This was a phase 3 multicenter, randomized, masked, sham-controlled trial to evaluate the efficacy of CT-152 in adult participants diagnosed with MDD who were taking ADT. Participants took part in the trial for up to 13 weeks. The trial included a screening period of up to 3 weeks, a treatment period of 6 weeks, and an extension period of 4 weeks to assess the durability of the effect (Figure 1). Given the COVID-19 pandemic and the advantages of remote trials, especially for a DTx, MIRAI was run as a remote study. Except the screening visit, all visits took place remotely; the screening visit had an in-person or remote option per investigator and patient preference. Due to the DTx nature of this trial, participants needed routine access to their smartphones in their work and home life.

Key participant inclusion and exclusion criteria are listed in Textbox 1.

This trial was conducted in accordance with local laws, the International Conference on Harmonization Good Clinical Practice guidelines, and the Declaration of Helsinki. The protocol was reviewed and approved by the governing institutional review board (IRB) for each investigational site. Of the 54 study sites, IRB services were provided by Advarra (approval number 00000971) for 53 sites, and WCG IRB (approval number 1-1377994-1) for 1 site. IRB approvals were granted between November 2020 and January 2021. All participants were asked to provide consent to participate in this study before enrollment and agree to its privacy policy and terms of service required for the remote collection of their information. All information collected in the trial was considered confidential and was only used in accordance with the protocol. Participants were identified only by a unique identification number in eSource. However, regulatory officials or sponsor personnel could be allowed access to records, consistent with local requirements. Participants could withdraw from the trial at any time without justification. Participants received a stipend to help cover the costs associated with being in the study. Participants also had the option to have their ADT provided to them by the sponsor at no charge for the length of the study.

CT-152 is a smartphone app–based DTx that delivers a proprietary interactive cognitive-emotional and behavioral intervention, intended for use as an adjunct to ADT. The core components of CT-152 are the EFMT exercises; brief CBT-based lessons, each paired with an out-of-app activity or guided audio psychotherapy exercise targeting the most common MDD symptoms; and SMS text messaging to reinforce the CBT-based lessons and encourage engagement with the app. Each treatment session consisted of an EFMT exercise and a brief lesson with an activity. During the EFMT exercises, participants were presented with a series of pictures of faces, tasked with identifying the emotions they observed and remembering the sequence of emotions [35]. Using an N-back working memory training paradigm, for each face observed, participants indicated whether the emotion is the same as the emotion N faces back [35]. To date, CT-152 has only been developed in English.

A sham DTx served as a control and provided a cognitive training exercise designed to retain user interest while minimizing therapeutic effects. Each sham session consisted of a Shapes Memory Task (SMT) exercise. The SMT is a working memory task that provides users with an analogous structure, matched for time and attention, to the cognitive-emotional training exercise (EFMT) in CT-152. The sham DTx did not include EFMT exercises or lessons to retain its intended nontherapeutic nature. Participants in the control group also received supportive SMS text messages related to maintaining engagement with study treatment. The SMT exercise has been previously used in EFMT pilot studies as the control condition [35,36].

In order to be eligible to participate, during the screening period, participants had to complete an N-back working memory task simulating basic features of CT-152 to ensure they had the necessary skills to use it. This also confirmed their interest in participating in the study. After eligibility was confirmed, participants were randomized 1:1 to either CT-152 or sham DTx at the baseline visit (day 1). The participants progressed through a treatment schedule of 18 sessions at a rate of 3 sessions per week over the 6-week treatment period (day 1 to week 6). Participants had a remote study visit at weeks 2, 4, and 6 and were contacted by telephone at weeks 1, 3, and 5.

After week 6, participants continued in the trial during the extension period (weeks 7 to 10) to assess the durability of the effect. In the extension period, the smartphone app remained installed for each group; no new therapeutic content was introduced, and EFMT and SMT exercises were not available. The 2 groups each received brief supportive SMS text messages in the extension period with content related to the previously completed exercises. Participants had a remote study visit at weeks 8 and 10 and were contacted by the trial site by telephone at weeks 7 and 9.

