Our study presents a secondary analysis of the CHAMPS trial, a multicenter, 2-arm, parallel-group, pilot RCT. This pilot trial compared the cannabis harm-reduction e-intervention, CHAMPS, in conjunction with EIS (CHAMPS+EIS), to the standalone EIS in young adults with FEP who use cannabis. CHAMPS was co-designed with people with lived experience through a patient advisory committee [40]. Consenting participants were randomly allocated in a 1:1 ratio to the intervention or control group, using a digital, stratified, permuted block design with varying block sizes implemented by the CRCHUM data management team. Randomization was stratified by CUD status and sex. Because of the nature of the intervention, the research staff and clinicians involved in data collection were unblinded; however, the research staff conducting the analyses remained blinded.

This study trial is reported in accordance with the CONSORT-EHEALTH (Consolidated Standards of Reporting Trials of Electronic and Mobile Health Applications and Online Telehealth), version 1.6.1, guidelines (Checklist 1) and was registered prospectively at ClinicalTrials.gov (NCT04968275) in July 2021 [41]. The detailed protocol was described by Coronado-Montoya et al [40], and the pilot feasibility and acceptability analysis was published by our team [34]. CHAMPS was conducted in accordance with the Good Clinical Practice (GCP) guidelines and the Declaration of Helsinki, alongside Canadian standards (Tri-Council Policy Statement and Health Canada Division 5) and applicable provincial and institutional ethical guidelines (20.433). Approval for the protocol was obtained from the local research ethics boards (REBs) at each participating site. The study started in December 2021 and was completed in October 2023. No major changes were made to the protocol after trial initiation, other than adding 2 new sites.

Informed consent was obtained from each participant prior to enrollment, and the informed consent form was approved by each REB, allowing for secondary analysis without additional consent. All participant-related information, including case report forms, assessments, and reports, was kept strictly confidential, and participants were identified only by means of a numeric study identifier specific to each participant. All computerized databases identified participants using numeric codes and were password-protected. The link between participants’ identification numbers and their names was kept in a password-protected document accessible only to research staff, and the document is stored on the CRCHUM secure server for a duration of 10 years. No identification of individual participants is possible from any figures or supplementary materials presented in this manuscript. Participants were compensated CAD $30 (equivalent to US $21) per visit and CAD $150 (equivalent to US $106) in total for completing the study.

Young adults with FEP who were using cannabis and were open to changing their cannabis use practices (N=101) were recruited into the study through referrals from EIS psychiatrists or case managers. Recruitment took place at 6 specialized EIS sites for the treatment of early psychosis in Canada and is described in Coronado-Montoya et al [40]. Inclusion criteria included being 18 to 35 years old, having a diagnosis of any psychotic disorder according to DSM-5 (Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition) criteria, being followed for a minimum of 3 months at an early psychosis clinic, using cannabis (at least once in the past month), being interested in changing cannabis-related practices, and understanding French or English. Exclusion criteria were concurrent enrollment in another cannabis-related intervention or current pursuit of treatment for CUD aimed at reducing or ceasing cannabis consumption. Participants identified as seeking treatment for CUD were instead directed to a concurrent clinical trial [42] or to usual clinical care. Digital literacy was not assessed. Individuals were considered eligible if they met all the inclusion criteria and none of the exclusion criteria. Eligible participants were randomly assigned (in a 1:1 ratio) to either the intervention arm (n=51) or the EIS-only arm (n=50). Sample size justification is described elsewhere [40]. In this exploratory analysis, given that 5 participants (intervention arm, n=5) never started the intervention, data from the remaining 96 participants were used in a modified intention-to-treat (mITT) analysis since participants who did not initiate the intervention could not contribute to post-initiation data needed to estimate changes in outcomes over time [43].

Participants accessed the CHAMPS e-intervention through a mobile health (mHealth) application and completed 6 psychotherapeutic, fully automated online modules within 6 weeks, followed by an automated booster session 4 weeks after the final module. Each module was designed to be completed in less than 10 minutes. Participants were guided to reflect on their cannabis use practices and their impact on their lives (modules 1 and 2). Module 3 provided information on harm-reduction strategies, while module 4 taught goal-setting skills. Modules 4 to 6 helped participants set personalized harm-reduction–related goals. The booster module guided participants in reviewing the main components of CHAMPS, such as exploration of their cannabis use and their goal setting. Additional resources were available, including evidence-based articles and videos cocreated with individuals with lived experience. All participants received EIS. The complete study procedures, including detailed module content, have been reported elsewhere [40]. The pilot RCT lasted up to 22 weeks in total, including a 14-day screening period, an additional 14-day window for baseline assessments, a 6-week intervention period plus a 4-week period to complete the booster module, and a 4-week follow-up period. Follow-up assessments were conducted by research assistants at weeks 12 and 18.

