Increasing Treatment Response Rates in Depressed Adolescents Via Feedback-Informed IPDT

Study goals and objectives This is the third study in a larger project, the ERiCA project. The overarching aim of the project is to develop and evaluate cost-effective internet-delivered treatments for adolescent depression. The present study will try an adaptive treatment strategy with the goal of increasing treatment response rates. Based on data from our previous large non-inferiority trial (Mechler et al., 2022), an algorithm have been developed for prediction of treatment non-response at week 3. The specific objective of the present trial is to evaluate whether identifying participants at risk of non-response and then adapting treatment during the course of treatment can reduce risk of non-response. Additional goals are to investigate process factors that predict and mediate outcome for the treatment in order to learn more about mechanisms and processes of change and no-change.

Primary research question Research question 1: Can adapted treatment reduce risk of non-response for identified at-risk-participants in IPDT? Hypothesis 1: Adapted treatment will decrease risk of non-response in IPDT in comparison with non-adaptive treatment strategy.

Research question 2: Can non-response be accurately predicted? Hypothesis 2: Non-response will be accurately predicted using an algorithm at week 3.

Secondary research questions 3: Are there patient factors that affect efficacy of the treatment program (i.e. predictors/moderators) and the adapted treatment program? 4: Are there within-treatment factors that affect efficacy of the program and adaptation (mediators)? 5. Are there differences in the effects on the severity of depressive symptoms, anxiety symptoms, capacity for mentalized affectivity, emotion regulation, fear of negative emotions, defense mechanisms, and self-concept when comparing adaptive treatment strategies to non-adaptive?

Study Design The trial is a randomized controlled trial (RCT). All eligible participants (n = 240) will be entered into IPDT. At week 3 in treatment, participants at risk of non-response will be identified using our open-source algorithm. Identified participants at risk will be randomized to either adapted treatment or continued IPDT as usual.

Withdrawal criteria: Patients that deteriorate into becoming suicidal will be withdrawn from the project and referred to psychiatric care. Data will be collected at baseline, multiple time points (weekly) during treatment, as well as at follow up (6 and 12 months after termination).

The trial is scheduled to take place between 2024 and 2025.

Study setting The project is based at Stockholm University. As treatment is conducted over the internet, participants can be recruited nation-wide, giving opportunities for recruiting a larger as well as a more heterogeneous sample regarding several aspects such as geographic location and socioeconomic status. The project will use the well-developed infrastructure and the secure Internet platform used in many studies conducted by Per Carlbring at Stockholm University and Gerhard Andersson at Linköping University. The project is conducted in close collaboration with professor Gerhard Andersson at Linköping University who has vast experience in internet trials with adults as well as adolescents.

Intervention IPDT consists of 8 therapist-supported self-help modules delivered over 10 weeks on a secure online platform. Modules consist of texts and video followed by assignments which they send to their therapist and receive feedback within a few days.

The IPDT program has been developed specifically for this project and is based on similar principles as a treatment with shown efficacy for adults in several RCTs. The aim of the intervention is to decrease emotional avoidance and increase awareness and experience of emotions. Participants are encouraged to approach previously warded off feelings. They will also be taught how to link their emotional reactions to their symptoms. Another treatment goal is to acquire a greater capacity for anxiety regulation. The final part of the program contains material on how to express previously warded off affects in key relationships, in order to improve the participant's relationships. Compared to the existing treatment for adults, the modules in the age-adapted program for adolescents are shorter and easier to read, with vignettes more recognizable for the age group. Latter parts of the material give particular notice to developmental aspects especially relevant in adolescence.

Adapted treatment Participants randomized to adapted treatment are contacted by their assigned therapist and given information about treatment modification. Changes are implemented as of week 3 of treatment. The adaptations will be suggested by the therapist from a treatment manual with the goal of personalizing treatment and facilitating treatment engagement. The therapist and the participant will collaboratively agree on adaptations. Examples of adaptations are: added weekly text-chat (30 min), added bi-weekly phone calls (30 min), reminders, and shortened chapters or less exercises.

Therapists The therapists will be master students in the Swedish clinical program in psychology, in the final phase of their training as psychologists in which they have their psychotherapy courses. The therapists in the project will be trained in IPDT and supervised by an experienced psychologist. The therapists will be able to track patients' progress on the measures administered weekly.

Outcomes Primary outcome will be response to treatment according to the proportion improvement method (Hiller et al., 2012).

Research question 1 will be investigated by comparing the outcome of adapted IPDT for at-risk patients with standard IPDT for at-risk patients.

Research question 2 will be investigated by comparing the outcome of standard IPDT for on-track patients with standard IPDT for at-risk patients.

Furthermore, we will investigate possible predictors, moderators and mediators in terms of capacity for mentalized affectivity, identity, fear of negative emotions, demographics, expectations, symptom/distress profile, self-compassion and treatment alliance.

Participant timeline Participants contacting the project will be directed to an online website where they can access further information about the project, give informed consent to participate and subsequently get access to the screening forms for the trial. If fulfilling inclusion criteria and not fulfilling exclusion criteria according to the screening forms, a diagnostic telephone interview will be held, establishing fulfilment of inclusion criteria and non-fulfilment of any exclusion criteria. Eligible participants will be asked to confirm participation in the study, after which they will start in standard IPDT.

All instruments consist of online administered self-report questionnaires with proven validity and reliability. Five of the self-rating scales will be administered weekly, for tracking outcome and possible mediators. Two questionnaires will also be administered weekly to the treating therapists for tracking potential mediators. Other questionnaires will be administered only at baseline, termination and follow-up.

