It is of interest to know if temporary and persistent personality disorders are associated with different outcomes.
A cohort of 210 people with anxiety and depressive disorders was followed up on nine occasions over 12 years. During this study, personality status was assessed at baseline and after 2 years using two methods, one linked to the new International Classification of Diseases 11th Revision (ICD-11) severity codes. The impact on the symptomatic outcome and social function of temporary (i.e. personality disorder on one occasion only) and persistent personality disorder (personality disorder present on both occasions) was compared.
Of the 162 patients studied we identified four groups (no personality disorder at any time (n = 46), two with temporary personality disorder (baseline only (n = 33) and 2 years only (n = 28), and persistent personality disorder (n = 55). Those with persistent personality disorder had significantly worse outcomes than other groups for self-rated anxiety symptoms (p = 0.02) and overall social function (p < 0.001), 81% had a current DSM diagnosis at 12 years compared with 52–65% in the other groups (p < 0.03). Significantly, more patients with ICD-11 moderate or severe personality disorder at baseline had persistent personality disorder than had temporary disorders (p = 0.017).
Persistent personality disorder is associated with more severe personality dysfunction and has a negative impact on the outcome of common mental disorder and particularly on long-term social functioning. Copyright © 2016 John Wiley & Sons, Ltd.
There has been increasing interest in the impact of personality status on the outcome of all mental disorders in recent years. The collective evidence suggests that personality disturbance in the absence of any intervention designed to correct it has a negative impact on the outcome of depression (Newton-Howes et al., 2006; Newton-Howes et al., 2014; Gorwood, Rouillon, Even, Falissard, Corruvle & Moran, 2010; Renner et al., 2014), and there is now evidence that this applies equally to anxiety (Skodol et al., 2014). But, there have also been contradictory reports that suggest that personality disorder may not impair outcome with some treatments (Mulder, 2002). In conjunction with these findings there has also been increasing acknowledgment that a single personality assessment at one point in time is not sufficient to confirm a diagnosis of personality disorder (Clark, 2007; Tyrer et al., 2007).
In the course of a long-term cohort study linked to a randomised trial (Tyrer et al., 1988) with clinical assessment at 12 years we measured personality status at baseline and also after 2 years, so this gave the opportunity of testing whether those with persistent personality disorder (those who had personality disorder on both occasions) were associated with different outcomes that those who only had personality disorder diagnosed on one occasion only.
The original design of the study (the Nottingham Study of Neurotic Disorder) is described elsewhere (Tyrer et al., 1988). The participants were all patients seen at general practice psychiatric clinics in Nottingham, UK. These clinics were staffed by psychiatrists but patients were referred somewhat earlier than to out-patient clinics, often before full attempts at treatment had begun. The participants were involved in a randomised trial of drug treatment, cognitive behaviour therapy and self-help for the first 10 weeks of the study. The inclusion criteria were (a) a Diagnostic and Statistical Manual of Mental Disorders, 3rd Edition ( DSM-III) diagnosis of generalised anxiety disorder, dysthymia disorder or panic disorder (including anxiety–depression combinations of these—cothymia)(Tyrer et al., 2001); (b) taking no psychotropic drugs at the time of assessment; and (c) informed consent to take either drug or psychological treatment.
At baseline assessments were made of observer-rated anxiety and depressive symptoms using the Brief Scale for Anxiety (Tyrer et al., 1984), and the Montgomery & Åsberg Depression Rating Scale (Montgomery & Åsberg, 1979), total symptomatology using the Comprehensive Psychopathology Rating Scale ( Åsberg et al., 1978) and self-rated anxiety and depression using the Hospital Anxiety and Depression Rating Scale (HADS)(Zigmond & Snaith, 1983). Personality assessment was made at baseline and after 2 years using the Personality Assessment Schedule (PAS) (Tyrer & Alexander, 1979) . The PAS scores 24 personality traits on an eight point scale and separates personality disturbance into four levels of severity: no personality disorder, personality difficulty, simple personality disorder and complex (diffuse) personality disorder (Tyrer & Johnson, 1996) that shows some similarity with the proposed ICD-11 classification of personality disorder (Tyrer et al., 2011). Five raters assessed personality status using the PAS at baseline and were trained until they all reached satisfactory levels of reliability (Landis & Koch, 1977).
