Background
Borderline personality disorder (BPD) is a serious psychiatric condition characterized by high rates of self-harm and suicide risk (Hawton, Zahl, & Weatherall, Reference Hawton, Zahl and Weatherall2003; Paris & Zweig-Frank, Reference Paris and Zweig-Frank2001). Intentional, repetitive self-harm, with or without an intent to die, is a priority target in BPD treatments, including dialectical behavior therapy (DBT). Recent findings from a randomized-controlled trial (RCT) of 240 participants with BPD demonstrated that a brief, 6-month course of DBT for BPD is non-inferior to the standard 12 months of treatment for treating repetitive self-harm and other outcomes (McMain et al., Reference McMain, Chapman, Kuo, Dixon-Gordon, Guimond, Labrish and Streiner2022). Significant improvements in self-harm occurred in both treatment arms by 6 months and were sustained at 24-month follow-up (McMain et al., Reference McMain, Chapman, Kuo, Dixon-Gordon, Guimond, Labrish and Streiner2022). These results converge with a growing body of evidence which suggests that briefer forms of DBT for BPD are effective on a range of outcomes including self-harm (Keng et al., Reference Keng, Mohd Salleh Sahimi, Chan, Woon, Eu, Sim and Wong2021; Westad, Hagen, Jonsbu, & Solem, Reference Westad, Hagen, Jonsbu and Solem2021). By extension, lengthier treatments may not be necessary for all individuals with BPD (Griffiths et al., Reference Griffiths, Duffy, Duffy, Brown, Hockaday, Eliasson and Schwannauer2019).
However, there is considerable heterogeneity in BPD treatment response; approximately half of BPD patients are considered non-responders following DBT and other BPD interventions (Woodbridge, Townsend, Reis, Singh, & Grenyer, Reference Woodbridge, Townsend, Reis, Singh and Grenyer2022). Furthermore, it remains unknown which BPD patients are likely to improve in briefer (e.g. 6 months) v. the more standard 12 months of DBT. To tailor the length of treatment for different patients, research is needed to identify patient characteristics that moderate treatment outcomes.
Historically, research on moderators of self-harm outcomes in BPD patients has employed self-report questionnaires to measure symptom-level moderators, and findings are mixed. On one hand, some findings suggest that greater symptom severity at pre-treatment is associated with better self-harm outcomes in DBT and other interventions for self-harm (Adrian et al., Reference Adrian, McCauley, Berk, Asarnow, Korslund, Avina and Linehan2019; Andover, Schatten, Holman, & Miller, Reference Andover, Schatten, Holman and Miller2020; Barnicot et al., Reference Barnicot, Katsakou, Bhatti, Savill, Fearns and Priebe2012; Sahlin et al., Reference Sahlin, Bjureberg, Gratz, Tull, Hedman-Lagerlöf, Bjärehed and Ljótsson2019). Although, Bateman and Fonagy (Reference Bateman and Fonagy2013) found no association between baseline BPD symptom severity and self-harm outcomes in mentalization-based therapy (MBT). Finally, McMain et al. (Reference McMain, Fitzpatrick, Boritz, Barnhart, Links and Streiner2018) examined associations between baseline severity markers and trajectories of self-harm outcomes in 180 BPD patients treated with DBT or general psychiatric management (GPM) and identified distinct trajectories. Specifically, patients with the lowest frequency of baseline self-harm showed a ‘rapid and recovered’ self-harm trajectory, whereas a second class of patients had slow yet sustained improvements in self-harm. Moreover, a third class with rapid but unsustained improvements, had the highest levels of baseline self-harm.
Taken together, the findings are inconclusive and the field may benefit from examining other types of moderators (e.g. cognitive moderators). Indeed, BPD is characterized by a range of cognitive deficits in the domains of impulsivity and attentional control (Albert et al., Reference Albert, Lopez-Martin, Arza, Palomares, Hoyos, Carretie and Carrasco2019; Koudys & Ruocco, Reference Koudys and Ruocco2021; Thomsen, Ruocco, Carcone, Mathiesen, & Simonsen, Reference Thomsen, Ruocco, Carcone, Mathiesen and Simonsen2017a). Deficits in inhibitory control have also been associated with intentional self-harm in BPD (Drabble, Bowles, & Barker, Reference Drabble, Bowles and Barker2014; LeGris, Links, van Reekum, Tannock, & Toplak, Reference LeGris, Links, van Reekum, Tannock and Toplak2012).
Impulsivity and related cognitive dimensions can be assessed using performance-based tasks (e.g. continuous performance tests). Factor analytic findings provide strong evidence that performance-based measures assess unique dimensions (e.g. action impulsivity), relative to self-report measures of those constructs (e.g. trait impulsivity) (Cyders & Coskunpinar, Reference Cyders and Coskunpinar2011; Stahl et al., Reference Stahl, Voss, Schmitz, Nuszbaum, Tüscher, Lieb and Klauer2014). Performance-based measures therefore have the potential to assess different markers of functioning that may moderate self-harm outcomes in BPD.
