Exercise‐based cardiac rehabilitation for coronary heart disease

Lindsey Anderson; David R Thompson; Neil Oldridge; Ann‐Dorthe Zwisler; Karen Rees; Nicole Martin; Rod S Taylor

doi:10.1002/14651858.CD001800.pub3

Exercise‐based cardiac rehabilitation for coronary heart disease

Authors' declarations of interest

Version published: 05 January 2016 Version history

https://doi.org/10.1002/14651858.CD001800.pub3

Collapse all Expand all

Abstract

available in

Background

Coronary heart disease (CHD) is the single most common cause of death globally. However, with falling CHD mortality rates, an increasing number of people live with CHD and may need support to manage their symptoms and prognosis. Exercise‐based cardiac rehabilitation (CR) aims to improve the health and outcomes of people with CHD. This is an update of a Cochrane systematic review previously published in 2011.

Objectives

To assess the effectiveness and cost‐effectiveness of exercise‐based CR (exercise training alone or in combination with psychosocial or educational interventions) compared with usual care on mortality, morbidity and HRQL in patients with CHD.

To explore the potential study level predictors of the effectiveness of exercise‐based CR in patients with CHD.

Search methods

We updated searches from the previous Cochrane review, by searching Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, Issue 6, 2014) from December 2009 to July 2014. We also searched MEDLINE (Ovid), EMBASE (Ovid), CINAHL (EBSCO) and Science Citation Index Expanded (December 2009 to July 2014).

Selection criteria

We included randomised controlled trials (RCTs) of exercise‐based interventions with at least six months’ follow‐up, compared with a no exercise control. The study population comprised men and women of all ages who have had a myocardial infarction (MI), coronary artery bypass graft (CABG) or percutaneous coronary intervention (PCI), or who have angina pectoris, or coronary artery disease. We included RCTs that reported at least one of the following outcomes: mortality, MI, revascularisations, hospitalisations, health‐related quality of life (HRQL), or costs.

Data collection and analysis

Two review authors independently screened all identified references for inclusion based on the above inclusion and exclusion criteria. One author extracted data from the included trials and assessed their risk of bias; a second review author checked data. We stratified meta‐analysis by the duration of follow up of trials, i.e. short‐term: 6 to 12 months, medium‐term: 13 to 36 months, and long‐term: > 3 years.

Main results

This review included 63 trials which randomised 14,486 people with CHD. This latest update identified 16 new trials (3872 participants). The population included predominantly post‐MI and post‐revascularisation patients and the mean age of patients within the trials ranged from 47.5 to 71.0 years. Women accounted for fewer than 15% of the patients recruited. Overall trial reporting was poor, although there was evidence of an improvement in quality of reporting in more recent trials.

As we found no significant difference in the impact of exercise‐based CR on clinical outcomes across follow‐up, we focused on reporting findings pooled across all trials at their longest follow‐up (median 12 months). Exercise‐based CR reduced cardiovascular mortality compared with no exercise control (27 trials; risk ratio (RR) 0.74, 95% CI 0.64 to 0.86). There was no reduction in total mortality with CR (47 trials, RR 0.96, 95% CI 0.88 to 1.04). The overall risk of hospital admissions was reduced with CR (15 trials; RR 0.82, 95% CI 0.70 to 0.96) but there was no significant impact on the risk of MI (36 trials; RR 0.90, 95% CI 0.79 to 1.04), CABG (29 trials; RR 0.96, 95% CI 0.80 to 1.16) or PCI (18 trials; RR 0.85, 95% CI 0.70 to 1.04).

There was little evidence of statistical heterogeneity across trials for all event outcomes, and there was evidence of small study bias for MI and hospitalisation, but no other outcome. Predictors of clinical outcomes were examined across the longest follow‐up of studies using univariate meta‐regression. Results show that benefits in outcomes were independent of participants' CHD case mix (proportion of patients with MI), type of CR (exercise only vs comprehensive rehabilitation) dose of exercise, length of follow‐up, trial publication date, setting (centre vs home‐based), study location (continent), sample size or risk of bias.

Given the heterogeneity in outcome measures and reporting methods, meta‐analysis was not undertaken for HRQL. In five out of 20 trials reporting HRQL using validated measures, there was evidence of significant improvement in most or all of the sub‐scales with exercise‐based CR compared to control at follow‐up. Four trial‐based economic evaluation studies indicated exercise‐based CR to be a potentially cost‐effective use of resources in terms of gain in quality‐adjusted life years.

The quality of the evidence for outcomes reported in the review was rated using the GRADE method. The quality of the evidence varied widely by outcome and ranged from low to moderate.

Authors' conclusions

This updated Cochrane review supports the conclusions of the previous version of this review that, compared with no exercise control, exercise‐based CR reduces the risk of cardiovascular mortality but not total mortality. We saw a significant reduction in the risk of hospitalisation with CR but not in the risk of MI or revascularisation. We identified further evidence supporting improved HRQL with exercise‐based CR. More recent trials were more likely to be well reported and include older and female patients. However, the population studied in this review still consists predominantly of lower risk individuals following MI or revascularisation. Further well conducted RCTs are needed to assess the impact of exercise‐based CR in higher risk CHD groups and also those presenting with stable angina. These trials should include validated HRQL outcome measures, explicitly report clinical event outcomes including mortality and hospital admissions, and assess costs and cost‐effectiveness.

PICOs

Population

Intervention

Comparison

Outcome

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

See more on using PICO in the Cochrane Handbook.

Plain language summary

available in

Exercise‐based rehabilitation for coronary heart disease

Background

Coronary heart disease (CHD) is the single most common cause of death globally. However, with falling CHD mortality rates, an increasing number of people live with CHD and may need support to manage their symptoms and reduce the chances of future problems such as heart attacks. Exercise‐based cardiac rehabilitation aims to improve the health and outcome of people with CHD.

Study characteristics

We searched the scientific literature for randomised controlled trials (experiments that randomly allocate participants to one of two or more treatment groups) looking at the effectiveness of exercise‐based treatments compared with no exercise in people of all ages with CHD. The search is current to July 2014.

Key results
This latest update identified 16 trials (3872 participants). We included a total of 63 trials that studied 14,486 people with CHD, predominantly heart attack survivors and those who had undergone heart bypass surgery or angioplasty (a procedure which widens narrowed or obstructed arteries or veins). The findings of this update are consistent with the previous (2011) version of this Cochrane review and show important benefits of exercise‐based cardiac rehabilitation that include a reduction in the risk of death due to a cardiovascular cause and hospital admission and improvements in health‐related quality of life, compared with not undertaking exercise. There was a considerable variation across studies in the reporting of health‐related quality of life outcome. A small body of economic evidence was identified indicating exercise‐based cardiac rehabilitation to be cost‐effective. Further evidence is needed to understand the effect of exercise training in people with CHD who are higher risk and in those with established angina (chest pain).

Quality of evidence
Although the reporting of methods has improved in recent trials, lack of reporting made it difficult to assess the overall methodological quality and risk of possible bias of the evidence.

Authors' conclusions

Implications for practice

This review shows that while exercise‐based CR does not reduce total mortality, it does provide important benefits by reducing cardiovascular mortality and hospitalisation (and associated healthcare costs), and improving HRQL in younger men who have suffered MI or are post‐revascularisation. While there was an increase in the proportion of female and older individuals in more recent trials, the application of this evidence base to a more poorly represented group, particularly angina pectoris and higher risk CHD patients, and those with major co‐morbidities, remains a question of clinical judgement. There appears to be little to choose between exercise‐only or exercise in combination with psychosocial or educational CR interventions. In the absence of definitive cost‐effectiveness comparing psychosocial or educational approaches to exercise‐based CR, it would be rational to use cost considerations to determine practice.

Implications for research

In spite of incorporation of recent trial evidence including more older and female patients, the population of CHD patients studied in this review update remains predominately low risk middle‐aged males following MI or revascularisation. Therefore, well‐designed, and adequately reported RCTs of CR in groups of CHD patients more representative of usual clinical practice are still needed. These trials should include validated HRQL outcome measures, need to explicitly report clinical events including mortality and hospital admission, and assess costs and cost‐effectiveness. Furthermore, further details of the presentation and diagnoses of CHD participants and interventions offered and received should be reported in trials, so that results of future reviews can better stratify outcomes according to the range of CHD populations or types of CR interventions.

