Association of Statin Adherence With Mortality in Patients With Atherosclerotic Cardiovascular Disease

Key Points

Question  Is statin adherence associated with mortality in stable patients with atherosclerotic cardiovascular disease (ASCVD)?

Findings  In this cohort study of patients with ASCVD and overall high-statin adherence, we found a graded, inverse association between statin adherence and mortality. This association was observed across patient subgroups and by statin intensity.

Meaning  Clinicians should carefully track and promote adherence to statins in high-risk patients with ASCVD.

Importance  Statins decrease mortality in those with atherosclerotic cardiovascular disease (ASCVD), but statin adherence remains suboptimal.

Objective  To determine the association between statin adherence and mortality in patients with ASCVD who have stable statin prescriptions.

Design, Setting, and Participants  This retrospective cohort analysis included patients who were between ages 21 and 85 years and had 1 or more International Classification of Diseases, Ninth Revision, Clinical Modification codes for ASCVD on 2 or more dates in the previous 2 years without intensity changes to their statin prescription who were treated within the Veterans Affairs Health System between January 1, 2013, and April 2014.

Exposures  Statin adherence was defined by the medication possession ratio (MPR). Adherence levels were categorized as an MPR of less than 50%, 50% to 69%, 70% to 89%, and 90% or greater. For dichotomous analyses, adherence was defined as an MPR of 80% or greater.

Main Outcomes and Measures  The primary outcome was death of all causes adjusted for demographic and clinical characteristics, as well as adherence to other cardiac medications.

Results  Of 347 104 eligible adults with ASCVD who had stable statin prescriptions, 5472 (1.6%) were women, 284 150 (81.9%) were white, 36 208 (10.4%) were African American, 16 323 (4.7%) were Hispanic, 4093 (1.2%) were Pacific Islander, 1293 (0.4%) were Native American, 1145 (0.3%) were Asian, and 1794 (0.5%) were other races. Patients taking moderate-intensity statin therapy were more adherent than patients taking high-intensity statin therapy (odds ratio [OR], 1.18; 95% CI, 1.16-1.20). Women were less adherent (OR, 0.89; 95% CI, 0.84-0.94), as were minority groups. Younger and older patients were less likely to be adherent compared with adults aged 65 to 74 years. During a mean (SD) of 2.9 (0.8) years of follow-up, there were 85 930 deaths (24.8%). Compared with the most adherent patients (MPR ≥ 90%), patients with an MPR of less than 50% had a hazard ratio (HR; adjusted for clinical characteristics and adherence to other cardiac medications) of 1.30 (95% CI, 1.27-1.34), those with an MPR of 50% to 69% had an HR of 1.21 (95% CI, 1.18-1.24), and those with an MPR of 70% to 89% had an HR of 1.08 (95% CI, 1.06-1.09).

Conclusions and Relevance  Using a national sample of Veterans Affairs patients with ASCVD, we found that a low adherence to statin therapy was associated with a greater risk of dying. Women, minorities, younger adults, and older adults were less likely to adhere to statins. Our findings underscore the importance of finding methods to improve adherence.

Given the evidence of the effectiveness of statins in reducing atherosclerotic cardiovascular disease (ASCVD) events and mortality, the 2013 American College of Cardiology (ACC) and the American Heart Association (AHA) cholesterol treatment guidelines recommended high-intensity statins for most patients with ASCVD.¹ The recently updated 2018 cholesterol guidelines similarly recommend statins as the mainstay of treatment for patients with ASCVD and emphasize the importance of tracking adherence in routine follow-up.² Our prior work documented that using the highest statin intensity is associated with a survival benefit in patients with ASCVD.^3,4 However, many high-risk patients do not maintain statin use (of any intensity) beyond the initial ASCVD event.^5-7 In fact, contemporary registries reveal a substantial underuse and underdosing of statins, even following an acute coronary syndrome.^7-9 Although high-intensity statin use has increased following the release of the ACC/AHA cholesterol treatment guidelines, disparities persist for certain patient groups, including older adults, women, and minorities.¹⁰

