边缘性人格障碍患者的药物干预

Diagnostic criteria

Participants were diagnosed with BPD according to criteria from DSM III, DSM III R, DSM IV, DSM IV R or ICD‐10. BPD diagnoses were confirmed by one or more standardised means of assessment. Four trials used the Diagnostic Interview for Borderline personality disorder (DIB) ( Cowdry 1988 ; De la Fuente 1994 ; Soloff 1989 ; Soloff 1993 ), while four used the revised version (DIB‐R) ( Reich 2009 ; Zanarini 2001 ; Zanarini 2003 ; Zanarini 2004 ). Two of the included trials used the International Personality Disorder Examination scale (IPDE) ( Schmahl 2012a ; Schmahl 2012b ) while one used IPDE in addition to SCID‐I ( Crawford 2018 ). One trial used the Scales of independent behaviour (SIB) ( Goldberg 1986 ), and two trials used the Zanarini Rating Scale for Borderline Personality Disorder (ZAN‐BPD) ( Grant 2022 ; Kulkarni 2018 ). One trial used the DIPD‐IV scale ( Frankenburg 2002 ) while two used both DIPD‐IV and ZAN‐BPD ( Schulz 2007 ; Zanarini 2007 ). Three trials used an unspecified clinical interview ( Jariani 2010 ; Montgomery 1982a ; Montgomery 1982b ). One trial did not use a structured instrument but confirmed BPD diagnosis through consensus ( Amminger 2013 ). Four trials stated that BPD was diagnosed according to the formal DSM criteria but did not specify if a standardised assessment instrument was used to confirm the diagnosis ( Leone 1982 ; NCT00533117 ; Shafti 2010 ; Shafti 2014 ). Eleven of the trials used the Structured Clinical Interview for DSM‐IV axis II disorders (SCID II): AstraZeneca 2007 ; Bogenschutz 2004 ; Hallahan 2007 ; Hollander 2001 ; Linehan 2008 ; Loew 2006 ; Nickel 2004 ; Nickel 2005 ; Nickel 2006 ; Simpson 2004 ; Tritt 2005 . In addition to SCID‐II, nine trials used one or more additional means of assessment ( Bellino 2014 ; Bozzatello 2017 ; Markovitz 1995a ; Moen 2012 ; Pascual 2008 ; Rinne 2002 ; Salzman 1995 ; Soler 2005 ; Ziegenhorn 2009 ). One trial used SCID but did not specify which version ( Black 2014 ). An overview of assessment instruments used by the individual trials is found in Table 1 and Characteristics of included studies .

Participant exclusion criteria

Most trials had several exclusion criteria, with the most common one used in 38 trials, the exclusion of participants with current major depression, bipolar affective disorder or psychotic disorder. Thirty of the trials also excluded participants with alcohol or substance abuse or dependence. Other common exclusion criteria were organic illness, mental retardation, cognitive disorder or impairment, severe somatic illness, chronic medical conditions, and pregnancy or breastfeeding. For exclusion criteria in all trials, see Table 1 and Characteristics of included studies .

Duration of interventions

Trial duration lasted from four weeks ( Ziegenhorn 2009 ) to 52 weeks ( Crawford 2018 ; NCT00533117 ). Thirty‐five of the trials had a duration of three months or less, leaving only 11 trials with a duration between three months and one year.

Format of intervention

The majority of trial interventions tested one type of medication to alleviate borderline symptoms. Two trials had more than one treatment arm investigating the effect of various doses or treatment durations of the same medication ( Black 2014 ; Zanarini 2007 ). Two trials compared a medication in addition to treatment‐as‐usual with a placebo intervention in addition to treatment‐as‐usual ( Crawford 2018 ; Kulkarni 2018 ), and one trial intervention specifically combined Dialectic Behaviour Therapy (DBT) with the selective serotonin reuptake inhibitor fluoxetine ( NCT00533117 ). Ten trials were head‐to‐head trials comparing active medications to each other ( Cowdry 1988 ; Bellino 2014 ; Bozzatello 2017 ; Jariani 2010 : Leone 1982 ; Shafti 2010 ; Shafti 2014 ; Soloff 1989 ; Soloff 1993 ; Zanarini 2004 ); four of which were multiple‐arm trials, with an additional placebo comparator ( Cowdry 1988 ; Soloff 1989 ; Soloff 1993 ; Zanarini 2004 ).

Type of intervention

The trials included in this review investigated the following 27 medications:

1. Amitriptyline

2. Alprazolam

3. Aripiprazole

4. Asenapine

5. Brexpiprazole

6. Carbamazepine

7. Clonidine

8. Fluoxetine

9. Flupenthixol

10. Fluvoxamine

11. Haloperidol

12. Lamotrigine

13. Loxapine

14. Memantine hydrochloride

15. Mianserin

16. Naltrexone

17. Olanzapine

18. Omega‐3 fatty acids

19. Phenelzine

20. Quetiapine

21. Sertraline

22. Thiothixene

23. Topiramate

24. Tranylcypromine

25. Trifluoperazine

26. Valporate semi sodium

27. Ziprasidone

Concomitant medication

Eight trials did not allow concomitant medication ( Bellino 2014 ; Bozzatello 2017 ; Moen 2012 ; Nickel 2004 ; Nickel 2006 ; Salzman 1995 ; Shafti 2014 ; Soloff 1993 ). Another seven trials gave specific information that psychotropics were not allowed ( Frankenburg 2002 ; Loew 2006 , Nickel 2005 ; Rinne 2002 ; Shafti 2010 ; Tritt 2005 ; Zanarini 2001 ). Thirteen trials did not allow for psychotropic co‐medication with the exception of benzodiazepines, SSRIs or both ( Amminger 2013 ; AstraZeneca 2007 ; Black 2014 ; Kulkarni 2018 ; Leone 1982 ; NCT00533117 ; Pascual 2008 ; Reich 2009 ; Schulz 2007 ; Simpson 2004 ; Soler 2005 ; Soloff 1989 ; Ziegenhorn 2009 ). However, many of these had restrictions and limitations regarding dosage, stability and drug initiation of concomitant medication prior to and during the trial. One trial allowed for patients to continue with standard psychiatric care and have changes to their psychotropic medication as prescribed, which 53.1% of participants did ( Hallahan 2007 ). In another trial, 100% of participants used methadone, but there was no further mention of co‐medication ( Jariani 2010 ). One trial allowed antidepressants, mood stabilisers, and stimulants, but the use of any new psychotropic medication was an exclusion criterion ( Grant 2022 ). Lastly, one trial specifically allowed medication for stable, chronic medical conditions such as hypertension but gave no further mention of other admissible concomitant medications ( Bogenschutz 2004 ). Twelve trials did not give any information on concomitant medication ( Cowdry 1988 ; Goldberg 1986 ; Hollander 2001 ; Linehan 2008 ; Markovitz 1995a ; Montgomery 1982a ; Montgomery 1982b ; Schmahl 2012a ; Schmahl 2012b ; Zanarini 2003 , Zanarini 2004 ; Zanarini 2007 ). Further, two trials similarly gave no details on permissible medications in the full report; however, Crawford 2018 excluded patients that were already prescribed a mood stabiliser within the past four weeks. Additionally, both arms received usual care. Another had a psychotropic medication washout for at least 10 days, a 15‐day washout for antidepressants (TCAs and MAOIs), and no patient had taken neuroleptics two months prior to the trial ( De la Fuente 1994 ).

Concomitant treatment

Twelve trials did not allow for concomitant psychotherapeutic treatment or treatment‐as‐usual ( Black 2014 ; Bozzatello 2017 ; Frankenburg 2002 ; Hallahan 2007 ; Loew 2006 ; Nickel 2004 ; Nickel 2005 ; Nickel 2006 ; Reich 2009 ; Rinne 2002 ; Shafti 2010 ; Tritt 2005 ). Five trials offered or allowed for Dialectic Behavioural therapy (DBT) ( Linehan 2008 ; Moen 2012 ; NCT00533117 ; Simpson 2004 ; Soler 2005 ) and another eight trials allowed or provided unspecified psychotherapy, psychological and psychosocial interventions concomitant with trial intervention or supportive atheoretical psychotherapy: Amminger 2013 ; De la Fuente 1994 ; Kulkarni 2018 ; Montgomery 1982a ; Montgomery 1982b ; Pascual 2008 ; Schmahl 2012a ; Schmahl 2012b . One trial allowed psychotherapy if initiated a minimum of three months prior to randomisation ( Bogenschutz 2004 ). The remaining 20 trials, did not state whether any type of concomitant treatment was offered or allowed.

Control intervention

The control intervention was a placebo.

Active medication versus placebo

1. Antidepressants SSRIs, NaSSAs, TCAs and MAOIs versus placebo : tranylcypromine sulfate ( Cowdry 1988 ), fluoxetine ( Markovitz 1995a ; NCT00533117 ; Salzman 1995 ; Simpson 2004 ), mianserin ( Montgomery 1982b ), fluvoxamine ( Rinne 2002 ), amitriptyline ( Soloff 1989 ), phenelzine sulfate ( Soloff 1993 );

2. First‐generation antipsychotics versus placebo: trifluoperazine hydrochloride ( Cowdry 1988 ), thiothixene ( Goldberg 1986 ), loxapine ( Leone 1982 ), flupenthixole ( Montgomery 1982a ), haloperidol ( Soloff 1989 ; Soloff 1993 );

3. Second generation antipsychotics versus placebo : olanzapine ( Bogenschutz 2004 ; Linehan 2008 ; Schulz 2007 ; Soler 2005 ; Zanarini 2001 ; Zanarini 2007 ), aripiprazole ( Nickel 2006 ), ziprasidone ( Pascual 2008 ), quetiapine ( Black 2014 ), brexpiprazole ( Grant 2022 );

4. Anticonvulsants versus placebo: lamotrigine ( Crawford 2018 ; Reich 2009 ; Tritt 2005 ), topiramate ( Loew 2006 ; Nickel 2004 ; Nickel 2005 ), valproate semi sodium ( Frankenburg 2002 ; Hollander 2001 ; Moen 2012 ), carbamazepine ( Cowdry 1988 ; De la Fuente 1994 ), divalproex ( Moen 2012 );

5. Omega‐3 fatty acids versus placebo: ( Amminger 2013 ; Hallahan 2007 ; Zanarini 2003 );

6. Antidementia drug versus placebo: memantine hydrochloride ( Kulkarni 2018 );

7. Opiod antagonist versus placebo : naltrexone ( Schmahl 2012a ; Schmahl 2012b );

8. Antihypertensives/alpha‐2 adrenoceptor agonist versus placebo: clonidine ( Ziegenhorn 2009 );

9. Benzodiazepine versus placebo: alprazolam ( Cowdry 1988 ).

Active medication versus active comparator medication

1. First‐generation antipsychotic versus first‐generation antipsychotic: loxapine versus chlorpromazine ( Leone 1982 );

2. Second‐generation antipsychotic versus second‐generation antipsychotic: olanzapine versus aripiprazole ( Shafti 2014 ), and olanzapine versus asenapine ( Bozzatello 2017 );

3. Second‐generation antipsychotic versus first‐generation antipsychotic : olanzapine versus haloperidol ( Shafti 2010 );

4. First‐generation antipsychotic versus antidepressant: trifluoperazine hydrochloride versus tranylcypromine sulfate ( Cowdry 1988 ), haloperidol versus amitriptyline ( Soloff 1989 ), and haloperidol versus phenelzine sulfate ( Soloff 1993 );

5. Second‐generation antipsychotic versus antidepressant: olanzapine versus sertraline ( Jariani 2010 ), and olanzapine versus fluoxetine ( Zanarini 2004 );

6. Benzodiazepine versus mood stabiliser: alprazolam versus carbamazepine ( Cowdry 1988 );

7. Benzodiazepine versus first‐generation antipsychotic : alprazolam versus trifluoperazine hydrochloride ( Cowdry 1988 );

8. Benzodiazepine versus antidepressant: alprazolam versus tranylcypromine sulfate ( Cowdry 1988 );

9. Mood stabiliser versus first‐generation antipsychotic: carbamazepine versus trifluoperazine hydrochloride ( Cowdry 1988 );

10. Mood stabiliser versus antidepressant: carbamazepine versus tranylcypromine sulfate ( Cowdry 1988 ).

Active medication versus combination of medications

1. Second generation antipsychotic versus second generation antipsychotic plus antidepressant: olanzapine versus olanzapine plus fluoxetine ( Zanarini 2004 );

2. Antidepressant versus antidepressant plus second‐generation antipsychotic : fluoxetine versus fluoxetine plus olanzapine ( Zanarini 2004 );

3. Mood stabiliser versus mood stabiliser plus miscellaneous (omega‐3 fatty acid): valproic acid versus valproic acid plus eicosapentaenoic acid (EPA) plus docosahexaenoic acid (DHA) ( Bellino 2014 ).

Adverse effects

Adverse effects were registered in all but seven trials. Adverse effects were, in most cases, measured spontaneously, but some trials also used anthropometric and laboratory values as measurement. Five trials used the Abnormal Involuntary Movement Scale (AIMS) and the Simpson‐Angus Scale (SAS), and a few used the Side Effect Rating Scale for extrapyramidal side effects. See Appendix 5 for specifications on scales used in each trial.

1.1 BPD symptom severity

A total of 16 trials comparing medication with placebo reported data for BPD symptom severity.

Eight trials compared antipsychotics to placebo ( Black 2014 ; Cowdry 1988 ; Goldberg 1986 ; Grant 2022 ; Pascual 2008 ; Schulz 2007 ; Soloff 1993 ; Zanarini 2007 ). The evidence indicates little to no difference but is very uncertain about the effect of antipsychotics compared with placebo on BPD symptom severity at the end of treatment (SMD ‐0.18, 95% CI ‐0.45 to 0.08; P = 0.18; I ² = 70%; 8 trials, 951 participants; very low‐certainty evidence; Analysis 1.1 ). The TSA showed that the z‐curve ended in the futility area, which means that the anticipated effect can be rejected. See Figure 4 in Appendix 6 .

Figure 4

---

---

TSA borderline severity: antipsychotics

DARIS: Diversity‐adjusted required information size; MIREDIF: Minimal relevant difference; TSA: Trial Sequential Analysis

Two trials compared antidepressants to placebo post‐treatment ( Cowdry 1988 ; Soloff 1993 ). The evidence indicates little to no difference but is very uncertain about the effect of antidepressants compared with placebo regarding BPD symptom severity (SMD 0.27, 95% CI −0.65 to 1.18; P = 0.57, I ² = 73%; 2 trials, 87 participants; very low‐certainty evidence; Analysis 1.1 ).