Information on adverse events was collected electronically by study staff and assessed systematically. No serious adverse events were reported.

A total of 101 participants were randomized to either the intervention CHAMPS+EIS or the control group, EIS only. Given that 5 participants never started the intervention, data from the remaining 96 participants were used in our mITT analysis, as stated in the CONSORT flow diagram [41] (Figure 1).

Table 1 summarizes baseline demographics and clinical characteristics of the participants. On average, participants were 25 (SD 3.9) years old, and most had completed at least a secondary high school diploma (n=69, 72%). Participants were predominantly men (n=69, 72%), White (n=57, 59%), and reported using cannabis at least 5 days per week (n=74, 77%). Although most baseline characteristics were balanced across study arms, the overall sample distribution was asymmetric for a few variables, with a greater proportion of participants reporting postsecondary education, being White, a higher frequency of cannabis use, and men gender.

Selected baseline sociodemographic characteristics across engagement groups—low to moderate (n=18) and high (n=28)—are described in Table 2. Individuals with high engagement reported higher levels of baseline social support compared with those with low-to-moderate engagement (median 33.5, IQR 31.5-38.0) vs (median 28.5, IQR 27.3-32.8; P=.005). A greater proportion of participants in the high engagement group also had a higher educational level (23/28, 82% vs 8/18, 44%; P=.01). Other sociodemographic factors did not differ significantly between the groups after correction for multiple comparisons.

Table S1 in Multimedia Appendix 1 summarizes MPS scores, PBSM total scores, and days of reported cannabis consumption in the last 14 days, with corresponding trajectories depicted in Figure 2. Descriptive data, estimated marginal means (EMMs), and adjusted mean differences from the LMM are reported in Tables S1 and S2 Multimedia Appendix 1. Findings suggest overall lower MPS scores in the CHAMPS+EIS group at all follow-up time points, although these differences did not reach statistical significance (Tables S1 and S2 in Multimedia Appendix 1). For instance, at week 6, the adjusted mean difference in MPS between the CHAMPS+EIS arm and the control arm was 1.4 (95% CI −3.8 to 1.0), with similar trends at week 12 and week 18.

Linear regression analyses within CHAMPS+EIS found no statistically significant associations between module completion thresholds (ranging from ≥1 to ≥6 modules) and any primary outcome at week 6 after adjusting for the baseline outcome, sex, and CUD status (Table 3). Although the differences were not statistically significant, descriptive analyses indicated greater improvements among participants completing 5 or more modules for the MPS and PBSM outcomes (Figure 3). Sensitivity analyses showed findings in the same direction when the number of modules was treated as a continuous predictor (Table S3 in Multimedia Appendix 1).

In this exploratory pilot study, we examined which sociodemographic factors were associated with engagement and whether specific engagement thresholds were associated with improved cannabis-related outcomes in our digital health intervention, CHAMPS, among young individuals with FEP. Overall, participants in the CHAMPS+EIS group who completed more than 5 modules were more likely to have higher social support and higher educational levels at baseline than those who completed fewer modules. Interestingly, while descriptive analyses indicated a potential gradient of improvement with increasing module completion, no specific module completion threshold was found to be significantly associated with improved cannabis-related outcomes at the study end point. Taken together, these findings suggest that engagement with CHAMPS occurs across diverse profiles but is likely to be influenced by social support and education, and that setting a specific minimum number of completed modules may not capture the variability in how individuals derive benefit or not.

Although CHAMPS previously demonstrated overall acceptability and feasibility across a diverse sample [34], participants with more module completion (5 modules or more) tended to report greater baseline social support, a finding that aligns with the existing literature on digital mental health interventions. Higher social support (be it from clinicians or peers) is a well-documented facilitator of positive adherence to self-guided interventions by reinforcing motivation and accountability [21,23,24,54]. In other studies, individuals with higher baseline social support were also described as more prone to engage with the social elements of digital interventions, and therefore with the digital intervention itself [23]. Similarly, higher education has been repeatedly associated with increased engagement, as educational attainment could facilitate not only comprehension of psychoeducational content but also the internalization and application of harm-reduction strategies [55]. However, digital interventions, such as those delivered in self-guided formats, may require a certain level of cognitive engagement and autonomous learning, which can pose a barrier to responsiveness in the context of FEP where cognitive impairment is a core feature for many individuals [56,57]. Although the present digital intervention took this into account when designing the app, a self-guided intervention like CHAMPS may represent an additional barrier to intervention responsiveness in people with lower educational levels. Individuals with lower educational attainment may benefit from additional guided support while emphasizing the need of tailoring content to different literacy levels, including digital literacy and health literacy [58].