The failure to prescribe and intensify statins is not the only obstacle to improving guideline adherence. Even in well-conducted randomized clinical trials, as many as a third of high-risk patients who were prescribed statins discontinue statin therapy, with most of the discontinuations attributable to patient preference and not to drug adverse effects.^11,12 In real-world clinical registries, adherence to statin therapy is even lower, with adherence rates less than 50% at 1 year after starting therapy, declining to 30% at 2 years.^7,13 African American and Hispanic women are particularly susceptible to discontinuing statin therapy following a myocardial infarction.^14,15

The reasons for statin low adherence are complex and may include patient, clinician, and health care system factors that have not been well characterized. Understanding the consequences of low adherence may motivate the discovery and implementation of strategies to increase the appropriate use of statins. Additionally, to our knowledge, most studies on statin adherence have studied patients immediately following their ASCVD event or hospitalization during which they receive a new statin prescription. Thus, the objective of this study was to determine the association between statin adherence and all-cause mortality for patients with ASCVD who had a stable statin prescription. Second, we sought to determine if the association between adherence and mortality was modified by statin intensity or by patient-level or system-level characteristics.

Our study population included all outpatients within the Veterans Affairs (VA) Health System between age 18 and 85 years with established ASCVD, defined as coronary artery disease (International Classification of Diseases, Ninth Revision [ICD-9] codes 410-414), cerebrovascular disease (ICD-9: 430-438), or peripheral artery disease (ICD-9: 440) that was identified between January 1, 2013, and April 1, 2014, with at least 1 ASCVD code on at least 2 different dates and at least 1 outpatient visit in the VA system during the study period during the previous 2 years. We identified quarterly index dates during the study period (January 1, 2013; April 1, 2013; July 1, 2013; October 1, 2013; and January 1, 2014). We included patients if they filled at least 1 prescription for statins in the 6 months before and the 6 months following the index date. Thus, patients were excluded if they filled no VA medications during 6 months after the index date to limit patients who received most of their care outside of the VA system. We were interested in a prevalent cohort and thus excluded patients who did not have a statin prescription between 5 years and 6 months before the index date (18%). We also excluded patients who underwent dose reduction (27 093 [6%]) or intensification (25 344 [5%]) during the 6 months following the index date. The study was approved by the Stanford institutional review board for human subjects, which waived the need for patient consent.

Statin intensity was classified according to the ACC/AHA cholesterol treatment guidelines.^1,2 Statin use was defined as a filled prescription during the previous 6 months. Low-intensity statin therapy was defined as fluvastatin, 20 to 40 mg; lovastatin, 20 mg; simvastatin, 10 mg; pitavastatin, 1 mg; and pravastatin, 10 to 20 mg. Moderate-intensity statin therapy was defined as atorvastatin, 10 to 20 mg; fluvastatin, 40 mg twice daily or 80 mg once daily (extended-release formulation); lovastatin, 40 mg; pitavastatin, 2 to 4 mg; pravastatin, 40 to 80 mg; rosuvastatin, 5 to 10 mg; and simvastatin, 20 to 40 mg. High-intensity statin therapy was defined as atorvastatin, 40 to 80 mg, or rosuvastatin, 20 to 40 mg. While there were no new prescriptions for simvastatin, 80 mg, those patients who continued to take this dose were included in the high-intensity statin group because it typically lowers low-density lipoprotein cholesterol (LDL-C) levels by nearly 50%.