In particular, continuous performance tests (CPTs; e.g. Conners, Reference Conners2014) that require selective responses to target and non-target stimuli are valid measures of inhibitory and attentional control. Findings from CPT studies generally show impaired inhibitory control in BPD, as observed by elevated commission errors in adults with BPD or BPD traits (e.g. Feliu-Soler et al., Reference Feliu-Soler, Soler, Elices, Pascual, Perez, Martin-Blanco and Portella2013; Soler et al., Reference Soler, Valdeperez, Feliu-Soler, Pascual, Portella, Martin-Blanco and Perez2012; van Dijk et al., Reference van Dijk, Schellekens, van den Broek, Kan, Verkes and Buitelaar2014), although one study found no between-group differences in commission errors (Ferraz et al., Reference Ferraz, Vallez, Navarro, Gelabert, Martin-Santos and Subira2009). Relatedly, many studies using variants of CPTs, such as stop signal tasks, have observed impairments in inhibitory control in BPD (Albert et al., Reference Albert, Lopez-Martin, Arza, Palomares, Hoyos, Carretie and Carrasco2019; McCloskey et al., Reference McCloskey, New, Siever, Goodman, Koenigsberg, Flory and Coccaro2009; Rentrop et al., Reference Rentrop, Backenstrass, Jaentsch, Kaiser, Roth, Unger and Renneberg2008; Rubio et al., Reference Rubio, Jiménez, Rodríguez-Jiménez, Martínez, Iribarren, Jiménez-Arriero and Ávila2007; van Dijk et al., Reference van Dijk, Schellekens, van den Broek, Kan, Verkes and Buitelaar2014), although many have found no impairment (Barker et al., Reference Barker, Romaniuk, Cardinal, Pope, Nicol and Hall2015; Dinn et al., Reference Dinn, Harris, Aycicegi, Greene, Kirkley and Reilly2004; Hagenhoff et al., Reference Hagenhoff, Franzen, Koppe, Baer, Scheibel, Sammer and Lis2013; Jacob et al., Reference Jacob, Gutz, Bader, Lieb, Tüscher and Stahl2010; Lampe et al., Reference Lampe, Konrad, Kroener, Fast, Kunert and Herpertz2007). These mixed findings may reflect heterogeneity in cognitive functioning in BPD.
To date, only 2 BPD treatment studies have included performance-based cognitive measures. One study found improvements in sustained attention and perceptual reasoning in 18 patients with BPD who completed MBT, but not in untreated controls (Thomsen, Ruocco, Uliaszek, Mathiesen, & Simonsen, Reference Thomsen, Ruocco, Uliaszek, Mathiesen and Simonsen2017b). Another study found improvements in inhibitory control, hit reaction time, and detectability on the Conners Continuous Performance Test-II in 40 BPD patients who completed GPM plus DBT mindfulness skills training, but not in 19 patients who received GPM alone (Soler et al., Reference Soler, Valdeperez, Feliu-Soler, Pascual, Portella, Martin-Blanco and Perez2012). These findings suggest that, among BPD patients, cognition might improve with treatment, although the samples in these studies were small and require replication in larger samples.
Importantly, impulsivity-related deficits in inhibitory control have been associated with the frequency and lethality of intentional self-harm in BPD (Williams et al., Reference Williams, Daros, Graves, McMain, Links and Ruocco2015). Similarly, pre-treatment brain activation in inhibitory control-related frontal regions has been prospectively associated with self-harm improvements in DBT for BPD (Ruocco et al., Reference Ruocco, Rodrigo, McMain, Page-Gould, Ayaz and Links2016). Indeed, biomarker correlates of psychotherapy improvements in BPD converge on adaptive changes in inhibitory and attentional control-related brain function as potential mechanisms of change (Marceau, Meuldijk, Townsend, Solowij, & Grenyer, Reference Marceau, Meuldijk, Townsend, Solowij and Grenyer2018). BPD treatment non-completion has also been linked to lower baseline attentional control (Fertuck et al., Reference Fertuck, Keilp, Song, Morris, Wilson, Brodsky and Stanley2012).
Taken together, it is plausible that performance-based measures of impulsivity and related cognitive dimensions could provide unique prognostic information about individuals with BPD undergoing DBT. This is an understudied area with the potential to illuminate unidentified cognitive risk factors for unfavorable self-harm outcomes. However, no published study to our knowledge has examined cognition as a predictor of self-harm outcomes in BPD. Therefore, in this study, we examined whether performance on a well-established measure of inhibitory and attentional control (the CPT-3; Conners, Reference Conners2014) predicted self-harm outcomes among BPD patients receiving comprehensive DBT. Relatedly, another important area investigated was whether cognitive performance differentially predicted self-harm outcomes in briefer or more standard length of treatment. In the present research, we wished to examine whether CPT-3 performance could identify patient profiles that are most suited to different lengths of DBT.
