Patient education in the management of coronary heart disease

Summary of findings 1. Patient education for the management of coronary heart disease

Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No of Participants (studies)	Quality of the evidence (GRADE)	Comments
Patient education for the management of coronary heart disease
Patient or population: patients with coronary heart disease Settings: Centre or home‐based Intervention: Patient education
	Assumed risk	Corresponding risk
	Control	Patient education
Total mortality at the end of the follow‐up period No of deaths Follow‐up: median 18 months	Study population		RR 0.80 (0.60 to 1.05)	10075 (13 studies)	⊕⊕⊕⊝ Moderate¹
	46 per 1000	37 per 1000 (28 to 48)
	Moderate population
	43 per 1000	34 per 1000 (26 to 45)
Fatal and/or non‐fatal MI at the end of the follow up period Follow‐up: median 33 months	Study population		RR 0.63 (0.26 to 1.48)	209 (2 studies)	⊕⊝⊝⊝ very low² ³ ⁴
	118 per 1000	74 per 1000 (31 to 174)
	Moderate population
	106 per 1000	67 per 1000 (28 to 157)
Other fatal and/or non‐fatal cardiovascular events Follow‐up: median 21 months	Study population		RR 0.36 (0.23 to 0.56)	310 (2 studies)	⊕⊕⊝⊝ low² ⁴
	386 per 1000	139 per 1000 (89 to 216)
	Moderate population
	324 per 1000	117 per 1000 (75 to 181)
Total revascularisations (including CABG and PCI) Follow‐up: median 36 months	Study population		RR 0.58 (0.19 to 1.71)	456 (3 studies)	⊕⊝⊝⊝ verylow² ³ ⁴
	35 per 1000	20 per 1000 (7 to 60)
	Moderate population
	33 per 1000	19 per 1000 (6 to 56)
Hospitalisations (cardiac‐related)at end of follow up period Follow‐up: median 12 months	Study population		RR 0.93 (0.71 to 1.21)	14849 (5 studies)	⊕⊝⊝⊝ very low¹ ² ⁵
	79 per 1000	74 per 1000 (56 to 96)
	Moderate population
	141 per 1000	131 per 1000 (100 to 171)
All cause withdrawal at follow‐up Follow‐up: median 12 months	Study population		RR 1.04 (0.88 to 1.22)	10972 (17 studies)	⊕⊕⊝⊝ low² ⁶ ⁷
	92 per 1000	96 per 1000 (81 to 113)
	Moderate population
	70 per 1000	73 per 1000 (62 to 85)
HRQoL Various HRQoL measures Follow‐up: median 12 months	Not measurable	Not measurable	Not measurable	4393 (13 studies)	⊕⊕⊕⊝ moderate²	HRQoL in intervention > HRQoL in comparator, in then 9/99 domains

¹ 95% CIs include both no effect and appreciable benefit (i.e. CI < 0.75) ² Blinding of outcome assessors was poorly described in over 50% of included studies; bias likely ³ 95% CIs include both no effect, appreciate benefit and appreciable harm (i.e. CI < 0.75 and > 1.25) ⁴ The point estimate is likely to be imprecise due to very low event rates ⁵ I² > 40%; heterogeneity may be important ⁶ 95% CIs include both no effect and appreciate harm (i.e. CI > 1.25) ⁷ Evidence of funnel plot asymmetry therefore publication bias likely

Background

Description of the condition

Coronary heart disease (CHD) is the largest cause of death globally. In 2105, an estimated 8,8 million people died from CHD worldwide (WHO 2017). In the United Kingdom (UK), an estimated 2.3 million people live with CHD, and in 2014, the condition accounted for around 69,000 deaths (15% of male deaths and 10% of female deaths), and 3.4% of all inpatient episodes in men and 1.4% in women (BHF 2015). Most cardiovascular diseases can be prevented by addressing behavioural risk factors such as smoking, unhealthy diet and obesity, physical inactivity and harmful use of alcohol. Indeed, through early detection strategies, advanced medical treatment, lifestyle changes and risk factor reductions, UK age‐standardised CHD death rates declined by 73% for all ages, and 81% for those dying before the age of 75, between 1974 and 2013 (BHF 2015). Nonetheless, with falling CHD mortality rates, an increasing number of people live with CHD and may need support to manage their symptoms and improve prognosis.

Description of the intervention

Based on evidence from previous meta‐analyses and systematic reviews, exercise‐based cardiac rehabilitation following a cardiac event is a Class I recommendation from the American College of Cardiology/American Heart Association (Balady 2011; Kulik 2015) and the European Society of Cardiology (ESC 2012; ESC 2016; Smith 2011). Many definitions of cardiac rehabilitation have been proposed. The following definition encompasses the key concepts of cardiac rehabilitation: “The coordinated sum of activities required to influence favourably the underlying cause of cardiovascular disease, as well as to provide the best possible physical, mental and social conditions, so that the patients may, by their own efforts, preserve or resume optimal functioning in their community and through improved health behaviour, slow or reverse progression of disease” (BACPR 2012). Cardiac rehabilitation is a complex intervention that may involve a variety of therapies, including exercise, risk factor education, behaviour change, psychological support, and strategies that are aimed at targeting traditional risk factors for cardiovascular disease. Cardiac rehabilitation is an essential part of contemporary heart disease care and is considered a priority in countries with a high prevalence of CHD.

Patient education is defined as "the process by which health professionals and others impart information to patients that will alter their health behaviours or improve their health status" (Koongstvedt 2001). Self‐management education programmes are designed to allow people with chronic conditions to take an active part in managing their own condition (Foster 2007). They are complex behavioural interventions which target patient education and promote self‐care behaviour and risk‐factor modification and aim to improve health outcomes and decrease the incidence of complications for patients by supporting, not replacing, medical care (Walker 2003). This in turn may lead to reduced hospitalisations and medical appointments and an associated reduction in costs, both to the patient and the healthcare system (Ferri 2007). Educational interventions within cardiac care increase patients’ knowledge and facilitate behaviour change (Ghisi 2014). Educational interventions in cardiac care have been shown to increase physical activity, and lead to healthier dietary habits and smoking cessation, although any related improvement in response to cardiac symptoms, medication compliance or psychosocial well‐being is more equivocal (Ghisi 2014).

The delivery of patient education programmes can vary substantially, and may be classroom‐ or home‐based, group or individual, tailored or generic. Common topics include nutrition, exercise, risk factor modification, psychosocial well‐being, and medications (Ghisi 2014). Duration, frequency and ongoing maintenance or re‐inforcement also varies between programmes. Some programmes are developed according to validated educational theory and by trained professionals who are part of an interdisciplinary team, whilst others are delivered by peers. Some programmes may use adjunctive written materials or videotapes that supplement clinical consultations, while Internet‐ and mobile phone‐based applications may be used to deliver educational material and messages to patients (Neubeck 2009). Telephone follow‐up is increasingly used by healthcare providers to reach patients more frequently and in their own environment without the burden of a clinic visit (Phillips 2014).

Both the American College of Cardiology/American Heart Association and the European Society of Cardiology recognise education as an important component of comprehensive cardiac rehabilitation programmes and give a Class I recommendation that patients with non‐ST‐elevation acute coronary syndromes (ACS) and individuals with very high cardiovascular disease risk, should be educated about appropriate cholesterol management, blood pressure, smoking cessation, and lifestyle management (Amsterdam 2014; ESC 2016; Perk 2012). Exercise and psychological interventions are the subject of recent Cochrane systematic review updates (Anderson 2016; Richards 2017). Whilst these reviews have considered trials that have included education as a co‐intervention, this review update specifically focuses on the impact of the educational component of cardiac rehabilitation for patients with CHD.

Why it is important to do this review

Two meta‐analyses of education for people with CHD were published in the 1990s (Dusseldorp 1999; Mullen 1992). Mullen 1992 demonstrated a significant reduction in mortality associated with patient education (weighted average effect size 0.24 standard deviation units, 95% CI 0.14 to 0.33), which translated into a 19% improvement in mortality. The average effects for morbidity (re‐infarction and re‐hospitalisation) were not found to be significant. However, one randomised controlled trial (RCT) was excluded from analysis because it was an outlier as it demonstrated a large positive effect size (Rahe 1979). Dusseldorp 1999 investigated the co‐interventions of health education and stress management and concluded that these programmes yielded a mean reduction of 34% in cardiac mortality and a 29% reduction in re‐infarction. A Cochrane Review was subsequently published in 2011 which identified 13 RCTs randomising a total of 68,556 participants (Brown 2011). Brown 2011 incorporated new evidence and addressed concerns relating to the generalisability of the results of the two earlier meta‐analyses to the wider CHD population, and their applicability to policy formation, improved medical treatment of people with CHD, and the changing provision of cardiac rehabilitation services. Brown 2011 did not find evidence that education reduced total mortality, cardiac morbidity, revascularisation or hospitalisation compared to control, while there was some evidence to suggest that education may improve health‐related quality of life (HRQoL) and reduce overall healthcare costs. A more recent systematic review investigated the impact of education on patients’ knowledge and health behaviour change in people with CHD (Ghisi 2014), but to our knowledge, there have been no other recent meta‐analyses which have updated the evidence on the effect of education delivered as part of cardiac rehabilitation, on mortality, morbidity and HRQoL in this population.

The American College of Cardiology/American Heart Association and the European Society of Cardiology recognise education as an important component of comprehensive cardiac rehabilitation programmes and give a Class I recommendation that people with non‐ST‐elevation ACS and those with very high cardiovascular disease risk should be educated about appropriate cholesterol management, blood pressure, smoking cessation, and lifestyle management (Amsterdam 2014; ESC 2016; Perk 2012).

This update aimed to use additional RCT evidence published since the 2011 Cochrane Review to re‐assess the effectiveness of education compared with usual care on mortality, risk of hospital admission, myocardial infarction, revascularisation, HRQoL and healthcare costs in people with CHD.

Objectives

To assess the effects of patient education delivered as part of cardiac rehabilitation, compared with usual care on mortality, morbidity, health‐related quality of life (HRQoL) and healthcare costs in patients with CHD.
To explore the potential study level predictors of the effects of patient education in patients with CHD (e.g. individual versus group intervention, timing with respect to index cardiac event).

Methods

Criteria for considering studies for this review

Types of studies

To reflect contemporary coronary heart disease (CHD) practice we included randomised controlled trials (RCTs) published after 1990.

Types of participants

We included studies where participants were adults (aged ≥ 18 years):

who had experienced a myocardial infarction (MI);
who underwent revascularisation (coronary artery bypass grafting (CABG), percutaneous coronary intervention (PCI) or coronary artery stenting); or
who had angina pectoris or CHD defined by angiography.

We excluded studies of education programmes which included participants who:

had received heart valve surgery;
suffered from heart failure;
were heart transplantation recipients;
were implanted with cardiac‐resynchronisation therapy; or
were implanted with defibrillators.

Types of interventions

We identified RCTs where patient education was the primary intention of the cardiac rehabilitation intervention with a follow‐up period of at least six months. We excluded studies of cardiac rehabilitation where exercise or psychological intervention were the primary focus for investigation. These latter components of cardiac rehabilitation have been investigated in recently updated Cochrane Reviews of exercise‐based cardiac rehabilitation (Anderson 2016) and psychological cardiac rehabilitation interventions for people with CHD (Richards 2017).

For the purposes of this review, patient education was defined as the following:

instructional activities organised in a systematic way involving personal direct contact between a health professional and CHD patients with or without significant others: e.g. spouse, family member;
delivered as an inpatient, or outpatient in a community‐based intervention setting or programme;
included some form of structured knowledge transfer about CHD, its causes, treatments or methods of secondary prevention; and
delivered in a face‐to‐face format, in groups or on a one‐to‐one basis. We also included alternative interactive methods of educational delivery such as 'telehealth' (telephone, e‐mail, Internet and teleconference between educator and patient).

We included only study interventions that met all the above criteria.

We excluded general information provision, which is not organised in a systematic way (e.g. written guidance given to a patient on leaving the cardiac care unit or personal communication with a healthcare provider), which was considered to be usual care.

Given the multifaceted nature of cardiac rehabilitation we excluded studies where exercise and psychological therapies, or both, were provided and patient education was not stated to be a primary intervention.

We particularly sought studies designed to assess the independent effect of education (e.g. patient education plus usual care versus usual care alone; patient education, usual care and exercise versus usual care and exercise alone; patient education, usual care and psychological intervention versus usual care and psychological intervention alone).

Types of outcome measures

The aim of the review was to include studies that reported event data (e.g. mortality, cardiovascular events). We excluded studies that only measured alternative outcomes such as changes in smoking, diet, blood pressure or effect of education on patient knowledge. We elected not to include these alternative surrogate outcomes because we considered event rates to be more significant.

Primary outcomes

Total mortality.
Cardiovascular mortality.
Non‐cardiovascular mortality.
Fatal and/or non‐fatal myocardial infarction (MI).
Other fatal and/or non‐fatal cardiovascular events.

Secondary outcomes

Total revascularisations (including coronary artery bypass grafting (CABG) and percutaneous coronary intervention (PCI)).
Hospitalisations (total number of cardiac‐related patient admissions in the follow‐up period following the intervention).
Withdrawals.
Health‐related quality of life (HRQoL, using validated measures e.g. Short Form Health Survey SF‐36, Sickness Impact Profile, Nottingham Health Profile).
Adverse events.
Healthcare costs and cost‐effectiveness.

We excluded any study that did not measure one or more of these outcomes.

Search methods for identification of studies

Electronic searches

We searched the following databases on 30 June 2016:

CENTRAL Issue 6, 2016 (in the Cochrane Library);
MEDLINE (Ovid) 1946 to June week 3 2016;
Embase (Ovid) 1980 to 2016 week 26;
PsycINFO (Ovid) 1806 to June week 3 2016; and
CINAHL (EBSCO) 1937 to 30 June 2016.

The search strategies were designed with reference to those used previously (Brown 2011). We searched the databases using a strategy combining selected MeSH terms and free text terms relating to patient education and CHD, with filters applied to limit to RCTs. We used the Cochrane sensitivity‐maximising RCT filter for MEDLINE, and for Embase, terms recommended in the Cochrane Handbook were applied (Lefebvre 2011). Adaptations of this filter were applied to CINAHL and PsycINFO. We translated the MEDLINE search strategy for use in the other databases using the appropriate controlled vocabulary as applicable. We imposed no language or other limitations and gave consideration to variations in terms used and spellings of terms in different countries so that studies would not be missed by the search strategy because of such variations. See Appendix 1 for details of the search strategies used.

Ongoing trials were identified from searching the following trial registries in May 2016:

UK Clinical Trials Gateway (https://www.ukctg.nihr.ac.uk/)
ClinicalTrials.gov (https://clinicaltrials.gov)
ICTRP WHO International Clinical Trials Registry Platform (http://apps.who.int/trialsearch/)

Search results reporting was conducted in accordance with PRISMA (Moher 2009). A flow diagram is included, which provides information about the number of studies identified, included and excluded, and reasons for exclusions (Figure 1).

Figure 1

PRISMA flow diagram

Searching other resources

Reference lists of all eligible trials, systematic reviews and meta‐analyses were searched for additional studies. Attempts were made to contact all study authors to obtain relevant information not available in the published manuscript.

Data collection and analysis

Selection of studies

Titles and abstracts of studies identified by the search strategy were screened by two independent review authors (LA, RST) and obviously irrelevant studies were discarded. The full‐text reports of all potentially relevant abstracts were obtained (LA) and assessed independently for eligibility (LA, RST). Any disagreement was resolved by discussion. Excluded studies and reasons for exclusion are detailed in the Characteristics of excluded studies table.

Data extraction and management

One review author (LA) extracted study characteristics of included RCTs using a standardised data collection form which had been piloted on two RCTs included in the review. Data on patient characteristics (e.g. age, sex, CHD diagnosis) details of the intervention (including duration, frequency and delivery), description of usual care and length of follow‐up were extracted. A second author (HKR) checked all extracted data for accuracy. Two independent review authors (LA, HKR) extracted outcome data onto a standardised collection form. If data were presented numerically (in tables or text) and graphically (in figures), the numeric data were used because of possible measurement error when estimating from graphs. Any discrepancies were resolved by arbitration. One review author (LA) transferred extracted data into Review Manager 5.3 (RevMan 2014), and a second author (RST) checked data for accuracy against the systematic reviews.

If there were multiple reports of the same study, we assessed the duplicate publications for additional data. We extracted outcome results at all follow‐up points post‐randomisation. We contacted study authors where necessary to provide additional information.

Assessment of risk of bias in included studies

One review author (LA) assessed the risk of bias in included studies using Cochrane's recommended tool, which is a domain‐based critical evaluation of the following core risk of bias items: the quality of random sequence generation and allocation concealment, description of withdrawals, blinding of outcome assessment, and presence of selective reporting (Higgins 2011). We also assessed three further quality criteria: whether the study groups were balanced at baseline, if the study groups received comparable care (apart from the educational component of the intervention), and whether an intention‐to‐treat analysis was undertaken. The criteria used for assessing these last three risk of bias domains are as follows.