Four trials compared mood stabilisers to placebo ( Cowdry 1988 ; Crawford 2018 ; Moen 2012 ; Reich 2009 ). The evidence indicates little to no difference but is very uncertain about the effect of mood stabilisers compared with placebo at end of treatment (SMD 0.07, 95% CI −0.43 to 0.57; P = 0.78, I ² = 55%; 4 trials, 256 participants; very low‐certainty evidence; Analysis 1.1 ).

Four trials compared miscellaneous pharmacological therapies to placebo: Cowdry 1988 compared alprazolam to placebo; Kulkarni 2018 compared memantin‐hydrochloride to placebo; Schmahl 2012b compared naltrexone to placebo; and Ziegenhorn 2009 compared clonidine to placebo. There was no evidence that any of these miscellaneous medications had an effect on BPD symptom severity at the end of treatment compared to placebo; Cowdry 1988 : MD −0.58, 95% CI −1.63 to 0.47; P = 0.28; 1 trial, 25 participants; Analysis 1.2 ; Kulkarni 2018 : MD 2.00, 95% CI −1.62 to 5.62; P = 0.28; 1 trial, 33 participants; Analysis 1.2 ; Schmahl 2012b : MD 0.10, 95% CI −0.29 to 0.49; P = 0.62; 1 trial, 32 participants; Analysis 1.2 ; and Ziegenhorn 2009 : MD −13.11, 95% CI −65.36 to 39.14; P = 0.62; 1 trial, 34 participants; Analysis 1.2 .

1.2 Self‐harm

Five trials comparing medications with placebo reported data for self‐harm.

Two trials compared antipsychotics to placebo ( Linehan 2008 ; Nickel 2006 ). The evidence is very uncertain about the effect of antipsychotics compared with placebo at end of treatment (RR 0.66, 95% CI 0.15 to 2.84; P = 0.57, I ² = 67%; 2 trials, 76 participants; very low‐certainty evidence; Analysis 1.4 ).

One trial compared an antidepressant to placebo ( Simpson 2004 ). It may have little to no effect but the evidence is very uncertain about the effect of the antidepressant compared with placebo at the end of treatment (MD 0.45, 95% CI −10.55 to 11.45; P = 0.94; 1 trial, 20 participants; very low‐certainty evidence; Analysis 1.3 ).

One trial compared a mood stabiliser to placebo ( Crawford 2018 ). The evidence is very uncertain about the effect of mood stabiliser lamotrigine compared with placebo at the end of treatment (RR 1.08, 95% CI 0.79 to 1.48; P = 0.64; 1 trial, 276 participants; very low‐certainty evidence; Analysis 1.4 ).

One trial compared omega‐3 fatty acids to placebo ( Hallahan 2007 ). There was no clear evidence of a difference between omega‐3 fatty acids and placebo regarding self‐harm at the end of treatment (RR 1.23, 95% CI 0.51 to 2.97; P = 0.65; 1 trial, 49 participants; Analysis 1.4 ).

1.3 Suicide‐related outcomes

Thirteen trials comparing medications with placebo reported data for suicide‐related outcomes.

Eight trials compared antipsychotics to placebo ( Bogenschutz 2004 ; Cowdry 1988 ; Grant 2022 ; Linehan 2008 ; Montgomery 1982a ; Pascual 2008 ; Soler 2005 ; Schulz 2007 ; Zanarini 2007 ). The evidence indicates little to no effect but is very uncertain about the effect of antipsychotics compared with placebo at the end of treatment either with continuous outcome data (SMD 0.05, 95% CI ‐0.18 to 0.29; P = 0.67; I ² = 55%; 7 trials, 854 participants; very low‐certainty evidence; Analysis 1.5 ) or with dichotomous outcome data (RR 0.73, 95% CI 0.31 to 1.73; P = 0.47, I ² = 62%; 2 trials, 61 participants; very low‐certainty evidence; Analysis 1.7 ).

Three trials compared antidepressants to placebo ( Cowdry 1988 ; Montgomery 1982b ; Simpson 2004 ). The evidence indicates little to no effect but is very uncertain about the effect of antidepressants at end of treatment compared with placebo either with continuous data (SMD −0.26, 95% CI −1.62 to 1.09; P = 0.70, I ² = 80%; 2 trials, 45 participants; very low‐certainty evidence; Analysis 1.5 ), or with dichotomous data (RR 1.00, 95% CI 0.71 to 1.41; P = 1.00; 1 trial, 58 participants; very low‐certainty evidence; Analysis 1.7 ).

Two trials compared mood stabilisers to placebo ( Cowdry 1988 ; Hollander 2001 ). The evidence indicates little to no effect but is very uncertain about the effect of mood stabilisers compared with placebo at the end of treatment (SMD ‐0.36, 95% CI ‐1.96 to 1.25; P = 0.66, I ² = 81%; 2 trials, 44 participants; very low‐certainty evidence; Analysis 1.5 ).

One trial compared omega‐3 fatty acids to placebo ( Hallahan 2007 ). There was evidence that omega‐3 fatty acids may reduce suicide‐related outcomes more than placebo at the end of treatment (RR 0.52, 95% CI 0.28 to 0.95; P = 0.03; 1 trial, 49 participants; Analysis 1.7 ).

One trial also compared the benzodiazepine alprazolam to placebo ( Cowdry 1988 ). There was no clear evidence of a difference between alprazolam and placebo on suicide‐related outcomes at end of treatment (MD 0.75, 95% CI ‐0.18 to 1.68; P = 0.11; 1 trial, 25 participants; Analysis 1.6 ).

1.4 Psychosocial functioning

Seventeen trials comparing medications to placebo reported data for psychosocial functioning. For consistency, negative values indicate favourable results for the active medication as for the remaining outcomes.

Seven trials compared antipsychotics to placebo ( Black 2014 ; Goldberg 1986 ; Schulz 2007 ; Soler 2005 ; Soloff 1989 ; Soloff 1993 ; Zanarini 2007 ). The evidence indicates little to no difference in outcome but is very uncertain about the effect of antipsychotics compared with placebo on psychosocial functioning at end of treatment (SMD −0.16, 95% CI −0.33 to 0.00; P = 0.05, I ² = 23%; 7 trials, 904 participants; very low‐certainty evidence; Analysis 1.8 ).

Four trials compared antidepressants to placebo ( Salzman 1995 ; Simpson 2004 ; Soloff 1989 ; Soloff 1993 ). The evidence indicates little to no difference between antidepressants and placebo regarding psychosocial functioning at end of treatment, but the evidence is very uncertain (SMD ‐0.25, 95% CI −0.57 to 0.06; P = 0.11, I ² = 0%; 4 trials, 161 participants; very low‐certainty evidence; Analysis 1.8 ).

Two trials compared mood stabilisers to placebo ( Crawford 2018 ; De la Fuente 1994 ) with little to no difference in psychosocial functioning at the end of treatment as compared to placebo (SMD −0.01, 95% CI −0.28 to 0.26; P = 0.94, I ² = 0%; 2 trials, 214 participants; very low‐certainty evidence; Analysis 1.8 ). Hollander 2001 obtained a similar result (RR 0.64, 95% CI 0.37 to 1.11; P = 0.11; 1 trial, 16 participants; very low‐certainty evidence; Analysis 1.10 ).

One trial compared omega‐3 fatty acids to placebo ( Amminger 2013 ). There was evidence of a difference between mood stabilisers and placebo regarding psychosocial function at the end of treatment favouring placebo (MD 19.90, 95% CI 7.11 to 32.69; P = 0.002; 1 trial, 15 participants; Analysis 1.9 ).

1.5 Anger

Twenty trials comparing medications with placebo reported data for anger at the end of treatment.

Ten trials compared antipsychotics to placebo ( Black 2014 ; Bogenschutz 2004 ; Cowdry 1988 ; Goldberg 1986 ; Nickel 2006 ; Pascual 2008 ; Schulz 2007 ; Soloff 1989 ; Soloff 1993 ; Zanarini 2007 ). There was evidence of a difference between treatments regarding anger at the end of treatment favouring antipsychotics (SMD −0.37, 95% CI −0.55 to −0.18; P = 0.0001, I ² = 48%; 10 trials, 1025 participants; Analysis 1.11 ). Inspection of the funnel plot suggested potential bias (small asymmetry; see Figure 5 ; Appendix 6 ), but we found no evidence of possible significant publication bias: Egger’s regression intercept (bias) 2.10 (two tailed, P = 0.069).

Figure 5

---

---

Funnel plot of comparison: 1 Medications compared with placebo, outcome: 1.11 Secondary: Anger at end of treatment (continuous outcomes, SMDs)

Six trials compared antidepressants to placebo ( Cowdry 1988 ; Rinne 2002 ; Salzman 1995 ; Soloff 1989 ; Soloff 1993 ). There was evidence of a difference between treatments regarding anger at the end of treatment favouring antidepressants (SMD −0.37, 95% CI −0.64 to −0.11; P = 0.006, I ² = 0%; 6 trials, 224 participants; Analysis 1.11 ).

Eight trials compared mood stabilisers to placebo ( Cowdry 1988 ; De la Fuente 1994 ; Frankenburg 2002 ; Hollander 2001 ; Loew 2006 ; Nickel 2004 ; Nickel 2005 ; Tritt 2005 ). Three of these trials reported extraordinarily large SMDs ( Tritt 2005 : SMD −1.69, 95% CI −2.62 to −0.75; Loew 2006 : SMD −3.10, 95% CI −3.89 to −2.30; Nickel 2004 : SMD −2.80, 95% CI −3.89 to −1.71). We decided to exclude the outliers one by one from the primary analyses, until low heterogeneity was reached (see Analysis 26.1 for sensitivity analysis). After excluding outliers, there was evidence of a difference between treatments regarding anger at the end of treatment favouring mood stabilisers (SMD −0.67, 95% CI −1.10 to −0.24; P = 0.002, I ² = 26%; 5 trials, 135 participants; Analysis 1.11 ).

Two trials compared omega‐3 fatty acids to placebo ( Hallahan 2007 ; Zanarini 2003 ). There was a difference between treatments regarding anger at the end of treatment favouring omega‐3 fatty acids (SMD −0.48, 95% CI −0.95 to −0.01; P = 0.04, I ² = 0%; 2 trials, 76 participants; Analysis 1.11 ).

One cross‐over trial compared naltrexone to placebo ( Schmahl 2012b ). There was no evidence of a difference between treatments regarding anger at the end of treatment (MD 1.65, 95% CI −4.54 to 7.84; P = 0.60; 1 trial, 32 participants; Analysis 1.12 ).

Another cross‐over trial compared alprazolam to placebo ( Cowdry 1988 ). There was no evidence of a difference between treatments regarding anger at the end of treatment (MD −0.57, 95% CI −1.48 to 0.34; P = 0.22; 1 trial, 25 participants; Analysis 1.12 ).

1.6 Affective instability

Seven trials comparing pharmacological treatments to placebo reported data for affective instability.

Four trials compared antipsychotics to placebo ( Bogenschutz 2004 ; Pascual 2008 ; Schulz 2007 ; Zanarini 2007 ). There was a small difference between treatments regarding affective instability at the end of treatment favouring antipsychotics (SMD ‐0.16, 95% CI ‐0.31 to ‐0.01; P = 0.04; I ² = 0%; 4 trials, 691 participants; Analysis 1.13 ).

One trial ( Rinne 2002 ) compared an antidepressant to placebo. There was a small difference between treatments regarding affective instability at the end of treatment favouring the antidepressant (MD ‐1.66, 95% CI ‐3.26 to ‐0.06; P = 0.04; 1 trial, 38 participants; Analysis 1.14 ).

Two trials ( Reich 2009 and Crawford 2018 ) compared mood stabilisers to placebo. There was no difference between treatments regarding affective instability at the end of treatment (SMD −0.21, 95% CI −0.68 to 0.26; P = 0.38, I ² = 39%; 2 trials, 222 participants; Analysis 1.13 ).

1.7 Chronic feelings of emptiness

Four trials comparing pharmacological treatments to placebo reported data for chronic feelings of emptiness. All four trials compared antipsychotics to placebo ( Bogenschutz 2004 ; Pascual 2008 ; Schulz 2007 ; Zanarini 2007 ). There was no difference between treatments regarding chronic feelings of emptiness at the end of treatment (SMD −0.00, 95% CI −0.16 to 0.15; P = 0.96, I ² = 6%; 4 trials, 691 participants; Analysis 1.15 ).

1.8 Impulsivity

Sixteen trials comparing pharmacological treatments to placebo reported data for impulsivity.

Ten trials compared antipsychotics to placebo ( Black 2014 ; Bogenschutz 2004 ; Cowdry 1988 ; Grant 2022 ; Pascual 2008 ; Schulz 2007 ; Soler 2005 ; Soloff 1989 ; Soloff 1993 ; Zanarini 2007 ). There was no difference between treatments regarding impulsivity at end of treatment (SMD ‐0.08, 95% CI ‐0.20 to 0.04; P = 0.21; I ² = 0%: 10 trials, 1038 participants; Analysis 1.16 ). Inspection of the funnel plot suggested potential bias (small asymmetry; see Figure 6 ; Appendix 6 ), but we found no evidence of possible significant publication bias: Egger’s regression intercept (bias) 2.10 (two tailed, P = 0.069).

Figure 6

---

---

Funnel plot of comparison: 1 Medications compared with placebo, outcome: 1.16 Secondary: Impulsivity at end of treatment (continuous outcomes, SMDs)

Four trials compared antidepressants to placebo ( Cowdry 1988 ; Rinne 2002 ; Soloff 1989 ; Soloff 1993 ). There was no clear difference between treatments regarding impulsivity at end of treatment (SMD ‐0.17, 95% CI ‐0.49 to 0.15; P = 0.29; I ² = 13%; 4 trials, 182 participants; Analysis 1.16 ).

Five trials compared mood stabilisers to placebo ( Cowdry 1988 ; Crawford 2018 ; De la Fuente 1994 ; Moen 2012 ; Reich 2009 ). In the analysis using dichotomous data, there was no difference between treatments regarding impulsivity at the end of treatment (RR 0.88, 95% CI 0.53 to 1.46; P = 0.61; 1 trial, 20 participants; Analysis 1.18 ). In the analysis using continuous data, there was little to no difference between treatments regarding impulsivity at the end of treatment favouring mood stabilisers (SMD ‐0.56, 95% CI ‐1.46 to 0.35; P = 0.23; I ² = 84%; 4 trials, 265 participants; Analysis 1.16 ).

One trial also compared the benzodiazepine alprazolam to placebo ( Cowdry 1988 ). There was no difference between treatments regarding impulsivity at the end of treatment (MD 0.67, 95% CI −0.36 to 1.70; P = 0.20; 1 trial, 25 participants; Analysis 1.17 ).

1.9 Interpersonal problems

Fourteen trials comparing pharmacological treatments to placebo reported data for interpersonal problems.