Study aims and hypotheses
The current study involved a secondary analysis of cognitive performance and self-harm data from the FASTER study (The Feasibility of a Shorter Treatment and Evaluating Responses; NCT02387736), a large multi-site RCT comparing the effectiveness of a brief, 6-month course of DBT for BPD (DBT-6) to a standard 12-month course of treatment (DBT-12). This secondary study had two primary aims. Aim 1 was to characterize changes in inhibitory control and attention across DBT-6 v. DBT-12. We hypothesized that improvements in cognition would be observed in both treatment arms by 6 months of treatment, but we expected that DBT-12 participants would continue to improve up to 12 months (i.e. treatment end point), whereas DBT-6 participants would display no further improvements after 6 months. Aim 2 was to examine whether cognition at pre-treatment could predict self-harm outcomes in DBT-6 v. DBT-12. Given that a certain level of inhibitory and attentional control functioning may be required to learn the novel and sometimes complex information in DBT (e.g. mindfulness skills training), we hypothesized that lower cognitive performance at pre-treatment would be associated with better self-injury outcomes in DBT-12 v. DBT-6 (i.e. reflecting that the most cognitively impaired patients at baseline would require a longer treatment to show self-injury improvements). In an exploratory manner, we also investigated the extent to which a performance-based measure of inhibitory and attentional control could provide prospective information about self-harm outcomes, compared to a measure of self-reported impulsivity.
Methods
Participants and study design
Data from the full FASTER study sample of 240 participants were analyzed. The trial was completed at the Centre for Addiction and Mental Health (CAMH) and Simon Fraser University (SFU) in Canada. The protocol was approved by the CAMH and SFU ethics boards and all participants provided written informed consent prior to participation. A description of the main study methods and results can be found in the original report (McMain et al., Reference McMain, Chapman, Kuo, Dixon-Gordon, Guimond, Labrish and Streiner2022). Demographics for all participants randomized are presented in Table 1. Participants were randomized to DBT-6 (n = 120, 79% female, mean age = 28.3) or DBT-12 (n = 120, 79% female, mean age = 27.3). Participants randomized to both treatment arms received comprehensive DBT, including weekly individual therapy, two weekly DBT skills groups, and on-call phone coaching with a DBT therapist. According to the standard, a participant was considered to have dropped out of treatment if they missed four consecutive individual or group sessions. All study therapists attended weekly DBT team consultations. Participants were followed from pre-treatment to 2-year follow-up, and completed follow-up assessments every 3 months, including assessments of self-harm and self-reported impulsivity. A performance-based measure of inhibitory control and attention was administered at pretreatment, after 6 months of treatment, and at 12 months.
Table 1. Participant demographics and clinical characteristics for all participants randomized
Eligibility criteria
Eligibility assessments were completed by graduate- and postdoctoral-level assessors under the supervision of a licensed psychologist. Individuals who met the following criteria were eligible to participate: 18–60 years; a diagnosis of BPD according to the Diagnostic and Statistical Manual of Mental Disorder- Fourth Edition (DSM-IV; American Psychiatric Association, 1994); engaged in at least two incidents of intentional self-harm in the 5 years prior to study enrolment, including at least one incident in the 2 months preceding enrolment; proficient in English; and valid provincial health insurance. Individuals with the following exclusion criteria were not eligible: meeting DSM-IV criteria for bipolar I disorder, dementia, or a psychotic disorder; IQ ⩽ 70; a serious physical health problem that required hospitalization over the next year; received ⩾8 weeks of DBT in the preceding year; plans to move out of province during the trial.
Clinical and neurocognitive measures
Psychometric properties of measures are in Supplementary Material.
The International Personality Disorders Examination – BPD Section (IPDE-BPD; Loranger et al., Reference Loranger, Sartorius, Andreoli, Berger, Buchheim, Channabasavanna and Regier1994) was used to assess BPD. The Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-1-IV; First, Gibbon, & Williams, Reference First, Gibbon and Williams2002) and Axis II Disorders (SCID-II-IV; First, Gibbon, Williams, & Benjamin, Reference First, Gibbon, Williams and Benjamin1997) were used to assess concurrent disorders.
Consistent with the main outcome paper (McMain et al., Reference McMain, Chapman, Kuo, Dixon-Gordon, Guimond, Labrish and Streiner2022), self-harm was queried with the Suicide Attempt Self-Injury Interview (SASII; Linehan, Comtois, Brown, Heard, & Wagner, Reference Linehan, Comtois, Brown, Heard and Wagner2006), a clinician-administered measure to assess the total aggregate incidents of intentional self-harm, with or without intent to die, across the 3 months preceding each assessment.
