Study selection and evaluation
Of the 1562 citations originally identified in our search, we retrieved 72 that seemed potentially relevant. We excluded 55 of these studies after detailed evaluation: 16 were not randomised, 13 were primary prevention studies, eight did not report the primary outcomes, eight evaluated interventions that were not comprehensive disease management systems (such as exercise therapy alone), five did not report the outcomes for patients with coronary heart disease separately or included <50% patients with coronary heart disease, two tested an inpatient based intervention, two enrolled fewer than 50 patients, and one had flawed methods (patients excluded after randomisation). (A full list of excluded studies is available from FM on request.) Our search retrieved 10 trials not included in previous systematic reviews.
Disagreement among the reviewers (FM and FL) regarding eligibility of the studies occurred on four occasions, for a κ value of 0.85. All disagreements were resolved by consensus.
Of the randomised trials eligible for inclusion, three were reported in more than one publication. One trial reported different end points in two separate publications.10,11
One trial reported the outcomes for all patients enrolled (only 45% of whom had heart disease) and, in a separate publication, provided details of event rates in the subgroup of patients with heart disease.12,13
The World Health Organization trial included 24 collaborating centres, but the original investigators excluded seven sites because of poor follow up of participants and four sites because of significant differences at baseline between the intervention and control arms.14
We included the three year outcome data from the remaining 13 sites as one trial for the purposes of this analysis, an approach validated by the non-significant tests for statistical heterogeneity for all cause mortality (Q=15.7, 11 df, P=0.16) and myocardial infarction (Q=15.9, 11 df, P=0.15) and the fact that the summary risk ratios for both end points were identical under the random and fixed effects models. Although the two Finnish centres in the WHO trial published their results separately (and for multiple periods of follow up), we included only their three year outcome data with the other 11 WHO sites for consistency of data presentation.15–17
Table presents summary data from the 12 randomised trials eligible for this systematic review. In all of the trials, patients randomised to the control groups received usual care (this was generally undefined).
Studies included in analysis of secondary prevention programmes in coronary heart disease
Quantitative data synthesis
Reinfarction rate—None of the seven trials reporting this end point detected a significant difference between patients in intervention and control arms (fig ), and the summary risk ratio for all 7480 patients was 0.94 (95% confidence interval 0.80 to 1.10).
Figure 1 Impact of interventions on recurrent myocardial infarctions. Data for all trials except that of Campbell et al10 are for the combined end point of non-fatal and fatal myocardial infarction. Campbell et al collected data only on non-fatal reinfarction (more ...)
All cause mortality—Only one out of 10 trials reported a significant survival benefit with the intervention (fig ). The summary risk ratio of 0.91 (0.79 to 1.04) for all 10 trials (9718 patients) confirms that these interventions have not been shown to improve survival.
Impact of interventions on all cause mortality
Rate of admission to hospital—
Although only two10,19
of the six trials reporting admission rates10,13,19,23,25,26
found a significant difference, the summary risk ratio of 0.84 (0.76 to 0.94) is consistent with a beneficial impact of the interventions. Of the four trials that evaluated length of stay, twoshowed shorter lengths of stay or a reduced total of days in hospital in the intervention group.13,26
Furthermore, one trial reported that significantly fewer patients in the intervention arms had multiple readmissions.13
There was insufficient detail in these studies to permit analysis by type of admission (for example, cardiac v
Sensitivity analyses—Year of study completion, duration of intervention, and length of follow up had no effect on the observed results (data not shown). Because of the small number of trials and the poor reporting of the specific elements of the intervention in many of the trials, analyses by different components of the multidisciplinary interventions failed to detect any one component that was statistically beneficial, although there was a trend towards greater survival benefits in those programmes that included structured exercise (risk ratio 0.87 (0.71 to 1.05) v 0.94 (0.78 to 1.13)).
Processes of care—Seven trials tested the impact of the disease management programmes on cardiovascular risk factors; five showed significantly greater improvements in patients randomised to the interventions, although the effect sizes were generally small to moderate (table ). Of the seven trials that assessed the use of drugs proved to be efficacious, all but two showed significantly increased prescription of at least one of these treatments in the intervention group. Pooled data showed that patients in the intervention arm were more likely to be prescribed these treatments, with risk ratios of 2.14 (1.92 to 2.38) for lipid lowering drugs, 1.19 (1.07 to 1.32) for β blockers, and 1.07 (1.03 to 1.11) for antiplatelet agents. The increases in prescription of β blockers and antiplatelet agents are noteworthy, as these trials tested interventions designed largely to look at lipid lowering in patients with coronary heart disease.
Impact of interventions on other end points
Other end points—
Five of the eight trials evaluating quality of life or functional status showed better scores in the intervention arms, although these were generally small and achieved significance in only three studies (table ). Only three of these trials described the costs of the intervention.12,20,26
Two reported that their intervention was cost saving,12,26
but none performed formal cost effectiveness analyses.