The flowchart (Figure 1) shows the process of recruitment and follow-up of students in the trial. Potential participants will be recruited through an online survey delivered across 3 colleges in Hyderabad. The research coordinator will visit each site and deliver presentations about the study as well as send campus-wide emails. After raising awareness about the study across the campus, the research coordinator will circulate a link to an online survey via email, which will also be posted in the student Facebook page and other social media groups. The landing page will again provide information about the study and instruct interested individuals to move forward to complete an online consent form and screening survey. This landing page will also contain the contact information of the on-campus counselor, as this information is meant to be available to all students across campus.

Those who meet criteria for clinical or subthreshold GAD will be invited to participate in the study and use an online program. Those who are interested will be informed that they will be assigned to a condition within 4 weeks. Everyone will receive their assignment email within 4 weeks following baseline, and those assigned into the 2 treatment conditions may activate their program account and begin immediately. Students in the GAD Online or Lantern conditions will have access to the intervention over a 3-month period. Students assigned to GAD Online can access content at whatever pace they choose. Students assigned to Lantern can access up to1 session per day because the program is structured to build mastery. Program access will be disconnected following this period. If students still meet clinical criteria for GAD at the end of the intervention, they will receive a referral to visit the on-campus counselor. Students will also be given a list of free, publicly available online self-help resources for anxiety (eg, Mental Health Online, ThisWayUp). Continuation in the guided intervention will not be possible due to resource constraints around program coach availability in the context of this research trial. Assessments will be conducted completely online. At each assessment time point, students will receive an email with a link to an online survey hosted on Qualtrics, an online survey platform with industry-standard data security measures.

Eligible participants will be current students of the colleges, all of which are English medium, aged 18 or older who provide their email address, informed consent to complete the study, and meet criteria for clinical or subthreshold GAD. This age range was chosen because individuals younger than 18 require parental consent for participation in research. Based on self-report responses in the online survey, students are categorized into clinical, subthreshold, and asymptomatic for GAD. Those who meet DSM-V diagnostic criteria for GAD as measured by criterion-based scoring of the 4th edition of the Generalized Anxiety Disorder Questionnaire (GAD-Q-IV) [42] are classified as clinical, or more realistically,

“a probable diagnosis of GAD based on self-report.” To note, the diagnostic criteria did not change from DSM-IV to DSM-V, and therefore the GAD-Q-IV questions can still be used to assess DSM-V GAD criteria. Those who score 5.7 or above using dimensional scoring of the GAD-Q-IV but are not clinical are classified as “subthreshold.” All others are asymptomatic. It is also important to note that clinical diagnosis should be established using a clinical interview, which is not feasible in this study, thus the use of the self-report measures. Moreover, this study’s aim is not to provide diagnoses to students but rather to use gold-standard instruments to provide appropriate “matches” between students and available resources (ie, the research study or referral).

The PTSD Checklist for DSM-V (PCL-5) [45] will be used to refer individuals who self-report clinical symptoms of PTSD to targeted intervention or clinical services. CBT for GAD has been demonstrated to improve symptoms across comorbid disorders such as social anxiety disorder and depression [47], and therefore a “pure GAD” sample is not necessary. However, PTSD requires a more specific approach, and data on CBT for GAD’s impact on PTSD are not available. A more conservative approach was followed: PTSD “probability” based on the PCL-5 (ie, scoring 38 or above) will result in a referral to seek help from the on-campus counselor and to access the free, publicly available iPhone and Android mobile app, PTSD Coach [48]. It is difficult to get a “clean sample,” so a balance between safety and generalizability was pursued. No additional symptom-based exclusion criteria were imposed, because a primary goal of this study was to connect most students to a service they might be willing to access. Once a student is connected to a program, mechanisms exist to ensure referrals are given if symptoms are not adequately reduced and/or other symptoms present themselves (eg, program coaches give referrals to those in the guided Lantern intervention and postintervention survey feedback gives referrals to those who still report clinical symptoms). Any individuals who report currently receiving mental health treatment will also be excluded from enrolling in the study and receiving a program. Following rule-out, individuals with a clinical or subthreshold GAD classification will be invited to participate in the study evaluating online interventions. Those who accept will be randomized to one of three conditions: a self-help program (GAD Online), a guided self-help program (Lantern), and a wait-list control group. A complete list of inclusion/exclusion criteria can be found in Textbox 1.