A dedicated call center was available to support the participants and the trial sites on the initial downloading of and access to the smartphone app, as well as any technical issues with the app throughout the trial.

The primary objective of the trial was to evaluate the efficacy of CT-152 in improving depressive symptoms versus the sham DTx, which served as a control. The primary efficacy end point was the change from baseline to week 6 in the Montgomery-Åsberg Depression Rating Scale (MADRS) total score.

Other efficacy end points included assessment of the MADRS at weeks 2 and 4; Generalized Anxiety Disorder-7 (GAD-7) scores at weeks 2, 4, and 6; the Clinical Global Impression–Severity scale scores (CGI-S) at weeks 2, 4, and 6; and the Patient Health Questionnaire-9 (PHQ-9) scores at weeks 4 and 6. Functional outcomes were assessed by World Health Organization Disability Assessment Schedule 2.0 (WHODAS 2.0) scores and quality of life by EQ-5D-5L scores at week 6. To evaluate the durability of the effect of the DTx, the change from baseline in MADRS and GAD-7 total score as well as MADRS response and remission rates were evaluated at weeks 8 and 10. The MADRS was conducted by trained and calibrated independent raters. Trial site staff administered the CGI-S. Participants completed the GAD-7, PHQ-9, WHODAS 2.0, and EQ-5D-5L (for scale descriptions, refer to Table 1).

The safety objective was to assess the safety of CT-152 in adult participants with MDD who were concurrently taking ADT. Safety was evaluated by the frequency and severity of AEs, serious AEs, and discontinuations from the trial due to AEs. AEs related to smartphone use may include dizziness, fatigue, or headache and were anticipated as potential AEs reported by the participants [63,64]. Previous research did not assess AEs related to EFMT use [35,36]. Participants were required to be on ADT monotherapy, and the safety profile of ADTs used for MDD is well understood. Based on the known risks of ADT and smartphone use and a preliminary understanding of the risks of EFMT use, no additional risks were anticipated as related to CT-152 use. All participants were assessed weekly for both anticipated and unanticipated AEs; any participants reporting serious AEs were monitored.

Participant adherence was monitored weekly. Furthermore, the number of CT-152 or sham sessions completed over the entire trial course was recorded.

In addition to masking the participants and trial sites for the treatment assignment, masked independent raters conducted the MADRS assessment by telephone. The independent raters had no access to trial procedures, treatment assignment, trial data, or clinical information other than what was solicited for the assessment. In addition, both independent raters and participants were instructed to avoid discussion of the DTx content and other aspects of the trial.

To minimize potential bias, trial sites had designated administrators of the other efficacy assessments. Administrators had limited contact with the participants and did not perform adherence checks. Participants completed self-reported outcomes independently and remotely.

The hypothesis of this trial—CT-152 would lead to greater improvement in MDD symptoms versus the sham DTx—was masked, and participants were informed that either DTx may be effective. This approach was intended to optimize adherence in both arms and limit the risk of participant unmasking and associated expectation biases. At the end of the study, participants were informed of the trial hypothesis, but were not informed of which DTx they had received.

CT-152 meets the definition of a nonsignificant risk device per 21 Code of Federal Regulations § 812.3(m) and a low- to moderate-risk device per Software as a Medical Device [65,66]. The masking approaches used in this trial were only used because (1) CT-152 posed minimal risks, (2) masking did not adversely affect the participants, and (3) the trial could only be conducted by withholding the information on participant treatment assignment.

