We conducted a single-blinded, randomized controlled trial with 2 arms. Patients were allocated to either the CG or the IG in a 1:1 ratio using a computer-generated, permuted block randomization scheme. The randomization was stratified by the CR program. A statistician (AHP) performed the randomization and sent the schemes to one of the coauthors (BBN), who prepared sealed envelopes. The first author, who also recruited the patients and conducted baseline assessments, brought the sealed envelopes to the baseline assessments. The envelopes were opened immediately after each baseline assessment, and patients were informed of their group allocation. The IG received individualized follow-up through an app for 1 year, while the CG received usual care, which included recommendations to maintain or further improve their lifestyle. Reporting adhered to the CONSORT (Consolidated Standards of Reporting Trials) 2010 statement [11].

Patients were recruited from 2 CR centers in the eastern part of Norway. One center offers 1- and 4-week inpatient CR programs, while the other provides a 12-week outpatient CR program. The 12-week outpatient rehabilitation program includes exercise and education sessions 3 times a week, along with personalized support based on patients’ needs, such as medication optimization, psychological counseling, dietary advice, and assistance with smoking cessation. This program has previously been evaluated and shown to improve exercise capacity and quality of life [12]. The 1- and 4-week residential programs had the same content as the 12-week outpatient program. However, the 4-week residential program provided patients with more time to adapt to lifestyle changes and learn about their condition compared with the 1-week program. The 4-week residential program has previously been reported as standard care [13]. Cardiologists, physiotherapists, dietitians, psychologists, and nurses were part of the health care team in all CR programs. Patients were recruited from 3 different CR programs, with randomization stratified accordingly. Approximately one-third of the patients were recruited from each program. Patients attending these CR programs are referred by a physician for various forms of heart disease, with coronary artery disease (CAD) being the most common reason for referral.

During participation in CR, all patients received oral information about the study from the first author (PL). Patients who met the inclusion criteria and were willing and able to participate were scheduled for a baseline assessment. Baseline assessments were conducted after a cardiopulmonary exercise test (CPET), during which patients were screened for exclusion criteria. As previously described [10,14], inclusion criteria included patients completing one of the CR programs, aged ≥40 years, ownership and use of an Android (Google LLC/Alphabet Inc.) or Apple (Apple Inc.) smartphone, and the ability to read and understand Norwegian or English. Exclusion criteria included ischemia or arrhythmias identified during CPET that imposed restrictions equivalent to <80% of maximal heart rate or a Borg scale (6-20) score [15] <15 during exercise. Patients with muscular or skeletal disorders that significantly affected exercise capacity more than their heart disease were excluded. Additionally, patients with severe malignant diseases (ie, advanced cancer) that had a greater impact on life expectancy than their heart disease were also excluded.

Recruitment for the 5-year follow-up was conducted by sending email and SMS text messages to all patients included in the original study [10]. The mail included written information about this study and an appointment for the 5-year follow-up, sent 4-6 weeks before the scheduled appointment. Patients were encouraged to confirm their willingness to participate and notify the first author if they planned to attend. Efforts were made to accommodate each patient’s schedule.

After the baseline assessment, patients randomly assigned to the IG were given access to an app and trained on how to use it. The app allowed users to create and set goals, with tasks and automatic reminders. A detailed description of the intervention is provided elsewhere [10,14]. Following the baseline assessment, the app was personalized with each patient’s goals and tasks for the upcoming year. Patients could choose when and how often they wanted reminders for their tasks. Additionally, they could use the app for reflections or notes related to a specific goal and directly communicate any questions to the supervisor. The supervisor had access to an administrator interface, which allowed them to monitor each patient. The interface also enabled the supervisor to send short motivational messages (maximum 112 characters) directly to each patient through the app.

Patients in the IG received comprehensive individualized feedback based on their goals, completed tasks, pending tasks, and any notes they had written. This feedback was provided weekly for the first 12 weeks and monthly for the remainder of the year. Additionally, all patients received 1-3 short individualized motivational messages per week throughout the year. The supervisor, a specialized physiotherapist in this study, provided all feedback and motivational messages to the patients in the IG. Questions submitted to the supervisor were answered within 2 working days. This individualized follow-up continued until the 1-year follow-up assessment. After the 1-year follow-up, the app was removed from the patients’ smartphones, and the follow-up was discontinued. At this point, patients were encouraged to maintain or improve their healthy behaviors based on their individual risk profiles.

Patients allocated to the CG received usual care, which included visits to their general practitioner and cardiologist as needed. After the baseline assessment, they were encouraged to maintain or improve healthy behaviors based on their individual needs and goals. The same encouragement was provided at the 1-year follow-up assessment.

The sample size was calculated based on the primary outcome, assuming a clinically significant difference in relative VO_2peak between groups of 3.5 ml/kg/minute [25-27]. Using data from a previously conducted feasibility study [28], the SD was estimated to be 6 ml/kg/minute. With a power of 80% and a significance level of .05, the required sample size was determined to be 47 patients per group. To account for a 20% dropout rate at the 1-year follow-up, we aimed to include a total of 113 patients.