The trial protocol has been granted ethical approval by the Institutional Review Board at Stanford School of Medicine (protocol number 31629) and governing bodies at each of the participating colleges. The protocol is registered with ClinicalTrials.gov (NCT02410265).

In the online consent form, students are notified of how their privacy and confidentiality of their data will be maintained. All survey data will be gathered and stored on Qualtrics, an online survey platform with industry-standard data security measures. Students are informed that their participation in the survey or in a program is private and will not be shared with their college, their parents, their peers, etc. Furthermore, all data gathered will be aggregated and deidentified prior to use in any publications. Only the research coordinator has access to identifying student data. Program coaches assigned to work with students in the guided Lantern intervention connect with students via the secure program platform; no personal contact information (eg, email, phone number) is exchanged. As Lantern users receive email notifications when they receive a message from a program coach, Lantern has access to the student users’ emails; however, Lantern is HIPAA compliant and follows appropriate data management and sharing protocols. Students assigned to the GAD Online intervention are provided with generic account user names and passwords, meaning GAD Online never has need for or access to students’ email addresses.

After giving online consent, students complete the baseline assessment, which also serves as the screen for eligibility. Table 1 provides a summary of the measures that will be used.

Randomization will be carried out by the research coordinator after the baseline assessment. Random allocation to the treatment groups will occur within 4 weeks after the screening/baseline survey has been completed. The algorithm for random allocation will consist of a stratified block design, with stratification by level of symptoms (clinical or subthreshold) and a block size of 6. There will be 2 strata, corresponding to clinical/subthreshold symptom level. Allocation will be administered using Randomizer, a Web-based patient randomization service for clinical trials.

Risk management procedures were developed collaboratively with the participating universities, on-campus counselors, online program providers (ie, GAD Online and Lantern), and the research team. All students will be provided with the contact information for the respective on-campus counselor at the start and finish of the online survey. In this way, anyone exposed to the survey will be informed about currently accessible services. Furthermore, all universities have agreed to increase the current time of the existing on-campus counselor should student demand require it. In addition, for any individual in the guided intervention (Lantern), program coaches will monitor for worsening symptoms and express student preference to receive more intensive treatment and provide a referral to the on-campus counselor as necessary. Additionally, if a student in the guided program indicates thoughts of harm to self or others, the coach will notify the coach clinical supervisor (a clinical instructor at Stanford and licensed mental health care professional in the United States) via email and the campus-specific, on-site counselor via phone. Both will help determine how best to manage the situation, and the on-campus counselor will intervene and reach out to the student as necessary. In the consent, students are informed that confidentiality can be breached should the student indicate harm to self or others. Those in the unguided intervention will be explicitly informed that their actions in the program are not monitored and that they should contact the on-campus counselor should they want live support. They will again be reminded of the contact information of the on-campus counselor when they are assigned into this unguided intervention.

The sample size estimation is based on detecting differences in GAD symptom severity change (measured using dimensional scoring of the GAD-Q-IV) from baseline to post- and follow-up assessments between each active conditions and the wait-list control group. Treatment trials of online, CBT-based, guided self-help interventions for GAD have found an average large effect size of roughly 1.07 relative to a wait-list control group using the GAD-Q-IV [38]. In this study, more than half of the treated participants had recovered according to a structured interview, suggesting that a treatment effect size of 1.07 is clinically meaningful. Treatment trials of purely self-help interventions for subthreshold and clinical anxiety have found smaller yet still promising effect sizes ranging from 0.62 to 0.84, respectively, relative to a wait-list control group [31].

To enable detection of difference in mean change between each active intervention and the control group, the sample size is powered to detect at least a medium effect size of 0.5. With 80% power and assuming a 5% significance level, a minimum total sample size per group of 64 is required. Assuming the GAD-Q-IV SD is 3.46 [49], the proposed sample size provides 80% power to detect a between-group effect corresponding to a GAD-Q-IV change of 1.73 points (or a Cohen’s d of 0.5 or greater). By targeting 100 students recruited per condition (N=300), even allowing for nearly 30% attrition, which is a possibility in Internet-based intervention trials, GAD-Q-IV mean differences between each active condition compared with the wait-list control condition will be detected with sufficient power. The difference in mean change between the two active conditions is not a primary comparison, as they are two different programs.

