Coronary heart disease (CHD) is a leading cause of premature death and disability worldwide [1]. Improved diagnosis, treatment, and management have substantially reduced the mortality rate of individuals living with CHD [2,3]. Following a cardiac event, clinical guidelines recommend people should participate in cardiac rehabilitation (CR), which is a multicomponent program that educates and supports self-management for the secondary prevention of CHD. CR aims to encourage people to make healthy lifestyle changes for reducing subsequent cardiac events. Lifestyle changes typically include initiating and maintaining regular physical activity, eating a healthy diet, stopping smoking, reducing harmful alcohol intake, and taking medications as per the prescribed regimen [4]. Appropriate self-management is critical for people with CHD to maximize treatment benefits [5].

Despite its benefits, participation in CR has been shown to be inadequate in all countries in which it has been measured [6]. In response, various alternative modes of delivery, including home-based CR and telehealth, have been developed, with similar reductions in cardiovascular risk factors as compared with hospital-based programs [7]. Most recently, mobile health technology has been used to better support the long-term self-management in people with cardiovascular disease [8]. Mobile health has the potential to be an extremely powerful tool for influencing behavior at a population level because it is widely available globally, inexpensive, and allows instant delivery of information [9].

SMS text messaging is the most widely used mobile phone intervention [10,11]. The strongest evidence for SMS text messaging in CHD is from the TEXT ME randomized controlled trial (RCT; n=710) [12], which reported statistically significant positive effects on low-density lipoprotein cholesterol, with more sizable effects on various secondary outcomes. Despite its effectiveness, the study was limited to a single center in Australia and excluded many participants because they did not own a mobile phone, which limited generalizability. TextMe was also evaluated as a standalone strategy; thus, it was unclear whether the intervention was more or less beneficial to those in traditional programs.

To address these limitations, we conducted the Text4HeartII trial across 2 district health boards in Auckland, New Zealand. This trial extended our previous Text4Heart randomized controlled pilot trial (n=123) [13,14], which found a doubling of adherence to lifestyle behaviors at 3 months but not at 6 months. Here, we present findings from a larger effectiveness trial of Text4Heart using an objective measure of medication adherence.

This study aimed to determine the effectiveness of the Text4HeartII self-management program for improving adherence to medication and lifestyle changes in addition to usual care in people with an acute coronary syndrome (ACS) at 24 and 52 weeks.

A two-arm, parallel RCT was conducted in 2 large metropolitan hospitals in Auckland, New Zealand between July 2016 and November 2019. The study was approved by the New Zealand Health and Disability Ethics Committee (15/NTA/205), and the protocol was registered and published before the conclusion of recruitment (Australian New Zealand Clinical Trials Registry, ID: ACTRN12616000422426. Registered, April 1, 2016). The trial was developed and reported in accordance with the CONSORT (Consolidated Standards of Reporting Trials) statement (Multimedia Appendix 1). No changes were made to the methods after commencement of the trial.

Eligible participants were adults with an ACS (including those who had undergone a percutaneous coronary revascularization procedure), clinically stable, able to read English, and able to provide informed consent. Participants were excluded if they had untreated ventricular tachycardia, severe heart failure, life-threatening coexisting disease with life expectancy of less than 1 year, and significant exercise limitations other than cardiovascular disease. Given the high level of mobile phone penetration (there were 6.5 million mobile connections in New Zealand in 2020; population 5.0 million people), mobile phone ownership was not considered in the eligibility criteria [15].

Figure 1 presents the flow diagram of the progress through the phases of the trial. A total of 739 people were screened between July 2016 and September 2018. The last trial participant completed the 24-week follow-up on October 2, 2019. Study data collection was completed in November 2019. A total of 306 eligible participants were randomized (153 per group), 282 participants were followed up at 24 weeks (141 per group), and 291 participants were followed up at 52 weeks (intervention: n=145 and control: n=146). All participants had MPR data at both 24 and 52 weeks, which were obtained directly from national medication dispensing records using the linked encrypted NHIs.

The participants were predominantly male (113/306, 80.3%) and New Zealand European (210/306, 68.6%), with a mean age of 61 years (SD 11) years. The groups were comparable at baseline (Table 1). Most participants were employed full-time (115/306, 37.6%) or retired (132/306, 43.1%). Most participants were married or living with a partner (230/306, 75.2%).

Adherence to the 3 medication classes (aspirin, statin, and any antihypertensive) in the intervention group was worse than that in the control group at 24 weeks (OR 0.60, 95% CI 0.38-0.96; P=.03). Differences in adherence to all 4 medication classes (aspirin, statin, β-blocker, and ACEI/ARB; OR 0.67, 95% CI 0.42-1.05; P=.08) and for individual classes of aspirin, statins, and β-blockers (Table 2) were not statistically significant, except for adherence to ACEI/ARB (OR 0.49, 95% CI 0.30-0.82; P=.006), which was worse in the intervention group than in the control group.

