The present study examined different concepts of categorical and dimensional stability of four PDs over 24 months with prospective data obtained in a multiwave design. The blinded repeated DIPD–IV assessment conducted 24 months after baseline revealed remission rates (based on DSM–IV diagnostic thresholds—a less stringent definition than used in the survival analyses discussed later) ranging from 50% (AVPD) to 61% (STPD). Lifetable survival analyses with prospective data revealed that the PD groups had significantly lower remission rates than the MDD comparison group. Although characterized by greater categorical stability than MDD, a substantial proportion of participants with PDs had remissions during the 24 months of follow-up. If a 2-month definition (2 months with two or fewer criteria) is used, remission rates range from 33% (STPD) to 55% (OCPD). Applying a much more stringent definition of 12 consecutive months with two or fewer criteria reveals lower remission rates, ranging from 23% (STPD) to 38% (OCPD). Collectively, these findings suggest that for these four PDs, substantial improvements in symptomatology are not uncommon (i.e., 23%–38% of patients) even when a stringent definition is used.
In terms of clinical entities, comparison of our temporal stability findings against previous studies of the longitudinal diagnostic stability of PD is difficult because of the inherent limitations of much of the literature (Grilo & McGlashan, 1999
; Grilo et al., 1998
; McDavid & Pilkonis, 1996
) and the lack of comparable studies with sufficient numbers of different PD diagnoses prospectively assessed with standardized instrumentation at repeated time points. A global comparison of our temporal stability findings (at threshold for diagnosis) is possible for BPD (several studies reviewed by McDavid & Pilkonis, 1996
; Zanarini, Frankenburg, Hennen, & Silk, 2003
) and for one study of AVPD (Ferro, Klein, Schwartz, Kasch, & Leader, 1998
). For AVPD, our kappa of .37 for the 2-year follow-up is higher than that reported in a 30-month follow-up study (κ = .24; Ferro et al., 1998
) of 108 depressed outpatients (of which n
= 13 met AVPD criteria). For BPD, a general comparison revealed similar rates of stability despite the varied follow-up intervals. One review (McDavid & Pilkonis, 1996
) reported the mean percentage of diagnostic stability for BPD across 10 studies as 57%, and Zanarini et al. (2003)
recently reported 64.5% stability for BPD at 24 months. We found that 44% (67 of 154 cases) of the BPD group was reassigned the BPD diagnosis at the 24-month blinded readministration of the DIPD– IV. This approach to stability (comparing two retrospective assessments at two different time points 24 months apart) compares the proportion of cases above and below DSM–IV
threshold for diagnosis. Our lifetable survival analyses with prospective data revealed that 58% remained BPD according to the 2-month definition of remission, and 72% remained BPD according to the 12-month definition of remission. These later two definitions represent more stringent thresholds and may reflect a more meaningful change in clinical status.
When stability was considered dimensionally, we observed a significant decrease in the mean proportion of criteria met in each of the PD groups suggesting decreases in severity over time. In contrast, when the relative stability of individual differences was examined, we found a high level of consistency. Collectively, these findings suggest that although individuals are quite consistent in terms of their rank order of PD criteria (individual differences in PD features are stable), they may fluctuate in the severity or amount of PD features over time. These findings are consistent with those reported by Lenzenweger (1999)
, which were also based on a multiwave methodology for a nonclinical sample of college students.
Relatively few differences were observed among the four study PDs in terms of their stability. Again, here the nature of the differences seemed to vary by how stability was considered. Life-table analyses with the 2-month definition of remission, but not the 12-month definition, revealed some variability between the PD groups. With the 2-month definition, STPD had a significantly lower remission rate than OCPD. Trends existed for STPD to have lower remission than AVPD and for BPD to have lower remission than OCPD. It is possible that these few observed differences for STPD might result, in part, from the methods of carrying forward the admixture of observational and symptomatic criteria in the lifetable analyses. When considered dimensionally in terms of the proportion of criteria met over time, the four PD groups differed overall, with the AVPD and BPD groups retaining significantly higher proportions of criteria met than the STPD and OCPD groups. However, inspection of the correlations (which showed that individual differences in PD features are stable) revealed little variability between PDs.
