Customize your JAMA Network experience by selecting one or more topics from the list below.
Krumholz HM, Chen Y, Radford MJ. Aspirin and the Treatment of Heart Failure in the Elderly. Arch Intern Med. 2001;161(4):577–582. doi:10.1001/archinte.161.4.577
Copyright 2001 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2001
We sought (1) to determine how often aspirin is prescribed as a discharge medication among patients 65 years or older and hospitalized with both heart failure and coronary artery disease; (2) to identify patient characteristics associated with the decision to prescribe aspirin; and (3) to evaluate the association between aspirin prescription at discharge and 1-year survival.
We performed a retrospective cohort study of consecutive Medicare beneficiary survivors of a hospitalization for heart failure at 18 Connecticut hospitals (up to 200 hospitalizations per hospital) from 1994 to 1995.
Among the 1110 patients in the study sample who did not have a contraindication to aspirin, aspirin therapy was prescribed for 456 (41%) at discharge. Patients who were prescribed aspirin at discharge had a lower 1-year mortality after discharge than patients who were not prescribed aspirin (odds ratio, 0.71; 95% confidence interval, 0.54-0.94), even after adjustment for baseline differences in demographic, clinical, and treatment characteristics between the 2 groups.
This study has identified a strong association between the use of aspirin and lower mortality in older patients with both heart failure and coronary artery disease. The benefit of aspirin is consistent with that expected from randomized trials of other groups of patients with vascular disease.
HEART FAILURE, the most common discharge diagnosis among Medicare beneficiaries, is associated with a high mortality rate. Recent clinical practice guidelines have defined the best strategies for the treatment of patients with heart failure.1-4 These guidelines, and the performance measures that have derived from them, have focused on the use of angiotensin-converting enzyme (ACE) inhibitors, digoxin, and, more recently, β-blockers.
Notably, aspirin has received relatively little attention for the treatment of patients with heart failure even though most of these patients have coronary artery disease.5 Aspirin has great value for patients with coronary artery disease because of its effectiveness, low cost, safety profile, and lack of strong contraindications. Pooled analyses of more than 18 000 patients with coronary artery disease show that aspirin significantly reduces long-term vascular mortality by about 20%.6
The reluctance to endorse aspirin for patients with heart failure derives from concerns about its effects as a prostaglandin inhibitor. Aspirin is reported to attenuate the effect of antihypertensive therapy,7-9 reduce vasodilator reserve,10 antagonize the action of ACE inhibitors,11 and decrease glomerular filtration pressure.12 The clinical trials have provided relatively little information about the effect of aspirin for patients with heart failure.13 An analysis of the Studies of Left Ventricular Dysfunction (SOLVD) trial14 suggested that the use of antiplatelet agents (more than 95% of which were aspirin) is associated with improved survival. The most recent heart failure guidelines warn about the combined use of antiplatelet agents and diuretics.3,4
To address this issue, we examined the use of aspirin as a discharge medication among Medicare beneficiaries who survived a hospitalization for both heart failure and coronary artery disease. We identified patient characteristics associated with the decision to prescribe aspirin and compared the 1-year mortality of patients who were prescribed aspirin at discharge with those who were not, adjusting for baseline differences between the groups. We also evaluated the effect of aspirin among subgroups of patients by their clinical characteristics (eg, baseline renal function) and treatments (eg, ACE inhibitors). While observational studies of this type will not supplant randomized trials for establishing efficacy, they allow for the rigorous study of patients in actual clinical practice, including many patient groups that are not commonly represented in the clinical trials.
Patients for the study were identified from Medicare's national claims history file (1994-1995) as part of a collaborative quality improvement project in Connecticut that was coordinated by Qualidigm (formerly the Connecticut Peer Review Organization), Middletown, Conn, and VHA Inc, Irving, Tex. Patients with a principal discharge diagnosis of heart failure (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 428-428.1, 402.01, 402.91, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93) at 18 Connecticut hospitals from 1994 to 1995 were selected for retrospective chart review. Initially, 200 hospitalizations from each hospital were reviewed. Institutions with fewer than 200 hospitalizations had all cases reviewed. We restricted the sample to patients who were aged 65 years or older with heart failure confirmed either by symptoms of heart failure or radiographic findings consistent with the diagnosis.
