Study inclusion criteria

Studies were selected if they were RCTs which assessed an HBCTR programme in patients with CHD. The detailed inclusion criteria included: (i) population: patients aged ≥18years with a medical diagnosis of CHD, myocardial infarction (MI), acute coronary syndrome (ACS), angina pectoris, and/or those who have undergone revascularization (i.e. percutaneous coronary intervention or coronary artery bypass grafting) who have not previously received CR; (ii) intervention: any website-based and/or mobile health (mHealth) application used either as a stand-alone or supplemented with other delivery modes, such as text message, telephone or video calls, email or tele-monitoring to deliver Phase 2 CR or secondary prevention exclusively in the home setting; (iii) the HBCTR programme should target at least one of the following lifestyle behaviours—PA, healthy diet, smoking cessation, medication adherence, and stress management; (iv) control: supervised CBCR or UC (standard medical care that does not include any supervised or structured exercise training); (v) primary outcome: functional capacity (as measured by maximal or submaximal exercise testing to assess cardiorespiratory fitness); (vi) secondary outcome: behavioural (i.e. PA, smoking, medication management), physiological (i.e. cardiovascular risk factor control), and clinical (i.e. quality of life, depression, cardiac-related hospitalization, mortality). Studies were excluded if: (i) population: heart failure patients regardless of left ventricular ejection fraction; (ii) intervention: delivered in tandem with CBCR (i.e. hybrid CR) or after participants had completed Phase 2 CR (i.e. Phase 3 CR); (iii) control group: received components of web-based and or mHealth CR; and (iv) qualitative studies, book chapter reviews, abstracts-only journals, editorials, discussions papers, conference proceedings, and letters. We also excluded studies where the intervention was only text messaging, telephone calls, video conferencing, or telemonitoring and uploading measurements alone. While we did not restrict studies based on sample size or follow-up duration, we only included studies published in English.

Search strategy

A systematic electronic search was performed in the following databases—PubMed, EMBASE, CENTRAL, CINAHL, Scopus, and PsycINFO up to January 2021. No limits on publication status and date were imposed on the search. The reference lists of relevant reviews and key literature were manually searched to supplement our database search. Details of our database search are documented in Supplementary material online, Table S1. Our search strategy was peer-reviewed by a university resource librarian with expertise in systematic review searching.

Study selection process

Search results were managed using EndNote X9. After the removal of duplicates, two reviewers independently screened the title and abstract of studies against the eligibility criteria. The full texts of all relevant studies were downloaded and further evaluated for compliance with the eligibility criteria. Disagreements between the two reviewers regarding inclusion were resolved by consultation with a third reviewer.

Risk of bias (quality) assessment

The Cochrane Risk of Bias tool21 for randomized trials was used to guide the quality assessment of each included study and consists of the following domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias (e.g. whether study groups were comparable at baseline). Two independent reviewers conducted the quality appraisal separately. Discrepancies were resolved through discussion.

Data extraction

Two reviewers used standardized forms to independently extract study data, any discrepancies were resolved between the two reviewers to reach a consensus. Two authors of the included studies were contacted through email to obtain missing data. Both replied and provided the required information.

Study characteristics

A summary of the characteristics of included studies is presented in Table 1.25–38 All 14 studies were two-arm RCTs involving a total of 2869 participants (sample size ranging between 15 and 1000). Six studies were conducted in China,26–28^,32^,34^,38 three in Canada,29^,30^,36 two in Australia,33^,35 one in the New Zealand,37 one in the UK,25 and another was a multi-centre study across Europe.31 Participants included in this review were diagnosed with the following: angina,25^,26^,37 MI,26^,33^,36^,37 ACS,29–31 CHD,27^,32^,35^,37 or had undergone coronary revascularization.29^,31^,34^,36–38 The mean age of participants ranged from 45.8 to 73.6.36 Females accounted for 21.8% of the overall sample. Control groups included three supervised exercise and health education in a CBCR setting, while the remaining majority were UC that involved outpatient visits to a physician or nurse25–28 and regular CHD health education.26–30^,34^,36^,38 One of these studies had a waitlist control group that received UC.27

Intervention characteristics

HBCTR was delivered largely via web-based platforms in five studies25^,27^,29^,30^,36 and smartphone applications in nine studies.26^,28^,31–35^,37^,38 Ten studies used telemonitoring devices to support PA self-monitoring including pedometers,26^,30^,33 accelerometers,25 and heart-rate monitors.28^,29^,31^,32^,36^,37 Supplementary forms of communication, such as text messages, phone calls, video calls, and emails were utilized between the participants and intervention team, and contact varied from a weekly to monthly basis. Eight studies mentioned the use of secured web servers,31^,37 data portals,26^,38 and password-protected websites25^,29^,30^,36 for the transmission and storage of patient data.