The initial sample size was calculated to detect a 3-unit difference between the CT-152 plus ADT and sham DTx plus ADT in the change from baseline in MADRS total score, with 85% power at a 2-sided α=.05 level. The resulting sample size was 324 evaluable participants in total (162 in each arm). The full analysis set was defined as all randomized participants who received at least 1 CT-152 or sham DTx session and had at least 2 MADRS total score assessments, 1 of which was at baseline. To compensate for participants who failed to have evaluable assessments of MADRS total score in the full analysis set sample (estimated at up to 10% of all participants), a total of approximately 360 participants (180 in each arm) were planned to be randomized. The trial randomized eligible participants 1:1 across 54 sites. The sample size at any single trial site was capped at approximately 15% of the total number of randomized participants. Randomization was stratified by the trial site.

Due to the limitations of applying assumptions on the treatment effect size, and to ensure adequate power of the trial, an unmasked interim analysis was conducted by a data monitoring committee, whose members recommended continuing the trial with the planned sample size of 360 participants (actual enrollment included 386 participants).

Baseline and demographic characteristics, including age, sex, race, and ethnicity, were summarized using descriptive statistics.

This was the first investigation of CT-152 as an adjunct to ADT for the treatment of depression. CT-152 has both cognitive-emotional training and behavioral intervention components, which is a novel combination compared with other digital tools designed to support patients with MDD currently available under FDA enforcement discretion [34]. This protocol incorporated aspects of conventional phase 3 pharmaceutical trials, including using a randomized controlled design and gold-standard scales for evaluating effectiveness. In addition, the safety analyses were intended to evaluate if CT-152, designed to provide a clinical intervention, had a low safety risk and was well tolerated by participants. MIRAI also considered factors that are not typical for DTx study design. The use of a sham DTx comparator minimized the potential that any changes between groups were because of CT-152 and not due to the demand characteristics of using a DTx. Bias related to knowledge of study treatment assignment was managed through robust masking procedures. Adherence was evaluated for both CT-152 and the sham DTx arms.

Although DTx for mental disorders may be less likely to lead to AEs than pharmacotherapy, AEs, such as dizziness, fatigue, and headache, that are associated with mobile device use may also be associated with DTx [63,64,68]. Therefore, it is important to assess the safety of DTx. Other current clinical trials for DTx for MDD do not assess safety outcomes [69]. Among 11 protocols for clinical trials assessing DTx for MDD published recently, only 1 mentions assessing AEs along with suicide risk [70]; 5 others assess suicide risk only [71-75]. Unlike the pivotal trials for current FDA-cleared DTx for mental disorders, the MIRAI trial assessed safety outcomes [76-79]. CT-152 was expected to be well tolerated due to the digital delivery of therapy and the mechanisms of action of its core therapeutic modules.

In trials for pharmacological treatments, masking is typically achieved through a placebo that resembles the treatment being tested. In RCTs for DTx, having a placebo that allows for masking can be challenging [50]. Of 11 clinical trial protocols published in the past few years that evaluated digital interventions for MDD, only 5 involved masking; 1 of the 5 trials masked both participants and interviewers [80]. When a participant is waitlisted or given treatment as usual during a trial of digital intervention, they may become aware of not receiving the treatment under study, which could impact their involvement and expectations [81]. In conclusion, there are masking advantages to using a sham comparator.

The variability in comparator arms in DTx trials can lead to challenges in understanding the effects of digital treatments as a class and relative to one another. In total, 3 meta-analyses of digital interventions for depression found lower effect sizes in studies of smartphone interventions that included comparators that controlled for time and attention but did not provide therapeutic intervention, compared with those studies with inactive comparators or waitlist controls [81-83]. This suggests that smartphone use alone may confer psychological benefits [46,81-84]. Using a digital comparator may raise the evidentiary standard for evaluating a digital treatment [83,84]. Given these recent findings and the potential confounding factors when using a nondigital comparator for the control group (eg, waitlist or treatment as usual), a digital comparator is important for a more unbiased and accurate understanding of the effectiveness of a DTx [81-83].

