Systematic Review of the Effect of Diet and Exercise Lifestyle Interventions in the Secondary Prevention of Coronary Heart Disease

2.1. Participants

We included male and female adults of all ages (aged 18+) with a diagnosis of CHD. Patients included those who had experienced a myocardial infarction (MI), coronary artery bypass graft (CABG), or percutaneous transluminal coronary angioplasty (PTCA) and those with angina pectoris and coronary artery disease defined by angiography. For the purposes of this review we excluded patients who had had a heart transplant, heart valve surgery, or heart failure, unless it was clearly specified that the cause related to CHD.

2.2. Interventions

2.3. Exclusions

We have excluded from this review studies which were not randomized controlled trials, focused on primary prevention, involved patients with multiple diseases and/or outcomes which were related to diseases other than CHD, involved patients in a hospital in-patient setting, focused on drug therapy, had short (less than three months) or no follow-up, or reported outcomes which were not the focus of this review (see “Types of outcome measures” section) such as depression, cost-effectiveness, or service delivery.

2.4. Study Duration

Trials were included in this review if they reported a minimum postintervention follow-up of three months to allow some change to take place.

2.5. Settings

2.6. Comparators

In the included studies, the comparators are “normal care” or “usual care,” meaning standard clinical care or standard care given by a general practitioner (GP).

2.7. Types of Outcome Measures

All or any number of the following.

Only outcomes measured using a validated instrument were included.

2.8. Search Methods for Identification of Studies

We searched electronic databases Medline, Cinahl, and Embase for English language randomized trials in humans published since 1990. The search method for Medline is detailed below; it was modified as appropriate for Cinahl and Embase. Reference lists of review articles were also searched.

2.9. Medline Search Method

Selecting an Advanced Ovid search, the keyword “lifestyle” was inserted and the box ticked to select “Map term to subject heading.” After selecting “Search” a new page was presented entitled “Mapping Display.” Options indicating the Subject Headings “Life Style” and “lifestyle.mp. search as keyword” were chosen and “Continue” selected. This allowed further keywords to be entered.

The same process was followed for the keywords “exercise,” “physical activity,” and “diet.” In each case only the keyword as a subject heading and as a keyword were selected on the Mapping Display page.

A fifth search was commanded by using all four keywords above separated by the Boolean operator “OR.” This was done by selecting “click to expand” the “Search History” box, ticking each of the four boxes relating to the keywords, and selecting “Combine selections with OR.”

A sixth search term “secondary prevention” was added as a keyword in same way as above. A seventh “coronary heart disease” was inserted, then its abbreviation “CHD” and four conditions related to it and their abbreviations, where appropriate, all as separate keywords: “myocardial infarction,” “MI,” “coronary artery bypass graft,” “CABG,” “percutaneous transluminal coronary angioplasty,” “PTCA,” and “angina pectoris.” This group incorporating coronary heart disease and related conditions involved a total of nine separate searches, and they were together inserted as a 16th search, with each term separated by “OR.” This was done, as above, by expanding the search history, selecting the relevant boxes beside keywords and selecting “Combine selections with OR.”

The three components of the search were put together in the following way.

On the search history page, search five was selected by clicking on the box beside it (Lifestyle.mp. or Life Style/OR Exercise.mp. or Exercise/OR Physical activity.mp. OR Diet/or diet.mp.). Search six was then selected (Secondary prevention.mp. or Secondary Prevention/), and search 16 (Coronary heart disease.mp. or Coronary Disease/OR CHD.mp. OR Myocardial infarction.mp. or Myocardial Infarction/OR MI.mp OR Coronary Artery Bypass/or coronary artery bypass graft.mp. OR CABG.mp. OR Percutaneous transluminal coronary angioplasty.mp. or Angioplasty, Transluminal, Percutaneous Coronary/OR PTCA.mp. OR Angina pectoris.mp. or Angina Pectoris/). The option “Combine selections with AND” was selected.