Similar results were found for medication adherence at 52 weeks (secondary outcome), with higher adherence to 3 medication classes in the control group than in the intervention group (OR 0.56, 95% CI 0.35-0.89; P=.01). The difference in adherence to the 4 medication classes was not significant (OR 0.68, 95% CI 0.43-1.08; P=.10). Similarly, there were no significant differences for individual classes of aspirin, statins, and β-blockers (Table 2), but adherence to ACEI/ARB was less in the intervention group than in the control group (OR 0.42, 95% CI 0.25-0.71; P=.001). These data are mirrored to some extent by the self-reported medication adherence data. At 24 weeks, there was a small effect on self-reported medication adherence, which favored the control group (adjusted mean difference 0.15; 95% CI −0.15 to 0.45; P=.33), with a larger effect observed at 52 weeks (adjusted mean difference 0.30; 95% CI 0.01-0.59; P=.04.

The composite measure of adherence with lifestyle behaviors was similar for both the intervention and control groups at 24 weeks (OR 1.11, 95% CI 0.65-1.90; P=.70) and 52 weeks (OR 0.97, 95% CI 0.58-1.62; P=.90; Table 3). In terms of individual behaviors, the number of participants who did not smoke at 24 weeks was higher in the intervention group (n=138) than in the control group (n=132; OR 5.75, 95% CI 1.08-30.61; P=.04). No other differences were observed between these groups. At 52 weeks, there was a trend for differences in time spent in vigorous-intensity activity (mean difference 44.8 min; 95% CI −3.9 to 93.5; P=.07) and MVPA (mean difference 100.3 min; 95% CI −16.6 to 217.3; P=.09), which favored the intervention group. A total of 75 serious adverse events were reported during the trial; however, none of them were related to the study. Only 1 participant died during the study period.

Table 4 shows participants’ responses to questions regarding the Text4HeartII intervention. Overall, most participants (136/139, 97.8%) did not experience any technical issues and would recommend the program to others. Most (114/138, 82.6%) participants felt that the program helped them manage their heart disease, but perceptions of whether the program influenced the behavior were mixed.

The Text4HeartII trial extended our pilot trial to determine the effectiveness of an SMS text messaging–based intervention to improve adherence to medication and lifestyle behaviors at both 24 and 52 weeks. Overall, we found no evidence to support the effectiveness of the program on dispensed medication or on adherence to a composite measure of lifestyle change over and above usual care.

The strengths of this trial were the RCT design and the objective assessment of medication adherence. Our study addressed criticisms of previous SMS text messaging trials [32,33], including enhanced generalizability. Our trial was of sufficient duration to elicit behavior change, and the follow-up assessments (24 and 52 weeks) were long enough to determine the sustained effect of the intervention. An additional strength of this trial was the complete data on the primary outcome. A possible limitation could be using MPR as an outcome, which reflects the number of days a drug was assumed to be in a person’s possession (based on dispensed drugs) rather than a measure of whether a person took their medication or not. Nevertheless, previous studies have used MPR as a proxy measure of medication adherence [20].

The lack of a positive effect on mediation adherence was surprising and contrary to previous research. A meta-analysis of 16 RCTs (N=2742) showed that SMS text messaging significantly improved medication adherence across a range of conditions (OR 2.11, 95% CI 1.52-2.93; P<.001) [34]. The effect was not sensitive to study characteristics (intervention duration or type of disease) or text message characteristics (personalization, 2-way communication, or daily text message frequency). The most commonly used method to assess adherence was self-report (9 studies), followed by the medication event monitoring system (4 studies).

Our findings differ from previous SMS text messaging trials in people with cardiovascular disease. A recent review (Text2PreventCVD) involving 9 trials (N=3779) and individual participant data (5 trials; n=2612) meta-analysis of SMS text messaging interventions demonstrated positive effects across a host of risk factors (BMI, systolic blood pressure, and diastolic blood pressure) [35]. Without directly comparing the messages within the respective interventions, it is difficult to know where the differences in effect may exist. However, a recent systematic description and comparison of the development processes of the interventions included in that systematic review highlighted that the Text4HeartII development process was comparable with that of previous trials [36]. Moreover, our Text4Heart pilot trial was included in the Text2PreventCVD systematic review and comparison of development processes. Like similar trials, the development of Text4HeartII involved consultation with experts, users, or other stakeholders; was based on literature reviews; included relevant theory; and involved primary research with end users [36]. The behavior change techniques used in this study were similar to those used in previous studies [36].

