Scolaris Content Display Scolaris Content Display

Internet‐based interventions for the secondary prevention of coronary heart disease

Collapse all Expand all

Abstract

available in

Background

The Internet could provide a means of delivering secondary prevention programmes to people with coronary heart disease (CHD).

Objectives

To determine the effectiveness of Internet‐based interventions targeting lifestyle changes and medicines management for the secondary prevention of CHD.

Search methods

We searched the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, in December 2014. We also searched six other databases in October 2014, and three trials registers in January 2015 together with reference checking and handsearching to identify additional studies.

Selection criteria

Randomised controlled trials (RCTs) evaluating Internet‐delivered secondary prevention interventions aimed at people with CHD.

Data collection and analysis

Two review authors independently assessed risk of bias and extracted data according to the Cochrane Handbook for Systematic Reviews of Interventions. We assessed evidence quality using the GRADE approach and presented this in a 'Summary of findings' table.

Main results

Eighteen trials met our inclusion criteria. Eleven studies are complete (1392 participants), and seven are ongoing. Of the completed studies, seven interventions are broad, targeting the lifestyle management of CHD, and four focused on physical activity promotion. The comparison group in trials was usual care (n = 6), minimal intervention (n = 3), or traditional cardiac rehabilitation (n = 2).

We found no effects of Internet‐based interventions for all‐cause mortality (odds ratio (OR) 0.27, 95% confidence interval (CI) 0.04 to 1.63; participants = 895; studies = 6; low‐quality evidence). There was only one case of cardiovascular mortality in a control group (participants = 895; studies = 6). No incidences of non‐fatal re‐infarction were reported across any of the studies. We found no effects for revascularisation (OR 0.69, 95% CI 0.37 to 1.27; participants = 895; studies = 6; low‐quality evidence).

We found no effects for total cholesterol (mean difference (MD) 0.00, 95% CI ‐0.27 to 0.28; participants = 439; studies = 4; low‐quality evidence), high‐density lipoprotein (HDL) cholesterol (MD 0.01, 95% CI ‐0.06 to 0.07; participants = 437; studies = 4; low‐quality evidence), or triglycerides (MD 0.01, 95% CI ‐0.17 to 0.19; participants = 439; studies = 4; low‐quality evidence). We did not pool the data for low‐density lipoprotein (LDL) cholesterol due to considerable heterogeneity. Two out of six trials measuring LDL cholesterol detected favourable intervention effects, and four trials reported no effects. Seven studies measured systolic and diastolic blood pressure; we did not pool the data due to substantial heterogeneity. For systolic blood pressure, two studies showed a reduction with the intervention, but the remaining studies showed no effect. For diastolic blood pressure, two studies showed a reduction with the intervention, one study showed an increase with the intervention, and the remaining four studies showed no effect.

Five trials measured health‐related quality of life (HRQOL). We could draw no conclusions from one study due to incomplete reporting; one trial reported no effect; two studies reported a short‐ and medium‐term effect respectively; and one study reported both short‐ and medium‐term effects.

Five trials assessed dietary outcomes: two reported favourable effects, and three reported no effects. Eight studies assessed physical activity: five of these trials reported no physical activity effects, and three reported effectiveness. Trials are yet to measure the impact of these interventions on compliance with medication.

Two studies measured healthcare utilisation: one reported no effects, and the other reported increased usage of healthcare services compared to a control group in the intervention group at nine months' follow‐up. Two trials collected cost data: both reported that Internet‐delivered interventions are likely to be cost‐effective.

In terms of the risk of bias, the majority of studies reported appropriate randomisation and appropriate concealment of randomisation processes. A lack of blinding resulted in a risk of performance bias in seven studies, and a risk of detection bias in five trials. Two trials were at risk of attrition bias, and five were at risk for reporting bias.

Authors' conclusions

In general, evidence was of low quality due to lack of blinding, loss to follow‐up, and uncertainty around the effect size. Few studies measured clinical events, and of those that did, a very small number of events were reported, and therefore no firm conclusions can be made. Similarly, there was no clear evidence of effect for cardiovascular risk factors, although again the number of studies reporting these was small. There was some evidence for beneficial effects on HRQOL, dietary outcomes, and physical activity, although firm conclusions cannot yet be made. The effects on healthcare utilisation and cost‐effectiveness are also inconclusive, and trials are yet to measure the impact of Internet interventions on compliance with medication. The comparison groups differed across trials, and there were insufficient studies with usable data for subgroup analyses. We intend to study the intensity of comparison groups in future updates of this review when more evidence is available. The completion of the ongoing trials will add to the evidence base.

PICOs

Population
Intervention
Comparison
Outcome

The PICO model is widely used and taught in evidence-based health care as a strategy for formulating questions and search strategies and for characterizing clinical studies or meta-analyses. PICO stands for four different potential components of a clinical question: Patient, Population or Problem; Intervention; Comparison; Outcome.

See more on using PICO in the Cochrane Handbook.

Plain language summary

Internet‐based programmes for people with heart disease

Review question

Are Internet‐based support programmes for people with heart disease helpful in improving their heart disease condition?

Background

Heart disease is the most common cause of ill health and preventable death. Cardiac rehabilitation is a programme that helps people with heart disease gain better health. It is held in group classes that take place at hospitals or within the community. People attend these classes once or twice a week for around six to eight weeks. The classes usually involve exercising, and receiving advice on ways to improve their health. People needing these programmes are not always able to attend them. An alternative is to provide this programme through the Internet. In this review we looked at whether programmes delivered through the Internet are helpful in improving death rates, the need for surgery, repeated heart attacks, cholesterol levels, blood pressure, health‐related quality of life (HRQOL), diet, physical activity, medication compliance, healthcare usage, and costs.

Study characteristics

The evidence is current to December 2014. We included 18 studies. Eleven are complete, and seven are ongoing. In the completed studies, 1392 people with coronary heart disease were recruited. The average age of participants ranged from 54.9 to 66.27 years. The majority of people recruited were men. Studies were carried out worldwide, and in a variety of healthcare settings. Seven studies tested broad programmes targeting multiple lifestyle factors related to heart disease. Four studies tested programmes focused only on increasing levels of physical activity. The length of the programmes in the included studies ranged from six weeks to one year. These programmes were compared to no intervention in six studies, some support in three studies, and full traditional rehabilitation in two studies.

Key results

There is no evidence to date to suggest that Internet‐delivered programmes help reduce rates of death or future cardiac surgery, but this was based on a small number of studies. There is also no strong evidence to date suggesting a benefit of these programmes for lipid levels or blood pressure. There is some evidence to suggest improvements in HRQOL and behaviour change, but there is insufficient evidence to date to draw firm conclusions. Studies have not yet measured the impact of Internet‐delivered programmes on medication compliance. There was very limited information on healthcare utilisation and cost of interventions. The reporting of the seven ongoing studies will add to the evidence base.

Quality of the evidence

The evidence was generally of low quality. The included studies were at some risk of bias, with six studies judged at high risk of bias for some risk of bias domains. The results of this review therefore need to be interpreted cautiously.

There is currently limited evidence on the effects of Internet‐based interventions for the treatment of coronary heart disease. We identified seven ongoing trials, which we will incorporate into this review when the results are available.

Authors' conclusions

available in

Implications for practice

Due to the low‐quality evidence in study outcomes and limited findings to date, there are no implications for practice at present.

Implications for research

The current evidence on the use of the Internet in the secondary prevention of heart disease is still evolving and has shown mixed results. A number of questions have been raised. We particularly need to investigate the long‐term effects of web‐based interventions used in the secondary prevention of CHD on cardiovascular risk factor profiles and clinical events. There is also a need to determine the intensity and duration of the intervention required to achieve effective secondary prevention of CHD and the effective components of behavioural changes.

With regards to the socio‐demographic characteristics of the CHD population, future studies should focus interventions on a wide range of participants so that findings are generalisable and can also be tailored to specific populations if differences are found.

More rigorous studies comparing the long‐term effects of Internet interventions are needed in order to determine long‐term effectiveness of Internet interventions for the secondary prevention of CHD. There is also a need to measure outcomes objectively. Physical activity was assessed using self report measures in six out of eight trials. These measures are susceptible to overestimations, which the use of accelerometer technology would reduce. Future trials should also include cost‐effectiveness outcomes.

Summary of findings

Open in table viewer
Summary of findings for the main comparison. Internet‐based interventions compared to usual care or no care for prevention of coronary heart disease

Internet‐based interventions compared to usual care or no care for prevention of coronary heart disease: clinical outcomes

Patient or population: patients with coronary heart disease
Settings: healthcare settings
Intervention: Internet‐based interventions
Comparison: usual care or no care

Outcomes

Illustrative comparative risks* (95% CI)

Relative effect
(95% CI)

No of Participants
(studies)

Quality of the evidence
(GRADE)

Comments

Assumed risk

Corresponding risk

Usual care or no care

Internet‐based interventions

Total mortality

Study population

OR 0.27
(0.04 to 1.63)

895
(6 studies)

⊕⊕⊝⊝
low1,2

9 per 1000

2 per 1000
(0 to 15)

Moderate risk population

3 per 1000

1 per 1000
(0 to 5)

Revascularisation

Study population

OR 0.69
(0.37 to 1.27)

895
(6 studies)

⊕⊕⊝⊝
low1,2

58 per 1000

41 per 1000
(22 to 73)

Moderate risk population

18 per 1000

12 per 1000
(7 to 23)

*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; OR: Odds ratio

GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

1Maddison 2014, Reid 2012, Zutz 2007 ‐ outcomes may have been influenced by lack of blinding.
2Confidence interval crosses line of no effect (uncertainty around the magnitude of effect).