Groups balanced at baseline

Low risk of bias: the characteristics of the participants in the intervention and control groups at baseline is reported to be comparable or can be judged to be comparable in terms of likely main prognostic factors.
Uncertain risk of bias: it is not reported whether the participants' characteristics in the two groups are balanced at baseline and there is inadequate information reported to assess this.
High risk of bias: there is evidence of substantive imbalance in the baseline characteristics of the intervention and control groups with regard to likely major prognostic factors.

Intention‐to‐treat analysis

Low risk of bias: the trial reports that the analyses were conducted according to an intention‐to‐treat analysis, and includes all the principles of such an analysis, e.g. keeping participants in the intervention groups to which they were randomised, regardless of the intervention they actually received; and measures outcome data on all or the majority of participants (i.e. > 80% of those randomised) or includes imputation of all missing data in the analysis, using appropriate methodology, e.g. multiple imputation.
Uncertain risk of bias: it is unclear if the trial has performed an intention‐to‐treat analysis.
High risk of bias: the trial does not include an intention‐to‐treat analysis, or there is a substantive loss of outcome data (e.g. > 20%) and analyses are performed according to imputation methods known to create bias such as last observation carried forward.

Groups received comparable treatment (except exercise)

Low risk of bias: all co‐interventions were delivered equally across intervention and control groups.
Uncertain risk of bias: there was insufficient information to access whether co‐interventions were equally delivered across groups.
High risk of bias: the co‐interventions were not delivered equally across intervention and control groups.

All risk of bias assessments were checked by a second review author (HKR) and any discrepancies were resolved by arbitration. Details of the assessments of risk of bias for each included trial are shown in the Characteristics of included studies table.

Measures of treatment effect

For dichotomous variables, risk ratios (RR) and 95% confidence intervals (CI) were derived for each outcome. If any continuous variables had been reported, mean differences and 95% CI would have been calculated for each outcome.

Unit of analysis issues

In accordance with Section 9.3.1 of the Cochrane Handbook for Systematic Reviews of Intervention (Higgins 2011), we ensured that the analysis was appropriate to the level at which randomisation occurred. All studies included in this review were simple parallel group RCTs, and so there were no issues relating to unit of analysis.

Dealing with missing data

We contacted investigators or study sponsors to verify key study characteristics and obtain missing numerical outcome data where possible (for example when a study is identified as an abstract only). Had this not been possible, and the missing data were thought to introduce serious bias, we planned to explore the impact of including such studies on the overall assessment of results by a sensitivity analysis.

Assessment of heterogeneity

We explored heterogeneity amongst included studies qualitatively (by comparing the characteristics of included studies) and quantitatively (using the Chi² test of heterogeneity and I² statistic). We used a threshold of I² greater than 50% for both dichotomous and continuous outcomes to determine the statistical model to be used for meta‐analysis.

Assessment of reporting biases

The funnel plot and the Egger test were used to examine small study bias (Egger 1997).

Data synthesis

We processed data in accordance with Cochrane Handbook for Systematic Reviews of Interventions guidance (Deeks 2011). Where appropriate and possible, results from included studies were combined for each outcome to give an overall estimate of treatment effect. Given the degree of clinical heterogeneity seen in participant selection, interventions and comparators across studies, we decided it was appropriate to pool studies using random‐effects modelling.

With the exception of total mortality, the review did not identify sufficient data to allow stratified meta‐analysis at different common follow‐up timings (e.g. 6 or 12 months post‐randomisation). Instead, we pooled studies at their longest follow‐up unless otherwise stated.

Summary of findings table

Two independent review authors (LA, RST) used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach to interpret result findings and used GRADEpro GDT 2014 to import data from Review Manager to create a 'Summary of findings table'. We created a 'Summary of findings' table using the following outcomes: total mortality, fatal and/or non‐fatal MI, total revascularisations, other fatal and/or non‐fatal cardiovascular events, hospitalisations, withdrawals and HRQoL. We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the quality of a body of evidence as it relates to the studies that contribute data to the meta‐analyses for the prespecified outcomes. We used methods and recommendations described in Section 8.5 and Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions using GRADEpro GDT 2014 software (Higgins 2011). We justified all decisions to downgrade the quality of studies using footnotes, and made comments to aid readers' understanding of the review where necessary.

Subgroup analysis and investigation of heterogeneity

As stated in the protocol, we planned to undertake subgroup analysis and stratified meta‐analysis, sensitivity analysis and meta‐regression to examine potential treatment effect modifiers. We intended to test the following a priori hypotheses that there may be differences in the effect of education on total mortality and withdrawal across particular subgroups:

CHD case mix (MI‐only trials versus other trials).
Dose and nature of structured patient education. Assessed on the basis of the number and nature of education sessions e.g. extent of training of who delivers the education, a healthcare professional, or specific educational training, whether feedback or re‐inforcement were given (i.e. literature, audiovisual follow‐up material).
Method of structured educational delivery (one‐to‐one versus group versus combination).
Theoretical versus no theoretical basis to educational intervention.
Involvement of significant others (e.g. spouse, family member) in the education.
Timing of the education following the index event.
Length of the educational intervention.
Follow‐up period (≤ 12 months versus > 12 months).

For this update we also tested the following predictors of total mortality and withdrawal using univariate meta‐regression:

mean age of participants;
percentage of male participants;
type of cardiac rehabilitation (education only versus education plus e.g. exercise or psychological intervention);
study location (continent); and
setting (centre versus home).

Due to poor reporting, we were unable to examine the association of dose of education or timing following the index event, with the risk of total mortality or withdrawal.

Sensitivity analysis

We undertook a sensitivity analysis to examine the effect of risk of bias (low risk in ≥ five items versus < five items) and year of publication (before 2000 versus 2000 or later) of included studies on total mortality and withdrawal.

We decided it was appropriate to pool studies using random‐effects modelling, due to the clinical heterogeneity seen in participant selection, interventions and comparators across studies. However, we undertook a sensitivity analysis to examine the effect on the pooled data of conducting a fixed‐effect or a random‐effects model. The results of the random‐effects model are reported as default in the text, while the results from both models for all outcomes are reported in Table 1.

Table 1. Results of sensitivity analysis for fixed‐effect versus random‐effects models

Outcome or subgroup	Studies	Participants	Effect estimate (random‐effect) RR (M‐H, random, 95% CI)	Effect estimate (fixed‐effect) RR (M‐H, fixed, 95% CI)
1.1 Total mortality at the end of the follow up period	13	10,075	0.80 [0.60, 1.05]	0.80 [0.66, 0.97]
1.1.1 Studies with 12 months or less follow‐up	6	4063	0.78 [0.35, 1.78]	0.87 [0.56, 1.36]
1.1.2 Studies with more than 12 months follow‐up	7	6012	0.78 [0.60, 1.02]	0.79 [0.64, 0.97]
2.1 Myocardial Infarction at the end of the follow‐up period	2	209	0.63 [0.26, 1.48]	0.59 [0.25, 1.38]
2.2 Total revascularisations	3	456	0.58 [0.19, 1.71]	0.58 [0.20, 1.69]
2.3 Other fatal and/or non‐fatal cardiovascular events	2	310	0.36 [0.23, 0.56]	0.36 [0.23, 0.56]
3.1 Cardiac hospitalisations at end of follow‐up period	5	14,849	0.93 [0.71, 1.21]	1.02 [0.90, 1.15]
4.1 All cause withdrawal or drop‐out at follow‐up	17	10,972	1.04 [0.88, 1.22]	0.98 [0.88, 1.10]
4.1.1 Studies with 12 months or less follow‐up	10	4960	1.18 [0.93, 1.49]	1.18 [0.94, 1.49]
4.1.2 Studies with more than 12 months follow‐up	7	6012	0.98 [0.80, 1.20]	0.92 [0.81, 1.05]

Results

Description of studies

Results of the search

We identified 3918 records through our electronic database search. After de‐duplication, 2749 titles and abstracts were screened for inclusion, of which 2639 were excluded. We identified two additional records from searching the citations of publications identified as meeting our inclusion criteria, and a further 10 studies listed on trial registries. We retrieved and assessed 122 full text reports for eligibility and subsequently excluded 84 publications. Sixteen studies were ongoing and five were determined as awaiting classification because the full text was not available and authors did not respond to repeated requests for study information. In total, we included an additional nine randomised controlled trials (RCTs) (17 publications) to the 13 RCTs (24 publications) from the 2011 review, totaling 22 RCTs (41 publications). Details of the exclusion process and reasons for exclusion are summarised in a PRISMA flow diagram (Figure 1) and in the Characteristics of excluded studies table.

Included studies

The previous version of this review (Brown 2011) included 13 RCTs (24 publications) (Clark 1997; Clark 2000; Clark 2009; Cupples 1994; Esposito 2008; Hanssen 2007; Lie 2009; Lisspers 1999; P.RE.COR Group 1991; Peikes 2009; Pogosova 2008; Southard 2003; Tingström 2005). We included an additional nine RCTs for this update (16 publications, 8215 participants) (Chow 2015; Cohen 2014; Dracup 2009; Furuya 2015; Jorstad 2013; Melamed 2014; Mooney 2014; Moreno‐Palanco 2011; Park 2013). We included a total of 22 studies reporting data for a total of 76,864 participants in this update.

Details of included studies are listed in the Characteristics of included studies table. Seventeen studies compared education‐only interventions with a comparator and five studies reported on an education plus counselling or behaviour change intervention (Dracup 2009; Hanssen 2007; Lie 2009; Lisspers 1999; Peikes 2009). No studies included interventions which comprised exercise as a co‐intervention.

Eleven studies were undertaken in Europe (Cohen 2014; Cupples 1994; Hanssen 2007; Jorstad 2013; Lie 2009; Lisspers 1999; Melamed 2014; Mooney 2014; Moreno‐Palanco 2011; P.RE.COR Group 1991; Tingström 2005); six were performed in the USA (Clark 1997; Clark 2000; Clark 2009; Esposito 2008; Peikes 2009; Southard 2003); and one each in Russia (Pogosova 2008), Australia (Chow 2015), South America (Furuya 2015) and Asia (Park 2013) and one was undertaken in sites in USA, Australia and New Zealand (Dracup 2009). Fourteen studies were multicentre RCTs (Clark 1997; Clark 2000; Clark 2009; Cohen 2014; Cupples 1994; Dracup 2009; Esposito 2008; Jorstad 2013; Melamed 2014; Mooney 2014; P.RE.COR Group 1991; Peikes 2009; Southard 2003; Tingström 2005); and eight were single centre RCTs (Chow 2015; Furuya 2015; Hanssen 2007; Lie 2009; Lisspers 1999; Moreno‐Palanco 2011; Park 2013; Pogosova 2008).

Sixteen studies reported sources of funding; six did not report funding sources (Clark 1997; Esposito 2008; Lie 2009; Melamed 2014; Moreno‐Palanco 2011; Pogosova 2008). One study was funded by an industrial sponsor (Jorstad 2013), four by health insurance companies (Chow 2015; Cohen 2014; Lisspers 1999; Peikes 2009) and 11 by government or public sources (Clark 2000; Clark 2009; Cupples 1994; Dracup 2009; Furuya 2015; Hanssen 2007; Mooney 2014; P.RE.COR Group 1991; Park 2013; Southard 2003; Tingström 2005).

Most trials were relatively small in sample size (median 454 participants, range: 63 to 46,606). Two large trials (Esposito 2008; Peikes 2009) contributed 85% (65,008 participants) of all included participants.

The median duration of trial intervention was six months (range 1 to 36 months) with median follow‐up of 12 months (range 6 to 60 months).

The case mix of participants recruited to the included trials varied considerably; six studies recruited mixed populations of people with CHD (Chow 2015; Clark 1997; Clark 2000; Clark 2009; Melamed 2014; Pogosova 2008); four studies recruited participants with myocardial infarction (MI) or angina (Cohen 2014; Jorstad 2013; Mooney 2014; Park 2013); and the remaining studies recruited participants post‐revascularisation (Furuya 2015; Lie 2009; Lisspers 1999); with coronary heart disease (CHD) or heart failure (Esposito 2008; Peikes 2009; Southard 2003); MI (Hanssen 2007; P.RE.COR Group 1991); acute coronary syndromes (ACS) (Dracup 2009; Moreno‐Palanco 2011); angina (Cupples 1994); or MI or post‐revascularisation (Tingström 2005).

The mean age of trial participants ranged from 51.0 to 72.8 years. Although all but two trials included women (20 studies, 91%), only 25% of participants recruited were women.

The two largest studies (65,008 participants) (Esposito 2008; Peikes 2009) included some participants who were outside the scope of this review (i.e. trialists considered people with congestive cardiac failure and diabetes). However, participants with CHD contributed 69% and 61% respectively, to these studies. Both studies reported hospitalisation, health‐related quality of life (HRQoL) and cost data. Only hospitalisation data from Esposito 2008 contributed to the meta‐analysis, and these data were reported separately for participants with CHD (Esposito 2008). Southard 2003 included participants with cardiac failure as well as those with CHD.

Four studies involved group sessions (Clark 1997; Clark 2000; Pogosova 2008; Tingström 2005); 12 involved individual education sessions (Chow 2015; Cohen 2014; Cupples 1994; Dracup 2009; Esposito 2008; Hanssen 2007; Lie 2009; Melamed 2014; Moreno‐Palanco 2011; Mooney 2014; Park 2013; Peikes 2009); three used both group and individual sessions (Lisspers 1999; P.RE.COR Group 1991; Southard 2003); one study compared group and individual approaches (Clark 2009); and one study did not report on the mode of teaching (Jorstad 2013). Eighteen studies involved face‐to‐face sessions (Cohen 2014; Dracup 2009; Clark 1997; Clark 2000; Clark 2009; Cupples 1994; Esposito 2008; Furuya 2015; Jorstad 2013; Lie 2009; Lisspers 1999; Melamed 2014; Moreno‐Palanco 2011; Mooney 2014; P.RE.COR Group 1991; Park 2013; Pogosova 2008; Tingström 2005); three were reliant on telephone contact (Esposito 2008; Hanssen 2007; Peikes 2009); three used face‐to‐face sessions as well as telephone follow‐up (Furuya 2015; Mooney 2014; Park 2013); one involved interactive use of the Internet (Southard 2003); and one used text messages via a mobile phone (Chow 2015). The educational intervention was delivered by a wide variety of personnel, with nine interventions delivered by nurses (Cohen 2014; Dracup 2009; Esposito 2008; Furuya 2015; Hanssen 2007; Jorstad 2013; Lie 2009; Mooney 2014; Moreno‐Palanco 2011); four by trained educators (Clark 1997; Clark 2000; Clark 2009; Tingström 2005); three by physicians (Melamed 2014; P.RE.COR Group 1991; Pogosova 2008), and one each by a care co0ordinator (Peikes 2009), case manager (Southard 2003), and a researcher (Furuya 2015). The person delivering the intervention was not described in one study (Park 2013). The intensity of the education programme varied substantially from just one 40 minute face‐to‐face session plus a 15 minute follow‐up call (Dracup 2009) to a four‐week residential stay reinforced with 11 months of nurse‐led follow‐up sessions (Lisspers 1999). Description of the educational content of the programmes was mostly brief and lacked detail. Table 2 summarises educational interventions.