These findings suggest that PDs may be characterized by maladaptive trait constellations that are stable in their structure but that can change in severity or expression over time. Indeed, our correlational coefficients for proportion of criteria met (reflecting rank order stability) are similar to the range (.61–.70) for total number of PD criteria reported by Lenzenweger (1999)
for college students and are comparable with those generally reported for normal personality trait dimensions (McCrae & Costa, 1990
Significant decreases in mean counts observed over time are frequently observed in longitudinal studies. This is a complex issue and might reflect, in part, numerous clinical and methodological issues. One possible interpretation is that some of these changes simply reflect regression to the mean (Nesselroade, Stigler, & Baltes, 1980
). Two findings are relevant to this issue. First, our analyses with reliability-adjusted criterion counts revealed that the mean expected number of criteria for follow-up assessments, although lower, was not low enough to bring participants below the threshold of two criteria needed to achieve remission. Indeed, a very small proportion of the remissions defined as two or fewer criteria present could be accounted for on the basis of unreliability alone. Second, although the greatest drop in mean scores was observed from the baseline to the 6-month time point, inspection of the mean changes in criterion counts across the repeated time points in this multiwave design (Nesselroade et al., 1980
), that is, from the 6- to 12-month and the 12- to 24-month assessments, revealed further drops. Of course, in such longitudinal studies with repeated assessments it is certainly possible that reduced responsiveness to interviews over time can result in reduced reports of symptoms.
Another issue that concerns naturalistic studies of patients is the potential for confounding by treatment. Differences in treatment histories previously reported (Bender et al., 2001
) for the PD groups (BPD and STPD groups had greater treatment use than did the AVPD and OCPD groups) are not associated with the categorical stability observed here. Moreover, in the present study, our analyses revealed no significant effects of treatment intensity on the (dimensional) stability of PD criteria. This is consistent with analyses suggesting that in naturalistic studies the amount of treatment received is driven by problem severity (Cochran, 1983
We note several issues pertaining to generalizability. Our study focused on clinical entities or diagnoses and thus recruited patients and treatment-seeking participants. Although there exist inherent complexities in studying treatment-seeking patients (e.g., Berkson’s bias; Berkson, 1946
), this represents to our knowledge the first prospective multiwave study of several PDs. A previous prospective multiwave study (Lenzenweger, 1999
) could not evaluate categorical stability of PDs because of their low frequencies in the nonclinical sample. Our sampling across diverse clinical settings affiliated with universities in four northeastern urban settings produced a heterogeneous adult study group that approximated U.S. norms for ethnicity. Our recruitment targeted adults age 18 – 45 years. Findings may differ for younger (Grilo, Becker, Edell, & McGlashan, 2001
) or older (Seivewright, Tyrer, & Johnson, 2002
) patient groups, who may show different patterns of change over time.
This report provides support for the stability of general maladaptive personality trait constellations while highlighting that potentially meaningful clinical changes are possible and not uncommon in patients with PD. In this ongoing CLPS study, future analyses will attempt to examine important unresolved questions about the nature of PDs. Some of the complex patterns observed in this report may reflect, in part, the heterogeneity of the criteria sets both within and across PDs. For example, some criteria are clearly more trait-like, in contrast to others that are more behaviorally anchored (Sanislow & McGlashan, 1998
). Hence, the meaning of stability may be different for the different types of criteria, highlighting the need for a clearer understanding and identification of the important symptom dimensions and personality traits that may underlie PDs (Shea, 1992
). Moreover, it is important to recognize the possibility that even if it appears that a similar number of criteria are met at different time points, it is possible that these might be different criteria. Future studies will consider such issues by examining (a) the nature of stability and change of specific criteria, (b) their predictive validity (their diagnostic efficiency and associations with psychiatric disorders and with broader aspects of psychosocial functioning), and (c) different or alternative groupings of criteria. Studies will evaluate the prospective stability of various degrees of PD persistence and will also attempt to delineate factors that may predict stability or change, such as stressful life events and changes in Axis I psychopathology or social functioning (Grilo, McGlashan, & Skodol, 2000
). Although this report considered several approaches to testing stability, we do not claim that the analyses presented here are exhaustive. Future waves of this ongoing prospective study will allow for consideration of different conceptual and statistical approaches to stability and change. Collectively, these empirical studies may inform the development and operationalization of optimal definitions for remission, recovery, and relapse for PDs that have facilitated research and clinical judgments in other areas (Frank et al., 1991