Open in table viewer
Summary of findings 2. Internet‐based interventions compared to usual care or no care for prevention of coronary heart disease

Internet‐based interventions compared to usual care or no care for prevention of coronary heart disease: cardiovascular risk factors

Patient or population: patients with coronary heart disease
Settings: healthcare settings
Intervention: Internet‐based interventions
Comparison: usual care or no care

Outcomes

Illustrative comparative risks* (95% CI)

Relative effect
(95% CI)

No of Participants
(studies)

Quality of the evidence
(GRADE)

Comments

Assumed risk

Corresponding risk

Usual care or no care

Internet‐based interventions

Total cholesterol

The mean total cholesterol in the intervention groups was
0 higher
(0.27 lower to 0.28 higher)

439
(4 studies)

⊕⊕⊝⊝
low1

LDL cholesterol

See comment

See comment

Not pooled

437
(4 studies)

⊕⊕⊝⊝
low1

High levels increase risk, while low levels reduce risk. Data not pooled due to unexplained considerable (I2 = 77%) heterogeneity

HDL cholesterol

The mean HDL cholesterol in the intervention groups was
0.01 higher
(0.06 lower to 0.07 higher)

437
(4 studies)

⊕⊕⊝⊝
low1

High levels reduce risk, while low levels increase risk

Triglycerides

The mean triglycerides in the intervention groups was
0.01 higher
(0.17 lower to 0.19 higher)

439
(4 studies)

⊕⊕⊝⊝
low1

High levels increase risk, while low levels reduce risk

Systolic blood pressure

See comment

See comment

Not pooled

623
(5 studies)

⊕⊕⊝⊝
low2

Systolic blood pressure measured in 5 studies. Data not pooled due to unexplained substantial (I2 = 63%) heterogeneity

Diastolic blood pressure

See comment

See comment

Not pooled

622
(5 studies)

⊕⊕⊝⊝
low2

Diastolic blood pressure measured in 5 studies. Data not pooled due to unexplained substantial (I2 = 58%) heterogeneity

*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval;HDL: High‐density lipoprotein
; LDL: Low‐density lipoprotein

GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

1Zutz 2007 ‐ outcomes may have been influenced by lack of blinding; Varnfield 2014 ‐ loss to follow‐up > 20% in both experimental arms.
2Devi 2014, Zutz 2007 ‐ outcomes may have been influenced by lack of blinding; Varnfield 2014 ‐ loss to follow‐up > 20% in both experimental arms.

Background

available in

Description of the condition

Cardiovascular diseases are the number one cause of death globally. It has been estimated that the cumulative cost of cardiovascular disease to the European Union economy is EUR 196 billion a year (European Society of Cardiology 2012). Secondary prevention interventions can favourably modify cardiac risk factors in people with coronary heart disease (CHD) (McAlister 2001), and have a positive effect on physical activity, exercise training, and overall cardiorespiratory fitness (Lavie 2009). Furthermore, there is also strong evidence showing benefits in terms of clinical events and health‐related quality of life (HRQOL) (Cole 2011; McAlister 2001; O'Connor 1989; Oldridge 1988). Previous Cochrane reviews have highlighted the potential benefits in terms of effectiveness of both exercise‐based interventions and psychological interventions for CHD (Heran 2011; Whalley 2011). A further Cochrane review illustrates that smoking cessation effectively alters the course of CHD (Critchley 2003).

Not all individuals with CHD take part in secondary prevention programmes. A recent annual survey found that only 43% of people with CHD took part in cardiac rehabilitation in England, Wales, and Northern Ireland between 2012 and 2013 (NACR 2014). Referral, uptake, and adherence to traditional cardiac rehabilitation programmes are often poor, particularly in older patients, women, ethnic minorities, and in patients with angina or heart failure (Beswick 2004). The reasons for this lack of uptake are complex, some related to the organisation and system of delivery (O'Driscoll 2007), and others to individual choice. Factors related to rehabilitation non‐attendance that appear frequently in the literature are employment commitments, difficulties with transport, lack of time, distance to travel to rehabilitation, and embarrassment related to attending rehabilitation (De Vos 2013; McKee 2013; Neubeck 2012).

Home‐based interventions are an alternative way to improve access to secondary prevention programmes. A meta‐analysis of 36 trials of home‐based secondary prevention interventions demonstrated that they improve HRQOL, lower systolic blood pressure, lower total cholesterol, reduce smoking rates, and reduce depression in people with CHD (Clark 2010). A Cochrane review confirmed the effectiveness of home‐based compared with centre‐based cardiac rehabilitation programmes(Taylor 2010), reporting no differences between programmes in the number of clinical events, exercise capacity, blood pressure, total cholesterol, proportion of smokers, or HRQOL. A recent approach has been to embrace technology and use the Internet as an option to support lifestyle change important for the secondary prevention of CHD.

Description of the intervention

Using the Internet offers an alternative way to deliver secondary prevention interventions. Internet interventions can overcome inconveniences such as the time and expense involved in travelling to intervention locations (Griffiths 2006; Neville 2009; Nguyen 2004). Users also benefit from having information and support available 24 hours per day. With advanced website programming, it is also possible to create highly interactive interventions that incorporate theoretical constructs of health behaviour change and evidence‐based 'behaviour change techniques' (Ciccolo 2008; Michie 2013).  Patients are also able to communicate with health professionals through the use of various communication channels such as email, instant chat, or discussion forums (Griffiths 2006; Murray 2008). Interventions may not include all of these features, however there is considerable potential to design interventions with as many features as possible. Further practical advantages for service providers include cost‐effectiveness, in Murray 2008, and the ability to reach large, geographically dispersed populations without time or location restrictions (Eng 1999; Eysenbach 2001; Griffiths 2006). Web‐based interventions also have the potential to store large volumes of information and can be easily updated as new research becomes available (Murray 2008). 

How the intervention might work

Online interventions have been shown to be effective for general health behaviour change (Wantland 2004). Several trials have examined the effectiveness of Internet‐based interventions for promoting healthy nutrition and weight loss and increasing physical activity (Moore 2008; Sternfeld 2009). These findings are consistent with reviews that also report considerable benefit of Internet‐based interventions in increasing physical activity (Davies 2012).  

Internet‐based interventions have been shown to be effective for people with multiple sclerosis in reducing medication discontinuation and increasing patients’ intentions towards medication persistency (Liang 2006). A review has examined telehealth interventions for secondary prevention of CHD (Neubeck 2009). The review included 11 studies, of which two used Internet‐based interventions; the remaining nine studies evaluated interventions delivered via telephone. The overall findings suggest that telehealth interventions are useful in the secondary prevention of CHD, showing improvements in the risk factor profile of patients with the intervention. The two Internet‐based studies included in this review present positive findings both in terms of clinical events, in Southard 2003, and cardiovascular risk factor profile (Zutz 2007). Munro 2013 assessed interventions for heart disease populations that compared Internet‐based cardiac rehabilitation to usual care in a systematic review. This review was broad, including studies of heart failure populations, and cohort study designs. The review included nine studies, which demonstrated improvements in clinical outcomes, physical activity, and psychosocial outcomes. A review recently carried out by Widmer 2015 assessed the benefit of digital health interventions on cardiovascular disease outcomes, and included studies delivering interventions through various digital technologies. Thirteen studies included in this review reported no significant improvements in weight, diastolic blood pressure, triglyceride levels, total cholesterol, low‐density lipoprotein cholesterol, or glucose in secondary prevention populations, but did demonstrate significant reductions in body mass index. A recent study on the feasibility of using Internet‐based interventions in this patient population showed that over 60% of participants surveyed who were eligible for cardiac rehabilitation had Internet access and were confident in opening links and navigating websites (Neubeck 2010), therefore demonstrating potential to reach this population via the Internet. No adverse effects of Internet‐based interventions have been observed in populations studied to date.

Why it is important to do this review

To date, evidence suggests that traditional secondary prevention interventions are effective in reducing adverse outcomes in people with CHD (Heran 2011; Taylor 2010), but that access to services in terms of provision, uptake, and adherence is limited (NACR 2014). Internet‐based interventions may address some of these limitations and be an effective alternative method of providing secondary prevention to this patient group (Griffiths 2006). No systematic reviews have specifically focused on examining the effectiveness of Internet‐based interventions for the secondary prevention of CHD, nor has there been an assessment of Internet intervention effects on HRQOL, lifestyle factors related to CHD, or cost‐effectiveness of these interventions.

Objectives

available in

To determine the effectiveness of Internet‐based interventions targeting lifestyle changes and medicines management for the secondary prevention of CHD.

Methods

available in

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs).

Types of participants

Adults (18 years of age or older) with CHD, including those having experienced a myocardial infarction, a revascularisation procedure (including stent, coronary artery bypass grafting, or percutaneous transluminal coronary angioplasty), those with angina, or angiographically defined CHD.

Types of interventions

We considered all Internet‐based interventions designed to promote a healthy lifestyle and medicines management and reduce cardiovascular risk in people with CHD.

We defined Internet‐based interventions as individually targeted interactive computer‐mediated applications available via the Internet. We only included interventions delivered via the Internet, and therefore studies considered for this review were primarily computer based, although we did consider that Internet‐based interventions may also be delivered via smartphone technology.

We excluded interventions delivered via other technologies that did not require an Internet connection.

We specifically excluded Internet‐based interventions that focused on smoking cessation, as this was the subject of a Cochrane review registered with the Tobacco Addictions group (Civljak 2010). This did not include trials where smoking cessation formed part of a package of care.

We only considered trials where the comparison group was usual care or no intervention, and where follow‐up was reported at least three months postintervention.

Types of outcome measures

Primary outcomes

Clinical outcomes:

  • Mortality (cardiovascular and overall)

  • Non‐fatal re‐infarction

  • Revascularisation

Cardiovascular risk factors:

  • Lipid levels (total cholesterol, high‐density lipoprotein (HDL), low‐density lipoprotein (LDL), and triglycerides)

  • Blood pressure (systolic blood pressure and diastolic blood pressure)

  • HRQOL

Secondary outcomes

  • Lifestyle changes in diet and physical activity. Where possible, we focused on objective measures of lifestyle change.