Table 2. Educational content of programs in included studies

Study ID	Description of Intervention	Theoretical basis	Tailored	Duration	One‐to‐one	Group	Face‐ to‐face	Telephone	Internet	Notes
Chow 2015	Text message‐based prevention program delivering regular semi‐personalised messages providing advice, motivation, and information to improve diet, increase physical activity, and encourage smoking cessation	NR	Y	4 messages per week for 24 weeks	Y	N	N	Y (text messages)	N	Content for each participant was selected using a prespecified algorithm dependent on key baseline characteristics
Clark 1997	*PRIDE	Y	Y	Once weekly for 4 weeks		Y	Y			Taught by health educator. Videotape and workbook aids
Clark 2000	*PRIDE	Y	Y	Once weekly for 4 weeks		Y	Y			Taught by health educator. Videotape and workbook aids
Clark 2009	*PRIDE	Y	Y	Once weekly for 6 weeks	Y	Y	Y			3 groups (self‐directed and group intervention and a control)
Cohen 2014	"House of Education" with individualised consultations with e.g. smoking cessation nurse	NR	Y	At least 6 sessions in 12 months	Y	N	Y	N	N	Consultations content was individualised according to a patient’s risk factors
Cupples 1994	Practical tailored advice on cardiovascular risk factors and appropriate health education	NR	Y	3 times a year for 2 years	Y		Y			Delivered at home by health visitor
Dracup 2009	Patients received education on ACS, anticipated emotional issues and social factors that could affect delay	Y	Y	55 mins (40 min face‐to‐face plus 15 min follow‐up call)	Y	N	Y	N	N	Delivered by a nurse with expertise in cardiology
Esposito 2008	Predesigned scripts to provide education on various aspects of care, geared to personalised clinical goals	NR	Y	Average 1.1 contacts per month for 18 months	Y		Y	Y		Nurse case manager, primarily by telephone but also face‐to‐face
Furuya 2015	Three booklets and three telephone follow‐up calls aimed at helping patient understand his cardiac condition, PCI and how to cope with CAD	Y	N	2 face to face sessions and 3 telephone calls over 16 weeks	Y	N	Y	Y	N	The first booklet was discussed with participants before undergoing PCI procedure
Hanssen 2007	Individualised education from a menu of topics to be covered	Y	Y	6 months (8 sessions in total)	Y			Y		Structured element and an on‐call element
Jorstad 2013	Outpatient clinic visits to a cardiovascular nurse	NR	Y	6 months (4 sessions)	NR	NR	Y	N	N	Nurse‐coordinated: provided general lifestyle advice, and individual counselling
Lie 2009	A psycho‐educative intervention. Structured information and psychological support	NR	N/S	2 visits (1 hour each)	Y		Y			Critical care nurse, home based
Lisspers 1999	Health education and achievement of behavioural change	NR	Y	4 week residential then 11 month one‐to‐one individual sessions	Y	Y	Y			Trained nurses (personal coaches). Seminars, lectures, discussion and skills sessions
Melamed 2014	Lesson materials consisted of a patient brochure, teaching cards and curriculum poster/wall chart set	NR	N	NR	Y	N	Y	N	N	Patients were given an exercise diary to enable them to document their daily physical activity
Mooney 2014	Education intervention aimed at reducing total prehospital delay time	Y	Y	6 months (1 face‐to‐face session, 1 telephone call and one reinforcement letter at 6 months)	Y	N	Y	Y	N	Research nurses used preprinted flip charts and prescriptive scripts as educational aids
Moreno‐Palanco 2011	Health education on the meaning of patients' disease and the importance of treatment	NR	NR	3 years (at least 5 sessions)	Y	N	Y	N	N	Each visit consisted of a nursing intervention and a medical assessment
P.RE.COR Group 1991	Education and counselling on management of cardiovascular risk factors and exercise	NR	Y	1 group session, 1 individual session with cardiologist	Y	Y	Y			Multidisciplinary input to group. Cardiologist tailors therapy
Park 2013	Psycho‐educational intervention comprising tailored face‐to‐face education and telephone‐delivered health coaching	NR	Y	12 weeks (6 sessions)	Y	N	Y	Y	N	Patients made choices about risk factors they wanted to lower and participated in goal setting
Peikes 2009	Variable ‐ nurse provision of patient education	NR	NR	1 to 2.5 times a month for an average of 30 months	Y			Y		15 different programs, majority telephone, one‐to‐one
Pogosova 2008	Structured program addressing different risk factors in each session	Y	NR	6 sessions (twice weekly, 90 min)		Y	Y
Southard 2003	Modular internet sessions, Interactive multiple choice and self tests followed by feedback	NR	NR	Once weekly for 6 months (at least 30 min)	Y	Y			Y	Communication with case manager and online discussion group
Tingström 2005	Problem based rehabilitation to teach a planned curriculum	Y	NR	13 sessions over 1 year		Y	Y			Trained facilitator

PRIDE = Problem Identification, Researching one's routine, Identifying a management goal, Developing a plan to reach it, Expressing one's reactions and Establishing rewards for making progress.

Y = Yes; N = No; NR = not reported

Excluded studies

We excluded 84 full text publications because they did not meet the inclusion criteria for this review (see Characteristics of excluded studies). Seven studies were not RCTs, 13 studies had follow‐up periods of less than six months, three studies included populations who were irrelevant for this Review, 36 studies investigated interventions that were not relevant to this Review, six included inappropriate comparators and 19 studies did not report outcomes that were relevant to this Review.

Ongoing studies

The details of 16 ongoing studies that appear to meet the inclusion criteria for this Review are presented in Characteristics of ongoing studies (ACTRN12613000395730; ACTRN12613000793718; ACTRN12616000426482; Brewer 2015; Dwinger 2013; IRCT201307162621N13; ISRCTN15839687; Kärner 2012; Lai 2016; Lynggaard 2014; NCT01028066; NCT01275716; NCT01925079; NCT02185391; NTR2388; Shah 2011).

Studies awaiting classification

The details of five studies that are awaiting classification are presented in Characteristics of studies awaiting classification.. One study was published in the IIOAB Journal, but we were not able to access any of this journal's web pages (Gao 2011). We were unable to find the full text or trace the authors of the remaining four studies ( Licina 2010; Soliman 2013; Vona 2009; Xiaolin 2012).

Risk of bias in included studies

Several studies did not report sufficient methodological detail to enable assessment of potential risk of bias. Details of random sequence generation, concealment of random allocation and blinding were the most frequent poorly reported parameters. Risk of bias results are summarised in Figure 2 and Figure 3.

Figure 2

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

Figure 3

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

Allocation

Sixteen studies were judged to provide evidence of adequate random sequence generation (Chow 2015; Cohen 2014; Clark 1997; Clark 2000; Clark 2009; Cupples 1994; Furuya 2015; Hanssen 2007;Jorstad 2013; Lie 2009; Mooney 2014; Moreno‐Palanco 2011; Park 2013; Peikes 2009; Southard 2003; Tingström 2005). Fourteen studies reported adequate concealment methods (Chow 2015; Clark 1997; Clark 2009; Cohen 2014; Cupples 1994; Furuya 2015; Hanssen 2007; Jorstad 2013; Lie 2009; Melamed 2014; Mooney 2014; Moreno‐Palanco 2011; Peikes 2009; Tingström 2005).

Blinding

Due to the nature of the educational intervention, it was not possible to blind education providers or trial participants. We investigated evidence to ascertain if those collecting, assessing or analysing outcome data were blinded to group allocation. Blinding of this nature was confirmed in 10 studies (Chow 2015; Cohen 2014; Clark 1997; Clark 2000; Clark 2009; Cupples 1994; Dracup 2009; Furuya 2015; Jorstad 2013; Moreno‐Palanco 2011).

Incomplete outcome data

Sixteen studies clearly stated withdrawal or numbers lost to follow‐up (Table 3). Overall, 11.3% of participants in intervention groups and 11.5% of control group participants were lost to follow‐up. Most authors failed to assess participants who were lost to follow‐up for systematic differences when compared to those who completed the study.

See: Summary of findings 1 Patient education for the management of coronary heart disease

Table 3. All‐cause withdrawal or drop‐out at follow‐up

Study ID		Number randomised	Number lost at follow‐up*	Notes
Chow 2015	Intervention	352	33	20 excluded from analysis, 9 unable to contact, 4 died
Chow 2015	Control	358	25	21 excluded from analysis, 3 unable to contact, 1 died
Clark 2000	Intervention	309	51	36 withdrew, 14 died, 1 data missing
Clark 2000	Control	262	42	33 withdrew, 8 died, 1 data missing
Clark 2009	Intervention	201	37	Self‐directed program; 33 withdrew, 4 died
	Intervention	190	24	Group format; 19 withdrew, 5 died
	Control	184	23	15 withdrew, 8 died
Cohen 2014	Intervention	251	48	6 did not meet inclusion criteria, 7 died, 23 follow‐up refusal, 10 lost to follow‐up, 2 in another protocol
Cohen 2014	Control	251	36	4 did not meet inclusion criteria, 7 died, 13 follow‐up refusal, 12 lost to follow‐up
Cupples 1994	Intervention	342	92	45 defaulted, 47 died; 21 defaulted at 2 years
Cupples 1994	Control	346	109	44 defaulted, 65 died; 25 defaulted at 2 years
Dracup 2009	Intervention	1777	197	89 lost to follow‐up, 41 withdrawn, 67 died
Dracup 2009	Control	1745	238	94 lost to follow‐up, 69 withdrawn, 75 died
Furuya 2015	Intervention	34	4	4 unable to contact by telephone at follow‐up (90 participants were originally randomised (45 in each group), but 24 participants were excluded immediately after randomisation as they were indicated for surgery or enrolled in another study)
Furuya 2015	Control	32	2	2 did not return for 6 month follow‐up
Hanssen 2007	Intervention	156	55	40 withdrew, 7 died, 8 missing data
Hanssen 2007	Control	132	38	21 withdrew, 7 died, 10 missing data
Jorstad 2013	Intervention	375	23	9 did not receive intervention, 3 died, 2 had early discontinuation of intervention, 9 had incomplete data
Jorstad 2013	Control	379	35	12 were excluded from the study, 10 died, 1 lost to follow‐up, 7 didn't attend 12 month follow‐up, 5 had incomplete data
Lie 2009	Intervention	101	8	6 withdrew, 2 medical exclusions
Lie 2009	Control	102	10	5 withdrew, 5 medical exclusions
Melamed 2014	Intervention	202	21	"patients were exclude (for example, because of missed training appointments)"
Melamed 2014	Control	205	19	"patients were excluded (for example, because of missed training appointments)"
Mooney 2014	Intervention	972	35	14 withdrew, 21 died
Mooney 2014	Control	972	27	10 withdrew, 17 died
Moreno‐Palanco 2011	Intervention	121	3	3 lost to follow‐up, 0 died
Moreno‐Palanco 2011	Control	126	5	5 lost to follow‐up, 0 died
P.RE.COR Group 1991	Intervention	60	0	Counseling program without exercise
	Intervention	61	0	Comprehensive cardiac rehabilitation
	Control	61	0	Usual care
Park 2013	Intervention	31	3	3 withdrew, 0 died
Park 2013	Control	32	2	2 withdrew, 0 died
Southard 2003	Intervention	53	4	Reasons for drop‐out stated: relocation, dietary intervention instead, psychiatric diagnosis, loss of interest
Southard 2003	Control	51	0
Tingström 2005	Intervention	104	3	7 lost to follow‐up: 2 died, 5 did not attend
Tingström 2005	Control	103	4	7 lost to follow‐up: 2 died, 5 did not attend
Combined results	Intervention	5692	641	11.3%
Combined results	Control	5341	615	11.5%

* All causes of drop out from follow0up included (including mortality)

Selective reporting

We compared the reported outcomes in the results sections to the outcomes described in the methods of published papers. No attempt was made to identify original study protocols and compare these to reported outcomes. Only one study demonstrated selective reporting by not reporting the results of a HRQoL measure (Southard 2003).

Other potential sources of bias

Baseline balance

Eighteen studies had a good balance of subject baseline characteristics between intervention and control groups. Four studies demonstrated a statistically significant imbalance between groups at baseline (Clark 2000; Dracup 2009; Mooney 2014; Peikes 2009). There were differences in baseline disease symptoms and weight in Clark 2000. Peikes 2009 highlighted 11 differences in 255 baseline characteristics compared between groups, which they qualified as "less than the expected number of statistical significant differences than would be observed by chance (Peikes 2009). In Dracup 2009, there were differences in baseline body mass index, gender (with more females in the experimental group than control, P = 0.02), and insurance for ambulance use. In Mooney 2014, there were some significant differences between characteristics and prognostic factors of the two groups at baseline including age.

Intention‐to‐treat analysis

Fifteen studies analysed results on an intention‐to‐treat basis (Chow 2015; Cohen 2014; Clark 2000; Clark 2009; Cupples 1994; Esposito 2008; Hanssen 2007; Lie 2009; Lisspers 1999; Mooney 2014; Moreno‐Palanco 2011; P.RE.COR Group 1991; Peikes 2009; Southard 2003; Tingström 2005). In most cases, this involved analysing those participants remaining at follow‐up according to initial randomisation. Clark 1997 did not present intention‐to‐treat data, but presented data for participants who had attended at least one of the four intervention sessions.

Comparative care

We specifically sought to investigate the impact of education. However, in addition to education (the primary intervention), participants appeared to receive other co‐interventions such as exercise or psychological therapy in a number of studies. It was unclear how much of these co‐interventions were received by control group participants, posing potential for performance bias (Dracup 2009; Esposito 2008; Hanssen 2007; Lisspers 1999; Park 2013; Peikes 2009; Southard 2003).

Effects of interventions

Primary outcomes

Total mortality

Thirteen studies (10,075 participants) reported total mortality. Two studies reported deaths at six months (Chow 2015; Furuya 2015); five at 12 months (Clark 2000; Cohen 2014; Dracup 2009; Jorstad 2013; Mooney 2014); two at 18 months (Clark 2009; Hanssen 2007); four at 24 months (Clark 2000; Cupples 1994; Lisspers 1999; P.RE.COR Group 1991); one at 36 months (Moreno‐Palanco 2011) and two at 60 months (Cupples 1994; Lisspers 1999). Only one study demonstrated a significant difference in total mortality between education and control, where the cumulative survival rate at three years was 97.4% in the intervention group and 85.5% in the control group (P = 0.003) (Moreno‐Palanco 2011). At the longest reported follow‐up, there was no difference in effect of education‐based interventions on total mortality (random effects RR 0.80, 95% CI 0.60 to 1.05; participants = 10,075; studies = 13; Analysis 1.1).

Quality of the evidence for this outcome was judged to be moderate (summary of findings Table 1).

When data were stratified by length of follow‐up, there was similar uncertainty of the effect of education‐based interventions on total mortality for those studies with a mean length of follow‐up of more than 12 months (random effects RR 0.78, 95% CI 0.60 to 1.02; participants = 6012; studies = 7). There was no evidence of a reduction in total mortality in studies with a mean length of follow‐up of 12 or fewer months (random effects RR 0.78, 95% CI 0.35 to 1.78; participants = 4063; studies = 6).

Cardiovascular mortality

Individual causes of mortality were poorly or not reported across studies. We were therefore unable to report separate data for cardiovascular mortality.

Non‐cardiovascular mortality

Individual causes of mortality were not reported across studies. We were therefore unable to report separate data for non‐cardiovascular mortality.

Fatal and/or non‐fatal myocardial infarction (MI)

Two studies reported fatal and/or non‐fatal MI (Lisspers 1999; P.RE.COR Group 1991). There was no evidence of a reduction in morbidity with education‐based interventions for fatal and/or non‐fatal MI (random effects RR 0.63, 95% CI 0.26 to 1.48; participants = 209; studies = 2; Analysis 1.2).

Quality of the evidence for this outcome was judged to be very low (summary of findings Table 1).

Other fatal and/or non‐fatal cardiovascular events

Two studies reported other fatal or non‐fatal cardiovascular events (Moreno‐Palanco 2011; Park 2013, 310 participants). There was some evidence of a reduction in other fatal or non‐fatal cardiovascular events with education‐based interventions (random effects RR 0.36, 95% CI 0.23 to 0.56; participants = 310; studies = 2; Analysis 1.3).

Quality of the evidence for this outcome was judged to be low (summary of findings Table 1).

Southard 2003 reported a difference in "major cardiovascular‐related events"; fewer events occurred in the intervention group (P = 0.053). These were defined as events needing hospitalisation either as an inpatient or from the emergency department. As other cardiovascular events may have occurred that did not meet this definition, it was not appropriate to include these data in the pooled analysis.

Secondary outcomes

Total revascularisations (coronary artery bypass graft (CABG) or percutaneous coronary intervention (PCI))

Three studies (456 participants), reported subsequent revascularisation (CABG or PCI) (Lisspers 1999; Moreno‐Palanco 2011; P.RE.COR Group 1991). There was no evidence of a reduction in morbidity with education for total revascularisations (random effects RR 0.58, 95% CI 0.19 to 1.71; participants = 456; studies = 3; Analysis 1.4).

Quality of the evidence for this outcome was judged to be very low (summary of findings Table 1).

Hospitalisations

Seven studies reported cardiac‐related hospitalisations (Clark 2000; Esposito 2008; Hanssen 2007; Lisspers 1999; Mooney 2014; Peikes 2009; Southard 2003).

We pooled the results of the five studies (14,849 participants) that reported numbers of participants who were hospitalised for cardiac‐related events (Esposito 2008; Hanssen 2007; Lisspers 1999; Mooney 2014; Southard 2003). There was no evidence of a reduction in hospitalisation with education‐based interventions (random effects RR 0.93, 95% CI 0.71 to 1.21; participants = 14,849; studies = 5; Analysis 1.5.

Quality of the evidence for this outcome was judged to be very low (summary of findings Table 1).

Due to the method of reporting hospitalisations in four studies, it was not possible to include these in the pooled analysis (Clark 2000; Dracup 2009; Jorstad 2013; Peikes 2009).

Using intention‐to‐treat analysis Clark 2000 found no statistically significant difference in the total number of hospitalisations between intervention and control. Analysis of the "heart‐related admissions" in those participants who attended at least one intervention session revealed statistically significant reductions in the intervention group: participants in the intervention group had 41% fewer "heart‐related admissions" (P = 0.05) and 61% fewer "heart‐related" inpatient days (P = 0.02) than in the control group (Clark 2000).

