Which factors are associated with duration of untreated illness in borderline personality disorder?

Borderline personality disorder (BPD) is a prevalent condition associated with high rates of hospitalizations. The purpose of this manuscript was to detect the factors associated with duration of untreated illness (DUI) in BPD.


| INTRODUCTION
Borderline personality disorder (BPD) represents the most diagnosed personality disorder, reaching a prevalence of 10%-12% among outpatients of community mental health services (Ellison et al., 2018).It is characterized by a pervasive pattern of instability in interpersonal relationships, mood oscillations, marked impulsivity and tendency to dangerous behaviours (APA, 2013).Available literature on this topic reports that about 76% of subjects with BPD attempt suicide in the course of life, up to 10% die by suicide, and the prevalence of crimes is significantly above the average (Goodman et al., 2017;Paris, 2019).
Moreover, these patients experience recurrent hospitalizations, accounting for important costs to health systems (Leichsenring et al., 2011), and this can be related both to the intrinsic characteristics of BPD and to the concomitant presence of mood disorders (Ceresa et al., 2021;Soloff & Chiappetta, 2017).Despite this, surprisingly few studies have so far investigated which factors affect the prognosis of BPD.Type of symptoms, substance misuse, unemployment, childhood maltreatment and the comorbidity with anxiety disorders have been associated with poor outcome in BPD (Woodbridge et al., 2021;Zanarini et al., 2018).Furthermore, some preliminary findings indicate that epigenetic modifications (e.g., in methylation status of brain-derived neurotrophic factor gene) may modulate response to psychotherapy (Knoblich et al., 2018).
In this context, some authors highlighted the importance of an early diagnosis and treatment of BPD (Bohus et al., 2021).Different factors can contribute to a delay of diagnosis, such as the lack of attribution of BPD in adolescence, lack of awareness of this condition, the accessibility to psychiatric services, the personal attitude towards treatment and the stigma related to mental conditions (Li et al., 2016;Miller et al., 2008;Sheehan et al., 2016).A consistent literature established that duration of untreated illness (DUI) is a variable influencing negatively the course of several mental disorders, such as bipolar disorder, depression, obsessive-compulsive disorder and schizophrenia (Altamura, Buoli, et al., 2015;Buoli et al., 2020;Dell'Osso et al., 2010;Howes et al., 2021;Murru & Carpiniello, 2018).
To our knowledge no studies have been published till now about the specific role of DUI in affecting outcome and course of patients affected by BPD (Bozzatello et al., 2019).Objectives of the present research were to identify factors associated with a longer DUI in patients affected by BPD and to verify whether a longer DUI is related to an unfavourable course in BPD, as in other psychiatric conditions.

| METHODS
A sample of 152 patients with a diagnosis of BPD according to diagnostic and statistical manual of mental disorders (APA, 2013) criteria was identified among outpatients followed up at psychiatric services of Fondazione IRCCS Policlinico (Milan, Italy) (N = 107) and ASST Monza (Monza and surroundings) (N = 45) between January 2007 and February 2021.All diagnoses were made by an expert senior psychiatrist through a clinical interview.Clinical data were collected referring to the last visit and they were obtained primarily through a screening of the clinical charts and Lombardy electronic database (psicheweb), and through interviews with patients and their relatives if data were missing.Inclusion criteria consisted of: age ≥18, diagnosis of BPD, contact at the aforementioned outpatient psychiatric services for at least 3 months.The only exclusion criterion was contact with the psychiatric services for less than 3 months.In case of psychiatric comorbidity, BPD represented the disorder that affected the patients for the longest time and was responsible for more disability (restriction of daily activities and social function as a result of BPD symptoms).The protocol of this study was approved by the local Ethical Committees (identification number of the study: 2064).
We collected socio-demographic and clinical data, which can be seen in detail in Tables 1 (socio-demographic variables) and 2 (clinical variables).Duration of illness was calculated as the time between age at onset (start of symptoms of BPD) and current age (Altamura, Serati, & Buoli, 2015).DUI was considered as the time elapsing between the onset of the disorder and the prescription of a proper treatment (pharmacotherapy with antidepressants, mood stabilizers, antipsychotics or dynamic/cognitive psychotherapy or a combination of both according to predominant psychopathological dimension or biological vulnerability due to family history for specific psychiatric disorders) (Altamura et al., 2010;Stone, 2019).
Descriptive analyses of the total sample were performed.Correlation analyses were performed to analyse the relation between DUI and quantitative variables.The length of DUI was compared between groups defined by categorical variables through one-way analyses of variance (ANOVAs).The statistically significant variables in the univariate analyses were then inserted in regression analyses: one model was performed for categorical variables (binary logistic regression) with DUI as independent variable, and one for quantitative ones (multivariate linear regression model).The level of statistical significance was set at p ≤ .05.All statistical analyses were performed using SPSS (version 26; Chicago, Illinois, United States).

