The clinical trial was designed as a single-blinded, 2-arm randomized controlled trial with a total follow-up period of 9 months, comprising 6 months of intervention and 3 months of additional observation. The trial was conducted at 2 sites in Korea: a secondary general hospital (Sejong General Hospital) and a tertiary academic hospital (Korea University Guro Hospital). The trial was registered at Clinical Research Information Service (CRIS) (KCT0002361) [23]. Registration occurred prospectively on August 10, 2016, and the first participant was enrolled on December 17, 2016. This study is reported in accordance with the CONSORT (Consolidated Standards of Reporting Trials) 2025 statement and the CONSORT-EHEALTH (Consolidated Standards of Reporting Trials of Electronic and Mobile Health Applications and Online Telehealth) checklist (Checklist 1) for electronic health and mobile health randomized trials [24,25], and the completed CONSORT 2025 checklist is provided in Checklist 2.

The study protocol was approved by the Institutional Review Board of Korea University Guro Hospital (approval number MD16037-002). All participants provided written informed consent prior to enrollment; the consent process included permission for use of collected data for the prespecified analyses reported in this manuscript. Study data were deidentified and stored on encrypted, access-restricted institutional servers, and only authorized study personnel had access. Participants were provided with contact details to report any privacy-related concerns. Participants did not receive financial compensation for their participation in this study. No identifiable images of individual participants or users are included in the manuscript or supplementary materials.

A detailed description of the AnSim app development and usability testing has been published previously [26]. Briefly, AnSim was developed through multidisciplinary collaboration (cardiology, behavioral science, nursing, pharmacy, nutrition, rehabilitation, and app development), informed by focus group interviews, and implemented as a structured message bank spanning 5 domains (CV health/medications, nutrition, physical activity, stress management, and smoking cessation). Messages were tailored according to simplified transtheoretical model stages and mapped to 26 behavior change techniques [27]; iterative refinement incorporated expert review and feedback from potential users, followed by pilot testing of the delivery system.

All analyses were performed on an intention-to-treat basis using SPSS version 20.0 (IBM Corp). Continuous variables are presented as mean (SD); categorical variables are reported as n (%). Normality was assessed using the Shapiro-Wilk test and variance homogeneity by F test. Between-group comparisons for continuous variables were performed using independent-samples t tests (2-tailed). Categorical variables were compared using Pearson chi-square or Fisher exact tests, as appropriate. To examine group-by-time interactions for repeated measures, 2-way mixed ANOVA was used. Relative risks and 95% CIs for outcomes were estimated using Cox proportional hazards models. All statistical tests were 2-sided, and P<.05 was considered statistically significant.

A total of 120 patients were enrolled between December 2016 and April 2018 and randomized into the intervention group (n=60) and the control group (n=60). Participants were recruited consecutively from the participating PCI clinics during the enrollment period (ie, a convenience sample of eligible patients presenting to the sites), and randomization occurred after written informed consent and baseline assessment. Follow-up completion rates were high: 96.7% (58/60) in the intervention group and 95% (57/60) in the control group at 6 months, and 90% (54/60) and 93.3% (56/60), respectively, at 9 months. By 6 months, 2 participants in the intervention group and 3 in the control group did not complete follow-up. Between 6 and 9 months, additional attrition occurred in 4 and 1 participants, respectively. Overall, reasons for attrition included consent withdrawal (2 in each group) and loss to follow-up (4 in the intervention group, 2 in the control group) (Figure 1).

Baseline characteristics are shown in Table 1. The mean age of participants was 58.5 (8.8) years, and 84.2% (101/120) were male. However, the prevalence of diabetes mellitus was significantly lower in the intervention group (21.7% vs 41.7%, P=.02), resulting in a lower baseline HbA_1c (6.13% vs 6.60%, P=.048). No significant differences were observed in baseline blood pressure, lipid profiles, BMI, medication usage, or proportion achieving guideline-recommended goals. The medications prescribed in each group at the time of discharge after index coronary intervention were similar. Baseline clinical variables were nearly complete (mean missingness 0.83%). Missingness increased at follow-up (mean missingness 12.40% at 6 mo and 23.14% at 9 mo across the prespecified continuous outcomes) and was broadly similar between groups (eg, SBP at 9 mo: 16.7% vs 20%; HbA_1c at 9 mo: 23.3% vs 30%). While the Little MCAR test across 24 continuous clinical variables did not reject the MCAR assumption (χ²₁₃₇=152.83, P=.17), indicating insufficient evidence of a systematic missingness pattern, this result suggests rather than definitively proves that the data are missing completely at random.