Dracup 2009 reported the number of participants who presented to the emergency department with symptoms of ACS (565 participants (16.0%) and a total of 842 admissions). Of the 565 participants, 305 (54%) were in the intervention group and 260 (46%) were in the control group. Of the 842 emergency department admissions, 408 (48%) were in the control group and 434 (52%) were in the intervention group.

Jorstad 2013 reported the cumulative number of re‐admissions in 12 months. In total, there were 86 rehospitalisations in the intervention group and 132 in the control group (P = 0.023) (Jorstad 2013). This difference was driven by a 67% reduction in re‐admissions for non‐ACS chest pain (12 admissions versus 36 admissions, P < 0.001); re‐admissions for ACS and elective interventions were comparable in both groups.

Peikes 2009 reported the rate of hospitalisations across 15 different USA study sites. Overall, there was no clear evidence of effect of intervention, with only two of 15 sites showing a significant difference in hospital admissions. One reported an increase in admissions in the intervention group and the other reported an increase in the control group. No between‐group statistical difference was found in average annualised admission rates 0.91 (intervention) versus 0.95 (control) (P = 0.145).

Withdrawals

Studies varied in their reporting of participants who withdrew or dropped out of the study, the analysis or both. Most studies failed to report the number of participants who withdrew because they were unable to complete the intervention. Therefore, we reported withdrawals at follow‐up (Table 3). There was no evidence of a difference in the number of withdrawals from the education‐based intervention or control groups (random effects RR 1.04, 95% CI 0.88 to 1.22; participants = 10,972; studies = 17; Analysis 1.6).

Results remained equivocal when data were stratified by length of follow‐up (mean follow‐up ≤ 12 months: random effects RR 1.18, 95% CI 0.93 to 1.49; participants = 4960; studies = 10; mean follow‐up > 12 months: random effects RR 0.98, 95% CI 0.80 to 1.20; participants = 6012; studies = 7).

Quality of the evidence for this outcome was judged to be low (summary of findings Table 1).

Clark 1997 reported a combined withdrawal of 181 participants from both groups. A differential breakdown was not given, but there was "no appreciable differences in withdrawal rates between the intervention and control group" demonstrated (Clark 1997).

Numbers lost to follow‐up were unclear in four studies (Esposito 2008; Lisspers 1999; Peikes 2009; Pogosova 2008).

Health‐related quality of life (HRQoL)

Fifteen studies reported HRQoL (Cohen 2014; Clark 1997; Clark 2000; Clark 2009; Cupples 1994; Esposito 2008; Furuya 2015; Hanssen 2007; Lie 2009; Lisspers 1999; Melamed 2014; Park 2013; Pogosova 2008; Southard 2003; Tingström 2005). These studies used several generic HRQoL instruments, i.e. SF‐12 (Cohen 2014; Furuya 2015), SF‐36 (Furuya 2015; Hanssen 2007; Lie 2009; Pogosova 2008; Tingström 2005), Nottingham Health Profile (Cupples 1994), Sickness Impact Profile (Clark 1997; Clark 2000), a five‐point patient assessment scale of quality of life (Cupples 1994) and two disease‐specific HRQoL instruments i.e. Seattle Angina Questionnaire (Lie 2009; Park 2013), AP‐QLQ (Angina Pectoris‐Quality of Life Questionnaire) (Lisspers 1999) and the MacNew Heart Disease Quality of Life Questionnaire (Melamed 2014). The wide variation in HRQoL outcomes and methods of reporting meant we were unable to meta‐analyse results. Instead, we undertook a detailed tabulation of the overall and domain HRQoL scores from each of the trials with a particular focus on intervention‐control differences at follow‐up. To provide some level of overall synthesis, for each study we assessed whether total and domain HRQoL between‐group differences were statistically different and, if so, the direction of effect (Table 4; Table 5; Table 6; Table 7; Table 8; Table 9; Table 10; Table 11; Table 12; Table 13; Table 14; Table 15; Table 16; Table 17 ).

Table 4. Summary of HRQoL scores at follow‐up: Clark 1997

Sickness Impact Profile+++ at 12 months
	Absolute mean outcome values at follow‐up++			Comparison
	Education	Comparator	Between group P value	Comparison
Clark 1997(12 months)
Total score	7.26	8.09	NS	Education = comparator
Psychosocial dimension	5.52	7.05	≤ 0.05	Education > comparator
Physical dimension	5.89	6.00	NS	Education = comparator
Sickness Impact Profile+++ at 18 months
	Absolute mean outcome values at follow‐up++			Comparison
	Education	Comparator	Between group P value	Comparison
Total score	7.93	7.41	NS	Education = comparator
Psychosocial dimension	6.05	6.23	NS	Education = comparator
Physical dimension	6.40	5.25	NS	Education = comparator

++ for mean scores at follow‐up (adjusted for baseline scores)
+++ lower score higher HRQoL

NS: No significant difference demonstrated
Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up.
Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up.
Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up.
Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported.
Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 5. Summary of HRQoL scores at follow‐up: Clark 2000

Sickness Impact Profile at 12 months
Clark 2000(12 months)	Absolute means at follow‐up++			Comparison
Clark 2000(12 months)	Education	Comparator	Between group P value	Comparison
Psychosocial dimension	5.15	5.91	0.144	Education = comparator
Physical dimension	7.09	7.66	0.05	Education > comparator

Means were adjusted to take account of baseline values

Table 6. Summary of HRQoL scores at follow‐up: Clark 2009

Sickness Impact Profile at 12 months
	Absolute means (SD) at follow‐up				Comparison
	Education	Education self directed	Comparator	Between group P value	Comparison
Total score	8.13 (8.63)	9.79 (10.17)	9.49 (9.46)	NS	Education = comparator
Psychosocial dimension	5.84 (8.02)	7.31 (10.74)	6.75 (9.39)	NS	Education = comparator
Physical dimension	8.07 (9.63)	9.46 (10.11)	9.85 (10.79)	NS	Education = comparator
Sickness Impact Profile at 18 months
Total score	8.44 (9.13)	8.98 (10.29)	9.64 (9.45)	NS	Education = comparator
Psychosocial dimension	5.74 (9.68)	6.16 (8.20)	7.17 (10.40)	NS	Education = comparator
Physical dimension	8.27 (10.02)	8.98 (9.33)	9.65 (10.19)	NS	Education = comparator

Note: analysis of these data was reported, but the individual results were not. These were obtained by contacting the author directly

Table 7. Summary of HRQoL scores at follow‐up: Cohen 2014

SF‐12 (Short Form 12 item survey) at 6 months
	Mean (SD) outcome values at follow‐up		Between group P value	Comparison
	Education	Comparator	Between group P value	Comparison
Mental component summary	47.5 (11.2)	47.7 (11.2)	0.43	Education = comparator
Physical component summary	47.5 (9.3)	47.3 (9.4)	0.44	Education = comparator

Negative baseline‐follow‐up difference favours intervention and positive favours control

Table 8. Summary of HRQoL scores at follow‐up: Cupples 1994

Nottingham Health Profile+ at 24 months
	MD (95% CI) between groups in change from baseline at follow‐up	Between group P value	Comparison
Emotional reaction	0.0 (‐5.2 to 5.2)	NS	Education = comparator
Energy	0.5 (‐7.8 to 8.8)	NS	Education = comparator
Physical mobility	‐0.4 (‐5.2 to 4.5)	NS	Education = comparator
Pain	0.5 (‐4.7 to 5.6)	NS	Education = comparator
Sleep	3.0 (‐4.0 to 9.9)	NS	Education = comparator
Social isolation	‐2.2 (‐6.6 to 2.1)	P < 0.05	Education > comparator
Nottingham Health Profile+ at 60 months
	MD (95% CI) between groups in change from baseline at follow‐up	Between group P value	Comparison
Emotional reaction	‐2.1 (‐7.5 to 3.3)	NS	Education = comparator
Energy	‐4.7 (‐13.2 to 3.7)	NS	Education = comparator
Physical mobility	‐1.3 (‐6.3 to 3.6)	< 0.05	Education > comparator
Pain	‐3.4 (‐9.2 to 2.3)	< 0.05	Education > comparator
Sleep	‐2.4 (‐9.3 to 4.5)	NS	Education = comparator
Social isolation	0.0 (‐4.3 to 4.3)	NS	Education = comparator

+ Higher scores reflect poorer quality of life
The value quoted is the mean difference (MD) (CI) between groups from baseline to follow‐up
P related to t‐tests (two tailed)

Table 9. Summary of HRQoL scores at follow‐up: Cupples 1994

Participant' self assessment of quality of life on a five‐point scale at 24 months
	Initial scores (% participants)		Follow‐up scores (% participants)		Between group P value	Comparison
	Education	Comparator	Education	Comparator	Between group P value	Comparison
					P < 0.03	Education > comparator
Poor	6.3	5.3	6.9	8.3
Fair	27.8	23.3	18.9	21.7
Average	35	39	33.1	33.7
Good	22.7	22.7	29.3	25.3
Very good	8.2	9.7	11.7	11

Note: the between group P value represents the overall "comparison of change in individuals' assessment for intervention and control groups" the significant difference being in favour of the intervention group

Table 10. Summary of HRQoL scores at follow‐up: Furuya 2014

*SF‐12 (Short Form 12 item survey) at 6 months**
	Mean (SD) outcome values at follow‐up		Between group P value	Comparison
	Education	Comparator	Between group P value	Comparison
Mental component summary	51.7 (9.5)	48.4 (9.2)	0.73	Education = comparator
Physical component summary	43.3 (10.6)	41.0 (11.0)	0.28	Education = comparator
*SF‐36 (Short Form 36 item survey)**
	Mean (SD) outcome values at follow‐up		Between group P value	Comparison
	Education	Comparator	Between group P value	Comparison
Social functioning	79.2 (25.1)	64.2 (28.4)	0.1	Education = comparator
Mental health	70.9 (22.7)	70.1 (19.1)	0.98	Education = comparator
Physical functioning	72.5 (23.9)	64.5 (27.8)	0.2	Education = comparator
General health	66.1 (19.8)	63.9 (20.0)	0.61	Education = comparator
Vitality	69.7 (20.6)	62.5 (20.7)	0.52	Education = comparator
Bodily pain	63.8 (28.5)	55.7 (24.2)	0.22	Education = comparator
Role–emotional	77.8 (36.4)	64.4 (36.0)	0.72	Education = comparator
Role–physical	52.5 (40.7)	50.0 (44.0)	0.96	Education = comparator

*Negative baseline ‐ follow‐up difference favours intervention; positive difference favours control

Table 11. Summary of HRQoL scores at follow‐up: Hanssen 2007

*SF‐36 (Short Form 36 item survey) at 6 months**
	Between group difference in mean change from baseline (95% CI) at follow‐up	Between group P value	Comparison
Overall physical	‐2.33 (‐4.54 to ‐0.12)	0.039	Education = comparator
Physical functioning	‐1.16 (‐3.28 to 0.95)	0.28	Education = comparator
Role physical	‐1.84 (‐5.32 to 1.64)	0.299	Education = comparator
Bodily pain	‐1.74 (‐4.54 to 1.05)	0.22	Education = comparator
General health	‐0.36 (‐2.64 to 1.91)	0.752	Education = comparator
Overall mental	1.07 (‐1.71 to 3.86)	0.447	Education = comparator
Vitality	‐0.07 (‐2.23 to 2.10)	0.951	Education = comparator
Social functioning	0.36 (‐2.96 to 3.67)	0.832	Education = comparator
Role‐emotional	0.78 (‐3.29 to to 4.84)	0.706	Education = comparator
Mental health	0.4 (‐1.81 to 2.60)	0.723	Education = comparator
*SF‐36 (Short Form 36 item survey) at 18 months**
	Between group difference in mean change from baseline (95% CI) at follow‐up	Between group P value	Comparison
Overall physical	‐1.44 (‐3.89 to 1.02)	0.25	Education = comparator
Physical functioning	‐0.79 (‐3.06 to 1.48)	0.491	Education = comparator
Role physical	‐0.94 (‐4.76 to 2.88)	0.627	Education = comparator
Bodily pain	‐0.77 (‐4.00 to 2.47)	0.641	Education = comparator
General health	0.25 (‐2.15 to 2.64)	0.838	Education = comparator
Overall mental	1.65 (‐1.35 to 4.65)	0.28	Education = comparator
Vitality	0.58 (‐1.95 to 3.12)	0.65	Education = comparator
Social functioning	0.55 (‐3.95 to 2.85)	0.751	Education = comparator
Role‐emotional	2.59 (‐1.58 to 6.77)	0.221	Education = comparator
Mental health	0.31 (‐2.11 to 2.73)	0.8	Education = comparator

*Negative baseline ‐ follow‐up difference favours intervention; positive difference favours control

Table 12. Summary of HRQoL scores at follow‐up: Lie 2009

Seattle Angina Questionnaire at 6 months
	Absolute mean (SD) outcome values at follow‐up				Comparison
	Education	P value	Comparator	P value	Comparison
Physical limitation	86.4 (15.6)	P < 0.001	83.2 (18.7)	P < 0.001	Education = comparator
Angina frequency	91.7 (16.6)	P < 0.001	90.8 (18.9)	P < 0.001	Education = comparator
Treatment satisfaction	89.2 (15.4)	NS	88.0 (16.1)	NS	Education = comparator
Disease perception	77.8 (20.2)	P < 0.001	73.9 (24.2)	P < 0.001	Education = comparator
SF‐36 (Short Form 36 item survey) at 6 months
	Absolute mean (SD) outcome values at follow‐up				Comparison
	Education	P value	Comparator	P value	Comparison
Overall physical	47.4 (9.6)	P < 0.001	47 (10)	P < 0.001	Education = comparator
Physical functioning	82.2 (19.2)	P < 0.001	82.3 (19.8)	P < 0.001	Education = comparator
Role physical	64 (41.2)	P < 0.001	57.2 (43.3)	P < 0.001	Education = comparator
Bodily pain	77.2 (22.3)	P < 0.001	78.5 (25.2)	P < 0.001	Education = comparator
General health	69.9 (23.3)	NS	65.7 (27.2)	NS	Education = comparator
Overall mental	52.1 (10.7)	P < 0.05	50.5 (10.8)	NS	Favours education
Vitality	61.9 (23.9)	P < 0.001	60.5 (21.6)	P < 0.001	Education = comparator
Social functioning	86.3 (21.4)	P < 0.001	84.3 (21.9)	P < 0.001	Education = comparator
Role‐ emotional	73.3 (38.2)	P < 0.01	67.4 (41.6)	P < 0.01	Education = comparator
Mental health	81.9 (17.3)	P < 0.001	78.5 (21)	P < 0.01	Education = comparator

Higher scores indicate better HRQoL

Table 13. Summary of HRQoL scores at follow‐up: Lisspers 1999

Angina Pectoris ‐ Quality of Life Questionnaire (AP‐QLQ) at 24 months
	Mean (SD) score at follow‐up		Between group P value	Comparison
	Education	Comparator	Between group P value	Comparison
QLQ (total)	4.7 (0.8)	4.3 (1.0)	NS	Education = comparator
Somatic symptoms	4.8 (1.0)	4.3 (1.1)	NS	Education = comparator
Physical activity	4.8 (1.0)	4.1 (1.2)	NS	Education = comparator
Emotional distress	4.8 (0.8)	4.6 (1.1)	NS	Education = comparator
Life satisfaction	4.2 (1.0)	3.9 (1.2)	NS	Education = comparator

Figures quoted represent absolute scores on a self‐rating scale

Table 14. Summary of HRQoL scores at follow‐up: Melamed 2014

MacNew Heart Disease Quality of Life Questionnaire (MacNew) at 220 days
	Mean (SD) outcome values at follow‐up		Between group P value	Comparison
	Education	Comparator	Between group P value	Comparison
Rank sum	5.75 (0.87)	5.74 (0.83)	0.056	Education = comparator

Table 15. Summary of HRQoL scores at follow‐up: Park 2013

Seattle Angina Questionnaire‐Korean (SAQ‐K) at 6 months
	Mean (SD) outcome values at follow‐up		Between group P value	Comparison
	Education	Comparator	Between group P value	Comparison
Physical symptoms
Physical limitation	90.77 (9.97)	85.74 (15.37)	0.901	Education = comparator
Angina stability	78.57 (20.09)	64.17 (23.38)	0.037	Education > comparator
Angina frequency	94.29 (7.90)	89.33 (14.84)	0.543	Education = comparator
Treatment satisfaction	86.38 (12.15)	73.13 (16.09)	0.021	Education > comparator
Diseases perception	74.40 (16.03)	52.78 (15.98)	0.005	Education > comparator

Higher scores indicate better HRQoL

Table 16. Summary of HRQoL scores at follow‐up: Pogosova 2008

SF‐36 (Short Form 36 item survey) at 12 months
	Mean change from baseline P value		Comparison
	Education	Comparator	Comparison
Overall physical	P > 0.05	P ≤ 0.05	Favours education
Physical functioning	P > 0.05	P ≤ 0.05	Favours education
Bodily pain	P > 0.05	P ≤ 0.05	Favours education
Overall mental	P > 0.05	P ≤ 0.05	Favours education
Vitality	P > 0.05	P ≤ 0.05	Favours education
Social functioning	P > 0.05	P ≤ 0.05	Favours education
Mental health	P > 0.05	P ≤ 0.05	Favours education

There were no significant changes demonstrated in the control group but no statistical comparison of the mean change between groups was reported

Table 17. Summary of HRQoL scores at follow‐up: Tingström 2005

*SF‐36 (Short Form 36 item survey) at 12 months**
	Mean change from baseline (SD)		Between group P value⁺	Comparison
	Education	Comparator	Between group P value⁺	Comparison
Physical functioning	3.6 (17.6)	4.4 (15.1)	0.749	Education = comparator
Role physical	38.2 (46.9)	33.8 (42.4)	0.504	Education = comparator
Bodily pain	5.69 (31.1)	6.18 (29.1)	0.911	Education = comparator
General health	1.4 (15.9)	1.8 (16.3)	0.862	Education = comparator
Vitality	5.3 (22.7)	4.9 (21.8)	0.921	Education = comparator
Social functioning	9.7 (24)	9.1 (25.3)	0.869	Education = comparator
Role emotional	15.8 (48.1)	16.5 (41.1)	0.913	Education = comparator
Mental health	2.9 (16.6)	4.2 (17.8)	0.566	Education = comparator

*Positive values indicate improvement in HRQL from baseline
+ P values are calculated on the difference between groups at pre‐test and on the mean change (post test minus pre‐test).

