Digital CBT-I was delivered using SleepioRx (Big Health, Inc). SleepioRx was cleared by the US Food and Drug Administration for the treatment of insomnia in August 2024 and was formerly known as Sleepio. SleepioRx delivers CBT-I and includes sleep consolidation, stimulus control, cognitive therapy, relaxation techniques, and SHE. Content is delivered in 6 engaging levels (5‐15 min each). The program is interactive, automated, and tailored to participant progress. Participants receive reminders to complete each level. In March 2025, SleepioRx was replatformed with the new app, only accessible via smartphone and tablet applications (iOS [Apple Inc] and Android [Google]), whereas the previous version was also available via a web app. All therapeutic content, techniques, and sequencing were unchanged.

Engagement is monitored by the study team. Each week, Big Health, Inc sent a file detailing each participant’s use of SleepioRx (eg, session start and end time and number of sleep diaries completed). Unmasked study staff members uploaded this file to REDCap (Research Electronic Data Capture) to monitor adherence to intervention protocols. Unmasked staff members contacted participants who were overdue ≥4 days on a session to identify and troubleshoot barriers.

Additionally, participants in this arm received supplemental content about prenatal, postpartum, and infant sleep that was previously evaluated in a trial of therapist-delivered CBT-I for prenatal insomnia [11]. This content was delivered via 3 separate emails at randomization, 36 weeks gestation, and 3 months post partum, and included sleep hygiene recommendations specific for pregnancy, suggestions for improving postpartum sleep, instructions on how to adjust the time in bed prescription during the postpartum period, psychoeducation about infant sleep, and tips for promoting healthy sleep development among babies 0‐3 months and 3‐6 months of age.

Our research question is whether digital CBT-I outperforms a clinically relevant comparator. SHE has face validity, is frequently used in clinical practice [23,24], and is rated as credible as CBT-I [25]. SHE matches the CBT-I intervention in delivery format (digital), frequency (weekly), and number of sessions (n=6). SHE was delivered via link to a REDCap survey page. It includes information about sleep hygiene, the impact of sleep on performance, healthy sleep habits, lifestyle influences on sleep, and creating a sleep-friendly bedroom. At the end of each session, participants were asked: “What key points from today’s session do you want to keep in mind this week?” They received via email a PDF of the handout and their response after submitting the questionnaire. Study staff members did not monitor SHE adherence. As a result, participants in the digital CBT-I condition had the opportunity for increased program support and staff contact.

Concomitant nonstudy treatment for insomnia or depression is not prohibited; use of nonstudy treatments is tracked in a nonstudy treatment questionnaire at each assessment time point.

Table 2 displays all primary, secondary, and exploratory measures, time points, and their allowable response windows. Per SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) 2025 guidelines [26], we focus the following description on primary and secondary outcomes.

We use the depression module of the Structured Clinical Interview for DSM-5 (Diagnostic and Statistical Manual of Mental Disorders [Fifth Edition]), Research Version (SCID-5-RV) [21] to assess our primary clinical outcome, incidence of perinatal depression. This binary outcome measures whether a participant experienced a depressive episode at any point since randomization. To reduce recall bias, we conduct interviews at 6 follow-up time points. We assess current and past major depressive episodes with the stem question revised from “Have you ever...” to “Since the last interview on [date], have you ever…”

We use the Edinburgh Postnatal Depression Scale (EPDS) [27] to quantify depressive symptom severity [27]. Participants complete the EPDS at each follow-up time point, allowing us to measure changes in depressive symptoms from baseline to 12 months post partum. The EPDS is a 10-item self-report measure that omits depressive symptoms that can be conflated with normal perinatal symptoms. It is frequently used and validated to assess depressive symptom severity during pregnancy [28,29], with estimates of sensitivity and specificity for detecting major depression ranging from 70%‐100% and 74%‐97%, respectively, depending on cutoff and sample [30]. Among postpartum samples, estimates of sensitivity and specificity are 86% and 78%, respectively [27]. We also use the Patient Health Questionnaire (PHQ-9) to quantify depressive symptom severity [31]. This 9-item self-report measure is frequently used in research and clinical practice.