Although our linear regression with the predictor of module completion threshold in the intervention arm only did not yield significant results, visual inspection showed a potential gradient of improvement with increasing module completion. As described by Perski et al [23,59], this observation is consistent with the literature on digital intervention research that describes a dose-response link between higher engagement levels and optimal response. The CHAMPS intervention included personalized feedback, psychoeducational techniques, and the evidence-based use of protective strategies [40]. Participants engaged in goal-setting tasks in modules 4 and 6, and were guided to set personalized harm reduction–related goals between modules 4 and 6. These observations highlight the need to consider robust in-app strategies, such as reminders, involvement of clinicians, and other support, to help users reach a selected threshold of module or content and thereby maximize the benefits of the intervention [28].

Drawing on the theoretical foundations of the BCW [32], CHAMPS underscored the interplay among capability, opportunity, and motivation. Our results may also reflect that adopting harm-reduction behaviors is an engaging process and behavioral engagement that requires more than mere exposure to the intervention; it also involves active effort and internalization [60]. In this sense, engagement is not a passive process but one that may be facilitated by baseline factors such as those described in this study [23]. Digital intervention research has also shown that merely sharing content with individuals is not enough to change their behavior. The timing and method of delivery play an important role in this [61]. A sequential and cumulative module structure, in which each module builds upon previous information while introducing more strategies, could be hindered by incomplete engagement. As emphasized in recent digital intervention literature, engagement is not a unidimensional construct but is rather shaped by the interplay among contextual elements, diverse components (behavioral, cognitive, and affective), individual factors, and the design of the intervention [23,54,60 ]. This may explain the heterogeneous effects observed across some subgroups in our research and suggests that simple usage metrics may provide only an incomplete representation of engagement [62].

Finally, this pilot trial was not powered to test the efficacy of CHAMPS in EIS, and the intervention arm did not demonstrate statistically significant differences in cannabis-related outcomes compared with the control group. However, descriptive data showed consistently lower scores on the cannabis-related problems outcome (MPS score) by about 2 points, which could be considered clinically relevant, at all follow-up time points in the CHAMPS+EIS group [63]. These differences, although not statistically significant, warrant further investigation in a properly powered efficacy trial to formally test CHAMPS’s potential benefits in this population.

While the present study was underpowered to fully test complex 3-way interactions, findings suggest that some subgroups could require higher levels of exposure to achieve comparable outcomes, but this may only capture a facet of a more complex engagement process. Qualitative interviews conducted in our pilot trial are expected to provide more nuanced insights into engagement motives, facilitators, and barriers that cannot be captured by objective usage data alone [64]. Together, these points further extend the growing call in the literature for personalizing digital interventions and avoiding “one-size-fits-all” delivery models [65].

Several limitations should be considered. First, the modest sample size limited the generalizability of the findings and did not allow for a formal efficacy analysis. Second, the mixed-model analyses yielded mostly nonsignificant values, and the reported difference in social support between engagement groups was not adjusted for potential confounding variables (eg, education level, racial or ethnic composition), both reflecting limited statistical power. Third, as expected for a pilot trial, some attrition occurred, with progressively fewer participants completing each follow-up assessment. This loss-to-follow-up may reflect diverse factors, including low cannabis problem severity at baseline (participants with low MPS and high PBSM scores at baseline may perceive limited relevance of the intervention), digital fatigue, technical barriers, and app usability issues. This differential attrition could have introduced retention bias, as those who remained in the study may have differed from dropouts, potentially underestimating or overestimating engagement and related outcomes. Finally, additional bias may stem from data missing not at random, for example, if participants who responded more poorly were more likely not to complete some assessments and/or to disengage from the digital intervention.

In this sample of youth with FEP in EIS care, most individuals who had higher social support and a higher educational level engaged more with our digital cannabis harm-reduction intervention. No specific completion threshold was robustly associated with improvements in outcomes, suggesting that engagement is a multifactorial construct that goes beyond quantitative metrics (module completion). Interpretation of the findings should remain cautious, given their exploratory nature, the number of comparisons performed, and the possibility that some groups may have been underrepresented. Given the high prevalence of cannabis use among individuals with FEP, expanding and enhancing existing interventions could help mitigate the individual and societal impact of psychosis while easing the workload of clinicians in EIS. Further research, including a large-scale efficacy trial and qualitative methods, could explore whether refinements such as social support features (eg, family or peer prompts and/or in-app reminders, adapted and flexible support from clinicians), engagement-boosting strategies (eg, closer clinician follow-up, motivating content, etc), and literacy-tailored supports (eg, guided journal prompts, clinician support, etc) would improve intervention uptake through engagement and result in more favorable cannabis-related outcomes.