Eight trials compared antipsychotics to placebo ( Bogenschutz 2004 ; Goldberg 1986 ; Nickel 2006 ; Pascual 2008 ; Schulz 2007 ; Soloff 1989 ; Soloff 1993 ; Zanarini 2007 ). Low‐certainty evidence suggests that antipsychotics may slightly reduce interpersonal problems (SMD −0.21, 95% CI −0.34 to −0.08; P = 0.002, I ² = 0%; 8 trials, 907 participants; low‐certainty evidence; Analysis 1.19 ). The TSA analysis showed that the DARIS was reached (n = 386), and that there was no risk of type 1 error (TSA adjusted CI −0.60 to 0.08). See Figure 7 ; Appendix 6 .

Figure 7

---

---

TSA interpersonal problems: antipsychotics

DARIS: Diversity‐adjusted required information size; MIREDIF: Minimal relevant difference; TSA: Trial Sequential Analysis

Two trials compared antidepressants to placebo ( Soloff 1989 ; Soloff 1993 ). There was little to no difference between treatments regarding interpersonal problems at the end of treatment (SMD −0.07, 95% CI −0.69 to 0.55; P = 0.82, I ² = 66%; 2 trials, 119 participants; low‐certainty evidence; Analysis 1.19 ).

Four trials compared mood stabilisers with placebo ( Crawford 2018 ; De la Fuente 1994 ; Frankenburg 2002 ; Loew 2006 ). This low‐certainty evidence suggested that mood stabilisers may result in a reduction of interpersonal problems, compared with placebo at end of treatment on interpersonal (SMD −0.58, 95% CI −1.14 to −0.02; P = 0.04, I ² = 73%; 4 trials, 300 participants; low‐certainty evidence; Analysis 1.19 ).

1.10 Abandonment

Four trials comparing pharmacological treatments to placebo reported continuous data for fear of abandonment. All four trials compared antipsychotics to placebo ( Bogenschutz 2004 ; Pascual 2008 ; Schulz 2007 ; Zanarini 2007 ). There was no difference between treatments regarding fear of abandonment at the end of treatment (SMD −0.01, 95% CI −0.17 to 0.14; P = 0.88, I ² = 5%; 4 trials, 691 participants; Analysis 1.20 ).

1.11 Identity disturbance

Four trials comparing pharmacological treatments to placebo reported continuous data for identity disturbance. All four trials compared antipsychotics to placebo ( Bogenschutz 2004 ; Pascual 2008 ; Schulz 2007 ; Zanarini 2007 ). There was no difference between treatments regarding identity disturbance at the end of treatment (SMD −0.09, 95% CI −0.25 to 0.07; P = 0.28, I ² = 7%; 4 trials, 691 participants; Analysis 1.21 ).

1.12 Dissociation and psychotic‐like symptoms

Thirteen trials comparing pharmacological treatments to placebo reported data for dissociation and psychotic‐like symptoms.

Eight trials compared antipsychotics to placebo ( Bogenschutz 2004 ; Goldberg 1986 ; Nickel 2006 ; Pascual 2008 ; Schulz 2007 ; Soloff 1989 ; Soloff 1993 ; Zanarini 2007 ). There was a small difference between treatments regarding dissociation and psychotic‐like symptoms at the end of treatment favouring antipsychotics (SMD −0.28, 95% CI −0.50 to −0.06; P = 0.01, I ² = 55%; 8 trials, 907 participants; Analysis 1.22 ).

Three trials compared antidepressants to placebo ( Simpson 2004 ; Soloff 1989 ; Soloff 1993 ). There was no difference between treatments regarding dissociation and psychotic‐like symptoms at the end of treatment (SMD −0.22, 95% CI −0.62 to 0.18; P = 0.29, I ² = 25%; 3 trials, 139 participants; Analysis 1.22 ).

Three trials compared mood stabilisers to placebo ( Crawford 2018 ; De la Fuente 1994 ; Loew 2006 ). There was no difference between treatments regarding dissociation and psychotic‐like symptoms at the end of treatment (SMD −0.23, 95% CI −0.66 to 0.20; P = 0.30; I ² = 51%; 3 trials, 270 participants; Analysis 1.22 ).

One trial compared omega‐3 fatty acids to placebo ( Amminger 2013 ). There was no difference between treatments regarding dissociation and psychotic‐like symptoms at the end of treatment (MD −2.80, 95% CI −5.70 to 0.10; P = 0.09; 1 trial, 15 participants; Analysis 1.23 ).

1.13 Depression

Twenty‐six trials comparing pharmacological treatments to placebo reported data for depression.

Twelve trials compared antipsychotics to placebo ( Black 2014 ; Cowdry 1988 ; Goldberg 1986 ; Grant 2022 ; Linehan 2008 ; Nickel 2006 ; Pascual 2008 ; Schulz 2007 ; Soler 2005 ; Soloff 1989 ; Soloff 1993 ; Zanarini 2007 ). There was a difference between treatments regarding depression at end of treatment favouring antipsychotics (SMD ‐0.22, 95% CI ‐0.42 to ‐0.01; P = 0.04; I ² = 59; 12 trials, 1138 participants; Analysis 1.24 ).

Five trials compared antidepressants to placebo ( Cowdry 1988 ; Salzman 1995 ; Simpson 2004 ; Soloff 1989 ; Soloff 1993 ). There was no clear difference between treatments regarding depression at end of treatment (SMD −0.37, 95% CI −0.82 to 0.08; P = 0.11, I ² = 52%; 5 trials, 187 participants; Analysis 1.24 ).

Six trials compared mood stabilisers to placebo ( Cowdry 1988 ; Crawford 2018 ; De la Fuente 1994 ; Frankenburg 2002 ; Hollander 2001 ; Loew 2006 ). There was a difference between treatments regarding depression at the end of treatment favouring mood stabilisers (SMD −0.44, 95% CI −0.80 to −0.08; P = 0.02, I ² = 46%; 6 trials, 344 participants: Analysis 1.24 ).

Six trials compared miscellaneous pharmacological treatments to placebo.

Three of these trials used a parallel design and compared omega‐3 fatty acids to placebo ( Amminger 2013 ; Hallahan 2007 ; Zanarini 2003 ). The analysis with dichotomous data showed that there was a difference between omega‐3 fatty acids and placebo regarding depression at the end of treatment favouring omega‐3 fatty acids (RR 0.48, 95% CI 0.28 to 0.81; P = 0.006; 1 trial, 49 participants; Analysis 1.26 ), while in the analysis with continuous data there was no difference between treatments regarding depression at end of treatment (SMD −0.54, 95% CI −1.18 to 0.11; P = 0.10, I ² = 0%; 2 trials, 42 participants; Analysis 1.24 ).

The other three trials used cross‐over designs. The first trial compared clonidine to placebo ( Ziegenhorn 2009 ), and found no difference between clonidine and placebo regarding depression at the end of treatment (MD −2.54, 95% CI −10.27 to 5.19; P = 0.52; 1 trial, 34 participants; Analysis 1.25 ). The second trial compared naltrexone to placebo ( Schmahl 2012b ), and found no difference between naltrexone and placebo regarding depression at the end of treatment (MD 2.50, 95% CI −4.22 to 9.22; P = 0.47; 1 trial, 32 participants; Analysis 1.25 ). The third trial compared the benzodiazepine alprazolam to placebo ( Cowdry 1988 ), and found no difference between alprazolam and placebo on depression at the end of treatment (MD 0.27, 95% CI −0.73 to 1.27; P = 0.60; 1 trial, 25 participants; Analysis 1.25 ).

1.14 Attrition

Thirty‐two trials comparing medications to placebo reported data on attrition.

Thirteen trials compared antipsychotics to placebo ( Black 2014 : Bogenschutz 2004 ; Goldberg 1986 ; Grant 2022 ; Linehan 2008 ; Montgomery 1982a ; Pascual 2008 ; Schulz 2007 ; Soler 2005 ; Soloff 1989 ; Soloff 1993 ; Zanarini 2001 ; Zanarini 2007 ). There was no clear difference between treatments regarding attrition at the end of treatment (RR 1.11, 95% CI 0.89 to 1.38; P = 0.34; I ² = 35%; 13 trials, 1216 participants; low‐certainty evidence ; Analysis 1.27 ). However, the TSA showed that the DARIS was not reached (n = 2008), and that there was a potential risk of type 1 error (TSA‐adjusted CI 0.74 to 2.13). See Figure 8 in Appendix 6 . Inspection of the funnel plot suggested potential bias (small asymmetry; see in Figure 9 in Appendix 6 ), but we found no evidence of possible significant publication bias: Egger’s regression intercept (bias) 1.89 (two tailed, P = 0.089).

Figure 8

---

---

TSA attrition: antipsychotics

DARIS: Diversity‐adjusted required information size; Pc: proportion with an outcome in the control group; RRR: Relative risk ratio; TSA: Trial Sequential Analysis

Figure 9

---

---

Funnel plot of comparison: 1 Pharmacotherapies compared with placebo, outcome: 1.19 Secondary: Attrition at end of treatment

RR: Relative Risk; SE(log[RR]): Standard Error of the logarithmic Risk Ratio

Six trials compared antidepressants to placebo ( Montgomery 1982b ; NCT00533117 ; Rinne 2002 ; Simpson 2004 ; Soloff 1989 ; Soloff 1993 ). There was no difference between treatments regarding attrition at the end of treatment (RR 1.07, 95% CI 0.65 to 1.76; P = 0.78, I ² = 0%; 6 trials, 289 participants; low‐certainty evidence; Analysis 1.27 ). The TSA, however, showed that the DARIS was not reached (n = 1816), and that there was a potential risk of type 1 error (TSA‐adjusted CI 0.07 to 14.98). See Figure 10 in Appendix 6 .

Figure 10

---

---

TSA attrition: antidepressants

DARIS: Diversity‐adjusted required information size; Pc: proportion with an outcome in the control group; RRR: Relative risk ratio; TSA: Trial Sequential Analysis

Nine trials compared mood stabilisers to placebo ( Crawford 2018 ; De la Fuente 1994 ; Frankenburg 2002 ; Hollander 2001 ; Loew 2006 ; Nickel 2004 ; Nickel 2005 ; Reich 2009 ; Tritt 2005 ). The evidence is very uncertain about the effect of mood stabilisers compared with placebo on attrition at the end of treatment (RR 0.89, 95% CI 0.69 to 1.15; P = 0.37, I ² = 0%; 9 trials, 530 participants; very low‐certainty evidence; Analysis 1.27 ). The TSA showed that the DARIS was not reached (n = 1300), and that there was a potential risk of type 1 error (TSA‐adjusted CI 0.37 to 2.23). See Figure 11 in Appendix 6 .

Figure 11

---

---

TSA attrition: mood stabilisers

DARIS: Diversity‐adjusted required information size; Pc: proportion with an outcome in the control group; RRR: Relative risk ratio; TSA: Trial Sequential Analysis

Four trials compared miscellaneous medications to placebo. Of these, two trials compared omega‐3 fatty acids to placebo ( Hallahan 2007 ; Zanarini 2003 ). There was no difference between omega‐3 fatty acids and placebo regarding omega attrition at end of treatment (RR 0.61, 95% CI 0.21 to 1.79; P = 0.37, I ² = 0%; 2 trials, 79 participants; Analysis 1.27 ). One trial compared memantine hydrochloride to placebo ( Kulkarni 2018 ). There was no difference between memantine hydrochloride and placebo regarding attrition at end of treatment (RR 1.57, 95% CI 0.45 to 5.52; P = 0.48; 1 trial, 33 participants; Analysis 1.27 ). Another trial compared clonidine to placebo ( Ziegenhorn 2009 ). There was no difference between clonidine and placebo regarding attrition at end of treatment (RR 0.67, 95% CI 0.13 to 3.50; P = 0.63; 1 trial, 34 participants; Analysis 1.27 ).

1.15 Adverse events

Seven trials comparing medications to placebo reported dichotomous data on total non‐serious adverse events at end of treatment. Five trials compared antipsychotics to placebo ( Black 2014 ; Grant 2022 ; Pascual 2008 ; Schulz 2007 ; Zanarini 2007 ). There was no clear difference in the presence of adverse events between the two groups receiving antipsychotics and placebo (RR 1.07, 95% CI 0.90 to 1.29; P = 0.43; I ² = 57%; 5 trials, 814 participants; very low‐certainty evidence; Analysis 1.28 ). The TSA showed that the z‐curve ended in the fidelity area. See Figure 12 in Appendix 6 .

Figure 12

---

---

TSA non‐serious adverse events: antipsychotics

DARIS: Diversity‐adjusted required information size; Pc: proportion with an outcome in the control group; RRR: Relative risk ratio; TSA: Trial Sequential Analysis

One trial compared a mood stabiliser to placebo ( Crawford 2018 ). There was no difference in the presence of adverse events between the two groups (RR 0.84, 95% CI 0.70 to 1.01; P = 0.07; 1 trial, 276 participants; very low‐certainty evidence; Analysis 1.28 ).

Another trial compared memantine hydrochloride to placebo ( Kulkarni 2018 ). There was no difference in any adverse events between the two groups (RR 1.41, 95% CI 0.79 to 2.52; P = 0.24; 1 trial, 33 participants; Analysis 1.28 ).

Two trials compared pharmacological treatment to placebo and reported dichotomous data on total serious adverse events. Kulkarni 2018 compared memantine hydrochloride to placebo and reported no serious adverse events in either the experimental or the placebo group ( Analysis 1.29 ). Grant 2022 compared brexpiprazole to placebo. One serious adverse event (mild suicidal ideation without a plan, considered serious as it necessitated hospitalisation) occurred in the placebo group. There was no statistical difference in the occurrence of serious adverse events between the two groups (RR 3.00, 95% CI 0.13 to 71.51; 1 trial, 80 participants; Analysis 1.29 ). No other trials specifically reported data on serious adverse events.

Please refer to Table 2 for an overview of adverse events.

Central nervous system

Headache

Six trials comparing pharmacological treatment to placebo reported data on headache as a non‐serious adverse event.

Four trials compared antipsychotics to placebo ( Black 2014 ; Grant 2022 ; Schulz 2007 ; Zanarini 2007 ). There was no difference in the presence of headache between the two groups (RR 1.01, 95% CI 0.63 to 1.62; P = 0.97; I ² = 32%; 4 trials, 754 participants; Analysis 2.1 ).

One trial compared a mood stabiliser to placebo ( Loew 2006 ). There was no difference in the presence of headache between the two groups (RR 1.00, 95% CI 0.15 to 6.61; P = 1.00; 1 trial, 56 participants; Analysis 2.2 ).

