Review of Research Evidence Supporting Guideline Statements

On the basis of the limitations of the evidence for assessment of a patient with possible BPD, no grading of the body of research evidence is possible.

On the basis of the limitations of the evidence for use of quantitative measures, no grading of the body of research evidence is possible.

On the basis of the limitations of the evidence for evidence-based treatment planning, no grading of the body of research evidence is possible.

Two randomized controlled trials (RCTs; N =50 and N = 80), rated as having a moderate risk of bias, assessed the effectiveness of psychoeducation compared with a wait-list control over 12 weeks (Zanarini and Frankenburg 2008; Zanarini et al. 2018). Psychoeducation consisted of an internet-based program detailing the latest information on BPD in one study (Zanarini et al. 2018) and a single workshop in the other (Zanarini and Frankenburg 2008). Participants received psychoeducation in addition to treatment as usual (TAU). Participants in the control group were on a wait-list for psychoeducation and continued with TAU only.

All participants were female, and the majority were White. The mean age was 21 years (Zanarini et al. 2018) and 19 years (Zanarini and Frankenburg 2008). Only one study reported the severity of BPD at baseline (Zanarini et al. 2018). Participants were mildly ill at baseline with mean Zanarini Rating Scale for BPD (ZAN-BPD) scores ranging from 10.13 to 12.13 (Zanarini et al. 2018).

Detailed information on main study characteristics and treatment effects is presented in Appendix D. presents certainty-of-evidence ratings.

Both studies assessed the severity of BPD on the ZAN-BPD (Zanarini and Frankenburg 2008; Zanarini et al. 2018) and reported nonsignificant differences between the psychoeducation and the wait-list groups. In addition, one RCT reported similar treatment effects between groups on the Borderline Evaluation of Severity Over Time (BEST) scale (Zanarini et al. 2018). This RCT reported significantly better scores for the psychoeducation group after 12 months of follow-up (Zanarini et al. 2018). The investigators, however, tested 10 outcome measures and did not adjust for multiple comparisons.

The larger of the two RCTs (N = 80; Zanarini et al. 2018) employing internet-based psychoeducation reported no significant differences between intervention and wait-list groups for anxiety and depressive symptoms. Participants in the psychoeducation group, however, achieved significantly better scores on the Social Adjustment Scale than participants in the wait-list group. As mentioned earlier, however, this study tested 10 outcome measures and did not adjust for multiple testing.

One study (N = 80; Zanarini et al. 2018) reported similar effects and no significant differences on the Sheehan Disability Scale after 12 weeks and 12 months.

None of the studies reported on the incidence of adverse events, serious adverse events, or withdrawal due to adverse events (Zanarini and Frankenburg 2008; Zanarini et al. 2018).

One RCT (Bozzatello and Bellino 2020) evaluated the efficacy of interpersonal psychotherapy compared with wait-list plus clinical management. The study included 43 participants in Italy who were assessed at 10 months. This study was rated as having a moderate risk of bias. The trial was funded by the Italian government.

The majority of the study participants were female; race was not reported. The overall mean age of participants was 35 years of age. The study excluded patients receiving psychiatric services or who had existing schizophrenia, bipolar disorder, mental impairment, or drug or alcohol dependence.

The intervention group received 22 sessions in the first 20 weeks and 20 sessions in the last 20 weeks. Each session lasted 50 minutes. TAU consisted of case management provided by hospital and primary and community care services.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. presents certainty-of-evidence ratings.

One RCT (Morton et al. 2012) evaluated the efficacy of acceptance and commitment therapy (ACT) in addition to TAU compared with TAU alone. The Australian study included 41 participants who were followed for a duration of 13 weeks. The study was rated as having a moderate risk of bias because of high attrition. The trial did not report funding.

Almost all of the study participants were female. The mean age of the ACT group was 36 years, while the mean age of the TAU group was 34 years. The study excluded participants with psychotic symptoms (besides “reactive psychotic symptoms” associated with BPD [not specified further]), with intellectual disability, with cognitive impairment, or who were a significant risk to other participants.

ACT was delivered as weekly group sessions that included performing mindfulness exercises, doing emotions skills training, focusing on awareness of one’s values, and identifying choice points for action. TAU consisted of case management provided by public mental health services in Australia.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. presents certainty-of-evidence ratings.