Self-reported impulsivity was assessed with the Barratt Impulsiveness Scale (BIS-11; Patton, Stanford, & Barratt, Reference Patton, Stanford and Barratt1995). The BIS-11 contains 30 items rated on a scale from 0 (‘rarely/never’) to 4 (‘almost always/always’) and produces a total score out of 120 representing the severity of impulsive behaviors. A score of 72 suggests high impulsivity, whereas a score between 51 and 71 reflects impulsivity within a normal range (Stanford et al., Reference Stanford, Mathias, Dougherty, Lake, Anderson and Patton2009). Scores below 51 suggest an invalid response profile (Helfritz et al., Reference Helfritz, Stanford, Greve, Villemarette-Pittman, Houston and Conklin2006) or an over-controlled personality presentation. In the current study, pre-treatment BIS-11 scores demonstrated good internal consistency (Cronbach's alpha = 0.83).
Performance-based measures of inhibitory control and attention were assessed with the Conners Continuous Performance Test – 3rd Edition (CPT-3; Conners, Reference Conners2014), a standardized, computer-administered task that requires selective responding to target and non-target stimuli. Participants were presented with letters of the alphabet and instructed to press a space bar when they saw any letter appear on the screen except the letter ‘X,’ in which case they were instructed to withhold their response. The speed of stimulus presentation and the interstimulus interval varied across the 10-min task, with non-target stimuli presented approximately 10% of the time. A description of each CPT-3 outcome variable is provided below. T-scores are based on CPT-3 normative data derived from the scores of individuals of the same age and gender of the participant. A T-score <60 indicates performance in the normal range, whereas a T score ⩾60 reflects clinically elevated (i.e. impaired) performance.
Inhibitory control
Commissions: T-scores for total mean commissions (i.e. ‘false-alarms’, or pressing the space bar for non-target stimuli) reflects inhibitory control performance. Higher T-scores suggest greater difficulties inhibiting a prepotent response.
Impulsivity index: A separate ‘impulsivity index’ raw score was calculated using an equation from Soler et al. (Reference Soler, Valdeperez, Feliu-Soler, Pascual, Portella, Martin-Blanco and Perez2012). Compared to commissions, the impulsivity index takes into account hit reaction time, total commission errors, and total omission errors: [1/hit reaction time] * [omissions/commissions]. As the score is not a standardized CPT-3 scale, a normatively referenced T-score is not available. Higher scores indicate higher impulsivity.
Attention
Detectability (d′): d′ T-scores represent the ability to discriminate between target and non-target stimuli. Higher scores indicate more difficulty differentiating targets from non-targets.
Attentional consistency (hit reaction time – standard deviation): T-scores for hit reaction time – standard deviation (HRT-SD) measure the consistency of response speed to target stimuli across the task. Higher scores reflect greater inconsistency and less task engagement.
Vigilance (hit reaction time interstimulus interval change): T-scores for hit reaction time – interstimulus interval change (HRT-ISI) reflect the slope of change in HRT across the three ISIs in the CPT-3 task (i.e. 1, 2, and 4 s). A positive slope reflects decelerating HRTs at longer ISIs, whereas a negative slope reflects accelerating HRTs at longer ISIs. Higher scores represent a loss of vigilance in the context of changing task demands.
Analyses
An intent-to-treat analysis was used and included data from all 240 participants in the trial, including dropouts. Missing data increased over time as participants were lost to follow-up (online Supplementary Table S1). The results of analyses to explore patterns of missingness supported the assumption that data were missing at random (online Supplementary Material). As such, pairwise deletion was used to handle missing data.
Multilevel modeling: does cognitive performance and self-reported impulsivity improve in DBT-6 and DBT-12?
Multilevel models were created for each CPT-3 outcome variable using the ‘nlme’ package in R to investigate Aim 1, evaluating if cognitive performance and self-reported impulsivity change over treatment, and whether changes depend on treatment arm. A time by treatment arm interaction term was included in all models, and no significant interactions were found (p > 0.05). As such, all analyses modeled CPT-3 performance as a function of time plus treatment arm but did not include an interaction term. A Bonferroni statistical correction was applied, resulting in a significance threshold of p < 0.003. Post-hoc pairwise comparisons were examined for all participants collapsed across treatment arms. A family-wise statistical correction was applied to each post-hoc test, resulting in a significance threshold of p < 0.017. Details pertaining to the construction of each model and the calculation of statistical correction thresholds are in Supplementary Material.
Generalized estimating equations: Is pre-treatment neurocognition associated with self-harm outcomes in DBT-6 and DBT-12?
Generalized estimating equations (GEEs) were created using the ‘geeglm’ package in R to investigate Aim 2, assessing whether CPT-3 performance at pretreatment was associated with the frequency of total self-harm across treatment and up to 2-year follow-up, and if associations depended on treatment arm.