  • Compliance with medication

  • Healthcare utilisation and costs

  • Adverse intervention effects

Search methods for identification of studies

Electronic searches

We searched the following electronic databases between October and December 2014:

  • Cochrane Central Register of Controlled Trials (CENTRAL, Issue 11 of 12, 2014) on the Cochrane Library

  • MEDLINE (OVID, 1946 to November week 4 2014)

  • EMBASE Classic and EMBASE (OVID, 1947 to 2014 December 22)

  • PsycINFO (OVID, 1806 to October week 3 2014)

  • CINAHL on EBSCOhost (to 17 October 2014)

  • Science Citation Index Expanded (SCI‐EXPANDED), Social Sciences Citation Index (SSCI), and Conference Proceedings Citation Index‐Science (CPCI‐S) on Web of Science (Thomson Reuters, 1970 to 15 October 2014)

  • Health Technology Assessment (HTA), Database of Abstracts of Reviews of Effects (DARE), and NHS Economic Evaluation Database (NEED) on the Cochrane Library (Issue 3 of 4, 2014)

We also used medical subject headings (MeSH) or equivalent and text word terms. We applied no language restrictions.

We have listed the search strategies in Appendix 1. We used the Cochrane sensitivity‐maximising RCT filter for MEDLINE and adaptations of it for use in the other databases (Lefebvre 2011), except CENTRAL and PsycINFO.

Searching other resources

We checked reference lists of reviews and retrieved articles for additional studies.

In January 2015, we further searched the metaRegister of controlled trials (mRCT) (www.controlled‐trials.com/mrct), ClinicalTrials.gov (www.clinicaltrials.gov), and the WHO International Clinical Trials Registry Platform (ICTRP) (http://apps.who.int/trialsearch/) for ongoing trials. We used different combinations of the following search terms: coronary heart disease, cardiovascular disease, Internet, web‐based, world wide web, and online.

We handsearched the Journal of Medical Internet Research and proceedings from the World Congress on Medical and Health Informatics (MEDINFO) for additional studies from the last five years. We contacted authors where necessary for additional information.

Data collection and analysis

Selection of studies

Two review authors (RD and shared between EI, EF, and SS) independently screened the titles and abstracts of all records using a checklist to identify relevant papers. We then obtained the full‐text reports of potentially relevant studies and applied our inclusion criteria to select studies for inclusion. Other review authors (JP or KR) were consulted when there were disagreements between review authors about study selection.

Data extraction and management

Two review authors (RD and shared between EI, EF, and SS) independently extracted data using a proforma. We contacted chief investigators for additional information if necessary. We extracted details regarding the study methodology, participant characteristics, study setting, intervention design (frequency, duration, intensity, level of interactivity, and the focus of the intervention), outcome data (including details of outcome assessment), adverse effects, and methodological quality (randomisation, blinding, attrition) from each of the included studies. Other review authors (JP or KR) were consulted when there were disagreements between review authors about data extraction.

Assessment of risk of bias in included studies

Two review authors (RD and shared between EF, SS, and JP) assessed risks of selection, performance, detection, attrition, and reporting bias using the Cochrane 'Risk of bias' tool (Higgins 2011).

Measures of treatment effect

We processed data in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We expressed dichotomous outcomes as odds ratios or risk ratios, and we calculated 95% confidence intervals for each study. We compared net changes for continuous variables (that is intervention group minus control group differences) and calculated a weighted mean difference or standardised mean difference and 95% confidence intervals for each study.

Assessment of heterogeneity

We carried out tests of heterogeneity (using the Chi2 test of heterogeneity and I2 statistic) for each outcome. In the case of no heterogeneity, we performed a fixed‐effects meta‐analysis. If we detected significant heterogeneity, we looked for possible explanations (for example participants and intervention). If the heterogeneity was not explainable, the review authors considered the following options: provide a narrative overview and not aggregate the studies at all, or use a random‐effects model with appropriate cautious interpretation. We used the Cochrane Handbook for Systematic Reviews of Interventions as a guide to interpret the I2 statistic where taken in consideration with the magnitude and direction of effect and strength of evidence for heterogeneity from the confidence interval for the I2 statistic or P value from the Chi2 test are as follows: 0% to 40% might not be important, 30% to 60% may represent moderate heterogeneity, 50% to 90% may represent substantial heterogeneity, and 75% to 100% considerable heterogeneity.

Subgroup analysis and investigation of heterogeneity

We intended to conduct subgroup analyses on:

  1. Multi‐component Internet‐based interventions versus single‐component interventions, however all the interventions contained multiple components, and none were single‐component interventions. Therefore we were unable to carry out this subgroup analysis.

2. Internet interventions as part of a broader package of care including non‐Internet‐based interventions versus Internet‐only interventions.

We planned to examine the effect of intensity and duration of the intervention (in terms of number of contacts, support given, and interactivity) and period of follow‐up using stratified analyses or meta‐regression. However, the number of studies with usable data was insufficient to explore this formally.

Usual care was defined by the study and included some measures focused on secondary prevention. We intended to examine the intensity of secondary prevention measures in the comparison group compared to that in the experimental group, but there were insufficient trials included for us to do this.  

Sensitivity analysis

We intended to conduct sensitivity analyses excluding studies at high risk of bias and to produce funnel plots and tests of asymmetry to assess possible publication bias (Egger 1997), but the number of included trials was insufficient for us to do this.

Results

Description of studies

See Characteristics of included studies and Characteristics of excluded studies.

Results of the search

The searches yielded a total of 21,459 potentially relevant studies, which we reduced to 14,841 after removing duplications. From these we short‐listed 111 studies. We examined the full papers to these studies, which resulted in including 16 published papers (11 completed trials reported in 12 publications and four trial protocols reported in four published articles).

We searched the reference lists of included studies and relevant review articles, which resulted in no additional studies being included. We also searched trial registers and identified three additional ongoing trials. We have summarised this process in Figure 1.


Study flow diagram.

Study flow diagram.

Overall, we included 11 completed trials (12 publications), Antypas 2014, Devi 2014, Frederix 2015, Lear 2014, Lindsay 2009, Maddison 2014, Reid 2012, Southard 2003, Varnfield 2014, Vernooij 2012, and Zutz 2007, and seven ongoing trials, Dale 2014, ISRCTN29243064, NCT02228603, NCT02350192, Redfern 2014, Reinwand 2013, and Shah 2011. We have outlined full study details and risks of bias in completed trials in the Characteristics of included studies table. We have outlined details of the ongoing trials in the Characteristics of ongoing studies table.

Recruitment

Included studies that are complete with data available were conducted in the USA (Southard 2003; Zutz 2007), Canada (Lear 2014; Reid 2012), the UK (Devi 2014; Lindsay 2009), the Netherlands (Vernooij 2012), Belgium (Frederix 2015), Norway (Antypas 2014), New Zealand (Maddison 2014), and Australia (Varnfield 2014). Four trials recruited participants from conventional cardiac rehabilitation programmes (Antypas 2014; Frederix 2015; Varnfield 2014; Zutz 2007). One trial was set in tertiary care and recruited people who had undergone percutaneous coronary revascularisation and were not planning on taking part in cardiac rehabilitation (Reid 2012). One trial recruited cardiac inpatients from both a tertiary and a regional hospital (Lear 2014). One trial recruited participants from two metropolitan hospitals, through outpatient clinics (Maddison 2014). Two trials recruited participants from primary care general practitioner practices (Devi 2014; Lindsay 2009), one trial recruited from primary and community health services (Varnfield 2014), and two trials recruited from both primary and secondary care (Southard 2003; Vernooij 2012).

A total of 1392 participants were recruited, with sample sizes ranging from 15 to 330 participants. Ten studies reported the mean age of study participants, which ranged from 54.9 to 66.27 years, and one trial reported the median age of participants in the control and experimental group as 58.4 years and 61.7 years, respectively (Lear 2014). In six studies, over 80% of participants were male (Frederix 2015; Lear 2014; Maddison 2014; Reid 2012; Varnfield 2014; Zutz 2007), and in five studies over 70% of participants were male (Antypas 2014; Devi 2014; Lindsay 2009; Southard 2003; Vernooij 2012).

Southard 2003 recruited participants diagnosed with CHD, congestive heart failure, or both. The medical characteristics of participants recruited are described, amalgamating those with multiple diagnoses. In the sample 9.6% had a past medical history of congestive heart failure, and therefore the sample consisted of mostly people with CHD. Reid 2012 recruited a mixture of percutaneous coronary intervention (PCI), acute myocardial infarction (AMI), and coronary artery bypass grafting (CABG) patients. Vernooij 2012 sample included a mixture of coronary artery disease, cerebrovascular disease, abdominal aortic aneurysm, and peripheral vascular disease, and Frederix 2015 sample included post‐PCI and post‐CABG patients. Lear 2014 recruited cardiac inpatients admitted for either acute coronary syndrome or revascularisation procedure. Zutz 2007 recruited a mixture of myocardial infarction (MI), PCI, CABG, and diabetes mellitus patients. Devi 2014 recruited a stable angina population, Maddison 2014 recruited both angina and MI patients, and Varnfield 2014 recruited post‐MI patients. Lindsay 2009 and Antypas 2014 did not provide specific details of CHD diagnosis. Three trials described participants' ethnicity: Southard 2003 and Devi 2014 samples consisted of 97.1% and 91% white participants, respectively, and Maddison 2014 recruited predominately European New Zealanders (76%).