The Lantern Anxiety Program is a cognitive-behavioral intervention (CBT) that can be accessed via any Internet-enabled computer, mobile phone, or tablet via the Go Lantern website (Figure 2). The program includes psychoeducational content, interactive tools and exercises, symptom monitoring, and an online program coach who can monitor a user’s progress in the program and provide personalized feedback and encouragement via in-program messaging and voice calls. The program is based on an evidence-based, 14-session CBT for GAD intervention, developed by Dr Michelle Newman [36,50,51]. CBT for GAD has been demonstrated to produce the largest effect sizes when compared to other therapy conditions such as analytic psychotherapy and nondirective therapy [52]. The CBT for GAD treatment includes (1) applied relaxation training, which involves the identification of early cues of anxiety, learning the skills of progressive relaxation (PR) and other relaxation techniques and learning how to apply relaxation to anxiety cues (called “applied relaxation,” [AR]); (2) imaginal rehearsal via self-control coping desensitization, using AR and alternative, nonanxious coping thoughts; and (3) cognitive therapy methods to help change how clients perceive, interpret, and believe, so that they will see less threat in the world and feel more confidence in their abilities to cope with the future.

Content is divided into 8 units, each with 5 sessions, resulting in 40 daily 10-minute sessions. The 8 units comprise an introduction to anxiety, automatic thoughts, cognitive reframing, introduction to behavior change, imaginal exposure, situational exposure, mindfulness, and habit formation. There are both cognitive (eg, worry tool, mindfulness) and behavioral (eg, progressive muscle relaxation, deep breathing, guided imagery) techniques taught in each module and always accessible thereafter. Users can access one session per day, with the next session unlocking only after the prior session is completed; therefore, less engaged users may not be exposed to all content. The program and program coach prompt users to schedule reminders to facilitate self-monitoring, encourage completion of scheduled activities, and promote use of tools and techniques in the program. Users can schedule reminders to complete sessions and to use techniques or self-monitoring entries at their desired frequency. They are prompted to do so after learning a new technique. Coaches can also schedule reminders for themselves to send users messages.

Students using Lantern will be guided through the program by mental health workers in India who have been trained to serve as online program coaches. To be eligible to be a coach, individuals must have some background and/or experience in psychology, have at least 6 months of supervised, in-person counseling experience, own a personal computer or smart device and check email daily, be comfortable with technology, and have a minimum of 3 hours per week to dedicate to coaching. A total of 20 potential program coaches were recruited, all of whom had completed a 6-month training course in the basic principles of psychological counseling from the Hyderabad Academy of Psychology based in Hyderabad, India.

A pretraining assessment was used to assess existing skill levels and inform training content development. To begin, 4 weeks of remote training covering basic foundational concepts was conducted. Manuals and papers reviewing CBT for GAD, panic disorder, and social anxiety were assigned, and a virtual discussion session was conducted each week via online collaboration platform GoToMeeting. In addition, coaches were provided a demo user account of the Lantern Anxiety Program so that they could familiarize themselves with the intervention content. Following remote training primarily reviewing content and basic concepts, a 3-day in-person training was conducted in Hyderabad, India, in December 2014. During this training, the Lantern coaching platform was introduced and coaches learned how to effectively interact with the coaching dashboard. The coaching dashboard enables coaches to monitor the activity and engagement of the students assigned to them as well as to send messages to them. Coaches were instructed on guided self-help coaching best practices, introductory phone call and messaging protocols, motivational interviewing, and risk management procedures.

Following training, coaches were assessed for comfort with technology via direct observation of their interaction with the Lantern coaching dashboard during interactive role plays. Coaches who demonstrated low technology fluency and experienced usability issues interacting with the coaching software in comparison with the group were asked to attend additional training and practice sessions to continue. Behavioral rehearsal tasks were used to evaluate coaches’ fidelity to the coaching protocol. Coaches were required to complete a mock introductory phone call with a control student (played by the research coordinator or volunteer students following a script), and fidelity to the phone call procedural checklist was assessed. Coaches were also required to respond to message prompts designed to assess their mastery of core competencies of CBT for GAD in the context of a guided self-help intervention. Their responses were rated against a messaging best practices checklist that was developed by the authors. Finally, coaches had to complete an established “e-therapist assessment” developed by the National eTherapy Centre (NeTC) at Swinburne University of Technology [53] and score at least 80%. Of the 20 trainees who began training, 65% (n=13) cleared the post-training evaluation and moved forward to become program coaches for this trial.

