Efficacy of digital cognitive behavioural therapy for symptoms of generalised anxiety disorder: a study protocol for a randomised controlled trial

This study is a parallel-group superiority RCT which will examine the efficacy of a novel smartphone-delivered digital CBT intervention for GAD compared to waitlist control for treating symptoms of GAD in adults with moderate-severe symptom severity. The DeLTA (DigitaL Therapy for Anxiety) trial will be conducted entirely online.

The sample size calculation was conducted using G*Power 3 [31]. The study aims to recruit 242 participants with GAD, with 121 participants randomised to each arm. This recruitment target accounts for approximately 30% attrition, based on the range of dropout rates from previous RCTs comparing smartphone-based digital interventions for anxiety to waitlist (19% to 45%; [32–34]). The sample size was determined based on an estimated medium effect on post-intervention scores on the seven-item Generalised Anxiety Disorder questionnaire (GAD-7) (Cohen’s d = 0.50) between the digital CBT and waitlist arms, with 90% power, a 1:1 allocation ratio, and with α set at 0.05. The effect size is based on a recent meta-analysis of RCTs of smartphone-delivered psychological treatment for anxiety versus waitlist which demonstrated a small-to-moderate (Hedges’ g = 0.33) between-group post-intervention effect on anxiety symptoms when compared to any active or waitlist/inactive control condition [28]. When comparing smartphone interventions with only waitlist/inactive control conditions, Hedges’ g increased to 0.45. The effect size of 0.50 corresponds to a minimally clinically significant 2-point change on the GAD-7 and a standard deviation of 4 in this population [35].

For inclusion in this trial, participants must: 1) be aged 18 years or older; 2) score 10 or higher on the GAD-7 [35] indicating at least moderate GAD symptom severity; 3) screen positive for a GAD diagnosis on a digital version of the Mini-International Neuropsychiatric Interview (MINI) version 7 for DSM-5 [36], followed by telephone verification (see Procedure below for details); 4) be either not on prescription medication for anxiety, depressive symptoms, or poor sleep, or on a stable dose for at least 4 weeks; and 5) must not be currently receiving or have previously received CBT for anxiety in the last 12 months (this inclusion criterion was included after a modification was made by the ethical review committee after recruitment start). We will exclude participants who self-report: 1) being diagnosed with any of the following conditions: schizophrenia, psychosis, bipolar disorder, seizure disorder, substance use disorder; 2) having recent trauma to the head or brain damage; 3) having severe cognitive impairment; 4) having serious physical health concerns necessitating surgery or with a prognosis of less than 6 months; or 5) being pregnant. Individuals with other anxiety or related disorders (e.g. panic disorder, social anxiety disorder, and so forth) were included in the study as long as they endorsed worry (consistent with a GAD diagnosis) as their primary concern assessed at telephone screening.

Participants will be recruited online from the UK and the USA. A number of recruitment strategies will be used, including online, print, and broadcast media advertisements. Participants will receive 10 GBP in Amazon.com gift vouchers (or USD equivalent at the time of payment) in exchange for completing each of the core assessments following baseline (mid-intervention, post-intervention, and follow-up), and a bonus of 10 GBP (or USD equivalent) in vouchers for completing all assessments. Therefore, each participant can receive up to a total maximum of 40 GBP (or USD equivalent) in vouchers. Payment will be made at the end of the core study period for each participant (after 10 weeks from randomisation). If participants withdraw from the study, payment will be made for completion of any assessments at the point of withdrawal. Participants may withdraw themselves from the study for any reason at any time. The study investigator may also withdraw participants from the study to protect their safety and/or if they are unwilling or unable to comply with required study procedures after consultation with the study investigator team.

Participants will be assigned to the digital CBT or waitlist arm by blocked randomisation with a 1:1 allocation ratio. Randomisation will be carried out and the allocation sequence generated automatically upon completion of baseline measures using the randomisation function within Qualtrics Survey Software (Qualtrics, Provo, UT, USA, [37]). Members of the research team will be unable to influence randomisation and will be concealed from future assignments.

All assessments will be completed by participants online, independent from members of the research team, which will reduce the risk of bias associated with researcher-administered assessments. Participants will be informed of their randomly allocated condition (digital CBT or waitlist control) and will not be blind to group allocation. The trial coordinator will not be blind to group allocation as they will monitor uptake (download) of the intervention by participants. Participants who do not download the intervention will be reminded first by a daily email reminder for up to 3 days and then contacted by telephone. All other members of the research team will be blind to allocation. Participant contact with the trial coordinator will be limited to standardised emails containing instructions for completing online assessments and accessing digital CBT and will not cover treatment content or clinical support. Data analyses will be conducted by an independent external statistician who is not involved in the management of this trial. The statistician will be blind during the study and subgroup-unblind (groups labelled as ‘A’ and ‘B’) during the statistical analysis.