Summary of findings

Open in table viewer

Summary of findings for the main comparison. Exercise‐based cardiac rehabilitation for coronary heart disease

Exercise‐based cardiac rehabilitation for coronary heart disease
Patient or population: Patients with coronary heart disease Intervention: Exercise‐based cardiac rehabilitation
Outcomes	No of Participants (Number of studies)	Number of Events / Participants		Risk Ratio (95% CI)	Statistical Heterogeneity I² statistic Chi²‐test (P value)	GRADE Quality of the evidence
		Intervention	Comparator
Total mortality (All Studies)	12455 (47)	838/6424	865/6031	RR 0.96 [0.88 to 1.04]	0% (0.58)	⊕⊕⊕⊝ moderate¹
Follow‐up of 6 to 12 months	8800 (29)	226/4573	238/4227	0.88 [0.73, 1.05]	0% (0.82)
Follow‐up of > 12 to 36 months	6823 (13)	338/3495	417/3328	0.89 [0.78, 1.01]	0% (0.47)
Follow‐up longer than 3 years	3828 (11)	476/1902	493/1926	0.91 [0.75, 1.10]	35% (0.12)
CV mortality (All Studies)	7469 (27)	292/3850	375/3619	RR 0.74 [0.64 to 0.86]	0% (0.70)	⊕⊕⊕⊝ moderate¹
Follow‐up of 6 to 12 months	4884 (15)	105/2561	107/2323	0.90 [0.69, 1.17]	0% (0.72)
Follow‐up of > 12 to 36 months	3833 (7)	199/1971	239/1862	0.77 [0.63, 0.93]	5% (0.38)
Follow‐up longer than 3 years	1392 (8)	56/690	100/702	0.58 [0.43, 0.78]	0% (0.91)
Fatal and/or non‐fatal MI (All Studies)	9717 (36)	356/4951	387/4766	RR 0.90 [0.79 to 1.04]	0% (0.48)	⊕⊕⊝⊝ low^{1, 2}
Follow‐up of 6 to 12 months	6911 (20)	126/3543	139/3368	0.85 [0.67, 1.08]	0% (0.58)
Follow‐up of > 12 to 36 months	5644 (11)	251/2877	222/2767	1.09 [0.91, 1.29]	0% (0.72)
Follow‐up longer than 3 years	1560 (10)	65/776	102/784	0.67 [0.50, 0.90]	0% (0.67)
CABG (All Studies)	5891 (29)	208/3021	212/2870	RR 0.96 [0.80 to 1.16]	0% (0.86)	⊕⊕⊕⊝ moderate¹
Follow‐up of 6 to 12 months	4563 (21)	123/2351	121/2212	0.99 [0.77, 1.26]	0% (0.83)
Follow‐up of > 12 to 36 months	2755 (8)	122/1379	123/1376	0.98 [0.78, 1.25]	0% (0.93)
Follow‐up longer than 3 years	675 (4)	19/333	29/342	0.66 [0.34, 1.27]	18% (0.30)
PCI (All Studies)	4012 (18)	171/2013	197/1999	RR 0.85 [0.70 to 1.04]	0% (0.59)	⊕⊕⊕⊝ moderate¹
Follow‐up of 6 to 12 months	3564 (13)	90/1778	99/1786	0.92 [0.64, 1.33]	16% (0.30)
Follow‐up of > 12 to 36 months	1983 (6)	114/996	116/987	0.96 [0.69, 1.35]	26% (0.24)
Follow‐up longer than 3 years	567 (3)	28/281	37/286	0.76 [0.48, 1.20]	0% (0.81)
Hospital admissions (All Studies)	3030 (15)	407/1556	453/1474	RR 0.82 [0.70 to 0.96]	34.5% (0.10)	⊕⊕⊝⊝ low^{1, 2}
Follow‐up of 6 to 12 months	1120 (9)	82/574	116/546	0.65 [0.46, 0.92]	37% (0.14)
Follow‐up of > 12 to 36 months	1916 (6)	322/984	330/932	0.95 [0.84, 1.07]	0% (0.50)
Follow‐up longer than 3 years	0 (0)	0/0	0/0	Not estimable	Not estimable
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
¹ Random sequence generation, allocation concealment or blinding of outcome assessors were poorly described in over 50% of included studies; bias likely, therefore quality of evidence downgraded by one level. ² Funnel Plots and / or Egger test suggest evidence of asymmetry, therefore quality of evidence downgraded by one level.

Background

Description of the condition

Coronary heart disease (CHD) is the single most common cause of death globally, with 7.4 million deaths in 2013, accounting for one‐third of all deaths (WHO 2014). In the United Kingdom (UK), an estimated 2.3 million people live with CHD and the condition accounts for one in five deaths in men and one in ten deaths in women (Nichols 2012; Townsend 2012). Although the mortality rate from CHD has been falling in the UK, primarily due to evidence‐based treatments and reductions in major risk factors, principally smoking (Unal 2004), it has fallen more slowly in those aged less than 55 years, and less than in many other developed countries (Nichols 2012; Townsend 2012). With falling CHD mortality rates, an increasing number of people live with CHD and may need support to manage their symptoms and prognosis.

Description of the intervention

Many definitions of cardiac rehabilitation (CR) have been proposed. The following definition encompasses the key concepts of CR: “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. Indeed, based on evidence from previous meta‐analyses (Clark 2005; Piepoli 2004; Taylor 2004), CR following a cardiac event is a Class I recommendation from the European Society of Cardiology, the American Heart Association and American College of Cardiology, with exercise therapy consistently identified as a central element (Balady 2011; Perk 2012; Smith 2011). However, despite the recommendations for exercise‐based CR as an integral component of comprehensive cardiac care of patients with CHD (particularly those following myocardial infarction (MI), revascularisation or with angina pectoris) and heart failure, a substantial proportion of patients do not receive it (Bethell 2008). Service provision, though predominantly hospital‐based, varies markedly, and referral, enrolment and completion are sub‐optimal, especially among women and older people (Beswick 2004, Clark 2012). Home‐based CR programmes have been increasingly introduced to widen access and participation (Taylor 2010), and interventions aimed at improving patient uptake and adherence to CR programmes have been adopted (Karmali 2014).

Exercise‐based CR is remarkably safe. An observational study of more than 25,000 patients undergoing CR reported one cardiac event for 50,000 hours of exercise training, equivalent to 1.3 cardiac arrests per million patient‐hours (Pavy 2006). An earlier study reported one case of ventricular fibrillation per 111,996 patient‐hours of exercise and one MI per 294,118 patient‐hours (Van Camp 1986). Patients with unstable angina, uncontrolled ventricular arrhythmia, and severe heart failure (New York Heart Association level 4) have been considered at high risk, with careful assessment recommended before they engage in the exercise component of CR. (BACPR 2012).

How the intervention might work

Exercise training has been shown to have direct benefits on the heart and coronary vasculature, including myocardial oxygen demand, endothelial function, autonomic tone, coagulation and clotting factors, inflammatory markers, and the development of coronary collateral vessels (Clausen 1976; Hambrecht 2000). However, findings of the original Cochrane review of exercise‐based CR for CHD (Jolliffe 2001) supported the hypothesis that reductions in mortality may also be mediated via the indirect effects of exercise through improvements in the risk factors for atherosclerotic disease (i.e. lipids, smoking and blood pressure) (Taylor 2006).

Why it is important to do this review

This is an update of a Cochrane review published in 2011 which identified 47 randomised controlled trials (RCTs) randomising a total of 10,794 patients (Heran 2011). A reduction in overall and cardiovascular mortality (risk ratio (RR): 0.87, 95% CI 0.75 to 0.99 and RR 0.74, 95% CI 0.63 to 0.87) and hospital admissions (RR 0.69,95% CI 0.51 to 0.93) in the shorter term (trials with follow up ≤12 months follow‐up) was reported with no evidence of heterogeneity of effect across trials. Exercise‐based CR was not found to reduce the risk of morbidity in terms of the risk of recurrent myocardial infarction or risk of revascularisation. Given both the heterogeneity in outcome measure and methods of reporting findings, a meta‐analysis was not undertaken for health‐related quality of life (HRQL) outcomes, although there was evidence of a higher level of HRQL with exercise‐based CR than usual care in the seven (out of 10) trials reporting validated HRQL outcome measures.

The 2011 review identified a number of limitations in the available RCT evidence, the most notable of which are listed below.