Throughout the intervention period, coaches will have on-going monitoring and support. Coaches must attend a weekly, live 1-hour supervision, during which coaches can present any challenging cases to the group and receive feedback from the coach clinical supervisor (a clinical instructor at Stanford and a licensed mental health care professional in the United States). Coaches will also be asked to submit 1 challenging message exchange per week to a clinical psychology doctoral student for individual feedback on messaging style and content. Finally, coaches can post questions and messages on a group listserv to get feedback from the supervisor as well as from peers.

The GAD Online program is also CBT based and can be accessed via any Internet-enabled computer, mobile phone, or tablet via the Mental Health Online website (Figure 3). Mental Health Online is a suite of online mental health programs developed and maintained by the NeTC at Swinburne University of Technology and funded by the Australian Federal Government Department of Health. The GAD Online program delivered in this trial, which was previously reported on as an Anxiety Online module [33], provides psychoeducation about GAD and techniques to reduce GAD symptoms (eg, relaxation, challenging thoughts, coping with worry, problem solving). The content covered in this program is comparable to that delivered in the Lantern GSH program. Content is divided into 12 modules, with 1 module per week being the suggested use; however, users can progress at their own pace. Unlike the design of the Lantern program, users can access other modules in the GAD Online program even if they do not complete the first module. Students are explicitly told they are not monitored and should contact the on-campus counselor should they require real-time support from a professional.

Students randomized to the wait-list control condition will complete surveys at the postintervention time point (3 months) and the follow-up time point (9 months). Following successful completion of these surveys, they will be allocated to one of the online interventions. Students will be assessed for whether they sought help from the on-campus counselor at each time point. The counselor’s contact information is available to all students across the campus.

The primary outcome is GAD symptom severity as measured by dimensional scoring of the 4th edition of the Generalized Anxiety Disorder Questionnaire (GAD-Q-IV) [42]. GAD caseness, as measured by categorical scoring of the GAD-Q-IV, is a secondary measure. The GAD-Q-IV is a 9-item self-report measure designed as an initial screen for the presence of GAD based on the DSM-IV. The GAD-Q-IV showed 89% specificity and 83% sensitivity when compared to structured interview diagnoses of individuals with GAD, social phobia, panic disorder, and a nonanxious comparison group. The GAD-Q-IV has demonstrated good test-retest reliability in a college sample over a 2-week assessment, with 92% of the sample showing stability across time with respect to GAD diagnosis [42].

The Patient Reported Outcomes Measure Information System (PROMIS) Emotional Distress from Anxiety measure [56] is a reliable 8-item self-report measure to evaluate emotional distress from anxiety. This measure was included as a potential global outcome measure of improved quality of life.

The Penn State Worry Questionnaire (PSWQ) [57] is a 16-item self-report measure of the frequency and intensity of worry. The PSWQ has been shown to distinguish individuals with GAD from individuals with other anxiety disorders [58]. This measure was included to assess change in worry, a primary symptom of GAD.

The Panic Disorder Self Report (PDSR) [44] is a 22-item self-report measure designed to diagnose panic disorder based on DSM-IV criteria. The PDSR showed 100% specificity and 89% sensitivity when compared with clinician-based ADIS-IV-L diagnoses of individuals diagnosed with panic disorder, GAD, social phobia, and a nonanxious comparison group. This measure, along with the other measures of anxiety disorders and depression, was included to assess both prevalence of this type of anxiety and the impact of the active interventions on symptoms associated with typically co-occurring disorders.

The Social Phobia Diagnostic Questionnaire (SPDQ) [43] is a 10-item self-report measure designed to diagnose social phobia based on DSM-IV criteria. The SPDQ showed a specificity of 82% and a sensitivity of 85% when compared with clinician-based structured interview diagnoses of individuals meeting criteria for social phobia, panic disorder, and a nonanxious comparison group.

The Post-Traumatic Stress Disorder Checklist for DSM-V (PCL-5) [45] is a 20-item self-report measure that assesses the 20 DSM-V symptoms of PTSD. It was designed to monitor symptom change during and after treatment, screen individuals for PTSD, and make provisional PTSD diagnoses.

The LEVEL 2—Repetitive Thoughts and Behaviors—Adult (adapted from the Florida Obsessive-Compulsive Inventory [FOCI] Severity Scale [Part B]) [59] assesses obsessive-compulsive disorder symptoms. This emerging measure of the DSM-V was developed to be administered at the initial patient interview and to monitor treatment progress.