This gave search 17. Search 17 was selected by ticking the box, and limits were imposed in an effort to narrow the search. Below the search results, within the “Limits” options, “English language,” “Humans,” and “Publication Year-1990 to Current” were selected. “Additional Limits” was then selected, search 17 was selected, and in the “Age Groups” options “All Adult (19 plus years)” was selected. No other limits were imposed. Then, “Limit A Search” was selected.

To refine the search further, terms relating to a randomized controlled trial study design were inserted. The terms entered were randomized controlled trial.mp. or Randomized Controlled Trial/, controlled clinical trial.mp. or Controlled Clinical Trial/, random allocation.mp. or Random Allocation/, double blind method.mp. or Double-Blind Method/, single blind method.mp. or Single-Blind Method. Each of these terms was combined with “OR,” giving search 24.

Selecting search 18 and search 24, and combining them with “AND” gave the final selection. The Medline search strategy is detailed in the Appendix.

2.10. Data Analysis

From the searching, titles and abstracts were screened by two reviewers (MC, JC), and potentially relevant references were retrieved. The two reviewers (MC, JC) then independently selected trials to be included in the review. Reference lists of relevant systematic reviews were also screened for potential papers to include.

After the final selection of trials was agreed upon, data including study characteristics, outcome measures and results were extracted.

In addition, the quality of trials was assessed in relation to randomization method, loss to follow up, intention to treat analyses, and blinding measures.

Dichotomous outcomes for each study have been expressed as risk ratios, where appropriate, and 95% confidence intervals (CIs). Meta-analyses were performed using random effects models, and data were presented as forest plots using Revman software [15]. We were only able to perform meta-analysis on three outcomes—total mortality, cardiovascular mortality, and nonfatal cardiac events.

Continuous variables have been expressed as the difference between intervention and control groups at study completion and the standard deviation difference where reported.

3.1. Study Selection and Evaluation

Figure 1 shows the selection, screening, and identification of studies for this review. Of the 266 papers originally identified through searching the three electronic databases, we identified 39 that were potentially eligible for inclusion, and full text versions of these studies were retrieved and assessed. Twelve of the 39 papers were excluded because they included diseases other than CHD, were drug trials, did not focus on lifestyle, were not randomized controlled trials, had a short follow-up period, were conducted within a hospital in-patient setting, had outcomes which were not of interest, or were retrieved more than once from the three search engines. Two papers were added after screening reference lists of relevant systematic reviews, and one trial was added which reported different outcomes of an identical study which emerged in our searches (Figure 1).

3.2. Description of Studies

Table 1 shows summary data of 30 papers from the 21 randomised controlled trials (RCTs) found to be eligible for this review [16–45]. We have presented them in the six categories described in the “Methods” section (exercise, dietary, psychological, educational, multifactorial, and organisational). The category which had the largest number of trials was multifactorial (10 in total). We also found one which focused on exercise only, one dietary intervention, three which had a psychological approach, four educational, and two organisational.

In all the trials except where stated, patients randomised to the control group received usual care, which was not defined by every study but usually meant standard clinical care or standard care given by a GP.

3.3. Risk of Bias in Included Studies

The methodological quality of the included studies is presented in Table 2. The quality of the trials, in terms of the method of randomisation, whether groups were similar at baseline, losses to follow up, whether intention to treat analyses were used and blinding of outcome assessors, is as reported in the papers.

In seven of the 21 original studies (excluding follow-up papers), the randomisation method was not clearly described. In six of these trials, the method was not stated at all [16, 25, 26, 39, 43, 45]. The majority of the remaining studies described various randomisation techniques including the use of computer-generated random numbers, random numbers tables, and sealed envelopes. Lewin et al. [29] used block randomisation, while Carlsson [42] allocated patients in groups of 20. Murphy et al. [17] used cluster randomisation because their intervention was aimed to alter practitioners’ behaviour which could contaminate their interactions with control patients.

