Trial design

The delivery of schema therapy and SPSP was planned in either 25 weekly sessions or 50 twice-weekly sessions over a period of 9–12 months. We hypothesised that 50 sessions would lead to both a greater reduction in the severity of depression over time and higher remission rates for depression and personality disorder after 1 year. A secondary aim was to examine whether SPSP and schema therapy are feasible and effective treatment modalities for people with co-occurring depression and personality disorder. The study was registered with The Netherlands Trial Register (now International Clinical Trials Registry Platform; registration number NTR5941). Details of the study design have been published elsewhere.^{Reference Kool, Van, Bartak, de Maat, Arntz and van den Eshof12} In summary, the protocol described the study as a single-centre, pragmatic randomised controlled trial with a 2 × 2 factorial design involving 200 patients with a depressive disorder and personality disorder(s), recruited from a Dutch mental healthcare centre for personality disorders. Participants would be randomised by therapy dosage (25 v. 50 sessions in a year) and type of therapy (schema therapy v. SPSP). The primary clinical outcome measures were depression severity and remission.

The protocol was followed but with some violations. Owing to limited therapist availability the allocation ratio (1:1:1:1) was temporarily changed to 1:3 for SPSP:schema therapy and later reversed to 3:1. In this paper we report on data collected from baseline up to the end of treatment. Two other minor corrections made to the protocol article, concerning type of randomisation and sample size, are described elsewhere.^{Reference Kool, Van, Bartak, de Maat, Arntz and van den Eshof13}

Participants

Participants were recruited from regular referrals to a specialised centre for personality disorders in Amsterdam, The Netherlands. They were adult out-patients who met the diagnostic criteria for DSM-IV depression or dysthymia and had one or more personality disorders, including non-specified or other groups (PD-NOS; OSPD) according to DSM-IV and DSM-5, based on a minimum of five personality disorder traits (DSM-5 was introduced in The Netherlands during the trial). Patients were excluded if they had psychotic symptoms, a bipolar disorder, inadequate mastery of the Dutch language or an indication for immediate hospital admission or intensive treatment, such as acute suicidality. Individuals with a history of addiction were excluded in case of current alcohol or substance dependence (including benzodiazepines). Individuals without a history of addiction were excluded if the intake clinician decided that the current addiction needed treatment before (or in parallel with) depression/personality disorder treatment. In addition, the treatment centre excluded individuals with a main personality disorder diagnosis in cluster A or antisocial personality disorder. In line with the pragmatic nature of the trial the use of antidepressants was allowed, as were changes in prescription during the course of treatment.

Procedure

Participants were recruited through regular assessment procedures. Once informed about the trial, they had at least 48 h to consider participation, after which they were further assessed for eligibility using a semi-structured interview for depression (section A for Depression and section B for Dysthymia in the Mini-International Neuropsychiatric Interview-Plus; MINI-Plus).^{Reference Sheehan, Lecrubier, Harnett-Sheehan, Janavs, Weiller and Bonora14} As part of the regular assessment procedure individuals were screened for personality disorder (SCID-PQ/SCID-5-SPQ) and those with scores at or above the cut-off minus 1 on the personality disorder sections were further assessed with the Structured Clinical Interview for DSM-IV Axis II Personality Disorders (SCID-II) or DSM-5 Personality Disorders (SCID-5-PD).^{Reference First, Williams, Benjamin and Spitzer15–Reference First, Gibbon, Spitzer, Williams and Benjamin18} The interviews were conducted by research assistants with a bachelor's degree in psychology and additional training for the semi-structured interviews. Interrater agreement for the SCID interviews was excellent (intraclass correlation coefficient ICC = 0.76 for trait scores by personality disorder and ICC = 0.86 for sum scores by personality disorder based on 27 double-rated interviews).

Randomisation took place after the diagnostic assessment. To ensure allocation concealment, randomisation codes were generated by an independent researcher, not working at one of the clinical departments. Research assistants assessing outcomes were masked for trial condition. Masking was successful: correctly guessed treatment allocations did not differ significantly from random guessing (30 and 25% respectively, P = 0.294). Participants were stratified according to depression severity assessed using the Beck Depression Inventory-II (low: BDI-II ≤ 29; high: BDI-II ≥ 30) and randomly assigned to one of four groups (25 or 50 sessions; schema therapy or SPSP) with random allocation sequences that were generated using the SPSS random number generator (SPSS, Chicago). All participants signed informed consent forms. A difference of 0.45 in the post-treatment effect size for depressive symptoms was estimated as clinically meaningful.^{Reference Cuijpers, Huibers, Ebert, Koole and Andersson8} Using this difference, a minimum of 211 participants were needed for adequate power, taking 25% drop-out into account (α = 0.05, power (1 − ß) = 0.80, two-tailed). (Owing to a miscalculation, the required minimum that was reported in the protocol paper was 200 participants. This was rectified in consultation with, and with the approval of, the medical ethics committee. The appropriate amendment was made in the trial register and a correction to the protocol article was published^{Reference Kool, Van, Bartak, de Maat, Arntz and van den Eshof13}).