One trial compared memantine hydrochloride to placebo ( Kulkarni 2018 ). There was no difference in the presence of headache between the two groups (RR 1.29, 95% CI 0.71 to 2.36; P = 0.40; 1 trial, 33 participants; Analysis 2.3 )

Dizziness

Four trials comparing pharmacological treatment to placebo reported data on dizziness as a non‐serious adverse event.

Two trials compared antipsychotics to placebo ( Black 2014 ; Pascual 2008 ). There was no difference in the presence of dizziness between the two groups (RR 3.07, 95% CI 0.40 to 23.45; P = 0.28, I ² = 9%; 2 trials, 68 participants; Analysis 2.1 ).

One trial compared a mood stabiliser to placebo ( Loew 2006 ). There was no difference in the presence of dizziness between the two groups (RR 1.50, 95% CI 0.27 to 8.30; P = 0.64; 1 trial, 56 participants; Analysis 2.2 ).

One trial compared memantine hydrochloride to placebo ( Kulkarni 2018 ). There was no difference in the presence of dizziness between the two groups (RR 1.69, 95% CI 0.72 to 3.98; P = 0.23; 1 trial, 33 participants; Analysis 2.3 ).

Fatigue

Five trials comparing pharmacological treatment to placebo reported data on fatigue as a non‐serious adverse event.

Three trials compared antipsychotics to placebo ( Grant 2022 ; Schulz 2007 ; Zanarini 2007 ). There was no difference in the presence of fatigue between the two groups (RR 1.50, 95% CI 0.58 to 3.89; P = 0.40; I ² = 56%; 3 trials, 692 participants; Analysis 2.1 ).

One trial compared a mood stabiliser to placebo ( Loew 2006 ). There was no difference in the presence of fatigue between the two groups (RR 2.00, 95% CI 0.40 to 10.05; P = 0.40; 1 trial, 56 participants; Analysis 2.2 ).

One trial compared memantine hydrochloride to placebo ( Kulkarni 2018 ). There was no difference in the presence of fatigue between the two groups (RR 1.32, 95% CI 0.52 to 3.31; P = 0.56; 1 trial, 33 participants; Analysis 2.3 ).

Somnolence

Three trials comparing pharmacological treatment to placebo reported data on somnolence as a non‐serious adverse event.

Two trials compared antipsychotics to placebo ( Schulz 2007 ; Zanarini 2007 ). There was a difference between the groups on somnolence, with an increased risk of somnolence using antipsychotics (RR 2.97, 95% CI 1.75 to 5.03; P < 0.0001, I ² = 0%; 2 trials, 615 participants; Analysis 2.1 ).

One trial compared memantine hydrochloride to placebo ( Kulkarni 2018 ). There was no difference in the presence of somnolence between the two groups (RR 1.65, 95% CI 0.59 to 4.57; P = 0.34; 1 trial, 33 participants; Analysis 2.3 ).

Sedation

Four trials comparing pharmacological treatment to placebo reported continuous data on sedation as a non‐serious adverse event. All four trials compared antipsychotics to placebo ( Black 2014 ; Pascual 2008 ; Schulz 2007 ; Zanarini 2001 ). There was a very slight difference in sedation between the two groups in favour of placebo (RR 2.66, 95% CI 0.99 to 7.12; P = 0.05, I ² = 67%; 4 trials, 445 participants; Analysis 2.1 ).

Anxiety

One trial comparing pharmacological treatment to placebo reported data on anxiety as a non‐serious adverse event. Schulz 2007 compared an antipsychotic to placebo. There was no difference in anxiety between the two groups (RR 0.90, 95% CI 0.33 to 2.42; P = 0.83; 1 trial, 314 participants; Analysis 2.1 ).

Insomnia

Two trials comparing pharmacological treatment to placebo reported data on insomnia as a non‐serious adverse event. Both trials compared antipsychotics to placebo ( Schulz 2007 ; Zanarini 2007 ). There was no difference in effects on insomnia between the two groups (RR 0.68, 95% CI 0.33 to 1.37; P = 0.28, I ² = 15%; 2 trials, 615 participants; Analysis 2.1 ).

Hyperinsomnia

One trial comparing pharmacological treatment to placebo reported data on hyperinsomnia as a non‐serious adverse event. Black 2014 compared an antipsychotic to placebo. There was no difference in hyperinsomnia between the two groups (RR 2.34, 95% CI 0.69 to 8.01; P = 0.17; 1 trial, 62 participants; Analysis 2.1 ).

Increased appetite

Three trials comparing pharmacological treatment to placebo reported data on increased appetite as a non‐serious adverse event. Grant 2022 , Schulz 2007 and Zanarini 2007 compared antipsychotics to placebo. There was an effect on increased appetite in favour of placebo, (RR 2.68, 95% CI 1.71 to 4.19; P < 0.0001; I ² = 0%, 3 trials, 692 participants; Analysis 2.1 ).

Change in appetite

One trial comparing pharmacological treatment to placebo reported dichotomous data on changes in appetite as a non‐serious adverse event. Black 2014 compared an antipsychotic to placebo. There was no difference in changes in appetite between the two groups (RR 0.65, 95% CI 0.10 to 4.06; P = 0.64; 1 trial, 17 participants; Analysis 2.1 ).

Forgetfulness or confusion

One trial comparing pharmacological treatment to placebo reported data on forgetfulness/confusion as a non‐serious adverse event. Black 2014 compared an antipsychotic to placebo. There was no difference in forgetfulness/confusion between the two groups (RR 1.46, 95% CI 0.38 to 5.60; P = 0.58; 1 trial, 62 participants; Analysis 2.1 ).

Disturbances in attention

One trial comparing pharmacological treatment to placebo reported data on disturbances in attention as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference in disturbances in attention between the two groups (RR 11.37, 95% CI 0.63 to 203.81; P = 0.10; 1 trial, 301 participants; Analysis 2.1 ).

Restlessness

One trial comparing pharmacological treatment to placebo reported data on restlessness as a non‐serious adverse event. Grant 2022 compared an antipsychotic to placebo. There was no difference in restlessness between the two groups (RR 0.93, 95% CI 0.20 to 4.30; P = 0.92; 1 trial, 77 participants; Analysis 2.1 ).

Hallucinations

One trial comparing pharmacological treatment to placebo reported data on hallucinations as a non‐serious adverse event. Grant 2022 compared an antipsychotic to placebo. There was no difference in hallucinations between the two groups (RR 0.19, 95% CI 0.01 to 3.74; P = 0.27; 1 trial, 77 participants; Analysis 2.1 ).

Sleep problems

One trial comparing pharmacological treatment to placebo reported data on sleep problems as a non‐serious adverse event. Grant 2022 compared an antipsychotic to placebo. There was no difference in sleep problems between the two groups (RR 0.19, 95% CI 0.01 to 3.74; P = 0.27; 1 trial, 77 participants; Analysis 2.1 ).

Tremor

One trial comparing pharmacological treatment to placebo reported data on tremor as a non‐serious adverse event. Grant 2022 compared an antipsychotic to placebo. There was no difference in tremor between the two groups (RR 0.31, 95% CI 0.01 to 7.36; P = 0.47; 1 trial, 77 participants; Analysis 2.1 ).

Memory problems

One trial comparing pharmacological treatment to placebo reported dichotomous data on memory problems as a non‐serious adverse event. Loew 2006 compared a mood stabiliser to placebo. There was no difference in memory problems between the two groups (RR 2.00, 95% CI 0.55 to 7.22; P = 0.29; 1 trial, 56 participants; Analysis 2.2 ).

Paraesthesia

One trial comparing pharmacological treatment to placebo reported dichotomous data on paraesthesia as a non‐serious adverse event. Loew 2006 compared a mood stabiliser to placebo. There was no difference in paraesthesia between the two groups (RR 3.00, 95% CI 0.33 to 27.12; P = 0.33; 1 trial, 56 participants; Analysis 2.2 ).

Gait/balance disturbances

One trial comparing pharmacological treatment to placebo reported dichotomous data on gait/balance disturbances. Kulkarni 2018 compared memantine hydrochloride to placebo. There was no difference in gait/balance disturbances between the two groups (RR 2.35, 95% CI 0.53 to 10.45; P = 0.26; 1 trial, 33 participants; Analysis 2.3 ).

Nervous system disorders

One trial comparing pharmacological treatment to placebo reported dichotomous data on nervous system disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in the appearance of nervous system disorders between the two groups (RR 1.05, 95% CI 0.68 to 1.62; P = 0.83; 1 trial, 276 participants; Analysis 2.2 ).

Psychiatric disorders

One trial comparing pharmacological treatment to placebo reported data on psychiatric disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in the appearance of psychiatric disorders between the two groups (RR 0.94, 95% CI 0.64 to 1.37; P = 0.74; 1 trial, 276 participants; Analysis 2.2 ).

Cardiovascular and respiratory system

Cold/flu symptoms

One trial comparing pharmacological treatment to placebo reported dichotomous data on cold/flu symptoms as a non‐serious adverse event. Black 2014 compared an antipsychotic to placebo. There was no difference in cold/flu symptoms between the two groups (RR 1.54, 95% CI 0.50 to 4.73; P = 0.45; 1 trial, 62 participants; Analysis 3.1 ).

Nasopharyngitis

One trial comparing pharmacological treatment to placebo reported dichotomous data on nasopharyngitis as a non‐serious adverse event. Schulz 2007 compared an antipsychotic to placebo. There was no difference in nasopharyngitis between the two groups (RR 0.62, 95% CI 0.23 to 1.66; P = 0.34; 1 trial, 301 participants; Analysis 3.1 ).

Sweating

One trial comparing pharmacological treatment to placebo reported dichotomous data on sweating as a non‐serious adverse event. Grant 2022 compared an antipsychotic to placebo. There was no difference in sweating between the two groups (RR 0.31, 95% CI 0.01 to 7.36; P = 0.47; 1 trial, 77 participants; Analysis 3.1 ).

Blood and lymphatic system disorders

One trial comparing pharmacological treatment to placebo reported dichotomous data on blood and lymphatic system disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in blood and lymphatic system disorders between the two groups (RR 0.68, 95% CI 0.11 to 3.99; P = 0.67; 1 trial, 276 participants; Analysis 3.3 ).

Cardiac disorders

One trial comparing pharmacological treatment to placebo reported dichotomous data on cardiac disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in cardiac disorders between the two groups (RR 0.34, 95% CI 0.01 to 8.23; P = 0.51; 1 trial, 276 participants; Analysis 3.3 ).

Endocrine disorders

One trial comparing pharmacological treatment to placebo reported dichotomous data on endocrine disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in endocrine disorders between the two groups (RR 0.34, 95% CI 0.01 to 8.23; P = 0.51; 1 trial, 276 participants; Analysis 3.3 ).

Respiratory, thoracic and mediastinal disorders

One trial comparing pharmacological treatment to placebo reported dichotomous data on respiratory, thoracic and mediastinal disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in respiratory, thoracic and mediastinal disorders between the two groups (RR 1.80, 95% CI 0.83 to 3.94; P = 0.14; 1 trial, 276 participants; Analysis 3.3 ).

Diastolic blood pressure in standing position (mean change from baseline to endpoint)

One trial comparing pharmacological treatment to placebo reported continuous data on diastolic blood pressure in standing position (mean change from baseline to endpoint) as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference in mean change between the two groups (MD −0.28, 95% CI −2.29 to 1.73; P = 0.78; 1 trial, 290 participants; Analysis 3.2 ).

Diastolic blood pressure in supine position (mean change from baseline to endpoint)

One trial comparing pharmacological treatment to placebo reported continuous data on diastolic blood pressure in supine position (mean change from baseline to endpoint) as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference in mean change between the two groups (MD −0.11, 95% CI −2.28 to 2.06; P = 0.92; 1 trial, 290 participants; Analysis 3.2 ).

Systolic blood pressure in standing position (mean change from baseline to endpoint)

One trial comparing pharmacological treatment to placebo reported continuous data on systolic blood pressure in standing position (mean change from baseline to endpoint) as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference in mean change between the two groups (MD 0.35, 95% CI −2.39 to 3.09; P = 0.80; 1 trial, 290 participants; Analysis 3.2 ).

Systolic blood pressure in supine position (mean change from baseline to endpoint)

One trial comparing pharmacological treatment to placebo reported continuous data on systolic blood pressure in supine position (mean change from baseline to endpoint) as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference in mean change between the two groups (MD −1.31, 95% CI −4.00 to 1.38; P = 0.34; 1 trial, 290 participants; Analysis 3.2 ).

Pulse in standing position (mean change from baseline to endpoint)

One trial comparing pharmacological treatment to placebo reported continuous data on pulse in standing position (mean change from baseline to endpoint) as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference in mean change between the two groups (MD 0.85, 95% CI −1.65 to 3.35; P = 0.50; 1 trial, 290 participants; Analysis 3.2 ).

Pulse in supine position (mean change from baseline to endpoint)

One trial comparing pharmacological treatment to placebo reported continuous data on pulse in supine position (mean change from baseline to endpoint) as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference in mean change between the two groups (MD −0.11, 95% CI −2.28 to 2.06; P = 0.92; 1 trial, 290 participants; Analysis 3.2 ).

Gastrointestinal system

Nausea

Five trials comparing pharmacological treatment to placebo reported dichotomous data on nausea and vomiting as a non‐serious adverse event.

Four trials compared antipsychotics to placebo ( Black 2014 ; Grant 2022 ; Schulz 2007 ; Zanarini 2007 ). There was no difference in nausea and vomiting between the two groups ((RR 0.80, 95% CI 0.49 to 1.29; P = 0.36; I ² = 0%; 4 trials, 754 participants; Analysis 4.1 ).

One trial compared memantine hydrochloride to placebo ( Kulkarni 2018 ). There was no difference in nausea and vomiting between the two groups (RR 1.00, 95% CI 0.45 to 2.23; P = 1.00; 1 trial, 34 participants; Analysis 4.4 ).

Uneasy feeling

One trial comparing pharmacological treatment to placebo reported dichotomous data on feelings of uneasiness as a non‐serious adverse event. Pascual 2008 compared an antipsychotic to placebo. There was no difference in the feeling of uneasiness between the two groups (RR 7.00, 95% CI 0.38 to 129.93; P = 0.19; 1 trial, 60 participants; Analysis 4.1 ).

Constipation

Two trials comparing pharmacological treatment to placebo reported dichotomous data on constipation as a non‐serious adverse event. One trial compared an antipsychotic to placebo ( Zanarini 2001 ). There was no difference in constipation between the two groups (RR 6.50, 95% CI 0.41 to 104.20; P = 0.19; 1 trial, 28 participants; Analysis 4.1 ). Another trial compared memantine hydrochloride to placebo ( Kulkarni 2018 ). There was no difference in constipation between the two groups (RR 1.65, 95% CI 0.59 to 4.57; P = 0.34; 1 trial, 33 participants; Analysis 4.4 ).