Medication adherence was measured by the medication possession ratio (MPR). The MPR is calculated by dividing the number of days of outpatient statin supplied during a 12-month period by the number of days that the patient was not hospitalized and alive during the 12-month period. To account for potential stockpiling from multiple prescriptions, MPR was capped at 100%. Adherence was categorized with MPR values less than 50%, 50% to 69%, 70% to 89%, and 90% or greater. For dichotomous analyses, adherence was defined by previous literature as an MPR of 80% or greater.^16,17

The primary outcome was all-cause mortality. Secondary outcomes were 1-year mortality, 1-year hospitalization for ischemic heart disease, or ischemic stroke. In a sensitivity analysis, we also determined the association between statin adherence and hospitalizations for gastrointestinal bleeding and pneumonia. All-cause mortality during follow-up was extracted from the VA Vital Status file. Outcomes were determined following the end of the 1-year period to determine adherence.

Race/ethnicity was self-reported in the VA Health System Records. Participants were classified as non-Hispanic white, Hispanic, African American, Asian, Pacific Islander, other race, or missing (2093 [0.6%]). Clinical comorbidities included a diagnosis (an encounter during the prior 2 years) of heart failure, diabetes, chronic obstructive pulmonary disease, hypertension, ischemic cerebrovascular disease, peripheral arterial disease, malignancy, and renal disease. Values for LDL-C levels and serum creatinine and vital signs (systolic blood pressure, diastolic blood pressure, pulse, respirations, pulse oximetry, and weight) were included if performed within the 6 months before the index date. A follow-up 1-year LDL-C level was identified if performed between 6 and 18 months following the end of the 1-year period to determine adherence. This LDL-C level was classified as more than 150 mg/dL (to convert to millimoles per liter, multiply by 0.0259), 101 to 150 mg/dL, 71 to 100 mg/dL, 51 to 70 mg/dL, 50 mg/dL or less, and missing.

Age was categorized into several groups: 35 years or younger, 35 to 44 years, 45 to 54 years, 55 to 64 years, 65 to 74 years, 75 to 84 years, and 85 years or older. For the regression models, the group aged 65 to 74 years was used as the reference because this was the most adherent group and included the median age of the study cohort. Adults younger than 75 years and those 75 years and older were also examined in a subgroup analysis on the association of statin adherence with mortality.

Hospital characteristics were obtained from the American Hospital Association database. These characteristics included geographic region (divided into Northeast, Midwest, South, and West) and academic teaching status defined as membership in the Council of Teaching Hospitals and Health Systems.

Baseline characteristics of participants were compared by MPR adherence categories (<50%, 50%-69%, 70%-89%, and ≥90%) in unadjusted analyses using either analysis of variance for continuous variables or the Mantel-Haenzsel χ² test for trend. For multivariate analyses, we imputed continuous variables using the mean and most common value for rare (<1%) missing categorical variables. For nonrare missing categorical variables (Council of Teaching Hospitals and Health Systems in 6%), we created a separate missing category.

We used a logistic regression model to identify predictors of adherence, defined as an MPR of 80% or greater. To determine the association of adherence with survival, we first performed an unadjusted proportional hazards analysis. To attempt to control for the “healthy adherer” effect, we adjusted for adherence to β-blockers and angiotensin-converting enzyme inhibitors. We then performed a fully adjusted model that controlled for all patient and facility characteristics described previously.

In sensitivity analyses we also adjusted for follow-up LDL-C levels and used a cohort with a first prescription for statins during the 6 months before the index date. We also determined the propensity to adhere to statins (MPR ≥ 80%) using the logistic regression model defined previously. In a sensitivity analysis, the proportional hazards analysis was repeated using inverse probability weighting with the propensity to adhere (MPR ≥ 80%). We adjusted for the clustering of patients within facilities. Because this had little change in the confidence intervals and no change (to within 2 significant digits) for the mortality hazard ratio (HR), the primary results are shown without an adjustment for clustering. All statistical testing was 2-sided at a significance level of P < .05. Analyses were performed using SAS statistical software (version 9.4; SAS Institute Inc).

We identified 487 812 adults with documented ASCVD in the cohort with filled prescriptions for statins at baseline and during follow-up. We excluded patients whose baseline prescription was a new prescription (88 271 [15%]) and those who had deintensification (27 093 [6%]) or intensification (25 433 [5%]) of their statin dose between baseline and follow-up, leaving a population for analysis of 347 104.