After 12 years patients were assessed again (by HT) without any knowledge of the results of previous records and all the original questionnaires and schedules repeated in a standard sequence. A high follow-up rate was achieved, and the conclusions of the data remained robust after multiple imputation of missing values (Longford et al., 2006). In addition the current psychiatric diagnosis was assessed using both the original DSM -III diagnosis and the DSM-III-R version of SCID (Spitzer & Williams, 1987). Assessment of social function was also made using the Social Functioning Questionnaire, a self-rated eight-item questionnaire (Tyrer et al., 2005). A longitudinal assessment of outcome, the Neurotic Disorder Outcome Scale (NDOS) (Tyrer, Seivewright & Johnson, 2004), itself adapted from a similar scale constructed by Surtees and Barkley (1994)(the Depression Outcome Scale), was also completed. The NDOS includes 10 measures suggestive of poor outcome, and each is rated by a single point (a score of 3 or more indicates a poor outcome) .
The study was approved by the Nottingham Ethics Committee.
STATA 13 was used to perform all the analyses. Descriptive statistics are presented as means and standard deviations (SD) for normally distributed variables, frequency (percentage) for categorical variables and median (inter-quantile range) for skewed variables. For group comparison, one-way ANOVA were used for normally distributed variables, Kruskal–Wallis test were performed for skewed variables and χ2 test was applied for categorical variables.
Baseline differences between personality groups
Of the 210 patients assessed at baseline 198 had their personality status recorded, but after 2 years only 162 had assessments at both baseline and 2 years. The data were therefore confined to these 162 patients.
The main demographic features of the four groups separated by personality disorder status are shown in the succeeding texts (Table 1). Only 46 patients (28.4%) had no personality disorder diagnosed at both baseline and 2 years. More patients with dysthymia and cothymia, and with moderate or severe personality disorder at baseline had persistent personality disorder. Of those with moderate and severe personality disorder (groups combined as only 3 had severe personality disorder), all but one were in the persistent personality disorder group (Χ2 = 30.9, degrees of freedom 3, p < 0.001).Demographic, diagnostic and baseline characteristics of each group
|No personality disorder at either 0 or 2 years Group A||Temporary personality disorder 1 (baseline only) Group B||Temporary personality disorder 2 (2 yrs only) Group C||Persistent personality disorder (present at 0 and 2 years) Group D||Significance of differences (P)|
|Age (mean, SD)||39.3(14.8)||37.3(11.4)||35.4(8.1)||38.9(13.4)||0.51|
|Initial diagnosis of GAD (%)||16(30.8)||15(28.9)||9(17.3)||12(23.08)||0.025 for all diagnoses|
|Initial diagnosis of dysthymia (%)||9(18.8)||8(18.75)||5(10.4)||25(52.08)|
|Initial diagnosis of panic disorder (%)||21(33.9)||9(14.5)||14(22.6)||18(29.03)|
|Mixed anxiety and depressive disorders (%)||10 (21.7)||11 (33.3)||5 (17.9)||29 (52.7%)||0.002|
|Proportion with moderate or severe pd (%) >=3||0||1 (5.9%)||0||16 (94.1%)||<0.001|
Outcome at 12 years
In all analyses carried out with the outcome data at 12 years there was evidence of greater symptomatology in those with persistent personality disorder, with most changes shown in the anxiety section of the HADS (p < 0.03), and in social dysfunction (p < 0.001), with those with persistent personality disorder having a mean score of over 10, equivalent to those with severe mental illness (Tyrer & Simmonds, 2003). Persistence of personality disorder also was associated with higher scores on the Neurotic Disorder Outcome Scale (Table 2). More patients with persistent personality disorder had a DSM diagnosis at 12 years. When the data were analysed by baseline ICD-11 status (Tyrer et al., 2011; Tyrer Crawford, Sanatinia et al. 2014) those with moderate or severe personality disorder at baseline, like the PAS diagnostic groups, were significantly over-represented in the persistent personality disorder group (p = 0.017) (Table 3).Descriptive statistics with scores at baseline and outcome at 12 years
|Outcome||Patient Group||Mean (base)||SD (base)||Means (12 years)||SDs (12 years)||Mean change||SD (chg)||P value of change comparison|
|(P25, p75)||Temporary 2||1||(1,3)|
|Severity of personality disorder using ICD-11 coding at baseline|
|Type of personality disorder||Mild (%)||Moderate/severe (%)||Total|
|Temporary||13 (39.4)||2 (9.5)||15|
|Persistent||20 (60.6)||19 (90.5)||39|
|Total (%)||33 (100.0)||21 (100.0)||54|
The finding of a graded impact of personality disorder on the outcome of anxiety and depressive disorders, with those with persistent personality disorder associated with the worst outcome, temporary personality disorder somewhat better, and no personality disorder with the best outcome supports the main hypothesis. The differences were most marked for social functioning and long-term morbidity, and this is in keeping with other evidence that impaired social function is a more consistent finding in personality disorder than behaviour and symptoms (Seivewright et al., 2004; Gunderson et al., 2011).