Loss to follow up was not reported by some authors. Where it was reported, it was defined differently. Some authors included as losses to follow up deaths and withdrawals due to poor health, while others defined it as people who could not be accounted for or contacted at follow-up. Follow-up as a quality marker relates only to original studies and varied considerably for longer-term follow-up studies as would be expected.

Thirteen of the 21 studies stated that analyses were conducted using intention to treat principles. The Vestfold Heartcare Study Group [28] used the last recorded value of a variable from the previous visit if there were missing data.

Only seven out of 21 studies reported that outcome assessors were blind to group allocation. Overall, the quality of many trials was poor with a majority not having blinded outcome assessors and many not describing the method of randomisation.

3.4. Effects of Interventions

3.4.1. All-Cause Mortality

Six studies reported total mortality and are presented in Figure 2. Overall, four studies showed benefits for intervention patients compared to controls in relation to total mortality. Because of the large differences in follow-up periods, we have used data from the studies which range from 3 to 5 years to allow easier comparison of studies. For all interventions except two, this constituted their study endpoint.

Murchie et al. [36] reported a significant survival effect for intervention patients compared to controls at their 4.7 year follow-up. Within the same study, Delaney et al. [38] reported that, at 10 years, the observed difference was no longer significant between the groups.

Hamalainen et al. [45] reported a significant survival effect for intervention patients compared to controls at their 4.7 year follow-up.

The I² test for heterogeneity combined with the Chi² nonsignificant P value suggests that the level of heterogeneity between studies is not important.

3.4.2. Cardiovascular Mortality

Eight studies reported cardiovascular mortality and are presented in Figure 3. Three studies showed significant survival effects. Data were collected at times ranging from 2 to 5 years. Delaney et al. [38] reported a 10-year follow-up study but we have presented their data collected at 4.7 years to allow for easier comparison with other studies. For the same reason we have used data reported by Hamalainen et al. [45] at three years of follow-up instead of the authors’ results at 15 years.

The three studies reporting significant survival effects were Cupples and McKnight [22], two-year follow-up results from an educational intervention, De Lorgeril et al. [33], a Mediterranean diet study with four-year follow-up, and Hamalainen et al. [45], a trial of a multifactorial intervention.

Hamalainen et al. [45], who reported coronary mortality, also reported significant survival effects for intervention patients compared to controls in their 15-year follow-up study. Delaney et al. [38] reported CV and coronary deaths at both 4.7- and 10-year follow-up points. However, the mortality figures were combined with nonfatal MI to calculate a proportional hazard ratio. Murchie et al. [36], a paper from the same study at 4.7 years of follow-up, reported an adjusted hazard ratio of 0.76 for coronary events (coronary deaths plus nonfatal MIs) (0.58 to 1.00, P = .049).

The I² test for heterogeneity indicated a moderate level of heterogeneity across studies.

3.4.3. Nonfatal Cardiac Events

This was reported by nine studies. In Figure 4, we have presented data for major nonfatal events (MI, CABG, PCI, coronary angioplasty). Significant benefits for intervention patients compared to controls were shown by five studies. Four studies reported nonfatal events separately [21, 33, 41, 44] while five reported the number of patients who had an event.

Lisspers et al. [27] included mortality in “event” results, therefore although they reported figures for various events and mortality they did not report a separate statistic for nonfatal events. Delaney et al. [38] reported coronary events which included coronary deaths and nonfatal MIs; however separate figures for each variable were not reported.

Gianuzzi et al. [21] reported nonfatal stroke, heart failure, and angina data but none of the values was significant.

Lewin et al. [29] reported the frequency of angina attack in number of episodes per week. Intervention group patients had a reduction of three attacks per week compared to a reduction of 0.4 attacks among control subjects (P = .016).

In addition to nonfatal MI and coronary revascularisation, De Lorgeril et al. [33] reported a number of major and minor secondary endpoints. They calculated risk ratios for nonfatal AMI combined with CV deaths, major secondary endpoints (periprocedural infarction, unstable angina, heart failure, stroke, pulmonary embolism, peripheral embolism) combined with primary endpoints, and primary and major secondary endpoints combined with minor secondary endpoints (stable angina, post-PCTA restenosis, and thrombophlebitis). Separate risk ratios for nonfatal CV events were not reported.