We assess suicidal ideation and anxiety because perinatal insomnia and depression are associated with suicidal ideation and anxiety [32-34]. We use the Since Last Contact version of the Columbia Suicide Severity Rating Scale [35] to assess suicidal ideation, intensity of ideation, and suicidal behavior since the last interview.

We use the Generalized Anxiety Disorder 7-item scale (GAD-7) to assess anxiety symptom severity [36]. The GAD-7 is a 7-item self-report measure with total scores ranging from 0 to 27. Among pregnant and postpartum women, a cutoff of 13 on the GAD-7 has a sensitivity of 61% and a specificity of 73% for detecting diagnoses of generalized anxiety disorder [37].

We assess insomnia symptom severity using the Insomnia Severity Index (ISI), a 7-item self-report measure that assesses difficulty initiating or maintaining sleep, satisfaction with sleep, impairment, distress, and the extent to which others have noticed symptoms over the last 2 weeks [38]. The ISI had high internal consistency in a clinical sample seeking treatment for insomnia (Cronbach α=0.91) [17]. At screening, we excluded women who were unlikely to meet criteria for insomnia disorder using a cutoff validated for identifying participants in clinical trials (ISI<11) [17]. At baseline, we included those with elevated insomnia symptom severity, defined as ISI≥10 [17], which has been validated for detecting DSM-5 insomnia disorder during pregnancy [19].

We also use the consensus sleep diary to assess sleep metrics, such as average sleep onset latency, wake after sleep onset, sleep efficiency, and sleep duration. The consensus sleep diary is the gold standard measure of subjective sleep. Research suggests that the consensus sleep diary is associated with subjective and objective measures of sleep, differentiates good sleepers from those with insomnia disorder, and is sensitive to improvements over the course of CBT-I treatment [39].

We use the EPDS to assess our primary moderator, baseline depressive symptom severity. We selected baseline depressive symptom severity as our primary moderator based on our strong preliminary data, as well as similar findings in a nonperinatal population [14].

People interested in participating completed an electronic consent form describing the screening survey, baseline sleep diaries, and orientation session. Those who consented to the screening process completed the online screening survey to determine potential eligibility and provided demographic data. Those who were potentially eligible (eg, ISI≥11) were invited to complete baseline sleep diaries.

Each morning for 7 consecutive days, potential participants received a link to a brief online sleep diary. The sleep diaries served as a behavioral run-in to identify participants who are likely to complete study measures after randomization. Those who completed at least 4 of 7 sleep diaries were invited to complete an orientation session. In addition, for those who were eventually included in the study, the data provided baseline sleep metrics (eg, sleep onset latency and sleep duration).

In a 1-hour orientation session via videoconference, study staff members discussed the importance of the study question, the required commitments, what to expect if randomized to digital CBT-I or digital SHE, the rationale for random assignment, and the detrimental effects of attrition bias on the success of the study.

Potential participants were required to provide documentation of their pregnancy, and could opt to do so at the end of the screening survey or after the orientation session. Acceptable forms of documentation included an ultrasound or an after-visit summary with the participant’s name, estimated due date or gestational age, and appointment date. Documentation was reviewed by research coordinators for validity before scheduling the consent and clinical interview. For example, research coordinators confirmed that the participant’s estimated due date and gestational age reported on the screening survey matched the information provided in the pregnancy documentation.

The day before the consenting and clinical interview, participants completed the ISI; those with an ISI total score less than 10 were excluded.

Study staff administered the revised version of the Structured Clinical Interview for Sleep Disorders overview and insomnia disorder sections. Participants who did not meet the criteria for insomnia disorder were excluded, and the appointment was ended. All others reviewed the study consent form with staff, and those who consented to participate completed the remainder of the clinical interviews. Staff ended the interview if a participant met an exclusion criterion (eg, current major depressive episode). If staff were uncertain about whether a participant met criteria for an excluded disorder, the interview was completed in its entirety, and the segments in question were reviewed by the coauthor (R Manber) before informing the participant of their eligibility.

