Blended Treatment för PTSD: A New Format of Computer-Assisted Trauma-Focused Cognitive Behavioral Therapy

Evaluation of Blended CBT for PTSD in Regular Care: A New Format of Computer-Assisted Trauma-Focused Cognitive Behavioral Therapy: A Randomized Controlled Non-inferiority Trial

A Randomized Controlled Non-inferiority Trial testing if the new experimental Blended Trauma focused cognitive behavioral therapy (B-TF-CBT) for post traumatic stress disorder (PTSD), is non-inferior to gold standard control treatment Prolonged Exposure (PE).

Research question and hypothesis

Primary

1. Is bTF-CT non-inferior to prolonged exposure (PE) when delivered in routine clinical care in terms of PTSD symptoms on the PCL-5 post-treatment?

   Secondary:

2. Is bTF-CT non-inferior to prolonged exposure (PE) when delivered in routine clinical care in terms of PTSD symptoms on the PCL-5 at 6 and 12 months follow up?

3. Are there significant differences between bTF-CT and PE when delivered in routine clinical care in terms of depression, anxiety, insomnia, functionality, complex PTSD, other PTSD-measures and treatment satisfaction post-treatment and at 6 and 12 months follow up?

The study will be conducted at 3-6 outpatient clinics in Region Stockholm, Sweden.

To study effects, the investigators aim to conduct a randomized controlled trial (RCT) comparing blended treatment to an evidence-based face-to-face TF-CBT protocol, Prolonged exposure.

Data will be collected concerning symptoms of PTSD with the PTSD Checklist for Diagnostic and Statistical Manual 5 (DSM-5; PCL-5) which will be the primary outcome measure. Further, symptoms of PTSD will be measured with Clinician-administered PTSD Scale for DSM-5 (CAPS-5), depression The PHQ-9, symptoms of anxiety with GAD-7 , quality of life with Work and social adjustment scale (WSAS), and sleep with the insomnia severity index (ISI). Outcome measures will be distributed before, during, after, as well as 12-months following treatment.

In accordance with CONSORT recommendations for non-inferiority trials, both intention-to-treat (ITT) and per-protocol (PP) analyses will be performed and reported for the primary non-inferiority comparisons. Secondary outcomes will be analyzed using ITT principles.

ITT population All randomized participants who completed the pre-treatment assessment and participated in the inclusion session.

PP population The PP population will include participants who have received at least 50% of the planned total treatment dose, defined across both digital modules and face-to-face sessions in the blended format.

Demographic variables Basic demographic variables (e.g., age, gender, education, trauma history, employment, and comorbid psychiatric disorders) will be collected at baseline.

Therapist competence and fidelity Therapist competence and adherence to PE and blended TF-CBT protocols will be assessed using structured fidelity checklists. A subset of recorded sessions will be rated by independent assessors.

Inter-rater reliability for CAPS-5 Interrater reliability will be assessed by having all CAPS raters independently score the same recorded role-play CAPS interview, with agreement evaluated using intraclass correlation coefficients (ICC), both between raters and relative to a predefined reference rating

Protocol deviations and adverse events Protocol deviations and adverse events (e.g., treatment interruptions, medication changes, unexpected events) will be documented and summarized descriptively.

Treatment dose and engagement Treatment engagement (e.g., number of sessions, digital module completion, total therapeutic time, and treatment duration) will be recorded to describe dose and support per-protocol analyses.

Concurrent treatments Participants will be asked about any concurrent psychological or pharmacological treatment at follow-up assessments.

To analyze the data, a non-inferiority analysis will be applied for the primary outcome variable. Step one in this analysis will be to determine what difference in mean scores on the PCL-5 between the two treatments is tolerated to conclude that the experimental intervention is non-inferior to the standard treatment (non-inferiority margin). Continuous outcomes will be analysed using linear mixed models, and non-inferiority will be assessed based on the 95% confidence interval for the estimated mean difference between treatments. Non-inferiority will be concluded if the lower bound of the confidence interval for the experimental treatment (bTF-CT) is above the predefined non-inferiority margin. If this criterion is met, the experimental treatment will be interpreted as non-inferior to the standard treatment (TF-CBT).

The non-inferiority margin was determined using procedures recommended in methodological guidelines for non-inferiority trials. One approach was based on preservation of a proportion of the established effect of the active control. Specifically, a pooled effect size of 1.0 for prolonged exposure therapy (PE) was identified from a published meta-analysis. In accordance with recommended practice, 50% of this effect was retained to define the non-inferiority margin, corresponding to a Cohen's d of 0.5. To adopt a more conservative approach and to ensure that the margin remained below the conventional threshold for a medium-sized effect, the non-inferiority margin was therefore set to Cohen's d = 0.4. Assuming a standard deviation of 15.98 on the PCL-5, based on post-treatment values from a comparable reference study, this corresponds to an absolute mean difference of approximately 7 points on the PCL-5. This value was therefore selected as the non-inferiority margin. As an additional check of clinical plausibility, the selected margin was compared with previously established estimates of the minimal important difference (MID) for the PCL-5, which have been reported to be approximately 9 points, indicating that the chosen margin was conservative.