In addition to MI, PTCA, and CABG, Ornish et al. [41] reported the frequency, duration, and severity of angina chest pain at one year and five years. One significant value out of three outcomes reported at the two different time points each was reported: chest pain severity (scale 1–7, baseline to one year: intervention group 1.5 (1.5) to 0.7 (1.2); control group 0.6 (0.8) to 1.4 (1.2); P < .001).

The I²  test for heterogeneity indicated that there may be substantial heterogeneity across studies so results of this meta-analysis have to be interpreted with caution.

4.1. Effectiveness of Interventions

Overall, a small number of studies showed a significant reduction in deaths and nonfatal MIs while several reported positive results relating to adherence to lifestyle change. Regarding total mortality, four studies out of the six reporting this outcome showed significant benefits in the relatively short time scale studied (3–5 years). The Lyon Diet Heart Study [33] showed a significant impact on All-cause mortality and fewer deaths from CV causes among intervention patients compared to controls. The protective effect of the intervention was maintained for up to four years. This was a noteworthy result because patients were seen annually, indicating that they were able to maintain the diet for long periods alone and without professional motivation. The other trials showing significant benefits were of an educational intervention [22] and two multifactorial [36, 45]. Only three out of eight studies showed significant results for cardiovascular mortality. These were a dietary intervention [33], educational [22] and multifactorial ones [45], all of which had also shown significant effects on total mortality.

Concerning nonfatal CV events, we presented 16 outcomes from nine studies. There were significant improvements in five of these trials, which were of dietary [33], psychological [27], and multifactorial [21, 39, 41] interventions. Of the five studies which reported hospital admissions, only one, Murphy et al. [17], reported significant benefit for intervention patients compared to controls, both for overall and CV admissions.

In terms of risk factors for CHD, these were assessed by greatly differing methods which made comparison difficult. Diet, PA, smoking status, blood pressure, and cholesterol were widely reported, with varying results. In relation to diet, significant results were reported for the dietary intervention, one psychological, three educational, and six multifactorial interventions. Significant results relating to PA were reported for one exercise trial, one educational, two psychological, and seven multifactorial. Of the five studies which showed significant benefits for intervention patients compared to controls in relation to smoking, four were multifactorial and one psychological. All five interventions which showed significant results relating to BP were multifactorial. Concerning cholesterol and lipid levels, the interventions which showed significant results were one psychological, one educational, and eight multifactorial.

In relation to quality of life, one psychological trial reported a significant benefit for intervention patients compared to controls, as did one educational and two multifactorial.

Few trials reported medication intake or adherence. However, as appropriate therapy is a key aspect of secondary prevention of CHD [8], this should surely be given greater consideration. The trials which reported significant results relating to medication were classified as psychological (one), multifactorial (two), and educational (three).

Likewise, self-efficacy, a patient’s ability, and confidence to manage their own condition, was only reported in one study. This multifactorial intervention reported a significant outcome for intervention patients compared to controls.

4.2. Relevance of Lifestyle and Risk Factor Modification

Recent evidence has shown the importance of focusing on lifestyle to effect positive changes relating to CHD. Bennett et al. [48] used the IMPACT CHD model to examine the decline in CHD mortality in Ireland between 1985 and 2000, and possible reasons for this. The mortality rate fell during this period by some 47%, representing 3673 fewer observed CHD deaths. The authors found that 48% of this decrease was due to a reduction in major risk factors including smoking and cholesterol levels. Conversely, an upward trend was seen in obesity, diabetes, and PA which were said to contribute an additional 500 deaths in 2000. These results were similar to those found by Palmieri et al. [49], who also used the IMPACT model and investigated the decline in CHD mortality in Italy between 1980 and 2000. They concluded that over half of the mortality fall was due to risk factors, mainly blood pressure and cholesterol, and just under half was due to medical therapies.