Interventions

Of the 11 completed trials, seven interventions were broad, targeting the general management of CHD (Devi 2014; Lear 2014; Lindsay 2009; Southard 2003; Varnfield 2014; Vernooij 2012; Zutz 2007), and four interventions were focused on physical activity promotion (Antypas 2014; Frederix 2015; Maddison 2014; Reid 2012). Seven interventions were delivered using the Internet only (Devi 2014; Lear 2014; Lindsay 2009; Reid 2012; Southard 2003; Vernooij 2012; Zutz 2007), and four interventions were delivered through both the Internet and mobile telephone technology (Antypas 2014; Frederix 2015; Maddison 2014; Varnfield 2014).

The intervention evaluated by Lindsay 2009 was an online heart care support community, where participants interacted with each other in one of five discussion forums moderated by researchers. The intervention also contained information resources about CHD, diet, exercise, and smoking. The web‐based programme evaluated by Southard 2003 was based around educational modules and involved interactive features such as multiple‐choice self test questions, an online discussion group, and a feature that allowed participants to upload health information, for example exercise and blood pressure. The health information provided by the user was then used to produce graphic feedback which showed the user's progress over time. The intervention evaluated by Vernooij 2012 was a personalised website containing an overview of participants' risk factors and self management information about different CHD risk factors. The intervention evaluated by Devi 2014 involved tailored goal‐setting for exercise, diet, anxiety and emotions, and smoking. Depending on the participant’s performance, these goals were modified/made increasingly difficult throughout the programme. Programme users also had to complete an online interactive exercise diary, uploading the daily number of exercise minutes carried out. The intervention evaluated by Varnfield 2014 used a smartphone to monitor health and exercise and to deliver motivational and educational materials via text messages and through audio or video files. The health and exercise daily diary entries were synchronised to a web portal, which mentors could access and review to give feedback during a weekly scheduled telephone consultation. The web‐based programme evaluated by Zutz 2007 comprised weekly education sessions, scheduled one‐on‐one chat sessions with various healthcare professionals, and monthly 'ask an expert' group chat sessions. Participants were also required to upload exercise levels, heart rate, weight, blood pressure, and glucose levels (if diabetic) data to the website. The trial carried out by Lear 2014 evaluated an intervention consisting of weekly education sessions (in the form of interactive slide presentations), a feature to upload participant health data, progress notes (for healthcare professionals), scheduled one‐on‐one chat sessions with a healthcare professional, and monthly 'ask an expert' group chat sessions.

Trials carried out by Antypas 2014, Frederix 2015, Maddison 2014, and Reid 2012 focused on promoting physical activity only. Both Antypas 2014 and Frederix 2015 investigated the effectiveness of an Internet‐ and mobile phone‐based intervention offered to participants after completing traditional cardiac rehabilitation. Antypas 2014 offered participants tailored motivation and support through both a website and text messaging, physical activity goal‐setting, and access to generic information. In Frederix 2015, participants wore a physical activity monitor, and uploaded weekly data via a USB connection to an online participant account, from which they received automated personalised feedback weekly via email or text messages. Similar to both Antypas and Frederix, Maddison 2014 aimed to increase moderate and vigorous aerobic exercise in those with CHD with an intervention comprised of personalised automated text messages, pedometer‐based step counts feedback, personalised feedback on a website, video messages, motivational messages, and weekly health and exercise tips. Reid 2012 aimed to increase levels of physical activity in those not taking part in traditional cardiac rehabilitation; intervention participants uploaded their daily physical activity data onto the intervention website and completed a series of online tutorials, which generated new physical activity plans.

The length of the interventions ranged from six weeks, in Devi 2014, to one year (Antypas 2014; Vernooij 2012). The duration of the other interventions was three months (Zutz 2007), four months (Lear 2014), four and a half months (Frederix 2015), six months (Maddison 2014; Reid 2012; Southard 2003), and nine months (Lindsay 2009). The intervention evaluated by Varnfield 2014 consisted of a six‐week intervention, which was followed by a six‐month self management phase. Participants received training on how to use the intervention in 10 trials (Antypas 2014; Devi 2014; Frederix 2015; Lear 2014; Lindsay 2009; Maddison 2014; Reid 2012; Varnfield 2014; Vernooij 2012; Zutz 2007). Details of how the intervention was introduced to participants was not described by Southard 2003.

In eight interventions, participants were able to initiate communication with a healthcare professional (Antypas 2014; Devi 2014; Lear 2014; Reid 2012; Southard 2003; Varnfield 2014; Vernooij 2012; Zutz 2007). Communication was through email access (Devi 2014; Reid 2012; Zutz 2007), private‐messaging function on the website (Antypas 2014; Southard 2003; Vernooij 2012; Zutz 2007), one‐to‐one chat facility (Lear 2014), a synchronised group chat (Devi 2014; Lear 2014; Zutz 2007), an online discussion forum (Lindsay 2009; Southard 2003), or telephone consultations (Varnfield 2014). The healthcare professionals communicating with participants were exercise specialists (Lear 2014; Reid 2012; Zutz 2007), dietitians (Lear 2014; Southard 2003; Zutz 2007), nurse practitioners (Lear 2014; Vernooij 2012; Zutz 2007), cardiac rehabilitation specialists (Devi 2014), or a physiotherapist (Antypas 2014). Varnfield 2014 did not describe the professional background of those delivering the telephone consultation component of the intervention. Only one intervention included a prompt feature, in which a nurse practitioner could message a participant, or telephone participants who had not recently logged on to the programme (Vernooij 2012).

The intervention user was able to communicate with other intervention users in four studies, either through online discussion forums (Lindsay 2009; Southard 2003), messaging on other users' profile pages (Antypas 2014), via an online synchronised group chat (Devi 2014; Lear 2014), or through email (Southard 2003).

Control groups

In all studies, the web‐based intervention was evaluated in two‐arm trials.

Usual care was the comparison group in six trials. One study did not provide details of usual care (Southard 2003), and in the others it was usual general practitioner care (Devi 2014; Lear 2014; Vernooij 2012), no intervention (Frederix 2015), or wait‐list control (Zutz 2007).

In three trials, control group participants received a minimal intervention consisting of a static, non‐tailored web‐based programme (Antypas 2014), weekly drop‐in sessions (Lindsay 2009), or general physical activity guidance and an educational booklet (Reid 2012). In two trials, the comparison group was traditional cardiac rehabilitation (Maddison 2014; Varnfield 2014).

Use of the intervention by participants

Five trials reported the frequency of website login. Southard 2003 reported that on average participants logged into the website 58 times over the six‐month intervention period, which was equivalent to twice per week. Devi 2014 reported an average of 19 logins per participant over the six‐week intervention, with an average of three logins per week. In Zutz 2007, the average number of logins over the 12‐week programme was 50, which averaged to 4.2 times per week. Vernooij 2012 and Lear 2014 did not report average values, instead reporting median and range values, respectively. In Vernooij 2012, participants logged in a median of 56 (interquartile range 35 to 83) times during the 12‐month intervention, and in Lear 2014, weekly logins ranged from one participant not logging in to the website at all to other participants logging in more than eight times per week.

Excluded studies

We have presented reasons for study exclusion in the Characteristics of excluded studies table.

Risk of bias in included studies

We judged the risk of bias in the 11 completed trials; Figure 2 and Figure 3 outline summaries of our judgements presented as percentages across all studies, and for each included study, respectively.


Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.


Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

All studies randomly allocated participants to study conditions.

In eight studies there was a low risk of bias in the method used to generate randomisation sequence (Antypas 2014; Devi 2014; Lear 2014; Maddison 2014; Reid 2012; Southard 2003; Varnfield 2014; Vernooij 2012), and in three studies there was an unclear risk (Frederix 2015; Lindsay 2009; Zutz 2007). The unclear risk was due to insufficient details provided in reports to judge adequate randomisation.

In seven studies there was a low risk of bias in the methods used to conceal participant allocation (Antypas 2014; Devi 2014; Frederix 2015; Lear 2014; Maddison 2014; Reid 2012; Varnfield 2014). In four studies there was an unclear risk (Lindsay 2009; Southard 2003; Vernooij 2012; Zutz 2007), as they did not describe the measures taken to ensure concealment of group allocation.

Blinding

Participants were not blinded in 10 trials (Devi 2014; Frederix 2015; Lear 2014; Lindsay 2009; Maddison 2014; Reid 2012; Southard 2003; Varnfield 2014; Vernooij 2012; Zutz 2007). We judged this lack of blinding to be at high risk of bias for study outcomes in seven studies (Devi 2014; Lindsay 2009; Maddison 2014; Reid 2012; Southard 2003; Varnfield 2014; Zutz 2007). In contrast, we judged this to be at low risk of causing bias in three studies (Frederix 2015; Lear 2014; Vernooij 2012), as the study outcome assessments were not likely to be influenced by the lack of blinding. One trial compared two Internet intervention conditions, the web‐based intervention that was under trial and the control group, which received a non‐tailored version of the programme, and therefore in this study it was possible to blind participants to study conditions (Antypas 2014).

In terms of blinding outcome assessors, five studies blinded outcome assessors to group allocation (Antypas 2014; Lear 2014; Maddison 2014; Southard 2003; Vernooij 2012), of which one, Southard 2003, reported inadequate blinding. The outcome assessor was not blinded in six trials (Devi 2014; Frederix 2015; Lindsay 2009; Reid 2012; Varnfield 2014; Zutz 2007), of which five were judged to be at high risk of bias as study outcomes may have been influenced (Devi 2014; Lindsay 2009; Reid 2012; Varnfield 2014; Zutz 2007), whereas this may have not been the case in Frederix 2015 due to the nature of physiological outcome measures used.

Incomplete outcome data

We were unable to judge attrition bias in Lindsay 2009 due to a discrepancy in participant drop‐out reported in the published paper. Attempts made to contact the authors were unsuccessful.

In the remaining 10 trials, participant drop‐out varied, ranging from 4%, in Southard 2003, to 72%, in Antypas 2014.