One U.S. RCT (Weinberg et al. 2006) evaluated the efficacy of manual-assisted cognitive therapy (MACT), compared with TAU. Overall, the study provided data on 30 participants. The study was rated as having a moderate risk of bias. Follow-up duration was 6 months after treatment. The study was supported by a Young Investigator Award from the Borderline Personality Disorder Research Foundation. The majority of the study participants were female and White and had a mean age of 28 years. The study did not report on baseline severity. The study excluded participants with psychotic disorders, substance abuse disorder, or risk of suicide.

MACT was administered as an adjunctive intervention to TAU and comprised six sessions, over 6–8 weeks, incorporating elements of dialectical behavior therapy (DBT), cognitive-behavioral therapy (CBT), and bibliotherapy, modified to focus on deliberate self-harm. Each session was structured around a chapter of a booklet, covering functional analysis of episodes of parasuicide (defined as deliberate self-harm or suicide attempts), emotion regulation strategies, problem-solving strategies, management of negative thinking, management of substance use, and relapse prevention strategies. TAU consisted of standard care.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. presents certainty-of-evidence ratings.

The Borderline Personality Disorder Study of Cognitive Therapy (BOSCOT) RCT (Davidson et al. 2006) evaluated the efficacy of CBT in addition to TAU compared with TAU only. The study included 106 participants in the United Kingdom who were followed for a duration of 24 months. The study was rated as having a moderate risk of bias. The trial was funded by a public foundation.

The majority of the study participants were female, and all of them were White (Davidson et al. 2006). The overall mean age of participants was 32 years of age. The study excluded patients receiving psychiatric services or who had existing schizophrenia or bipolar disorder, mental impairment, or drug or alcohol dependence.

The intervention group received an average of 27 sessions of CBT over 12 months in addition to TAU (Davidson et al. 2006). Each session lasted 1 hour. TAU consisted of case management provided by hospital and primary and community care services.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. presents certainty-of-evidence ratings.

Six studies, four RCTs (Carter et al. 2010; Feigenbaum et al. 2012; McMain et al. 2017; Verheul et al. 2003), a nonrandomized trial (Bohus et al. 2004), and a retrospective cohort study (Gregory and Sachdeva 2016), evaluated the efficacy of DBT compared with TAU. Overall, these studies provided data on 483 participants. Three studies were rated as having a high risk of bias, two as moderate risk of bias, and one as low risk of bias. Reasons for ratings of high risk of bias were lack of intention-to-treat analysis and high attrition. Follow-up durations ranged from 3 months to 12 months. One trial was funded by a health insurance company; the other studies were publicly funded or did not report source of funding.

The majority of study participants were female, and mean ages ranged from 25 years to 35 years. Only one study, in which the majority of participants were White, reported on race or ethnicity. Likewise, only one study reported the severity of BPD at baseline (Gregory and Sachdeva 2016). In this retrospective cohort study, participants were moderately ill at baseline, with BEST scores of 45 to 49. Studies excluded patients with psychiatric comorbidities such as schizophrenia, major depressive disorder (MDD), alcohol or substance use disorder, and bipolar disorder.

DBT combines weekly individual psychotherapy sessions, weekly skills training groups, and weekly supervision and consultation meetings for the therapists. One study assessed brief DBT with skills training only over 20 weeks (McMain et al. 2017). All studies enrolled outpatients, except a study from Germany, which conducted DBT as an inpatient treatment (Bohus et al. 2004).

TAU consisted of a range of individualized service provisions and professional mental health care. All except one study (Gregory and Sachdeva 2016) employed a wait-list design in which participants of the TAU groups were offered DBT at the end of the study.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. presents certainty-of-evidence ratings.

One nonrandomized clinical trial (Barnicot and Crawford 2019), conducted in the United Kingdom and rated as having a high risk of bias, compared DBT with mentalization-based treatment (MBT) in 90 patients with BPD. The majority of participants were female (72%), with a mean age of 31 years. More than one-third (36%) were Black or belonged to a minority ethnic group. Mean baseline BPD severity ranged from 40.7 points to 44.8 points on the BEST scale. Reasons for the high risk of bias included selection bias and confounding.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. presents certainty-of-evidence ratings.