To examine clinically meaningful cognitive predictors, the T-score for each CPT-3 variable was used in the GEE analyses. In each GEE model, CPT-3 T-scores were binarized according to standardized CPT-3 norms. Accordingly, a T-score <60 represents baseline cognitive performance in the normal range and T ⩾ 60 represents performance in a likely clinically impaired range (Conners, Reference Conners2014). Regarding T-scores for commission errors, 161 participants fell in the normal range, and 59 participants in the clinically impaired range. For d′ Tscores, 172 participants were in the normal range and 48 were in the clinically impaired range. Due to a restricted range of T-scores for the HRT-SD and HRT-ISI variables (most participants exhibited baseline performance in the normal range), these predictors were not examined. As normative data are not available for the impulsivity index, this variable was binarized according to the 50th percentile of the sample scores, resulting in 110 participants in the low impulsivity index group and 110 participants in the high impulsivity index group. Finally, pre-treatment BIS-11 scores were binarized according to published guidelines, whereby a score of ⩾72 suggested elevated self-reported impulsivity, and a score of 52–71 indicated self-reported impulsivity in the average range (Stanford et al., Reference Stanford, Mathias, Dougherty, Lake, Anderson and Patton2009).
An ANOVA of each GEE model examined the three-way interaction between pre-treatment cognitive performance, time, and treatment arm. A Bonferroni statistical correction was applied to all ANOVAs, resulting in a significance threshold of p < 0.013. Post-hoc pairwise comparisons explored differences in self-harm between DBT-6 and DBT-12 participants at each time point, using the Tukey's HSD adjustment for multiple comparisons (Tukey, Reference Tukey1953), which controls for the experiment-wise error rate, α EW (Maxwell & Delaney, Reference Maxwell and Delaney2004). Details pertaining to the construction of GEE models are in Supplementary Material.
Results
Descriptive statistics and pre-treatment associations
Table 2 presents descriptive statistics for pre-treatment cognitive performance. There were no significant between-arm differences. A Pearson R correlation matrix of CPT-3 scores across assessment timepoints is in Supplementary Material (online Supplementary Table S2).
Table 2. Descriptive statistics for pre-treatment neurocognition
Aim 1: Does cognitive performance change during DBT, and do changes depend on treatment arm?
Inhibitory control
Commissions. A significant reduction in commissions was observed at the 6-month (β = −2.73 [95% CI −3.92 to −1.54], s.e. = 0.61, t[315] = −4.49, p < 0.001) and 12-month time points (β = −3.13 [95% CI −4.73 to −1.88], s.e. = 0.64, t[315] = −4.92, p < 0.001). There was no effect of treatment arm on reductions in commissions (β = 1.50 [95% CI −1.02 to 4.02], s.e. = 1.28, t[235] = 1.17, p = 0.24). Figure 1 displays pairwise comparisons collapsed across treatment arms showing a significant improvement in commissions at 6 months (β = −2.73, s.e. = 0.61, z = −4.51, p < 0.001) and 12 months (β = −3.13, s.e. = 0.63, z = −4.94, p < 0.001) relative to baseline. No improvement was observed between 6 and 12 months (β = −0.39, s.e. = 0.64, z = −0.61, p = 0.81).
Figure 1. Neurocognitive performance across treatment, collapsed across arms for commissions (left top panel), impulsivity (middle top panel), d′ (right top panel), and self-reported impulsivity (bottom right panel). ***p < 0.001. Asterisks indicate significance at the corrected alpha threshold.
Impulsivity index. Reductions in impulsivity index scores were observed at the 6-month (β = −0.0008 [95% CI −0.0016 to 0.0005], s.e. = 0.0004, t[315] = −1.95, p = 0.05) and 12-month time points (β = −0.001 [95% CI −0.002 to −0.0009], s.e. = 0.0005, t[315] = −2.04, p = 0.04), which were not significant at the corrected threshold (p > 0.003; Figure 1). There was no effect of treatment arm on improvements in the impulsivity index (β = 0.001 [95% CI −0.0005 to 0.04], s.e. = 0.0009, t[235] = 1.30, p = 0.20).
Attention
Detectability. A significant improvement in d′ was observed at the 6-month (β = −3.25 [95% CI −4.51 to −1.99], s.e. = 0.64, t[315] = −5.06, p < 0.001) and 12-month time points (β = −3.28 [95% CI −4.60 to −1.97], s.e. = 0.67, t[315] = −4.90, p < 0.001). There was no effect of treatment arm on d′ (β = 0.87 [95% CI −1.39 to 3.13], s.e. = 1.15, t[235] = 0.76, p = 0.45). Figure 1 shows pairwise comparisons collapsed across treatment arms demonstrating improvements in d′ at 6-months (β = −3.25, s.e. = 0.64, z = −5.07, p < 0.001) and 12-months (β = −3.28, s.e. = 0.67, z = −4.92, p < 0.001) relative to baseline. No improvement in d′ was observed between 6 and 12 months (β = −0.04, s.e. = 0.68, z = −0.05, p = 0.99).
Attentional consistency. The effect of time on HRT-SD T-scores was not significant at either the 6-month (β = −1.11 [95% CI −2.26 to 0.04], s.e. = 0.59, t[315] = −1.89, p = 0.06) or 12-month time points (β = −1.05 [95% CI −2.25 to 0.15], s.e. = 0.61, t[315] = −1.71, p = 0.09). There was no effect of treatment arm on HRT-SD T-scores (β = 0.21 [95% CI−1.99 to 2.41], s.e. = 1.12, t[235] = 0.18, p = 0.85).