The findings are also consistent with other long-term follow-up data that show clinical assessments of personality disorder status to be less reliable than self-rated trait based dimensional assessments (Morey et al., 2012). Clinical assessments of personality disorder may be persistent over variable periods (Lopez-Castroman et al., 2012), but in ordinary clinical practice they are less persistent than other mental disorders (Baca-Garcia et al., 2007). But, a lot of this instability is related to the existing categorical classification. The differences in outcome shown between moderate and mild personality disorder in the ICD-11 classification show the advantages of a single-dimensional rating of personality disorder, and offset criticisms that such a simplified version of personality assessment is inadequate for clinicians prevented from using formal diagnostic categories (Pedersen et al., 2013). The evidence that there is good discrimination between the outcomes of those with mild and moderate personality disorder, without the need for further categories and qualified only by domain traits (Kim et al., 2015), shows an important advantage of the ICD-11 dimensional system that is likely to be copied in future editions of DSM.
There have now been many studies that show a surprising degree of improvement in personality disorder over a long time period, of which the McLean Hospital studies by Mary Zanarini (2012) and the Collaborative Longitudinal Personality Disorder Study (Skodol et al., 2010) are the most prominent. But, these did not record personality disorder by severity, and it is hoped that further studies will now do so.
Personality disorder is not as easy to assess in clinical practice as disorders of mental state, although it may be becoming easier with changes in classification that allow general physicians to make the diagnosis (Tyrer et al., 2015). Because of concern over the pejorative nature of the condition, there is often reluctance in making a diagnosis that is in danger of becoming a permanent label. The results of this study emphasise the importance of not being certain about a personality disorder diagnosis at one contact only and to take care in repeating the assessment later before concluding the exact status of the personality. But, this concern is to some extent offset by the finding that those with moderate or severe personality disorder at baseline were more likely to be in the persistent group. A thorough assessment that suggests a more severe personality disorder may therefore be sufficient for a confident diagnosis. Further enquiries on what leads to long-term persistence would also be helpful.
There is increasing evidence that personality disorders are frequently highly comorbid with other psychiatric and medical illness (Yang et al., 2010; Hengartner et al., 2014; Sanatinia et al., 2015) and so it is reasonable to consider whether persistent personality pathology is the cause of greater symptoms or a consequence of them. The tendency in much of the literature is to assume primacy of the mental state diagnosis but this is now being challenged (Munjiza et al., 2014; Tyrer, 2015). There is also concern that the evidence base for the long-term course of personality disorders is nit helped by studies that examine the outcome of categorical personality disorder only. As Newton-Howes and colleagues (2015) aptly put it: ‘A major clinical difficulty is the relative absence of randomised controlled trials for interventions aimed at personality trait domains and functional outcomes, as opposed to specific personality disorder types such as borderline personality disorder’ (Newton-Howes, Clark & Chanen, 2015). The classification of personality disorder on a single dimension in this study has shown the advantages of the dimensional approach and its even longer-term value is being tested in the 30-year follow-up of this cohort, which developed from a randomised trial of treatment that has already shown long-term interactions between personality status and treatment allocation (Tyrer et al., 1993).
This research was funded by the Mental Health Foundation, the Nicola Pigott Memorial Fund, the Research Committee of the Trent Regional Health Authority, the National Offender Management Service and the National Institute for Health Research: Imperial Biomedical Research Centre. We thank Tony Johnson for statistical advice, Lee Anna Clark for pertinent comment and Mike Crawford, Scott Weich, Katarina Miloseska-Reid and Maria Zauter-Tutt for additional support. This paper was first presented at the 16th Annual Meeting of the British and Irish Group for the Study of Personality Disorders in March 2015.