Interventions

Both treatment modalities consider depression and enduring personality patterns to be interwoven and both aim to alleviate depression and the underlying susceptibility to depression. However, they differ in some key elements: SPSP is a psychodynamic psychotherapy, originally proven effective for treating depression, with a concurrent positive effect on personality measures.^{Reference de Jonghe, de Maat, Hendriksen, Nooteboom, Dekker and Van19,Reference Kool, Dekker, Duijsens, de Jonghe and Puite20} Schema therapy is an integrative approach that combines elements from cognitive–behavioural therapy, attachment and object relation theories, Gestalt therapy and experiential therapies.^{Reference Young, Klosko and Weishaar21} It has proven effective in treating personality disorder and has yielded promising results for the treatment of (persistent) depression.^{Reference Carter, McIntosh, Jordan, Porter, Frampton and Joyce22,Reference Malogiannis, Arntz, Spyropoulou, Tsartsara, Aggeli and Karveli23} The manuals for SPSP and schema therapy for chronic depression were used and the interventions were provided in either 25 or 50 sessions.^{Reference de Jonghe, de Maat, Hendriksen, Nooteboom, Dekker and Van19,Reference Renner, Arntz, Leeuw and Huibers24} Participants in the 25-session condition received 16 weekly sessions, followed by 9 sessions every 2 weeks. Participants in the 50-session condition received 32 twice-weekly sessions, followed by 18 weekly sessions. Duration of therapy was at least 8 months but owing to holidays and sick days on both sides a duration of 9–12 months was expected in all conditions.

Treatment integrity

Therapists were psychiatrists or psychologists with at least a post-Masters degree in psychology, registered as SPSP or schema therapy therapists or who had at least completed a basic course in SPSP (3 days) or schema therapy (25 h), had 6 months of experience in the given form of treatment and had attended additional biweekly supervision sessions led by a registered supervisor for SPSP or schema therapy. All schema therapists received an additional 1-day course in schema therapy for depression. To improve adherence to the model, therapists were allowed to treat patients in only one treatment modality. To reduce potential therapist bias they were assigned to both dosage conditions.

Adherence (i.e. whether the protocol was followed) and competence (i.e. how well the therapy was performed) were rated by four masked experts (psychologists with a post-master's degree in psychology and 12–16 years of clinical experience), who scored a total of 160 audiotapes of sessions with 80 randomly selected participants. Further details on the adherence and competence scales and procedures are given in Data supplement 1, available at https://doi.org/10.1192/bjp.2024.56. In addition, adherence and competence were checked and enhanced in biweekly peer supervision at which current cases and audiotaped material were discussed. Issues that were not solved in supervision were discussed during biweekly meetings with research staff and supervisors on the trial's advisory committee.

Primary outcomes

The primary outcome was the change in depressive symptoms over time on the Dutch version of the revised Beck Depression Inventory-II (BDI-II-NL-R).^{Reference Beck, Steer and Brown25} The BDI-II assessments were made at the start of treatment, at 1, 2, 3 and 6 months, and at the end of treatment (9–12 months).