Overall, the mean (SD) statin adherence in this population taking a stable statin intensity was 87.7% (17%). Table 1 provides characteristics of the patients by statin adherence levels. This cohort included 87 524 patients (25%) taking high-intensity, 217 570 (63%) taking moderate-intensity, and 42 010 (12%) taking low-intensity statin therapy (Table 1). Patients taking moderate-intensity statin therapy were slightly more likely to adhere compared with patients taking low-intensity or high-intensity statins. More adherent patients had lower LDL-C values (77.2 mg/dL for MPR ≥90% compared with 92.1 mg/dL for MPR <50%).

Table 2 shows the characteristics that are independently associated with statin adherence (MPR ≥ 80%). Women were significantly less likely to adhere to statins compared with men. All minority groups were less adherent compared with non-Hispanic white patients. Patients with peripheral artery disease and cerebrovascular disease were less adherent to statins than those with coronary artery disease. Younger and older patients were less likely to adhere compared with patients aged 65 to 74 years. Patients with clinical comorbidities, including heart failure, diabetes, and hypertension, were more likely to adhere, as were those who were hospitalized within the year before the index date. Statin adherence was significantly associated with lower systolic and diastolic blood pressure and higher weight.

In the year following the index date, ischemic heart disease or stroke hospitalizations in the VA system were more frequent in those who were less adherent to statins. The proportion of patients with a hospitalization for ischemic heart disease or ischemic stroke was 13.4% (n = 2653) for an MPR of less than 50%, 13.1% (n = 4018) for an MPR of 50% to 69%, 11.5% (n = 8729) for an MPR of 70% to 89%, and 11.5% (n = 25 434) for an MPR of 90% or greater (P < .001). The association remained significant after adjustment for baseline characteristics (MPR <50: odds ratio [OR], 1.08; 95% CI, 1.03-1.14; MPR, 50-69: OR, 1.02; 95% CI, 0.99-1.07; MPR, 70-89; OR, 1.09; 95% CI, 1.06-1.12; all P < .001). In contrast, the association between statin MPR and 1-year hospitalization for gastrointestinal bleeding was nonsignificant (156 [0.79%] for MPR <50%, 232 [0.76%] for MPR 50%-69%, 557 [0.73%] for MPR 70%-89%, and 1630 [0.74%] for MPR ≥90%; P = .45). Similarly, MPR was not significantly associated with 1-year hospitalization for pneumonia (0.32% for MPR <50%, 0.34% for MPR 50%-69%, 0.25% for MPR 70%-89%, and 0.31% for MPR ≥90%; P = .45). The association remained nonsignificant between statin MPR and hospitalization for gastrointestinal bleeding and pneumonia after adjustment for patient and facility characteristics (data not shown).

There were 85 930 deaths (24.8%) during a mean (SD) of 2.9 (0.8) years of follow-up. During the first year there were 21 396 deaths (6.2%), which was associated with adherence levels (8.8% 1-year mortality for MPR <50%, 7.5% for MPR 50%-69%, 6.3% for MPR 70%-89%, and 5.7% for MPR ≥90%; P < .001). In an unadjusted survival analyses, patients with the lowest adherence levels (MPR <50%) had an HR for mortality of 1.36 (95% CI, 1.34-1.38) as compared with the most adherent patients (Table 3). Adjusting for adherence to other cardiac medications attenuated the effect size but did not change the dose-response association between adherence and mortality. However, additional adjustments for patient characteristics strengthened the association between statin low adherence and mortality (Table 3). When LDL-C levels during follow-up were added as a categorical variable (including a category for missing), the effect size was attenuated but remained significant.

There was a graded association between statin adherence levels and mortality (Figure). The graded association between statin low adherence and mortality was greatest for those taking high-intensity statins (Table 4). The association between adherence and mortality was also noted for subgroups defined by age, sex, and race/ethnicity. The association between low adherence and high mortality appeared stronger for men than women and for non-Hispanic white than African American or Hispanic patients. When those who were not taking statins during follow-up were included, this group had the highest HR for mortality compared with the most adherence group (1.46; 95% CI, 1.42-1.50).

