Rates are during follow-up among cohort members with dyslipidemia eligible for lipid-lowering treatment as defined by National Cholesterol Education Panel Adult Treatment Panel III risk-based criteria who had no known prior statin use, overall, and stratified by the presence or absence of known coronary heart disease (CHD) at study entry. P<.001 for all comparisons of statin therapy vs no statin therapy. Error bars indicate 95% confidence intervals.
Go AS, Lee WY, Yang J, Lo JC, Gurwitz JH. Statin Therapy and Risks for Death and Hospitalization in Chronic Heart Failure. JAMA. 2006;296(17):2105–2111. doi:10.1001/jama.296.17.2105
Author Affiliations: Division of Research, Kaiser Permanente of Northern California, Oakland (Drs Go and Lo and Ms Yang); Departments of Epidemiology, Biostatistics, and Medicine, University of California, San Francisco (Dr Go); Department of Medicine, Tufts School of Medicine, Boston, Mass (Ms Lee); Division of Endocrinology, Department of Medicine, San Francisco General Hospital, University of California, San Francisco (Dr Lo); and Meyers Primary Care Institute, University of Massachusetts Medical School, Fallon Clinic Foundation and Fallon Community Health Plan, Worcester (Dr Gurwitz).
Context Whether statin therapy has beneficial effects on clinical outcomes in patients with heart failure is unclear.
Objective To evaluate the association between initiation of statin therapy and risks for death and hospitalization among adults with chronic heart failure.
Design, Setting, and Patients Propensity-adjusted cohort study of adults diagnosed with heart failure who were eligible for lipid-lowering therapy but had no previous known statin use, within an integrated health care delivery system in northern California between January 1, 1996, and December 31, 2004. Statin use was estimated from filled outpatient prescriptions in pharmacy databases.
Main Outcome Measures All-cause death and hospitalization for heart failure during a median of 2.4 years of follow-up. We examined the independent relationships between statin therapy and risks for adverse events overall and stratified by the presence or absence of coronary heart disease after multivariable adjustment for potential confounders.
Results Among 24 598 adults diagnosed with heart failure who had no prior statin use, those initiating statin therapy (n = 12 648; 51.4%) were more likely to be younger, male, and have known cardiovascular disease, diabetes, and hypertension. There were 8235 patients who died. Using an intent-to-treat approach, incident statin use was associated with lower risks of death (age- and sex-adjusted rate of 14.5 per 100 person-years with statin therapy vs 25.3 per 100 person-years without statin therapy; adjusted hazard ratio, 0.76 [95% confidence interval, 0.72-0.80]) and hospitalization for heart failure (age- and sex-adjusted rate of 21.9 per 100 person-years with statin therapy vs 31.1 per 100 person-years without statin therapy; adjusted hazard ratio, 0.79 [95% confidence interval, 0.74-0.85]) even after adjustment for the propensity to take statins, cholesterol level, use of other cardiovascular medications, and other potential confounders. Incident statin use was associated with lower adjusted risks of adverse outcomes in patients with or without known coronary heart disease.
Conclusion Among adults diagnosed with heart failure who had no prior statin use, incident statin use was independently associated with lower risks of death and hospitalization among patients with or without coronary heart disease.
Randomized controlled trials have proven that hydroxymethyl glutaryl coenzyme A reductase inhibitors (statins) are highly effective for prevention of atherosclerotic vascular events and death.1- 4 However, patients with heart failure were largely excluded from these trials. Despite therapeutic advances, outcomes for the growing number of predominantly elderly patients with heart failure remain poor,5,6 and additional strategies are needed to reduce the high rate of death and morbidity in these patients.
Statins effectively lower low-density lipoprotein cholesterol level. However, increasing attention has focused on other potentially favorable “pleotropic” effects that may apply in the setting of heart failure.7 For example, statins may induce angiogenesis by recruiting bone marrow stem cells,8 reduce levels of inflammatory factors, and improve endothelial function.9,10 On the other hand, epidemiological studies have observed a higher risk of adverse events with low levels of low-density lipoprotein cholesterol in persons with heart failure.11,12 Statins may diminish the ability of lipoproteins to bind endotoxins leading to stimulation of proinflammatory cytokines13; reduce levels of coenzyme Q1014 and selenoproteins,15 which could adversely affect cardiac muscle and function; and have deleterious interactions with medications commonly used for heart failure, such as digoxin.16
Therefore, we examined the association of incident statin use on the risks of death and hospitalization within a large, diverse population of patients with heart failure who were considered eligible for lipid-lowering therapy.