Five studies27^,32^,35^,37^,38 reported the use of a theoretical framework. With the exception of one study,38 all included studies had a PA component as part of their intervention. Five studies26^,28^,30^,33^,34 had comprehensive HBCTR that included all the core components of secondary prevention (CHD risk factors management, PA, smoking cessation, medication adherence, and stress management), while three studies31^,32^,37 focused solely on PA. Stress management was the least addressed area and was covered in only six studies.

Duration of HBCTR ranged from 6weeks to 6months. Eight studies monitored participants’ intervention compliance via website or mobile application logins,25^,29^,35^,36^,38 exercise and monitoring data uploading,29^,33^,35^,36 number of chat sessions attended,29^,36 and progress of education modules and reading materials.26^,34 Details of intervention characteristics can be found in Table 1 and Supplementary material online, Table S3.

Risk of bias of included studies

Of the 14 included RCTs, random sequence generation was clearly detailed in 11 studies.25^,26^,29–35^,37^,38 Six studies27^,28^,32^,35^,36 had an unclear risk of selection bias due to failure to describe how allocation concealment was undertaken. Due to the nature of the intervention, blinding of participants and personnel was not possible in all studies,25–38 rendering a high risk of performance bias. Detection bias was avoided in seven studies26^,29–31^,35–37 through blinding of outcome assessors; however, two studies were rated as unclear risk of bias as one study’s27 contactless online data collection precluded any direct involvement of outcome assessors and most of the outcomes in both studies27^,32 were unlikely to be influenced by a lack of blinding. Ten studies26^,28^,29^,31^,32^,34–38 had a low risk of bias as attrition rates were <20% and dropouts did not differ largely between treatment groups. Nine studies25^,26^,29–31^,33^,35^,37^,38 reported having a study protocol and were deemed low risk of reporting bias. Three studies29^,31^,36 had risk of bias due to imbalances in baseline characteristics between treatment groups, including type II diabetes, hypertension, total cholesterol, prior revascularization, family history of cardiovascular, self-reported PA, and total treadmill time. A summary of the risk of bias is provided in Figure 2.

Open in a separate window

Figure 2

Risk of bias assessment of included studies.

HBCTR vs. UC

Our finding of a statistically significant mean difference of 25.95m in the 6MWT distance between HBCTR and UC is clinically relevant, as it reached the minimal clinically importance difference of 25m for the 6MWT in CHD patients undergoing CR.39 Unexpectedly, we did not observe a significant difference in symptom-limited exercising testing between HBCTR and UC. The European Society of Cardiology recommends the evaluation of personal preferences and goals in the prescription of individualized approaches to exercise training to achieve long-term adherence and optimized health benefits.2 Closer evaluation of our analysis revealed that all but one study31 showed a trend of improved functional capacity in favour of HBCTR. Subsequent omission of this study by leave-out sensitivity analysis changed the results into a significant difference in favour of HBCTR. A possible explanation for this is that all the HBCTR participants in Snoek et al.31 received a standardized exercise training plan of 30min of moderate activity for five times/week, whereas participants in the other three studies29^,32^,36 included in the meta-analysis provided tailored exercise prescriptions. This suggests that if patients are not given alternatives to their training plan that align with their preferences, they may not adhere sufficiently to the minimum PA levels required to obtain significant health benefits of improved functional capacity. However, as there were only four studies in this meta-analysis, more research is required for firmer conclusions.

HBCTR also showed some evidence of improved PA behaviour compared with UC in objective (step count/day) and subjective (proportion of patients categorized as physically active) measures, but not in another subjective measure (energy expenditure kcal/week). Similarly, narrative synthesis of three included trials27^,31^,37 produced polarizing results on the effects of HBCTR on PA behaviour. Our observation of positive PA improvements in some studies and not others are due to a few reasons. First, as included studies were inconsistent in their PA definition and measurements, we were limited in our ability to reasonably pool all available data into a meta-analysis. Second, the use of subjective PA measures and the classification of continuous data into categories of PA could have resulted in measurement recall bias and a loss of data sensitivity, respectively.40 Therefore, in order to evaluate the potential superiority of HBCTR in improving PA behaviours, we recommend that future studies rely on objective measures of PA [such as percentage of peak heart rate, heart rate reserve and peak VO₂ or metabolic equivalent of task (MET) via wearable accelerometers, and heart-rate monitors], and report outcomes of PA intensity, duration, and frequency that are paralleled with recommended national guidelines.2^,41 Meta-analysis by Rawstorn et al.15 found significant improvement in PA behaviour but not in functional capacity when comparing telehealth exercise-based CR with UC. The following may offer possible explanations for this inconsistency: (i) our review focused on interactive web or smartphone HBCTR as the major delivery platform, whereas Rawstorn et al.15 included land-based telephone services; and (ii) variation in PA intervention intensity, duration, frequency, and engagement. It is likely that differences in these programme characteristics have an influence on intervention delivery and effectiveness and may have contributed to this variation in findings.