Whilst overall we found no consistent difference in HRQoL total or domain score at follow‐up between intervention and comparator, a number of studies reported statistically significant differences in HRQoL domains in favour of intervention (Clark 1997; Clark 2000; Cupples 1994; Park 2013). Pogosova 2008 demonstrated an improvement in all SF‐36 domain scores and Lie 2009 an improvement in the overall mental score in the intervention groups. No studies reported HRQoL scores that favoured the comparator group.

Although Southard 2003 reported Dartmouth COOP Quality of Life scores at trial entry, there were no reports of this outcome at follow‐up. Esposito 2008 reported on a HRQoL questionnaire undertaken in a randomly selected subgroup of patients from the overall trial. No significant differences were found between the intervention and control groups in a number of measures of mental and physical status, including: "Primary condition interfered a lot or somewhat with enjoyment of life in the last 4 weeks" (between‐group difference ‐3.6% (in favour of intervention) P = 0.379); "Beneficiary felt primary condition placed a burden on family in the past 4 weeks" (between‐group difference 0.5% P = 0.897); "Beneficiary felt depressed about living with primary condition in the past 4 weeks" (between‐group difference 1.2% (in favour of control) P = 0.766).

Adverse events

Few studies reported on adverse events other than mortality and cardiovascular‐related morbidity and hospitalisations. Cohen 2014 reported numbers of participants in the intervention and usual care groups with arrhythmia (1.6% and 2.8% respectively); coronary angiography (7.3% and 8.9%); scheduled angioplasty (4.9% and 3.2%); ACS, stent thrombosis, or chest pain (13.5% and 12.6%); and dyspnoea, lung oedema, or congestive heart failure (3.7% and 2.4%). Esposito 2008 reported that there were no significant differences between the groups in the number of preventable events such as hospitalisations for pneumonia or exacerbations of heart failure, or lower‐extremity amputations in patients with diabetes.

No study reported any intervention‐related adverse events such as prohibitive time or travel demands which would prevent participation in the intervention.

Healthcare costs and cost‐effectiveness

Five studies reported healthcare utilisation and costs (Clark 2000; Cupples 1994; Esposito 2008; Peikes 2009; Southard 2003). Given that cost results were presented in different currencies and incurred in different years, it was difficult to directly compare studies. Furthermore, although studies assessed healthcare costs, there was variation in the particular aspects of healthcare costs that were quantified. Components of costs considered included inpatient admissions, primary care visits, emergency attendances, use of drugs, investigations and subsequent procedures performed. To compare studies and gain an overall impression of the differences in healthcare between intervention and control, we undertook a detailed tabulation of the overall and component healthcare costs for each of the included studies (Table 18).

Table 18. Cost summary of intervention and comparison of healthcare costs incurred by intervention and control groups during follow‐up period

Variable	Clark 2000	Cupples 1994	Esposito 2008				Southard 2003	Peikes 2009
Follow‐up	24 months	24 months	6 months	7 to 12 months	12 months	18 months	6 months	25 months
Year	2000	NR	2005 to 2006				NR	2002 to 2005
Currency	USD	GBP	USD				USD	USD
Mean cost of cardiac rehabilitation program per patient
Total costs	USD 187	GBP 49.72	USD 162				USD 453	USD 196
Costs considered	Personnel, instructional materials, telephone supplies, ongoing staff training	Direct costs by health visitors (staff time), Travel Costs	Average monthly fee paid to the program per member				Nurse salary Overheads Subscription costs	Average monthly fee paid to the program per member
Comments	Participating site overheads were not measured, a "conservatively high" estimate of these was taken to double the treatment cost to USD 374	Costs of the health visitor also included time spent recording data collection for the study						Cost varied among the 15 included studies. Negotiated locally with Medicare and Medicaid Services. (Range USD 50 to USD 444)
Mean total healthcare costs per patient
Total cost (intervention)	USD ~3300 (calc)	GBP 1801	USD 1627	USD 2356	USD 2288	USD 1793	USD 635	USD 1283^*
Total cost (control)	USD ~6500	GBP 1812	USD 1632	USD 2464	USD 2372	USD 1818	USD 2053	USD 1314^*
Between group difference	USD ~1800*	GBP 9.60	USD 5	USD 107	USD 84	USD 25	USD 1418	USD 144 (80% CI 99 to 188)
P value	NR	NS	0.895	0.077	0.132	0.365	NR	< 0.001
Cost saving per pt (when cost of intervention taken into account§)	USD ~1610 or USD ~1420 if estimated overheads were included	GBP 40	USD ‐157	USD ‐55	USD ‐78	USD ‐137	USD 965	USD ‐52
Additional healthcare costs considered	Number of admissions (heart related), number of inpatient days, In patient cost' emergency department costs	Prescription of drugs, GP visits, visits to hospital as inpatients and outpatients, all tests investigations and treatments carried out	Medicare medical claims				Cardiovascular‐related emergency department visits and hospitalisations
Comments	Expenditure was calculated from differences in % utilisation of hospital services. i.e. hospital charges for participants were on average 49% lower and the average annual expenditure was USD 6500. * There was a calculated saving of a hospital charge of USD 3200, the ratio of payments to charges was 0.56 therefore USD 1800 actual saving	There was a difference in the drug usage at baseline which is not accounted for in these figures although this would make minimal impact to the results. The intervention group were more costly for drugs, procedures and service use	Claims quoted are per member per month					*Expenditure/pt/month enrolled Overall costs were increased by 11% when the care coordination fees were taken into account
Summary difference between groups	Favours Rx	Rx = Control	Rx = control (for all time periods studied)				Favours Rx	Favours control

§ = Negative mean difference indicates a net cost of the intervention group
NR = not recorded
NS = not significant

Reflecting the different education modalities and intensities of the interventions, the reported cost of provision per patient varied from GBP 49 (Cupples 1994) to USD 453 (Southard 2003). The largest trials, investigating the efficiency of the Medicare system in the USA (Esposito 2008; Peikes 2009), did not investigate the cost of providing the intervention but instead reported the charge associated with providing this service negotiated by the supplier (care co‐ordination fee). Peikes reported a mean of USD 196 per month (Peikes 2009), while Esposito reported a mean of USD 162 per month (Esposito 2008).

Two studies reported an overall average net saving, after subtracting costs of intervention provision, of USD 965 per patient at six‐months follow‐up (Southard 2003) and USD 1420 per patient at 24‐months follow up (Clark 2000). Peikes 2009 reported an increase in average net costs of USD 52 per patients; six of the 15 programmes investigated had higher costs for the intervention group. Two trials found no difference in between‐group net costs (Cupples 1994; Esposito 2008).

Meta‐regression and stratified meta‐analysis

Predictors of total mortality and all‐cause withdrawal were examined across the longest follow‐up of each individual study using univariate meta‐regression (Table 19; Table 20). We found no evidence that total mortality risk was associated with case mix, age of participants, percentage of male participants, type of cardiac rehabilitation, method of delivery, duration of intervention, theoretical basis of intervention, involvement of family members, study location, setting, or length of follow‐up (Table 19). Similarly, we found no associations between predictors of withdrawal, with the exception of evidence of an increased risk of withdrawal in studies with shorter follow‐up periods (Table 20). Due to poor reporting we were unable to examine the association of dose of education or timing following the index event with the risk of total mortality or withdrawal.

Table 19. Results of univariate meta‐regression analysis for total mortality

Explanatory variable (n trials)	Exp(slope)*	95% CI univariate; P value	Proportion of variation explained	Interpretation
Case mix (% myocardial infarction patients) (n = 11)	RR = 1.004	0.988 to 1.020 P = 0.631	‐190.36%	No evidence that RR is associated with case mix
Age of participants (n = 13)	RR = 1.005	0.940 to 1.074 P = 0.876	‐28.26%	No evidence that RR is associated with the age of participants
Percentage of male participants (n = 13)	RR = 0.991	0.986 to 1.012 P = 0.882	‐25.27%	No evidence that RR is associated with the percentage of male participants
Type of CR (education only vs. education plus e.g. exercise or psychological intervention) (n = 13)	RR = 0.181	0.014 to 2.321 P = 0.168	28.25%	No evidence that RR is associated with type of CR
Method of structured educational delivery (one‐to‐one vs. group versus combination) (n = 13)	RR = 1.010	0.728 to 1.401 P = 0.948	‐28.28%	No evidence that RR is associated with method of delivery
Duration of intervention (n = 12)	RR = 0.978	0.948 to 1.010 P = 0.152	3.69%	No evidence that RR is associated with duration of intervention
Theoretical vs. no theoretical basis to educational intervention (n = 13)	RR = 1.473	0.750 to 2.895 P = 0.233	‐0.31%	No evidence that RR is associated with theoretical basis
Involvement of significant others (e.g. spouse, family member) in the education programme (n = 13)	RR = 1.245	0.890 to 1.722 P = 0.166	7.06%	No evidence that RR is associated with family involvement
Study location (n = 13)	RR = 1.050	0.714 to 1.543 P = 0.787	‐59.78%	No evidence that risk ratio is associated with study location
Setting (centre vs. home) (n = 13)	RR = 1.171	0.773 to 1.774 P = 0.421	‐44.55%	No evidence that RR is associated with centre status
Length of follow‐up (n = 13)	RR = 0998	0.964 to 1.033 P = 0.924	‐22.64%	No evidence that RR is associated with length of follow‐up

CR ‐ cardiac rehabilitation; RR ‐ risk ratio

Table 20. Results of univariate meta‐regression analysis for withdrawal

Explanatory variable (n trials)	Exp(slope)*	95% CI univariate P value	Proportion of variation explained	Interpretation
Case mix (% myocardial infarction patients) (n = 12)	RR = 1.002	0.992 to 1.013 P = 0.611	‐9.25%	No evidence that RR is associated with case mix
Age of participants (n = 17)	RR = 0.998	0.963 to 1.034 P = 0.903	‐15.62%	No evidence that RR is associated with the age of participants
Percentage of male participants (n = 17)	RR = 0.999	0.992 to 1.005 P = 0.621	‐28.64%	No evidence that RR is associated with the percentage of male participants
Type of CR (education only vs. education plus e.g. exercise or psychological intervention) (n = 17)	RR = 0.752	0.260 to 2.174 P = 0.575	‐16.99%	No evidence that RR is associated with type of CR
Method of structured educational delivery (one‐to‐one vs. group versus combination) (n = 17)	RR = 1.033	0.860 to 1.242 P = 0.714	‐29.18%	No evidence that mortality risk is associated with method of delivery
Duration of intervention (n = 16)	RR = 0.993	0.964 to 1.023 P = 0.625	‐25.06%	No evidence that mortality risk is associated with duration of intervention
Theoretical vs. no theoretical basis to educational intervention (n = 17)	RR = 1.031	0.690 to 1.541 P = 0.874	‐24.70%	No evidence that RR is associated with theoretical basis
Involvement of significant others (e.g. spouse, family member) in the education (n = 17)	RR = 1.016	0.829 to 1.245 P = 0.872	‐33.62%	No evidence that RR is associated with family involvement
Study location (n = 17)	RR = 0.942	0.801 to 1.109 P = 0.449	24.47%	No evidence that RR is associated with study location
Setting (centre vs. home) (n = 17)	RR = 1.096	0.873 to 1.374 P = 0.404	‐13.38%	No evidence that RR is associated with centre status
Length of follow‐up (n = 17)	RR = 0.976	0.955 to 0.998 P = 0.035	90.79%	Significant evidence that risk of withdrawal is increased in studies with a shorter follow‐up

CR ‐ cardiac rehabilitation; RR ‐ risk ratio

Sensitivity analysis

Sensitivity analysis found no evidence that total mortality risk was associated with year of publication (before 2000 versus 2000 or later), or risk of bias (Table 21). Similarly, we found no associations between risk of withdrawal and year of publication, but did see evidence of an increased risk of withdrawal in studies with higher risk of bias (low risk in ≥ five items versus < five items) (Table 22).

Table 21. Results of sensitivity analysis for total mortality

Explanatory variable (n trials)	Exp(slope)*	95% CI univariate P value	Proportion of variation explained	Interpretation
Year of publication (n = 13)	RR = 0.998	0.950 to 1.047 P = 0.913	‐61.7%	No evidence that RR is associated with year of publication
Risk of bias (low risk in ≥ 5 items vs. < 5 items) (n = 13)	RR = 1.105	0.421 to 3.831 P = 2.899	‐84.29%	No evidence that RR is associated with risk of bias

RR ‐ risk ratio

Table 22. Results of sensitivity analysis for withdrawal

Explanatory variable (n trials)	Exp(slope)*	95% CI univariate P value	Proportion of variation explained	Interpretation
Year of publication (before 2000 vs. 2000 or later) (n = 17)	RR = 1.017	0.982 to 1.052 P = 0.327	‐7.02%	No evidence that RR is associated with year of publication
Risk of bias (low risk in ≥ 5 items vs. < 5 items) (n = 17)	RR = 1.437	1.069 to 1.931 P = 0.020	15.35%	Significant evidence that risk of withdrawal is increased in studies with higher risk of bias

Data for all outcomes were pooled using both random‐effects and fixed‐effect modelling (Table 1). With the exception of hospitalisations and withdrawals at the longest follow‐up, the direction of effect of all outcomes was the same, regardless of the model used. Similarly, with the exception of total mortality, the choice of model used did not change whether or not the confidence intervals included the null hypothesis.

Small study bias

With the exception of total mortality and withdrawals, there were too few studies and outcome data to assess small study bias by means of funnel plots or Egger's test. There was no evidence of funnel plot asymmetry or statistically significant Egger's test for total mortality (Figure 4 P = 0.83) or withdrawals (Figure 5; P = 0.10).

Figure 4

Funnel plot of comparison: 4 Education versus no education, outcome: 4.1 Total mortality at the end of the follow up period.

Figure 5

Funnel plot of comparison: 1 Education versus no education, outcome: 1.6 Withdrawals.

Quality of evidence from randomised controlled trials

The quality of the evidence for outcomes reported in the review was rated using the GRADE method (Schünemann 2011). The quality of the evidence varied widely by outcome and ranged from very low to moderate (see summary of findings Table 1). The reasons for downgrading evidence of outcomes included poor reporting of blinding of outcome assessors in at least 50% of the studies which contributed data to the evidence, evidence of heterogeneity (I² > 50%), or imprecision around the point of effect.

Discussion

Summary of main results

We included 22 randomised controlled trials (RCTs) involving 76,864 participants with coronary heart disease (CHD) where education was the primary interventional intent of cardiac rehabilitation. The 'dose' of the education intervention varied substantially across studies from just one 40 minute face‐to‐face session plus a 15 minute follow‐up call (Dracup 2009) to a four‐week residential stay re‐inforced with 11 months of nurse‐led follow up sessions (Lisspers 1999). Control participants typically received usual medical care without a formalised cardiac rehabilitation or secondary prevention education programme.