The Kaiser Permanente Chronic Heart Failure cohort included all adults (age ≥20 years) diagnosed with heart failure between January 1, 1996, and December 31, 2004, within Kaiser Permanente of Northern California, a large integrated health care delivery system. Using previously described methods,17 patients were included if they met 1 or more of the following criteria based on a diagnosis of heart failure found in health plan databases: 1 or more hospitalizations with a principal diagnosis of heart failure (International Classification of Diseases, Ninth Edition [ICD-9] codes 398.91, 402.01, 402.11, 402.91, 428.0, 428.1, or 428.9); 2 hospitalizations with a secondary diagnosis of heart failure in which the principal diagnosis is related to the disease (eg, coronary heart disease [CHD]); 3 or more hospitalizations with secondary diagnosis of heart failure; 2 or more outpatient diagnoses; 3 or more emergency department visit diagnoses; or 2 or more inpatient secondary diagnoses plus 1 outpatient diagnosis. The index date was assigned at the first qualifying diagnosis. Medical record review of 9533 patients meeting these criteria confirmed serial physician-assigned heart failure diagnoses in 97% of cases. We previously showed that 96% of patients with a primary discharge diagnosis of heart failure18 were confirmed by chart review using Framingham clinical criteria.19
Given concern for treatment selection bias with inclusion of prevalent statin users20 and those who did not meet criteria for lipid-lowering therapy, we focused on the association of incident statin use during follow-up with outcomes only in patients who were not receiving statin therapy at the study entry date and who were eligible for treatment based on national guidelines.21 The Kaiser Foundation Research Institute's institutional review board approved the study and waived the requirement for written informed consent.
Receipt of any statin during the 120 days before the index date and throughout follow-up was identified based on filled prescriptions found in pharmacy databases. We also controlled for use of angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, diuretics, β-blockers, calcium channel blockers, other lipid-lowering agents, spironolactone, direct vasodilators, and recombinant erythropoietin. Using previously described methods,17 longitudinal medication use was estimated from drug refill patterns using the calculated day supply for each prescription. For any 2 consecutive prescriptions, the patient was classified as continually taking the medication if the second prescription was filled within 30 days of the projected end date of the first. If the second prescription was filled more than 30 days after the projected end date of the first, the patient was classified as not taking the medication from day 31 until the start date of the next prescription. If 2 prescriptions for the same drug were filled on the same day, we used the prescription with the longest calculated day supply to determine the projected end date. If a nonfatal hospitalization occurred during follow-up, the length of stay (in days) was added to the estimated day supply for any prescription crossing that hospitalization because patients were unlikely to take their own medications while hospitalized.
Age, sex, and self-reported race/ethnicity were identified from health plan databases. Race/ethnicity was included because studies suggest it may be associated with differential treatment or outcomes in cardiovascular diseases. Socioeconomic status was estimated from 2000 US Census data. Low education was defined as living in a census block where more than 25% of those aged 25 years or older had less than a 12th-grade education; low income was defined as living in a block where annual household income is less than $35 000 per year.22 We ascertained information on coexisting illnesses based on diagnoses or procedures using ICD-9 codes, laboratory results, or specific therapies from health plan hospitalization discharge, ambulatory visit, laboratory, and pharmacy databases; diabetes mellitus registry23; and regional cancer registry.24 This included baseline and follow-up diagnoses of CHD, cerebrovascular disease, peripheral arterial disease, diabetes, hypertension, malignancy, thyroid disease, liver disease, lung disease, human immunodeficiency virus infection, valvular disease, dementia, depression, ventricular arrhythmias, and atrial fibrillation/flutter (ICD-9 codes available on request).