One Canadian RCT (McMain et al. 2012; described in three publications), rated as having a high risk of bias, compared DBT with well-specified general psychiatric management (GPM) in 180 patients with BPD. The majority of participants were female (86%), with a mean age of 30 years. Race and ethnicity were not reported. Mean baseline BPD severity ranged from 14.9 points to 15.5 points on the ZAN-BPD. Reasons for high risk of bias included high attrition (38%) at 12 months.

Treatment duration was 12 months, and the study was funded through the Canadian Institutes for Health Research (McMain et al. 2012). DBT included weekly individual therapy and group skills training, weekly telephone coaching with explicit focus on self-harm and suicidal behavior, and weekly therapist team consultation. Manualized GPM consisted of weekly individual therapy that was expanded away from focusing on self-harm and suicidal behaviors and included medication management. Generalized psychiatric therapy also included mandated therapist supervision weekly meetings.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–8 presents certainty-of-evidence ratings.

One nonrandomized clinical trial (Guillén Botella et al. 2021) conducted in Spain, rated as having a high risk of bias, compared DBT with Systems Training for Emotional Predictability and Problem-Solving (STEPPS) in 72 patients with BPD. The overwhelming majority of participants were female (94%), and all were White, with a mean age of 32 years. Mean baseline BPD severity ranged from 35.8 points to 38.6 points on the BSL-23. The study was rated as having a high risk of bias due to high attrition (32%).

Treatment duration was 6 months (Guillén Botella et al. 2021). DBT included weekly individual therapy and group skills training, telephone skills coaching, and team consultation. STEPPS included group therapy, a reinforcement team, telephone consultations with relatives, consultations with other professionals, and weekly clinician meetings. The study funding source was not reported.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–9 presents certainty-of-evidence ratings.

One three-armed retrospective cohort study (Gregory and Sachdeva 2016; reported in two publications), conducted in the United States and rated as having a high risk of bias, compared DBT with dynamic deconstructive psychotherapy (DDP) and TAU in 68 patients with BPD. The majority of participants were female (81%) and White (88%), with a mean age of 31 years. Mean baseline BPD severity ranged from 45.5 points to 49.2 points on the BEST scale. Reasons for the rating of high risk of bias included high attrition (53%) and confounding.

Treatment duration for both DBT (N = 25) and DDP (N = 27) was 12 months (Gregory and Sachdeva 2016). DBT included weekly individual therapy, weekly group sessions, and telephone skills coaching. DDP included weekly individual sessions that combined elements of translational neuroscience, object relations theory, and deconstructionist philosophy. The study was supported by the American Psychoanalytic Association.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–10 presents certainty-of-evidence ratings.

One three-armed RCT (Clarkin et al. 2007), rated as having a high risk of bias and conducted in the United States, compared DBT with transference-focused psychotherapy (TFP) and supportive therapy in 90 patients with BPD and reported results for patients for whom they had at least three data points (N = 62). The majority of participants were female (92%), White (68%), and with a mean age of 31 years. Mean baseline BPD severity was not reported. We rated the study as having high risk of bias due to the randomization process and high attrition (31%). Treatment duration was 12 months. DBT included weekly individual therapy, weekly group sessions, and telephone skills coaching. TFP included two individual weekly sessions focused primarily on the dominant affect-laden themes that emerge in the patient-therapist relationship. Supportive treatment included one weekly session supplemented with additional sessions as needed. The study was supported by the Borderline Personality Disorder Research Foundation.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–11, A3–12 and A3–13 present certainty-of-evidence ratings for the different comparisons.

DBT is a multifaceted cognitive-behavioral treatment approach that includes individual therapy, group skills training, telephone coaching, and a consultation team meeting for therapists. Three studies (one nonrandomized clinical trial [Andión et al. 2012], one RCT [Linehan et al. 2015], one prospective cohort study [Lyng et al. 2020]) assessed the comparative value of individual therapy components of DBT. Together, these studies provided data on 238 participants. One study (Andión et al. 2012) compared the individual therapy component of DBT with combined individual and group therapy. Another (Lyng et al. 2020) compared the stand-alone group skills component with 6 months of the full four-component DBT program. A third three-armed study (Linehan et al. 2015) compared 12 months of standard DBT (i.e., the full four-component program) with stand-alone group skills training and individual therapy with an activities group. All three studies were rated as having a high risk of bias. Reasons for ratings of high risk of bias included high overall attrition or high differential attrition, bias due to deviations from the intended intervention, and bias due to confounding (Andión et al. 2012; Linehan et al. 2015; Lyng et al. 2020).