Dry mouth

Four trials comparing pharmacological treatment to placebo reported dichotomous data on dry mouth as a non‐serious adverse event. They all compared antipsychotics to placebo ( Black 2014 ; Grant 2022 ; Schulz 2007 ; Zanarini 2007 ). There was a difference between the two groups, with an increased risk of dry mouth for antipsychotics (RR 2.60, 95% CI 1.46 to 4.64; P = 0.001; I ² = 0%; 4 trials, 754 participants; Analysis 4.1 ).

Gastrointestinal disorders

One trial comparing pharmacological treatment to placebo reported dichotomous data on gastrointestinal disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was a slight difference between the two groups, with an increased risk of gastrointestinal disorders for placebo (RR 0.70, 95% CI 0.50 to 0.98; P = 0.04; 1 trial, 276 participants; Analysis 4.3 ).

General disorders and administration site conditions

One trial comparing pharmacological treatment to placebo reported dichotomous data on general disorders and administration site conditions as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in general disorders and administration site conditions between the two groups (RR 1.01, 95% CI 0.50 to 2.05; P = 0.97; 1 trial, 276 participants; Analysis 4.3 ).

Hepatobiliary disorders

One trial comparing pharmacological treatment to placebo reported dichotomous data on hepatobiliary disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in hepatobiliary disorders between the two groups (RR 3.04, 95% CI 0.13 to 74.07; P = 0.49; 1 trial, 276 participants; Analysis 4.3 ).

Metabolism and nutrition disorders

One trial that compared a pharmacological treatment to placebo reported dichotomous data on metabolism and nutrition disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in metabolism and nutrition disorders between the two groups (RR 2.03, 95% CI 0.19 to 22.12; P = 0.56; 1 trial, 276 participants; Analysis 4.3 ).

Liver function: alanine aminotransferase (ALT)/serum glutamic‐pyruvic transaminase (SGPT) baseline to endpoint mean change (u/L)

Two trials that compared a pharmacological treatment to placebo reported continuous data on mean change in ALT/SGPT as a non‐serious adverse event. Schulz 2007 and Zanarini 2007 compared antipsychotics to placebo. There was a difference between the two groups regarding mean change in ALT/SGPT favouring placebo (SMD 0.46, 95% CI 0.29 to 0.63; P < 0.001, I ² = 0%; 2 trials, 530 participants; Analysis 4.2 ).

Liver function: aspartate aminotransferase (AST)/serum glutamic‐oxaloacetic transaminase (SGOT) baseline to endpoint mean change (u/L)

Two trials that compared a pharmacological treatment to placebo reported continuous data on mean change in AST/SGOT as a non‐serious adverse event. Schulz 2007 and Zanarini 2007 compared antipsychotics to placebo. There was a difference between the two groups regarding mean change in AST/SGOT favouring placebo (SMD 0.35, 95% CI 0.18 to 0.52; P < 0.001, I ² = 0%; 2 trials, 526 participants; Analysis 4.2 ).

Liver function: total bilirubin baseline to endpoint mean change (μmol/L)

One trial that compared a pharmacological treatment to placebo reported continuous data on mean change in total bilirubin as a non‐serious adverse event. Schulz 2007 compared an antipsychotic to placebo. There was a difference between the two groups regarding mean change in total bilirubin favouring the antipsychotic (MD −0.98, 95% CI −1.80 to −0.16; P = 0.02; 1 trial, 264 participants; Analysis 4.2 ).

Liver function: direct bilirubin baseline to endpoint mean change (μmol/L)

One trial that compared a pharmacological treatment to placebo reported continuous data on mean change in direct bilirubin as a non‐serious adverse event. Schulz 2007 compared an antipsychotic to placebo. There was a difference between the two groups regarding mean change in direct bilirubin favouring the antipsychotic (MD −0.30, 95% CI −0.51 to −0.09; P = 0.004; 1 trial, 258 participants; Analysis 4.2 ).

Liver function: gamma‐glutamyl transferase (GGT) baseline to endpoint mean change

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in GGT as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was a difference between the two groups regarding mean change in GGT favouring placebo (MD 2.96, 95% CI 0.22 to 5.70; P = 0.03; 1 trial, 268 participants; Analysis 4.2 ).

Lipids: total cholesterol baseline to endpoint change (mmol/L)

Two trials comparing pharmacological treatment to placebo reported continuous data on mean change in total cholesterol as a non‐serious adverse event. Both trials compared antipsychotics to placebo ( Schulz 2007 ; Soler 2005 ). There was a difference between the two groups regarding mean change in total cholesterol favouring placebo (SMD 0.42, 95% CI 0.20 to 0.64; P = 0.0002, I ² = 0%; 2 trials, 327 participants; Analysis 4.2 ).

Lipids: low‐density lipoprotein (LDL) cholesterol baseline to endpoint mean change (mmol/L)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in LDL cholesterol as a non‐serious adverse event. Schulz 2007 compared an antipsychotic to placebo. There was a difference between the two groups regarding mean change in LDL cholesterol favouring placebo (MD 0.21, 95% CI 0.06 to 0.36; P = 0.005; 1 trial, 259 participants; Analysis 4.2 ).

Lipids: high‐density lipoprotein (HDL) cholesterol (dextran precip.) baseline to endpoint mean change (mmol/L)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in HDL cholesterol as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was a difference between the two groups regarding mean change in HDL cholesterol favouring the antipsychotic (MD −0.06, 95% CI −0.11 to −0.01; P = 0.02; 1 trial, 269 participants; Analysis 4.2 ).

Lipids: triglycerides, fasting, baseline to endpoint mean change (mmol/L)

One trial compared a pharmacological treatment to placebo reported continuous data on mean change in triglycerides as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was a difference between the two groups regarding mean change in triglycerides favouring placebo (MD 0.27, 95% CI 0.07 to 0.47; P = 0.009; 1 trial, 203 participants; Analysis 4.2 ).

Prolactin: baseline to endpoint mean change (μg/L)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in prolactin as a non‐serious adverse event. Schulz 2007 compared an antipsychotic to placebo. There was a difference between the two groups regarding mean change in prolactin favouring placebo (MD 7.10, 95% CI 1.64 to 12.56; P = 0.01; 1 trial, 259 participants; Analysis 4.2 ).

Platelet count baseline to endpoint mean change (GI/L)

Two trials comparing a pharmacological treatment to placebo reported continuous data on mean change in platelet count as a non‐serious adverse event. Both trials compared antipsychotics to placebo ( Schulz 2007 ; Zanarini 2007 ). There was no difference in platelet count between the two groups (SMD 0.03, 95% CI −0.53 to 0.59; P = 0.91, I ² = 90%; 2 trials, 517 participants; Analysis 4.2 ).

Erythrocyte count baseline to endpoint mean change (TI/L)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in erythrocyte count as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference in erythrocyte count between the two groups (MD −0.05, 95% CI −0.12 to 0.02; P = 0.14; 1 trial, 262 participants; Analysis 4.2 ).

Leukocyte count baseline to endpoint mean change (GI/L)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in leukocyte count as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was a difference between the two groups regarding mean change in leukocyte count favouring the antipsychotic (MD −0.70, 95% CI −1.12 to −0.28; P = 0.001; 1 trial, 262 participants; Analysis 4.2 ).

Neutrophils, segmented, baseline to endpoint mean change (GI/L)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in neutrophils as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was a difference between the two groups regarding mean change in neutrophils favouring the antipsychotic (MD −0.60, 95% CI −0.97 to −0.23; P = 0.002; 1 trial, 262 participants; Analysis 4.2 ).

Basophils baseline to endpoint mean change (GI/L)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in basophils as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was a difference between the two groups regarding mean change in basophils favouring the antipsychotic (MD −0.01, 95% CI −0.02 to −0.00; P = 0.02; 1 trial, 262 participants; Analysis 4.2 ).

Monocytes baseline to endpoint mean change (GI/L)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in monocytes as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was a difference between the two groups regarding mean change in monocytes favouring the antipsychotic (MD −0.04, 95% CI −0.07 to −0.01; P = 0.02; 1 trial, 262 participants; Analysis 4.2 ).

Haemoglobin baseline to endpoint mean change (mml/L‐F)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in haemoglobin as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference between the two groups regarding mean change in haemoglobin (MD −0.11, 95% CI −0.24 to 0.02; P = 0.09; 1 trial, 262 participants; Analysis 4.2 ).

Mean cell haemoglobin concentration (MCHC) baseline to endpoint mean change (mml/L‐F)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in MCHC as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference between the two groups regarding MCHC (MD 0.02, 95% CI −0.17 to 0.21; P = 0.84; 1 trial, 260 participants; Analysis 4.2 ).

Calcium baseline to endpoint mean change (mmol/L)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in calcium as a non‐serious adverse event. Schulz 2007 compared an antipsychotic to placebo. There was a difference between the two groups regarding mean change in calcium favouring the antipsychotic (MD −0.03, 95% CI −0.05 to −0.01; P = 0.006; 1 trial, 268 participants; Analysis 4.2 ).

Albumin baseline to endpoint mean change (g/L)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in albumin as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference between the two groups regarding mean change in albumin (MD −0.67, 95% CI −1.42 to 0.08; P = 0.08; 1 trial, 269 participants; Analysis 4.2 ).

Creatine phosphokinase baseline to endpoint mean change (u/L)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in creatine phosphokinase as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference between the two groups regarding mean change in creatine phosphokinase (MD −44.81, 95% CI −95.39 to 5.77; P = 0.08; 1 trial, 268 participants; Analysis 4.2 ).

Urea nitrogen baseline to endpoint mean change (mmol/L)

One trial comparing a pharmacological treatment to placebo reported continuous data on mean change in urea nitrogen as a non‐serious adverse event. Zanarini 2007 compared an antipsychotic to placebo. There was no difference between the two groups regarding mean change in urea nitrogen (MD −0.17, 95% CI −0.46 to 0.12; P = 0.25; 1 trial, 269 participants Analysis 4.2 ).

Musculoskeletal system

Bodily pain

One trial comparing a pharmacological treatment to placebo reported dichotomous data on bodily pain as a non‐serious adverse event. Black 2014 compared an antipsychotic to placebo. There was no difference in bodily pain between the two groups (RR 0.88, 95% CI 0.47 to 1.64; P = 0.69; 1 trial, 62 participants; Analysis 5.1 ).

Musculoskeletal and connective tissue disorders

One trial comparing a pharmacological treatment to placebo reported dichotomous data on musculoskeletal and connective tissue disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in musculoskeletal and connective tissue disorders between the two groups (RR 1.16, 95% CI 0.43 to 3.11; P = 0.77; 1 trial; 276 participants; Analysis 5.4 ).

Body weight change

Twelve trials comparing a pharmacological treatment to placebo reported continuous data on body weight change as a non‐serious adverse event.

Seven trials compared antipsychotics to placebo ( Bogenschutz 2004 ; Linehan 2008 ; Schulz 2007 ; Soler 2005 ; Soloff 1993 ; Zanarini 2001 ; Zanarini 2007 ). There was a difference between the two groups regarding body weight change favouring placebo (SMD 0.78, 95% CI 0.44 to 1.12; P < 0.001, I ² = 74%; 7 trials, 810 participants; Analysis 5.2 ).

Soloff 1993 also compared an antidepressant to placebo. There was no difference between the two groups regarding body weight change (MD 0.09, 95% CI −0.31 to 0.49; P = 0.66; 1 trial, 62 participants; Analysis 5.3 ).

Five trials compared mood stabilisers to placebo ( Frankenburg 2002 ; Loew 2006 ; Nickel 2004 ; Nickel 2005 ; Tritt 2005 ). There was no difference between the two groups regarding body weight change (SMD −0.26, 95% CI −0.72 to 0.20; P = 0.27, I ² = 55%; 5 trials, 184 participants; Analysis 5.5 ).

Sensory system

Eye disorders

One trial comparing a pharmacological treatment to placebo reported dichotomous data on eye disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in eye disorders between the two groups (RR 0.17, 95% CI 0.02 to 1.39; P = 0.10; 1 trial, 276 participants; Analysis 6.1 ).

Reproductive system

Pregnancy, puerperium and perinatal conditions

One trial comparing a pharmacological treatment to placebo reported dichotomous data on pregnancy, puerperium and perinatal conditions as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in pregnancy, puerperium and perinatal conditions between the two groups (RR 1.52, 95% CI 0.26 to 8.97; P = 0.64; 1 trial, 276 participants; Analysis 7.1 ).

Reproductive system and breast disorders

One trial comparing pharmacological treatment to placebo reported dichotomous data on the reproductive system and breast disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in the reproductive system and breast disorders between the two groups (RR 3.04, 95% CI 0.32 to 28.90; P = 0.33; 1 trial, 276 participants; Analysis 7.1 ).

Menstrual pain

One trial comparing a pharmacological treatment to placebo reported dichotomous data on menstrual pain as a non‐serious adverse event. Loew 2006 compared a mood stabiliser to placebo. There was no difference in menstrual pain between the two groups (RR 1.67, 95% CI 0.44 to 6.31; P = 0.45; 1 trial, 56 participants; Analysis 7.1 ).

Other

Injury, poisoning or procedural complications

One trial comparing a pharmacological treatment to placebo reported dichotomous data on injuries, poisonings and procedural complications as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was a difference between the two groups regarding injuries, poisonings and procedural complications, with an increased risk of injuries, poisonings and procedural complications in the placebo group (RR 0.44, 95% CI 0.26 to 0.74; P = 0.002; 1 trial, 276 participants; Analysis 8.1 ).

Skin and subcutaneous tissue disorders

One trial comparing a pharmacological treatment to placebo reported dichotomous data on skin and subcutaneous tissue disorders as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in skin and subcutaneous tissue disorders between the two groups (RR 1.15, 95% CI 0.75 to 1.75; P = 0.53; 1 trial, 276 participants; Analysis 8.1 ).

Social circumstances

One trial comparing a pharmacological treatment to placebo reported dichotomous data on social circumstances as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no evidence of a difference in social circumstances between the two groups (RR 1.01, 95% CI 0.06 to 16.06; P = 0.99; 1 trial, 279 participants; Analysis 8.1 ).

Surgical and medical procedures

One trial comparing a pharmacological treatment to placebo reported dichotomous data on surgical and medical procedures as a non‐serious adverse event. Crawford 2018 compared a mood stabiliser to placebo. There was no difference in surgical and medical procedures between the two groups (RR 4.06, 95% CI 0.46 to 35.85; P = 0.21; 1 trial, 276 participants; Analysis 8.1 ).

Withdrawal due to adverse events

One trial comparing pharmacological treatment to placebo reported dichotomous data on withdrawal due to adverse events. Kulkarni 2018 compared memantine hydrochloride to placebo. There was no difference in withdrawal due to adverse events between the two groups (RR 2.82, 95% CI 0.33 to 24.43; P = 0.35; 1 trial, 33 participants; Analysis 9.1 ).