We also identified outpatient measurements of total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, and hemoglobin17 from laboratory databases during the 12 months before study entry and throughout follow-up. We classified kidney function using the Modification of Diet in Renal Disease equation for estimated glomerular filtration rate based on outpatient determinations of serum creatinine.25,26 We categorized glomerular filtration rate (in units of mL/min per 1.73 m2) as 60 or greater, 45 to 59, 30 to 44, 15 to 29, less than 15 not requiring dialysis, and receiving maintenance dialysis.22 Categories of 60 mL/min per 1.73 m2 or greater were not delineated given the higher degree of error in estimating the glomerular filtration rate above this level.27,28
We ascertained information on systolic function status from health plan databases. If available, we used a previously described approach18 to define reduced systolic function as left ventricular ejection fraction of less than 40% or a qualitative description of moderate or severely reduced systolic function; preserved left ventricular systolic function was defined by left ventricular ejection fraction of 40% or higher or a qualitative description of normal or only mildly reduced systolic function.
Finally, as a proxy for intensity of care, we identified the number of visits to a cardiologist before study entry and during follow-up. Information on participation in heart failure case management programs was not available.
Patients were censored at health plan disenrollment or end of follow-up on December 31, 2004. Disenrollment was defined as a gap in membership of 90 days or longer. Primary outcomes included death from any cause and hospitalization for heart failure. Deaths were identified from health plan databases for proxy report and deaths that occurred in the emergency department or hospital, Social Security Administration files,29 and California mortality files30 through 2004, which were the most recent data available at the time of the analysis. Hospitalization for heart failure was based on primary discharge diagnoses (ICD-9 codes 398.91, 402.01, 402.11, 402.91, 428.0, 428.1, and 428.9) found in hospital discharge and billing claims databases.
Analyses were performed using SAS software version 9.13 (SAS Institute Inc, Cary, NC), and a 2-sided P value of less than .05 was considered significant. We compared baseline characteristics among all patients during follow-up using the t test or Wilcoxon rank sum test for continuous variables and the χ2 test for categorical variables. Age- and sex-adjusted rates of death and hospitalization were calculated using Poisson regression with generalized estimating equations to account for repeated measures within individuals.
We examined the independent association between incident statin use and outcomes using extended Cox regression with time-dependent covariates. Because certain patients may stop therapies toward the end of life, which could yield a potentially overly optimistic estimate of treatment effects using a time-varying definition of drug exposure, our primary approach was similar to the intent-to-treat method in which treatment-eligible persons who initiated statin therapy during follow-up were considered to always be taking statins through the follow-up period even if they did not refill prescriptions. A secondary analysis incorporated time-varying estimates of statin therapy and assigned exposure status at the time of an outcome event based on our medication exposure algorithm. Stratified models were also performed in patients with or without known CHD at study entry given the known benefit of statins for secondary prevention.
We also adjusted for the likelihood of receiving statin therapy by using a continuous propensity score.31,32 Our propensity score logistic model (c statistic = 0.79) considered candidate variables shown in Table 1, Table 2, and Table 3. Additional variables selected for the final models included those known to be associated with the outcomes of interest, as well as any covariates that differed between patients in the 2 treatment groups during follow-up at a P value of less than .05. In analyzing hospitalizations for heart failure, a robust sandwich variance estimator was applied in the calculation of the 95% confidence intervals (CIs) to account for multiple hospitalizations by the same individual.33
We identified 24 598 adults with diagnosed heart failure who had no prior known statin use and were considered eligible for lipid-lowering therapy. During follow-up, 12 648 (51.4%) initiated statin therapy. Patients initiating statin therapy were more likely than those not initiating statin therapy to be younger and male, with no clinically relevant differences in the distribution of race/ethnicity, education, or income level (Table 1). The most notable differences in other baseline characteristics between patients who did or did not initiate statin therapy were a higher prevalence of known CHD, diabetes, and hypertension in those initiating statin therapy (Table 2). Patients initiating statin therapy were also more likely to undergo testing for left ventricular function but there was no significant difference in the prevalence of reduced systolic function among tested patients (Table 3). Those initiating statin therapy had more visits to a cardiologist before and after study entry and had higher levels of total cholesterol and low-density lipoprotein cholesterol but no clinically relevant differences in levels of high-density lipoprotein cholesterol or hemoglobin. Baseline use of angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers, calcium channel blockers, β-blockers, other lipid-lowering drugs, vasodilators, and nitrates was higher among patients who initiated statin therapy (Table 3).
