Temporal Patterns in the Medical Treatment of Congestive Heart Failure With Angiotensin-Converting Enzyme Inhibitors in Older Adults, 1989 Through 1995 FREE

THE EVIDENCE supporting the beneficial effect of angiotensin-converting enzyme (ACE) inhibitor treatment in patients with congestive heart failure (CHF) has been accumulating since the publication of the Cooperative North Scandinavian Enalapril Survival Study (CONSENSUS)¹ in 1987. The CONSENSUS demonstrated a significant 40% reduction in mortality for participants treated with enalapril maleate (mean final dose was 18.4 mg/d) compared with those treated with placebo. The largest study published since CONSENSUS, the Studies of Left Ventricular Dysfunction (SOLVD) Treatment Trial,² randomly assigned 2569 participants with CHF to receive either enalapril or placebo. A significant 26% reduction in mortality was found in patients who received enalapril (mean final dose of 11.2 mg/d) compared with those who received placebo, and subjects with the lowest ejection fractions (EFs) appeared to benefit most from treatment.

A recently published meta-analysis³ of 32 clinical trials of patients with CHF demonstrated a 35% overall reduction in mortality or hospitalizations for CHF (odds ratio [OR], 0.65; 95% confidence interval [CI], 0.57-0.74) among the ACE inhibitor group compared with the placebo group. The benefit was seen across all 8 ACE inhibitors. Treatment effects were even more pronounced among patients with EFs of 0.25 or less (OR, 0.53; 95% CI, 0.43-0.65) compared with those with EFs above 0.25 (OR, 0.85; 95% CI, 0.69-1.04).

The extent to which these findings have influenced clinical practice is unknown. In this study, we hypothesized that the use of ACE inhibitors for the treatment of CHF would increase over time, especially among participants with low EFs.

The Cardiovascular Health Study is a population-based, prospective cohort study of risk factors for cardiovascular and cerebrovascular disease.⁴ Participants were recruited from 4 US communities (Washington County, Maryland; Pittsburgh [Allegheny County, Pennsylvania]; Forsyth County, North Carolina; and Sacramento County, California) on the basis of a randomly generated sampling frame from Health Care Financing Administration files. Annual examinations began in June 1990 and are ongoing. This study included data on the original cohort from the baseline examination (study entry, June 10, 1989, through May 31, 1990) and the first 5 years of follow-up, through May 31, 1995 (study end, June 1, 1994, through May 31, 1995).

The cohort consisted of 5201 community-dwelling adults aged 65 years or older. All subjects gave informed consent for their participation according to guidelines created by the appropriate institutional review board.

Participants with CHF at entry into the study, in 1989-1990, were identified primarily by self-report at baseline examination. Eighty-nine percent of these cases were validated by medical record histories or physician interviews.⁵ Participants who had a clinical history of CHF at study entry and who did not report this during the baseline examination were subsequently identified when medical records were reviewed during hospitalizations that occurred during the 5 years of cohort follow-up.⁶

Incident cases of congestive heart failure were identified during semiannual contacts during which participants were asked about physician-diagnosed CHF in the previous 6 months. Participants who reported CHF, angina, myocardial infarction, claudication, stroke, or transient ischemic attacks had their medical records reviewed by the Cardiovascular Health Study Events Committee, which classified all events. The classification of CHF required (1) a CHF diagnosis and (2) confirming clinical evidence, such as radiographic findings or drug treatment with diuretics, digitalis glycosides, or vasodilators to support the diagnosis.

Left ventricular EFs were based on qualitative readings of echocardiograms that were conducted at study entry. The EFs were assessed as normal, borderline, or low. For this analysis, we grouped borderline and normal readings. Interreader and intrareader agreement on a 10% sample of echocardiogram readings was fair (κ=0.35) and very good (κ=0.92), respectively. Ejection fraction measures were updated for participants who had cardiac catheterizations subsequent to baseline readings. Angiogram readings with EF of 0.40 or less were considered low.

Medication use was ascertained by inventory at annual clinic visits where participants brought all prescription medications used within the last 2 weeks.⁷ Interviewers transcribed drug information from the medication containers, which included drug name, unit strength, and prescribed dosing instructions.