Further, to estimate the required sample size, rigorous power calculations were carried out in collaboration with an expert in statistical analysis, applying a simulation-based approach using a 2-level linear mixed-effects model, using estimates from the reference TF-CBT, looking at six assessment points. Variance components that were incorporated into the analysis were random intercept, random slopes and residual variance. The power calculation showed that with 78 participants in each treatment group and a 20% dropout rate, 80% power is reached, given a non-inferiority margin of 7 points and α = 0.05.

An interim power analysis was also conducted by an external analyst, using model parameters from observed data in the power analysis, while also updating the number of assessment points to 16 and using α = 0.025. This resulted in 70 participants per group with power > 80%. To allow for per-protocol analysis, we will aim for 160 participants in total.

Continuous outcomes will be analyzed using linear mixed models (LMM) with an appropriate distribution family. Model checks for residuals and outliers will be performed using the R packages 'DHARMa' and 'performance' with default settings. If outliers are identified, this will be handled by using a robust LMM. Time in weeks will be treated as a continuous linear variable. The baseline measurement will be used as a covariate, modeled with a restricted cubic or natural spline. If time is found to be non-linear, it will also be modeled using splines. Confidence intervals will be generated using bootstrap. The mean difference between intervention groups (average treatment effect) will be estimated for the post-treatment measurement point using g-estimation. In the main analysis, treatment site will be added as a fixed effect covariate interacted with the treatment variable, and a therapist variable using a random slope only. This analysis will be conducted both using intent-to-treat (ITT) and per-protocol (PP). Missing data will be handled by multiple imputation with chained equations, imputing 20 datasets for analysis and pooling results. If there are differences in adherence between treatment groups for the PP analysis, a more robust method will be used, such as inverse probability of treatment weighting of instrumental variables(1). The investigators will also calculate effect sizes using Cohen's d based on pooled standard deviations.

Following regulatory recommendations (2), a baseline covariate-adjusted analysis will also be conducted and reported. This model will use the ANHECOVA approach, where all covariates interact with the treatment variable. Planned covariates that are added compared to the main analysis are: age, gender, education level, depression score, insomnia severity index score, work and social adjustment scale score, and trauma during childhood. Continuous baseline variables will be grand mean centered. Bootstrap will be used for confidence intervals. This analysis will only use ITT.

Sensitivity analyses will be conducted. One model will include only baseline, mid-treatment and post-treatment measurement points, since these are expected to have much lower attrition compared to weekly measurements. We will use a mixed model for repeated measurements (MMRM) with an unstructured covariance matrix. This model will also be fitted using both ITT and PP. The credibility and expectancy scale (CEQ), measured 2 weeks post-treatment start, will be used as a covariate in one model to evaluate potential effects of CEQ on the primary outcome. This model will also include the list of covariates specified in the previous paragraph.

Dichotomous variables that will be analyzed as outcomes: Remission rates will be calculated using a cut off of 31 on the PCL-5, across time in weeks; having a PTSD diagnosis according to CAPS-5 at time points pre, post, 6- and 12-months follow-up; and the proportion of patients who will show treatment response on the PCL-5, defined as a reduction of at least 10 points from baseline. All dichotomous outcomes will be analyzed using mixed model logistic regressions and comparisons between treatment groups will be reported as incident rate ratios and/or risk differences. Dichotomous data of remission will also be analyzed with survival analysis where time to remission is calculated and compared between groups.

1. Dodd M, Carpenter J, Thompson JA, et al. Assessing efficacy in non-inferiority trials with non-adherence to interventions: Are intention-to-treat and per-protocol analyses fit for purpose? Stat Med 2024; 43: 2314-2331. DOI: https://doi.org/10.1002/sim.10067.
2. Ye T, Shao J, Yi Y, et al. Toward Better Practice of Covariate Adjustment in Analyzing Randomized Clinical Trials. Journal of the American Statistical Association 2023; 118: 2370-2382. DOI: 10.1080/01621459.2022.2049278.

Inclusion Criteria:

• PTSD according to Diagnostic and Statistical Manual of mental disorders 5 as primary problem
• ≥ 18 years
• Basic knowledge of speaking and writing Swedish
• No other psychological treatment at the same time
• No ongoing threats or violence
• If antidepressant medication, stable dose 6 weeks before starting treatment
• Have access to the internet and a computer
• Willingness/ability to participate after receiving information about what participation entails

Exclusion Criteria:

• Other diagnosis or problem is assessed as primary to PTSD
• Moderate to severe suicidal risk
• PTSD from childhood trauma that the person cannot remember