To decrease the heterogeneity of the sample, we excluded patients with severe aortic stenosis, severe mitral valve stenosis, or heart failure secondary to an acute medical illness (eg, sepsis). These exclusions resulted in an initial study sample of 2445 patients. For the current analysis we also excluded patients who died during hospitalization (n = 169), were transferred to another institution (n = 78), or had a documented terminal illness at discharge (n = 37). A terminal illness was considered present if (1) a patient was documented as "terminally ill" or as having a life expectancy of less than 6 months; (2) the admission orders indicated that the patient should be given palliative care only; or (3) a "do not resuscitate" order was written. We further excluded patients who did not have an indication for aspirin use, ie, patients without coronary artery disease (n = 920). Patients who had the following contraindications to long-term aspirin use were also excluded: documented allergy to aspirin (n = 103), platelet count lower than 100 × 109/L (n = 35), hematocrit level lower than 0.30 (n = 274), or bleeding (n = 19). The resulting sample consisted of 1110 patients.
Patient characteristics and clinical information (including the prescription of aspirin at discharge) were obtained from medical record review using a standardized data abstraction form. Trained nurses and medical record technicians abstracted the hospital records. Strategies to decrease abstraction errors and variability included training sessions and detailed data definitions for each field.
Coronary artery disease was considered present before discharge if myocardial infarction, angina, coronary artery disease, percutaneous transluminal coronary angioplasty (PTCA), and/or coronary artery bypass graft surgery (CABG) were documented in the medical record before or during the course of hospitalization. Patients were also considered to have coronary artery disease if they were hospitalized for ischemic heart disease (ICD-9-CM codes 410-414) in the year before the index hospitalization.
Left ventricular ejection fractions (LVEFs) were obtained from medical record review based on 1 of the following 3 methods: radionuclide ventriculography, cardiac catheterization (contrast ventriculography), or echocardiography (prioritized in that order for patients in whom LVEF was measured using more than 1 method). Qualitative measures of left ventricular function were translated into quantitative measures as follows: normal function was assigned a value of 55% or higher; mild or mild-moderate depression, 40% to 54%; moderate or moderate-severe depression, 20% to 39%; and severe depression, lower than 20%. Missing values were indicated with a dummy variable.
Other study variables included demographic characteristics (sex, age, and race); medical history (diabetes, hypertension, heart failure, myocardial infarction, angina, PTCA, CABG, atrial fibrillation, and preadmission aspirin and β-blocker use); admission characteristics (blood pressure, rales, pulmonary edema on chest radiograph, potassium level, creatinine level, and albumin level); major complications during hospitalization (a composite variable including cardiac arrest, myocardial infarction, and shock); major procedures during hospitalization (a composite variable including CABG, PTCA, and cardiac catheterization); and discharge characteristics (mobility, mental status, creatinine level, ratio of serum urea nitrogen [SUN] level to creatinine level, and the prescription of β-blockers, calcium channel blockers, ACE inhibitors, nitrates, nonsteroidal anti-inflammatory drugs [NSAIDs], and warfarin).
Outcome variables (including death, all-cause readmission, heart failure–related readmission, and death and readmission combined within 1 year after discharge) were obtained from administrative data. Information about readmission was derived from the Health Care Financing Administration's Medicare provider analysis and review file (1994-1996). This file contains discharge abstracts for all Medicare inpatients from Connecticut hospitals. Information about mortality was obtained from the Medicare Enrollment Database.
Initially, bivariate analyses were performed to examine the association between aspirin prescription and patient demographic and clinical characteristics. For presentation purposes, continuous variables were dichotomized or categorized as given in the tables. Missing data for categorical variables were considered as characteristics not present. χ2 Statistics or the Fisher exact test was used to compare proportions. Multivariate logistic regression models and stepwise selection procedures with an entry level P value less than .20 and an exit level P value greater than .10 were used to identify factors independently associated with the prescription of aspirin at discharge. For these models we did not include the prescription of aspirin at admission as a predictor variable.
Subsequently, associations between the prescription of aspirin and 1-year mortality postdischarge were examined. Multivariate analyses were performed using the Cox proportional hazards model, accounting for confounding effects. Assumptions of proportionality were examined and satisfied for the factors studied. Potential confounding factors were selected based on their significance in the bivariate associations with aspirin prescription and clinical judgment. The impact of aspirin use on 1-year mortality was also examined in the following subgroups: sex, age (65-74 and ≥75 years), presence of diabetes, creatinine level at admission (<176.80 µmol/L and ≥176.80 µmol/L), the prescription of ACE inhibitors at discharge, the prescription of β-blockers at discharge, and measures of LVEF (<40%, ≥40%, and unknown). All analyses were performed using PC-SAS version 6.12 (SAS Institute Inc, Cary, NC). All P values reported were 2-sided, and a P value less than .05 was considered to indicate statistical significance.