| RESULTS
The total sample included 152 patients, whose mean DUI was 3.50 (±6.83) years.The results of descriptive and correlation analyses as well as the values of DUI according to categorical variables are reported in Tables 1 (socio-demographic variables) and 2 (clinical variables).
Patients' current main pharmacological treatment is reported in Table 3.With regard to the relation between a longer DUI and an unfavourable course of BPD, the only significant result was a longer DUI in patients with post-onset substance abuse disorders than those without post-onset substance use disorders (F = 3.86, p = .05).A binary logistic regression analysis was then performed to confirm this result.The model was reliable (test di Hosmer-Lemeshow: χ 2 = 3.08, p = .38),allowing for a correct classification of 60.5% of the cases.

| DISCUSSION
To our knowledge this is the first study investigating specifically the role of DUI in influencing clinical features, course and outcome of patients affected by BPD.Even though not confirmed by regression analysis, ANOVAs showed that patients with the presence of substance misuse after the onset of BPD had a longer DUI compared to patients without post-onset substance misuse.This finding is not surprising because patients who wait to schedule a visit to mental health services often resort to substances as an attempt to selfmedicate their symptoms, and this worsens their prognosis (Heath et al., 2018).It must also be considered that many outpatient clinics do not accept to manage patients with comorbid substance use disorders, addressing them to services specialized in the treatment of substance-related disorders (Vitali et al., 2018).Another correlation that emerged in the present study (although not confirmed by regression analysis) was between longer DUI and an earlier age at onset.
Indeed, the diagnosis of BPD can be challenging in adolescents, especially because it is still debated whether it is legitimate to diagnose BPD before 18 years and which is the optimal treatment for these patients (Guilé et al., 2018).
The utility of early intervention in BPD patients is not clear.A recent review has documented the effectiveness of early intervention for BPD, although the data should be taken with caution due to the limitations of the included studies (Chanen et al., 2020).Of note, there is a lack of international agreement about the proper treatment for this condition, and few evidence-based recommendations can help the clinicians to manage efficaciously patients affected by BPD (Simonsen et al., 2019).Psychotherapy is always recommended, while the use of pharmacotherapy is debated and should be limited to patients with moderate to severe symptoms or psychiatric comorbidity (Stone, 2019).In light of this, the prescribed treatments are not uniform, as emerged from both the literature and our sample (Magni et al., 2021).
According to our data, unlike other psychiatric disorders, DUI did not seem to affect clinical course and outcome of BPD.The mean DUI of our sample was 3.50 years, a value that is similar to that reported in studies investigating the effects of DUI in samples of patients affected by schizophrenia or mood disorders, which is respectively 4 and >2 years (Altamura et al., 2010;Zoghbi et al., 2022).The dissimilar effect of DUI in BPD compared to other psychiatric disorders can be explained by the different clinical course: while schizophrenia and bipolar disorder are characterized by a progressive worsening of symptoms (Altamura, Buoli, et al., 2015;Buoli et al., 2017), several studies documented how patients with BPD often experience a remission of their symptoms as well as a greater clinical stabilization with the progress of age (Videler et al., 2019).In the light of the results of our study, there seems no doubt that prospective studies with robust outcome measures are needed to clarify the association between DUI and outcome for this population.
The present study has some limitations: the retrospective collection of data, the absence of a follow-up, the limited number of out- hospitalizations and suicidal attempts in BPD.However, other factors, potentially modifying DUI in patients with BPD, should be investigated by future research, including the availability of specialized centres for the treatment of this disorder, the patients' attitude towards medical treatment or other cultural aspects such as the fear of the stigma associated with mental disorders (Nakanishi et al., 2021).For these reasons, prospective studies with larger samples are needed to confirm the results of the present study.
come variables, and the lack of validated rating scales to assess patients' symptom severity.Finally the results of this article can be biased by the fact that our sample had a long duration of illness, while an ideal sample should include patients with recent onset of illness to have an optimal evaluations of the effects of DUI on the course of BPD.Taken as a whole, the present study indicates that delayed treatment might not affect some outcome variables such as number of T A B L E 3 Patients' current main pharmacological treatment.
Socio-demographic variables of the total sample and values of DUI according to these factors.In bold statistically significant p for the following analyses: analyses of variance tests for categorical variables and correlation analysis for age.Mean ± SD of total sample are reported for age.The percentages referred to the total sample are reported into brackets for categorical variables.Abbreviations: DUI, duration of untreated illness; SD, standard deviation.Clinical variables of the total sample and values of DUI/Pearson's r according to these factors.
T A B L E 2Note: In bold statistically significant p for the following analyses: analyses of variance for categorical variables and correlation analysis with DUI for quantitative ones.Mean ± SD of total sample are reported for quantitative variables.The percentages referred to the total sample are reported into brackets for categorical variables.Poly-comorbidity is defined as the concomitance presence of 2 or more medical or psychiatric disorders in comorbidity with borderline personality disorder.Abbreviations: DUI, duration of untreated illness; SD, standard deviation.Considering the factors associated with a longer DUI in patients affected by BPD, DUI was significantly correlated only with older age (r = .27,p < .01),earlier age at onset (r = À.17, p = .04)and longer duration of illness (r = .52,p < .01).Multivariate linear regression model with age, age at onset and duration of illness as independent variables and DUI as dependent one was reliable (Durbin-Watson statistic: 1.80), but it did not confirm the statistically significant correlations (age: p = .18;age at onset: p = .18;duration of illness: p = .48).