Six trials achieved follow‐up of 80% or more (Frederix 2015; Lear 2014; Maddison 2014; Southard 2003; Vernooij 2012; Zutz 2007), of which four reported reasons for participant drop‐out and were judged as unlikely to be at risk of bias (Frederix 2015; Lear 2014; Maddison 2014; Southard 2003). Two of these trials did not report reasons for participant drop‐out (Vernooij 2012; Zutz 2007), although due to the low level of attrition this was unlikely to have caused bias.

We also judged other trials with attrition rates of 23%, in Devi 2014, and 31%, in Reid 2012, to be at low risk of attrition bias. In Devi 2014, the number of and reasons for dropouts were balanced across groups. In Reid 2012, the number of dropouts was balanced across groups, and missing data was replaced using multiple imputations.

In contrast, we judged both Antypas 2014 and Varnfield 2014 to be at high risk of attrition bias. Antypas 2014 reported a high attrition rate of 72%, and does not describe reasons for missing data, and the drop‐out rate in Varnfield 2014 was 40% and judged to be related to the trial's primary outcome measure (uptake, adherence, and completion rates of the intervention).

Selective reporting

The risk of selective reporting was unclear in four studies where the study protocol was not available (Frederix 2015; Lindsay 2009; Southard 2003; Zutz 2007). We judged two studies to be at low risk (Antypas 2014; Lear 2014), as all prespecified outcomes outlined in the protocol were reported. We judged five studies to be at high risk, as some variables described in trial protocols were not reported (Devi 2014; Maddison 2014; Reid 2012; Varnfield 2014; Vernooij 2012), although four studies did report the primary outcome (Devi 2014; Maddison 2014; Reid 2012; Vernooij 2012), and in one trial the reported primary outcome measure differed from the primary outcome measure described in the study protocol (Varnfield 2014).

Other potential sources of bias

The information provided in the included studies was insufficient to determine other potential sources of bias.

Effects of interventions

See: Summary of findings for the main comparison Internet‐based interventions compared to usual care or no care for prevention of coronary heart disease; Summary of findings 2 Internet‐based interventions compared to usual care or no care for prevention of coronary heart disease

Effects of interventions on clinical outcomes

Seven studies reported clinical outcomes (Frederix 2015; Lear 2014; Maddison 2014; Reid 2012; Southard 2003; Vernooij 2012; Zutz 2007). One group reported overall cardiovascular events but did not break this down further (Southard 2003), so we were unable to combine data from this study in the meta‐analyses (contact was made with the authors, however they no longer have access to the data). This study reported that two and eight participants in the intervention group and control group, respectively, experienced a cardiovascular event. The difference between groups was of borderline statistical significance, P = 0.053.

Mortality

Six trials reported this as an outcome, with a total of 895 participants randomised (Analysis 1.1). A total of four deaths were reported across three of these trials, all in the control groups (Lear 2014; Reid 2012; Vernooij 2012). Vernooij 2012 reported one death described as a "fatal cerebrovascular event", Lear 2014 reported one death as a non‐cardiovascular disease death, and Reid 2012 reported two cases of mortality with no reasons provided. The odds ratio (OR) was 0.27 (95% confidence interval (CI) 0.04 to 1.63), P = 0.15 (Analysis 1.1), with low‐quality evidence (summary of findings Table for the main comparison).

Only one trial reported a case of cardiovascular‐related mortality; Vernooij 2012 reported one cardiovascular‐related death in the control group.

Non‐fatal re‐infarction

No studies reported any incidences of non‐fatal re‐infarction.

Revascularisation

Six studies contributed to the analysis with 895 participants randomised (Analysis 1.2). In total, 18 revascularisations were reported amongst the intervention groups in three studies (Lear 2014; Maddison 2014; Vernooij 2012), and 26 in the control groups across five studies (Frederix 2015; Lear 2014; Maddison 2014; Reid 2012; Vernooij 2012), with no evidence of an effect of the intervention (OR 0.69 (95% CI 0.37 to 1.27), P = 0.23) (Analysis 1.2, Figure 4), and with low‐quality evidence (summary of findings Table for the main comparison).


Forest plot of comparison: 1 Cardio events, outcome: 1.1 Revascularisation.

Forest plot of comparison: 1 Cardio events, outcome: 1.1 Revascularisation.

Effects of interventions on cardiovascular risk factors

Lipid levels

Six trials assessed the impact of web‐based interventions on total cholesterol, HDL cholesterol, LDL cholesterol, and triglycerides (Frederix 2015; Lear 2014; Southard 2003; Varnfield 2014; Vernooij 2012; Zutz 2007). We were unable to combine data from Southard 2003 and Lear 2014. Southard 2003 did not report variance values, and we were unable to obtain these from the authors, and Lear 2014 reported data in terms of medians and interquartile ranges. At six months' follow‐up, Southard 2003 reported no statistically significant changes in cholesterol levels, and Lear 2014 reported statistically significant group differences between the control and intervention group for total cholesterol (P = 0.026) and LDL cholesterol (P = 0.022), although not for HDL cholesterol (P = 0.075) and triglycerides (P = 0.715). For total cholesterol, the median values in the control group at baseline, four months, and 16 months were 3.45 mmol, 3.77 mmol, and 3.66 mmol, respectively, and experimental group median values were 3.54 mmol, 3.68 mmol, and 3.60 mmol, respectively, P = 0.026. For LDL cholesterol, the median values in the control group at baseline, four months, and 16 months were 1.79 mmol, 1.99 mmol, and 1.82 mmol, respectively, and experimental group values were 1.74 mmol, 1.79 mmol, and 1.69 mmol, respectively, P = 0.022.

There were four studies where data could be combined, there was moderate heterogeneity for the outcome total cholesterol and a random‐effects model was used (I2 = 41%) showing no effect of the intervention on total cholesterol (mean difference (MD) 0.00 (95% CI ‐0.27 to 0.28) mmol/L, P = 0.98, four studies, 439 participants) (Analysis 2.1, Figure 5) with low‐quality evidence (summary of findings Table 2). Similarly, there were no intervention effects for HDL cholesterol (MD 0.01 (95% CI ‐0.06 to 0.07) mmol/L, P = 0.82, four studies, 437 participants) (Analysis 3.1, Figure 6) with low‐quality evidence (summary of findings Table 2). There was considerable heterogeneity for the LDL cholesterol outcome, and results were not pooled statistically (I2 = 77%) (Analysis 4.1, Figure 7), low‐quality evidence (summary of findings Table 2). Of these studies, one reported a difference in LDL cholesterol between groups at 12 months' follow up (MD ‐0.3 (95% CI ‐0.5 to ‐0.1) (Vernooij 2012) other trials reported no effect on LDL levels (Frederix 2015; Varnfield 2014; Zutz 2007). For triglycerides, there were again no intervention effects (MD 0.01 (95% CI ‐0.17 to 0.19) mmol/L, P = 0.91, four studies, 439 participants) (Analysis 5.1, Figure 8).


Forest plot of comparison: 2 Total cholesterol, outcome: 2.1 Total Cholesterol.

Forest plot of comparison: 2 Total cholesterol, outcome: 2.1 Total Cholesterol.


Forest plot of comparison: 3 HDL cholesterol, outcome: 3.1 HDL Cholesterol.

Forest plot of comparison: 3 HDL cholesterol, outcome: 3.1 HDL Cholesterol.


Forest plot of comparison: 4 LDL cholesterol, outcome: 4.1 LDL Cholesterol.

Forest plot of comparison: 4 LDL cholesterol, outcome: 4.1 LDL Cholesterol.


Forest plot of comparison: 5 Triglycerides, outcome: 5.1 Triglycerides.

Forest plot of comparison: 5 Triglycerides, outcome: 5.1 Triglycerides.

Blood pressure

In total, seven studies measured systolic (SBP) and diastolic (DBP) blood pressure (Devi 2014; Lear 2014; Maddison 2014; Southard 2003; Varnfield 2014; Vernooij 2012; Zutz 2007). We were unable to combine data from two studies due to variance data not being reported (Southard 2003), and due to data being presented using median values (Lear 2014). Southard 2003 reported no effects for SBP or DBP between groups at a six months' follow‐up (P values not provided), and Lear 2014 reported a between‐group difference over time for SBP (P = 0.051), although not for DBP (P = 0.776). The median SBP values in the control group at baseline, four months', and 16 months' follow‐up were 112 mmHg, 114 mmHg, and 117 mmHg, and the experimental group median values were 121 mmHg, 126 mmHg, and 121 mmHg, respectively.

For the remaining five trials, heterogeneity was substantial (I2 = 63% for SBP, 58% for DBP), and so we did not pool results statistically (Analysis 6.1, Figure 9; Analysis 7.1, Figure 10). In Vernooij 2012, the difference in SBP at 12 months' follow‐up in the intervention group was ‐3 (standard deviation (SD) = 17.52) mmHg, and in the control group the difference was 2 (SD = 18.52) mmHg; this was reported within a 95% CI of ‐7.6 to 0.2. For DBP, the difference at a 12 months' follow‐up in the intervention group and the control group was ‐1 (SD = 9.54) mmHg, and 1 (SD = 10) mmHg, respectively; this was reported with a 95% CI of ‐4.4 to 0.4 (Vernooij 2012). Varnfield 2014 reported six weeks' follow‐up data in the published findings; we contacted the authors for the six months' follow‐up data, however this was not made available. At six weeks' follow‐up, Varnfield 2014 reported an intervention effect for DBP (P = 0.03), while the intervention effect on SBP was not significant (P = 0.4). Maddison 2014 measured DBP and SBP, which is not reported in the published findings; the authors were contacted and the findings provided. At six months' follow‐up, the change in SBP in the intervention group was 4.77 (SD = 13.39) mmHg, and in the control group 0.29 (SD = 13.37) mmHg. The change in DBP at six months' follow‐up in the intervention group was 1.23 (SD = 9.27) mmHg and in the control group ‐1.73 (SD = 10.16) mmHg. Both Devi 2014 and Zutz 2007 reported no significant SBP and DBP effects between the intervention and control groups at six months', and 12 weeks' follow‐up, respectively.


