A major goal of the CLPS has been to empirically test the stability of personality disorders. We have found that fewer than half of PD patients remain at or above full criteria every month over intervals as short as 1 or 2 years (
Shea et al., 2002;
Grilo, Shea et al., 2004). More than half of PD patients show what we have called a “remission,” defined as at least 12 consecutive months with no more than two criteria of their baseline disorder within the first 2 years of follow-up. Ten percent of patients with BPD remitted in the first 6 months, most often in association with situational changes, such as leaving stressful relationships, raising questions about whether certain PDs are more temporally fluctuating than previously assumed (
Gunderson et al., 2003). Personality disorders were, however, more stable than MDD. These changes in PD diagnoses have been examined to see if they were the result of methodological artifacts, such as the effects of concurrent MDD, measurement by interview versus self-report, the influence of repeated interviews, or rater unreliability or drift. None of these accounted for the degree of improvement observed.
Viewed as dimensions, however, PDs showed considerably more stability. Although the mean number of criteria decreased over time for each group, a continuous measure of number of criteria met was highly correlated across assessments during the first 2 years of follow-up. These findings suggest that PDs may be characterized by maladaptive trait constellations that are stable in their structure (individual differences), but can change in severity or expression over time.
Furthermore, some diagnostic criteria for PDs are more stable than others—findings that are important in the search for phenotypes. For example,
affective instability was the most stable of the BPD criteria over the first 2 years of follow-up, followed by
inappropriate, intense anger (
McGlashan et al., 2005). The least stable BPD criteria were
frantic efforts to avoid abandonment and
self-injury. For AVPD, the most stable criteria were
feels socially inept and
feels inadequate and the least stable was
avoids jobs with interpersonal contact. These findings have led us to hypothesize that PDs may be reconceptualized as hybrids of two elements: (1) stable personality traits that may have normal variants, but that in PDs are pathologically skewed or exaggerated, and (2) dysfunctional behaviors that are attempts at adapting to, defending against, coping with, or compensating for these pathological traits (e.g., self-cutting to reduce affective tension, avoiding work situations involving many people because of shyness).
We have also demonstrated that traits of general personality functioning (e.g., Five-Factor traits) tend to be stable, with stability estimates in the
r = .70 to .80 range over 2 years (
Warner et al., 2004). However, when study patients change on these traits, the changes are followed by lagged changes in PD psychopathology across the spectrum of PDs. Importantly, these relationships are nonreciprocal, in that changes in PD features are not predictive of subsequent changes in personality traits.
The association in the course of co-occurring personality disorders and Axis I disorders over time was examined to test predictions of specific longitudinal associations derived from the model of psychobiological dimensions hypothesized by
Siever and Davis (1991) to cut across the psychopathology represented by both Axis I and II (
Shea et al., 2004). Using time varying analyses, we showed that despite substantial comorbidity with Axis I disorders, the timing of course changes in PDs has been relatively independent of changes in the course of Axis I disorders. The exceptions are BPD and both MDD and PTSD, and AVPD and social phobia. These disorders may share fundamental underlying dimensions of psychopathology in the affective and anxiety domains, respectively. Although changes in both Axis I and Axis II disorders could each be shown to precede changes in the other, improvements in BPD were considerably more often followed by improvements in MDD than vise versa (
Gunderson et al., 2004). Therefore, clinicians should not ignore BPD in hopes that treatment of MDD will be followed by improvement of borderline psychopathology.
These findings lead to several critical questions to answer in the future. Among those patients with PDs who “remit,” what is the probability of “relapsing?” If relapse rates are low, this would suggest that a substantial proportion of individuals, who meet criteria for a PD using state of the art measures, cannot be characterized by the general DSM PD criterion of an “enduring” or “stable” pattern of maladaptive behaviors or traits. Our ability to characterize patients who turn out to be “false positives,” using the traditional definition of a PD, would be valuable, particularly for family or genetic studies. On the other hand, if relapse rates are substantial, this would suggest a subgroup of PDs with a fluctuating course, in contrast to one that is uniformly stable. Such relapses would also provide the opportunity to examine the role of mediating factors, such as positive and negative life events or minority status, in altering the course of PDs.
A second critical question is, do subjects who have not yet remitted remain ill indefinitely, or do they simply remit more slowly? If future rates of remission are low, this observation would be consistent with the notion of an enduring pattern of maladaptive traits and behaviors at least for a core group of patients with PDs.
Finally, will the initial longitudinal relationships of pathological traits, PDs, and Axis I psychopathology hold up over time? Such relationships point strongly toward shared endophenotypes, whose identification is critical for genetic studies, treatment development, and classification.
The publisher's final edited version of this article is available at 