Secondary outcomes

Secondary outcomes included remission of depression (MINI-Plus) and remission of personality disorder (during the past 6 months, assessed using the SCID-II/SCID-5-PD) measured at the end of treatment.^{Reference Sheehan, Lecrubier, Harnett-Sheehan, Janavs, Weiller and Bonora14,Reference First, Gibbon, Spitzer, Williams and Benjamin16,Reference First, Williams, Benjamin and Spitzer17} Secondary personality outcomes were improvement over time in terms of psychodynamic and schema therapy constructs assessed using the Severity Indices of Personality Problems (SIPP), the Developmental Profile Inventory (DPI), the Young Schema Questionnaire-Short Form (YSQ-SF) and the Schema Mode Inventory (SMI) planned at inclusion (SIPP), baseline (YSQ-SF, SMI), 6 months (SIPP, DPI, YSQ-SF, SMI) and at the end of treatment (SIPP, DPI, YSQ-SF, SMI).^{Reference Verheul, Andrea, Berghout, Dolan, Busschbach and van der Kroft26–Reference Young, Arntz, Atkinson, Lobbestael, Weishaar and Vreeswijk29} Assessments of the reduction of general psychological symptoms (using the Brief Symptom Inventory (BSI) and the Outcome Questionnaire's Symptom Distress subscale (OQ-SD)) and improvement in quality of life (EuroQol 5-Dimension, EQ-5D) were planned at inclusion (BSI, OQ-SD), baseline (BSI, EQ-5D), 3 months (BSI, OQ-SD, EQ-5D), 6 months (BSI, OQ-SD, EQ-5D) and at the end of treatment (BSI, OQ-SD, EQ-5D).^{Reference Derogatis and Melisaratos30–Reference Brooks, Rabin and De Charro32} The happiness item was included in every measurement.^{Reference Veenhoven33} For a detailed description of these instruments and psychometric properties, we refer to our protocol paper.^{Reference Kool, Van, Bartak, de Maat, Arntz and van den Eshof12}

Data analysis

An outline of the analysis strategy is provided in the published protocol.^{Reference Kool, Van, Bartak, de Maat, Arntz and van den Eshof12} Data Supplement 1 describes the analysis of treatment integrity. Differences in number of sessions, session frequency and therapy duration were tested with Mann–Whitney tests. Primary analyses used intention to treat. To investigate the effect of psychotherapy dosage on depression (BDI-II) multilevel regression analyses with restricted maximum likelihood estimation were conducted. Interventions were represented by two dichotomous variables: 25 (0) versus 50 (1) sessions and SPSP (0) versus schema therapy (1). The initial basic model was a two-level model with repeated measurements (level 1) nested within participants (level 2) with two two-way interactions testing the difference in the change in BDI-II scores over time (in days) for the different psychotherapy dosages (time × dosage) and treatments (time × treatment). When available (BDI, SIPP, OQ-SD), we used the inclusion measurement as a covariate.

The addition of random levels for therapist was tested, as well as the addition of random slopes for time. Time in days was used as the time variable. Various covariance structures for the repeated part were tested (SPSS options AR1, ARMA11, CS, ARH1, CSH). Potential additional quadratic functions of time were compared with the linear model for the best model fit. The additional quadratic parameters were tested in the following sequence: first time-squared, then time-squared and time-squared × dosage, and then time-squared and time-squared × treatment (eventually combined with time-squared × dosage if that led to significantly better fit in the previous step). To explore whether the effect of psychotherapy dosage differed between SPSP and schema therapy, the addition of a three-way interaction (dosage × treatment × time) was tested separately. To investigate differences in remission from depression/dysthymia (MINI-Plus), the groups were analysed with χ² tests.

Estimated marginal means were derived from the linear mixed models for all continuous measures. Within-group effect sizes from start to 1 year (Cohen's d) were computed: (estimated mean at start of treatment − estimated mean at 1 year)/(observed s.d. at start of treatment). Between-group effect sizes at 1 year were derived from linear mixed model analysis. For the SIPP and OQ-SD, measurement at start of treatment was not available to calculate the effect sizes, and we therefore used the observed measures at inclusion as the reference point and the standard deviation of the inclusion measurement as the denominator. For the BDI we also reported the effect sizes after 308 days, which was the mean treatment duration. Reliable change (response) was defined as a minimum decrease of 9 BDI-II points, based on Jacobson & Truax.^{Reference Jacobson and Truax34} Response differences between treatment groups were tested using generalised estimating equations (GEE). Differences in remission from personality disorder between the groups were analysed using χ². A sensitivity analysis was performed with completers only (>72% session attendance). Differences in drop-out rates were analysed using χ². Differences in time until drop-out were tested using Kaplan–Meier survival analysis with log rank tests. This analysis was done for (a) completers + participants who did not start the intervention + those who dropped out during treatment and for (b) completers + those who dropped out during treatment. Significance levels were set at P < 0.05. All statistical analyses were performed in SPSS Statistics 27.0 for Windows. The results are reported in line with CONSORT guidelines.^{Reference Jainer and Onalaja35}

Ethics statement

The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. All procedures involving patients were approved by the Medical Ethical Committee of Vrije Universiteit Amsterdam (registration number NL55916.029.15).