Cluster B personality disorders (PDs) are severe and chronic mental health conditions characterized by relational and affective instability, identity disorder, and marked impulsivity (1). People with cluster B PDs present several difficulties in their relational and occupational functioning, decreasing their quality of life (1). A recent systematic review and meta-analysis estimated the lifetime prevalence of cluster B PDs at 5.5%, with differences between subtypes (0.8% for histrionic, 1.2% for narcissistic, 1.9% for borderline, and 3.1% for antisocial) (2). In Quebec, Canada, we estimated cluster B PDs lifetime prevalence at 2.6% (3). Individuals with cluster B PDs are high users of medical services, more than schizophrenia patients (4). They also have high rates of comorbid substance use disorders, particularly alcohol, opioid, and cocaine use disorders (5). Additionally, they exhibit a reduced life expectancy, and 20.5% of premature mortality is attributable to suicide (3, 6, 7). Indeed, borderline personality disorder may specifically indicate a predisposition to suicidal behavior disorder diagnoses (8).

Treatment of cluster B PDs is a clinical challenge because of the high suicidal behaviors and the resistance of clinical symptoms to pharmacological treatments (9, 10). Current clinical guidelines, such as the 2009 United Kingdom’s National Institute for Health Care and Excellence (NICE) recommendations (11) or the European guidelines for PDs (12), and a recent systematic review of clinical recommendations on the treatment of PD patients made by different mental health organizations worldwide (13) do not recommend the use of pharmacotherapy for the treatment of cluster B PD individuals because of the lack of evidence to support their use. Systematic reviews have been conducted in the last decade to identify which medications could benefit patients with cluster B PDs (10, 14–16). A 2010 Cochrane review of randomized clinical trials (RCTs) assessing the effectiveness and safety of different psychotropic medication classes found only fragmentary evidence supporting the use of antipsychotics (especially first generation) in reducing anger (14). More recent reviews (10, 16) and Cochrane updates (17, 18) did not find substantial differences from newer RCTs evaluating second-generation antipsychotics, antidepressants, mood stabilizers, or various medications (e.g., the antiepileptic lamotrigine or the anti-dementia drug memantine). Indeed, no medication has been officially approved for treating patients with cluster B PDs.

Although systematic reviews (10, 18, 19), expert opinions (9), and clinical practice guidelines (11, 12) recognize psychotherapy as the first-line treatment for cluster B PDs, there is evidence of a gap between evidence-based recommendations on pharmacotherapy and current clinical practice, with different psychotropic medications prescribed to treat this condition despite the lack of evidence for their efficacy. Indeed, some studies have reported high consumption of psychotropics in individuals with cluster B PDs (20–31). Nevertheless, most of these studies were cross-sectional (21, 22, 26, 28, 30), focused on borderline individuals (21–23, 25, 26, 28, 31), were conducted in small samples (26, 31), or considered only hospitalized patients (25, 26, 31), those participating in mental health programs (20, 22, 23, 29), or those followed by a psychiatrist (21).

The aim of this study was thus to draw a portrait of the use of psychotropic medications in individuals with cluster B PDs from a publicly managed care system enrolling all 8.5 million inhabitants of the Canadian province of Quebec. Specific objectives were (1) to describe psychotropic medication use in the year before and after cluster B PD diagnosis; (2) to identify trends and patterns in psychotropic use over 16 years; and (3) to identify factors associated with the exposure to psychotropic medication classes.

We identified 211,974 individuals aged 14 years and above with a first diagnosis of cluster B PD between the fiscal years 2002–2003 and 2018–2019. Among them, we excluded those not fully covered by the public drug plan between 1 year before and one year after the diagnosis. Thus, we gathered a group of 87,778 people, with 6.7% of them diagnosed during a hospital stay, while the remaining 93.3% diagnosed during an outpatient visit (45.7% by a general practitioner and 44.3% by a psychiatrist) and classified according to the fiscal year of their first diagnosis. The general practitioners are commonly available in the Emergency Rooms of hospitals where PD are more likely to visit than patients with other diseases, such as schizophrenia (4). The cohort was predominantly female and socially and materially deprived. Characteristics of the study population are reported in Table 1 according to the individual’s age group at the time of cluster B PD diagnosis.