Additional medication data were collected for incident CHF cases either from (1) hospitalization records for those diagnosed in-hospital or (2) physician surveys for cases diagnosed in outpatient settings. These medications were prescribed at the time of the CHF diagnosis and therefore temporally preceded the inventory medication data that may have been collected up to 1 year after diagnosis. A participant was considered an ACE inhibitor user if (1) an ACE inhibitor was administered within the first 48 hours after diagnosis or admission for hospitalized cases or (2) an ACE inhibitor was prescribed as therapy to treat incident CHF for outpatient cases.

Trends in the use of ACE inhibitors were examined among prevalent and incident CHF cases. Prevalent cases were defined at each examination as those CHF cases that had been diagnosed at any point before the examination. At baseline, this included only those who entered the study with a previous CHF diagnosis. At each additional examination, cases of CHF that were diagnosed within the preceding year were added to the prevalent cases. Incident CHF cases were defined as those cases newly diagnosed only within the year before each of the 5 follow-up examinations. Participants with CHF who had died, who were too ill to participate further in the study, or who were otherwise not available for follow-up were excluded from analysis, since no medication data were available.

Change in the use of ACE inhibitors over time to treat prevalent CHF cases was analyzed by comparing the percentage who used ACE inhibitors at the baseline examination, 1989-1990, to the percentage who used ACE inhibitors at the last follow-up examination, 1994-1995. Comparisons were made among all prevalent cases and among those with normal and low EFs. Since a subject could appear in both the baseline and 5-year follow-up examination, percentages were compared by means of a bootstrap technique that accounts for the nonindependence of the study population.⁸ Odds ratios were calculated to compare the odds of ACE inhibitor use among cases with low EF with those of cases with normal EF at baseline and again at year 5 follow-up.

Change in the use of ACE inhibitors over time to treat incident CHF cases was assessed by means of a test for trend in proportions that used data from each of the 5 years of follow-up.⁹ Separate analyses for EF-stratified data were not conducted because of small numbers. Odds ratios were calculated to compare the odds of ACE inhibitor use among cases with low EF with those of cases with normal EF at year 1 and again at year 5 follow-up.

We analyzed the association between clinical and sociodemographic characteristics and the use of ACE inhibitors by means of multivariate logistic regression among incident cases.¹⁰ Candidate characteristics included age at diagnosis (65-74 years vs 75 years and older), sex, race (white vs nonwhite), education (less than high school degree vs high school degree and beyond), supplemental medical insurance coverage apart from Medicare, usual site for medical care (physician's office vs inpatient or outpatient hospital clinic, health department, emergency department, or home visit) county of residence (3 indicator variables for Washington, Allegheny, and Sacramento counties vs Forsyth County), and history of myocardial infarction (inclusive of co-occurring event), hypertension, diabetes, serum creatinine level of 132.6 mmol/L (1.5 mg/dL) or more, whether the CHF diagnosis was precipitated by a cardiovascular procedure (eg, bypass surgery) as opposed to a cardiovascular event (eg, myocardial infarction), and year of diagnosis. Medical insurance and care data were collected at the fourth follow-up examination, resulting in 20% of CHF cases having missing values. To maximize the number of observations in our logistic regression analysis, we imputed values for these missing medical insurance and site of medical care variables on the basis of their distribution among levels of ACE inhibitor use. We also analyzed the data by means of a complete case approach for comparison purposes. We collapsed the county of residence variable in the final model, since risks for the 3 counties were similar and differed from that of Forsyth County.

Of the 5210 participants enrolled at baseline, 250 (4.8%) were identified as having prevalent CHF (Table 1). With new CHF cases developing during the 5 years of follow-up and some CHF participants dying, the number of prevalent cases increased to 431 (9.5%) at the fifth follow-up examination. The number surviving to each follow-up examination is listed in Table 1. Among these survivors, medication data were available from 86%, 93%, 83%, 81%, and 84% of CHF cases at each of the 5 follow-up examinations, respectively.

Baseline estimation of EF by echocardiography was completed for 5152 (99%) of the participants (Table 2). Among the 250 with CHF at baseline, 57 (23%) were found to have low EFs compared with 133 (3%) of those without baseline CHF. Of the 133, 43 (32%) developed CHF during the 5-year follow-up.