Median follow up was 2.4 years (interquartile range, 1.0-4.3 years), during which time there were 8235 patients who died and 9215 who were hospitalized for heart failure. The age- and sex-adjusted rate of death was lower with statin therapy (14.5 per 100 person-years) vs without statin therapy (25.3 per 100 person-years) (P<.001; Figure). The lower rate of death associated with statin therapy was observed in the presence or absence of known CHD. The age- and sex-adjusted rate of hospitalization for heart failure was also lower with statin therapy (21.9 per 100 person-years) vs without statin therapy (31.1 per 100 person-years; P<.001), and seen in the presence or absence of known CHD (Figure).
In our primary analysis using an intent-to-treat approach, incident statin use was associated with a 24% lower relative risk (RR) of death compared with not taking a statin, even after adjustment for sociodemographic characteristics, comorbidities, longitudinal use of other therapies, and propensity to take a statin (Table 4). In a secondary analysis, time-dependent exposure to statins was associated with an even lower adjusted risk of death compared with periods when not taking statins. The favorable association between time-varying statin use and death was observed among patients with known CHD (adjusted hazard ratio [HR] 0.66; 95% CI, 0.61-0.71) and in those without known CHD (HR, 0.60; 95% CI, 0.54-0.67) at study entry.
Similarly, using an intent-to-treat approach, incident statin use was associated with a 21% lower adjusted RR of hospitalization for heart failure compared with no statin use; the effect size was greater when examining statin use as a time-varying exposure (Table 4). Time-varying statin use was also associated with lower adjusted risks of hospitalization in patients with known CHD (HR, 0.73; 95% CI, 0.67-0.78) and in those without known CHD (HR, 0.74; 95% CI, 0.66-0.83).
Statins are effective for primary and secondary prevention of atherosclerotic events in the general population.1 While alterations of lipoprotein levels are considered a major reason for this benefit, other non–lipid-related effects such as reducing levels of inflammatory factors and detrimental cytokines, improving endothelial function, and stabilizing coronary plaque may be particularly beneficial in patients with heart failure.7 On the other hand, theoretical downsides to statin use in the setting of heart failure include reductions in coenzyme Q1014 and selenoprotein15 levels, which could have adverse effects on myocyte structure and function, as well as decreased ability of lipoproteins to bind endotoxins leading to excessive inflammation.13 In the present study, we found that within a large population of adults with heart failure who were eligible for lipid-lowering therapy, initiation of statin therapy was associated with lower risks for death and hospitalization, even after adjusting for expected differences in patients taking or not taking a statin with regard to cholesterol levels, other potential confounders, concurrent therapies, and the propensity to take a statin. The observed beneficial associations were prominent among patients with or without known CHD.
A recent systematic review of observational studies suggested that receipt of statins is associated with lower mortality and morbidity in patients with heart failure in most published studies with a wide range of potential effectiveness.34 For example, among 28 282 patients hospitalized for heart failure, receipt of any statin within 90 days after discharge was associated with a 28% lower adjusted RR of a composite cardiovascular outcome.35 In a post hoc analysis of 1153 patients with severe systolic heart failure in the Prospective Randomized Amlopidine Survival Evaluation (PRAISE) trial,36 the adjusted RR of death was 48% lower in the 134 patients receiving statins at baseline or during follow-up. In addition, among 551 patients with advanced systolic heart failure referred to a tertiary heart failure clinic, statin users had a 57% lower RR of death or urgent transplant.37 However, these and other published observational studies largely focused on selected subgroups including patients who were hospitalized for heart failure, had advanced heart failure, and/or impaired systolic function. Additional limitations include inclusion of prevalent statin users, only cross-sectional data on statin exposure, limited or no information on patients' eligibility for lipid-lowering therapy, minimal ethnic diversity, and concerns about residual confounding and incomplete adjustment for the effects of concurrent treatments that may be differentially used.34