Forest plot of comparison: 6 Systolic blood pressure, outcome: 6.1 Systolic Blood Pressure.

Forest plot of comparison: 6 Systolic blood pressure, outcome: 6.1 Systolic Blood Pressure.


Forest plot of comparison: 7 Diastolic blood pressure, outcome: 7.1 Diastolic Blood Pressure.

Forest plot of comparison: 7 Diastolic blood pressure, outcome: 7.1 Diastolic Blood Pressure.

Health‐related quality of life

Five studies measured changes in health‐related quality of life (HRQOL) (Devi 2014; Maddison 2014; Reid 2012; Southard 2003; Varnfield 2014). Due to a lack of homogeneity in instruments used across studies, we could not combine these findings in a meta‐analysis. Southard 2003 used the Dartmouth COOP, however only baseline data was described, and no follow‐up findings reported. We contacted the authors, and unfortunately this data is no longer available. Both Devi 2014 and Reid 2012 used the MacNew to measure HRQOL; we were unable to combine this data as Reid 2012 did not report baseline scores. Devi 2014 showed that compared to a control group the experimental group demonstrated statistically significant improvements in emotional HRQOL (P = 0.04, 95% CI 0.01 to 0.54) at six weeks' follow‐up and statistically significant improvements in social HRQOL at six months' follow‐up (P = 0.018, 95% CI 0.05 to 0.54). In addition, Maddison 2014 reported statistically significant intervention effects in the general health domain of the 36‐Item Short Form Health Survey at 24 weeks (mean difference 2.1, 95% CI 0.1 to 4.1; P = 0.03), while there were no statistically significant differences in other HRQOL domains. Varnfield 2014 reported a statistically significant improvement between groups in HRQOL measured using the EQ‐5D index at six weeks' follow‐up (adjusted MD ‐0.08, 95% CI ‐0.14 to ‐0.02, P = 0.01), however this improvement was not maintained at six months' follow‐up. Reid 2012 reported that the differences between groups over time at six and 12 months' follow‐up in HRQOL domains were not statistically significant.

Effects of interventions on lifestyle changes

Diet

Five trials assessed dietary outcomes (Devi 2014; Lear 2014; Lindsay 2009; Southard 2003; Varnfield 2014), and findings were inconclusive. Southard 2003 used MEDFICTS, a measure of fat and cholesterol intake, and reported no statistically significant changes at six months' follow‐up. Devi 2014 measured diet using the DINE (Dietary Instrument for Nutrition Education) and reported no statistically significant dietary effects at both six weeks' and six months' follow‐up. Varnfield 2014 assessed diet using Dietary Habits Questionnaire; this study did not report six months' follow‐up data, and at six weeks reported no statistically significant differences between groups in fat intake (P = 0.4), fibre intake (P = 0.7), sodium (P = 0.4), or alcohol (P = 0.6). Lear 2014 reported no significant group differences over time in carbohydrate (P = 0.224) and fat (P = 0.451) intake, but reported statistically significant intervention effects in protein (P = 0.044) and saturated fat (P = 0.018) intake. Lindsay 2009 also demonstrated a positive finding. This study measured the frequency of unhealthy foods eaten using variables from the Health Survey for England and reported that after six months of using the intervention, the experimental group ate unhealthy foods less often compared to the control group (P = 0.014); this change was not sustained at nine months' follow‐up (P = 0.517).

Physical activity

Eight trials assessed physical activity effects. Six studies used self report measures (Antypas 2014; Lear 2014; Lindsay 2009; Maddison 2014; Southard 2003; Zutz 2007), one used an objective measure (Devi 2014), and one used both an objective and a self report measure (Reid 2012).

Of the studies using self report measures, two used the International Physical Activity Questionnaire (IPAQ) (Antypas 2014; Maddison 2014), two used the Minnesota Leisure Time Physical Activity Questionnaire (Lear 2014; Zutz 2007), and two used unstandardised measures (Lindsay 2009; Southard 2003). In both Southard 2003 and Lindsay 2009, there were no significant physical activity effects.

Even though Maddison 2014 and Antypas 2014 both used the IPAQ, we could not combine the data, as Maddison 2014 reported means and standard deviations, and Antypas 2014 reported median and interquartile range values. Maddison 2014 reported a statistically significant increase in self reported leisure time physical activity (mean difference 110.2 min/week, 95% CI ‐0.8 to 221.3; P = 0.05) and walking (mean difference 151.4 min/week, 95% CI 27.6 to 275.2; P = 0.02) at 24 weeks in favour of the intervention group, which represents increases of 40% and 42%, respectively. Maddison 2014 reported no statistically significant differences for the other activity domains (total activity, active transport, domestic/gardening, and reduced sitting time). Antypas 2014 reported that at three months' follow‐up the intervention group had a significantly higher IPAQ score than the control group, P = 0.02, and higher levels of walking than the control group (P value not reported). There were no significant differences between groups in moderate and vigorous activity or time spent sitting at a one and three months' follow‐up (Antypas 2014).

Although Lear 2014 and Zutz 2007 both used the Minnesota Leisure Time Physical Activity Questionnaire, we were unable to combine their findings as Zutz 2007 reported mean values, and Lear 2014 reported median values. Zutz 2007 reported no statistically significant differences between groups (P value not reported). Similarly, Lear 2014 reported that the group differences over time in leisure time physical activity were not statistically significant (P = 0.191).

Two of the included studies used an objective measure to evaluate physical activity. Devi 2014 used an accelerometer to measure a range of outcomes, and reported statistically significant improvements in daily step count (intervention group n = 35, control group n = 40, P = 0.016, 95% CI 263 to 2451), energy expenditure (intervention group n = 35, control group n = 40, P = 0.01, 95% CI 43.93 to 309.98), duration of sedentary activity (intervention group n = 35, control group n = 40, P = 0.012, 95% CI ‐55.01 to ‐7.01), and duration of moderate activity (intervention group n = 35, control group n = 40, P = 0.014, 95% CI 6.01 to 51.20) at six weeks' follow‐up. There were no statistically significant effects at a six months' follow‐up (Devi 2014). Reid 2012 used both an objective (pedometer) and a self reported measure (a modified version of the Godin Leisure‐Time Exercise Questionnaire); however this study did not collect pedometer data at baseline, and only collected data at the six‐ and 12‐month follow‐ups. This study reported that the difference between groups over time in pedometer‐measured activity (P = 0.656) and self reported moderate and vigorous physical activity levels (P = 0.782) was not significant.

Effects of interventions on compliance with medication

No studies have yet measured the impact of web‐based interventions on compliance with medication.

Effects of interventions on healthcare utilisation and cost

Two studies collected data on healthcare utilisation. One study reported that at six months' follow‐up there were no differences between study groups in healthcare utilisation (P = 0.757), and at nine months' follow‐up the intervention group had statistically significantly higher levels of health visits than the control group (P = 0.044) (Lindsay 2009). The other study reported no statistically significant differences between groups in emergency room visits (P = 0.349) (Lear 2014).

Two studies reported on intervention cost‐effectiveness. Maddison 2014 collected information on the cost of implementing and delivering the intervention and described the intervention as likely to be cost‐effective in increasing metabolic equivalent (MET) hours (walking and leisure activity) per week, and for improving HRQOL. Southard 2003 also reported cost‐effectiveness data, and described the estimated cost of the intervention as USD 453 per participant. Based on the medical cost associated with the cardiovascular events that occurred in both study groups (USD 104,684 and USD 31,110 in the control and intervention group, respectively), there was a gross cost savings of USD 1418 per person, and the net cost savings was USD 965 per person. These figures project an estimated return of 213% on the investment.

Adverse intervention effects

An adverse intervention effect was reported in 1 trial. Lindsay 2009 reported statistically significant higher levels of health visits to a GP, nurse, specialist or other health provider in the intervention group at a 9 month follow up, compared to the control group (P = 0.044) (Lindsay 2009).

Interventions including non‐Internet‐based components versus Internet‐only interventions

The Internet‐based interventions tested in Antypas 2014 and Frederix 2015 were provided to patients after they had completed traditional CR. Antypas 2014 aimed to enhance the maintenance of PA, and Frederix 2015 aimed to improve patients' physical fitness. In relation to the outcomes of interest in this review, Antypas 2014 reported the effects on physical activity, and Frederix 2015 reported the effects on clinical outcomes and cardiovasular risk factors. Due to heterogeneity in reported outcomes between both of these studies we were unable to do any meaningful comparisons of these with Internet‐only interventions at this stage. This will be further examined in an update of this review when more evidence has accrued.

Discussion

available in

With the rising prevalence of heart disease and economic pressures to produce low‐resource‐intensive/cost‐saving solutions, Internet‐delivered interventions have the potential to produce high impact. Internet interventions are not restrained by time or geographical location, and an increasing proportion of retired people over the age of 65 are using the Internet (Dutton 2013), reflecting the typical CHD population.