A total of 1110 patients who had coronary artery disease, survived a hospitalization for heart failure, and did not have a contraindication to aspirin use were included in the analysis. The sample included more women (55%) than men. The mean ±SD age was 78.3 ±7.3 years, and most patients were white (89%). Comorbidity was common in the study sample: 68% had prior heart failure; 60% had hypertension; 39% had diabetes; and 59% had a myocardial infarction before admission. Aspirin was used in 38% of the patients before the index hospitalization. The mortality at 1 year postdischarge was 24%. More than half (64%) of the patients were rehospitalized within 1 year after discharge (Table 1).
Aspirin was prescribed at discharge for 456 (41%) of the patients. The strongest bivariate associations (P<.01) were found between aspirin prescription and hypertension; prior myocardial infarction; prior atrial fibrillation; preadmission aspirin and β-blocker use; and the prescription of β-blockers, digoxin, nitrates, and warfarin at discharge (Table 1).
Patient and clinical characteristics independently associated with the decision to prescribe aspirin at discharge were identified using logistic regression models and stepwise selection procedures. Prior CABG; admission with chest pain; SUN level at admission; and the prescription of β-blockers, calcium channel blockers, nitrates, ACE inhibitors, and warfarin at discharge were significantly associated with aspirin prescription at discharge (Table 2). Patients with a SUN level greater than 14.28 mmol/L at admission and patients who were prescribed warfarin at discharge were less likely to be prescribed aspirin. Overall, the model showed a good discriminant ability (c-statistic, 0.75) in differentiating those who were and were not prescribed aspirin and did not indicate a lack of goodness of fit (P = .87). The model excluding treatments at discharge (with only demographic and clinical variables) had much less discriminant value (c-statistic, 0.61).
Patients who were prescribed aspirin had a significantly lower mortality at 1 year after discharge. The association between aspirin and long-term mortality remained significant after adjusting for confounding factors including sex; age; presence of diabetes, hypertension, or prior atrial fibrillation; albumin level, SUN level, and chest pain at admission; LVEF; creatinine level and SUN/creatinine ratio at discharge; and the prescription of β-blockers, calcium channel blockers, nitrates, warfarin, and ACE inhibitors at discharge (Table 3). Patients who were prescribed aspirin at discharge had a 29% lower mortality by 1 year after discharge compared with those who were not prescribed aspirin at discharge (relative risk [RR], 0.71; 95% confidence interval [CI], 0.54-0.94). Aspirin was not associated with rates of readmission or mortality and readmission combined at 1 year after discharge (Table 1).
To examine the strength of the association between aspirin and mortality, we performed subgroup analyses by sex, age, presence of diabetes, creatinine level at admission, the prescription of ACE inhibitors and β-blockers at discharge, and LVEF (Table 3). In general, the beneficial effect of aspirin on mortality persisted within each subgroup, except for patients who did not receive a prescription of ACE inhibitors at discharge and patients who had a normal LVEF (>40%). Not all of the comparisons were statistically significant, but the effect sizes were consistent. Prescriptions of other NSAIDs were found in 26 patients. An analysis combining aspirin and NSAID prescriptions did not substantially change the results.
The current American Heart Association guidelines for patients with established coronary artery disease recommend the use of aspirin.15 This recommendation is based on the results of numerous randomized controlled trials that have indicated the efficacy of aspirin and other antiplatelet agents.6 Despite this recommendation, the use of aspirin for patients with heart failure, many of whom have coronary artery disease, has been controversial.13 The randomized trials for the treatment effect of aspirin generally did not report a subgroup analysis of patients with heart failure. The trials that presented data for patients with heart failure did not provide definitive results.16,17 An analysis of a randomized trial population of patients with heart failure suggested that there is a benefit in using the medication.14 Physiology-based studies have provided a rationale for a potential harm of inhibiting prostaglandins in these patients. As a result, guidelines do not endorse the use of aspirin for patients with heart failure, even for those with coronary artery disease.
This uncertainty seems to have affected practice. In our study, among the patients who had coronary artery disease and did not have a contraindication to aspirin, only 41% were prescribed aspirin at discharge, which goes against the recommendations of the American Heart Association's secondary prevention guidelines.15 Patients who were less likely to be prescribed aspirin included those with a history of atrial fibrillation and a high SUN level at admission. However, demographic and clinical characteristics provided relatively little information about the decision to prescribe aspirin, suggesting that there is great variability in practice that is not dictated by the patient's condition.