The majority of participants were female, with a mean age across studies ranging from 26 years to 33 years. Race was reported in just one of three studies, in which more than 70% of participants were White (Linehan et al. 2015). Two studies were conducted in Europe (Andión et al. 2012; Lyng et al. 2020) and one in the United States (Linehan et al. 2015). Just one study provided baseline information on BPD severity, reporting a mean score on the BSL-23 of 2.7 points (Lyng et al. 2020). Treatment durations ranged from 6 months (Lyng et al. 2020) to 1 year (Andión et al. 2012; Linehan et al. 2015). One study followed patients through 18 months (6 months after the end of the intervention) (Andión et al. 2012), and another study followed patients through 2 years (12 months following the end of treatment) (Linehan et al. 2015). Studies were generally funded by public funds with no commercial funding.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–14, A3–15, and A3–16 present certainty-of-evidence ratings for different comparisons.

One RCT (N = 111) (Linehan et al. 2006), rated as having a high risk of bias, compared DBT with community therapy offered by nonbehavioral psychotherapy experts over 1 year. All participants were female, with a mean age of 29 years, who had at least two suicide attempts; the majority were White (87%). The severity of BPD at baseline was not reported.

We rated the study as having a high risk of bias because of lack of intention-to-treat analysis. The follow-up duration was 2 years, and the study was funded by the National Institute of Mental Health (Linehan et al. 2006).

The intervention group received standard DBT for 1 year, including weekly individual psychotherapy sessions, weekly group skills training, and telephone consultation as needed (Linehan et al. 2006). Community treatment by experts involved selected psychotherapists who were matched with therapists administering DBT by controlling for sex, availability, expertise, allegiance, training, and experience.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–17 presents certainty-of-evidence ratings.

A retrospective cohort study (Gregory and Sachdeva 2016) evaluated the efficacy of DDP compared with TAU. The study provided data on 44 participants. The study was rated as having a high risk of bias because of confounding and attrition. The follow-up duration was 12 months, and the study was funded by the American Psychoanalytic Association. The majority of the study participants were female and White, and the mean age was 28 years. This study reported baseline BEST scores ranging from 46 to 49. The study excluded patients with schizophrenia, intellectual disabilities, or dementia.

DDP involved weekly individual sessions over a 12-month period and combined elements of translational neuroscience, object relations theory, and deconstruction philosophy (Gregory and Sachdeva 2016). TAU consisted of unstructured psychotherapy.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–18 presents certainty-of-evidence ratings.

One RCT (Beck et al. 2020) evaluated the efficacy of MBT compared with TAU alone. This Danish study included 112 participants who were followed for a duration of 12 months. The study was rated as having a high risk of bias because of high attrition. The trial reported no commercial funding.

Almost all of the study participants were female, with the exception of one person (Beck et al. 2020). The mean age was 16 years. The study excluded participants with comorbid diagnosis of pervasive developmental disorder, learning disability, anorexia, current psychosis, schizophrenia or schizotypal personality disorder, antisocial personality disorder, any other mental disorder other than BPD considered the primary diagnosis, current (past 2 months) substance use disorder (SUD; but not substance abuse), and current psychiatric inpatient treatment.

MBT, delivered over 12 months, consisted of 3 introductory sessions, 37 weekly group sessions (90 minutes each), 5 individual case formulation sessions, and 6 sessions for caregivers (Beck et al. 2020). TAU consisted of at least 12 individual supportive sessions, one per month, comprising psychoeducation, counseling, and crisis management and sessions as needed.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–19presents certainty-of-evidence ratings.

Three RCTs, described in four articles, compared MBT with supportive therapy (Bateman and Fonagy 2009; Bateman et al. 2021; Carlyle et al. 2020; Jørgensen et al. 2013). Together, these studies provided data on 317 participants. Supportive therapy was not identical across the studies, but all included group sessions that focused on supportive techniques such as problem-solving. Two studies were rated as having a moderate risk of bias (Bateman and Fonagy 2009; Carlyle et al. 2020), and the other as a high risk of bias (Jørgensen et al. 2013). Reasons for ratings of high risk of bias included high attrition and deviations from the intended intervention.