Under‐representation of women, elderly people, and other cardiac groups (post revascularisation and angina pectoris).
Poor reporting of methodology and results in many trial publications. The method of randomisation, allocation, concealment, or blinding of outcomes assessment was rarely described. Furthermore, incomplete outcome data (primarily due to losses to follow‐up or dropouts) were insufficiently addressed in most trials. Losses to follow‐up were relatively high across trials (approximately one third of trials reported a greater than 20% loss to follow‐up) but reasons for dropout were often not reported.
Several trials excluded significant numbers of patients post‐randomisation, and thus in an intention‐to‐treat analysis, these patients were regarded as dropouts. This may be partly explained by the fact that the majority of trials were not designed (or powered) to assess treatment group differences in mortality and morbidity but instead surrogate measures of treatment efficacy, such as exercise capacity or cardiac risk factor levels.
Lack of robust evidence for the impact on HRQL, costs and cost‐effectiveness.

The 2011 review authors concluded that well designed and adequately reported RCTs in groups of CHD patients more representative of usual clinical practice are needed. It was also recommended that these trials should include validated HRQL outcome measures, explicitly report clinical events including hospital admission, and assess costs and cost‐effectiveness.

Using additional RCT evidence published since the 2011 Cochrane review, the aim of this update was to reassess the effectiveness of exercise‐based CR compared to usual care on mortality, risk of hospital admissions, myocardial infarction, revascularisation, HRQL, and costs and cost‐effectiveness in patients with CHD.

Changes in this update review

In addition to updating the searches, given the increased number of RCTs reporting longer follow up, this update review has stratified the results of meta‐analyses according to time of follow‐up: short‐term, 6‐12 months; medium‐term,13‐36 months; and long‐term, > 36 months (follow‐up is likely to be a key driver of intervention effect), and has assessed the quality of the evidence for reported outcomes using the GRADE framework (Schünemann 2011).

Objectives

To assess the effectiveness and cost‐effectiveness of exercise‐based CR (exercise training alone or in combination with psychosocial or educational interventions) compared with usual care on mortality, morbidity and HRQL in patients with CHD.
To explore the potential study level predictors of the effectiveness of exercise‐based CR in patients with CHD.

Methods

Criteria for considering studies for this review

Types of studies

We sought RCTs of exercise‐based CR versus usual care with a follow‐up period of at least six months.

Types of participants

We included men and women of all ages, in both hospital‐based and community‐based settings, who have had a MI, or who had undergone revascularisation (coronary artery bypass grafting (CABG), percutaneous coronary intervention (PCI)) or who have angina pectoris or coronary artery disease defined by angiography.

We excluded studies which only included participants following heart valve surgery, with heart failure, with atrial fibrillation, with heart transplants, or implanted with either cardiac‐resynchronisation therapy (CRT) or implantable defibrillators (ICD). These indications are the subject of other Cochrane reviews (Risom 2014; Sibilitz 2014; Taylor 2014). We also excluded studies of participants who had completed a CR programme prior to randomisation.

Types of interventions

Exercise‐based CR is defined as a supervised or unsupervised inpatient, outpatient, community‐ or home‐based intervention which includes some form of exercise training that is applied to a cardiac patient population. The intervention could be exercise training alone or exercise training in addition to psychosocial or educational interventions, or both (i.e. "comprehensive CR").

Usual care could include standard medical care, such as drug therapy, but without any form of structured exercise training or advice.

Types of outcome measures

We included studies which reported one or more of the following outcomes:

Primary outcomes

Mortality
- Total
- Cardiovascular

MI
- Fatal MI
- Non‐fatal MI

Revascularisations
- CABG
- PCI

Hospitalisations

Secondary outcomes

Health‐related quality of life assessed using validated instruments (e.g. SF‐36, EQ‐5D)

Costs and cost‐effectiveness

Search methods for identification of studies

The search from the previously published Cochrane review (Heran 2011) was updated by searching the Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library (Issue 6, 2014), Database of Abstracts of Reviews of Effects (DARE) (Issue 2, 2014), Health Technology Assessment (HTA) (Issue 2, 2014), MEDLINE & Medline in Process (OVID) (1946 to 2nd July 2014), EMBASE (OVID) (1980 to week 26, 2014) and CINAHL Plus (EBSCO) (1937 to 3 July 2014). Conference proceedings were searched on Science on Web of Science Core Collection (Thomson Reuters) (1970 to June 2014). We hand‐searched reference lists of retrieved articles and systematic reviews published since the last update, for any studies not identified by the electronic searches. We searched trial registers (WHO's ICTRP and Clinicaltrials.gov) for on‐going clinical trials and also sought expert advice.

We designed search strategies with reference to those of the previous systematic review (Heran 2011). We added new search terms to expand the search to include percutaneous coronary intervention (PCI) and related interventions, and also angina‐related conditions such as acute coronary syndrome (ACS). We also added terms relating to education and psychological interventions to better reflect the comprehensive nature of CR. We searched MEDLINE, EMBASE and CINAHL using a strategy combining selected MeSH terms and free text terms relating to exercise‐based rehabilitation and coronary heart disease with filters applied to limit to humans and RCTs.The RCT filter for MEDLINE was the Cochrane sensitivity‐maximising RCT filter, and for EMBASE, terms recommended in the Cochrane Handbook were applied (Lefebvre 2011). Adaptations of this filter were applied to CINAHL and Web of Science. We translated the MEDLINE search strategy into the other databases using the appropriate controlled vocabulary as applicable. We applied date limits to the previously used search terms, and limited searches in The Cochrane Library by publication years 2009‐2014. We applied the new terms without time limits. 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 were not missed by the search strategy because of such variations. See Appendix 1 for details of the search strategies used.

Data collection and analysis

Selection of studies

Two reviewers (LA and RST) independently examined the titles and abstracts of citations identified by the electronic searches for possible inclusion and coded them as 'retrieve' (eligible or potentially eligible/unclear) or 'do not retrieve’. We retrieved full text publications of potentially relevant studies (and had them translated into English where required) and two reviewers (LA and RST) then independently determined study eligibility using a standardised inclusion form. We resolved any disagreements about study eligibility by discussion and, if necessary, a third reviewer (ADZ) was asked to arbitrate.

Data extraction and management

One reviewer (LA) extracted study characteristics of included RCTs and outcome data using a standardised data collection form which had been piloted on two RCTs included in the review. A second author (RST) checked all extracted data for accuracy. We resolved disagreements by consensus. 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. A second reviewer (RST) confirmed all numeric calculations and extractions from graphs or figures. Any discrepancies were resolved by consensus. One author (LA) transferred extracted data into Review Manager 5.3 (RevMan 2014), and a second author (RST) spot‐checked data for accuracy against the systematic reviews.

Data on patient characteristics (e.g. age, sex, CHD diagnosis) and details of the intervention (including mode of exercise, duration, frequency and intensity), description of usual care and length of follow‐up were also extracted.

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 reviewer (LA) assessed the risk of bias in included studies using the Cochrane Collaboration'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 drop‐outs and 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 exercise 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 (e.g. baseline data reported in Table 1) 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 (e.g. no Table 1) 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 reviewer (RST) and any discrepancies were resolved by consensus. Details of the assessments of risk of bias for each included trial are shown in the Characteristics of included studies table.

Quality of evidence in included reviews

One author (LA) used GRADEProfiler software to assess the quality of evidence for outcomes reported in the review (GRADEpro GDT 2015), based on the following factors: indirectness of evidence, unexplained heterogeneity, publication bias, risk of bias due to study design limitations and imprecision of results (Balshem 2011). A second author (RST) checked the assessment.

Data analysis

We processed data in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011). Dichotomous outcomes for each comparison have been expressed as risk ratios (RR) with 95% confidence intervals (CI). If there was a statistically significant absolute risk difference, the associated number needed to treat for an additional beneficial or harmful outcome was calculated. Heterogeneity amongst included studies was explored qualitatively, by comparing the characteristics of included studies, and quantitatively, using the Chi² test of heterogeneity and I² statistic (Higgins 2003). Given the clinical heterogeneity of the included trials, we pooled data from each study using a random effects model. Compared with a fixed‐effects, this model provides a more conservative statistical comparison of the difference between intervention and control by typically providing a wider confidence interval around the effect estimate. If a statistically significant difference was present using the random‐effects model, we also reported the fixed effect pooled estimate and 95% CI because of the tendency of smaller trials, which are more susceptible to publication bias, to be over weighted with a random effects analysis (Heran 2008a; Heran 2008b). We planned to pool the results for HRQL using a standardised mean difference (SMD) but this was not possible due to the heterogeneity in outcome measures and methods of reporting findings.