Online assessments for the core study period will occur at 0 (baseline), 3 (mid-intervention), 6 (post-intervention) and 10 weeks (follow-up) from randomisation. Waitlist control participants will receive access to the digital CBT app after the follow-up assessment point (after 10 weeks from randomisation). In order to understand if effects persist over time, participants randomised to the intervention arm will also be invited to complete an optional longer-term uncontrolled assessment online at 6 months from randomisation without compensation. Table 1 presents details of the self-reported measures administered at each assessment point. After randomisation, participants will have up to 2 weeks to complete each questionnaire assessment battery. The trial coordinator will make every reasonable effort to follow the participant for the entire study period (from consent to the final follow-up assessment) using both email and telephone calls. To promote outcome data completeness, where necessary participants will be reminded to complete assessments first by a daily email reminder for up to 3 days and then by telephone for each overdue assessment. Participants who do not complete a study assessment will not be withdrawn from the study and will continue to be asked to complete any subsequent assessments. All online surveys will be hosted and stored on Qualtrics. Potential errors with data entry will be minimised as data will be entered by participants online.

The Consolidated Standards of Reporting Trials (CONSORT; [38]) diagram displaying participant flow through the trial is shown in Fig. 1. Participants who respond to advertisements will be directed to the online participant information sheet and asked to consent to and complete an online screening survey as a first step in assessing eligibility against inclusion and exclusion criteria. Screening questions will assess age, GAD symptom severity (GAD-7), GAD diagnosis (MINI for GAD), medical and psychiatric history, prescription medication use and pregnancy. Eligible participants who meet inclusion criteria assessed above by the online screening survey will then be contacted by a member of the research team to take part in a brief 10-min telephone call. During this call, study procedures will be explained, any questions about the study will be answered, and we will verify participant reports of GAD against criteria from the Structured Clinical Interview for DSM-5 [39], ensuring that the participant feels their GAD symptoms are their current primary area of concern. All potential participants were informed (through study advertising, the participant information sheet and during the telephone conversation) that they should aim to use the app daily throughout the study period. Eligible participants who are willing to participate are then emailed a link to the online consent form.

Participants who consent will be allocated a unique identification code to use when completing all subsequent online surveys to ensure that identifiable information provided as part of consent is not directly linked to survey responses. This unique identification code will be used for data files to ensure confidentiality. Participants will receive an email asking them to complete an online baseline survey containing a battery of self-report measures (see Table 1). Upon completion of the baseline survey, participants will be automatically randomised to either the digital CBT or waitlist control condition and informed of their allocation by automated email. All participants allocated to digital CBT will receive immediate access to the intervention and access will not be withdrawn (i.e. participants are welcome to continue using the app after conclusion of their participation in the study). All waitlist control participants will be given access to digital CBT after the final controlled follow-up assessment point (after 10 weeks from randomisation).

The intervention involves interactive and tailored delivery of digital CBT for worry and anxiety via a smartphone-based app, Daylight (https://www.bighealth.com/daylight↗). The app provides an interactive and media-rich experience and includes supportive visuals and brief animations. Daylight is a voice-led experience, in which a virtual therapist guides the user through intervention content. Throughout the intervention, participants are asked to complete questions within the app about their anxiety and other aspects of their experience (e.g. mood, sleep). Personalisation is built in using algorithms to tailor the intervention based on responses to questions and user’s progress.

Daylight was developed in collaboration with leading experts in the area of CBT for GAD and anxiety disorders. The content is based on evidence-based CBT techniques for the treatment of GAD, including psychoeducation, stimulus control, applied relaxation, cognitive restructuring, imaginal exposures and self-monitoring of progress. The programme is designed to be self-paced and includes four modules, each lasting up to 20 min, and shorter practice versions of the same techniques (lasting approximately 5 min). Modules are accessed sequentially, with access to a subsequent module available on completion of the previous module. The app provides feedback and troubleshooting based on the user’s feedback during the exercises as well as personalised recommendations for how the techniques may be applied in the user’s life. At the start of the programme, Daylight recommends that users aim to use the app daily to practice techniques; accordingly, the modules and practice exercises can be repeatedly accessed (as per the user’s preference) for reinforcement of specific techniques. Optional notifications in the form of emails (once weekly for the first few weeks of use), text messages (twice weekly) and in-app push notifications (of varied frequency) are used to prompt use of the app. The digital CBT intervention allows for monitoring of engagement through the automatic capture of objective usage statistics including the number of modules completed, frequency of access, and length of time taken to complete all modules.