Toward that end, the present study attempts to overcome many of these methodological challenges. Our heart failure population was large and sociodemographically diverse and included patients diagnosed with heart failure in both ambulatory and hospital settings. Reliance on an insured population with equal access to care also removed an important confounder. Our ascertainment of hospitalizations for heart failure was 100% because the health plan was financially responsible for admissions to network and out-of-network facilities. Identification of deaths included use of health plan databases that recorded inpatient and proxy-reported outpatient deaths, state death files, and Social Security Administration vital status. Our primary analysis used an intent-to-treat approach that focused on the association between initiation of statin therapy and risk of clinical outcomes in the relevant subgroup of patients with heart failure who were all recommended to receive lipid-lowering therapy based on national guidelines.21 Secondary analyses that more accurately characterized the timing and duration of exposure to statins suggested even greater benefit. We also statistically adjusted for cholesterol levels, a broad set of potential confounding variables including socioeconomic status, time-updated comorbidity and longitudinal use of other cardiovascular medications, as well as the propensity31 to initiate statin therapy in an attempt to further reduce the impact of treatment selection bias. Finally, we conducted analyses stratified by the presence or absence of CHD at study entry, given that statins and other lipid-lowering medications are indicated for secondary prevention in the general population. Overall, statin therapy remained a robust predictor of improved outcomes. Yet, because of the observational design of our study and the possibility of residual confounding, the magnitude of potential benefit from statin therapy may be overestimated given the effect sizes demonstrated in randomized controlled trials of other therapies for heart failure.38
Our study had several limitations. As an observational study of clinical practice, we cannot completely exclude residual confounding or selection bias as an alternative explanation of our findings, although we were able to adjust for a wide range of clinical and sociodemographic patient characteristics and other therapies received when evaluating incident statin use20 in comparable patients recommended to receive statin therapy.21 Despite use of previously validated approaches for detection of heart failure and relevant comorbidity from our databases,17,18,22- 24 as well as laboratory measures (ie, levels of lipoproteins, hemoglobin, serum creatinine), and filled medication prescriptions, we cannot rule out possible misclassification for all coexisting illnesses. We did not have data on left ventricular function for all patients, and information on functional status and quality of life was not available. Our study could not address specific mechanisms through which statins exert a beneficial effect on outcomes. Finally, as our study was conducted among insured adults in northern California, our results may not be completely generalizable to uninsured persons and other health care or geographic settings.
Limited experimental data on the efficacy of statins are available in patients with heart failure, and existing randomized comparisons in relatively small samples have yielded mixed results for intermediate outcomes related to inflammatory markers, left ventricular ejection fraction and other echocardiographic parameters, heart rate variability and QT duration and variability.39- 42 Therefore, given the clinically relevant effect size associated with receipt of statins in our study, results from ongoing (Controlled Rosuvastatin Multinational Study in Heart Failure [CORONA],43 Gruppo Italiano per lo Studio della Sopravvivenza nell’Insufficienza Cardiaca-Heart Failure [GISSI-HF]44) and future randomized controlled trials involving clinical outcomes—particularly among patients with nonischemic heart failure not otherwise recommended to receive lipid-lowering therapy—are needed to clarify the role of statins in the management of heart failure.
Corresponding Author: Alan S. Go, MD, Division of Research, Kaiser Permanente of Northern California, 2000 Broadway St, Third Floor, Oakland, CA 94612 (Alan.S.Go@kp.org).
Author Contributions: Dr Go had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Go, Lee, Yang, Gurwitz.
Acquisition of data: Go, Yang.
Analysis and interpretation of data: Go, Yang, Lo, Gurwitz.
Drafting of the manuscript: Go, Lee.
Critical revision of the manuscript for important intellectual content: Go, Lee, Yang, Lo, Gurwitz.
Statistical analysis: Go, Yang.
Obtained funding: Go.
Administrative, technical, or material support: Go.
Study supervision: Go.
Financial Disclosures: Dr Go and Ms Yang reported receiving research support for this study from Amgen Inc. Dr Go also reported receiving research support from Wyeth. Dr Lo reported receiving research support from Novartis. None of the other authors reported disclosures.
Funding/Support: The research for this article was funded by a grant from Amgen Inc.
Role of the Sponsor: Amgen Inc had no role in the design, collection, analysis, and interpretation of data; in the writing of the manuscript; or in the preparation or decision to submit the manuscript for publication.
Independent Statistical Analysis: All analyses were performed by Jingrong Yang, MA, a biostatistician and SAS analyst who works exclusively for the Kaiser Division of Research. She received statistical consultation from Lynn Ackerson, PhD, who also works exclusively for Kaiser's Division of Research.
Acknowledgment: We thank Lynn M. Ackerson, PhD, for her statistical input and Krista L. Lepper, BA, for her expert technical assistance. Both Dr Ackerson and Ms Lepper received research support from Amgen Inc for their work.