As length of follow‐up was anticipated to be a driver of intervention effect, we stratified meta‐analysis of each outcome according to the length of trial duration i.e. 'short‐term' follow up (6 to 12 months); 'medium‐term' follow‐up (13 to 36 months), and 'long‐term' follow‐up ( > 36 months). Univariate meta‐regression was undertaken to explore heterogeneity and examine potential treatment effect modifiers. We tested nine hypotheses that there may be differences in the effect of exercise‐based CR on total mortality, cardiovascular mortality, total MI, revascularisation (CABG and PCI) and hospitalisation across particular subgroups: (1) CHD case mix (MI‐only trials versus other trials); (2) type of CR (exercise‐only CR versus comprehensive CR); (3) 'dose' of exercise intervention [dose = number of weeks of exercise training x average number of sessions/week x average duration of session in minutes] (dose ≥ 1000 units versus dose < 1000 units); (4) follow‐up period; (5) year of publication; (6) sample size; (7) setting (home‐ or centre‐based CR); (8) risk of bias (low risk of bias in < 5 out of 8 domains) ; and (9) study location (continent). Hypotheses (1) to (5) were defined a priori and (7) to (9) during this update. Given the relatively small ratio of trials to covariates, meta‐regression was limited to univariate analysis (Deeks 2011). The permute option in STATA was used to allow for multiple testing in meta‐regression (StataCorp 2013).

We used the funnel plot and the Egger test to examine small study bias (Egger 1997). We processed data in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011). We completed data synthesis and analyses using Review Manager 5.3 software (RevMan 2014) and STATA version 13.0 (StataCorp 2013).

Results

Description of studies

The previous version of this review (Heran 2011) included 30 trials (55 publications) from the original Cochrane review (Andersen 1981; Bell 1998; Bengtsson 1983; Bertie 1992; Bethell 1990; Carlsson 1998; Carson 1982; DeBusk 1994; Engblom 1996; Erdman 1986; Fletcher 1994; Fridlund 1991; Haskell 1994; Heller 1993; Holmbäck 1994; Kallio 1979; Leizorovicz 1991; Lewin 1992; Miller 1984; Oldridge 1991; Ornish 1990; Schuler 1992; Shaw 1981; Sivarajan 1982; Specchia 1996; Stern 1983; Vecchio 1981; Vermeulen 1983; WHO 1983; Wilhelmsen 1975) and an additional 17 studies (26 publications) identified by the updated search (Belardinelli 2001; Bäck 2008; Dugmore 1999; Giallauria 2008; Hofman‐Bang 1999; Kovoor 2006; La Rovere 2002; Manchanda 2000; Marchionni 2003; Seki 2003; Seki 2008; Ståhle 1999; Toobert 2000; VHSG 2003; Yu 2003; Yu 2004; Zwisler 2008). This 2015 update identified an additional 16 trials (20 publications) (Aronov 2010; Bettencourt 2005; Briffa 2005; Hambrecht 2004; Higgins 2001; Houle 2012; Maddison 2014; Maroto 2005; Munk 2009; Mutwalli 2012; Oerkild 2012; Reid 2012; Roman 1983; Sandström 2005; Wang 2012; West 2012) as well as one publication (Dorn 1999) which provided further follow‐up data of a study included in the original review (Shaw 1981). The study selection process is summarised in the PRISMA flow diagram shown in Figure 1 (Liberati 2009). A total of 63 studies reporting data for a total of 14,486 patients have been included in this review update.

Figure 1

Summary of study selection process

Details of the studies included in the review are listed in the Characteristics of included studies table. Thirty eight studies compared comprehensive programmes (i.e. exercise plus education or psychological management, or both), while 24 reported on an exercise‐only intervention. In addition, one study randomised patients to a comprehensive programme, exercise only intervention or usual care (Sivarajan 1982). The majority of studies (37 studies, 59%) were undertaken in Europe, either as single (n = 45) or multicentre (n = 18) studies. Most trials were relatively small in sample size (median 126, range: 28‐2304). Two large trials (WHO 1983;West 2012) contributed about 30% (4997 participants) of all included participants. The median duration of trial intervention was six months (range 1 to 48) with median follow‐up of 12 months (range 6 to 120) months. Patients with MI alone were recruited in 31 trials (49%); the remaining trials recruited patients suffering exclusively from angina (five trials), post‐revascularisation patients (two trials) or a mixed population of patients with CHD. The mean age of patients within the trials ranged from 49.3 to 71.0 years. Although over half of the trials included women (42 studies, 67%), women accounted for less than 15% of the patients recruited overall. More recent trials were less dominated by MI patients and were more likely to include older and female participants. The average mean age of patients within trials rose from 56.3 years for trials published prior to 2005, to 61.7 years for trials published since 2005. In this time, the proportion of women in trials increased from 12.7% to 20.7%.

The CR programmes were commonly delivered in either an exclusively supervised centre‐based setting or a centre‐based setting in combination with some home exercise sessions. Fifteen studies were conducted in an exclusively home‐based setting (Bäck 2008; Bell 1998; DeBusk 1994; Fletcher 1994; Haskell 1994; Heller 1993; Higgins 2001; Houle 2012; Lewin 1992; Maddison 2014; Miller 1984; Mutwalli 2012; Oerkild 2012; Reid 2012; Wang 2012), with two of these studies randomising patients to usual care, or to an electronically‐delivered intervention designed to increase exercise behaviour, accessed via a mobile phone or the internet (Maddison 2014;Reid 2012).The mode of exercise training in CR programmes was aerobic in nature and most commonly static cycling, walking or circuit training.The dose of exercise ranged considerably across trials, in overall duration (range 1 to 48 months), frequency (1 to 7 sessions/week), session length (20 to 90 minutes/session) and intensity (50% to 85% of maximal heart rate; 50% to 95% of maximal oxygen uptake (VO₂ max); Borg rating of 11 to 15). Due to poor and inconsistent reporting of adherence and fidelity to exercise programmes in the RCTs, we were not able to consider the actual amount of exercise that the participants received or performed in this review. In general, comparator groups were described as receiving usual or standard care, which might have included medication, education and advice about diet and exercise, or psychosocial support, or both, but no formal exercise training. One trial (Hambrecht 2004) compared exercise training to stent angioplasty for patients with stable angina, while another (Kovoor 2006) compared exercise training to an "early return to normal activities group" where patients returned to work two weeks following a MI, without a formal CR programme.

Sixty eight publications identified in the current search were excluded for reasons listed in the Characteristics of excluded studies table. The most common reasons for exclusion were a failure to report any of the pre‐specified outcomes of this review update, or that the study was not a RCT. The status of ongoing trials which meet the inclusion criteria of this review are detailed in the Characteristics of ongoing studies table.

Risk of bias in included studies

The overall risk of bias was low or unclear. A number of trials failed to give sufficient detail to assess their potential risk of bias although the quality of reporting was generally higher in more recently published trials (Figure 2).

Figure 2

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

Allocation

Nearly all the trial publications reported that the trial was 'randomised' but did not provide sufficient details to assess whether the method was appropriate. A total of 16/63 (25%) studies reported details of appropriate generation of the random sequence (Andersen 1981; Bell 1998; Bethell 1990; Briffa 2005; Erdman 1986; Hambrecht 2004; Haskell 1994; Holmbäck 1994; Houle 2012; Maddison 2014; Munk 2009; Oerkild 2012; Reid 2012; Wang 2012; Wilhelmsen 1975; Zwisler 2008) and 13/63 (21%) studies reported appropriate concealment of allocation (Bell 1998; Briffa 2005; Haskell 1994; Holmbäck 1994; Kovoor 2006; Maddison 2014; Munk 2009; Oerkild 2012; Reid 2012; Schuler 1992; VHSG 2003; West 2012; Zwisler 2008).

Blinding

Given the nature of the exercise‐based CR intervention, it is not possible to blind participants or programme personnel. Only 16/63 studies (25%) reported adequate details of blinding of outcome assessment (Fletcher 1994; Giallauria 2008; Hambrecht 2004; Holmbäck 1994; Lewin 1992; Maddison 2014; Manchanda 2000; Marchionni 2003; Munk 2009; Ornish 1990; Reid 2012; Sandström 2005; Schuler 1992; West 2012; Wilhelmsen 1975; Zwisler 2008).