Vigilance. The effect of time on HRT-ISI T-scores was not significant at the 6-month (β = 0.72 [95% CI −0.41 to 1.85], s.e. = 0.58, t[315] = 1.25, p = 0.21) or 12-month time points (β = 0.99 [95% CI −0.35 to 2.34], s.e. = 0.69, t[315] = 1.45, p = 0.15). There was no effect of treatment arm on HRT-ISI T-scores (β = −0.17 [95% CI −2.18 to 1.84], s.e. = 1.02, t[235] = −0.17, p = 0.87).
Exploratory aim: Does self-reported impulsiveness change during DBT and do changes depend on treatment arm?
Bis-11 scores
The effect of time on BIS scores was significant at the 6-month (β = −3.17 [95% CI −4.34 to −1.99], s.e. = 0.60, t[376] = −4.49, p < 0.001) and the 12-month time points (β = −4.29 [95% CI −5.47 to −3.06], s.e. = 0.64, t[376] = −6.72, p < 0.001). There was no effect of treatment arm on self-reported impulsivity (β = 1.67 [95% CI −1.08 to 4.37], s.e. = 1.28, t[238] = 1.17, p = 0.24). Figure 1 displays pairwise comparisons collapsed across treatment arms, showing a significant improvement in self-reported impulsivity at 6-months (β = −3.17, s.e. = 0.60, z = −5.30, p < 0.001), and at 12-months (β = −4.27, s.e. = 0.61, z = −6.94, p < 0.001), relative to baseline. No further improvements were observed between the 6- and 12-month timepoints (β = −1.10, s.e. = 0.63, z = −1.75, p = 0.19).
Aim 2: Is pre-treatment cognition associated with self-harm outcomes, and do associations depend on treatment length?
Pre-treatment inhibitory control and self-harm outcomes
Commissions. A significant three-way interaction between pretreatment commissions, time, and treatment arm was found (χ2 = 35.0, df = 8, p < 0.001; Table 3), indicating that the association between pre-treatment commissions and self-harm outcomes depended on treatment arm. Figure 2a displays post-hoc pairwise comparisons for self-harm outcomes in each treatment arm up to the 2-year follow-up for participants with normal v. clinically impaired pre-treatment commissions. An examination of Fig. 2a reveals that participants with normal-range pre-treatment commissions displayed a lower frequency of self-harm when assigned to DBT-12 compared to DBT-6, during active treatment (p < 0.001 for between-arm contrasts at 9 and 12 months), and at 18-month follow up (p < 0.05). Although, by the 2-year follow-up, self-harm outcomes did not significantly differ between participants with normal commission scores assigned to either DBT-6 or DBT-12 (p > 0.05). On the other hand, participants falling within the clinically impaired range for commissions at pre-treatment displayed significantly lower frequencies of self-harm when assigned to DBT-6 v. DBT-12. In fact, significantly lower frequencies of self-harm in DBT-6 v. DBT-12 were observed during active treatment (p < 0.05 for between-arm contrasts at 6 and 12 months) and remained significant up to 2-year follow-up (p < 0.05 for between-arm contrasts at 15 and 18 months and p < 0.001 at 24 months). Taken together, poor inhibitory control at pre-treatment was associated with better self-harm outcomes, but only when assigned to brief v. standard length DBT.
Table 3. GEE analyses: significant three-way interactions between pre-treatment cognition, time, and treatment length
Figure 2. (a) Self-injury outcomes as a function of normal (left) or clinically impaired (right) pre-treatment inhibitory control performance. *p < 0.05, ***p < 0.001. (b) Self-injury outcomes as a function of low (left) or high (right) performance-based impulsivity at pre-treatment. *p < 0.05, **p < 0.01, ***p < 0.001
Impulsivity index. A significant three-way interaction between pre-treatment impulsivity index, time, and treatment arm was observed (χ2 = 41.1, df = 8, p < 0.001; Table 3), suggesting that the association between impulsivity index at pretreatment and self-harm outcomes depended on treatment arm. Figure 2b displays post-hoc pairwise comparisons of self-harm outcomes in each treatment arm up to the 2-year follow-up in participants with low v. high pre-treatment impulsivity index scores. Examining Fig. 2b reveals that participants with low pre-treatment impulsivity index scores displayed a lower frequency of self-harm when assigned to DBT-12 v. DBT-6 (between-arm contrasts were significant by 6 months at a significance threshold of p < 0.05, and remained significant at 18 months at a threshold of p < 0.01). However, by the 2-year follow-up, self-harm outcomes did not significantly differ between DBT-6 and DBT-12 in participants with low baseline impulsivity (p > 0.05). In contrast, participants with high pre-treatment impulsivity index scores displayed a significantly lower frequency of self-harm when assigned to DBT-6 v. DBT-12, although these between-arm differences were only significant at long-term follow-up time points (p < 0.01 at 15 months and p < 0.001 at 24 months). Taken together, participants with high impulsivity at pre-treatment demonstrated significantly lower frequencies of self-harm at some long-term timepoints, but only when assigned to brief v. standard length DBT.
