The results of this study confirm the view that compliance and persistence with cardiovascular medication is poor, regardless of the method used for data collection (10
). In terms of compliance, patients filled 72% of prescriptions in the first year of treatment. Thus, almost 30% of days ‘on therapy’ were not actually covered by medication. Furthermore, only 59% of patients had medication for more than 80% of their days ‘on therapy’ in the year. The results of one study with a follow-up of more than 2 years showed that compliance decreases at first but then reaches a plateau (17
Persistence also decreased with time, but with wide variability. The reasons for the variability in persistence rates are unclear. The 10% point difference between European and North-American persistence rates may have been due to differences in the average follow-up time between the different studies. However, other factors are clearly involved. In a study comparing compliance with statin therapy in Italy and Denmark, 91% of patients remained persistent after 2 years in Denmark, but only 48% remained persistent at this time in Italy (18
). No single explanation for the difference emerged from the study, although differences in prescribing practices between the two countries could have played a role.
This systematic review confirmed previous findings that the definitions and measures of compliance vary considerably between studies (12
). However, there were some encouraging signs of a move towards using standard methodology, especially in retrospective analyses of pharmacy claims, in which MPR was the measure of choice in almost all cases. As studies move towards the use of a standard measure for compliance, it should become easier to compare estimates of compliance, enabling the influence of patient or regimen characteristics on compliance to be determined.
Many of the studies analysed in this review used the definition of MPR proposed by Steiner and Prochazka (16
). It has been argued that it is more informative to report the percentage of patients above a certain threshold of MPR rather than mean MPR (19
). However, the authors of this review believe that graphic presentation of the whole distribution as a histogram along with mean, SD or quintiles is preferable. Presenting compliance and persistence over time graphically is also very informative, as charted in 25 of the studies included in this review [see for example Sturkenboom et al. (20
The most important reason for investigating these issues is that poor compliance and lack of persistence with medications for hypertension, hyperlipidaemia and diabetes potentially lead to suboptimal health outcomes. Around one-third of studies in this review investigated the effect of compliance on outcomes, and the majority (73%) showed that compliance has a positive influence on outcome. Only one instance of a marginally negative effect of compliance on outcome (raised systolic blood pressure) was identified (21
) and only a few studies found compliance to have a neutral effect on outcome (22
). In addition, two of the studies reviewed (27
) found mortality rates to be lower in more compliant patients.
A number of recent studies have shown that patients who are compliant with therapy are likely to have better outcomes. In a meta-analysis, which included cardiovascular studies, a consistent association between good adherence to drug therapy and reduced mortality was found (29
). Similarly, increased compliance and persistence with long-term cardiovascular therapy have been shown to reduce the risk of adverse cardiovascular outcomes; compliant patients (MPR ≥ 80%) showing a reduced predicted relative risk of 4.6% for men and 16.4% for women (11
). Another study showed that patients who were highly compliant with antihypertensive therapy were 45% more likely to achieve blood pressure control than those with medium or low compliance (odds ratio 1.45, p = 0.026) (30
In another study investigating the relationship between persistence (defined as < 60 days gap between refills) with antihypertensive therapy and the risk of MI/stroke, multivariate analysis showed that persistent users were at significantly lower risk of MI/stroke than non-persistent patients (relative risk 0.88, 95% confidence interval: 0.82–0.94) (31
). Better compliance with antihypertensive therapy has also been shown to reduce the risk of hospitalisation (32
), while better compliance with antidiabetic medication has been shown to reduce emergency room visits by 26% over a 2- to 3-year period (33
Over the 5-year period of the present study, a trend towards more retrospective studies using data collected from pharmacy claims databases was seen. This was not surprising given that such studies take less time and money than prospective studies and potentially provide larger numbers of patients.
One limitation of the present review is that one person selected the studies and extracted the data. Thus, relevant studies may have been missed or incorrectly categorised. Another limitation is that although study characteristics and compliance measures were examined at the study level, differences between treatment arms within studies (e.g. different classes of AHTs, different doses) were not investigated. In studies with multiple treatment arms, population-weighted averages were used to calculate compliance. Meta-analysis was beyond the scope of this review. Because of the number of papers included in the review, study characteristics were presented only in tabular form. In addition, only means and SDs of compliance measures were calculated, without multivariate analysis. Finally, semi-transparency of definitions in some papers made it difficult to determine whether methods, and consequently compliance measures, were truly comparable.
Prospective real-world studies that use standard definitions and measures of compliance, and focus on objective outcomes, such as mortality, are needed to further our understanding of the issue of compliance. Large retrospective studies that analyse existing databases, identify appropriate stratification subgroups and use modelling exercises would also be useful.