A sensitivity analysis that adjusted for the propensity to adhere to statins using inverse probability weighting showed a similar association between low adherence and mortality. After adjusting for the propensity to adhere (MPR ≥80%), not adhering to statins (MPR <80%) was associated with an HR for mortality of 1.23 (95% CI, 1.22-1.25). Adjusting for a potential clustering of patients within facilities had a minimal effect on the confidence interval for the association between adherence and mortality (HR, 1.23; 95% CI, 1.21-1.26).

In this large, national study of VA patients with ASCVD, we found overall high rates of statin adherence in outpatients taking stable statin doses during an average of 3 years of follow-up. This differs markedly from patients receiving a first prescription, for which prior studies have documented poor adherence. However, few had low adherence and we documented an inverse dose-response relationship between statin adherence and mortality. This association was strongest for those taking high-intensity statins. Predictors of low adherence included female sex, nonwhite race/ethnicity, and the youngest and oldest age groups.

Although statins are among the most effective drugs for the secondary prevention of ASCVD, low adherence is a common problem.^4,5,18 Adherence to secondary prevention medications has been shown to improve outcomes, decrease mortality, and decrease readmissions.^19-21 To our knowledge, few studies have specifically examined the effects of statin adherence by intensity and survival in real-world, nonclinical trial populations that extend beyond the initial statin prescription following an acute event. Our findings complement and extend prior work that has shown that discontinuing statins is associated with a higher risk of adverse cardiovascular events, including recurrent stroke and myocardial infarction.^6,22 As expected, low adherence was associated with higher LDL-C levels and higher rates of hospitalization for stroke and ischemic heart disease.

However, high adherence itself may be a marker for overall healthy behavior, a term known as the healthy adherer effect.²³ In fact, high adherence to placebo has been associated with improved outcomes after myocardial infarction and lower cardiovascular mortality.^24,25 It is possible that patients with low adherence are also likely to engage in detrimental health behaviors, such as smoking, or have poor psychosocial support. To account for this, we controlled for adherence to other cardiac drugs (β-blockers and angiotensin-converting enzyme inhibitors) as well as blood pressure, which was higher among those with lower adherence. While this attenuated the association, there remained a strong dose-response association between the adherence to statins and survival. Although this healthy adherer effect may partially confound the association between statin adherence and overall mortality, it can help clinicians identify patients who are more likely to benefit from statin adherence and maintaining a healthy lifestyle.

The finding that women and minority groups were less likely to adhere to statins is also noteworthy and consistent with prior literature. Peters et al²⁶ documented that women are less likely to fill a prescription for a high-intensity statin following a hospitalization for acute myocardial infarction. Prior work has also shown that women underuse statins and other secondary prevention medications for ASCVD compared with men.^27,28 It is possible that women may experience more statin adverse effects, such as myalgias,²⁹ or that women are perceived (or perceive themselves to be) at lower risk compared with men.³⁰ Importantly, statins have been shown to be equally effective in women in reducing cardiovascular events and all-cause mortality.^29,31 Similarly, studies have documented that nonwhite patients are 50% less likely to adhere to statins,³² but the reasons for this disparity have not been fully elucidated. One study from the Patient and Provider Assessment of Lipid Management Registry showed that African Americans were less likely to receive guideline-recommended statin therapy and that this may be partly explained by negative beliefs about statin therapy.³³ We found that low adherence was greatest for African American patients, followed by Hispanic and Asian patients. Minorities have not been well represented in statin-related trials and more work is needed to implement strategies to improve guideline adherence in these populations.