Participants who were eligible after completing the clinical interviews completed baseline self-report measures online via REDCap before randomization (Table 2).

Randomization was completed during a telephone call with an eligible participant. At the beginning of the call, staff confirmed the participant’s willingness to complete either intervention and follow-up measures regardless of intervention assignment. After receiving an affirmative response, staff revealed the intervention assignment, informed the participant, and provided information about how to access their intervention. Study staff encouraged the participant to set aside time once a week to complete their intervention sessions. Study staff reviewed the schedule of follow-up measures and instructed the participant not to reveal their intervention assignment to masked follow-up interviewers. Before ending the call, participants completed a brief measure of treatment expectations and credibility. Study staff then emailed all information that was discussed on the call.

At 5 weeks post randomization (ie, halfway to the 10-week postrandomization time point), participants completed brief online measures of insomnia and depressive symptom severity. Subsequently, participants completed follow-up measures via online questionnaires and clinical interviews via videoconferencing at 10 weeks post randomization, 36 weeks of gestation (unless randomized after 24 weeks of gestation or if they gave birth before 35 weeks of gestation), and every 3 months after delivery (ie, 3, 6, 9, and 12 months post partum).

At enrollment, participants were instructed to make immediate contact with study personnel with any questions or concerns regarding worsening depression symptoms or suicidal ideation. Additionally, we provided contact information for therapeutic resources and crisis hotlines.

When completing self-report measures online via REDCap, participants who score ≥10 on the EPDS or PHQ-9 immediately view a pop-up with mental health resources and instructions to call 911 or go to the nearest emergency room if they are in imminent danger of hurting themselves. This information is also sent by email. Email triggers immediately notify study staff members about score elevations. Study staff members subsequently conduct the SCID-5-RV to determine if the participant meets criteria for a major depressive episode. Study staff recommends that any participant who meets criteria for a major depressive episode notify their obstetrician or primary care provider.

Similarly, participants who endorse the suicidality item on the EPDS (item 10 score >0) and/or the PHQ-9 (item 9 score >0) complete the Columbia Suicide Severity Rating Scale self-report survey. Information on suicide and mental health resources is automatically displayed on the same page. Email triggers immediately notify study staff members, who then follow the suicide risk assessment protocol.

Based on our preliminary data, we expect a 16% lower rate of depression among women in the digital CBT-I condition relative to women in the digital SHE condition. We conservatively power the study to detect a between-group difference of 10% in case the effect is smaller than expected. Standard power calculations for the comparison of 2 proportions show that with α set at .05, chi-square tests with 199 women per group would have power greater than 80% to detect this difference (hypothesis 1a) [40].

The sample size and number of repeated measures per participant are large enough to allow us to detect important differences in the magnitude of within-woman change in depressive symptom severity, measured by the EPDS, between the 2 conditions with sufficient power (hypothesis 1b). We will measure depression outcomes at 7 time points throughout the study period. The statistical analyses will compare within-subject changes in the EPDS over time between the 2 intervention conditions using linear mixed effects models. Sample size calculations for repeated measures studies require estimates of the magnitude of the within-subject correlation of the outcome; data from our previous trial of digital CBT-I provide us with an estimate of 0.5 for this correlation for the EPDS outcome [10]. Our previous trial also provides an estimate of 9.2 units for the within-subject variance of the EPDS measure. Standard power calculations for linear mixed effects models show that with an α set at the .05 level and with 2 conditions of 199 subjects each, we will have power greater than 80% to detect a difference in within-subject slopes of 0.04 (units of EPDS) per month between the 2 intervention groups [41]. In our previous trial, the estimated difference in slopes between digital CBT-I and standard care was 0.12.