Outcomes are self-reported and measured online at screening or baseline (week 0; immediately preceding randomisation), mid-intervention (3 weeks), post-intervention (6 weeks) and follow-up (10 weeks). Measures included are indicated below and in Table 1.

The trial investigators and the trial statistician will have access to the final cleaned dataset. The project dataset will be collected and stored online in the study web portal created for the study by Qualtrics Survey Software (Qualtrics, Provo, UT, USA), and all data will be password protected. Data integrity will be enforced through a variety of mechanisms including data rules, range checks and consistency checks against data already stored in the database. All data will be treated as confidential. To ensure confidentiality, data dispersed to project team members will be blinded of any identifying participant information. Questionnaire data will be de-identified by a unique coded participant identification number only (assigned at consent into the study) to maintain confidentiality. Participant data linkage information will be stored in a single digital file separate from questionnaire research data. This file will be encrypted and stored on a secure server with limited user access and password protection for members of the study team only. Minimal personal identifiable information (e.g. name, email) will be captured and will only be made accessible (password protected) to the research study team. These data will be deleted at the completion of the study on publication or public release of the results. We will only hold contact information of participants who give optional consent to being contacted again in the future and will retain these records for 7 years after the release of this work. The final anonymised research data will be archived with the Oxford Research Archive (ORA-Data) at the University of Oxford for longer term storage (at least 7 years) after publication or public release of the results.

Analyses will be carried out using Stata [52]. In accordance with the CONSORT statement [38], we will report all participant flow in each arm (see Fig. 1). Descriptive statistics of participant recruitment, study dropout and intervention engagement will be reported. Baseline characteristics will be presented by randomised group and summarised by means and standard deviations for continuous measures, or number and percentage for categorical measures.

To test the primary hypothesis, treatment effects on GAD symptom severity (GAD-7) will be estimated using linear mixed models fitted to data at all core time points (baseline, 3 weeks mid-intervention, 6 weeks post-intervention, 10 weeks follow-up). Fixed effects will be baseline assessment for GAD-7, randomised group, time (categorical), and time by group interactions. Participant ID will be included as random intercepts to allow for repeated measures. Marginal treatment effects will be estimated for GAD-7 at each time point and reported separately as mean-adjusted differences in scores between the randomised treatment groups with 95% confidence intervals and two-sided P values. The analysis will use statistical techniques for handling missing data under a missing-at-random assumption. Secondary hypotheses will be tested using analogous analyses. For the binary secondary outcome (whether participants achieved remission and reliable remission on GAD-7, PHQ-9 and SCI-8 outcomes), the same approach will be followed using logistic mixed models. All primary and secondary analyses will be conducted on an intention-to-treat basis, including all participants who were randomised, with no planned interim analysis.

Additional sensitivity analyses will be conducted to test primary and secondary hypotheses using data from participants who receive and engage with the intervention, defined as those who: 1) download the app; 2) complete at least two modules of digital CBT; and 3) complete all modules. Separate sets of analyses will be conducted for each of these three categories of intervention completion.

Exploratory analyses will examine mechanisms underlying the effect of digital CBT compared to waitlist control on GAD symptom severity using causal mediation methods based on parametric regression models. This will examine whether changes in worry (PSWQ), depressive symptoms (PHQ-9) and sleep difficulty (SCI-8) during the intervention period (from baseline to 3 weeks mid-intervention) significantly mediate the effect of digital CBT compared to waitlist on changes in GAD symptom severity (GAD-7) from baseline to 6 weeks post-intervention. Exploratory moderation analyses will also be conducted to examine whether the between-group effect on GAD symptom severity is moderated by baseline variables (e.g. demographic variables, baseline levels of anxiety).

All analyses will be pre-specified in a detailed statistical analysis plan.

We will publish the results of this study in peer-reviewed journals, irrespective of magnitude or direction of effect. Findings will also be presented at national or international scientific meetings. A lay report of the findings will be produced and disseminated to participants upon request. Items in this protocol comply with the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) checklist [53] (see Additional file 1; and Fig. 2 for the SPIRIT figure).

Any participant who is discovered to have another medical condition will be referred to their medical practitioner by a study clinical psychologist. Participants who complete or withdraw at any time for any reason during the study period will be referred by a study clinical psychologist at discretion, and/or if required or requested by the participant, to their medical practitioner for further management.

This is protocol version 1.1, September 2019. Recruitment started on 2 August 2019.

At the time of manuscript submission, recruitment for this study is not yet complete. It is anticipated that recruitment will end in November 2019.

Please see Fig. 2 for a copy of the SPIRIT figure. The SPIRIT checklist can be found in Additional file 1.

The full study protocol, statistical code and participant-level dataset for the current study may be available from the corresponding author on reasonable request.