Our study has several important clinical implications. The graded dose-response associations between statin adherence and survival shows the benefit of strict statin adherence. Until now, most studies have focused on short-term adherence to a first prescription following an acute event (ie, myocardial infarction or stroke) with adherence decreasing after the first 6 months of treatment. We show that even in a relatively adherent prevalent cohort of stable patients with ASCVD who received consistent statin prescriptions, a significant minority do not refill them as recommended, and this reduced adherence had clinical implications. This association with mortality was small but significant for those with adherence that some may consider reasonable (MPR, 70%-89%). Lack of adherence had the greatest association with mortality for those taking high-intensity statins, likely because high-intensity statin therapy confers a greater survival benefit than lower-intensity statin regimens.³ Our findings suggest that statin adherence is an important metric that should be carefully tracked by clinicians and health systems. This is now emphasized in the 2018 updated American College of Cardiology/American Heart Association cholesterol guidelines that recommend that clinicians routinely monitor adherence by checking LDL-C levels at specific intervals and regularly engage in discussions with patients about adhering to pharmacotherapy and healthy lifestyles.²

Our study has several strengths including the use of a large, national, well-characterized patient cohort at the VA, which offers an opportunity to study administrative and clinical variables, including laboratory values and vital signs. Unlike prior studies that have focused on short-term statin adherence following acute coronary syndromes or stroke, we included patients taking stable statin intensities with longer follow-up. Several participants died during follow-up (nearly 25% of the sample), which gave us the power to detect differences in outcomes by statin adherence level. Our results should also be interpreted in the context of several limitations. Because of the observational nature of our study, we could not fully account for residual confounding that may explain the association between low adherence to statins and mortality. We were also unable to determine the cause of death. Unlike mortality in which multiple government sources of data were available, we were unable to capture hospitalizations that occurred outside of the VA system. Medication possession ratios were used as a proxy for adherence; however, as with all pharmacy databases, we can only determine if the prescription was dispensed and not if the patient actually took the medication. This limitation would bias the results toward the null, suggesting that the association between adherence and mortality that we observed was a lower estimate. Medication possession ratios have a high specificity for medication adherence and are widely used in the literature.³⁴ The association of higher MPRs with lower LDL-C levels in our sample supports the validity of this metric. Because we relied on administrative codes for diagnosing ASCVD, coding errors may have affected our findings, although these errors are likely to be nondifferential by statin adherence group.

We found an inverse, graded association between long-term statin adherence and all-cause mortality in a national sample of VA patients with ASCVD. These findings suggest there is a substantial opportunity for improvement in the secondary prevention of ASCVD through optimization of statin adherence.

Accepted for Publication: December 17, 2018.

Corresponding Author: Fatima Rodriguez, MD, MPH, Division of Cardiovascular Medicine, Stanford University, 870 Quarry Rd, Falk CVRC, Stanford, CA 94305-5406 (frodrigu@stanford.edu).

Published Online: February 13, 2019. doi:10.1001/jamacardio.2018.4936

Author Contributions: Dr Heidenreich had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Knowles, Virani, Maron, Heidenreich.

Study concept and design: Rodriguez.

Acquisition, analysis, or interpretation of data: Rodriguez, Maron, Lin, Heidenreich.

Drafting of the manuscript: Rodriguez.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Lin, Heidenreich.

Obtained funding: Heidenreich.

Administrative, technical, or material support: Rodriguez, Lin.

Supervision: Heidenreich.

Conflict of Interest Disclosures: Dr Virani reported grants from Department of Veterans Affairs, honorarium from the National Lipid Association, and serving as a member of the steering committee for the Patient Provider Assessment of Lipid Management Registry for Duke Clinical Research Institute during the conduct of the study and honorarium from PRIME CME outside the submitted work. No other disclosures were reported.

Funding/Support: The study was conducted using departmental funds. Dr Knowles received funding from the Doris Duke Charitable Trust. Dr Virani has received research funding from the Department of Veterans Affairs Health Services Research and Development Service.

Role of the Funder/Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The views expressed in this article represent those of the authors and do not necessarily represent the views of the US Department of Veterans Affairs.