapy is used to treat both state symptoms during periods of acute decompensation and trait vulnerabilities, which represent the biologic diathesis to future episodes. This duality of effect raises serious methodological and ethical issues concerning the appropriate definition and assessment of target symptoms, measurement of change, and duration of treatment (e.g., how long does one treat temperament)? By targeting the neurotransmitter physiology that regulates cognition, perception, affect, and impulse, pharmacotherapy may modify temperament, allowing for significant relearning in interpersonal behavior. Pharmacotherapy is not an appropriate treatment for chronic, dysfunctional attitudes about the self or others and is best viewed as an adjunctive treatment, not a substitute for psychotherapy. Medication cannot cure character.

Target symptoms for pharmacotherapy in the PD patient may be derived from the symptomatic expressions of cognitive-perceptual, affective, and impulsive-behavioral dysregulation. This theoretical model follows studies of symptom response patterns in pharmacotherapy trials of severely impaired inpatients with borderline personality disorder (BPD), a strategy known as pharmacological behavioral dissection. In one such study, Soloff and associates (1989) defined three broad symptom response patterns following treatment with amitriptyline, haloperidol, or placebo among severely impaired BPD inpatients. The three broad response patterns defined by factor analysis of symptom changes were termed "global depression," "hostile depression," and "schizotypal" symptom-change factors, after the dominant assessment measures in each factor. The change patterns defined recognizable clinical presentations of patients with BPD. Patients with the global depression response pattern had a preponderance of "atypical" depressive symptoms, anxiety, and somatic complaints. They presented as withdrawn, pervasively depressed individuals, clearly distinguishable from the angry, demanding, entitled patients who made up the hostile depressed group. The schizotypal pattern defined the transient cognitive-perceptual, psychotic-like experiences of the borderline patient under stress. Symptoms of impulsive-behavioral dyscontrol were not initially identified as a separate symptom change pattern in this factor analytic study but were associated with both affective and cognitive dysregulation. Subsequent pharmacological research (detailed later in this article) has demonstrated the independence of impulsive-behavioral dysregulation as a separable symptom domain with discrete responsiveness to pharmacotherapy independent of affective or cognitive symptoms.

The neurobiology of personality dimensions transcends our definitions of Axis I and II disorders. The categorical definitions that separate Axis I and II are statistical constructs and are not based on etiology or neurobiology. By targeting personality dimensions for pharmacotherapy, we are making the basic assumption that closely related symptoms in Axis I and Axis II disorders may share a common etiology in neurotransmitter physiology. For example, we assume that the pathophysiology of mild thought disorders in Axis II patients may be related to the same dysfunction found in more severe thought disorders of Axis I patients. Similarly, disinhibition of affect and impulse may be mediated by a common neurotransmitter in BPD and some bipolar disorders. Severity and other disease factors clearly separate the clinical disorders; however, the common elements of pathophysiology suggest the possibility of shared responsiveness to medication. This principle remains the most rational guide for pharmacotherapy trials in the PD patient.

Using a dimensional perspective and a psychobiological model of personality dimensions, pharmacotherapy of the PD patient may be reframed as symptom-specific treatments. Treatment algorithms can be defined from research and case experience to provide assistance to the clinician in selecting medications for the principal target dimensions in the PD patient: cognitive-perceptual symptoms, affective dysregulation, and symptoms of impulsive-behavioral dyscontrol.

Development of a treatment algorithm for personality dimensions

Algorithms are defined as "rule-based deductive systems that operate with inputs, sequences, timeframes and outputs" (Jobson &: Potter, 1995, p. 457). In the setting of clinical medicine, algorithms provide decision trees assisting clinical judgments involving diagnoses and treatments. They reveal our decision-making processes, requiring that we justify each step in a treatment course. Rush and Prien (1995) divide this process into strategic decisions--determining what is wrong, whether to treat, and what treatment to use--and tactical decisions, concerned with how to conduct the treatment and what to do following success (or failure). Algorithms must incorporate an empirical database of research studies, but must also be guided by actual clinical experience, reflecting the judgments required in "real-life" applications to specific patients and situations. Patients seen in clinical practice rarely fit the artificial standards imposed for selection into controlled research trials. The practicing clinician is often faced with comorbid medical and psychiatric illnesses, wide heterogeneity in syndrome presentations, intolerance of side effects, patient preferences in choice of treatment, and social and economic factors affecting compliance, which are "controlled out" of the research trial (see Rush 6c Prien, 1995, for review).

Jobson and Potter (1995) have described a model for development of treatment algorithms primarily intended for Axis I affective disorders, which relies on reviews of treatment literature and consensus opinions of experts well versed in both research literature and clinical experience. These methods may be applied to address the treatment of Axis II personality dimensions. This article represents an initial offering of treatment algorithms for personality dimensions, based on literature review and the author's clinical experience. These algorithms are intended to promote interest, discussion, and research, which, it is hoped, will lead to formation of expert panels to further refine the algorithms.

Algorithms are rule-based systems. Three rules were applied to clinical decisions at each level: (1) Preference is given to medications for which efficacy has been most strongly supported by empirical research; (2) safer medications are given preference over those that incur more risk in overdose, abuse, or noncompliance; and (3) preference is given to rapidly acting drugs where clinical need is acute, although a slower acting agent might have more specificity (e.g., rapid treatment of psychotic symptoms with neuroleptics might take priority over treatment of an underlying depressed mood with antidepressants, although the psychotic symptom might be a direct product of the depressed mood). To be most broadly applicable and most conservative, the algorithms assume treatment in an outpatient setting.

We followed Jobson and Potter's (1995) convention of assigning a rating for each recommendation on the strength of available research evidence (e.g., A = supported by "multiple randomized, controlled trials," B = supported by "at least one randomized, controlled clinical trial," and C = "supported by opinion, case reports, and studies that do not meet randomized, control-trial criteria"). Well-controlled clinical trials with negative outcomes reduce the overall level of a recommendation based on several positive outcome studies. There are several important limitations to this method. Recommendations of efficacy based on single-trial studies (especially treatment of "first episodes" in previously medication-naive subjects) are not truly comparable to trials in patients who have failed multiple medication treatments (e.g., patients on the second or third step of a treatment algorithm). Given the diversity in methods, patient characteristics, and assessment measures found in the treatment literature, one can generalize claims for efficacy only after multiple replications. Statistically significant differences in research measures do not always translate into clinically important differences between patients on drug and placebo. In this regard, it is important to keep in mind the large effect sizes achieved by placebo treatments in pharmacotherapy trials in PD, although the placebo condition may be reported as less improved than the active drug.