Pre-treatment detectability and self-injury outcomes
There was a marginal three-way interaction between pre-treatment d′, time, and treatment arm (χ2 = 15.4, df = 8, p = 0.05) that was not statistically significant after correcting for multiple comparisons (p > 0.013; Table 3). Post-hoc comparisons were not examined due to the non-significant omnibus F-test.
Exploratory aim: pre-treatment self-reported impulsivity and self-injury outcomes
A three-way interaction between pre-treatment self-reported impulsivity on the BIS-11, time, and treatment arm (χ2 = 16.0, df = 8, p = 0.04) was not statistically significant after correcting for multiple comparisons (p > 0.013; Table 3). Post-hoc comparisons were not examined due to the non-significant omnibus F-test.
Discussion
There is a lack of research on changes in cognitive functioning during BPD treatments and on cognitive moderators of BPD treatment outcomes. The current findings suggest two things. First, they suggest that inhibitory control and attention may improve after at least 6 months of DBT for BPD, but due to the lack of a waitlist control arm, it is not possible to definitively conclude this. That is, it is possible that improvements in cognition could be due to practice effects on the CPT-3, instead of treatment-specific improvements. Three prior studies have reported on cognitive changes following specialist BPD therapy. Notably, our findings converge with Soler et al. (Reference Soler, Valdeperez, Feliu-Soler, Pascual, Portella, Martin-Blanco and Perez2012), who found improvements in detectability scores, in addition to inhibitory control and sustained attention, in participants with BPD who completed GPM + DBT mindfulness skills, but not GPM-alone, suggesting that DBT may impact attentional processes via the uptake of mindfulness skills. Soler et al. (Reference Soler, Elices, Pascual, Martin-Blanco, Feliu-Soler, Carmona and Portella2016) also found improvements in delay discounting (a component of impulsivity), but not on the CPT-II impulsivity index, in participants with BPD randomized to a mindfulness skills training group, relative to those randomized to interpersonal effectiveness skills training, which suggests that specific DBT skills may impact particular facets of impulsivity in BPD. On the other hand, Thomsen et al. (Reference Thomsen, Ruocco, Uliaszek, Mathiesen and Simonsen2017b) observed changes in attention and perceptual reasoning, but not inhibitory control, following 6 months of MBT for BPD, which contrasts with our findings of improved inhibitory control after 6 months of DBT. Future research may investigate whether there are changes in cognitive processes driven by the different treatment approaches of DBT and MBT. As DBT is a behaviorally rooted treatment, it is possible that it may exert a stronger influence on behavioral inhibition, whereas MBT targets social cognition (i.e. mentalizing) and may have a stronger impact on attention and perception.
Second, we also found that CPT-3 inhibitory control at pre-treatment was associated with different self-harm outcomes in DBT-6 v. DBT-12. Specifically, for participants with average inhibitory control at pre-treatment, self-harm outcomes were significantly better at 9–18 months into treatment, but only when assigned to DBT-12. Given that these between-arm differences did not emerge until 9 months into treatment, it is possible that participants with average inhibitory control may have benefited more from a longer (standard) 12-month course of DBT v. a brief course of DBT-6. As such, early desisting of DBT may not be appropriate in all cases and further study is needed in this area. Although, it is noteworthy that by the 2-year follow-up, self-harm in these participants did not differ in DBT-6 and DBT-12, suggesting that by 2 years, treatment length was not associated with self-harm outcomes in participants with average inhibitory control, In contrast, participants with clinically impaired inhibitory control at pre-treatment showed significantly lower rates of self-harm at post-treatment that were further potentiated up to 2-year follow-up, but only when assigned to the briefer DBT-6 arm. Taken together, these findings suggest that, on average, DBT-6 may be superior to DBT-12 for reducing self-harm in participants with impaired inhibitory control, but not for participants with average inhibitory control. Importantly, these findings require replication in a clinical trial with a waitlist control group to conclude that improvements in self-harm were treatment-specific.
There are several reasons why participants with impaired inhibitory control may fare better in a shorter treatment, although future work is required to explore these possibilities. Response inhibition (a component of inhibitory control strongly associated with CPT-3 commissions) may be a non-specific pre-requisite cognitive ability for optimal outcomes in long-term psychotherapy. For example, impaired inhibitory control may translate to difficulty suppressing distractors to maintain attention and process novel information across a 12-month course of psychotherapy. As a result, a briefer treatment may have felt more tolerable for participants with impaired inhibitory control initiating DBT who have such difficulties. Similarly, cognitive impairment may be associated with feelings of lower self-efficacy or poor treatment expectancies to DBT-12 v. DBT-6, which may have impacted treatment engagement or skills acquisition. These hypotheses parallel findings from at least one prior study showing that better performance-based executive control at pre-treatment was predictive of more weeks in treatment (i.e. engagement) across self-harming patients with BPD who were completing 12-month treatments (Fertuck et al., Reference Fertuck, Keilp, Song, Morris, Wilson, Brodsky and Stanley2012). Future RCTs should include participant experience measures to explore moderators of treatment response.