2.1 BPD symptom severity

Three trials compared one medication with another medication and reported continuous data for BPD symptom severity.

Bozzatello 2017 compared two second‐generation antipsychotics, olanzapine and asenapine. There was no clear difference between the two treatments regarding the effects on BPD symptom severity (MD −2.23, 95% CI −8.04 to 3.58; P = 0.45; 1 trial, 51 participants; Analysis 10.1 ).

Soloff 1993 compared the antipsychotic haloperidol with the antidepressant phenelzine sulfate. There was no clear difference between the two treatments regarding BPD symptom severity (MD 5.73, 95% CI −0.33 to 11.79; P = 0.06; 1 trial, 64 participants; Analysis 10.1 ).

Cowdry 1988 had a cross‐over design with four active treatment arms: an arm with a benzodiazepine (alprazolam), an arm with a mood stabiliser (carbamazepine), an arm with an antipsychotic (trifluoperazine hydrochloride) and an arm with an antidepressant (tranylcypromine sulfate). From these, it was possible to make six comparisons: 1) alprazolam compared to carbamazepine; 2) alprazolam compared to trifluoperazine hydrochloride; 3) alprazolam compared to tranylcypromine sulfate; 4) carbamazepine compared to trifluoperazine hydrochloride; 5) carbamazepine compared to tranylcypromine sulfate; and 6) trifluoperazine hydrochloride compared to tranylcypromine sulfate.

There was a difference in BPD symptoms at end of treatment between alprazolam and carbamazepine, favouring alprazolam (MD −1.64, 95% CI −2.71 to −0.57; P = 0.003; 1 trial, 27 participants; Analysis 10.1 ), as well as between alprazolam and tranylcypromine sulfate, favouring alprazolam (MD −1.58, 95% CI −2.76 to −0.40; P = 0.009; 1 trial, 24 participants; Analysis 10.1 ).

There were no differences between treatments on BPD symptom severity at end of treatment for any of the other comparisons (alprazolam compared to trifluoperazine hydrochloride: MD −0.80, 95% CI −2.21 to 0.61; P = 0.27; 1 trial, 22 participants; Analysis 10.1 ; carbamazepine compared to trifluoperazine hydrochloride: MD 0.84, 95% CI −0.41 to 2.09; P = 0.19 ; 1 trial, 25 participants; Analysis 10.1 ; carbamazepine compared to tranylcypromine sulfate: MD 0.06, 95% CI −0.92 to 1.04; P = 0.90; 1 trial, 27 participants; Analysis 10.1 ; trifluoperazine hydrochloride compared to tranylcypromine sulfate: MD −0.78, 95% CI −2.13 to 0.57; P = 0.26; 1 trial, 22 participants; Analysis 10.1 ).

2.2 Self‐harm

One trial compared a medication with another medication and reported continuous data for self‐harm. Bozzatello 2017 compared two antipsychotics: olanzapine and asenapine. There was no evidence of a difference between the two antipsychotics regarding self‐harm (MD 0.21, 95% CI −0.58 to 1.00; P = 0.60; 1 trial, 51 participants; Analysis 10.2 ).

There were no other trials comparing medications to another medication that reported data on self‐harm.

2.3 Suicide‐related outcomes

One trial comparing medications to other medications reported continuous data on suicide‐related outcomes.

Cowdry 1988 had a cross‐over design with four active treatment arms (see above in 2.1).

There was a difference in suicide‐related outcomes at end of treatment between alprazolam and carbamazepine, favouring carbamazepine (MD 2.12, 95% CI 1.06 to 3.18; P < 0.001; 1 trial, 27 participants; Analysis 10.3 ); between alprazolam and trifluoperazine hydrochloride, favouring trifluoperazine hydrochloride (MD 1.73, 95% CI 0.62 to 2.84; P = 0.002; 1 trial, 22 participants; Analysis 10.3 ); and between alprazolam and tranylcypromine sulfate, favouring tranylcypromine (MD 2.00, 95% CI 0.89 to 3.11; P = 0.0004; 1 trial, 24 participants; Analysis 10.3 ).

There was no difference between treatments on suicide‐related outcomes at end of treatment for any of the other comparisons (carbamazepine compared to trifluoperazine hydrochloride: MD −0.39, 95% CI −1.53 to 0.75; P = 0.50; 1 trial, 25 participants; Analysis 10.3 ; carbamazepine compared to tranylcypromine sulfate: MD −0.12, 95% CI −1.26 to 1.02; P = 0.84; 1 trial, 27 participants; Analysis 10.3 ; trifluoperazine hydrochloride compared to tranylcypromine sulfate: MD 0.27, 95% CI −1.00 to 1.54; P = 0.68; 1 trial, 22 participants; Analysis 10.3 ).

2.4 Psychosocial functioning

Five trials compared a medication with another medication and reported continuous data for psychosocial functioning.

Three trials compared one antipsychotic with another antipsychotic ( Bozzatello 2017 ; Shafti 2010 ; Shafti 2014 ). There was no difference between the treatments for any of the comparisons regarding psychosocial functioning; for Bozzatello 2017 comparing olanzapine to asenapine (MD = 0.20, 95% CI −0.23 to 0.63, P = 0.36; 1 trial, 51 participants; Analysis 10.4 ); for Shafti 2010 comparing olanzapine to haloperidol (MD 0.35, 95% CI −0.45 to 1.15; P = 0.39; 1 trial, 28 participants; Analysis 10.4 ); or for Shafti 2014 comparing olanzapine to aripiprazole (MD 0.12, 95% CI −0.58 to 0.82; P = 0.74; 1 trial, 24 participants; Analysis 10.4 ).

Two trials compared an antipsychotic to an antidepressant ( Soloff 1989 ; Soloff 1993 ). Soloff 1989 compared haloperidol to amitriptyline. There was no difference between the treatments regarding psychosocial functioning (MD −3.87, 95% CI −10.67 to 2.93; P = 0.26; 1 trial, 57 participants; Analysis 10.4 ). One study, Soloff 1993 , compared haloperidol to phenelzine sulfate and found a difference in psychosocial functioning following treatment that favoured phenelzine sulfate (MD 5.15, 95% CI 0.29 to 10.01; P = 0.04; 1 trial, 64 participants; Analysis 10.4 ).

2.5 Anger

Six trials compared a medication with another medication and reported continuous data for anger.

Three trials compared an antipsychotic to another antipsychotic: Shafti 2010 compared olanzapine to haloperidol, Shafti 2014 compared olanzapine to aripiprazole, and Bozzatello 2017 compared olanzapine to asenapine. There was no difference regarding anger at the end of treatment between olanzapine and haloperidol (MD 0.21, 95% CI −8.90 to 9.32; P = 0.96; 1 trial, 28 participants; Analysis 10.5 ) or between olanzapine and aripiprazole (MD −0.40, 95% CI −8.05 to 7.25; P = 0.92, 1 trial, 24 participants; Analysis 10.5 ). However, there was a difference between asenapine and olanzapine regarding anger at end of treatment, favouring asenapine (MD 1.14, 95% CI 0.31 to 1.97; P = 0.007; 1 trial, 51 participants; Analysis 10.5 ).

Three trials compared an antipsychotic to an antidepressant. Jariani 2010 compared olanzapine to sertraline, Soloff 1989 compared haloperidol to amitriptyline, and Soloff 1993 compared haloperidol to phenelzine sulfate. There was no difference regarding anger at end of treatment between haloperidol and amitriptyline (MD −0.34, 95% CI −0.82 to 0.14; P = 0.16; 1 trial, 57 participants; Analysis 10.5 ), or between haloperidol and phenelzine sulfate (MD 0.06, 95% CI −0.31 to 0.43; P = 0.75; 1 trial, 64 participants; Analysis 10.5 ). However, there was a superior effect of olanzapine compared to sertraline regarding anger at the end of treatment (MD −0.33, 95% CI −0.48 to −0.18; P < 0.001; 1 trial, 120 participants; Analysis 10.5 ).

Cowdry 1988 had a cross‐over design with four active treatment arms (see above in 2.1).

There was a difference in anger at end of treatment between alprazolam and carbamazepine, favouring carbamazepine (MD 1.65, 95% CI 0.80 to 2.50; P = 0.0001; 1 trial, 27 participants; Analysis 10.5 ), as well as between alprazolam and tranylcypromine sulfate, favouring tranylcypromine sulfate (MD 1.41, 95% CI 0.24 to 2.58; P = 0.02; 1 trial, 24 participants; Analysis 10.5 ).

There was no difference between treatments in anger at the end of treatment for any of the other comparisons: alprazolam compared to trifluoperazine hydrochloride (MD 0.58, 95% CI −0.77 to 1.93; P = 0.40; 1 trial, 22 participants; Analysis 10.5 ); carbamazepine compared to trifluoperazine hydrochloride (MD −1.07, 95% CI −2.28 to 0.14; P = 0.08; 1 trial, 25 participants; Analysis 10.5 ); carbamazepine compared to tranylcypromine sulfate (MD −0.24, 95% CI −1.23 to 0.75; P = 0.64; 1 trial, 27 participants; Analysis 10.5 ); and trifluoperazine hydrochloride compared to tranylcypromine sulfate (MD 0.83, 95% CI −0.62 to 2.28; P = 0.26; 1 trial, 22 participants; Analysis 10.5 ).

2.6 Affective instability

One trial compared a medication with another medication and reported continuous data for affective instability. Bozzatello 2017 compared two antipsychotics, olanzapine and asenapine. There was a difference in effects between the two antipsychotics regarding affective instability favouring asenapine (MD 2.28, 95% CI 1.51 to 3.05; P < 0.001; 1 trial, 51 participants; Analysis 10.6 ).

No other trials comparing medications to alternate medications reported data on affective instability.

2.7 Chronic feelings of emptiness

One trial compared a medication with another medication and reported continuous data for chronic feelings of emptiness. Bozzatello 2017 compared the two antipsychotics, olanzapine and asenapine. There was no difference in effects between the two antipsychotics regarding chronic feelings of emptiness (MD −0.54, 95% CI −1.29 to 0.21; P = 0.16, 1 trial, 51 participants; Analysis 10.7 ).

No other trials comparing a medication with another medication reported data on chronic feelings of emptiness.

2.8 Impulsivity

Five trials compared a medication with another medication and reported continuous data for impulsivity.

One trial compared two antipsychotics, olanzapine and asenapine ( Bozzatello 2017 ). There was no difference between olanzapine and asenapine regarding impulsivity at the end of treatment (MD −0.78, 95% CI −1.59 to 0.03, P = 0.06; 1 trial, 51 participants; Analysis 10.8 ).

Three trials compared an antipsychotic with an antidepressant. Soloff 1989 compared haloperidol to amitriptyline, Soloff 1993 compared haloperidol to phenelzine sulfate, and Zanarini 2004 compared olanzapine to fluoxetine. There was no difference between treatments regarding impulsivity at the end of treatment for any of the comparisons (haloperidol versus amitriptyline: MD 3.52, 95% CI −5.52 to 12.56; P = 0.45; 1 trial, 57 participants; Analysis 10.8 ; haloperidol versus phenelzine sulfate: MD 3.29, 95% CI −14.52 to 21.10; P = 0.72; 1 trial, 64 participants; Analysis 10.8 ; and olanzapine versus fluoxetine: MD −4.31, 95% CI −19.72 to 11.10; P = 0.58; 1 trial, 29 participants; Analysis 10.8 ).

Cowdry 1988 had a cross‐over design with four active treatment arms (see above in 2.1).

There was a difference in impulsivity at end of treatment in four of the comparisons, specifically between: alprazolam and carbamazepine, favouring carbamazepine (MD 2.18, 95% CI 1.20 to 3.16; P < 0.001; 1 trial, 27 participants; Analysis 10.8 ); alprazolam and tranylcypromine sulfate, favouring tranylcypromine sulfate (MD 1.83, 95% CI 0.66 to 3.00; P = 0.002; 1 trial, 24 participants; Analysis 10.8 ); carbamazepine and trifluoperazine hydrochloride, favouring carbamazepine (MD −1.73, 95% CI −2.87 to −0.59; P = 0.003; 1 trial, 25 participants; Analysis 10.8 ); and trifluoperazine hydrochloride and tranylcypromine sulfate, favouring tranylcypromine sulfate (MD 1.38, 95% CI 0.08 to 2.68; P = 0.04; 1 trial, 22 participants; Analysis 10.8 ).

There was no difference between treatments in impulsivity at end of treatment for any of the other two comparisons (alprazolam compared to trifluoperazine hydrochloride: MD 0.45, 95% CI −0.87 to 1.77; P = 0.50; 1 trial, 22 participants; Analysis 10.8 ; and carbamazepine compared to tranylcypromine sulfate: MD −0.35, 95% CI −1.31 to 0.61; P = 0.48; 1 trial, 27 participants; Analysis 10.8 ).

2.9 Interpersonal problems

Three trials compared a medication with another medication and reported continuous data for interpersonal problems.

One trial ( Bozzatello 2017 ) compared two antipsychotics, olanzapine and asenapine. There was no difference between olanzapine and asenapine regarding interpersonal problems at the end of treatment (MD 0.40, 95% CI −0.35 to 1.15, P = 0.29; 1 trial, 51 participants; Analysis 10.9 ).

Two trials compared an antipsychotic with antidepressants. Soloff 1989 compared haloperidol to amitriptyline and Soloff 1993 compared haloperidol to phenelzine sulfate. There was no difference between the two treatments regarding interpersonal problems at end of treatment, either when comparing haloperidol to amitriptyline (MD −0.13, 95% CI −0.62 to 0.36; P = 0.60; 1 trial, 57 participants; Analysis 10.9 ), or to phenelzine sulfate (MD −0.33, 95% CI −0.68 to 0.02; P = 0.06; 1 trial, 64 participants; Analysis 10.9 ).

2.10 Abandonment

One trial compared a medication with another medication and reported continuous data for fear of abandonment. Bozzatello 2017 compared two antipsychotics. There was no difference between the two antipsychotics (olanzapine and asenapine) regarding avoidance of abandonment (MD −0.40, 95% CI −1.07 to 0.27, P = 0.24; 1 trial, 51 participants; Analysis 10.10 ).

No other trials comparing a medication with another medication reported data on fear of abandonment.

2.11 Identity disturbance

One trial compared a medication with another medication and reported continuous data for identity disturbance. Bozzatello 2017 compared the two antipsychotics olanzapine and asenapine. There was no difference in effects between olanzapine and asenapine regarding identity disturbance (MD 0.68, 95% CI −0.12 to 1.48, P = 0.10; 1 trial, 51 participants; Analysis 10.11 ).