We found no difference in effect of education on total mortality compared with control. Pooled data for total mortality translates to a possible reduction of 18 deaths per 1000 people, to an increase in two deaths per 1000 people, compared with the assumed risk in the control group of 46 deaths per 1000 (summary of findings Table 1). As individual causes of mortality were not reported across studies, we were unable to report separate data for cardiovascular mortality or non cardiovascular mortality. We found no evidence that education reduced fatal and/or non‐fatal myocardial infarction (MI). We found some evidence of a reduction with education in other fatal and/or non‐fatal cardiovascular events, although this was based only on two studies (310 participants). We found no evidence that education reduced revascularisations, or hospitalisation, compared to control groups receiving no education.

Univariate meta‐regression analysis shows that the impact of education on total mortality appears to be largely consistent across trials irrespective of case mix (percentage of post‐MI participants), type of rehabilitation (exercise‐only versus comprehensive), dose of education, duration of follow‐up, study location, study setting or risk of bias. As most studies had a relatively short follow‐up, these results were based on very few events, and therefore, our meta‐analysis lacks sufficient statistical power to make definitive conclusions about the impact of educational interventions on events in people with CHD.

Although health‐related quality of life (HRQoL) was reported by almost all included studies, we were unable to pool findings due to the heterogeneity of measures. Whilst there was some evidence of higher HRQoL in some domain scores, overall there was no consistent evidence of superior HRQoL following education compared to control. Many studies used generic HRQoL measures that are known to lack sensitivity with cardiac treatment, particularly in comparison with disease‐specific measures (Cohen 2014; Furuya 2015; Oldridge 2003; Taylor 1998).

The intention of including analysis of withdrawal from the intervention was to use it as a surrogate for the adverse effects of the intervention, e.g. the educational intervention was so demanding that it could not be completed by participants. However, despite withdrawals being reported by most studies, the reasons for withdrawal were not always clearly described. We found no increase in withdrawals with education compared to control.

The search for this update identified no new studies that reported healthcare costs or cost‐effectiveness data. As previously reported (Brown 2011), different currencies and the years in which studies were conducted, made it difficult to directly compare healthcare costs across studies. The cost of the educational intervention varied widely (between GBP 49 and USD 453 per patient) reflecting the differing intensity and requirements for provision of the interventions investigated. There was some evidence that when compared to usual care, patient education may be cost‐saving as a result of a reduction in downstream healthcare utilisation.

Overall completeness and applicability of evidence

The scope of this review was limited in its design in three specific ways:

Inclusion only of studies published in 1990 or later;
Inclusion only of studies where the educational component was the primary intention of the intervention; and
Inclusion only of studies that reported event data (e.g. total mortality) as opposed to intermediate outcomes (e.g. blood pressure, exercise tolerance).

These limitations in scope were crucial in addressing the specific question of what is the 'added value' of patient education in the context of contemporary cardiovascular management. The interpretation of previous systematic reviews of patient education have been confounded by including multicomponent rehabilitation interventions, of which education was only an element, and reporting on studies using surrogate outcomes such as health knowledge or blood pressure. Indeed, many of the trials identified and considered in this review process only investigated outcomes such as cardiovascular risk factor reduction, or pre‐hospital delay from time of symptom onset to hospital arrival, and have not been included in this review.

In spite of the focus of this review, there was considerable heterogeneity of participants and interventions. Several studies included CHD in combination with comorbidities such as diabetes, hypertension or a degree of heart failure (Esposito 2008; Peikes 2009; Southard 2003) and interventions varied substantially in content, mode of delivery and dose. It could be argued that a benefit of this heterogeneity is that the results are more likely to be applicable to the wider population of people with CHD and clinical practice. However, we must acknowledge that different components of the educational intervention may contribute to the composite effect of the education to varying extents. Similarly, fidelity (whether the intervention was delivered as intended) and dose (the quantity of intervention implemented) are important aspects of the delivery of a complex intervention such as cardiac rehabilitation, which were not reported by any included studies.

Previous reviews of patient education, and more broadly cardiac rehabilitation, have identified the paucity of research into outcomes in women and the elderly. However, this review includes several studies with a substantive proportion of women (Clark 2000) and older people (Clark 2009) specifically addressing this disparity. Nevertheless, ethnic minorities remain under‐represented, with a mean of 84% of participants described as Caucasian in studies that reported ethnicity.

Quality of the evidence

The general lack of reporting of methods in some of included RCTs made it difficult to assess methodological quality and thereby judge risk of bias. Details were often poorly reported and confirmation of methodology needed to be sought from some authors. Interestingly, reporting of methods was inferior in some of the more recent studies to many of the older studies, leading to a higher risk of selection and attrition bias than was reported in the previous version of this review (Brown 2011). The area of greatest potential risk of bias in this review was detection bias (lack of outcome assessment blinding), with less than half the studies providing sufficient details to judge if outcomes were assessed by researchers blinded or independent to the trial.

Due to this poor reporting, the quality of the evidence for outcomes was assessed as moderate at best. Other reasons for downgrading the quality of evidence included inconsistency (hospitalisations), and imprecision (total mortality, hospitalisations, subsequent MI, revascularisations and withdrawal).

The other area of potential risk of bias was the imbalance of co‐interventions received by intervention and control subjects, which made it difficult to investigate the specific impact of education on outcomes. We specifically selected studies on the basis of education being the primary intervention. However, a number of studies appeared to include additional elements (e.g. behaviour modification or psychological support) in the educational intervention arm, which led to a risk of performance bias. Whilst the decision to include studies was made independently by two review authors, the decision was ultimately one of judgement based on the description of the intervention provided by the authors. During correspondence, the lead author of one included study stated: "I would not define our program as 'patient education' (at least according to the way I define this term) ‐ more as a 'behaviour change program'...we ...very much tried to develop active program components which actually and concretely supported the behaviour change process in the short term and for the long‐term maintenance" (Lisspers 1999). We would argue that a key objective of patient education is to change behaviour, i.e. through education, patients learn to understand the reasoning for improved diet, exercise regime and compliance with medication and are, therefore, more likely to modify their behaviour. This objective is consistent with adult learning theory; learning is the outcome of education and can be defined as "a relatively permanent change in behaviour as a result of experience, training or practice" (Reece 2007).

Potential biases in the review process

We believe this is the most comprehensive systematic review to date of RCT‐based evidence for the impact of education‐based cardiac rehabilitation for people with CHD. However, our review has some limitations. Given the inconsistent reporting of outcomes, we were unable to judge the degree of publication bias for most outcomes, although there was no evidence of funnel plot asymmetry or statistically significant Egger's test for total mortality or all‐cause withdrawal. Although a specific goal of this updated review was to clarify the impact of education programmes on clinical events, many of the included trials were relatively small and had short‐term follow‐up periods so that the number of deaths and hospitalisations reported by most trials was small. Indeed, in many studies, we located event data in the trial descriptions of losses to follow‐up and exclusions, rather than as stated primary or secondary outcomes. We also acknowledge that the median outcome follow‐up of 12 months is limited when assessing for impact on total mortality and morbidity outcome measures. However, reassuringly, our results were consistent when pooling was limited to RCTs with a follow‐up more than 12 months.

In this updated review, we had hoped to categorise the CHD diagnoses of trial participants according to a more detailed framework based on Braunwald's classification of CHD (Braunwald 2011) and current clinical management of CHD. However, given the lack of details of trial participants, this more detailed assessment of diagnostic groups was not possible. All participants in the included studies had documented CHD; most had experienced MI or undergone revascularisation. As with the previous version of this review, we combined these different participant groups because there were insufficient data to stratify trials by CHD type. Finally, the mean age of participants in the included trials ranged from 51.0 years to 72.8 years. Elderly participants in clinical trials are under‐represented with only 14% of participants in clinical trials reported to be aged over 75 years (Dodd 2011). Yet, Myocardial Ischaemia National Audit Project registry data indicate that around 40% of people with ACS are in that age group (Zaman 2014). Clinical trial participants are therefore unlikely to be truly representative of, and are likely to have relatively fewer comorbidities than, the more frail, older population seen in clinical practice (Alexander 2007; Zaman 2014).

Agreements and disagreements with other studies or reviews

The findings of this updated review are largely in accord with the previous version (Brown 2011). In this update, we identified nine new RCTs looking at the effect of education on people with CHD. Disappointingly, few studies reported additional total mortality or event data, and consequently, the results of this update remain largely unchanged. While the previous version showed no evidence for the effect of education when delivered as part of cardiac rehabilitation on total mortality in people with CHD, this review found evidence of a reduced risk for this outcome (random effects RR 0.80, 95% CI 0.60 to 1.05; fixed effect RR 0.80, 95% CI 0.66 to 0.97, moderate quality evidence). Similarly, early systematic reviews by Mullen 1992 and Dusseldorp 1999, reported a statistically significant reduction in mortality and morbidity in people with CHD following an educational intervention. These earlier reviews included studies with multidimensional interventions (e.g. education plus psychological interventions), non‐randomised studies, and studies conducted before 1990 prior to the era of optimal medical therapy. Given the proven survival advantage of contemporary medical treatments, and the limited opportunity for mortality gain in this patient cohort, any incremental total mortality benefit with education is likely to be small.

A more recent systematic review that investigated the impact of education on patients’ knowledge and health behaviour change in patients with CHD reported that educational interventions in cardiac care increased patients’ knowledge and facilitated behaviour change (Ghisi 2014). Educational interventions were found to lead to increases in physical activity, healthier dietary habits and smoking cessation, but no associations between education and cardiac symptoms, medication adherence or psychosocial well‐being were found (Ghisi 2014).

Figure 1

PRISMA flow diagram

Figure 2

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

Figure 3

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

Figure 4

Funnel plot of comparison: 4 Education versus no education, outcome: 4.1 Total mortality at the end of the follow up period.

Figure 5

Funnel plot of comparison: 1 Education versus no education, outcome: 1.6 Withdrawals.

Analysis 1.1

Comparison 1: Education versus no education, Outcome 1: Total mortality at the end of the follow up period

Analysis 1.2

Comparison 1: Education versus no education, Outcome 2: Fatal and/or non‐fatal MI

Analysis 1.3

Comparison 1: Education versus no education, Outcome 3: Other fatal and/or non‐fatal cardiovascular events

Analysis 1.4

Comparison 1: Education versus no education, Outcome 4: Total revascularisations (including CABG and PCI)

Analysis 1.5

Comparison 1: Education versus no education, Outcome 5: Hospitalisations

Analysis 1.6

Comparison 1: Education versus no education, Outcome 6: Withdrawals

Summary of findings 1. Patient education for the management of coronary heart disease

Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No of Participants (studies)	Quality of the evidence (GRADE)	Comments
Patient education for the management of coronary heart disease
Patient or population: patients with coronary heart disease Settings: Centre or home‐based Intervention: Patient education
	Assumed risk	Corresponding risk
	Control	Patient education
Total mortality at the end of the follow‐up period No of deaths Follow‐up: median 18 months	Study population		RR 0.80 (0.60 to 1.05)	10075 (13 studies)	⊕⊕⊕⊝ Moderate¹
	46 per 1000	37 per 1000 (28 to 48)
	Moderate population
	43 per 1000	34 per 1000 (26 to 45)
Fatal and/or non‐fatal MI at the end of the follow up period Follow‐up: median 33 months	Study population		RR 0.63 (0.26 to 1.48)	209 (2 studies)	⊕⊝⊝⊝ very low² ³ ⁴
	118 per 1000	74 per 1000 (31 to 174)
	Moderate population
	106 per 1000	67 per 1000 (28 to 157)
Other fatal and/or non‐fatal cardiovascular events Follow‐up: median 21 months	Study population		RR 0.36 (0.23 to 0.56)	310 (2 studies)	⊕⊕⊝⊝ low² ⁴
	386 per 1000	139 per 1000 (89 to 216)
	Moderate population
	324 per 1000	117 per 1000 (75 to 181)
Total revascularisations (including CABG and PCI) Follow‐up: median 36 months	Study population		RR 0.58 (0.19 to 1.71)	456 (3 studies)	⊕⊝⊝⊝ verylow² ³ ⁴
	35 per 1000	20 per 1000 (7 to 60)
	Moderate population
	33 per 1000	19 per 1000 (6 to 56)
Hospitalisations (cardiac‐related)at end of follow up period Follow‐up: median 12 months	Study population		RR 0.93 (0.71 to 1.21)	14849 (5 studies)	⊕⊝⊝⊝ very low¹ ² ⁵
	79 per 1000	74 per 1000 (56 to 96)
	Moderate population
	141 per 1000	131 per 1000 (100 to 171)
All cause withdrawal at follow‐up Follow‐up: median 12 months	Study population		RR 1.04 (0.88 to 1.22)	10972 (17 studies)	⊕⊕⊝⊝ low² ⁶ ⁷
	92 per 1000	96 per 1000 (81 to 113)
	Moderate population
	70 per 1000	73 per 1000 (62 to 85)
HRQoL Various HRQoL measures Follow‐up: median 12 months	Not measurable	Not measurable	Not measurable	4393 (13 studies)	⊕⊕⊕⊝ moderate²	HRQoL in intervention > HRQoL in comparator, in then 9/99 domains

¹ 95% CIs include both no effect and appreciable benefit (i.e. CI < 0.75) ² Blinding of outcome assessors was poorly described in over 50% of included studies; bias likely ³ 95% CIs include both no effect, appreciate benefit and appreciable harm (i.e. CI < 0.75 and > 1.25) ⁴ The point estimate is likely to be imprecise due to very low event rates ⁵ I² > 40%; heterogeneity may be important ⁶ 95% CIs include both no effect and appreciate harm (i.e. CI > 1.25) ⁷ Evidence of funnel plot asymmetry therefore publication bias likely

Summary of findings 1. Patient education for the management of coronary heart disease

Table 1. Results of sensitivity analysis for fixed‐effect versus random‐effects models

Outcome or subgroup	Studies	Participants	Effect estimate (random‐effect) RR (M‐H, random, 95% CI)	Effect estimate (fixed‐effect) RR (M‐H, fixed, 95% CI)
1.1 Total mortality at the end of the follow up period	13	10,075	0.80 [0.60, 1.05]	0.80 [0.66, 0.97]
1.1.1 Studies with 12 months or less follow‐up	6	4063	0.78 [0.35, 1.78]	0.87 [0.56, 1.36]
1.1.2 Studies with more than 12 months follow‐up	7	6012	0.78 [0.60, 1.02]	0.79 [0.64, 0.97]
2.1 Myocardial Infarction at the end of the follow‐up period	2	209	0.63 [0.26, 1.48]	0.59 [0.25, 1.38]
2.2 Total revascularisations	3	456	0.58 [0.19, 1.71]	0.58 [0.20, 1.69]
2.3 Other fatal and/or non‐fatal cardiovascular events	2	310	0.36 [0.23, 0.56]	0.36 [0.23, 0.56]
3.1 Cardiac hospitalisations at end of follow‐up period	5	14,849	0.93 [0.71, 1.21]	1.02 [0.90, 1.15]
4.1 All cause withdrawal or drop‐out at follow‐up	17	10,972	1.04 [0.88, 1.22]	0.98 [0.88, 1.10]
4.1.1 Studies with 12 months or less follow‐up	10	4960	1.18 [0.93, 1.49]	1.18 [0.94, 1.49]
4.1.2 Studies with more than 12 months follow‐up	7	6012	0.98 [0.80, 1.20]	0.92 [0.81, 1.05]

Table 1. Results of sensitivity analysis for fixed‐effect versus random‐effects models