Incomplete outcome data

Although losses to follow‐up and drop‐out were relatively high (ranging from 21% to 48% in trials where losses to follow up were reported), follow‐up of 80% or more was achieved in 45/63 (71%) studies (Andersen 1981; Aronov 2010; Belardinelli 2001; Bell 1998; Bethell 1990; Bettencourt 2005; Bäck 2008; Briffa 2005; Carlsson 1998; Dugmore 1999; Engblom 1996; Giallauria 2008; Hambrecht 2004; Haskell 1994; Heller 1993; Holmbäck 1994; Kallio 1979; La Rovere 2002; Leizorovicz 1991; Lewin 1992; Maddison 2014; Manchanda 2000; Marchionni 2003; Maroto 2005; Miller 1984; Munk 2009; Oerkild 2012; Oldridge 1991; Roman 1983; Sandström 2005; Schuler 1992; Seki 2003; Shaw 1981; Specchia 1996; Stern 1983; Ståhle 1999; Toobert 2000; Vermeulen 1983; VHSG 2003; Wang 2012; West 2012; Wilhelmsen 1975; Yu 2003; Zwisler 2008). However, reasons for loss to follow‐up and dropout were often not reported, and only 25/63 (40%) of studies were judged to have adequately reported reasons for loss to follow‐up, thus having a low risk of bias, with 36/63 (57%) studies judged as having a high risk of bias and two studies having an unclear risk of bias.

Selective reporting

While the majority (56/63; 89%) of trials reported all outcomes listed in the methods section, six trials failed to report all outcomes at all time points collected (La Rovere 2002; Manchanda 2000; Oerkild 2012; Ornish 1990; Specchia 1996; Toobert 2000) and one trial was judged as having an unclear risk of bias as it didn't clearly describe the outcomes to be collected in the methods section (Wilhelmsen 1975). A number of the included studies were not designed to assess treatment group differences in morbidity and mortality (as these were not the primary outcomes of these trials) and, therefore, may not have fully reported all clinical events that occurred during the follow‐up period. All studies collecting validated HRQL outcomes fully reported these outcomes.

Other potential sources of bias

Groups balanced at baseline

The majority of studies (47/63; 75%) reported the baseline characteristics of participants in the intervention and comparator groups to be comparable in terms of likely main prognostic factors, or provided sufficient data for them to be judged to be comparable. In 13 studies there was evidence of substantive imbalance in the baseline characteristics of the intervention and control groups with regard to likely major prognostic factors (Bäck 2008; Carson 1982; Fletcher 1994; Haskell 1994; Hofman‐Bang 1999; Kovoor 2006; Lewin 1992; Manchanda 2000; Specchia 1996; Stern 1983; Toobert 2000; WHO 1983; Wilhelmsen 1975), while three further studies reported inadequate information to assess whether the two groups were balanced (Bell 1998; Carlsson 1998; Vermeulen 1983).

Intention‐to‐treat analysis conducted

Twenty nine of the studies (46%) reported that their analysis was conducted according to an intention‐to‐treat analysis, and measured outcome data on all or the majority of participants, or included imputation of all missing data using appropriate methodology in the analysis (Andersen 1981; Bäck 2008; Belardinelli 2001; Bengtsson 1983; Bettencourt 2005; Briffa 2005; Carlsson 1998; DeBusk 1994; Engblom 1996; Fletcher 1994; Fridlund 1991; Hambrecht 2004; Heller 1993; Houle 2012; Maddison 2014; Kovoor 2006; La Rovere 2002; Leizorovicz 1991; Oerkild 2012; Reid 2012; Roman 1983; Sandström 2005; Shaw 1981; Specchia 1996; Vecchio 1981; Vermeulen 1983; Wang 2012; West 2012; Zwisler 2008). Eight studies provided insufficient detail to determine if the trial performed an intention‐to‐treat analysis (Aronov 2010; Bell 1998; Dugmore 1999; Manchanda 2000; Munk 2009; Seki 2008; Toobert 2000; WHO 1983), while the remaining 26 studies did not conduct an intention‐to‐treat analysis, or use appropriate methodology for imputation of missing data.

Groups received comparable treatment

Thirty three studies (52%) were judged to have a low risk of bias, with all co‐interventions being described as being delivered equally to both the intervention and comparator groups (Andersen 1981; Aronov 2010; Bäck 2008; Belardinelli 2001; Bertie 1992; Bethell 1990; Bettencourt 2005; Briffa 2005; Carson 1982; Dugmore 1999; Engblom 1996; Erdman 1986; Fletcher 1994; Giallauria 2008; Hambrecht 2004; Holmbäck 1994; Houle 2012; La Rovere 2002; Maddison 2014; Marchionni 2003; Miller 1984; Munk 2009; Mutwalli 2012; Oerkild 2012; Sandström 2005; Schuler 1992; Shaw 1981; Specchia 1996; Ståhle 1999; Stern 1983; Vecchio 1981; Wang 2012; Wilhelmsen 1975). Twenty eight studies (44%) were judged to have a high risk of bias as the intervention included additional components that were not received by the comparator group. Two studies did not describe the intervention and comparator groups with sufficient detail to assess if the two groups received comparable treatment (Bell 1998; Roman 1983).

Effects of interventions

See: Summary of findings for the main comparison Exercise‐based cardiac rehabilitation for coronary heart disease

Clinical Events

Mortality

Forty seven (N = 12,455 participants) of the included studies reported total mortality (Analysis 1.1, Figure 3). Four trials contributed mortality data at more than one follow‐up period (WHO 1983; Wilhelmsen 1975; Shaw 1981;West 2012). Compared with control, while there was a reduction in total mortality with exercise‐based CR in trials with short‐ (29 trials, RR: 0.88, 0.73 to 1.05) and medium‐term follow‐up (13 trials, RR 0.89, 95% CI 0.78 to 1.01) this failed to reach statistical significance. There was no evidence of a difference between groups in the long‐term follow‐up (11 trials, RR 0.91, 95% CI 0.75 to 1.10) or across all trials reporting this outcome (47 trials, RR 0.96, 95% CI 0.88 to 1.04).

Figure 3

Total mortality for all studies at their longest follow‐up.

Filled diamonds represent the risk ratio (RR) for individual studies at the longest reported follow‐up. The boxes are proportional to the weight of each
study in the analysis and the lines represent their 95% confidence interval (CI). The open diamond represents the pooled RR, and its width represents its
95% CI.

Twenty seven trials (N = 7469 participants) reported cardiovascular mortality (Analysis 1.2, Figure 4). One trial reported both short‐ and medium‐term follow‐up (WHO 1983). While there was a reduction in cardiovascular mortality in the short‐term (15 trials, RR 0.90, 95% CI 0.69 to 1.17) this only became statistically significant in the medium‐ (7 trials, RR 0.77, 95% CI 0.63 to 0.93) and long‐term (8 trials, RR 0.58, 95% CI 0.43 to 0.78) follow‐up. A reduction in cardiovascular mortality was also seen across all trials reporting this outcome (27 trials, RR 0.74, 95% CI 0.64 to 0.86) .

Figure 4

CV mortality for all studies at their longest follow‐up.

Twenty studies reported both mortality outcomes. Results for mortality outcomes in this sub‐group were consistent with the overall meta‐analysis results (all‐cause mortality RR: 0.91, 95% CI: 0.82 to 1.01; CV mortality RR: 0.78, 95%: 0.67 to 0.90). There was no evidence of statistical heterogeneity across trials for either total or cardiovascular mortality.

Myocardial infarctions

Thirty six trials (N = 9717 participants) reported the risk of fatal or non‐fatal MI (Analysis 1.3,Figure 5). Although there was no statistically significant difference in the risk of total MI in trials with follow‐up in the short‐ (20 trials, RR 0.85, 95% CI 0.67 to 1.08) or medium‐term (11 trials, RR 1.09, 95% CI 0.91 to 1.29), or across all trials reporting this outcome (36 trials, RR 0.90, 95% CI 0.79 to 1.04), there was evidence of a significant reduction in risk in studies with long‐term follow‐up (10 trials, RR 0.67, 95% CI 0.50 to 0.90). There was no evidence of statistical heterogeneity across trials.

Figure 5

Fatal and / or nonfatal MI for all studies at their longest follow‐up.