Summary of main results

We identified 11 completed trials with data available. In terms of study outcomes, seven studies measured clinical outcomes, eight assessed cardiovascular risk factors, five measured HRQOL, five measured impact on diet, and eight assessed physical activity. Six of the eight studies that measured physical activity relied upon self reported measures, which could have been affected by social desirability or poor recall. There was heterogeneity between studies, which prevented statistical pooling for some outcomes. For each analysis there were few studies that contributed, and no overall effects were seen for clinical events, although follow‐up was relatively short. In terms of cardiovascular risk factor outcomes, there were no statistically significant effects for total cholesterol, HDL cholesterol, and triglycerides. It was not possible to pool results from studies measuring LDL cholesterol. Of the four trials measuring LDL cholesterol, one reported favourable intervention effects (Vernooij 2012). Five studies measured HRQOL, with three studies finding evidence for improvements (Devi 2014; Maddison 2014; Varnfield 2014), demonstrating positive effects at six weeks, in Devi 2014 and Varnfield 2014, and at six months, in Devi 2014 and Maddison 2014. In terms of diet, one trial found an effect at six months, which was not maintained at a longer follow‐up (Lindsay 2009), and another trial demonstrated effects in protein and saturated fat intake (Lear 2014). There was some evidence to show that Internet‐based interventions have positive effects on physical activity. Eight studies measured physical activity effects, of which three reported improvements. Maddison 2014 reported improved self reported leisure time physical activity and walking at a six months' follow‐up, Antypas 2014 reported improved IPAQ score and walking at a three months' follow‐up, and Devi 2014 reported improved steps, energy expenditure, duration of sedentary activity, and duration of moderate activity at six weeks' follow‐up. No studies have been conducted yet that measure the effects of web‐based interventions on compliance with medication. Two studies measured healthcare utilisation (Lear 2014; Lindsay 2009). One study reported higher levels of healthcare visits in the intervention group compared to the control group at nine months' follow‐up (Lindsay 2009), the other study reported no differences between groups in emergency room visits (Lear 2014). Two studies measured the cost‐effectiveness of the intervention and reported positive findings in favour of the intervention (Maddison 2014; Southard 2003). There was one adverse intervention effect detected, and this was higher levels of healthcare visits at a 9 month follow up in comparison to a control group (Lindsay 2009).

Overall completeness and applicability of evidence

This is a relatively new area of research, with the first trial published in 2003 (Southard 2003), and then 2007 thereafter (Zutz 2007). A variety of interventions were studied, of which seven were broad, targeting the general management of coronary risk factors, and four focused on promoting physical activity. The length of follow‐up and participant characteristics varied between trials. Three trials had a long‐term follow‐up of 12 months, in Lear 2014, Reid 2012, and Vernooij 2012, six trials had a medium‐term follow‐up of six months, in Devi 2014, Maddison 2014, Reid 2012, Southard 2003, and Varnfield 2014, and nine months, in Lindsay 2009, and three trials had short‐term follow‐ups of three months, in Antypas 2014 and Zutz 2007, and 4.5 months, in Frederix 2015. Participant types varied across studies. Three trials recruited participants with a relatively recent manifestation of heart disease. Maddison 2014 recruited participants who in the last three to 12 months had a diagnosis of ischaemic heart disease; in the study by Reid 2012, over half (64.6%) of the sample had their first cardiac event; and in the study by Varnfield 2014, all participants were post‐myocardial infarction patients. One study recruited a primary care angina population (Devi 2014). Two studies recruited mixed CHD populations (Southard 2003; Vernooij 2012). Frederix 2015 recruited a post‐percutaneous coronary intervention and post‐coronary artery bypass grafting population, and, similarly, Lear 2014 recruited cardiac inpatients for either acute coronary syndrome or revascularisation. Zutz 2007 recruited a mixture of myocardial infarction, percutaneous coronary intervention, coronary artery bypass grafting, and diabetes mellitus patients, Antypas 2014 and Lindsay 2009 provided no specific details of CHD diagnosis.

We found no trials of that combined the Internet intervention with a face‐to‐face intervention component. In 10 out of 11 trials, participants were introduced to the intervention using a face‐to‐face consultation, and one trial did not describe how the intervention was introduced to participants (Southard 2003). We found no trials that recruited participants online, and therefore the feasibility of recruitment through the Internet is not yet known.

Studies were conducted in a wide range of countries, and therefore the structure of health care would have differed, limiting our ability to draw generalisable conclusions. The majority of participants were male, with mean ages across 10 studies ranging from 54.9 to 66.27 years, and studies providing details on ethnicity reported that the majority of participants were white, in Devi 2014 and Southard 2003, or New Zealand European (Maddison 2014). This again limits the extent to which these results can be generalised widely. Some studies did provide details on participant marital status (Lear 2014; Southard 2003), education level (Antypas 2014; Lear 2014; Reid 2012; Southard 2003), income (Lear 2014; Lindsay 2009; Southard 2003), and employment (Devi 2014; Southard 2003). Future studies should collect more participant demographics data to enable us to draw conclusions regarding applicability of evidence to wider populations in future updates of this review.

Due to unrestricted access to the Internet, a challenge remaining with web‐based intervention trials is the difficulty in determining the contribution of a specific web‐based programme. Intervention users may use multiple websites to search for information related to the disease, and therefore participants may well be using the intended intervention in conjunction with other sites. Similarly, the control group may also be using the Internet to search for information related to the disease.

The interventions evaluated varied. Seven interventions were delivered through the Internet only, and in four trials the intervention was delivered using both the Internet and mobile phone technology. With the increased use of smartphones and tablets, it is likely that future web‐based interventions will be used through a combination of different technologies, such as through computers, smartphones, and tablets. It is also possible that future interventions may also incorporate and capitalise on the rise in social networking.

In terms of participant engagement in interventions, five trials reported the frequency of participant logins. For a six‐month, three‐month, and six‐week intervention, the average numbers of weekly logins were 2 (Southard 2003), 4.2 (Zutz 2007), and 3 (Devi 2014), respectively. The study by Vernooij 2012 reported a median of 56 logins during a 12‐month intervention, and the study by Lear 2014 reported a range of 0 to greater than 8 logins per week over a four‐month intervention. It is possible that with longer‐length interventions there is a reduction in participant engagement. More trials assessing user engagement are required to enable firm conclusions.

Quality of the evidence

All of the studies included in this review were randomised controlled trials, and we assessed the quality of evidence using the GRADE approach for evidence synthesis. The evidence for the outcomes analysed was generally of low quality as a result of lack of blinding, uncertainty around the magnitude of effect, and loss to follow‐up (summary of findings Table for the main comparison; summary of findings Table 2). The majority of studies provided details about the generation of random sequence and the appropriate concealment of allocation. We were unable to judge the quality of the randomisation method used in three studies (Frederix 2015; Lindsay 2009; Zutz 2007), as these details were not provided. In addition, four trials did not describe the method used to conceal treatment allocation (Lindsay 2009; Southard 2003; Vernooij 2012; Zutz 2007). Ten studies did not blind participants to study groups; we judged this to be likely to cause bias in seven studies (Devi 2014; Lindsay 2009; Maddison 2014; Reid 2012; Southard 2003; Varnfield 2014; Zutz 2007). The outcome assessor was blinded in five trials (Antypas 2014; Lear 2014; Maddison 2014; Southard 2003; Vernooij 2012), of which one was judged to be at high risk of detection bias due to inadequate blinding (Southard 2003). In the six trials where the outcome assessor was not blinded, we judged five to be at high risk of bias as study outcomes may have been influenced (Devi 2014; Lindsay 2009; Reid 2012; Varnfield 2014; Zutz 2007). We also judged the likelihood of attrition bias. We judged two studies to be at high risk of attrition bias due to large attrition rates with no reasons for missing data provided (Antypas 2014), and because attrition was likely to be related to the trial's primary outcome measure (Varnfield 2014). In contrast, we judged eight studies judged to be at low risk of attrition bias (Devi 2014; Frederix 2015; Lear 2014; Maddison 2014; Reid 2012; Southard 2003; Vernooij 2012; Zutz 2007). We were unable to assess attrition bias in one trial due to a discrepancy detected in the published findings (Lindsay 2009). In terms of reporting bias, we judged five studies to be at high risk, as not all the measures outlined in the study protocol had been reported (Devi 2014; Maddison 2014; Reid 2012; Varnfield 2014; Vernooij 2012), although Devi 2014, Maddison 2014, Reid 2012, and Vernooij 2012 did report their primary outcome measures. Antypas 2014 and Lear 2014 were at low risk of reporting bias as all prespecified outcomes outlined in the protocol were described in the trial write‐up. It was not possible to judge risk in Frederix 2015, Lindsay 2009, Southard 2003, and Zutz 2007, as trial protocols were not available.

Potential biases in the review process

The searching for this review was extensive involving a number of different databases, and all review processes were conducted in duplicate to minimise bias. Although we looked for unpublished data, we were unable to find any unpublished randomised controlled trials that fulfilled our inclusion criteria, therefore the review contains published data only. Interventions targeting the secondary prevention of heart disease are often multi‐componential and complex, and due to the nature of cardiac risk factors involve changing lifestyle. When participants make lifestyle changes, various cognitions and psychological aspects are involved, and the complex nature of this means there are a large number of primary and secondary outcomes within trials that are of interest. However, we have only reported on prespecified outcomes as described in the protocol.

The comparison groups differed across trials, consisting of usual care (n = 6), minimal intervention (n = 3), or traditional cardiac rehabilitation (n = 2). We intended where possible to examine the intensity of secondary prevention measures in the comparison group compared to that in the experimental group. However, the number of studies with usable data for meta‐analyses was insufficient to explore this formally in subgroup analyses. The review authors intend to formally study the intensity of the comparison group in subgroup analysis in an update of this review when more evidence has accrued.

The protocol for this review was constructed at a time when smartphone technology was not as widely used as it is today, and therefore the review authors' primary focus was on interventions delivered using Internet websites. In our search we found that more recently conducted trials delivered interventions that combined smartphone and Internet site technology. This shows that these interventions are evolving, and in future updates of this review we intend to distinguish between the level of smartphone and Internet site contributions in the design of interventions.

The conclusions we can draw from this review are currently limited by the small number of included studies and the heterogeneity between studies in terms of the intervention and participant characteristics and length of follow‐up. More trials with longer follow‐ups are required to be able to determine the effects of the interventions on clinical events and whether effects on intermediate outcomes are sustained following the end of the intervention period.