The MIRAI trial design is innovative because it included a sham DTx as the comparator, which helps to ensure that the comparison reflected a true change in symptoms (if any), rather than a nonspecific benefit related to digital interventions more broadly [46,81,82]. The sham DTx more easily allowed for masking, with less risk of participants and investigators becoming unmasked.

An important consideration for any treatment is understanding the relationship between improvement in patient symptoms and increased patient adherence to the treatment; this is a novel consideration for digital interventions [85,86]. For MDD specifically, among 11 study protocols involving DTx for MDD, 9 protocols described adherence assessments [80]. While data on adherence were collected, none of the protocols defined adherence or how it would be evaluated.

A systematic literature review on adherence and engagement with digital interventions for MDD defined adherence as the percentage of sessions completed (the number of sessions completed divided by the total number of sessions) and found that only 38 out of 94 studies reviewed had this information available [80]. Their findings highlight the importance of establishing a standardized metric for adherence to digital interventions to better compare results across studies and different types of interventions [80]. The MIRAI study was designed to collect data that can be used to measure the percentage of sessions completed, which will contribute to the body of literature measuring adherence to digital interventions in a way that can be compared across studies.

Another novel consideration for DTx that does not often apply to pharmacological treatments is the consideration of patient satisfaction with and the ability to easily use the intervention. This can include features such as an easy onboarding practice, personalizing app content, and text reminders for patients [87]. CT-152 includes text messages to remind participants of what they learned and to engage them with treatment. In addition, MIRAI had both a participant and an HCP satisfaction survey to gain further insights into how easy CT-152 was to use and how satisfied participants were with their experience.

While this study design addressed many common concerns about trials for digital interventions, it had limitations, such as not including adults younger than 22 years or older than 65 years. Therefore, potential benefits for these populations will need to be explored in future research [88]. The inclusion criterion for MDD severity was a Hamilton Rating Scale for Depression, 17-item score of ≥18, potentially excluding participants with milder MDD. A further limitation is that the MIRAI trial assessed CT-152 as an adjunct to ADT, and thus the results cannot be generalized to using the digital intervention without ADT.

There were also potential limitations related to remote trials and digital interventions, such as limited access to technology or low digital literacy. In this study, participants were required to own a smartphone and had to complete an onboarding module to show they had an interest in study participation and the skills necessary to use CT-152 or the sham DTx. These requirements limited the potential for low digital access or literacy to impact participation.

In addition, a common occurrence in trials for digital interventions or that use remote data collection is participant attrition [89,90]. Although most clinical trials have some participants discontinue or be unresponsive to follow-up, this is more likely to happen in studies involving DTx [89,90]. Participant attrition was evaluated after the trial but was expected to be similar to what is seen in the literature. Participant engagement and adherence to digital interventions were also concerns [80]. Both CT-152 and the sham DTx had text messages designed to encourage participant adherence and engagement throughout the study. To encourage overall continued adherence to the study program, participants were informed before enrolling that remote visits would be required, and they reviewed the visit schedule with trial site staff once randomized to a treatment arm. Adherence checks were also performed during the treatment period and adherence was monitored remotely through the DTx. Any signs of nonadherence were addressed by the investigator to remind and encourage participants to continue adhering to the study program.

DTx has the potential to address many barriers to MDD treatment, including increasing access to more treatment options, allowing for remote access, easily fitting into a patient’s schedule, and easing stigma concerns [26-28,30]. However, it is important that the effectiveness and safety of DTx are evaluated in RCTs with rigorous study design [51-53].

Many considerations in conventional phase 3 trial design, such as safety end points, adherence, and the use of a sham comparator, have been incorporated into this protocol to assess the efficacy of CT-152 for MDD, although the way these considerations are assessed is often different for pharmacological treatment. Incorporating safety and adherence considerations, along with participant masking to the treatment, are not yet typical for DTx evaluation. Moreover, rigorous assessments of these outcomes will provide a more comprehensive understanding of the effectiveness of CT-152.