With respect to the effect of HBCTR vs. UC on QoL, our review revealed varied results. Meta-analysis of the effects on QoL measured by Short-Form questionnaire revealed significant results in favour of HBCTR, but this was not echoed in our narrative synthesis. While the reason for our discrepancy in findings is unclear, it warrants exploration in future studies as QoL has been known to be significantly impaired in patients with CHD, and improvements in QoL have been associated with a reduction in rehospitalization,42 re-cardiac events, and mortality.43 Hence, improvements in QoL remains an important therapeutic goal in CHD management, although the mechanism of this prognostic effect is not completely understood.44

Our finding of significant effects on depression scores in favour of HBCTR is encouraging, as depression is common in the CHD population and has been associated with re-cardiac event, cardiac-related mortality, and reduced participation in CR programmes.45 Favourable effects were also seen in medication adherence but given the insufficiency of data and employment of distinctive methodology, this outcome effect was difficult to quantify and could not be confirmed with a meta-analysis.

While it was unexpected that we did not find statistically significant effects of HBCTR on physiological cardiovascular risk factor outcomes, smoking, mortality, and cardiac-related hospitalization when compared with UC, our findings are similar to a previous meta-analysis.15 Advancements in early diagnostic and therapeutic coronary revascularizations coupled with the introduction of statins and the systematic use of cardio-protective pharmacotherapy have led to substantial improvements in risk factor control and overall patient outcomes in this modern era of cardiology.46 Hence, against this backdrop of reduced risk profile, any impact derived from participation in secondary prevention programmes like CR on long-term physiological and clinical outcomes could be potentially attenuated.46 Evidently, this was observed in our current review, where all our included trials belonged to this modern era. Furthermore, given the paucity of adverse events like current smoking status, mortality, and hospitalization and our low number of included studies with relatively short follow-up duration, we are prevented from detecting a difference in outcome effect between groups and are unable to confidently arrive at an evidenced-based conclusion on the effect of HBCTR vs. UC for these outcomes.

HBCTR vs. CBCR

Meta-analysis of the three included studies that compared HBCTR to CBCR found equivalent effects on functional capacity, PA behaviour, smoking, physiological risk factors, QoL, depression, and cardiac-related hospitalization. While our small number of included CBCR-controlled studies may have been insufficient to detect intervention effects, our findings are consistent with previous reviews.14^,15 This suggests that HBCTR may serve as a suitable alternative to Phase 2 CBCR, especially when the need for CBCR cannot be met due to existing restrictions and resource limitations brought about by the COVID-19 pandemic.

Challenges of HBCTR

Our review also identified distinct methodological and implementation challenges of HBCTR. Firstly, the very nature of HBCTR necessitates a greater responsibility on the patient to self-manage and change long-standing habits of unhealthy behaviours.7 Accordingly, HBCTR programmes should be developed to support patients with the necessary skills to successfully undertake this behaviour change. Only five of the included studies in this review stated the use of a behaviour change theory in their intervention development. However, with the exception of one study,27 the extent to which the theoretical framework informed the intervention components to effectively support behaviour change related to CHD is unclear. Secondly, as interventions like HBCTR are self-administered, methods to ensure patients receive an appropriate dose and complete the intervention are challenging but important.47 Information on levels of engagement and receipt of treatment could serve as an instructive adjunct to determine why an intervention outcome is effective or not.47 Attempts to measure intervention compliance (Table 1) varied and were not consistently reported across included studies. There is a need for future studies to monitor intervention compliance and investigate if HBCTR improves uptake and completion of CR programmes compared with CBCR. Additionally, quality assurance of HBCTR programmes should be considered if HBCTR is to see favourable and cost-effective outcomes, and available quality metrics outlined by Randal et al.7 may serve to guide future programmes in this regard. Lastly, CR providers should acknowledge the value that patients place on their privacy and its consequences on their willingness to engage in digital platforms like HBCTR.8^,48 Less than half of our included studies specified the use of secure password-protected intervention platforms. Future programmes should ensure the provision of a transparent privacy policy and secure health information and clinical data storage systems to encourage patients to participate in technology-led healthcare interventions.

The authors acknowledge the assistance of Ms Annelissa Chin with the peer-review database search strategy.