Agreements and disagreements with other studies or reviews

A previous review conducted by Munro 2013 examined the impact of patient Internet‐based approaches to cardiac rehabilitation. This review included nine randomised controlled trials and cohort studies and reported positive findings for clinical outcomes and physical activity with the intervention. Due to the heterogeneity between studies, the authors stated that their results should be interpreted with caution. Widmer 2015 assessed the benefit of digital health interventions on cardiovascular disease outcomes and reported no significant improvements in weight, systolic blood pressure, triglyceride levels, total cholesterol, or low‐density lipoprotein cholesterol in secondary prevention populations. This current review reports similar findings to Widmer 2015.

Study flow diagram.
Figures and Tables -
Figure 1

Study flow diagram.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Figures and Tables -
Figure 2

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Figures and Tables -
Figure 3

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Forest plot of comparison: 1 Cardio events, outcome: 1.1 Revascularisation.
Figures and Tables -
Figure 4

Forest plot of comparison: 1 Cardio events, outcome: 1.1 Revascularisation.

Forest plot of comparison: 2 Total cholesterol, outcome: 2.1 Total Cholesterol.
Figures and Tables -
Figure 5

Forest plot of comparison: 2 Total cholesterol, outcome: 2.1 Total Cholesterol.

Forest plot of comparison: 3 HDL cholesterol, outcome: 3.1 HDL Cholesterol.
Figures and Tables -
Figure 6

Forest plot of comparison: 3 HDL cholesterol, outcome: 3.1 HDL Cholesterol.

Forest plot of comparison: 4 LDL cholesterol, outcome: 4.1 LDL Cholesterol.
Figures and Tables -
Figure 7

Forest plot of comparison: 4 LDL cholesterol, outcome: 4.1 LDL Cholesterol.

Forest plot of comparison: 5 Triglycerides, outcome: 5.1 Triglycerides.
Figures and Tables -
Figure 8

Forest plot of comparison: 5 Triglycerides, outcome: 5.1 Triglycerides.

Forest plot of comparison: 6 Systolic blood pressure, outcome: 6.1 Systolic Blood Pressure.
Figures and Tables -
Figure 9

Forest plot of comparison: 6 Systolic blood pressure, outcome: 6.1 Systolic Blood Pressure.

Forest plot of comparison: 7 Diastolic blood pressure, outcome: 7.1 Diastolic Blood Pressure.
Figures and Tables -
Figure 10

Forest plot of comparison: 7 Diastolic blood pressure, outcome: 7.1 Diastolic Blood Pressure.

Comparison 1 Clinical outcomes, Outcome 1 Total Mortality.
Figures and Tables -
Analysis 1.1

Comparison 1 Clinical outcomes, Outcome 1 Total Mortality.

Comparison 1 Clinical outcomes, Outcome 2 Revascularisation.
Figures and Tables -
Analysis 1.2

Comparison 1 Clinical outcomes, Outcome 2 Revascularisation.

Comparison 2 Total cholesterol, Outcome 1 Total Cholesterol.
Figures and Tables -
Analysis 2.1

Comparison 2 Total cholesterol, Outcome 1 Total Cholesterol.

Comparison 3 HDL cholesterol, Outcome 1 HDL Cholesterol.
Figures and Tables -
Analysis 3.1

Comparison 3 HDL cholesterol, Outcome 1 HDL Cholesterol.

Comparison 4 LDL cholesterol, Outcome 1 LDL Cholesterol.
Figures and Tables -
Analysis 4.1

Comparison 4 LDL cholesterol, Outcome 1 LDL Cholesterol.

Comparison 5 Triglycerides, Outcome 1 Triglycerides.
Figures and Tables -
Analysis 5.1

Comparison 5 Triglycerides, Outcome 1 Triglycerides.

Comparison 6 Systolic blood pressure, Outcome 1 Systolic Blood Pressure.
Figures and Tables -
Analysis 6.1

Comparison 6 Systolic blood pressure, Outcome 1 Systolic Blood Pressure.

Comparison 7 Diastolic blood pressure, Outcome 1 Diastolic Blood Pressure.
Figures and Tables -
Analysis 7.1

Comparison 7 Diastolic blood pressure, Outcome 1 Diastolic Blood Pressure.

Summary of findings for the main comparison. Internet‐based interventions compared to usual care or no care for prevention of coronary heart disease

Internet‐based interventions compared to usual care or no care for prevention of coronary heart disease: clinical outcomes

Patient or population: patients with coronary heart disease
Settings: healthcare settings
Intervention: Internet‐based interventions
Comparison: usual care or no care

Outcomes

Illustrative comparative risks* (95% CI)

Relative effect
(95% CI)

No of Participants
(studies)

Quality of the evidence
(GRADE)

Comments

Assumed risk

Corresponding risk

Usual care or no care

Internet‐based interventions

Total mortality

Study population

OR 0.27
(0.04 to 1.63)

895
(6 studies)

⊕⊕⊝⊝
low1,2

9 per 1000

2 per 1000
(0 to 15)

Moderate risk population

3 per 1000

1 per 1000
(0 to 5)

Revascularisation

Study population

OR 0.69
(0.37 to 1.27)

895
(6 studies)

⊕⊕⊝⊝
low1,2

58 per 1000

41 per 1000
(22 to 73)

Moderate risk population

18 per 1000

12 per 1000
(7 to 23)

*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; OR: Odds ratio

GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

1Maddison 2014, Reid 2012, Zutz 2007 ‐ outcomes may have been influenced by lack of blinding.
2Confidence interval crosses line of no effect (uncertainty around the magnitude of effect).

Figures and Tables -
Summary of findings for the main comparison. Internet‐based interventions compared to usual care or no care for prevention of coronary heart disease
Summary of findings 2. Internet‐based interventions compared to usual care or no care for prevention of coronary heart disease

Internet‐based interventions compared to usual care or no care for prevention of coronary heart disease: cardiovascular risk factors

Patient or population: patients with coronary heart disease
Settings: healthcare settings
Intervention: Internet‐based interventions
Comparison: usual care or no care

Outcomes

Illustrative comparative risks* (95% CI)

Relative effect
(95% CI)

No of Participants
(studies)

Quality of the evidence
(GRADE)

Comments

Assumed risk

Corresponding risk

Usual care or no care

Internet‐based interventions

Total cholesterol

The mean total cholesterol in the intervention groups was
0 higher
(0.27 lower to 0.28 higher)

439
(4 studies)

⊕⊕⊝⊝
low1

LDL cholesterol

See comment

See comment

Not pooled

437
(4 studies)

⊕⊕⊝⊝
low1

High levels increase risk, while low levels reduce risk. Data not pooled due to unexplained considerable (I2 = 77%) heterogeneity

HDL cholesterol

The mean HDL cholesterol in the intervention groups was
0.01 higher
(0.06 lower to 0.07 higher)

437
(4 studies)

⊕⊕⊝⊝
low1

High levels reduce risk, while low levels increase risk

Triglycerides

The mean triglycerides in the intervention groups was
0.01 higher
(0.17 lower to 0.19 higher)

439
(4 studies)

⊕⊕⊝⊝
low1

High levels increase risk, while low levels reduce risk

Systolic blood pressure

See comment

See comment

Not pooled

623
(5 studies)

⊕⊕⊝⊝
low2

Systolic blood pressure measured in 5 studies. Data not pooled due to unexplained substantial (I2 = 63%) heterogeneity

Diastolic blood pressure

See comment

See comment

Not pooled

622
(5 studies)

⊕⊕⊝⊝
low2

Diastolic blood pressure measured in 5 studies. Data not pooled due to unexplained substantial (I2 = 58%) heterogeneity

*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval;HDL: High‐density lipoprotein
; LDL: Low‐density lipoprotein

GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

1Zutz 2007 ‐ outcomes may have been influenced by lack of blinding; Varnfield 2014 ‐ loss to follow‐up > 20% in both experimental arms.
2Devi 2014, Zutz 2007 ‐ outcomes may have been influenced by lack of blinding; Varnfield 2014 ‐ loss to follow‐up > 20% in both experimental arms.

Figures and Tables -
Summary of findings 2. Internet‐based interventions compared to usual care or no care for prevention of coronary heart disease
Comparison 1. Clinical outcomes

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

1 Total Mortality Show forest plot

6

895

Odds Ratio (M‐H, Fixed, 95% CI)

0.27 [0.04, 1.63]

2 Revascularisation Show forest plot

6

895

Odds Ratio (M‐H, Fixed, 95% CI)

0.69 [0.37, 1.27]

Figures and Tables -
Comparison 1. Clinical outcomes
Comparison 2. Total cholesterol

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

1 Total Cholesterol Show forest plot

4

439

Mean Difference (IV, Random, 95% CI)

0.00 [‐0.27, 0.28]

Figures and Tables -
Comparison 2. Total cholesterol
Comparison 3. HDL cholesterol

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

1 HDL Cholesterol Show forest plot

4

437

Mean Difference (IV, Fixed, 95% CI)

0.01 [‐0.06, 0.07]

Figures and Tables -
Comparison 3. HDL cholesterol
Comparison 4. LDL cholesterol

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

1 LDL Cholesterol Show forest plot

4

Mean Difference (IV, Random, 95% CI)

Totals not selected

Figures and Tables -
Comparison 4. LDL cholesterol
Comparison 5. Triglycerides

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

1 Triglycerides Show forest plot

4

439

Mean Difference (IV, Fixed, 95% CI)

0.01 [‐0.17, 0.19]

Figures and Tables -
Comparison 5. Triglycerides
Comparison 6. Systolic blood pressure

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

1 Systolic Blood Pressure Show forest plot

5

Mean Difference (IV, Fixed, 95% CI)

Totals not selected

Figures and Tables -
Comparison 6. Systolic blood pressure
Comparison 7. Diastolic blood pressure

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

1 Diastolic Blood Pressure Show forest plot

5

Mean Difference (IV, Fixed, 95% CI)

Totals not selected

Figures and Tables -
Comparison 7. Diastolic blood pressure