Moreover, it is important to highlight that there are diverse ways to measure dimensions of impulsivity (i.e. performance-based v. self-report measures). It is notable that prior research has mostly relied on self-report or demographic variables to predict treatment response. For example, one study could not identify differences in early and late self-harm responders to DBT when using such predictors (Westad et al., Reference Westad, Hagen, Jonsbu and Solem2021). The lack of prognostic utility of such variables converges with our finding that trajectories of self-harm in DBT-6 v. DBT-12 could not be discerned using self-reported impulsivity on the BIS-11. One explanation for this finding is that the BIS-11 measures trait impulsivity, which is elevated in BPD (Stanford et al., Reference Stanford, Mathias, Dougherty, Lake, Anderson and Patton2009) As such, BIS-11 scores in BPD samples may not possess enough variability to identify distinct subgroups for prediction. Indeed, approximately 73% of participants in the current study had pre-treatment BIS-11 scores in the elevated range (Stanford et al., Reference Stanford, Mathias, Dougherty, Lake, Anderson and Patton2009). In contrast, the CPT-3 assesses the ability to deliberately withhold a prepotent response (which falls in the domain of action impulsivity), and only about one-third of participants in the current study scored in the clinically elevated range. With this said, at least two studies have shown that subgrouping BPD patients based on profiles of self-reported executive function can lead to the identification of subgroup-specific symptom profiles (Hoermann, Clarkin, Hull, & Levy, Reference Hoermann, Clarkin, Hull and Levy2005; Kalpakci, Ha, & Sharp, Reference Kalpakci, Ha and Sharp2018) and the prognostic utility of these self-reported measures for predicting treatment outcomes may be investigated in the future.
Our study also has some notable strengths. This is the first study to examine performance-based cognitive predictors of self-injury outcomes in BPD treatment and to identify discernable patterns of treatment response based on pre-treatment cognitive performance. It is also the first study to examine how treatment length may interact with cognitive functioning to affect self-harm outcomes. Importantly, we employed a performance-based measure of attention and inhibitory control with standardized clinical cut-offs and clinically interpretable predictors of self-harm outcomes. Additionally, our sample of N = 240 was highly powered to detect effects among randomized participants, resulting in no between-arm differences in pre-treatment cognition or self-harm.
Clinical implications and limitations
Our findings preliminarily suggest that cognitive measures of impulsivity may have predictive utility for patients with BPD who are beginning DBT and challenge the assumption that more impaired patients require lengthy treatments to make self-harm gains. In fact, we found that brief DBT may be better suited for patients with impaired response inhibition at pre-treatment, although the limitations of our study preclude any definitive clinical implications. First, because we did not employ a waitlist control group, we cannot conclude that the observed cognitive or self-harm improvements were treatment-specific. Second, our analysis drew group-level conclusions, precluding personalized predictions of individual outcome trajectories. Finally, most participants in the trial were women, which reflect the large gender disparity in treatment-seeking patients with BPD. Future work should prioritize recruiting more gender-diverse samples into a waitlist-controlled randomized trial and may employ machine learning to examine whether CPT-3 performance can predict individual patient outcomes.
Conclusions
The current findings suggest that 6 months of DBT may be associated with improvements in inhibitory control and attention. Moreover, clinically impaired inhibitory control at pre-treatment was predictive of significantly better self-harm outcomes in brief v. standard DBT. This work represents an initial step toward an improved understanding of patient profiles that are best suited to briefer v. lengthier treatments for BPD. However, before clinical implications can be drawn from these findings, they require replication in a clinical trial with a waitlist control arm.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291723003197.
Acknowledgments
We extend our thanks to the participants in this trial for their time and contribution to BPD research. We would also like to thank CAMH biostatistician Marcos Sanches for providing statistical consultation.
Funding statement
This study was funded by the Canadian Institutes of Health Research (CIHR FRN 133428). The funder had no role in the study design, preparation of data, or writing of the manuscript. JMT was supported by a CIHR Fellowship award (2019–2021).
Competing interests
Dr McMain received personal fees from reimbursement for giving seminars and workshops on dialectical behavior therapy (DBT) and is co-owner of a practice offering DBT. Dr Chapman received payment for providing seminars and workshops on DBT and royalties for books on DBT and is co-owner of a practice offering DBT. Dr Kuo receives payment for providing consultation and workshops on DBT and is co-owner of practices offering DBT and DBT training. The other authors do not have conflicts of interest to report.
Ethical standards
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.
Open access