IBM SPSS Statistics (version 29) and Stata (version 18; StataCorp LLC) were used for statistical analysis. Continuous, normally distributed baseline data were analyzed using an independent t test to assess differences between groups, while the Pearson chi-square test was used for categorical data. Baseline differences between participants with and without 5-year outcome data were analyzed using the same statistical tests. Missing data were assessed according to established strategies for handling missing data in clinical trials [29]. Differences between groups in primary and secondary outcomes were evaluated using a mixed model for repeated measurements. This model included a random intercept of the patient, as well as group, follow-up time, the interaction between group and follow-up time, and the baseline outcome measure as fixed effects. The Mann-Whitney U test was used to analyze differences between groups in IPAQ data at the 5-year follow-up. Independent t tests and Pearson chi-square tests were used to assess differences between participants with and without 5-year primary outcome data. This analysis was conducted to determine whether imputation of missing values was necessary, in accordance with strategies for managing missing data in clinical trials [29]. All analyses were performed on an intention-to-treat basis, and all tests were 2-sided. Data are presented as mean (SD) unless otherwise specified. A P value <.05 was considered statistically significant.

The Regional Committee for Medical and Health Research Ethics (South-East, ID: 2016-1476) approved the study protocol, and the study was conducted in accordance with the Declaration of Helsinki. An updated study protocol for this follow-up study was submitted and approved. All patients provided written, informed consent before inclusion in the study. The study data reported in this manuscript are anonymous. The original and updated study protocols were registered on ClinicalTrials.gov with the IDs NCT03174106 and NCT05697120, respectively. The study protocol and 1-year results have been previously published [10,14].

To the best of our knowledge, this is the first study to evaluate the long-term effects of individualized follow-up using an app during the first year after CR. Our primary finding indicates that the beneficial effects on VO_2peak and exercise performance observed 1 year after CR [10] had diminished 4 years after the intervention. For the primary outcome (VO_2peak), we maintained adequate statistical power to detect a true difference between the groups at the 5-year follow-up. However, the results for secondary outcomes should be interpreted with caution.

Numerous factors may explain why the effects observed after 1 year of individualized follow-up [10] were no longer present at the 5-year follow-up, and we can only speculate whether these effects might have been sustained if the follow-up had continued. We believe the relationship between the supervisor and the patients in our study played a pivotal role in fostering the successful adherence to healthy behaviors observed 1 year after CR [10]. This was also highlighted by the patients in a qualitative study investigating their experiences with the follow-up provided [30]. The patients emphasized that the supervisor, the person behind the app, was the most significant factor in promoting adherence to healthy behavior [30]. Successful adherence in the context of cardiovascular risk reduction is often attributed to an ongoing, collaborative relationship between the health care provider and the patient [31]. The idea that close monitoring or follow-up can enhance adherence is widely recognized [1,3]; however, only a few studies have demonstrated this in real-world settings. In our study, follow-up was discontinued after 1 year, and the observed effects appeared to diminish over time. The app-based follow-up provided to the IG during the first year after CR in this study was grounded in the Transtheoretical Model, also known as the Stages of Change Model [32]. According to this model, health behavior change involves 6 stages: precontemplation, contemplation, preparation, action, maintenance, and termination. The model emphasizes that behavior change is a process that takes time and is not linear. For example, patients in the maintenance stage may quickly regress to the preparation stage after CR due to factors such as a new cardiac event or other diseases or health problems that challenge their ability to maintain healthy behaviors. The need for support varies across different stages and should be tailored to increase the likelihood of successful behavior change [32]. Our findings suggest that reaching the termination phase—where individuals are confident they will never revert to their previous unhealthy behaviors and require minimal support [32]—is unlikely to be achieved within a 1-year follow-up period. Whether it would be possible with extended follow-up beyond 1 year remains an open question.

Patients in the IG reported exercising significantly more than those in the CG at the 5-year follow-up. However, there were no statistically significant differences between the groups in variables that reflect the effects of exercise (eg, VO_2peak and exercise performance). Notably, the difference in exercise habits between the groups at 5 years was similar to that observed at the 1-year follow-up. A plausible explanation for the lack of differences in VO_2peak and exercise performance at the 5-year follow-up, despite the IG reporting higher levels of exercise, could be the lower intensity or shorter duration of exercise sessions after the individualized follow-up was discontinued. As expressed by the patients in the IG [30], the individualized follow-up during the first year after CR provided more structure and focus to their exercise sessions than they were able to achieve on their own. Based on IPAQ data, the CG and IG appear to be equally physically active at the 5-year follow-up, although the CG reported significantly higher levels of walking METs compared with the IG. The IG reported higher levels of physical activity at both moderate and vigorous intensities than the CG, which may correspond to the significant difference observed in exercise habits, although differences observed in moderate and vigorous intensities were not statistically significant. Additionally, there was a statistically significant difference in triglycerides at the 5-year follow-up, favoring the CG. However, the uncertainty regarding the clinical relevance of the observed difference renders the result insignificant, and, on average, both groups remained within normal ranges [33].

