Discussion
Summary of Main Findings
We found that participation in the Heartwatch CVD prevention program, compared to non-participation, was associated with significantly greater improvements in CV risk factor profiles and use of secondary preventative therapies as well as significant reductions in the risks of the CV composite, CV death, and all-cause mortality on long-term follow-up.
Comparison With the Existing Literature
Previous cohort studies evaluating the Heartwatch program have reported significant reductions in blood pressure, favorable changes in lipid profiles, increased smoking cessation rates, and significant increases in the use of secondary preventative therapies such as lipid-lowering agents and antihypertensive medications. These benefits were maintained after 3.5 years of follow-up. Our study extends these findings by reporting an association between Heartwatch and the incidence of major vascular events, meaning that benefits in intermediate outcomes (risk factor modification) appear to translate into benefits for clinically meaningful outcomes. Our results are consistent with a meta-analysis of 63 randomized, controlled trials of similar secondary prevention programs for CVD that reported a significant reduction in the risk of recurrent myocardial infarction for participating patients (summary relative risk 0.83, 95% CI 0.74–0.94). However, participants in clinical trials may not fully represent patients in real-life clinical practice. Our findings provide evidence that the benefits reported in clinical trials are likely to be generalized to an unselected population of patients with CVD in the community.
The results of our study are particularly important in light of the recent findings of the PURE study, an epidemiological survey of the use of secondary preventative drugs in patients with a history of coronary artery disease or stroke, in both high-income and lower-income countries. In high-income countries, only 62% of patients with a previous history of coronary artery disease or stroke were on aspirin, and only 73.8% were on statin therapy, with even lower rates of β-blocker (40%) and ACE-inhibitor/angiotensin receptor blocker use (49.8%). Similar rates of prescription of these therapies were also observed in our cohort, aspirin (64.3%), lipid-lowering agents (60.7%), β-blockers (45.7%), and ACE-inhibitors (35.8%), with higher rates of prescription noted among Heartwatch participants ( Table II ). These results as well as those from other studies highlight the need for more effective strategies to increase uptake of effective CV therapies in high-risk patients and promote their sustained use. Heartwatch has been shown to be a cost-effective intervention (cost per life-year gained of 7,987) and is associated with a relatively low additional burden of work for general practices.
Strengths and Limitations of This Study
There are several limitations that require mention. First, this is an observational study and not a randomized, controlled trial, so we cannot make direct inferences about the effectiveness of the Heartwatch intervention. Second, only 60% of the practices randomly selected to participate in this study actually agreed to participate. Practices with fewer resources, such as single-handed practices or those without the support of ancillary staff, may have been less likely to participate in this study because of the associated time and cost constraints. This could potentially limit the generalizability of our results. However, we used stratified random sampling, which allowed a broad representation of patients and clinicians from routine clinical practice, which enhances the external validity of our results.
Third, although we used multivariable models to adjust for the effect of known confounders, our results are still subject to residual confounding from unmeasured or inadequately measured predictors of CV events. We were unable to adjust for the effect of obesity (body mass index) because there was a large amount of missing data for this variable. Furthermore, patient behaviors such as dietary habits and levels of physical activity were unmeasured, and so, we could not include them in our analysis. Patients and practices that agree to participate in Heartwatch may differ from those who do not participate. However, our analyses did adjust for the most important risk factors for recurrent major vascular events. In a sensitivity analysis that restricted the population to partnership general practices with a practice nurse, results were not materially altered, and a formal test for an interaction between Heartwatch and type of general practice or presence of a practice nurse was not significant, adding strength to our findings ( Table I ).
Fourth, some deaths may have been incorrectly classified as cardiac related. However, we do not suspect that such misclassification would have systematically biased our results as the cause of death was reported independently of our study and unlikely to be influenced by knowledge of whether the patients were participating in Heartwatch.
Finally, patients joined the Heartwatch program at varying times after the initial diagnosis of CVD, which may have resulted in variations in the uptake of secondary preventative therapies. In addition, patients participating in the Heartwatch program may have been followed up more closely than nonparticipants (surveillance bias), and thus, the detection of events may have been greater in this group. However, both of these factors would likely have biased our results toward the null hypothesis.
The strengths of our study include its minimal losses to follow-up, the broad representation of patients from routine clinical practice, the use of clinically meaningful outcome measures, and the use of multivariable model risk adjustment to control for the effect of a multitude of confounders.