We base sample size calculations on a statistical test to detect statistically significant mediation of the treatment effect on within-subject change in EPDS through within-subject change in ISI using linear regression. Standard sample size calculations for mediation using linear regression analysis require the specification of the variances of the depression outcome (EPDS), the treatment predictor, and the sleep mediator (ISI), the magnitude of the treatment effects without adjustment for the mediator, and the correlation of the mediator and treatment variables [42]. Data from our previous trial of digital CBT-I provide estimates for all of these quantities. Calculations [42] show that a standard test for mediation using our sample will have power greater than 80% to detect a proportion of treatment effect explained of 0.02. Using data from our previous trial of digital CBT-I for women and changes from baseline to 10 weeks, a linear regression with change in EPDS from baseline to 6 months post partum as the outcome and treatment as the single predictor yielded an estimated regression coefficient of −0.14 (units of EPDS). Adding a change in ISI to the model reduced the estimated treatment effect to −0.12 (units of EPDS), with an estimated proportion of treatment effect of 0.10, far exceeding 0.02.

We base sample size calculations on tests of the interaction of treatment condition with depressive symptoms at baseline (using an established cutoff of EPDS <10) in logistic regressions with incidence of probable major depression as the outcome. These sample size methods require the specification of several inputs that describe the anticipated prevalence of the outcome, predictors, and their relationships [43]. Calculations based on inputs from our previous digital CBT-I trial show that the sample of 398 participants (199 per group) provides 80% power to detect an interaction odds ratio (OR; OR for baseline EPDS <10 compared with baseline EPDS ≥10) of 4.1 with binary indicators for condition and depressive symptoms at baseline. Our previous trial yielded an estimated OR of 10, indicating that the detectable interaction OR of 4 is a realistic objective.

These calculations show that the sample size will allow us to detect clinically important differences in depression incidence and changes in depressive symptom severity between women in the 2 intervention conditions, as well as clinically meaningful mediation and moderation effects, with 80% power. Initially, we set our attrition rate at 20% based on our previous trial, yielding a total enrolled sample size of 498. In September 2025, with approval from the Data Safety Monitoring Board (DSMB) and the National Institute of Mental Health, we revised the attrition rate to 10% based on our observed higher-than-anticipated retention, yielding a required sample size of at least 443.

This study protocol was approved by the Institutional Review Board (IRB) in the Human Research Protection Program at the University of California, San Francisco (UCSF IRB 21‐35440). The Stanford IRB relied on the UCSF IRB as the single IRB of record. Important changes to the original protocol after trial commencement are described in Table 3.

Because this study involves no in-person visits, all consent occurred electronically. There were 3 consent forms—the screening consent, the future contact consent, and the study consent. Consent forms were written at an 8th-grade education level, used lay language, and provided information about the background and purpose of the study, scope and length of participation, study procedures, legal and ethical limits to confidentiality, clinical interview audio and videotaping procedures, risks, benefits, alternatives, ability to discontinue participation, privacy and confidentiality, compensation, and whom to contact with questions. Potential participants were assured that their decision to participate or decline participation in the study would have no effect on their current or future receipt of health care services at UCSF or affiliated clinics.

Screening consent: First, potential participants viewed an introduction page that briefly described the study. Next, the screening consent form described the enrollment procedures (ie, screening survey, daily sleep questionnaires, orientation appointment, consent, and clinical interview). Participants who declined to participate were automatically directed to the end of the survey and thanked for their time. Participants who agreed to participate then began the screening survey.

Future contact consent: After completing the screening survey, regardless of eligibility, participants were given the opportunity to provide their contact information to be contacted for future research opportunities.

Study consent: The orientation session provided another opportunity to engage participants in the informed consent process (refer to Participant Procedures and Timeline section for more details). Individuals who were interested in participating after the orientation session were scheduled for an individual appointment to complete the consent process and the clinical interview to determine eligibility. Consenting participants provided an electronic signature via REDCap and received a signed copy of their consent form via email.

All participant data were collected and stored in REDCap, a secure, HIPAA (Health Insurance Portability and Accountability Act)-compliant platform with role-based access controls, encrypted data transmission, isolated identifier storage, and full audit trails.