The Lyon Diet Heart Study [33] was the first clinical trial evidence in support of the Mediterranean diet, which comprises a high intake of fruit, vegetables, nuts, legumes, and grains. Other recent evidence has been found supporting the Mediterranean diet’s protective effect in the secondary prevention of CHD [50–52]. O’Connor et al. [6] examined RCTs of cardiac rehabilitation with exercise and found a moderate reduction of 20% in total and CV mortality after one year, with a reduced risk maintained for three years after infarction. In a recent review of interventions incorporating exercise as part of a cardiac rehabilitation programme, Jolliffe et al. [11] also observed a reduction in total mortality. Exercise-only interventions resulted in a 27% reduction in total mortality and 31% reduction in cardiac mortality while comprehensive rehabilitation reduced mortality to a lesser extent (26%).

The psychological interventions included in this review showed small beneficial effects. This is in keeping with similar findings in a recent review of 36 RCTs of psychological interventions by Rees et al. [12] which showed no evidence of effect on total or cardiac mortality, but found small reductions in anxiety and depression.

We included two organisational interventions in this review, and both reported disappointing results. Jolly et al. [40] investigated a programme to improve communication between hospital and GP practices using liaison nurses. The programme was effective in promoting follow-up care of patients in general practice; however health outcomes were not improved. Munoz et al. [24] examined an intervention involving postal reminders sent to patients to encourage them to visit their GP. Blood pressure and HDL-cholesterol levels were improved, but no effect was observed on mortality or morbidity.

4.3. Limitations of the Review

This review has a number of potential limitations. The relatively small number of studies which included mortality as an outcome, the heterogeneity between trials, and poor quality of reporting all arise from the primary data. Further, imprecise descriptions of the interventions and the limited data have made it difficult to determine the benefits of various components of the interventions. The majority of the population of the interventions was male and relatively low risk; however greater benefit could have been derived from the interventions by higher risk patients who were excluded on the basis of comorbidity.

The broad range of trials included, and the subsequent large number discovered in our search, may have made comparisons difficult. Nonetheless this reflects the current lack of clarity in respect of the optimal components of effective interventions for CHD secondary prevention.

4.4. Implications for Practice

We have found that lifestyle interventions promoting regular PA, a healthy diet, and adherence to medication have beneficial effects among patients with CHD. It is therefore reasonable to promote such a healthy lifestyle to patients similar to those included in the RCTs we investigated—mainly older adults who had suffered a coronary event. As practitioners endeavour to achieve target levels of blood pressure and cholesterol by altering their patients’ prescribed medication, the potential value of their advice regarding exercise and diet should not be overlooked.

4.5. Implications for Future Research

Overall, the current evidence suggests that lifestyle interventions have some beneficial effects on total and cardiac mortality, morbidity, and on behaviour change in relation to modifiable cardiac risk factors. Even where little or no effect was observed relating to mortality or morbidity, some trials reported benefits in terms of lifestyle behaviour change. That healthcare education and even small-scale interventions can lead to healthier lifestyle choices, as shown in this review and others, should be an encouragement to professionals in practice. Future studies should be designed carefully, with attention given to aspects of study quality, which we addressed in Table 2. For example, RCTs of a cluster design help to avoid contamination of control patients. Further, outcomes should be matched to intervention elements. In addition, it is important to incorporate concealment of allocation and blinding of outcomes. Maximal follow-up should be ensured and more consideration given to the underlying theoretical foundation for the intervention.

However, the fact that more profound and wide reaching benefits were not seen in our review is surprising considering that all guidelines focus on the importance of lifestyle interventions. Future research should perhaps focus on the components of interventions and what the ideal combination of measures, intervention intensity, and duration should be. In addition, investigations into the barriers to lifestyle change among patients with CHD may shed new light on why some well-designed and executed interventions have not resulted in expected benefits.