The majority of participants were female, and the mean age across the three studies was 31 years. Race was reported in two studies, in which the majority of participants were White (Bateman and Fonagy 2009; Carlyle et al. 2020). Two studies were conducted in Europe (Bateman and Fonagy 2009; Jørgensen et al. 2013) and one in New Zealand (Carlyle et al. 2020). No study reported severity of BPD at baseline; however, one study reported global severity of symptoms at baseline that ranged from 1.7 points to 2.0 points on the Symptom Checklist–90 (SCL-90) Global Severity Index scale (Jørgensen et al. 2013). Treatment durations ranged from 18 months (Bateman and Fonagy 2009; Carlyle et al. 2020) to 24 months (Jørgensen et al. 2013). No study had commercial funding; one was funded through a foundation grant (Bateman and Fonagy 2009).

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–20 presents certainty-of-evidence ratings.

Two studies, one RCT (Laurenssen et al. 2018) that was rated as having a moderate risk of bias and one observational study with a nonconcurrent control group (Bales et al. 2015) that was rated as having a high risk of bias, compared day-hospital MBT with another specialized psychotherapy. Together, these studies provided data on 299 participants. Day-hospital MBT differed from typical MBT in terms of intensity; it involved daily group psychotherapy and weekly individual therapy along with art and writing therapy. The specialized psychotherapy comparator groups consisted of a variety of treatments, settings, and durations that were explicitly not limited to supportive therapy. Reasons for the high risk of bias rating included confounding and measurement of outcomes.

The majority of participants were female, and the mean age across the two studies was 32 years (Bales et al. 2015; Laurenssen et al. 2018). Race was not reported in either study, both of which were conducted in the Netherlands. One study reported BPD severity ranging from 32.8 points to 34.3 points at baseline using the BPD Severity Index (Laurenssen et al. 2018). Treatment duration was 18 months in both studies, with one study following patients through 36 months (Bales et al. 2015). Neither study had commercial funding.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–21 presents certainty-of-evidence ratings.

Two RCTs (Blum et al. 2008; Bos et al. 2010) and one prospective cohort study (González-González et al. 2021) evaluated the efficacy of STEPPS compared with TAU. The studies provided data on 362 participants. One RCT (Bos et al. 2010) was rated as having a moderate risk of bias because of differential attrition, and one RCT (Blum et al. 2008) and one cohort study (González-González et al. 2021) were rated as having a high risk of bias for high overall attrition. Additionally, the cohort study had risks of bias from selection and confounding. The timing of the initial follow-up ranged from 20 weeks to 24 weeks for the two RCTs. Both reported 1-year outcomes. For one RCT, the primary endpoint was at 20 weeks (Blum et al. 2008); for the other, the primary endpoint was at 1 year (Bos et al. 2010). For the cohort study (González-González et al. 2021), the primary endpoint was at 2 years. Both RCTs were funded; neither had pharmaceutical industry support. The cohort study did not have specific funding. The majority of the study participants were female, and the mean age was 32 years in the RCTs and 34 years in the cohort study. Only one of the studies reported ethnicity; 94% of participants were White (Blum et al. 2008). One RCT reported mean baseline BEST scores ranging from 39 to 40 (Blum et al. 2008). The cohort study reported mean baseline BEST scores ranging from 50 to 52. Studies excluded patients with psychotic or primary neurological disorders, who were cognitively impaired, or who had participated in STEPPS previously.

STEPPS involved 18 or 20 weekly therapy sessions; components included psychoeducation about BPD, emotion management skills training, and behavior management skills training. TAU consisted of usual care such as individual psychotherapy, medication, and case management.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–22 presents certainty-of-evidence ratings.

One RCT (Doering et al. 2010) conducted in Austria and Germany evaluated the efficacy of TFP compared with TAU. The study provided data on 104 participants. The study was rated as having a high risk of bias because of high differential attrition from follow-up. Follow-up duration was 12 months. The trial was funded by the Austrian National Bank.