No other trials comparing a medication with another medication reported data on identity disturbance.

2.12 Dissociation and psychotic‐like symptoms

Five trials compared a medication with another medication and reported continuous data for dissociation and psychotic‐like symptoms.

Three trials compared an antipsychotic to another antipsychotic. Bozzatello 2017 compared olanzapine to asenapine, Shafti 2010 olanzapine to haloperidol, and Shafti 2014 olanzapine to aripiprazole. There was no difference between treatments regarding dissociation and psychotic‐like symptoms at the end of treatment for any of the comparisons: olanzapine and asenapine (MD −0.69, 95% CI −1.53 to 0.15; P = 0.11; 1 trial, 51 participants; Analysis 10.12 ), olanzapine and haloperidol (MD −2.30, 95% CI −10.15 to 5.55; P = 0.57; 1 trial, 28 participants; Analysis 10.12 ), and olanzapine and aripiprazole (MD −3.30, 95% CI −10.63 to 4.03; P = 0.38; 1 trial, 24 participants; Analysis 10.12 ).

Two trials compared an antipsychotic with antidepressants. Soloff 1989 compared haloperidol to amitriptyline and Soloff 1993 compared haloperidol to phenelzine sulfate. There was no difference between treatments for any of the comparisons regarding dissociation and psychotic‐like symptoms at the end of treatment (haloperidol versus amitriptyline: MD −0.28, 95% CI −0.69 to 0.13; P = 0.18; 1 trial, 57 participants; Analysis 10.12 ; and haloperidol versus phenelzine sulfate: MD 0.14, 95% CI −0.31 to 0.59; P = 0.54; 1 trial, 64 participants; Analysis 10.12 ).

2.13 Depression

Six trials compared a medication with another medication and reported continuous data for depression.

Bozzatello 2017 compared the two antipsychotics olanzapine and asenapine. There was a difference in effects between olanzapine and asenapine regarding depression, favouring asenapine (MD 2.90, 95% CI 0.88 to 4.92; P = 0.005; 1 trial, 51 participants; Analysis 10.13 ).

Four trials compared an antipsychotic with an antidepressant. Jariani 2010 compared olanzapine to sertraline, Soloff 1989 compared haloperidol to amitriptyline, Soloff 1993 compared haloperidol to phenelzine sulfate, and Zanarini 2004 compared olanzapine to fluoxetine. There was no difference between haloperidol and amitriptyline regarding depression at end of treatment (MD 0.88, 95% CI −5.12 to 6.88; P = 0.77; 1 trial, 57 participants; Analysis 10.13 ); however, a difference between treatments was observed in the comparisons between olanzapine and sertraline, favouring sertraline (MD 0.37, 95% CI 0.22 to 0.52; P < 0.001; 1 trial, 120 participants; Analysis 10.13 ); haloperidol and phenelzine sulfate, favouring phenelzine sulfate (MD 7.81, 95% CI 2.13 to 13.49; P = 0.007; 1 trial, 64 participants; Analysis 10.13 ); and olanzapine and fluoxetine, favouring olanzapine (MD −5.40, 95% CI −10.68 to −0.12; P = 0.04; 1 trial, 29 participants; Analysis 10.13 ).

Cowdry 1988 had a cross‐over design with four active treatment arms (see above in 2.1).

There was a difference in depression at the end of treatment between alprazolam and carbamazepine, favouring carbamazepine (MD 1.36, 95% CI 0.37 to 2.35; P = 0.007; 27 participants, 1 trial; Analysis 10.13 ), as well as between alprazolam and tranylcypromine sulfate, favouring tranylcypromine sulfate (MD 1.67, 95% CI 0.48 to 2.86; P = 0.006; 24 participants, 1 trial; Analysis 10.13 ).

There was no difference between treatments in terms of depression at end of treatment for any of the other comparisons (alprazolam compared to trifluoperazine hydrochloride: MD 0.90, 95% CI −0.49 to 2.29; P = 0.20; 1 trial, 22 participants; Analysis 10.13 ; carbamazepine compared to trifluoperazine hydrochloride: MD −0.46, 95% CI −1.81 to 0.89; P = 0.50; 1 trial, 25 participants; Analysis 10.13 ; carbamazepine compared to tranylcypromine sulfate: MD 0.31, 95% CI −0.82 to 1.44; P = 0.59; 1 trial, 27 participants; Analysis 10.13 ; and trifluoperazine hydrochloride compared to tranylcypromine sulfate: MD 0.77, 95% CI −0.73 to 2.27; P = 0.31; 1 trial, 22 participants; Analysis 10.13 ).

2.14 Attrition

Eight trials compared a medication with another medication and reported dichotomous data on attrition.

Four trials compared an antipsychotic to another antipsychotic. Bozzatello 2017 compared olanzapine to asenapine, Leone 1982 compared loxapine to chlorpromazine, Shafti 2010 compared olanzapine to haloperidol, and Shafti 2014 compared olanzapine to aripiprazole. There was no difference between treatments for any of the comparisons on attrition (olanzapine versus asenapine: RR 0.80, 95% CI 0.28 to 2.29; P = 0.68; 1 trial, 51 participants; Analysis 10.14 ; loxapine versus chlorpromazine: RR 1.14, 95% CI 0.46 to 2.85; P = 0.77; 1 trial, 80 participants; Analysis 10.14 ; olanzapine versus haloperidol: no events occurred in any of the treatment groups; Analysis 10.14 ; and olanzapine versus aripiprazole: no events occurred in any of the treatment groups; Analysis 10.14 ).

Four trials compared an antipsychotic to an antidepressant. Jariani 2010 compared olanzapine to sertraline, Soloff 1989 compared haloperidol to amitriptyline, Soloff 1993 compared haloperidol to phenelzine sulfate, and Zanarini 2004 compared olanzapine to fluoxetine. There was no difference between treatments for any of the comparisons on attrition (olanzapine versus sertraline: no events occurred in any of the treatment groups; Analysis 10.14 ; haloperidol versus amitriptyline: RR 2.90, 95% CI 0.32 to 26.38; P = 0.34; 1 trial, 61 participants; Analysis 10.14 ; haloperidol versus phenelzine sulfate: RR 1.58, 95% CI 0.49 to 5.15; P = 0.45; 1 trial, 75 participants; Analysis 10.14 ; and olanzapine versus fluoxetine: RR 0.29, 95% CI 0.01 to 6.69; P = 0.44; 1 trial, 30 participants; Analysis 10.14 ).

2.15 Adverse events

No trials comparing a medication with another medication reported data on non‐serious adverse events of the gastrointestinal system, the sensory system or the reproductive system; or on any other adverse events than those suggested in our analysis plan.

Three trials comparing a medication with another medication reported dichotomous data on total non‐serious adverse events.

All trials compared an antipsychotic to another antipsychotic. Leone 1982 compared loxapine to chlorpromazine, Shafti 2010 compared olanzapine to haloperidol, and Shafti 2014 compared olanzapine to aripiprazole. There was no difference between treatments regarding adverse events in any of the comparisons i.e. loxapine versus chlorpromazine (RR 1.27, 95% CI 0.66 to 2.45; P = 0.47; 1 trial, 80 participants, Analysis 10.15 ; olanzapine versus haloperidol (RR 0.75, 95% CI 0.35 to 1.60; P = 0.46; 1 trial, 28 participants; Analysis 10.15 ; and olanzapine versus aripiprazole (RR 0.75, 95% CI 0.38 to 1.50; P = 0.42; 1 trial, 24 participants; Analysis 10.15 ). No trials comparing a medication with another medication reported data on serious adverse events.

Withdrawal due to adverse events

One trial comparing a medication with another medication reported dichotomous data on withdrawal due to adverse events. Bozzatello 2017 compared the antipsychotics olanzapine and asenapine. There was no difference between the treatment groups in withdrawal due to adverse events (RR 0.96, 95% CI 0.15 to 6.31; P = 0.97; 1 trial, 51 participants; Analysis 11.1 ).

Central nervous system

Sedation

One trial comparing a medication with another medication reported dichotomous data on sedation as a non‐serious adverse event. Zanarini 2004 compared an antipsychotic with an antidepressant. There was an increased risk of sedation with antipsychotics compared to antidepressants (RR 3.50, 95% CI 1.23 to 9.92; P = 0.02; 1 trial, 30 participants; Analysis 12.1 ).

Restlessness

Two trials comparing a medication with another medication reported dichotomous data on restlessness as a non‐serious adverse event. Leone 1982 compared the antipsychotics loxapine and chlorpromazine, while Zanarini 2004 compared the antipsychotic olanzapine to the antidepressant fluoxetine. None of the analyses showed a difference in restlessness between the groups (loxapine versus chlorpromazine: (RR 1.50, 95% CI 0.26 to 8.50; P = 0.65; 1; I ² = 0%; 1 trial, 80 participants; Analysis 12.2 ; and olanzapine versus fluoxetine: RR 0.70, 95% CI 0.23 to 2.11; P = 0.53; 1 trial, 30 participants; Analysis 12.2 ).

Restlessness/anxiety

One trial comparing a medication with another medication reported dichotomous data on restlessness/anxiety as a non‐serious adverse event. Bozzatello 2017 compared two antipsychotics, olanzapine and asenapine. There was no difference in restlessness/anxiety between the two groups (RR 0.19, 95% CI 0.01 to 3.82; P = 0.28; 1 trial, 51 participants; Analysis 12.3 ).

Sleepiness/drowsiness

Two trials comparing a medication with another medication reported dichotomous data on sleepiness/drowsiness as a non‐serious adverse event. Leone 1982 compared the two antipsychotics loxapine and chlorpromazine. There was no difference in sleepiness/drowsiness between the two groups (RR 0.80, 95% CI 0.23 to 2.76; P = 0.72; 1 trial, 80 participants; Analysis 12.4 ). Bozzatello 2017 compared two second‐generation antipsychotics olanzapine and asenapine. There was no difference in sleepiness/drowsiness between the two groups (RR 6.74, 95% CI 0.37 to 124.21; P = 0.20; 1 trial, 51 participants; Analysis 12.4 ).

Fainting spells

One trial comparing a medication with another medication reported dichotomous data on fainting spells as a non‐serious adverse event. Leone 1982 compared the two antipsychotics loxapine and chlorpromazine. There was no difference in fainting spells between the groups (RR 0.14, 95% CI 0.01 to 2.68; P = 0.19; 1 trial, 80 participants; Analysis 12.5 ).

Akathisia

One trial comparing a medication with another medication reported dichotomous data on akathisia as a non‐serious adverse event. Bozzatello 2017 compared the antipsychotics olanzapine and asenapine. There was no difference in akathisia between the two groups (RR 0.19, 95% CI 0.01 to 3.82; P = 0.28; 1 trial, 51 participants; Analysis 12.6 ).

Moderate anxiety

One trial comparing a medication with another medication reported dichotomous data on moderate anxiety as a non‐serious adverse event. Bozzatello 2017 compared the antipsychotics olanzapine and asenapine. There was no difference in moderate anxiety between the two groups (RR 0.32, 95% CI 0.01 to 7.53; P = 0.48; 1 trial, 51 participants; Analysis 12.7 ).

Fatigue

One trial comparing a medication with another medication reported dichotomous data on fatigue as a non‐serious adverse event. Bozzatello 2017 compared the antipsychotics olanzapine and asenapine. There was no difference in fatigue between groups (RR 4.81, 95% CI 0.24 to 95.58; P = 0.30; 1 trial, 51 participants; Analysis 12.8 ).

Cardiovascular and respiratory system

Oral hypoaesthesia

One trial comparing a medication with another medication reported dichotomous data on oral hypoaesthesia as a non‐serious adverse event. Bozzatello 2017 compared the antipsychotics olanzapine and asenapine. There was no difference in oral hypoaesthesia between the two groups (RR 0.32, 95% CI 0.01 to 7.53; P = 0.48; 1 trial, 51 participants; Analysis 13.1 ).

Musculoskeletal system

Muscle spasms

One trial comparing a medication with another medication reported dichotomous data on muscle spasms as a non‐serious adverse event. Leone 1982 compared the two antipsychotics loxapine and chlorpromazine. There was no difference in muscle spasms between the groups (RR 3.00, 95% CI 0.33 to 27.63; P = 0.33; 1 trial, 80 participants; Analysis 14.1 ).

Body weight change

Two trials comparing a medication with another medication reported continuous data on body weight change as a non‐serious adverse event. Both compared an antipsychotic to an antidepressant. Soloff 1993 compared haloperidol to phenelzine sulfate and Zanarini 2004 compared olanzapine to fluoxetine. There was no difference between haloperidol and phenelzine sulfate on body weight change (MD −0.22, 95% CI −0.59 to 0.15; P = 0.25; 1 trial, 64 participants; Analysis 14.2 ). However, there was a difference between olanzapine and fluoxetine on body weight change favouring fluoxetine (MD 2.50, 95% CI 0.72 to 4.28; P = 0.006; 1 trial, 29 participants; Analysis 14.2 ).

Weight gain

One trial comparing a medication with another medication reported dichotomous data on weight gain as a non‐serious adverse event. Bozzatello 2017 compared the antipsychotics olanzapine and asenapine. There was no difference in weight gain between groups (RR 4.81, 95% CI 0.24 to 95.58; P = 0.30; 1 trial, 51 participants; Analysis 14.3 ).

3.1 BPD symptom severity

One trial compared one medication with a combination of medications and reported continuous data on BPD symptom severity. Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was a difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding BPD symptom severity at the end of treatment favouring the mood stabiliser alone (MD 8.48, 95% CI 3.39 to 13.57; P = 0.001; 1 trial, 34 participants; Analysis 15.1 ).

3.2 Self‐harm

One trial compared a pharmacotherapy with a combination of pharmacotherapies and reported continuous data on self‐harm. Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was a difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding self‐harm at the end of treatment favouring the mood stabiliser alone (MD 2.55, 95% CI 0.98 to 4.12; P = 0.001; 1 trial, 34 participants; Analysis 15.2 ).

3.3 Suicide‐related outcomes

One trial compared a pharmacotherapy with a combination of pharmacotherapies and reported continuous data on suicide‐related outcomes. Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was no difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding suicide‐related outcomes at the end of treatment (MD 0.23, 95% CI −0.74 to 1.20; P = 0.64; 1 trial, 34 participants; Analysis 15.3 ).

3.4 Psychosocial functioning

One trial compared a pharmacotherapy with a combination of pharmacotherapies and reported continuous data on psychosocial functioning. Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was no difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding psychosocial functioning at the end of treatment (MD 0.88, 95% CI −6.21 to 7.97; P = 0.81; 1 trial, 34 participants; Analysis 15.4 ).

3.5 Anger

Two trials compared a pharmacotherapy to a combination of pharmacotherapies and reported continuous data on anger.