Revascularisations

Twenty nine (N = 5891 participants), and 18 (N = 4012 participants) of the included trials reported the risk of CABG and PCI, respectively (Analysis 1.4, Figure 6; Analysis 1.5, Figure 7). There was no difference between exercise‐based CR and usual care for either CABG or PCI in trials with short‐ (CABG: 21 trials, RR 0.99, 95% CI 0.77 to 1.26; PCI: 13 trials, RR 0.92, 95% CI 0.64 to 1.33) or medium‐term (CABG: 8 trials, RR 0.98, 95% CI 0.78 to 1.25; PCI: 6 trials, RR 0.96, 95% CI 0.69 to 1.35) follow‐up, or across all trials reporting these outcomes (CABG: 29 trials, RR 0.96, 95% CI 0.80 to 1.16; PCI: 18 trials, RR 0.85, 95% CI 0.70 to 1.04). A reduction in revascularisation in the small number of trials reporting follow‐up longer than 36 months did not reach statistical significance (CABG: 4 trials, RR 0.66, 95% CI 0.34 to 1.27; PCI: 3 trials, RR 0.76, 95% CI 0.48 to 1.20). There was no evidence of major statistical heterogeneity across trials.

Figure 6

CABG for all studies at their longest follow‐up.

Figure 7

PCI for all studies at their longest follow‐up.

Hospitalisations

Fifteen (N = 3030 participants) studies reported hospital admissions (Analysis 1.6, Figure 8). One study reported follow‐up at both short‐ and medium‐term (Hofman‐Bang 1999). No trials with long‐term follow‐up reported data. Risk of hospital admissions was reduced with exercise‐based CR compared with usual care in the short term (9 trials, RR 0.65, 95% CI 0.46 to 0.92) with no significant difference in trials with medium‐term follow‐up (6 trials, RR 0.95, 95% CI 0.84 to 1.07). A significant reduction in risk was seen across all trials reporting hospitalisations (RR 0.82, 95% CI 0.70 to 0.96).

Figure 8

Hospital admissions for all studies at their longest follow‐up.

Health‐related quality of life

Twenty trials (N = 5060 participants) assessed HRQL using a range of validated generic (e.g. Short‐Form 36) or disease‐specific (e.g. HeartQOL) outcome measures (Table 1). Given both the heterogeneity in HRQL outcome measures and methods of reporting findings, a meta‐analysis was not undertaken. Although most trials demonstrated an improvement in HRQL at follow‐up compared to baseline following exercise‐based CR, a within‐group improvement was also often reported in control patients. Fourteen out of the 20 trials reported higher levels of quality of life in one or more sub‐scale with exercise‐based CR compared with control at follow‐up (Belardinelli 2001; Bell 1998; Bettencourt 2005; Briffa 2005; Engblom 1996; Heller 1993; Hofman‐Bang 1999; Houle 2012; Maddison 2014; Mutwalli 2012; Reid 2012; Toobert 2000; Wang 2012; Yu 2003), and in five trials there was evidence of a significantly higher level of quality of life in half or more ( ≥ 50%) of the sub‐scales (Belardinelli 2001; Bell 1998; Mutwalli 2012; Reid 2012; Wang 2012).