Table 2. Educational content of programs in included studies

Study ID	Description of Intervention	Theoretical basis	Tailored	Duration	One‐to‐one	Group	Face‐ to‐face	Telephone	Internet	Notes
Chow 2015	Text message‐based prevention program delivering regular semi‐personalised messages providing advice, motivation, and information to improve diet, increase physical activity, and encourage smoking cessation	NR	Y	4 messages per week for 24 weeks	Y	N	N	Y (text messages)	N	Content for each participant was selected using a prespecified algorithm dependent on key baseline characteristics
Clark 1997	*PRIDE	Y	Y	Once weekly for 4 weeks		Y	Y			Taught by health educator. Videotape and workbook aids
Clark 2000	*PRIDE	Y	Y	Once weekly for 4 weeks		Y	Y			Taught by health educator. Videotape and workbook aids
Clark 2009	*PRIDE	Y	Y	Once weekly for 6 weeks	Y	Y	Y			3 groups (self‐directed and group intervention and a control)
Cohen 2014	"House of Education" with individualised consultations with e.g. smoking cessation nurse	NR	Y	At least 6 sessions in 12 months	Y	N	Y	N	N	Consultations content was individualised according to a patient’s risk factors
Cupples 1994	Practical tailored advice on cardiovascular risk factors and appropriate health education	NR	Y	3 times a year for 2 years	Y		Y			Delivered at home by health visitor
Dracup 2009	Patients received education on ACS, anticipated emotional issues and social factors that could affect delay	Y	Y	55 mins (40 min face‐to‐face plus 15 min follow‐up call)	Y	N	Y	N	N	Delivered by a nurse with expertise in cardiology
Esposito 2008	Predesigned scripts to provide education on various aspects of care, geared to personalised clinical goals	NR	Y	Average 1.1 contacts per month for 18 months	Y		Y	Y		Nurse case manager, primarily by telephone but also face‐to‐face
Furuya 2015	Three booklets and three telephone follow‐up calls aimed at helping patient understand his cardiac condition, PCI and how to cope with CAD	Y	N	2 face to face sessions and 3 telephone calls over 16 weeks	Y	N	Y	Y	N	The first booklet was discussed with participants before undergoing PCI procedure
Hanssen 2007	Individualised education from a menu of topics to be covered	Y	Y	6 months (8 sessions in total)	Y			Y		Structured element and an on‐call element
Jorstad 2013	Outpatient clinic visits to a cardiovascular nurse	NR	Y	6 months (4 sessions)	NR	NR	Y	N	N	Nurse‐coordinated: provided general lifestyle advice, and individual counselling
Lie 2009	A psycho‐educative intervention. Structured information and psychological support	NR	N/S	2 visits (1 hour each)	Y		Y			Critical care nurse, home based
Lisspers 1999	Health education and achievement of behavioural change	NR	Y	4 week residential then 11 month one‐to‐one individual sessions	Y	Y	Y			Trained nurses (personal coaches). Seminars, lectures, discussion and skills sessions
Melamed 2014	Lesson materials consisted of a patient brochure, teaching cards and curriculum poster/wall chart set	NR	N	NR	Y	N	Y	N	N	Patients were given an exercise diary to enable them to document their daily physical activity
Mooney 2014	Education intervention aimed at reducing total prehospital delay time	Y	Y	6 months (1 face‐to‐face session, 1 telephone call and one reinforcement letter at 6 months)	Y	N	Y	Y	N	Research nurses used preprinted flip charts and prescriptive scripts as educational aids
Moreno‐Palanco 2011	Health education on the meaning of patients' disease and the importance of treatment	NR	NR	3 years (at least 5 sessions)	Y	N	Y	N	N	Each visit consisted of a nursing intervention and a medical assessment
P.RE.COR Group 1991	Education and counselling on management of cardiovascular risk factors and exercise	NR	Y	1 group session, 1 individual session with cardiologist	Y	Y	Y			Multidisciplinary input to group. Cardiologist tailors therapy
Park 2013	Psycho‐educational intervention comprising tailored face‐to‐face education and telephone‐delivered health coaching	NR	Y	12 weeks (6 sessions)	Y	N	Y	Y	N	Patients made choices about risk factors they wanted to lower and participated in goal setting
Peikes 2009	Variable ‐ nurse provision of patient education	NR	NR	1 to 2.5 times a month for an average of 30 months	Y			Y		15 different programs, majority telephone, one‐to‐one
Pogosova 2008	Structured program addressing different risk factors in each session	Y	NR	6 sessions (twice weekly, 90 min)		Y	Y
Southard 2003	Modular internet sessions, Interactive multiple choice and self tests followed by feedback	NR	NR	Once weekly for 6 months (at least 30 min)	Y	Y			Y	Communication with case manager and online discussion group
Tingström 2005	Problem based rehabilitation to teach a planned curriculum	Y	NR	13 sessions over 1 year		Y	Y			Trained facilitator
PRIDE = Problem Identification, Researching one's routine, Identifying a management goal, Developing a plan to reach it, Expressing one's reactions and Establishing rewards for making progress. Y = Yes; N = No; NR = not reported

Table 2. Educational content of programs in included studies

Table 3. All‐cause withdrawal or drop‐out at follow‐up

Study ID		Number randomised	Number lost at follow‐up*	Notes
Chow 2015	Intervention	352	33	20 excluded from analysis, 9 unable to contact, 4 died
Chow 2015	Control	358	25	21 excluded from analysis, 3 unable to contact, 1 died
Clark 2000	Intervention	309	51	36 withdrew, 14 died, 1 data missing
Clark 2000	Control	262	42	33 withdrew, 8 died, 1 data missing
Clark 2009	Intervention	201	37	Self‐directed program; 33 withdrew, 4 died
	Intervention	190	24	Group format; 19 withdrew, 5 died
	Control	184	23	15 withdrew, 8 died
Cohen 2014	Intervention	251	48	6 did not meet inclusion criteria, 7 died, 23 follow‐up refusal, 10 lost to follow‐up, 2 in another protocol
Cohen 2014	Control	251	36	4 did not meet inclusion criteria, 7 died, 13 follow‐up refusal, 12 lost to follow‐up
Cupples 1994	Intervention	342	92	45 defaulted, 47 died; 21 defaulted at 2 years
Cupples 1994	Control	346	109	44 defaulted, 65 died; 25 defaulted at 2 years
Dracup 2009	Intervention	1777	197	89 lost to follow‐up, 41 withdrawn, 67 died
Dracup 2009	Control	1745	238	94 lost to follow‐up, 69 withdrawn, 75 died
Furuya 2015	Intervention	34	4	4 unable to contact by telephone at follow‐up (90 participants were originally randomised (45 in each group), but 24 participants were excluded immediately after randomisation as they were indicated for surgery or enrolled in another study)
Furuya 2015	Control	32	2	2 did not return for 6 month follow‐up
Hanssen 2007	Intervention	156	55	40 withdrew, 7 died, 8 missing data
Hanssen 2007	Control	132	38	21 withdrew, 7 died, 10 missing data
Jorstad 2013	Intervention	375	23	9 did not receive intervention, 3 died, 2 had early discontinuation of intervention, 9 had incomplete data
Jorstad 2013	Control	379	35	12 were excluded from the study, 10 died, 1 lost to follow‐up, 7 didn't attend 12 month follow‐up, 5 had incomplete data
Lie 2009	Intervention	101	8	6 withdrew, 2 medical exclusions
Lie 2009	Control	102	10	5 withdrew, 5 medical exclusions
Melamed 2014	Intervention	202	21	"patients were exclude (for example, because of missed training appointments)"
Melamed 2014	Control	205	19	"patients were excluded (for example, because of missed training appointments)"
Mooney 2014	Intervention	972	35	14 withdrew, 21 died
Mooney 2014	Control	972	27	10 withdrew, 17 died
Moreno‐Palanco 2011	Intervention	121	3	3 lost to follow‐up, 0 died
Moreno‐Palanco 2011	Control	126	5	5 lost to follow‐up, 0 died
P.RE.COR Group 1991	Intervention	60	0	Counseling program without exercise
	Intervention	61	0	Comprehensive cardiac rehabilitation
	Control	61	0	Usual care
Park 2013	Intervention	31	3	3 withdrew, 0 died
Park 2013	Control	32	2	2 withdrew, 0 died
Southard 2003	Intervention	53	4	Reasons for drop‐out stated: relocation, dietary intervention instead, psychiatric diagnosis, loss of interest
Southard 2003	Control	51	0
Tingström 2005	Intervention	104	3	7 lost to follow‐up: 2 died, 5 did not attend
Tingström 2005	Control	103	4	7 lost to follow‐up: 2 died, 5 did not attend
Combined results	Intervention	5692	641	11.3%
Combined results	Control	5341	615	11.5%
* All causes of drop out from follow0up included (including mortality)

Table 3. All‐cause withdrawal or drop‐out at follow‐up

Table 4. Summary of HRQoL scores at follow‐up: Clark 1997

Sickness Impact Profile+++ at 12 months
	Absolute mean outcome values at follow‐up++			Comparison
	Education	Comparator	Between group P value	Comparison
Clark 1997(12 months)
Total score	7.26	8.09	NS	Education = comparator
Psychosocial dimension	5.52	7.05	≤ 0.05	Education > comparator
Physical dimension	5.89	6.00	NS	Education = comparator
Sickness Impact Profile+++ at 18 months
	Absolute mean outcome values at follow‐up++			Comparison
	Education	Comparator	Between group P value	Comparison
Total score	7.93	7.41	NS	Education = comparator
Psychosocial dimension	6.05	6.23	NS	Education = comparator
Physical dimension	6.40	5.25	NS	Education = comparator
++ for mean scores at follow‐up (adjusted for baseline scores) +++ lower score higher HRQoL NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 4. Summary of HRQoL scores at follow‐up: Clark 1997

Table 5. Summary of HRQoL scores at follow‐up: Clark 2000

Sickness Impact Profile at 12 months
Clark 2000(12 months)	Absolute means at follow‐up++			Comparison
Clark 2000(12 months)	Education	Comparator	Between group P value	Comparison
Psychosocial dimension	5.15	5.91	0.144	Education = comparator
Physical dimension	7.09	7.66	0.05	Education > comparator
Means were adjusted to take account of baseline values NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 5. Summary of HRQoL scores at follow‐up: Clark 2000

Table 6. Summary of HRQoL scores at follow‐up: Clark 2009

Sickness Impact Profile at 12 months
	Absolute means (SD) at follow‐up				Comparison
	Education	Education self directed	Comparator	Between group P value	Comparison
Total score	8.13 (8.63)	9.79 (10.17)	9.49 (9.46)	NS	Education = comparator
Psychosocial dimension	5.84 (8.02)	7.31 (10.74)	6.75 (9.39)	NS	Education = comparator
Physical dimension	8.07 (9.63)	9.46 (10.11)	9.85 (10.79)	NS	Education = comparator
Sickness Impact Profile at 18 months
Total score	8.44 (9.13)	8.98 (10.29)	9.64 (9.45)	NS	Education = comparator
Psychosocial dimension	5.74 (9.68)	6.16 (8.20)	7.17 (10.40)	NS	Education = comparator
Physical dimension	8.27 (10.02)	8.98 (9.33)	9.65 (10.19)	NS	Education = comparator
Note: analysis of these data was reported, but the individual results were not. These were obtained by contacting the author directly NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 6. Summary of HRQoL scores at follow‐up: Clark 2009

Table 7. Summary of HRQoL scores at follow‐up: Cohen 2014

SF‐12 (Short Form 12 item survey) at 6 months
	Mean (SD) outcome values at follow‐up		Between group P value	Comparison
	Education	Comparator	Between group P value	Comparison
Mental component summary	47.5 (11.2)	47.7 (11.2)	0.43	Education = comparator
Physical component summary	47.5 (9.3)	47.3 (9.4)	0.44	Education = comparator
Negative baseline‐follow‐up difference favours intervention and positive favours control NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 7. Summary of HRQoL scores at follow‐up: Cohen 2014

Table 8. Summary of HRQoL scores at follow‐up: Cupples 1994

Nottingham Health Profile+ at 24 months
	MD (95% CI) between groups in change from baseline at follow‐up	Between group P value	Comparison
Emotional reaction	0.0 (‐5.2 to 5.2)	NS	Education = comparator
Energy	0.5 (‐7.8 to 8.8)	NS	Education = comparator
Physical mobility	‐0.4 (‐5.2 to 4.5)	NS	Education = comparator
Pain	0.5 (‐4.7 to 5.6)	NS	Education = comparator
Sleep	3.0 (‐4.0 to 9.9)	NS	Education = comparator
Social isolation	‐2.2 (‐6.6 to 2.1)	P < 0.05	Education > comparator
Nottingham Health Profile+ at 60 months
	MD (95% CI) between groups in change from baseline at follow‐up	Between group P value	Comparison
Emotional reaction	‐2.1 (‐7.5 to 3.3)	NS	Education = comparator
Energy	‐4.7 (‐13.2 to 3.7)	NS	Education = comparator
Physical mobility	‐1.3 (‐6.3 to 3.6)	< 0.05	Education > comparator
Pain	‐3.4 (‐9.2 to 2.3)	< 0.05	Education > comparator
Sleep	‐2.4 (‐9.3 to 4.5)	NS	Education = comparator
Social isolation	0.0 (‐4.3 to 4.3)	NS	Education = comparator
+ Higher scores reflect poorer quality of life The value quoted is the mean difference (MD) (CI) between groups from baseline to follow‐up P related to t‐tests (two tailed) NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 8. Summary of HRQoL scores at follow‐up: Cupples 1994

Table 9. Summary of HRQoL scores at follow‐up: Cupples 1994

Participant' self assessment of quality of life on a five‐point scale at 24 months
	Initial scores (% participants)		Follow‐up scores (% participants)		Between group P value	Comparison
	Education	Comparator	Education	Comparator	Between group P value	Comparison
					P < 0.03	Education > comparator
Poor	6.3	5.3	6.9	8.3
Fair	27.8	23.3	18.9	21.7
Average	35	39	33.1	33.7
Good	22.7	22.7	29.3	25.3
Very good	8.2	9.7	11.7	11
Note: the between group P value represents the overall "comparison of change in individuals' assessment for intervention and control groups" the significant difference being in favour of the intervention group NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 9. Summary of HRQoL scores at follow‐up: Cupples 1994

Table 10. Summary of HRQoL scores at follow‐up: Furuya 2014

*SF‐12 (Short Form 12 item survey) at 6 months**
	Mean (SD) outcome values at follow‐up		Between group P value	Comparison
	Education	Comparator	Between group P value	Comparison
Mental component summary	51.7 (9.5)	48.4 (9.2)	0.73	Education = comparator
Physical component summary	43.3 (10.6)	41.0 (11.0)	0.28	Education = comparator
*SF‐36 (Short Form 36 item survey)**
	Mean (SD) outcome values at follow‐up		Between group P value	Comparison
	Education	Comparator	Between group P value	Comparison
Social functioning	79.2 (25.1)	64.2 (28.4)	0.1	Education = comparator
Mental health	70.9 (22.7)	70.1 (19.1)	0.98	Education = comparator
Physical functioning	72.5 (23.9)	64.5 (27.8)	0.2	Education = comparator
General health	66.1 (19.8)	63.9 (20.0)	0.61	Education = comparator
Vitality	69.7 (20.6)	62.5 (20.7)	0.52	Education = comparator
Bodily pain	63.8 (28.5)	55.7 (24.2)	0.22	Education = comparator
Role–emotional	77.8 (36.4)	64.4 (36.0)	0.72	Education = comparator
Role–physical	52.5 (40.7)	50.0 (44.0)	0.96	Education = comparator
*Negative baseline ‐ follow‐up difference favours intervention; positive difference favours control NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 10. Summary of HRQoL scores at follow‐up: Furuya 2014

Table 11. Summary of HRQoL scores at follow‐up: Hanssen 2007

*SF‐36 (Short Form 36 item survey) at 6 months**
	Between group difference in mean change from baseline (95% CI) at follow‐up	Between group P value	Comparison
Overall physical	‐2.33 (‐4.54 to ‐0.12)	0.039	Education = comparator
Physical functioning	‐1.16 (‐3.28 to 0.95)	0.28	Education = comparator
Role physical	‐1.84 (‐5.32 to 1.64)	0.299	Education = comparator
Bodily pain	‐1.74 (‐4.54 to 1.05)	0.22	Education = comparator
General health	‐0.36 (‐2.64 to 1.91)	0.752	Education = comparator
Overall mental	1.07 (‐1.71 to 3.86)	0.447	Education = comparator
Vitality	‐0.07 (‐2.23 to 2.10)	0.951	Education = comparator
Social functioning	0.36 (‐2.96 to 3.67)	0.832	Education = comparator
Role‐emotional	0.78 (‐3.29 to to 4.84)	0.706	Education = comparator
Mental health	0.4 (‐1.81 to 2.60)	0.723	Education = comparator
*SF‐36 (Short Form 36 item survey) at 18 months**
	Between group difference in mean change from baseline (95% CI) at follow‐up	Between group P value	Comparison
Overall physical	‐1.44 (‐3.89 to 1.02)	0.25	Education = comparator
Physical functioning	‐0.79 (‐3.06 to 1.48)	0.491	Education = comparator
Role physical	‐0.94 (‐4.76 to 2.88)	0.627	Education = comparator
Bodily pain	‐0.77 (‐4.00 to 2.47)	0.641	Education = comparator
General health	0.25 (‐2.15 to 2.64)	0.838	Education = comparator
Overall mental	1.65 (‐1.35 to 4.65)	0.28	Education = comparator
Vitality	0.58 (‐1.95 to 3.12)	0.65	Education = comparator
Social functioning	0.55 (‐3.95 to 2.85)	0.751	Education = comparator
Role‐emotional	2.59 (‐1.58 to 6.77)	0.221	Education = comparator
Mental health	0.31 (‐2.11 to 2.73)	0.8	Education = comparator
*Negative baseline ‐ follow‐up difference favours intervention; positive difference favours control NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 11. Summary of HRQoL scores at follow‐up: Hanssen 2007

