Although the general consensus among clinicians is that patients who present with comorbid depression and personality disorders have a worse outcome, this is not always supported by research findings. Several large clinical investigations did find evidence to support an adverse outcome in depressed patients with comorbid personality disorders.
In a study designed to examine the effects of personality traits (as determined by the Structured Clinical Interview for DSM-III Personality Disorders [SOD]) in depressed patients receiving desipramine over 4-5 weeks, responders had a significantly lower total personality trait score and a significantly lower Cluster C personality trait score than nonresponders (). Those with a sustained response after 6 months were found to have significantly lower Cluster B, Cluster C, and total personality scores. In a study assessing predictors of response and remission, Ezquiaga and colleagues (1998) found that the presence of a personality disorder, as assessed by DSM-III-R criteria, was a strong predictor of outcome and was associated with significantly less likelihood of response or remission. Bschor and colleagues (2001) examined retrospectively predictors of a good response to lithium augmentation in a group of depressed patients who failed to respond to tricyclic antidepressants (TCAs). Absence of comorbid personality disorders was a significant predictor of response to lithium augmentation.
In a large tertiary-based study involving over 600 patients and designed to investigate predictors of response to sertraline or imipramine in patients with chronic or double depression, Hirschfeld and colleagues (1998) reported that the presence of a comorbid personality disorder (measured by the SCID-II, based on DSM-III-R criteria) did not predict treatment response. When individual personality disorders were examined, however, a diagnosis of passive-aggressive personality disorder was predictive of a positive response to treatment. Similarly, Fava and colleagues (1994) evaluated the predictive value of personality disorder comorbidity in a group of patients who received fluoxetine treatment. Based on the Personality Diagnostic Questionnaire-Revised (PDQ-R), the presence of a Cluster B diagnosis before treatment predicted positive antidepressant response. However, no differences in response were found in patients with or without a Cluster A or C diagnosis.
In contrast, Joyce and colleagues (1994) found that neither individual personality disorders nor clusters (DSM-III-R) were predictive of treatment outcome, although there was a trend toward those with a borderline personality disorder having a poorer outcome.
Almost a decade later, Mulder (2002) published the most comprehensive review to date on the influence of both personality disorders and dimensions on pharmacotherapy and psychotherapy outcomes. This review of 27 studies using DSM Axis II criteria to examine the relation between personality comorbidity in MDD and treatment outcome found conflicting results. Fifteen studies reported a worse outcome, five reported a partially worse outcome, and seven reported no difference in treatment outcome. Overall, the author concluded that the best-designed studies reported the least effect of personality pathology on depression treatment outcome. With the same group of studies, Mulder and colleagues (2003) also examined the interaction between antidepressant class (selective serotonin reuptake inhibitor [SSRI; fluoxetine] and TCA [nortriptyline]), personality disorder, and outcome. They reported that the presence or absence of a comorbid personality disorder (assessed by the SCID-II, based on DSM-III-R criteria) did not significantly affect outcome. However they confirmed an interaction between presence of a comorbid personality disorder, drug type, and outcome. Although there was a trend for all patients treated with nortriptyline to have a less favorable outcome, those patients with a Cluster B personality disorder who received nortriptyline demonstrated significantly less improvement. These findings support the idea that individuals with Cluster B personality disorders, in particular those with borderline personality disorder, respond better to SSRIs than to TCAs. The findings are also consistent with previous reports that patients with a Cluster B diagnosis at baseline assessment demonstrated a significantly greater decrease in Hamilton Rating Scale for Depression (Ham-D) scores with fluoxetine treatment compared with those without a Cluster B diagnosis ().
The Seven-Factor Model of Temperament and Character
Several investigators have examined the influence of TPQ- and TCI-derived dimensions and temperament types on treatment response. Using the TPQ, Joyce and colleagues (1994) predicted differential treatment responses to clomipramine and desipramine in depressed outpatients. Although none of the dimensions of the Eysenck Personality Inventory predicted outcome, temperament as measured by the TPQ accounted for 35% of the variance in treatment outcome, which increased to 49% when the sample was limited to the severely depressed group. By contrast, less than 5% of the variance was predicted by clinical variables. Depressed patients with high reward dependence and harm avoidance scores, or both, had a good outcome regardless of sex or drug. However, in women high reward dependence was associated with a better response to clomipramine, whereas high harm avoidance was associated with a better response to desipramine. The authors speculated that reward dependence is related to features of atypical depression, including interpersonal sensitivity, and that patients with these temperamental traits may preferentially respond to serotonergic-acting medications, including clomipramine.
Unfortunately, these findings have proven difficult to replicate. Nelson and Cloninger (1997) evaluated more than 1,000 depressed subjects receiving open-label treatment with nefazodone. Although they also found that the levels of reward dependence and harm avoidance and their interaction had significant predictive value, this accounted for only 1.1% of the variance.