Open in table viewer

Table 1. Summary of health‐related quality of life (HRQL) scores at follow‐up

Measure of HRQL	Mean (SD) outcome values at follow‐up		P value	Difference between groups
	Exercise	Usual Care
Belardinelli 2001
SF‐36 at 6 months follow‐up:
Physical functioning	78 (19)	55 (20)	0.001	Exercise > Usual care
Physical performance	75 (13)	65 (14)	0.01	Exercise > Usual care
Bodily pain	4 (9)	22 (10)	0.001	Exercise > Usual care
General health	68 (14)	50 (19)	0.001	Exercise > Usual care
Vitality	NR	NR
Social functioning	66 (10)	69 (12)	0.14*	Exercise = Usual care
Emotional performance	NR	NR
Mental health	65 (12)	48 (15)	0.01	Exercise > Usual care
SF‐36 at 12 months follow‐up:
Physical functioning	82 (18)	54 (20)	0.001	Exercise > Usual care
Physical performance	76 (9)	58 (14)	0.01	Exercise > Usual care
Bodily pain	4 (9)	32 (12)	0.001	Exercise > Usual care
General health	70 (14)	50 (18)	0.001	Exercise > Usual care
Vitality	NR	NR
Social functioning	68 (11)	68 (12)	1.00*	Exercise = Usual care
Emotional performance	NR	NR
Mental health	70 (14)	45 (15)	0.001	Exercise > Usual care
Bell 1998
Nottingham health profile at 10.5 months follow‐up:
Energy	17.6 (27.1)	18.3 (29.8)	0.87**	Exercise = Usual care
Pain	2.8 (8.8)	4.82 (11.9)	< 0.05	Exercise > Usual care
Emotional reactions	6.4 (17.0)	12.2 (19.9)	< 0.001	Exercise > Usual care
Sleep	7.5 (18.4)	20.5 (27.8)	< 0.001	Exercise > Usual care
Social isolation	2.3 (10.6)	4.0 (13.3)	0.37*	Exercise = Usual care
Physical mobility	8.4 (11.1)	8.9 (14.5)	0.82**	Exercise = Usual care
Bettencourt 2005
SF‐36 at 1 year follow‐up:
Physical functioning	70	62	NS*	Exercise = Usual care
Physical performance	66	57	NS*	Exercise = Usual care
Bodily pain	73	65	NS*	Exercise = Usual care
General health	57	46	< 0.02	Exercise > Usual care
Vitality	62	47	< 0.02	Exercise > Usual care
Social functioning	73	66	NS*	Exercise = Usual care
Emotional performance	65	58	NS*	Exercise = Usual care
Mental health	87	75	NS*	Exercise = Usual care
Mental component	71	57	0.02	Exercise > Usual care
Physical component	63	57	NS*	Exercise = Usual care
Briffa 2005
SF‐36 at 6 months follow‐up:
	Δ (95% CI)	Δ (95% CI)
Physical functioning	7.1 (1 to 13)	15.9 (−8 to 23)	NS*	Exercise = Usual care
Physical performance	75 (0 to 100)	75 ( 0 to 100 )	NS*	Exercise = Usual care
Bodily pain	19.2 (11 to 27)	26.6 (18 to 35 )	NS*	Exercise = Usual care
General health	− 0.6 (‐5 to 4)	0.1 (−6 to 6 )	NS*	Exercise = Usual care
Vitality	3.7 (‐2 to 9)	7.1 (1 to 13 )	NS*	Exercise = Usual care
Social functioning	14.1 (7 to 21)	19.6 (10 to 29 )	NS*	Exercise = Usual care
Emotional performance	33.3 (33 to 100)	33.3 (0 to 100 )	NS*	Exercise = Usual care
Mental health	1.4 (‐3 to 5)	0.5 (−4 to 5)	NS*	Exercise = Usual care
SF‐36 at 1 year follow‐up:
	Δ (95% CI)	Δ (95% CI)
Physical functioning	6.8 (−1 to 14 )	17.6 (10 to 25)	0.04	Exercise > Usual care
Physical performance	75 (12 to 30 )	100 (0 to 100)	NS*	Exercise = Usual care
Bodily pain	20.9 (−2 to 7)	30.2 (23 to 37)	NS*	Exercise = Usual care
General health	2.2 (−2 to 7)	2.7 (−3 to 5)	NS*	Exercise = Usual care
Vitality	6.9 (1 to 12)	11.9 (6 to 18)	NS*	Exercise = Usual care
Social functioning	16.4 (9 to 23 )	23.6 (14 to 33)	NS*	Exercise = Usual care
Emotional performance	33.3 (33 to 100 )	33.3 (33 to 100)	NS*	Exercise = Usual care
Mental health	3.9 (0 to 8)	3.6 (−1 to 9)	NS*	Exercise = Usual care
Engblom 1992
Nottingham health profile at 5 years follow‐up:
Energy	18	25	0.08	Exercise = Usual care
Pain	12	18	0.07	Exercise = Usual care
Emotional reactions	14	21	0.27	Exercise = Usual care
Sleep	24	29	0.42	Exercise = Usual care
Social isolation	7	9	0.42	Exercise = Usual care
Physical mobility	6	14	0.005	Exercise > Usual care
Heller 1993
QLMI at 6 months follow‐up:
Emotional	5.4 (1.1)	5.2 (1.2)	0.04	Exercise > Usual care
Physical	5.4 (1.2)	5.2 (1.3)	0.17*	Exercise = Usual care
Social	5.9 (1.1)	5.8 (1.1)	0.35*	Exercise = Usual care
Hofman‐Bang 1999
AP‐QLQ at 12 months follow‐up:
Physical activity	4.9	4.3	<0.05	Exercise > Usual care
Somatic symptoms	NR	NR	NS	Exercise = Usual care
Emotional distress	NR	NR	NS	Exercise = Usual care
Life satisfaction	NR	NR	NS	Exercise = Usual care
Houle 2012
Quality of Life Index‐cardiac version III at 6 months follow‐up:
Health and functional score	26 (5.1)	24.5 (5.3)	0.048	Exercise > Usual care
Psychological/spiritual score	25.6 (5.8)	25.5 (3.8)	0.383	Exercise = Usual care
Social and economic score	25.7 (3.8)	25.4 (4.7)	0.392	Exercise = Usual care
Family score	28.1 (2.5)	26.7 (4.3)	0.048	Exercise > Usual care
Overall	26.2 (4.3)	25.8 (4.1)	0.057	Exercise = Usual care
Quality of Life Index‐cardiac version III at 12 months follow‐up:
Health and functional score	27.8 (2.0)	25.3 (4.6)	0.036	Exercise > Usual care
Psychological/spiritual score	27.4 (2.5)	26.2 (4.0)	0.336	Exercise = Usual care
Social and economic score	27.2 (3.0)	25.9 (5.2)	0.638	Exercise = Usual care
Family score	28 (2.6)	26.8 (5.0)	0.092	Exercise = Usual care
Overall	27.7 (2.1)	25.7 (4.2)	0.048	Exercise > Usual care
Maddison 2014
EQ‐5D at 24 weeks follow‐up:
	0.86	0.83	0.23	Exercise = Usual care
SF‐36 at 24 weeks follow‐up:
Physical functioning	52.9	51.9	0.20	Exercise = Usual care
Role physical	52.6	50.8	0.08	Exercise = Usual care
Bodily pain	52.4	51.9	0.71	Exercise = Usual care
General health	55.3	53.2	0.03	Exercise > Usual care
Vitality	55.7	55.9	0.79	Exercise = Usual care
Social Functioning	53.3	52.4	0.42	Exercise = Usual care
Role emotional	51.4	51.6	0.81	Exercise = Usual care
Mental health	54.6	54.0	0.61	Exercise = Usual care
Mutwalli 2012
SF‐36 Health status score at 6 months follow‐up:
	90.14 (4.83)	60.55 (16.21)	0.000	Exercise > Usual care
Oerkild 2012
SF‐36 at 12 months follow‐up:
	Δ (95% CI)	Δ (95% CI)
SF 12 PCS	‐1.1 (‐5.3 to 3.1)	‐1.4 (‐5.2 to 2.3)	NS*	Exercise = Usual care
SF 12 MCS	‐1.4 (‐6.1 to 3.3)	‐0.3 (‐4.6 to 4.0)	NS*	Exercise = Usual care
Oldridge 1991
QLMI at 4 months follow‐up:
Limitations	54	54	NS	Exercise = Usual care
Emotions	103	101	NS	Exercise = Usual care
QLMI at 8 months follow‐up:
Limitations	54	54	NS	Exercise = Usual care
Emotions	103	103	NS	Exercise = Usual care
QLMI at 12 months follow‐up:
Limitations	54	55	NS	Exercise = Usual care
Emotions	105	102	NS	Exercise = Usual care
Reid 2012
MacNew at 12 months follow‐up:
Global score	5.8 (0.6)	5.6 (0.8)	0.112	Exercise = Usual care
Emotional subscale	5.6 (0.6)	5.4 (0.7)	0.038	Exercise > Usual care
Social subscale	6.3 (0.8)	6.0 (1.0)	0.162	Exercise = Usual care
Physical subscale	6.0 (0.8)	5.8 (1.0)	0.031	Exercise > Usual care
Sandstrom 2005
Time Trade Off (TTO) at 12 months follow‐up:
	0.86 (0.23)	0.85 (0.21)	NS*	Exercise = Usual care
EuroQol Part one at 12 months follow‐up:
	0.87 (0.15)	0.86 (0.16)	NS*	Exercise = Usual care
EuroQol Part two at 12 months follow‐up:
	7.6 (1.46)	7.43 (1.46)	NS*	Exercise = Usual care
Stahle 1999
Karolinska Questionnaire at 12 months follow‐up:
Chest pain	0.6 (1.2)	0.4 (1.3)	NS	Exercise = Usual care
Shortness of breath	0.4 (1.1)	0.2 (1.0)	NS	Exercise = Usual care
Dizziness	‐0.1 (1.1)	0.2 (0.9)	NS	Exercise = Usual care
Palpitation	‐0.1 (1.0)	0.1 (0.9)	NS	Exercise = Usual care
Cognitive ability	‐0.1 (0.6)	0.0 (0.7)	NS	Exercise = Usual care
Alertness	0.0 (0.9)	0.1 (0.8)	NS	Exercise = Usual care
Quality of sleep	0.0 (0.5)	0.1 (0.5)	NS	Exercise = Usual care
Physical ability	0.2 (0.7)	0.1 (0.4)	NS	Exercise = Usual care
Daily activity	0.3 (0.5)	0.1 (0.5)	NS	Exercise = Usual care
Depression	0.1 (0.3)	0.1 (0.2)	NS	Exercise = Usual care
Self perceived health	0.5 (1.3)	0.3 (1.0)	NS	Exercise = Usual care
"Ladder of Life" present	1.2 (1.2)	0.9 (1.8)	NS	Exercise = Usual care
"Ladder of Life" future	0.8 (2.7)	0.4 (2.3)	NS	Exercise = Usual care
Fitness	0.6 (1.4)	0.4 (1.0)	NS	Exercise = Usual care
Physical ability	0.7 (1.0)	0.4 (1.1)	NS	Exercise = Usual care
Toobert 2000
SF‐36 at 24 months follow‐up:
Physical functioning	NR	NR	NS	Exercise = Usual care
Physical performance	NR	NR	NS	Exercise = Usual care
Bodily pain	NR	NR	NS	Exercise = Usual care
General health	NR	NR	<0.05	Exercise > Usual care
Vitality	NR	NR	NS	Exercise = Usual care
Social functioning	NR	NR	<0.05	Exercise > Usual care
Emotional performance	NR	NR	NS	Exercise = Usual care
Mental health	NR	NR	NS	Exercise = Usual care
Wang 2012
SF‐36 at 6 months follow‐up:
Physical functioning	80.8 (13.7)	73.2 (13.0)	<0.001	Exercise > Usual care
Physical performance	68.2 (17.3)	56.2 (46.8)	0.015	Exercise > Usual care
Bodily pain	68.2 (17.3)	63.5 (14.6)	0.012	Exercise > Usual care
General health	57.4 (20.3)	49.0 (16.2)	0.017	Exercise > Usual care
Vitality	66.3 (17.3)	56.4 (21.7)	0.002	Exercise > Usual care
Social functioning	71.3 (21.4)	65.8 (18.0)	0.031	Exercise > Usual care
Emotional performance	80.8 (37.9)	75.9 (39.7)	0.12	Exercise = Usual care
Mental health	73.5 (17.1)	65.4 (20.7)	0.011	Exercise > Usual care
MIDAS at 6 months
Physical Activity	37.7 (11.2)	42.6 (12.3)	<0.001	Exercise > Usual care
Insecurity	28.7 (9.7)	33.4 (13.8)	<0.001	Exercise > Usual care
Emotional reaction	30.4 (12.8)	34.8 (14.4)	0.008	Exercise > Usual care
Dependency	27.6 (9.4)	31.8 (16.6)	0.001	Exercise > Usual care
Diet	36.8 (15.4)	43.6 (20.7)	0.40	Exercise = Usual care
Concerns over meds	29.4 (12.6)	37.7 (18.0)	<0.001	Exercise > Usual care
Side Effects	28.2 (11.1)	30.8 (14.3)	0.30	Exercise > Usual care
West 2012
SF‐36 at 12 months follow‐up:
Physical function	65 (29)	64 (30)	NS*	Exercise = Usual care
Role physical	69 (31)	67 (33)	NS*	Exercise = Usual care
Role emotional	85 (23)	85 (25)	NS*	Exercise = Usual care
Social function	81 (28)	79 (29)	NS*	Exercise = Usual care
Mental health	76 (13)	76 (13)	NS*	Exercise = Usual care
Energy /vitality	65 (24)	65 (24)	NS*	Exercise = Usual care
Pain	69 (28)	68 (29)	NS*	Exercise = Usual care
Health Perception	58 (25)	57 (25)	NS*	Exercise = Usual care
Yu 2003
SF‐36 at 8 months follow‐up:
Physical functioning	88 (12)	82 (17)	0.03*	Exercise > Usual care
Physical performance	75 (33)	66 (35)	0.18*	Exercise = Usual care
Bodily pain	80 (25)	80 (25)	1.00*	Exercise = Usual care
General health	64 (26)	60 (28)	0.45*	Exercise = Usual care
Vitality	79 (18)	65 (17)	0.0001	Exercise > Usual care
Social functioning	89 (27)	82 (28)	0.15	Exercise = Usual care
Emotional performance	93 (18)	83 (35)	0.05	Exercise = Usual care
Mental health	84 (16)	80 (15)	0.2	Exercise = Usual care
SF‐36 at 24 months follow‐up:
Physical functioning	88 (13)	87 (9)	0.67*	Exercise = Usual care
Physical performance	80 (32)	79 (30)	0.87*	Exercise = Usual care
Bodily pain	81 (21)	85 (20)	0.33*	Exercise = Usual care
General health	64 (20)	61 (18)	0.43*	Exercise = Usual care
Vitality	73 (21)	73 (17)	1.00*	Exercise = Usual care
Social functioning	79 (30)	90 (18)	0.04*	Exercise > Usual care
Emotional performance	89 (25)	93 (25)	0.42*	Exercise = Usual care
Mental health	85 (14)	85 (12)	1.00*	Exercise = Usual care
Zwisler 2008
SF‐36 at 12 months follow‐up:
Physical Component Score	45.2 (9.8)	46.4 (9.8)	0.39*	Exercise = Usual care
Mental Component Score	50.6 (10.8)	48.4 (11.5)	0.16*	Exercise = Usual care