To the best of our knowledge, this is the first study comparing medication use before and after cluster B PD diagnosis in a large cohort of newly diagnosed individuals, with trends and patterns analyzed over a period of 16 years. One of the most striking findings in our study was the extremely high proportion of individuals with cluster B PD receiving psychotropic medications both before and after the first recorded diagnosis, with almost four out of five patients being exposed to at least one psychotropic medication in the year after the diagnosis and a mean 2.8 different psychotropics used. Other studies reported similar high proportions (20–23, 25–31). Moreover, the formal diagnosis of cluster B PD did not seem to reduce the use of psychotropic medications but rather increased the use of some psychotropic medication classes. For those diagnosed with cluster B PD in 2018–2019, the proportion of users remained relatively stable or increased only slightly between 1 year before and after the formal diagnosis for anxiolytics, mood stabilizers, and ADHD medications, with a small peak around the date of the diagnosis for anxiolytics. For antidepressants and antipsychotics, on the contrary, this proportion increased just around the date of the diagnosis and remained higher the entire year afterwards. The peak in the proportion of users around the date of the diagnosis may indicate that the diagnosis of cluster B PD was made after some crisis that required psychotropic medications to control for manifested symptoms, which was a prelude to formal cluster B PD diagnosis.

Nevertheless, while the proportion of users decreased after this crisis period for anxiolytics, it remained stable for antidepressants and, especially, for antipsychotics. Other studies reported extensive antipsychotic utilization, especially atypical, in individuals with PDs (22, 26, 28, 30, 31, 36). While the efficacy of antidepressants seems limited in cluster B PD patients (9), that of antipsychotics, like quetiapine, has been suggested for managing anger, impulsivity and aggressivity (9). This, along with using quetiapine for insomnia, could explain the higher proportion of antipsychotic users after cluster B PD diagnosis (37). The increase in antidepressants and antipsychotic users concurrently with the diagnosis of cluster B PD was especially manifest in younger individuals. Still, it was not neglectable also in older individuals who already had high use. Older individuals were also exposed to anxiolytics in a higher proportion than other age groups, but this was not (or only slightly) affected by cluster B PD diagnosis. It is thus possible that anxiolytics have been used for a long time in older individuals as their use was more prevalent in the past years (38).

Different patterns of psychotropic medication classes use occurred in the last decades. Antidepressants remained the most used medication class, even if their use remained relatively stable over the study period, with only slight differences in the proportion of users depending on the year of diagnosis. Depression and anxiety, the main indications for antidepressant medications (39), were the most frequent comorbid conditions in our sample, with a prevalence of 50.9% for depression and 49.7% for anxiety, often concurrently in the same individual. Nevertheless, no recent clinical trial has been conducted to assess the efficacy of antidepressants in treating cluster B PD (9). Consequently, beyond treating an anxious-depressive comorbid condition, the use of antidepressants for cluster B PDs may have been less frequent or relatively stable.

Antipsychotics showed increased use from 2002 to 2008, remained stable until 2014, and decreased afterwards. In our population, the prevalence of psychotic disorders in the five years before the cluster B PD diagnosis was estimated at 37.5%. For those who were prescribed at least one antipsychotic within a year following their PD diagnosis, the prevalence of schizophrenia, other psychosis, or bipolar disorders was 46.20%. This suggests that more than half of the newly diagnosed PD patients lacked any recorded diagnosis for which an antipsychotic is indicated. Thus, there is a discrepancy between antipsychotic use and recorded diagnoses, which may suggest overuse of medications. However, some differences may result from the lack of diagnoses in the database. Moreover, cluster B PD diagnosis seems to act as a catalyst for antipsychotic prescriptions because of the gap between the proportion of users of these medications before and after the diagnosis. Nonetheless, this gap narrowed slightly in recent years. This pattern could be due to better compliance with clinical guidelines recommending psychotherapy, rather than pharmacotherapy, for cluster B PD treatment and the lack of evidence on antipsychotic treatment efficacy in these individuals (11, 12, 40).