Changes in blood pressure over time were not significantly different between groups at either 6 or 9 months (Figure 2; Table 2). At 9 months, mean SBP was 126 (14.2) mmHg in the intervention group vs 128 (15.4) mmHg in the control group (P=.94); diastolic pressure was 80.8 (13.2) mmHg vs 81.3 (12) mmHg (P=.85). There was no significant interaction effect between time and group for systolic (F_2,95=0.029; P=.97) or diastolic (F_2,95=0.333; P=.72) blood pressure. Similarly, no significant differences were observed in secondary outcomes, including lipid levels, HbA_1c, BMI, walking distance, or smoking rates at either follow-up point. The proportion of patients achieving all 5 guideline-recommended targets (LDL-C <70 mg/dL, BP <140/90 mmHg, regular exercise, nonsmoking status, BMI <25 kg/m²) did not differ significantly between groups at 6 or 9 months (Table 3).

During the 9-month follow-up, 2 clinical events occurred, both in the control group (1 coronary revascularization and 2 readmissions), but the difference between groups was not significant (P=.50; Table S3 in Multimedia Appendix 1). No deaths, myocardial infarctions, or heart failure hospitalizations were reported in either group.

Since the intervention using the AnSim app was message-mediated, the efficacy of the intervention could be dependent on participant message use. Among intervention group participants, the frequency of message reading was analyzed as a marker of engagement. Patients were stratified into high (upper 50%) and low (lower 50%) message readers (Table 4). High readers accessed more than twice as many messages on average (130.5 vs 50.6 messages; P<.001) and had significantly greater engagement in health diary input (250.2 vs 46.8 entries; P=.001). Although no significant differences were observed in blood pressure or lipid levels between subgroups, a significantly larger proportion of high readers achieved at least 4 of 5 guideline-recommended goals at 9 months (69% vs 19.2%; P<.001).

In a subgroup analysis, patients in the intervention group were categorized based on whether they demonstrated improvement in blood pressure by 9 months (n=23) or not (n=33). Patients were categorized as blood pressure responders if their follow-up blood pressure was less than or equal to 130/80 mmHg at 9 months. Responders had significantly higher baseline systolic and DBPs but showed greater reductions over time. At 9 months, mean SBP was lower in the responder group (121.4 vs 130.8 mmHg; P=.02), and DBP was similarly reduced (75.9 vs 84.3 mmHg; P=.03; Table 5). Responders also read significantly more messages (110.3 vs 83.3 messages, P=.02) and achieved better overall CV risk factor control. They were more likely to meet LDL-C targets at 6 and 9 months (P=.06 and .05) and maintain BP less than 140/90 mmHg at 9 months (P=.052). At 6 months, 73.9% of the BP-improved group met at least 4 guideline-recommended goals compared to 31.4% in the nonimproved group (P=.04, Table S4 in Multimedia Appendix 1). At 9 months, this pattern persisted, with 71.4% of the improved group versus 30.3% of the nonimproved group achieving at least 4 of 5 goals (P=.045, Table S4 in Multimedia Appendix 1).

Among 53 intervention participants who completed the satisfaction survey at 6 months, overall acceptability of the messaging intervention was high. Most respondents reported that the messages were easy to understand and helpful (n=45, 84.9%), expressed willingness to continue receiving messages (n=42, 79.2%), and indicated that they would recommend the app to others (84.9%). Detailed item-level acceptability and usability ratings are provided in Table S5 in Multimedia Appendix 1. We further explored whether engagement varied by participant characteristics. Linear regression analyses showed weak, nonsignificant associations between age and engagement metrics: age was not significantly associated with message-reading behavior (R²=0.04; P=.13; Figure 3A) or with diary engagement (R²=0.02; P=.34; Figure 3B). However, message reading frequency differed by education level, with elementary school graduates reading significantly fewer messages (P=.045; Table S6 in Multimedia Appendix 1). Satisfaction and health diary input frequency did not significantly differ by education level.