Table 12. Summary of HRQoL scores at follow‐up: Lie 2009

Seattle Angina Questionnaire at 6 months
	Absolute mean (SD) outcome values at follow‐up				Comparison
	Education	P value	Comparator	P value	Comparison
Physical limitation	86.4 (15.6)	P < 0.001	83.2 (18.7)	P < 0.001	Education = comparator
Angina frequency	91.7 (16.6)	P < 0.001	90.8 (18.9)	P < 0.001	Education = comparator
Treatment satisfaction	89.2 (15.4)	NS	88.0 (16.1)	NS	Education = comparator
Disease perception	77.8 (20.2)	P < 0.001	73.9 (24.2)	P < 0.001	Education = comparator
SF‐36 (Short Form 36 item survey) at 6 months
	Absolute mean (SD) outcome values at follow‐up				Comparison
	Education	P value	Comparator	P value	Comparison
Overall physical	47.4 (9.6)	P < 0.001	47 (10)	P < 0.001	Education = comparator
Physical functioning	82.2 (19.2)	P < 0.001	82.3 (19.8)	P < 0.001	Education = comparator
Role physical	64 (41.2)	P < 0.001	57.2 (43.3)	P < 0.001	Education = comparator
Bodily pain	77.2 (22.3)	P < 0.001	78.5 (25.2)	P < 0.001	Education = comparator
General health	69.9 (23.3)	NS	65.7 (27.2)	NS	Education = comparator
Overall mental	52.1 (10.7)	P < 0.05	50.5 (10.8)	NS	Favours education
Vitality	61.9 (23.9)	P < 0.001	60.5 (21.6)	P < 0.001	Education = comparator
Social functioning	86.3 (21.4)	P < 0.001	84.3 (21.9)	P < 0.001	Education = comparator
Role‐ emotional	73.3 (38.2)	P < 0.01	67.4 (41.6)	P < 0.01	Education = comparator
Mental health	81.9 (17.3)	P < 0.001	78.5 (21)	P < 0.01	Education = comparator
Higher scores indicate better HRQoL NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 12. Summary of HRQoL scores at follow‐up: Lie 2009

Table 13. Summary of HRQoL scores at follow‐up: Lisspers 1999

Angina Pectoris ‐ Quality of Life Questionnaire (AP‐QLQ) at 24 months
	Mean (SD) score at follow‐up		Between group P value	Comparison
	Education	Comparator	Between group P value	Comparison
QLQ (total)	4.7 (0.8)	4.3 (1.0)	NS	Education = comparator
Somatic symptoms	4.8 (1.0)	4.3 (1.1)	NS	Education = comparator
Physical activity	4.8 (1.0)	4.1 (1.2)	NS	Education = comparator
Emotional distress	4.8 (0.8)	4.6 (1.1)	NS	Education = comparator
Life satisfaction	4.2 (1.0)	3.9 (1.2)	NS	Education = comparator
Figures quoted represent absolute scores on a self‐rating scale NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 13. Summary of HRQoL scores at follow‐up: Lisspers 1999

Table 14. Summary of HRQoL scores at follow‐up: Melamed 2014

MacNew Heart Disease Quality of Life Questionnaire (MacNew) at 220 days
	Mean (SD) outcome values at follow‐up		Between group P value	Comparison
	Education	Comparator	Between group P value	Comparison
Rank sum	5.75 (0.87)	5.74 (0.83)	0.056	Education = comparator
NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 14. Summary of HRQoL scores at follow‐up: Melamed 2014

Table 15. Summary of HRQoL scores at follow‐up: Park 2013

Seattle Angina Questionnaire‐Korean (SAQ‐K) at 6 months
	Mean (SD) outcome values at follow‐up		Between group P value	Comparison
	Education	Comparator	Between group P value	Comparison
Physical symptoms
Physical limitation	90.77 (9.97)	85.74 (15.37)	0.901	Education = comparator
Angina stability	78.57 (20.09)	64.17 (23.38)	0.037	Education > comparator
Angina frequency	94.29 (7.90)	89.33 (14.84)	0.543	Education = comparator
Treatment satisfaction	86.38 (12.15)	73.13 (16.09)	0.021	Education > comparator
Diseases perception	74.40 (16.03)	52.78 (15.98)	0.005	Education > comparator
Higher scores indicate better HRQoL NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 15. Summary of HRQoL scores at follow‐up: Park 2013

Table 16. Summary of HRQoL scores at follow‐up: Pogosova 2008

SF‐36 (Short Form 36 item survey) at 12 months
	Mean change from baseline P value		Comparison
	Education	Comparator	Comparison
Overall physical	P > 0.05	P ≤ 0.05	Favours education
Physical functioning	P > 0.05	P ≤ 0.05	Favours education
Bodily pain	P > 0.05	P ≤ 0.05	Favours education
Overall mental	P > 0.05	P ≤ 0.05	Favours education
Vitality	P > 0.05	P ≤ 0.05	Favours education
Social functioning	P > 0.05	P ≤ 0.05	Favours education
Mental health	P > 0.05	P ≤ 0.05	Favours education
There were no significant changes demonstrated in the control group but no statistical comparison of the mean change between groups was reported NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 16. Summary of HRQoL scores at follow‐up: Pogosova 2008

Table 17. Summary of HRQoL scores at follow‐up: Tingström 2005

*SF‐36 (Short Form 36 item survey) at 12 months**
	Mean change from baseline (SD)		Between group P value⁺	Comparison
	Education	Comparator	Between group P value⁺	Comparison
Physical functioning	3.6 (17.6)	4.4 (15.1)	0.749	Education = comparator
Role physical	38.2 (46.9)	33.8 (42.4)	0.504	Education = comparator
Bodily pain	5.69 (31.1)	6.18 (29.1)	0.911	Education = comparator
General health	1.4 (15.9)	1.8 (16.3)	0.862	Education = comparator
Vitality	5.3 (22.7)	4.9 (21.8)	0.921	Education = comparator
Social functioning	9.7 (24)	9.1 (25.3)	0.869	Education = comparator
Role emotional	15.8 (48.1)	16.5 (41.1)	0.913	Education = comparator
Mental health	2.9 (16.6)	4.2 (17.8)	0.566	Education = comparator
*Positive values indicate improvement in HRQL from baseline + P values are calculated on the difference between groups at pre‐test and on the mean change (post test minus pre‐test). NS: No significant difference demonstrated Education = Comparator: no significant difference (P > 0.05) in HRQoL between the education and the comparator groups at follow‐up. Education > Control: significant difference (P ≤ 0.05) in HRQoL in favour of the education group at follow‐up. Comparator > Education: significant difference (P ≤ 0.05) in HRQoL in favour of the comparator group at follow‐up. Favours education: Available evidence favours the intervention group but direct statistical comparison between intervention and control groups was not reported. Favours comparator: Available evidence favours the control group but direct statistical comparison between intervention and control groups was not reported.

Table 17. Summary of HRQoL scores at follow‐up: Tingström 2005

Table 18. Cost summary of intervention and comparison of healthcare costs incurred by intervention and control groups during follow‐up period

Variable	Clark 2000	Cupples 1994	Esposito 2008				Southard 2003	Peikes 2009
Follow‐up	24 months	24 months	6 months	7 to 12 months	12 months	18 months	6 months	25 months
Year	2000	NR	2005 to 2006				NR	2002 to 2005
Currency	USD	GBP	USD				USD	USD
Mean cost of cardiac rehabilitation program per patient
Total costs	USD 187	GBP 49.72	USD 162				USD 453	USD 196
Costs considered	Personnel, instructional materials, telephone supplies, ongoing staff training	Direct costs by health visitors (staff time), Travel Costs	Average monthly fee paid to the program per member				Nurse salary Overheads Subscription costs	Average monthly fee paid to the program per member
Comments	Participating site overheads were not measured, a "conservatively high" estimate of these was taken to double the treatment cost to USD 374	Costs of the health visitor also included time spent recording data collection for the study						Cost varied among the 15 included studies. Negotiated locally with Medicare and Medicaid Services. (Range USD 50 to USD 444)
Mean total healthcare costs per patient
Total cost (intervention)	USD ~3300 (calc)	GBP 1801	USD 1627	USD 2356	USD 2288	USD 1793	USD 635	USD 1283^*
Total cost (control)	USD ~6500	GBP 1812	USD 1632	USD 2464	USD 2372	USD 1818	USD 2053	USD 1314^*
Between group difference	USD ~1800*	GBP 9.60	USD 5	USD 107	USD 84	USD 25	USD 1418	USD 144 (80% CI 99 to 188)
P value	NR	NS	0.895	0.077	0.132	0.365	NR	< 0.001
Cost saving per pt (when cost of intervention taken into account§)	USD ~1610 or USD ~1420 if estimated overheads were included	GBP 40	USD ‐157	USD ‐55	USD ‐78	USD ‐137	USD 965	USD ‐52
Additional healthcare costs considered	Number of admissions (heart related), number of inpatient days, In patient cost' emergency department costs	Prescription of drugs, GP visits, visits to hospital as inpatients and outpatients, all tests investigations and treatments carried out	Medicare medical claims				Cardiovascular‐related emergency department visits and hospitalisations
Comments	Expenditure was calculated from differences in % utilisation of hospital services. i.e. hospital charges for participants were on average 49% lower and the average annual expenditure was USD 6500. * There was a calculated saving of a hospital charge of USD 3200, the ratio of payments to charges was 0.56 therefore USD 1800 actual saving	There was a difference in the drug usage at baseline which is not accounted for in these figures although this would make minimal impact to the results. The intervention group were more costly for drugs, procedures and service use	Claims quoted are per member per month					*Expenditure/pt/month enrolled Overall costs were increased by 11% when the care coordination fees were taken into account
Summary difference between groups	Favours Rx	Rx = Control	Rx = control (for all time periods studied)				Favours Rx	Favours control
§ = Negative mean difference indicates a net cost of the intervention group NR = not recorded NS = not significant

Table 18. Cost summary of intervention and comparison of healthcare costs incurred by intervention and control groups during follow‐up period

Table 19. Results of univariate meta‐regression analysis for total mortality

Explanatory variable (n trials)	Exp(slope)*	95% CI univariate; P value	Proportion of variation explained	Interpretation
Case mix (% myocardial infarction patients) (n = 11)	RR = 1.004	0.988 to 1.020 P = 0.631	‐190.36%	No evidence that RR is associated with case mix
Age of participants (n = 13)	RR = 1.005	0.940 to 1.074 P = 0.876	‐28.26%	No evidence that RR is associated with the age of participants
Percentage of male participants (n = 13)	RR = 0.991	0.986 to 1.012 P = 0.882	‐25.27%	No evidence that RR is associated with the percentage of male participants
Type of CR (education only vs. education plus e.g. exercise or psychological intervention) (n = 13)	RR = 0.181	0.014 to 2.321 P = 0.168	28.25%	No evidence that RR is associated with type of CR
Method of structured educational delivery (one‐to‐one vs. group versus combination) (n = 13)	RR = 1.010	0.728 to 1.401 P = 0.948	‐28.28%	No evidence that RR is associated with method of delivery
Duration of intervention (n = 12)	RR = 0.978	0.948 to 1.010 P = 0.152	3.69%	No evidence that RR is associated with duration of intervention
Theoretical vs. no theoretical basis to educational intervention (n = 13)	RR = 1.473	0.750 to 2.895 P = 0.233	‐0.31%	No evidence that RR is associated with theoretical basis
Involvement of significant others (e.g. spouse, family member) in the education programme (n = 13)	RR = 1.245	0.890 to 1.722 P = 0.166	7.06%	No evidence that RR is associated with family involvement
Study location (n = 13)	RR = 1.050	0.714 to 1.543 P = 0.787	‐59.78%	No evidence that risk ratio is associated with study location
Setting (centre vs. home) (n = 13)	RR = 1.171	0.773 to 1.774 P = 0.421	‐44.55%	No evidence that RR is associated with centre status
Length of follow‐up (n = 13)	RR = 0998	0.964 to 1.033 P = 0.924	‐22.64%	No evidence that RR is associated with length of follow‐up
CR ‐ cardiac rehabilitation; RR ‐ risk ratio

Table 19. Results of univariate meta‐regression analysis for total mortality

Table 20. Results of univariate meta‐regression analysis for withdrawal

Explanatory variable (n trials)	Exp(slope)*	95% CI univariate P value	Proportion of variation explained	Interpretation
Case mix (% myocardial infarction patients) (n = 12)	RR = 1.002	0.992 to 1.013 P = 0.611	‐9.25%	No evidence that RR is associated with case mix
Age of participants (n = 17)	RR = 0.998	0.963 to 1.034 P = 0.903	‐15.62%	No evidence that RR is associated with the age of participants
Percentage of male participants (n = 17)	RR = 0.999	0.992 to 1.005 P = 0.621	‐28.64%	No evidence that RR is associated with the percentage of male participants
Type of CR (education only vs. education plus e.g. exercise or psychological intervention) (n = 17)	RR = 0.752	0.260 to 2.174 P = 0.575	‐16.99%	No evidence that RR is associated with type of CR
Method of structured educational delivery (one‐to‐one vs. group versus combination) (n = 17)	RR = 1.033	0.860 to 1.242 P = 0.714	‐29.18%	No evidence that mortality risk is associated with method of delivery
Duration of intervention (n = 16)	RR = 0.993	0.964 to 1.023 P = 0.625	‐25.06%	No evidence that mortality risk is associated with duration of intervention
Theoretical vs. no theoretical basis to educational intervention (n = 17)	RR = 1.031	0.690 to 1.541 P = 0.874	‐24.70%	No evidence that RR is associated with theoretical basis
Involvement of significant others (e.g. spouse, family member) in the education (n = 17)	RR = 1.016	0.829 to 1.245 P = 0.872	‐33.62%	No evidence that RR is associated with family involvement
Study location (n = 17)	RR = 0.942	0.801 to 1.109 P = 0.449	24.47%	No evidence that RR is associated with study location
Setting (centre vs. home) (n = 17)	RR = 1.096	0.873 to 1.374 P = 0.404	‐13.38%	No evidence that RR is associated with centre status
Length of follow‐up (n = 17)	RR = 0.976	0.955 to 0.998 P = 0.035	90.79%	Significant evidence that risk of withdrawal is increased in studies with a shorter follow‐up
CR ‐ cardiac rehabilitation; RR ‐ risk ratio

Table 20. Results of univariate meta‐regression analysis for withdrawal

Table 21. Results of sensitivity analysis for total mortality

Explanatory variable (n trials)	Exp(slope)*	95% CI univariate P value	Proportion of variation explained	Interpretation
Year of publication (n = 13)	RR = 0.998	0.950 to 1.047 P = 0.913	‐61.7%	No evidence that RR is associated with year of publication
Risk of bias (low risk in ≥ 5 items vs. < 5 items) (n = 13)	RR = 1.105	0.421 to 3.831 P = 2.899	‐84.29%	No evidence that RR is associated with risk of bias
RR ‐ risk ratio

Table 21. Results of sensitivity analysis for total mortality

Table 22. Results of sensitivity analysis for withdrawal

Explanatory variable (n trials)	Exp(slope)*	95% CI univariate P value	Proportion of variation explained	Interpretation
Year of publication (before 2000 vs. 2000 or later) (n = 17)	RR = 1.017	0.982 to 1.052 P = 0.327	‐7.02%	No evidence that RR is associated with year of publication
Risk of bias (low risk in ≥ 5 items vs. < 5 items) (n = 17)	RR = 1.437	1.069 to 1.931 P = 0.020	15.35%	Significant evidence that risk of withdrawal is increased in studies with higher risk of bias

Table 22. Results of sensitivity analysis for withdrawal

Comparison 1. Education versus no education

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Total mortality at the end of the follow up period Show forest plot	13	10075	Risk Ratio (M‐H, Random, 95% CI)	0.80 [0.60, 1.05]

1.1.1 Studies with 12 months or less follow‐up	6	4063	Risk Ratio (M‐H, Random, 95% CI)	0.78 [0.35, 1.78]
1.1.2 Studies with more than 12 months follow‐up	7	6012	Risk Ratio (M‐H, Random, 95% CI)	0.78 [0.60, 1.02]
1.2 Fatal and/or non‐fatal MI Show forest plot	2	209	Risk Ratio (M‐H, Random, 95% CI)	0.63 [0.26, 1.48]

1.3 Other fatal and/or non‐fatal cardiovascular events Show forest plot	2	310	Risk Ratio (M‐H, Random, 95% CI)	0.36 [0.23, 0.56]

1.4 Total revascularisations (including CABG and PCI) Show forest plot	3	456	Risk Ratio (M‐H, Random, 95% CI)	0.58 [0.19, 1.71]

1.5 Hospitalisations Show forest plot	5	14849	Risk Ratio (M‐H, Random, 95% CI)	0.93 [0.71, 1.21]

1.6 Withdrawals Show forest plot	17	10972	Risk Ratio (M‐H, Random, 95% CI)	1.04 [0.88, 1.22]

1.6.1 Studies with 12 months or less follow‐up	10	4960	Risk Ratio (M‐H, Random, 95% CI)	1.18 [0.93, 1.49]
1.6.2 Studies with more than 12 months follow‐up	7	6012	Risk Ratio (M‐H, Random, 95% CI)	0.98 [0.80, 1.20]

Comparison 1. Education versus no education