For other metabolic risk factors, there were no statistically significant differences between the groups. The mean difference in body weight between the IG and CG at the 5-year follow-up was >1 kg. Although this difference did not reach statistical significance, we believe it is noteworthy, as each kilogram of weight gain has been shown to increase the risk of type 2 diabetes mellitus by 7.3% [34]. This suggests there may be a clinically relevant difference in body weight favoring the IG. However, given a minimal clinically important difference in body weight of 1 kg and an SD of 17 kg (based on our sample), more than 4500 patients would be required in each group to achieve 80% statistical power at a 5% significance level. Another noteworthy finding, although not statistically significant, is the LDL-cholesterol levels. The mean LDL-cholesterol at the 5-year follow-up was 2.3 mmol/l in the CG and 2.1 mmol/l in the IG. Given that the sample includes patients both with and without CAD, these levels are not alarming. When examining LDL-cholesterol specifically for those with CAD or acute coronary syndrome as the referral cause, the mean LDL-cholesterol levels at 5 years were 1.70 (SD 0.62) mmol/l in the CG and 1.88 (SD 0.62) mmol/l in the IG. The elevated LDL-cholesterol levels observed appear to be a common finding in studies evaluating maintenance programs after CR [35-38]. A reasonable explanation for this trend is that these post-CR programs [35-38] primarily focus on exercise rather than specifically addressing metabolic risk factors such as LDL-cholesterol and body weight. When comparing our results regarding LDL-cholesterol with a cross-sectional study by Peersen and colleagues (n=1127) [4], our sample at the 5-year follow-up is slightly better than their sample, with a mean LDL-cholesterol of 2.02 mmol/l 2-36 months (median 16 months) after a cardiac event. However, it is important to note that treatment targets for LDL-cholesterol have changed from <1.8 mmol/l to <1.4 mmol/l [39] since the completion of the intervention in this study. This indicates that there remains a need to optimize lipid management after CR. For the most part, lipid management after CR is handled by patients’ general practitioners. For decades, studies on lipid profiles following a cardiac event or CR have shown that few patients meet the guidelines for secondary prevention. This raises the timely question of whether this responsibility should be prioritized more by general practitioners or shifted back to the specialist health service. Consistent with the 1-year follow-up results [10], there were no statistically significant differences between the groups in HRQL at the 5-year follow-up. Nevertheless, it is encouraging that HRQL in our sample remained at a high level. This is particularly important, as reduced HRQL has been associated with an increased incidence of depression and anxiety [40,41], both of which are well-established psychosocial risk factors for poor prognosis in patients with cardiovascular disease [2]. Compared with the reference values for HeartQoL from the EUROASPIRE IV survey (EUROpean Action on Secondary and Primary prevention through Intervention to Reduce Events), our sample demonstrated higher scores in the global domain as well as the physical and emotional HRQL domains at both follow-up points [18].

A key strength of our study is the low dropout rate at both the 1- and 5-year follow-ups. Additionally, patients were recruited from 3 different CR programs, and the use of broad inclusion criteria further strengthens the generalizability of our findings. Our goal was to make the results applicable to the wider population of cardiac patients participating in CR. Given that CR programs vary widely both across and within countries, we included patients from 3 distinct CR programs delivered in different settings (residential and outpatient) and with varying durations (1, 4, and 12 weeks). Because of the uncertainty regarding whether adherence to healthy behaviors after CR is influenced by the type of CR program, randomization was stratified based on the specific CR program. By including all medically stable patients with cardiac diseases, we were able to capture a representative sample of a typical CR population. However, this broad inclusion criterion also posed challenges in reporting some outcomes, particularly metabolic risk factors. For example, patients treated for valve disease may not necessarily be advised to reduce LDL-cholesterol levels or body weight. To address this, descriptive statistics on subgroups and the clinical relevance of these outcomes have been provided. We believe that different subgroups of patients completing CR may derive varying benefits from the intervention. To explore this further, future studies could consider narrowing the inclusion criteria or significantly increasing the sample size to enable systematic subgroup analyses.

This study demonstrates that the beneficial and clinically important effects of a 1-year individualized follow-up via an app post-CR [10] had ceased by 4 years after the intervention. Despite the growing emphasis on addressing challenges related to adherence to healthy behavior and the use of technology in international research strategies [6,42,43], as well as in key public policies, we believe that significant work remains before mHealth can be fully implemented in this context. However, our findings on feasibility [28], 1-year follow-up outcomes [10], patients’ experiences [30], and the results presented in this study can provide valuable guidance for developing an implementation strategy for mHealth interventions aimed at promoting adherence to healthy behaviors in the post-CR setting. Before developing such a strategy, it is essential to conduct a comprehensive investigation into the potential barriers and facilitators associated with its implementation, involving all relevant stakeholders. To fully understand the multifaceted challenges of adherence to healthy behavior after CR, this investigation must include input from policy makers, CR program managers, health care professionals, and, most importantly, patients—both those struggling with adherence and those who have successfully maintained adherence to secondary prevention recommendations over the years.