All of the study participants were female and had a mean age of 28 years (Doering et al. 2010). The ethnicity of the participants was not reported. Authors noted that the study included participants with less severe BPD, with higher GAF scores, fewer comorbid Axis I and II disorders, and fewer self-harming acts than other treatment studies of BPD because patients with more severe symptoms would receive inpatient treatment in Austria and Germany. Studies excluded patients with schizophrenia; bipolar I and II disorder with a major depressive, manic, or hypomanic episode during the previous 6 months; SUD in the past 6 months; or organic pathology or intellectual disability.

TFP is a modified psychodynamic therapy and consists of two 50-minute sessions delivered every week by experienced clinical psychologists or medical doctors, along with medications as needed for 1 year of treatment (Doering et al. 2010). TAU consisted of individualized standard care from community psychiatrists.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–23 presents certainty-of-evidence ratings.

One RCT (described in two publications; Giesen-Bloo et al. 2006; Spinhoven et al. 2007), rated as having a high risk of bias and conducted in the Netherlands, compared TFP with schema-focused therapy (SFT) in 88 patients with BPD. The majority of participants were female (93%), with a mean age of 31 years. Race and ethnicity were not reported. Mean baseline BPD severity ranged from 33.5 points to 34.4 points on the BPD Severity Index. Reasons for a rating of high risk of bias included high attrition (39%) and measurement of outcomes.

Treatment duration was 3 years (Giesen-Bloo et al. 2006). Both TFP and SFT included two 50-minute sessions per week. The TFP focused on the patient-therapist relationship, while the SFT involved integrated cognitive therapy focused on four schema modes. The study was funded by a grant from the Dutch Health Care Insurance Board.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–24 presents certainty-of-evidence ratings.

Detailed information on main study characteristics and treatment effects is presented in Appendix D for nine studies that compared various psychotherapies within special populations. Overall, there is no evidence to support one psychotherapy over another for any of the special populations identified.

On the basis of the lack of data on harms in studies of psychotherapies in BPD, no grading of the body of research evidence is possible.

On the basis of the limitations of the evidence for assessment of patients with possible BPD, no grading of the body of research evidence is possible.

Nine double-blinded RCTs evaluated the efficacy of four SGAs (aripiprazole, olanzapine, quetiapine extended release [ER], ziprasidone) compared with placebo (Black et al. 2014; Bogenschutz and Nurnberg 2004; Linehan et al. 2008; Nickel et al. 2006; Pascual et al. 2008; Schulz et al. 2008; Soler et al. 2005; Zanarini and Frankenburg 2001; Zanarini et al. 2011b). Overall, these studies provided data on 1,124 participants. Two studies were rated as having a moderate (Black et al. 2014; Nickel et al. 2006) risk of bias and seven as having a high risk of bias (Bogenschutz and Nurnberg 2004; Linehan et al. 2008; Pascual et al. 2008; Schulz et al. 2008; Soler et al. 2005; Zanarini and Frankenburg 2001; Zanarini et al. 2011b). Reasons for ratings of high risk of bias were lack of intention-to-treat analysis and high attrition. Four trials employed fixed-dose designs assessing aripiprazole (15 mg/day; Nickel et al. 2006), olanzapine (2.5 mg/day or 5 mg/day; Linehan et al. 2008; Zanarini and Frankenburg 2001), and quetiapine ER (150 mg/day or 300 mg/day; Black et al. 2014); five trials used flexible-dose designs for olanzapine (2.5–20 mg/day; Bogenschutz and Nurnberg 2004; Schulz et al. 2008; Soler et al. 2005; Zanarini et al. 2011b) and ziprasidone (40–200 mg/day; Pascual et al. 2008). Follow-up durations ranged from 8 weeks to 6 months. All trials, except one (Nickel et al. 2006), were funded by the pharmaceutical industry.

The majority of trial participants were female and White; mean ages across studies ranged from 21 years to 34 years. Participants were moderately ill at baseline, with mean ZAN-BPD scores ranging from 14.6 to 17.7 and scores on the CGI scale modified for BPD from 4.3 to 4.8. Studies, in general, excluded patients with psychiatric comorbidities such as schizophrenia, MDD, alcohol or substance use disorder, or bipolar disorder.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–25 presents certainty-of-evidence ratings.