Similarly, Newman and colleagues (2000) failed to replicate these results in a study involving fluoxetine treatment for depressed outpatients. There was no correlation between harm avoidance, reward dependence, or novelty seeking and percent improvement in depression score. These authors state that their failure to replicate previous findings may in large part be because of the use of different classes of antidepressants (SSRIs as opposed to TCAs) as well as not maximizing fluoxetine dosages. Using the TCI to determine possible predictors of response to the heterocyclic antidepressant maprotiline, Sato and colleagues (1999) were also unable to confirm that any of the seven personality dimensions had predictive value.
In summary, although the original results of the Joyce and colleagues (1994) study are thought provoking, these findings have been difficult to replicate.
Neuroticism and Extraversion
The relationship between neuroticism and treatment response in MDD has received considerable attention. In a comprehensive review of 13 studies that have examined the relation between neuroticism and treatment outcome, Mulder (2002) found that high neuroticism scores generally predicted worse treatment outcome in depressed patients, although he also concluded that in the best-designed studies, personality pathology has the least effect on depression treatment outcome. In a large randomized, controlled trial involving over 300 patients in primary care who met criteria for dysthymia, those with high neuroticism scores, as assessed by the NEO PI (), were less likely to achieve remission (). This finding applied to both problem-solving and paroxetine treatment groups. The authors cite literature suggesting that individuals with high neuroticism report lower tolerance for stress, lower self-esteem, and a greater likelihood of feeling victimized and resentful. They also report that high neuroticism has been shown to be predictive of adverse life events and is associated with an increased perception of severity of life stressors ().
Some have shown neuroticism to be a predictor of poor long-term outcome, whereas others have failed to confirm such a relationship. Pe-tersen and colleagues (2002) found that neuroticism was not a significant predictor of clinical response in a 6-week open trial of fluoxetine in depressed outpatients. Similarly, Bagby and colleagues (1995) did not find that neuroticism predicted treatment outcome. Consistent with previous research, the mean neuroticism score for depressed patients pretreatment was almost two standard deviations higher than scores in a normative sample. The authors also reported that neuroticism score was significantly correlated with severity of depression scores, and there was a significant drop in neuroticism scores with treatment, although neuroticism scores for recovered patients remained more than one standard deviation higher than the normative sample (). These results suggest that elevated neuroticism scores in a sample of depressed patients may not be explained entirely by depressive symptomatology at the time of the initial assessment.
A similar pattern of results was described for extraversion, although the magnitude of change and correlation with measures of depression was less than that observed for neuroticism scores. Patients who ultimately recovered were less than one standard deviation lower on extra-version compared with the normative sample and less than half a standard deviation lower compared with the normative sample following treatment. Such results suggest that low extraversion may not be a specific vulnerability factor for MDD. However, extraversion may have an important role to play in course and/or treatment outcome (). Indeed, in a regression analysis using neuroticism and extraversion scores as predictors of treatment outcome, only extraversion scores predicted a significant reduction of depressive symptoms, with higher extraversion scores, in particular the “gregariousness” facet, predieting a greater reduction of symptoms. Moreover, extraversion scores at treatment entry for patients who did not recover were more than two standard deviations below the normative sample. These results were later replicated ().
Other cognitive variables, including rumination and distraction, also have been linked to vulnerability and recovery from depression. Bagby and Parker (2001) investigated the relationship between neuroticism and extraversion as measured by the NEO-PI-R and the cognitive constructs, rumination and distraction, as measured by the Response Style Questionnaire (RSQ) (), before the initiation of a 14-week trial of pharmacotherapy in 168 patients with major depression. They found that remission defined as both a decrease of at least 50% in Ham-D scores and a final score of less than 9 (after 14 weeks of treatment) was significantly associated with extraversion and distraction but not related to neuroticism or a ruminative response style. Partial correlations were calculated to determine whether the contributions of extraversion and distraction were independent of their shared variance with one another. After controlling for distraction, the relationship between extraversion and treatment outcome was preserved, although the correlation between distraction and outcome was no longer significant after controlling for extraversion.
Overall the above results suggest that neuroticism may be a vulnerability factor for MDD, whereas extraversion may moderate treatment outcome (). Differences across studies in the degree to which neuroticism predicts outcome may be partly explained by examining whether the investigators covaried out baseline depression. When baseline severity of depression is covaried from the predictor model, neuroticism does not predict outcome (). In summary, results of the relationship between personality comorbidity and MDD are quite mixed. In general, the better controlled the study, the less evidence there is for personality pathology affecting outcome ().
Selections from the book: Depression and personality : conceptual and clinical challenges (2005)