The trial excluded Cluster B, and obsessive-compulsive personality disorder became half the sample
- Secondary analysis of a randomised controlled trial, pre-registered on the Open Science Framework, published online on 24 November 2025 and carried in the March 2026 issue of Personality Disorders: Theory, Research, and Treatment, volume 17, issue 2, pages 178–188. First and corresponding author Molly Fennig Steinhoff of Washington University in St. Louis, with Rebecca Tillman and Kirsten Gilbert of the same university, Julian Baudinet of the Maudsley, Roelie Hempel of RO DBT Online B.V. and Thomas Lynch of the University of Southampton. Funded by the National Institute of Mental Health (R01 MH126984, F31 MH137979), the National Heart, Lung, and Blood Institute (T32 HL130357) and the Taylor Family Institute.
- The dataset is RefraMED, a three-site randomised trial run in the United Kingdom with recruitment from March 2012 to March 2015 (ISRCTN 85784627), which allocated 250 adults with treatment-refractory depression 15:8 to radically open dialectical behaviour therapy plus treatment as usual, or to treatment as usual alone. Personality disorders were diagnosed at baseline by trained assessors using SCID-I and SCID-II, categorically and by DSM-IV, blind to allocation.
- Eligibility excluded psychosis, bipolar disorder, a primary substance use diagnosis and any Cluster B personality disorder. The resulting Axis II table across all 250 participants: avoidant personality disorder 136 (54.4%), obsessive-compulsive 117 (46.8%), paranoid 57 (22.9%), schizoid 25 (10.2%), dependent 23 (9.3%), schizotypal 22 (8.8%), antisocial 3 (1.7%), borderline 3 (1.2%), histrionic 0, narcissistic 0.
- Among the 117 participants with obsessive-compulsive personality disorder, the treatment was associated at 12 months with better emotional approach coping (estimate 3.037, partial eta-squared 0.099) and lower psychological inflexibility (estimate -5.199, partial eta-squared 0.083), both at FDR p = 0.008. Depressive symptoms did not separate at 12 months (p = 0.912) and nothing survived correction at 7 months.
- Within the treated arm of 162 people, carrying the diagnosis predicted no outcome: corrected p values ran from 0.868 upward at 7 months and from 0.893 upward at 12, with partial eta-squared no higher than 0.004, and two one-sided tests against bounds of d = 0.2 rejected both null hypotheses in every regression. The authors state that RefraMED is the only randomised controlled trial of this therapy to date.
One randomised controlled trial stands behind radically open dialectical behaviour therapy, and few trials have tested treatment for obsessive-compulsive personality disorder, and the authors write that sample sizes and absent comparison conditions preclude conclusions about efficacy. RefraMED allocated 250 adults with treatment-refractory depression across three sites in the United Kingdom, recruiting from March 2012 to March 2015, and set radically open dialectical behaviour therapy added to treatment as usual against treatment as usual alone. A group at Washington University in St. Louis, working on grants from the National Institute of Mental Health and the National Heart, Lung, and Blood Institute, went back into that dataset and asked what had become of the 117 participants who met criteria for obsessive-compulsive personality disorder on the day they entered. The answer to that question is the controlled evidence on treating the disorder. No trial was ever built to produce it.
Who could enter was settled before anyone was randomised. A participant needed a current major depressive episode on the Structured Clinical Interview for DSM-IV, depression running two years or longer or two episodes that had not responded to six weeks of an antidepressant, and a Hamilton Depression Rating Scale score of at least 15. Four groups were turned away at the door: psychosis, bipolar disorder, a primary substance use diagnosis, and a Cluster B personality disorder. Anyone in the radically open arm already receiving standard dialectical behaviour therapy, or waiting for a place in it, was turned away as well. Personality disorders were then diagnosed by trained assessors on SCID-I and SCID-II, blind to allocation, categorically and by DSM-IV rather than by the alternative model or by ICD-11.
Ten diagnoses and two absences
Across all 250 participants the column reads: avoidant personality disorder in 136 people (54.4%), obsessive-compulsive in 117 (46.8%), paranoid in 57 (22.9%), schizoid in 25 (10.2%), dependent in 23 (9.3%), schizotypal in 22 (8.8%), antisocial in 3 (1.7%), borderline in 3 (1.2%), histrionic in none, narcissistic in none. The obsessive-compulsive diagnosis was present in 42 of the 88 people who received treatment as usual (47%) and in 75 of the 162 who received the therapy (46%).
One line of the eligibility criteria produced most of that shape. The disorder that carries this field's trials, its manuals and its trained workforce appears here as 3 people out of 250, and the disorder that the treatment literature has barely tested fills nearly half the sample. Whatever RefraMED could ever report about personality disorders was fixed at screening, and it could be reported only about the diagnoses the screen let through.
What it reports is narrow and specific. Among the 117 participants with obsessive-compulsive personality disorder, the therapy was associated at 12 months with better emotional approach coping, meaning the recognition and communication of emotion (estimate 3.037, partial eta-squared 0.099), and with lower psychological inflexibility (estimate -5.199, partial eta-squared 0.083), both surviving correction at FDR p = 0.008. Depressive symptoms, the parent trial's primary outcome, did not separate at 12 months (p = 0.912), and nothing survived correction at 7 months. Inside the treated arm, carrying the diagnosis predicted no outcome at all, with every corrected p value at 0.868 or above at 7 months and 0.893 or above at 12, and partial eta-squared never exceeding 0.004.
Set that beside what the authors describe as the existing evidence for what they call the second most common personality disorder. Psychodynamic therapy and cognitive behavioural therapy have been examined preliminarily, with sample sizes and absent comparison conditions that the authors say preclude conclusions about efficacy. Cognitive behavioural therapy for perfectionism reaches medium to large effects on perfectionism and on depression in meta-analysis, and its effects on obsessive-compulsive personality disorder were not measured. The evidence is thin because the trials were not run, which is a statement about the trials and not about the disorder.
Cluster B was an exclusion criterion, so borderline personality disorder appears in 3 of the 250 participants and narcissistic personality disorder in none, while obsessive-compulsive personality disorder reaches 117 people, and whatever this trial could ever report about personality disorders was fixed at that screening line.
The design is a secondary analysis of a trial built to test a depression treatment, and every constraint follows from that. Obsessive-compulsive personality disorder was assessed only at baseline, and its severity was never measured at any timepoint, so the paper cannot say whether the diagnosis changed. Everyone enrolled was seeking treatment for treatment-refractory depression, and the authors note it is unclear how the findings extend to the disorder in other contexts or to other personality disorders. The two outcomes that separated are self-report questionnaires of 7 and 8 items, while depression was clinician-rated. Missing data ran to 33.2–39.2% of 12-month values and were handled by multiple imputation across 25 datasets. The sensitivity analysis shows 80% power only for effects at or above partial eta-squared 0.077 in the first model, and the authors report that a simulation doubling the sample at the same effect size turned the non-significant depression result significant. The third aim was exploratory and not pre-registered. The sample was almost entirely White, and the trial ran in the United Kingdom between 2012 and 2015 on DSM-IV criteria. Two internal figures disagree: the text gives 164 women (65.3%) against 163 (65.2%) in the table, and the limitations give 244 White participants (98%) against 232 (97.5%) in the table. Thomas Lynch receives royalties for the RO DBT textbook and manual and is married to a co-owner of the training company Radically Open Ltd; Roelie Hempel is a co-owner of that company.