The benefit associated with aspirin in this study was substantial. Overall, patients who were prescribed aspirin at discharge had a 29% lower risk of mortality at 1 year than the other patients. This reduction of risk among patients with heart failure is comparable to that reported in the randomized trials of patients with vascular disease.6 The benefit demonstrated in our results is also remarkably close to that estimated from the patients enrolled in the SOLVD trial.14 Given the high risk of mortality among patients with heart failure, this relative reduction in mortality translates into a benefit of 60 lives saved per 1000 patients treated. This estimate compares with the results of the trials that found antiplatelet therapy to be associated with about 40 vascular events avoided for every 1000 patients treated.
In exploratory analyses, we examined potential interactions of aspirin with other patient characteristics and treatments. In general, we found a consistent effect across subgroups. In some cases, the effect was not statistically significant, but the effect size was similar to other groups and the overall result.
Studies have suggested that aspirin may interfere with the beneficial effects of ACE inhibitors.18-21 In patients with reduced LVEF, the SOLVD trial found that antiplatelet agents reduced the benefit from enalapril in the treatment trial but not in the prevention trial. The investigators concluded that antiplatelet agents were associated with a persistent though reduced benefit from enalapril.14 We found that aspirin was associated with a greater survival advantage among patients who were prescribed ACE inhibitors, as well as those with a reduced LVEF. We did not specifically examine the influence of aspirin on the benefit of ACE inhibitors because that would require restricting the sample to ideal candidates for ACE inhibitors, which would markedly reduce the power of the study. Future studies clearly need to address directly whether there are specific groups that do not benefit from aspirin use or in which aspirin antagonizes ACE inhibitors. We also were not able to assess the combined effect of aspirin and warfarin because so few patients were treated with both medications.
The study has several limitations. First, treatment was not randomly allocated. However, our methods minimized the problems in drawing inferences from observational data.22 We defined a cohort of patients who had coronary artery disease and were free of contraindications to aspirin. We designated discharge as the reference time at which baseline clinical status was determined and from which follow-up was ascertained. We collected detailed information to adjust for differences between the treatment groups in their risk of mortality. While we cannot exclude the possibility of residual confounding factors, our results are very close to what was reported from randomized trials of patients with cardiovascular disease.
Second, we focused on patients who were aged 65 years or older in Connecticut, and the generalizability of our findings to younger patients is not known. Nevertheless, our study is relevant to more than 80% of the patients hospitalized in the United States with heart failure and who are Medicare beneficiaries. Moreover, our sample was not selected and represents the patients who are seen in practice. The study, however, was conducted solely in Connecticut, which has a high rate of aspirin use among patients discharged after an acute myocardial infarction compared with the rest of the country.23 Consequently, our estimates of the use of aspirin for patients with heart failure may be higher than what might be found in other regions, and the opportunity to improve care may be even greater elsewhere.
Third, we ascertained the use of aspirin at discharge through retrospective chart review. It is possible that aspirin use may not have been properly recorded or that treatment may not have continued after hospital discharge. These limitations, however, would have tended to diminish the association between aspirin use and improved survival.
Fourth, we determined the presence of coronary artery disease based on information from medical records. This method may have led to some misclassification of patients. The inclusion of patients in our cohort who did not have coronary artery disease would have tended to dilute the effect of aspirin unless it is effective for heart failure that is caused by nonischemic factors.
Finally, we are limited in our ability to suggest a mechanism of the benefit. We did not find a difference in the all-cause readmission rate or the heart failure readmission rate. The analysis from the SOLVD trial found that the reduction in mortality from antiplatelet agents was due to a reduction in sudden death and a reduction in fatal myocardial infarction. No association was found between antiplatelet agents and death associated with worsening heart failure.14
In conclusion, this study, undertaken as part of a local effort to improve care for Medicare beneficiaries with heart failure, identified a strong association between the use of aspirin and lower mortality among patients with both heart failure and coronary artery disease. The benefit of aspirin is consistent with what would be expected from randomized trials of other groups of patients with vascular disease. None of the analyses suggested that aspirin was associated with a significant harm. Thus, increasing the use of aspirin at discharge may be an excellent opportunity to improve the care of elderly patients with heart failure. Further studies need to determine the impact of aspirin on the benefit of ACE inhibitors.
Accepted for publication August 31, 2000.
Reprints: Harlan M. Krumholz, MD, Yale University School of Medicine, 333 Cedar St, PO Box 208025, New Haven, CT 06520-8025 (e-mail: firstname.lastname@example.org).