Sample size Power calculation focusing on the primary research question was made based on binary outcomes (non-response/response) using Sealed Envelope. We based our power analysis on the results of adapted treatment in Forsell et al. (2019), where adapted treatment for at-risk patients was associated with a lessened risk for treatment failure with an odds ratio of 0.327. Furthermore, using an α of 0.95, and a power of 80%, 49 patients in each group is needed, meaning a randomized sample of 98 patients. As the algorithm developed for this project identifies approximately 50% of patients, this means a total sample of 196 patients. Based on an attrition rate of 20%, the total number of patients needed are 235. To obtain slightly broader margins for attrition, we will include a total number of 240 participants.

Recruitment Participants will be recruited primarily through advertisements in newspapers and on the Internet (including social media). Junior and senior high schools will also be contacted to inform about the study. User organizations will be involved in informing about the study and recruiting participants.

Blinding/masking For at-risk-participants allocated to adapted treatment, therapists cannot be blind to allocation. However, participants are not informed that they are classified as at risk for non-response. For all other participants, meaning participants classified as on-track as well as participants classified as at-risk randomly allocated to the IPDT as usual-arm, both participant and therapist will be blind to classification. All of the outcome measures are self-rating scales, rendering blinding of assessors irrelevant.

Data management Data management in the study will follow the established procedures for studies on Internet-based treatments conducted by the research groups of Per Carlbring at Stockholm University and Gerhard Andersson at Linköping University. These procedures have been used in a vast amount of studies conducted over several years. All individuals participating in the project will be assigned a unique study code, used throughout the project for communication via the Internet. All correspondence with the participants will be made using a secure communication system. Importantly, no communication with the participant will be made using standard email. All communication and data collected through Internet-based self-report measures will only be associated with the unique identifier, and no personal information. While the project is on-going all data (communication and data from self-report measures) will be stored in an encrypted database, only accessible by the primary study administrators. Personal information will be stored in a fireproof safe that only the study administrators can access. The key that links participants to their unique study code will be stored in a separate fireproof safe fulfilling the regulations concerning safe storage of data. The data management procedures adhere to the General Data Protection Regulation (GDPR).

Statistical methods Based on data from the large non-inferiority trial conducted in the present research project (Mechler et al., 2022), an algorithm has been developed to identify participants at risk of non-response or deterioration. Based on the data from Mechler et al., 2022 and using the F1-score as precision measure (Sasaki, 2007), this algorithm made predictions of non-response at treatment week 3 that could be trusted to 59% (i.e. a positive prediction value of 0.59) and with specificity of 80%. This algorithm is based on linear mixed models and uses data from weekly measurements of QIDS-A17-SR as well as baseline ratings of the Personality Inventory for the Diagnostic and Statistical Manual of mental disorders 5 Brief Form (PID 5-BF). Using stratified 5-fold cross validation, different mixed models were trained on data from Mechler et al., 2022 using Bayesian estimation methods with a Markov chain Monte Carlo (MCMC) algorithm (Dimitris, 2016). The best performing model for finding at-risk patients (evaluated by the F1 measure) had a logarithmic change term for modelling the weekly measurements of QIDS-A17-SR and baseline ratings of PID 5-BF as covariate. Using parameters from this model, Bayesian estimation with MCMC will be used in the same way as in the training step to make predictions of QIDS-A17-SR values for future time points in the present study. At-risk patients will be patients predicted by the model not to meet the two treatment response criteria defined by Hiller et al. (2012) at the last treatment session.

Primary outcomes will be assessed using odds ratios. Secondary outcomes exploring trajectories of change will use rates-of-change estimates based on linear mixed models. Differences in efficacy between conditions will be investigated by modelling interaction effects of group and time. These methods have been recommended for RCTs investigating internet interventions (Hesser, 2015). One important advantage is their ability to handle missing data, using full information maximum likelihood estimation (Gueorguieva \& Krystal, 2004). The process of change will be examined thoroughly by employing longitudinal mediation in a structural equation modelling (SEM) framework. In case mediators are found, causality will be tested by including the mediator as lagged time-varying covariate in the model, i.e. testing if change in an outcome variable at a certain time-point can be predicted by change in a mediating variable at a previous time-point.

Dissemination of Results and Publication Policy The primary outcome paper will present outcome data in a journal with open access publication. No outcome data will be published or presented before data collection is completed. The results will also be disseminated in popular science form through different media, partly with help of the user representatives from Suicide Zero.

Ethics and safety considerations Approval of the Swedish Ethical Review Authority was decided on 13th November, 2023. Participants give informed consent prior to screening. Correspondence takes place in a closed system accessible with double authentication, data encrypted, and stored in accordance with Swedish law. Data is analysed and reported at group-level. Thorough inclusion interviews ensure that psychiatric problems making the treatment unsuitable or indication of suicidality are ruled out and the young person will be referred to relevant care. An experienced psychiatrist is participating in the study, who partakes in assessing suitability versus unsuitability, and referring adolescents who need more specialized treatment. Participants who express increased suicidality or severe deterioration will be called up by their therapist or by the principal investigator and risk of suicidality or self-harm will be assessed. If a participant is considered as suicidal, the psychiatrist will make a second assessment and help the young person to relevant care. If participants under 18 express increased suicidal ideation, deteriorates severely or do not respond when contacted, legal guardian(s) will be contacted. All participants will be advised to inform their legal guardian(s) about their participation in the trial. All adverse events will be recorded and reported in the study. Those events judged as serious adverse events (such as suicide risk, substance misuse, or on-going maltreatment or sexual abuse) will lead to actions to ensure the safety of the patient, and that he/she receives adequate help. Participants are informed that they may withdraw from the study for any reason at any time.