Short Form‐36 (SF‐36); QLMI=Quality of Life After Myocardial Infarction questionnaire; AP‐QLQ=Angina Pectoris‐Quality of Life questionnaire; NR=not reported; NS=not significant

* Calculated by authors of this report based on independent two group t test.

** Adjusted for baseline difference between groups.

Exercise = Usual care: no statistically significant difference (P > 0.05) between exercise and usual care groups at follow up

Exercise > Usual care: statistically significant difference (P < 0.05) between exercise and usual care groups at follow up

NS*: The authors of this review have inferred a P value of > 0.05 based either on the 95% CI, or from narrative in the paper, rather than from directly observing the P‐value.

Costs and cost‐effectiveness

Seven of the included studies reported data on costs of CR and overall healthcare costs in both groups (Briffa 2005; Hambrecht 2004; Kovoor 2006; Maddison 2014; Marchionni 2003; Oldridge 1991; Yu 2004). These results are summarised in Table 2. While it was not possible to directly compare costs across studies due to differences in currencies and the timing of studies, it is possible to compare the within‐study costs for CR and control groups. Three studies showed no difference in total healthcare costs between groups, (Briffa 2005; Kovoor 2006; Yu 2004), one study found healthcare costs for rehabilitation lower (USD 2378 less per patient) compared to control (Hambrecht 2004), and the remaining three did not report a p‐value for the cost difference (Maddison 2014; Marchionni 2003; Oldridge 1991).

Open in table viewer

Table 2. Summary of costs of exercise‐based rehabilitation and usual care

Author/year	Briffa 2005	Hambrecht 2004	Kovoor 2006¹	Maddison 2014	Marchionni 2003	Oldridge 1991²	Yu 2004
Follow‐up (months)	12	12	12	6	14	12	24
Year of costs (currency)	1998 ($Aus)	NR	1999 ($AUD)	NR (€; Euros)	2000 ($USD)	1991 ($USD)	2003 ($USD)
Cost of rehabilitation
Mean cost/patient	$694	NR	$394	€127	$5246	$670	NR
Costs considered	Details of costed elements not provided	NR	staff, assessments, counselling, education, patient travel	NR	NR	space, equipment, staff, literature resources, operating costs, parking, patients costs	NR
Total healthcare costs
Rehabilitation mean cost/patient	$4937	$3708 ± 156	NR	NR	$17 272	NR	$15 292
Usual care mean cost/patient	$4541	$6086 ± 370	NR	NR	$12 433	NR	$15 707
Difference mean/patient*	$395	‐$2378	NR	NR	$4839	$480	‐$415
P value for cost difference	0.74	P < 0.001	P > 0.05 (see below)	NR	NR	NR	P > 0.05
Additional healthcare costs considered	hospitalisations, pharmaceuticals, tests, consultations, rehabilitation, patient expenses, ambulance	rehospitalisations, revascularisation, cycle ergometers, training facilities, and supervising staff	phone calls (P = 0.10); hospital admissions (P = 0.11); gated heart pool scan (P = 0.50); exercise stress test (P = 0.72); other diagnostics (P = 0.37); visits to general practitioner (P = 0.61), specialist doctor (P = 0.35), or health‐care professional (P = 0.31)	NR	NR	Service utilisation, physician costs, emergency costs, in‐patient days, allied health, other rehabilitation visits	hospitalisations; revascularisations; private clinic visit; cardiac clinic visits; public noncardiac visits; casualty visits; drugs
Cost‐effectiveness
Rehabilitation mean health care benefits	Utility‐Based Quality of life – Heart questionnaire: 0.026 (95% CI, 0.013 to 0.039)	NR	NR	NR	NR	NR	NR
Usual care mean health care benefit	Utility 0.010 (95% CI, −0.001 to 0.022)	NR	NR	NR	NR	NR	NR
Incremental mean health care benefit	Utility 0.013 (95% CI, NR), P = 0.38; +0.009 QALYS	NR	NR	NR	NR	0.052 QALYS (95% CI, 0.007 to 0.1)	0.06 QALYs
Incremental cost effectiveness ratio/patient	+$42,535 per QALY. Extensive sensitivity analyses reported	NR	NR	+€15,247 per QALY	NR	+$9,200 per QALY	‐$650 per QALY

¹Cost data for this study is reported in Hall 2002

² cost data for this study is reported in Oldridge 1993

Four studies (Briffa 2005; Maddison 2014; Oldridge 1991; Yu 2004) also reported cost‐effectiveness using a cost utility approach (i.e. cost per quality‐adjusted life year (QALY). The incremental cost ratio ranged from an additional cost with CR compared to control of USD 42,535 more per QALY (Briffa 2005) to a reduction in cost of USD 650 less per QALY (Yu 2004). Based on these analyses, authors consistently concluded CR to be a cost‐effective use of healthcare resources compared to usual care.

Meta‐regression

Predictors of total mortality, cardiovascular mortality, recurrent MI, revascularisation (CABG and PCI) and hospitalisation were examined across the longest follow‐up of each individual study, using univariate meta‐regression. No statistically significant associations were seen in any of the analyses (Table 3, Table 4, Table 5, Table 6, Table 7, Table 8).

Open in table viewer

Table 3. Results of univariate meta‐regression analysis for total mortality

Explanatory variable (n trials)	Exp(slope)*	95% Confidence interval Univariate P value	Proportion of variation explained	Interpretation
Case mix (% MI patients) (n = 41)	RR = 0.998	0.996 to 1.000 P = 0.93	0%	No evidence that risk ratio is associated with case mix
Dose of exercise (number of weeks of exercise training x average number of sessions/week x average duration of session in min) (n = 29)	RR = 1.000	1.000 to 1.000 P = 0.74	0%	No evidence that risk ratio is associated with increased dose of exercise
Type of CR (exercise only vs comprehensive CR) (n = 42)	RR = 1.084	0.909 to 1.292 P = 1.00	0%	No evidence that risk ratio is associated with type of CR
Duration of follow‐up (months) (n = 41)	RR = 1.001	1.000 to 1.002 P = 1.00	0%	No evidence that risk ratio is associated with duration of follow‐up
Year of publication (pre 1995 vs post 1995) (n = 42)	RR = 1.006	0.999 to 1.013 P = 1.00	0%	No evidence that risk ratio is associated with year of publication
Setting (centre vs home) (n = 42)	RR = 1.012	0.822 to 1.246 P = 1.00	0%	No evidence that risk ratio is associated with setting
Risk of bias (low risk in ≥ 5 items v < 5 items) (n = 42)	RR = 1.033	0.985 to 1.083 P = 1.00	0%	No evidence that risk ratio is associated with risk of bias
Study location (n = 42)	RR = 1.071	0.915 to 1.254 P = 1.00	0%	No evidence that risk ratio is associated with study location
Sample size (n = 42)	RR = 1.192	0.732 to 1.940 P = 1.00	0%	No evidence that risk ratio is associated with sample size