In this single-blinded randomized controlled trial, we evaluated the effectiveness of a patient-specific, smartphone-based messaging app (AnSim) on CV risk factor control in patients following PCI. Relative to our primary objectives, there were no significant between-group differences in changes in SBP or DBP. Within the intervention arm, greater engagement with messaging was associated with a higher likelihood of achieving guideline-recommended secondary prevention targets. Participants who experienced clinically relevant reductions in blood pressure at 9 months read more messages and demonstrated greater health behavior engagement, suggesting a potential dose–response relationship between digital engagement and risk factor control.

Several factors may have contributed to the lack of between-group differences in blood pressure. The participants had relatively well-controlled baseline blood pressure and lipid profiles, likely due to recent coronary intervention and optimal medical therapy. These ceiling effects may have limited the observable impact of the intervention. Moreover, adherence to medication and behavior change tend to be highest immediately following PCI, narrowing the margin for improvement and attenuating between-group differences [29]. A staged or adaptive approach, with reinforcement delivered later in recovery or targeted to patients with suboptimal baseline control, may be more likely to demonstrate incremental effects.

Several digital health interventions have demonstrated beneficial effects on CV outcomes, medication adherence, and behavior change [30,31]. Delivery of a comprehensive, app-supported rehabilitation program in the SMART-REHAB (SMARTphone-based early cardiac REHABilitation) trial was associated with improvements in physical recovery outcomes and greater enrollment and adherence to CR [32]. Compared to SMART-REHAB, which used structured exercise and real-time tracking, our AnSim intervention relied on passive, message-based behavioral prompting. This design distinction may explain differences in observed outcomes. Additionally, the baseline blood pressure of our participants was relatively well controlled due to recent PCI and optimal medical therapy, similar to the findings in SMART-REHAB where prescription rates for secondary prevention pharmacotherapy were consistently high. Accordingly, any incremental benefit of a low-intensity, message-only intervention on blood pressure is likely to be modest and may require longer follow-up or a higher-risk subgroup to be detectable.

Our results are consistent with the TEXT-ME study [18], which found that messaging helped maintain favorable risk profiles over time rather than lowering already well-controlled parameters. Additionally, our findings highlight the heterogeneity in responsiveness to digital interventions. Consistent with prior trials [33,34], our results demonstrate that engagement, not mere exposure, is a primary driver of digital intervention efficacy. Participants in the top half of message readership demonstrated greater health diary input and a higher likelihood of meeting at least 4 CV risk factor targets. Similarly, those who experienced improvements in blood pressure had significantly higher message exposure. This engagement-dependent effect parallels findings from a secondary analysis of a large CV text messaging trial, in which participants who actively engaged with the messages demonstrated significantly higher 12-month medication adherence than nonresponders, despite overall neutral trial results [35]. Such results support the view that the “dose” of digital engagement matters, and simply offering mobile health solutions may be insufficient without ensuring user uptake and sustained participation. Practically, it is essential to sustain participation through tailored content and timing, and to add a feedback loop that converts passive messages into actionable tasks (self-monitoring, goal tracking, and reinforcement). For users showing early drop-off, a stepped-care approach, such as automated reminders followed by brief coaching, may help restore engagement [36].

This trial represents an early real-world evaluation of AnSim, as acceptability and usability are central to interpreting scalability. Overall acceptability of the messaging content was high (eg, clarity/helpfulness, willingness to continue, and recommendation intent), supporting feasibility; however, overall usability ratings were comparatively lower, suggesting the presence of workflow friction that may limit sustained engagement [37]. To ensure digital equity, we implemented a simplified, user-friendly interface with large fonts and intuitive navigation, and provided face-to-face training at enrollment to overcome initial technical barriers. These measures likely contributed to our finding that age was not a significant determinant of message engagement. In contrast, education level remained an influencing factor; participants with lower education levels tended to read fewer messages, highlighting a persistent gap. This suggests that while simplified interfaces can bridge the age gap, addressing literacy barriers requires further tailoring [38]. Future interventions should focus on minimizing workflow friction and incorporating visual content with adaptive education levels to improve inclusivity and reduce disparities.