On the other hand, the use of anxiolytics and mood stabilizers has diminished markedly over the study period. The decrease in the utilization of these classes may be related to the increased focus on providing a comprehensive continuum of care for people with PDs. This approach includes psychotherapy, rehabilitation, and all other necessary services across various lines of care (41). Nonetheless, no significant change in the care of patients with mental disorders has been implemented in recent years in Quebec. Moreover, in Quebec, public institutions do not offer coverage for psychotherapy services, leaving residents without personal or work insurance to pay out of pocket for these services. As a result, some individuals cannot access the resources they require, with only a minority of individuals with a mental health problem reporting having sought and received support services (42–44). This significant change could thus be driven by recent clinical guidelines not recommending anxiolytics, such as benzodiazepines, for current mental disorders, with a preference for newer antidepressants (e.g., selective serotonin reuptake inhibitors [SSRIs]) and the consequent change in clinical practice. Indeed, benzodiazepines, the most used anxiolytics, have been associated with injurious falls, fractures, delirium, and long-term cognitive decline, especially in older adults due to their sedative, cognitive-impairing, and motor-impairing effects (45, 46). Benzodiazepines have also been associated in Scandinavian countries’ health administrative databases with the worst health outcomes for schizophrenia, personality disorders, and some substance use disorder (47–50) The decline in the use of anxiolytics has also been reported in older individuals with schizophrenia and in the general aged population in two studies analyzing trends in more than a decade in Quebec (38, 51). Nevertheless, the decline in anxiolytic use we noticed was not compensated for by an equivalent increase in antidepressant use. This suggests that anxiolytics were used more in later years to treat symptomatic aspects of cluster B PDs rather than comorbid anxiety. As for mood stabilizers, their efficacy has not been proven, and their use should be limited to patients with comorbid bipolar disorders (11, 12, 40).

Medications for ADHD showed a constant increase in prescriptions over the study period, from about 1% in 2002–2003 to almost 12% in 2018–2019. This increase aligns with the rise in ADHD diagnosis over the last years in Quebec among children and adults, independently of the diagnosis of cluster B PDs (52, 53). As expected, their use was exceptionally high in younger individuals in recent years, with a proportion of users reaching more than 20% in 2018–2019 (results not shown). Beyond a better recognition of this disorder during the study period, the increased use of ADHD medications could also be due to the favorable risk/benefit profile these medications have shown in clinical practice in reducing ADHD symptoms like hyperactivity and impulsivity (54), with more recent studies also supporting their effectiveness in lowering unintentional injuries (55) and mortality (56), also in PD patients (49).

Trends and patterns in of psychotropic classes combinations first showed that they were more common than single classes. Even if the patterns of combinations changed over time, the proportion of cluster B PD individuals exposed to more than one psychotropic medication only slightly decreased over time (71% in 2002–2003 to 67% in 2018–2019 among users of at least one psychotropic medication). Because of the lack of solid evidence of the effectiveness of psychotropics in cluster B PD individuals, the use of combinations of psychotropic medications raises concerns as they may complexify the therapy of these patients and increase the risk for non-adherence, adverse effects and drug–drug interactions. These concerns are exceptionally elevated in older individuals as they are already at higher risk for multimorbidity and polypharmacy (57–60), which can lead to possible interactions with psychotropic medications, adverse effects and mortality.

Many factors have been associated with exposure to different psychotropic medication classes. Younger age was associated with the use of ADHD medications, which is in line with trends and patterns of diagnosis and treatment of ADHD described in recent studies (61). Age was also associated with exposure to other medication classes, such as antidepressants, antipsychotics, anxiolytics, and mood stabilizers, but in these cases, older individuals were more likely to use them. Generally, those with the higher risk were individuals in the 50–64 years groups, but it was possible to identify a trend for anxiolytics, with an increased risk with increasing age. In a recent study on patients with cluster B PDs, the authors found that older individuals were more likely to use many medications (more elevated than those <65 years) and significantly higher than their peers without cluster B PD diagnosis (62). A cluster B PD diagnosis seems to be a risk factor for receiving many medications, more so than for patients with other psychiatric diagnoses, such as affective disorders (62).

Rather than social deprivation, material deprivation was associated with a higher likelihood of exposure to antipsychotics, antidepressants, and anxiolytics. Our cohort of publicly insured patients was already composed of generally more materially deprived individuals than the general population because, in Quebec, public insurance is offered to those without private employ-related insurance (e.g., unemployed), along with those retired. Nevertheless, the association we found between material deprivation and exposure to psychotropic medications aligns with recent reports for antipsychotic medications in a large retrospective study in the United Kingdom (36). In that study, less deprived individuals with cluster B PDs were less likely to receive antipsychotics than more deprived ones, with adjusted risk ratios between 0.56 and 0.66 (36). More affluent patients may have access to private insurance for psychotherapy since private psychologists are not covered by the public health plan like in the United Kingdom (53).

This study has many strengths. First, it was the first to estimate psychotropic medication use in a large cohort of cluster B PDs individuals throughout the Quebec province. Additionally, the assessment of medication usage prior to and following the initial diagnosis of cluster B personality disorder had not previously been conducted. We thus could find that a not negligible proportion of cluster B PD individuals seemed exposed to psychotropic medications in the period near cluster B PD diagnosis and that this proportion increased after the formal diagnosis. Furthermore, we had access to data spanning nearly two decades, allowing us to analyze trends and patterns in psychotropic medication use. Finally, to identify all the individuals with a registered cluster B personality disorder diagnosis in the QICDSS databases, we used a case definition previously developed through a consensus-leading procedure by a team of experts psychiatrists and psychologists treating PD and using ICD-9 and ICD-10 codes related to DSM symptoms and relevant for clinical practice in Quebec (3).