One industry-funded RCT (N = 45; Zanarini et al. 2004c), rated as having a moderate risk of bias, assessed differences in efficacy between olanzapine (2.5–7.5 mg/day), fluoxetine (10–30 mg/day), and a combination of fluoxetine and olanzapine. The study duration was 8 weeks. All trial participants were females between 18 years and 40 years of age; the majority were White. The severity of disease at baseline was not reported.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–26 presents certainty-of-evidence ratings.

One RCT (N = 51; Bozzatello et al. 2017) and one retrospective cohort study (N = 116; García-Carmona et al. 2021) compared SGAs with other SGAs.

The RCT, rated as having a high risk of bias, assessed differences in efficacy between asenapine (5–10 mg/day) and olanzapine (5–10 mg/day) (Bozzatello et al. 2017). The study duration was 12 weeks. All trial participants were between 18 years and 50 years of age; the majority were female (63%), with race being unreported.

The high risk of bias retrospective cohort study compared the effectiveness of oral SGAs (not specified) and long-acting injectable SGAs (aripiprazole, paliperidone, risperidone) (García-Carmona et al. 2021). The study used data from 116 outpatients in Spain with follow-up data from 1 month to 3 months.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–27 presents certainty-of-evidence ratings.

Nine double-blinded RCTs evaluated the efficacy of three anticonvulsant medications (divalproex sodium, lamotrigine, topiramate) compared with placebo (Crawford et al. 2018; Frankenburg and Zanarini 2002; Hollander et al. 2001; Loew et al. 2006; Moen et al. 2012; Nickel et al. 2004, 2005; Reich et al. 2009; Tritt et al. 2005). Overall, these studies provided data on 523 participants.

Two studies were rated as having a low risk of bias (Loew et al. 2006; Tritt et al. 2005), three as having a moderate risk of bias (Crawford et al. 2018; Nickel et al. 2004, 2005), and four as having a high risk of bias (Frankenburg and Zanarini 2002; Hollander et al. 2001; Moen et al. 2012; Reich et al. 2009). Reasons for ratings of high risk of bias were lack of intention-to-treat analysis and high attrition.

Four trials employed fixed-dosage designs assessing lamotrigine (200 mg/day) (Tritt et al. 2005) or topiramate (200 mg/day and 250 mg/day) (Loew et al. 2006; Nickel et al. 2004, 2005); five trials used flexible-dosage designs for divalproex sodium (Frankenburg and Zanarini 2002; Hollander et al. 2001; Moen et al. 2012) or lamotrigine (Crawford et al. 2018; Reich et al. 2009). Follow-up durations ranged from 8 weeks to 52 weeks. Four trials were funded by the pharmaceutical industry (Frankenburg and Zanarini 2002; Hollander et al. 2001; Moen et al. 2012; Reich et al. 2009); the others reported no funding or were supported by public institutions.

The majority of trial participants were female and White, and mean ages ranged from 25 years to 38 years. Participants were moderately ill at baseline, with mean scores on the ZAN-BPD ranging from 11.3 to 20.2. Studies, in general, excluded patients with psychiatric comorbidities such as schizophrenia, MDD, alcohol or substance use disorder, and bipolar disorder. An exception was the trial by Frankenburg and Zanarini (2002), which included participants with BPD and bipolar disorder.

Detailed information on main study characteristics and treatment effects is presented in Appendix D. A3–28 presents certainty-of-evidence ratings.

One industry-funded RCT (N = 25; Simpson et al. 2004), rated as having a high risk of bias, assessed differences in efficacy between fluoxetine (20–40 mg/day) and placebo. The study duration was 12 weeks. All trial participants were female; the majority were White. Participants in both treatment groups received individual DBT and were part of 2-hour weekly skills groups.

Detailed information on main study characteristics and treatment effects is presented in Appendix D.A3–29 presents certainty-of-evidence ratings.

One RCT (N = 9; Cailhol et al. 2014), rated as having a moderate risk of bias, assessed differences in efficacy between 10 sessions of repetitive transcranial magnetic stimulation (rTMS) and sham rTMS. The study duration was 3 months. The majority of trial participants were females between 20 years and 45 years of age, with race being unreported. The severity of disease at baseline was reported by the BPD Severity Index. The study was publicly funded. Detailed information on main study characteristics and treatment effects is presented in Appendix D.

On the basis of the limitations of the evidence for pharmacotherapy review in patients with possible BPD, no grading of the body of research evidence is possible.