Zanarini 2004 compared the antipsychotic olanzapine to olanzapine plus the antidepressant fluoxetine combined. There was no difference between the groups regarding impulsivity at the end of treatment (MD 0.46, 95% CI −12.93 to 13.85; P = 0.95; 1 trial, 29 participants; Analysis 15.5 ). Zanarini 2004 also compared fluoxetine to fluoxetine plus olanzapine. There was no difference between the groups at the end of treatment (MD 4.77, 95% CI −9.67 to 19.21; P = 0.52; 1 trial, 26 participants; Analysis 15.5 ).

Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was no difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding anger at end of treatment (MD 0.60, 95% CI −0.82 to 2.02; P = 0.41; 1 trial, 34 participants; Analysis 15.5 ).

3.6 Affective instability

One trial compared a medication with a combination of medications and reported continuous data on affective instability. Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was a difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding affective instability at the end of treatment favouring valproic acid combined with omega‐3 fatty acids (MD 1.72, 95% CI 0.68 to 2.76; P = 0.001; 1 trial, 34 participants; Analysis 15.6 )

3.7 Chronic feelings of emptiness

One trial compared a medication with a combination of medications and reported continuous data on chronic feelings of emptiness. Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was no difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding anger at the end of treatment (MD 0.03, 95% CI −0.97 to 1.03; P = 0.95; 1 trial, 34 participants; Analysis 15.7 ).

3.8 Impulsivity

Two trials compared a medication with a combination of medications and reported continuous data on impulsivity.

One trial ( Zanarini 2004 ) compared the antipsychotic olanzapine to olanzapine plus the antidepressant fluoxetine combined. There was no difference between treatments regarding impulsivity at the end of treatment (MD 0.46, 95% CI −12.93 to 13.85; P = 0.95; 1 trial, 29 participants; Analysis 15.8 ). Zanarini 2004 also compared fluoxetine to fluoxetine plus olanzapine. There was no difference between the groups regarding impulsivity at the end of treatment (MD 4.77, 95% CI −9.67 to 19.21; P = 0.52; 1 trial, 26 participants; Analysis 15.8 ).

Another trial ( Bellino 2014 ), compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was a difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding impulsivity at the end of treatment, favouring valproic acid plus omega‐3 fatty acids combined (MD 12.59, 95% CI 6.11 to 19.07; P = 0.0001; 1 trial, 34 participants; Analysis 15.8 ).

3.9 Interpersonal problems

One trial compared a medication with a combination of medications and reported continuous data on interpersonal problems. Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was no difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding interpersonal problems at the end of treatment (MD 0.47, 95% CI −0.41 to 1.35; P = 0.29; 1 trial, 34 participants; Analysis 15.9 ).

3.10 Abandonment

One trial compared a medication with a combination of medications and reported continuous data on fear of abandonment. Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was no difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding fear of abandonment at the end of treatment (MD 0.01, 95% CI −0.83 to 0.85; P = 0.98; 1 trial, 34 participants; Analysis 15.10 ).

3.11 Identity disturbance

One trial compared a medication with a combination of medications and reported continuous data on identity disturbance. Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was no difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding identity disturbances at the end of treatment (MD 0.70, 95% CI −0.34 to 1.74; P = 0.19; 1 trial, 34 participants; Analysis 15.11 ).

3.12 Dissociation and psychotic‐like symptoms

One trial compared a medication with a combination of medications and reported continuous data on dissociation and psychotic‐like symptoms. Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was no difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding dissociation and psychotic‐like symptoms at the end of treatment (MD 0.03, 95% CI −1.09 to 1.15; P = 0.96; 1 trial, 34 participants; Analysis 15.12 ).

3.13 Depression

Two trials compared pharmacotherapies with combinations of other pharmacotherapies and reported continuous data on depression.

Zanarini 2004 compared the antipsychotic olanzapine to olanzapine plus the antidepressant fluoxetine. There was no difference between olanzapine alone and olanzapine plus fluoxetine regarding depression at the end of treatment (MD −1.78, 95% CI −6.48 to 2.92; P = 0.46; 1 trial, 29 participants; Analysis 15.13 ). Zanarini 2004 also compared fluoxetine to fluoxetine combined with olanzapine. There was no difference between fluoxetine alone and fluoxetine plus olanzapine regarding depression at the end of treatment (MD 3.62, 95% CI −1.36 to 8.60; P = 0.15; 1 trial, 25 participants; Analysis 15.13 ).

Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was no difference between valproic acid alone and valproic acid plus omega‐3 fatty acids regarding depression at end of treatment (MD 1.30, 95% CI 0.00 to 2.60; P = 0.05; 1 trial, 34 participants; Analysis 15.13 ).

3.14 Attrition

Two trials compared a medication with a combination of medications and reported dichotomous data on attrition.

One trial compared the antipsychotic olanzapine to olanzapine plus the antidepressant fluoxetine ( Zanarini 2004 ). There was no difference between olanzapine alone and olanzapine combined with fluoxetine regarding attrition (RR 0.19, 95% CI 0.01 to 3.63; P = 0.27; 1 trial, 31 participants; Analysis 15.14 ). Zanarini 2004 also compared fluoxetine to fluoxetine plus olanzapine. There was no difference between fluoxetine alone and fluoxetine combined with olanzapine regarding attrition (RR 0.54, 95% CI 0.05 to 5.28; P = 0.59; 1 trial, 29 participants; Analysis 15.14 ).

One trial compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined ( Bellino 2014 ). There was no difference between valproic acid alone and valproic acid combined with omega‐3 fatty acids regarding attrition (RR 0.92, 95% CI 0.29 to 2.97; P = 0.89; 1 trial, 43 participants; Analysis 15.14 ).

3.15 Adverse events

No trials comparing pharmacotherapies to a combination of other pharmacotherapies reported data on total serious or non‐serious adverse events; non‐serious adverse events of the cardiovascular and respiratory system, the sensory system or the reproductive system; or any other non‐serious adverse events than those suggested in our analysis plan. Nor did they report data on withdrawal due to adverse events.

Central nervous system

Sedation

One trial comparing a medication with a combination of medications reported dichotomous data on sedation as a non‐serious adverse event. Zanarini 2004 compared the antipsychotic olanzapine to olanzapine plus the antidepressant fluoxetine. There was no difference in sedation between olanzapine alone and olanzapine plus fluoxetine combined (RR 1.61, 95% CI 0.87 to 2.96; P = 0.13; 1 trial, 31 participants; Analysis 16.1 ). Zanarini 2004 also compared fluoxetine to fluoxetine plus olanzapine. There was no difference between fluoxetine compared to fluoxetine plus olanzapine (RR 0.46, 95% CI 0.15 to 1.44; P = 0.18; 1 trial, 29 participants; Analysis 16.1 ).

Gastrointestinal system

Nausea

One trial comparing a medication with a combination of medications reported dichotomous data on nausea as a non‐serious adverse event. Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was no difference between valproic acid alone and valproic acid combined with omega‐3 fatty acids regarding nausea (RR 0.22, 95% CI 0.01 to 4.34; P = 0.32; 1 trial, 34 participants; Analysis 17.1 ).

Dyspepsia

One trial comparing a medication with a combination of medications reported dichotomous data on dyspepsia as a non‐serious adverse event. Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was no difference between valproic acid alone and valproic acid combined with omega‐3 fatty acids regarding dyspepsia (RR 1.13, 95% CI 0.08 to 16.55; P = 0.44; 1 trial, 34 participants; Analysis 17.2 ).

Muscloskeletal system

Akathisia

One trial comparing a medication with a combination of medications reported dichotomous data on akathisia as a non‐serious adverse event. Zanarini 2004 compared the antipsychotic olanzapine to olanzapine plus the antidepressant fluoxetine. There was no difference in akathisia between olanzapine alone and olanzapine plus fluoxetine combined (RR 0.75, 95% CI 0.25 to 2.28; P = 0.61; 1 trial, 31 participants; Analysis 18.1 ). Zanarini 2004 also compared fluoxetine to fluoxetine plus olanzapine. There was no difference in akathisia between fluoxetine alone and the combination of fluoxetine and olanzapine (RR 1.07, 95% CI 0.39 to 2.92; P = 0.89; 1 trial, 29 participants; Analysis 18.1 ).

Body weight change

Two trials comparing a medication with a combination of medications reported continuous data on body weight change as a non‐serious adverse event.

One trial compared the antipsychotic olanzapine to olanzapine plus the antidepressant fluoxetine ( Zanarini 2004 ). There was a difference in body weight change between olanzapine alone and olanzapine plus fluoxetine combined, which favoured the combination of olanzapine and fluoxetine (MD 1.50, 95% CI 0.09 to 2.91; P = 0.04; 1 trial, 29 participants; Analysis 18.3 ). Zanarini 2004 also compared fluoxetine to fluoxetine plus olanzapine. There was no difference between fluoxetine alone and the combination of fluoxetine and olanzapine in body weight change (MD −1.00, 95% CI −2.39 to 0.39; P = 0.16; 1 trial, 26 participants; Analysis 18.3 ).

Bellino 2014 compared the mood stabiliser valproic acid to valproic acid plus omega‐3 fatty acids combined. There was no difference between valproic acid alone and valproic acid plus omega‐3 fatty acids combined for change in body weight (RR 1.13, 95% CI 0.26 to 4.80; P = 0.87; 1 trial, 34 participants; Analysis 18.2 ).

4.1.1 Antipsychotics

4.1.1.1 First‐ versus second‐generation antipsychotics

Subgroup analyses were feasible for the primary outcomes of BPD severity, suicide‐related outcomes, and psychosocial functioning. The intervention effect varied according to subgroups of antipsychotics indicating inferiority of first‐generation antipsychotics (SMD 0.29, 95% CI −0.09 to 0.67; I ² = 0%; 3 trials, 108 participants) compared to second‐generation antipsychotics (SMD ‐0.36, 95% CI ‐0.66 to ‐0.05; I ² = 74%; 5 trials, 820 participants) regarding BPD severity (test for subgroup differences: Chi² = 6.74, df = 1 (P = 0.009), I² = 85.2%, Analysis 20.1 ). There was no indication that intervention effects varied according to subgroups of first‐ and second‐generation antipsychotics neither regarding suicide‐related outcomes (test for subgroup differences: Chi² = 0.24, df = 1 (P = 0.63), I² = 0%; Analysis 20.2 ), nor psychosocial functioning (test for subgroup differences: Chi² = 0.03, df = 1 (P = 0.86), I² = 0% Analysis 20.3 ).

4.1.1.2 Antipsychotics by substance

Subgroup analyses were feasible for the primary outcomes of BPD severity, suicide‐related outcomes, and psychosocial functioning.

For BPD severity, subgroups included one trial each (haloperidol, thiothixene, quetiapine, ziprasidone and brexpiprazole) except for olanzapine (two trials). The subgroup analysis revealed that the intervention effect varied according to type of substance (haloperidol; MD 3.95, 95% CI ‐2.66 to 10.56; 1 trial, 58 participants; olanzapine; SMD ‐0.15, 95% CI ‐0.41 to 0.10; I ² = 60%; 2 trials, 596 participants; thiothixene; MD 1.20, 95% CI ‐1.19 to 3.60; 1 trial, 50 participants; quetiapine; MD ‐0.70, 95% CI ‐2.80 to 1.40; 1 trial, 95 participants; ziprasidone; MD ‐0.42, 95% CI ‐0.87 to 0.03; 1 trial, 60 participants; brexpiprazole; MD ‐5.30, 95% CI ‐7.56 to ‐3.04; 1 trial, 69 participants). The test for subgroup differences indicated: Chi² = 19.72, df = 5 (P = 0.001), I² = 74.6%, Analysis 20.4 .

There was no indication that intervention effects varied according to subgroups of substances, olanzapine, ziprasidone, brexpiprazole and alprazolam, either regarding suicide‐related outcomes at end of treatment (test for subgroup differences: Chi² = 4.30, df = 3 (P = 0.23), I² = 30.2%; Analysis 20.6 ), or between subgroups of haloperidol, olanzapine, quetiapine and thiothixene regarding the outcome of psychosocial functioning (test for subgroup differences: Chi² = 1.05, df = 3 (P = 0.79), I² = 0%; Analysis 20.7 ).

4.1.2 Antidepressants

A subgroup analysis of class and substance of antidepressants was only feasible for the primary outcome of psychosocial functioning. There was no indication that the intervention effect on psychosocial functioning varied according to antidepressant class or substance for the subgroups tricyclic antidepressant (amitriptyline) and selective serotonin reuptake inhibitor (fluoxetine). The test for subgroup differences indicated: Chi² = 0.25, df = 1 (P = 0.61), I² = 0%; Analysis 20.8 .

4.1.3 Mood stabilisers

Data permitted a subgroup analysis only for the primary outcome of BPD severity. There was no indication that intervention effects varied according to mood stabiliser substances divalproex and lamotrigine (test for subgroup differences: Chi² = 0.01, df = 1 (P = 0.91), I² = 0%; Analysis 20.5 ).

5.3.1 Choice of data sources for Black 2014

For one trial, results were available from a journal publication ( Black 2014a ) as well as a trial registry entry ( NCT00880919 ). We decided to use the data from the peer‐reviewed journal publication, whenever possible. However, for one outcome (Zan‐BPD total), data were available from both sources. While those reported in the journal article resulted in a small effect of SMD −0.14 (95% CI −0.58 to 0.29), the data from the trial registry entry would have yielded a very large, clearly outlying large effect of SMD −1.95 (95% CI −2.47 to −1.43). This finding reinforced our decision to primarily use data from peer‐reviewed sources, whenever possible.

5.3.2 Exclusion of outliers

For the secondary outcome of anger, heterogeneity was high (87%) due to several trials reporting outstandingly high effect estimates, as the visual inspection of the forest plot indicated. We decided to exclude the outliers one‐by‐one, until heterogeneity was low. Therefore, primary analyses resulted in an effect estimate of mood stabilisers compared to placebo of SMD −0.67 (95% CI −1.10 to −0.24; I ² = 26%; 5 trials, 135 participants), once Loew 2006 , Nickel 2004 and Tritt 2005 had been excluded. If all outliers were included, the effect estimate would be more than twice as high, with  SMD −1.39 (95% CI −2.16 to −0.61; I ² = 84%; 8 trials, 247 participants; Analysis 26.1 ).

The substantial reduction of high statistical heterogeneity from 84% to 27% by excluding those trials with outstanding, exceptionally high effect estimates and the bisection of the pooled effect supports our decision not to include the outliers in the primary analyses in order to avoid an overestimation of intervention effects. All these trials originate from the same working group and had been discussed for possibly biased results previously ( NICE 2009 , p. 217: "The GDG [guideline development group] [...] took the decision not to consider these trials when drawing up their conclusions."