Nonetheless, the results should be interpreted considering some limitations. Since we used administrative data and claims as a proxy of medication use, we could have overestimated the proportion of people under psychotropic medications. However, our results are consistent with past studies in different countries and settings, suggesting high medication use in cluster B PD individuals. Even in case of an overestimation of medication use, the trends over the 16 years of the study would not have been affected, and the temporal changes observed over time should be accurate. Moreover, the identification of patients with a diagnosis of cluster B PD in the database was based on a case definition using ICD-9 and ICD-10 codes. Since there are no specific ICD codes for cluster B PDs, the codes were chosen through consensus among clinicians. However, the case definition was not formally validated but reflected the consensus reached by a team of experts in the field who treat patients with PDs in Quebec hospitals and clinics. This means that some individuals with cluster A or C may have been misclassified as cluster B PDs due to our definition. Notably, the medico-administrative databases of Quebec, such as QICDSS, are recognized for their high specificity in identifying chronic conditions (33, 63). A form of external validation comes from the important excess mortality we observed in a previous study associated with our definition (3), which is in line with the findings in other countries with comparable linked health administrative databases (64). Moreover, it is worth noting that certain PD patients’ first diagnosis may not have been recorded in the QICDSS because they received it before 1996, the year of developing the QICDSS database, or because they were previously diagnosed in the private sector. To counteract this potential bias, we used a 5-year period to search for diagnosis before the first cohort entry in 2002. Finally, we excluded individuals with private insurance in the two years around cluster B PD diagnosis since the QICDSS database registers only medication use in those under the public drug plan. We cannot thus exclude that medication use would be different in those with private insurance as they are generally less materially deprived and could therefore have better access to psychotherapy in the private sector.

The results of this study indicate that individuals diagnosed with cluster B personality disorders are being prescribed psychotropic medication to a significant extent, especially in the months around the date of diagnosis, which may not be consistent with current best practice recommendations and guidelines. The high use of psychotropic medications may reflect attempts to manage comorbid suicidality and substance use disorders, frequent in cluster B PD patients (3, 5, 6, 65). According to a recent Scandinavian health administrative database study, the long-term benefit of antipsychotics, antidepressants, or benzodiazepines in PD patients was not supported, while the exposure to ADHD medications, clozapine, and bupropion showed a positive effect on these patients (49). A similar and challenging finding was found for patients with methamphetamine use disorder (48), suggesting that the impulsivity dimension present in both PD and substance use disorder (SUD) was regulated by ADHD medication. Medication regulating the craving\sensation seeking dimension of PD and SUD shall also be similarly explored (66, 67). These findings suggest potential challenges to existing guidelines and reflection on the behavioral dimensions of PD and SUD (67), brought by long-term real-life observation made possible by registers of health administrative databases.

We could highlight different trends and patterns in specific psychotropic medication classes over the study period, suggesting essential changes in clinical practice related to certain changes in the management of patients with PDs and more general changes due to better knowledge of psychotropics’ efficacy and safety profiles. Similarly, combinations of psychotropics have been used frequently, with differences in clinical practice during the last sixteen years. In light of these findings, it is vital to conduct further research on the impact of the use of psychotropic medications on health-related outcomes in different healthcare settings, to explore the combination of medications as well as craving\sensation seeking pharmacotherapies.

The datasets presented in this article are not readily available because no permission is granted to use the Quebec Integrated Chronic Diseases Surveillance System (QICDSS) data. This means we will not be able to share data. Requests to access the datasets should be directed to Louis.Rochette@inspq.qc.ca.

This study is part of the continuous chronic disease surveillance mandate granted to the National Public Health Institute of Quebec (Institut National de Santé Publique du Québec - INSPQ) by the provincial Ministry of Health and Social Services. Ethical approval and written informed consent to participate in this study were not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and the institutional requirements.

AL, CL, and LC conceived the study and developed the protocol. CS, ER, and VM substantially contributed to the development of the methodology. EL and LR performed the statistical analyses. CL and CM wrote the first draft of the manuscript. AL, CS, EL, ER, EV, LC, LR, MK, RB, PD, PV, SR, and VM revised it critically and substantially contributed to the submitted manuscript. All the authors approved the final version of the manuscript.