We observed a sustained and increased effect of the intervention over follow-up, which may reflect delayed behavior adoption or reinforcement over time. As shown in Table 4, the proportion of participants in the intervention group achieving multiple guideline-recommended risk factor targets was sustained or even slightly improved after the cessation of the 6-month messaging intervention. In contrast, the control group showed a marked decline in target achievement over the same period. These patterns are hypothesis-generating and suggest that the behavioral improvements facilitated by the AnSim messaging intervention may persist beyond the active intervention phase, supporting the hypothesis of a delayed or sustained effect. Given that secondary prevention requires lifelong adherence, interventions that promote sustained behavior change and long-term engagement may be particularly valuable, warranting longer-term evaluation.

We used a randomized design with blinded outcome assessment and standardized follow-up to evaluate the incremental effect of a messaging intervention on risk-factor control after PCI. Strengths include randomized design, theory-based messaging content, integration of behavior change staging, and concurrent evaluation of clinical outcomes, acceptability, and intervention engagement. To support participant blinding, the Heart Keeper app was installed in both groups; Heart Keeper functioned as a simple diary-style tool for recording clinical information and did not deliver structured behavior-change messaging or coaching.

However, there are several limitations. First, the modest sample size and short follow-up period limited the power to detect differences in long-term CV outcomes. Second, we did not systematically capture detailed reasons for nonenrollment (including inability to use a smartphone), which limits the assessment of representativeness and generalizability, particularly for populations with lower digital access or literacy. Third, providing a diary-style logging application to both groups may have encouraged self-monitoring and reduced between-group separation. Fourth, our intervention lacked multimedia interactivity, real-time feedback, or structured exercise modules, which may have reduced efficacy. Regulatory restrictions in Korea also limited bidirectional communication, a component emphasized in SMART-REHAB as essential for behavior reinforcement. Furthermore, because Heart Keeper does not generate message-exposure metrics comparable to AnSim, objective engagement data were not available for the control group, precluding direct between-group comparisons of digital engagement intensity. Finally, we did not conduct qualitative interviews among participants with low engagement; future studies should incorporate user-centered qualitative methods to identify modifiable barriers and optimize equitable uptake.

The current study underscores the importance of personalization and engagement in digital secondary prevention support. While app-based interventions can extend the reach of CR programs, their impact depends heavily on user activation. Future versions of AnSim should incorporate co-design with patients and clinicians, formative usability testing, and stakeholder engagement to optimize fit-for-purpose delivery. Adaptive intervention timing (reinforcement initiated later in recovery or triggered by loss of control/engagement) may better align with patients’ evolving needs. Integrating app data with electronic medical records and care teams could facilitate more proactive medical adjustments, similar to nurse-led programs or digital case management models [39].

Larger trials with longer follow-up are needed to determine whether these interventions improve clinical outcomes such as hospital readmissions and mortality. Comparative effectiveness studies between full-scale digital rehabilitation programs and lighter interventions like AnSim will also help clarify the optimal design and intensity of mobile CR. Although messaging interventions may be low-cost per participant, implementation requires development, maintenance, and monitoring resources; future work should assess cost-effectiveness, including staff workload if escalation or bidirectional communication is introduced, and evaluate scalability in routine post-PCI care.

This randomized trial demonstrated that a personalized, smartphone-based messaging intervention did not significantly improve blood pressure or lipid levels after PCI compared with usual care. In exploratory post hoc analyses within the intervention group, greater engagement with the AnSim app was associated with better risk-factor target attainment, with signals suggesting a larger potential benefit among participants with higher baseline blood pressure. These findings should be interpreted as hypothesis-generating and underscore engagement as a key determinant of potential impact for low-intensity mobile health strategies. Unlike comprehensive digital CR programs that incorporate structured exercise modules, real-time monitoring, or bidirectional coaching, AnSim was intentionally designed as a lightweight, message-mediated intervention that can be deployed with minimal resource burden. As a practical implication, these results suggest that real-world impact will likely depend less on message availability and more on implementation strategies that improve uptake and sustained engagement (eg, adaptive timing, stepped support, and integration into routine follow-up workflows). Future studies should test interventions designed to improve uptake and sustained participation, evaluate targeted or adaptive delivery for patients at higher residual risk, ensure digital equity for individuals with lower digital literacy, and assess longer-term clinical outcomes.