The Depression Anxiety Stress Scales-Short Form (DASS21) [60] is a 21-item self-report measure designed to measure negative emotional states of depression, anxiety, and tension or stress. The DASS is divided into 3 self-report scales. The depression scale assesses dysphoria, hopelessness, devaluation of life, self-deprecation, lack of interest/involvement, anhedonia, and inertia. The Anxiety Scale assesses autonomic arousal, skeletal muscle effects, situational anxiety, and subjective experience of anxious affect. The Stress Scale is sensitive to levels of chronic nonspecific arousal, assessing difficulty relaxing, nervous arousal, and being easily upset/agitated, irritable/over-reactive, and impatient. All scales have been shown to have high internal consistency and to yield meaningful discriminations in a variety of settings.

The Insomnia Severity Index is a 7-item self-report questionnaire to assess insomnia severity [61]. It has adequate internal consistency and is a reliable self-report measure to evaluate perceived sleep difficulties. This measure, along with the other measures of general functioning, was included to assess the impact of the active interventions on functional impairments typically associated with GAD.

The Strengths and Difficulties Questionnaire [62] assesses self-reported difficulties with emotions, concentration, and relationships. It is a reliable short-form, self-report assessment of inattention, peer relationships, and pro-social behavior, factors that, if improved, could indicate positive impact of the active interventions.

The PROMIS Satisfaction with Social Roles and Activities measure [63] is an 8-item self-report assessment of satisfaction with one’s ability to perform daily activities and meet the needs of various relationships.

The Kessler Distress Measure [64] assesses nonspecific distress. The 10-question (K10) self-report scale has good precision as well as consistent psychometric properties across major sociodemographic subsamples. This measure was included as an assessment of a global outcome of reduced general distress.

The General Perceived Self-Efficacy scale (GSE) [65] is a 10-item self-report measure that assesses beliefs in one’s capability to handle new and difficult tasks and adaptive challenges after experiencing stressful life events. This measure was included because enhanced self-efficacy is considered one of the main psychological benefits of self-help interventions [66].

Chi-squares (categorical variables) and t tests (continuous variables) will be used to compare demographic variables and baseline scores on the outcome measures for the 3 groups. Findings will be reported according to the CONSORT guidelines [68], including a trial flowchart. This will include total students assessed for inclusion and exclusion criteria within period of screening, number of students meeting inclusion or exclusion criteria, number screened for eligibility, number agreeing to enter the intervention trial, and number refusing or excluded (with reasons). The number continuing through the trial, actively withdrawing, and passively lost to follow-up will be shown by arm. The outcome measures will be summarized at baseline/screening, at 3-month and 9-month follow-up by the intervention arm and overall.

A 3 condition × time mixed-design ANOVA will be conducted to examine both primary and secondary hypotheses. Using a mixed-design ANOVA will test for differences between 2 or more independent groups while subjecting participants to repeated measures. The model is a type of mixed effect model, with the fixed effects factor being a between-subjects variable and the random effects factor being a within-subjects variable. Follow-up paired comparisons will be conducted to specifically assess if there is significant differential change in GAD symptoms between each active condition and control from the baseline to the postcondition and follow-up assessments. Within- and between-group effect sizes will be calculated using Cohen’s d (based on the pooled standard deviation), specifically contrasting each active condition with the control condition for the main outcome.

Primary analyses will be undertaken using an intention-to-treat (ITT) approach, including all participants randomized regardless of treatment actually received, program engagement, or withdrawal from the trial. Using a mixed model design will allow us to include participants with missing data, which is likely to occur in large Internet-based trials.

Study site will be included as a covariate as this might have an influential effect given that the 3 sites were selected primarily based on accessibility versus between-site comparability. If some sites have poor recruitment, the data will be combined.

Potential mediators and moderators of the efficacy of the intervention will also be explored. For example, level of program engagement, defined by number of sessions completed and number of days logged on to the program, will be evaluated for mediation of outcomes.

Additionally, given that randomization was stratified between clinical and subthreshold populations, whether symptom severity influences the efficacy of the intervention to reduce GAD symptoms can be explored. Additional predictors of outcome that will be explored include medication use, motivation to work on anxiety, belief in efficacy of online programs, perceived social support, and grade point average. Medication use will be controlled for in the analyses if the sample is large enough; otherwise, the results will be provided qualitatively.