Given the existing body of evidence supporting SMS text messaging on medication adherence, it is clear that Text4HeartII did not have the desired effect on the population and setting in this trial. This may be attributed to personal factors associated with our cohort; randomized participants appeared to be more motivated to adhere than those in previous studies. The results showed that the difference between the control and intervention groups was relatively small and driven predominantly by MPR for antihypertensive medication. Despite similar days in hospital between groups at 24 weeks (approximately 1 day), adherence to all medication classes for both groups was higher than our proposed control rate (30%); approximately 41.5% (127/306) of the participants had an MPR >0.8. Our original control rate was based on previous research, which found that 59.3% (8028/13,520) of patients had an MPR >0.8 for statins only [18]. However, in this study, adherence to statins was higher; 82.6% (253/306) of participants had an MPR >0.8. Second, in our study, participants were recruited by cardiac nurses, and all participants were offered CR services, with good access to follow-up. The lack of marked differences between groups may have reflected organizational factors with higher levels of available support (including CR) at the 2 participating metropolitan hospitals. Thus, the effectiveness of SMS text messaging interventions may be strongly influenced by the population and context in which they are applied. This highlights the need to better understand the context in which SMS text messaging interventions are delivered and how individual- and organizational-level factors may affect the adoption, implementation, and effectiveness of these types of interventions. It also emphasizes the importance of evaluating interventions in the setting they are likely to be used before widespread adoption. Third, it is also possible that our text messages were not potent enough to evoke changes in medication adherence over and above usual cardiac care. Many previous SMS text messaging interventions have targeted single behaviors, such as smoking behavior, exercise, and medication adherence [32,34,37,38]. Thus, the lack of potency may have resulted from our approach to target multiple behaviors rather than targeting medication adherence alone. This is consistent with a previous qualitative research, which showed that conversations about changing multiple health behaviors were perceived to be overwhelming for patients and difficult to implement for health care professionals [39]. The lack of potency hypothesis is in line with qualitative responses from participants in this study, which showed that although 82.6% (114/138) of intervention participants responded that participating in this program had helped them recover from their heart condition, only 24.3% (19/78) felt the messages helped them to take their medications regularly. Furthermore, 56.5% (78/138) of the intervention participants responded that the program helped them change their health-related behaviors. Fourth, it is possible that the SMS text messaging program did not fully address the needs of participants [40]. Previous authors have highlighted the need for more personalized messages, depending on diagnosis (atherosclerosis, spontaneous coronary artery dissection, and Takotsubo cardiomyopathy) and key risk factors such as current smoking status. Although the proposed message library developed by Marshall et al [40] was similar to ours in terms of advice on a graduated exercise program, nutrition, smoking cessation, stress management, and medication, their proposed content also included messages around support groups and information about respective diagnoses. Patients could also send a message, which prompted a response from a CR nurse. The results of that study have not been published and were therefore not available for direct comparison with this study.

Surprisingly, the effects observed on lifestyle behaviors differed from those in our original pilot study [14]. In this study, we showed no clear differences in our composite measure of lifestyle change or individual behaviors, beyond those reported. The lack of effect may be attributed to the following issues. First, it is possible that there was a ceiling effect, with approximately 59.1% (181/306) of participants reporting adherence to 3 or more lifestyle behaviors at baseline. This could reflect self-reported bias, which was evident in the measurement of physical activity. Second, our measure of physical activity in this study differed from that in the pilot study. In this study, we used 3 items to capture the time spent in light-, moderate-, and vigorous-intensity activities. This approach was used to provide a better metric for adherence to guidelines for physical activity in adults (≥150 min of MVPA per day). Descriptive data suggested substantial overreporting of physical activity levels at all time points. For example, participants at baseline in both groups reported more than 300 minutes of daily MVPA, which is considerably higher than the standard population estimate for adults in New Zealand. Third, the lack of effect may have been related to the fact that both groups had well-managed risk factors at baseline (eg, few smokers, well-managed blood pressure), which limited the potential for change. Fourth, the lack of effect may also be attributed to changes made from the original intervention, which involved removing the website, which allowed people to set goals, review previous text messages, and access other resources related to CHD. We decided to remove the website component as it was infrequently used, and fewer than half of the participants in the pilot study felt using a website was a good way to deliver the program [14]. In the original study, we also issued pedometers and allowed people to track step counts, which was not a feature of this study. These features may have allowed for more interaction with the program but were removed because of pragmatic reasons (costs and distribution) related to the potential to scale as a national program.

Despite the lack of effect observed in our trial, SMS text messaging as an intervention has the potential to improve outcomes in people with CHD and other conditions. Findings from this study suggest that the context in which SMS text messaging interventions are delivered is important to consider and may have a significant impact on whether an intervention is effective or not. Future studies need to explore both individual- and organizational-level factors that may affect the adoption, implementation, and effectiveness of such interventions. For example, the RE-AIM (Reach, Effectiveness, Adoption, Implementation and Maintenance) framework [41,42] is a widely used evaluative framework for guiding the evaluation and reporting of health intervention effectiveness. It emphasizes collecting information about the reach, effectiveness, adoption, implementation, and maintenance of an intervention across both individual- and setting- or staff-level variables. RE-AIM aligns with systems-based approaches and allows for the assessment of vertical (eg, adoption decisions within a given organization) and horizontal (eg, adoption across different sectors) components. The application of RE-AIM or a similar framework would help in providing a clear understanding of the key factors that may affect the effectiveness of future text-based messaging interventions.

There was no evidence to support the effectiveness of Text4HeartII on dispensed medication or adherence to a favorable lifestyle over and above usual care. In its current form, Text4HeartII cannot be used to augment existing services. The findings of this study are in contrast with those of previous studies and highlight the importance of evaluating interventions in the setting they are likely to be used before widespread adoption. Changes to the intervention program are warranted to justify its future implementation.