The use of ACE inhibitors to treat CHF increased slightly over time in prevalence analyses: 26% were treated in 1989-1990 compared with 36% in 1994-1995 (Figure 1). This 10% increase in ACE inhibitor use was statistically significant (P<.01). The increases among participants with normal EF and low EF were comparable, yet only the increase among those with normal EF was statistically significant (P<.01). The relative odds of being treated with ACE inhibitors for participants with low EF compared with those with normal EF did not change from 1989-1990 to 1994-1995 (OR, 2.5; 95% CI, 1.3-4.6; vs OR, 2.1; 95% CI, 1.2-3.8, respectively).

The majority of prevalent CHF cases who used an ACE inhibitor were treated with captopril, and the mean dose was 54.2 mg/d (95% CI, 48.1-60.4 mg/d) (Table 3). Enalapril maleate and lisinopril had mean doses of 8.9 mg/d (95% CI, 7.1-10.6 mg/d) and 11.7 mg/d (95% CI, 9.5-14.1 mg/d), respectively.

Medication data from hospital records or physician surveys were available for 63%, 57%, 51%, 65%, and 66% of the incident cases in 1990-1991, 1991-1992, 1992-1993, 1993-1994, and 1994-1995, respectively. Eighty-seven percent of the data were collected from hospitalization records. Among those cases with data, ACE inhibitors were used in 26%, 53%, 50%, 46%, and 42% of participants immediately after a CHF diagnosis in 1990-1991, 1991-1992, 1992-1993, 1993-1994, and 1994-1995, respectively.

Forty-two percent, 37%, 38%, 27%, and 40% of those newly diagnosed with CHF in the preceding year were using ACE inhibitors in 1990-1991, 1991-1992, 1992-1993, 1993-1994, and 1994-1995, respectively (Figure 2). The 2% decrease in use was not statistically significant (P=.68). Although the relative odds of using ACE inhibitors among participants with low EF increased from 1.5 (95% CI, 0.32-7.0) in 1990-1991 to 4.5 (95% CI, 0.82-25.4) in 1994-1995, the CIs overlap and we cannot conclude that the odds differ.

For a subset of CHF participants, both medication data immediately after diagnosis and medication data at the next follow-up examination (average interval length, 173 days) were available. Among those using ACE inhibitors immediately after diagnosis (n=80), 47 (59%) continued to be treated with ACE inhibitors up to 1 year afterward. Among those not using ACE inhibitors immediately after diagnosis (n=97), 20 (21%) started treatment within the year.

Captopril was the more commonly prescribed ACE inhibitor for incident CHF cases, and the mean dose was 46.9 mg/d (95% CI, 34.8-59.0 mg/d) (Table 3). The mean doses for enalapril maleate and lisinopril were 7.7 mg/d (95% CI, 5.5-9.9 mg/d) and 11.5 mg/d (95% CI, 8.8-14.1 mg/d), respectively.

Predictors of ACE inhibitor use within a year of CHF diagnosis among participants in the Cardiovascular Health Study are listed in Table 4. Of the 288 participants with incident CHF and with medication data, 249 participants (86%) had complete covariate data for logistic regression analysis. Those with a low EF were 4.5 times more likely and hypertensive subjects were 2.5 times more likely to be using an ACE inhibitor within a year of their diagnosis after adjusting for age, sex, race, smoking, education, county of residence, history of hypertension, myocardial infarction, diabetes, serum creatinine level, procedure-related diagnosis, supplemental medical insurance, usual medical care site, and year of diagnosis. Younger age was marginally associated with an 80% increase in use. Other predictor variables could not be shown to be associated with ACE inhibitor use in the multivariate model. Complete-case analysis of these data produced similar findings.

Overall, the proportion of community-dwelling elderly persons with prevalent CHF who were treated with an ACE inhibitor never exceeded 40%. The use of ACE inhibitors increased modestly between 1989 and 1995 among all participants with a diagnosis of CHF but did not increase among incident cases. Among the subset of participants with immediate postdiagnosis medication data, there was also no evidence of a temporal trend. Interestingly, the near-doubling of ACE inhibitor use in this subset between the first and second follow-up examinations coincided with the publication of SOLVD. Participants with CHF who had a low EF were more likely to be treated with an ACE inhibitor than were participants with a normal EF; however, there was no evidence that those with a low EF were treated more often over time than were participants with a normal EF.

The modest increase in ACE inhibitor use among prevalent cases and absence of an increase among incident cases suggests that clinical trial data published in the early 1990s may have had at best a modest effect on prescribing patterns. No other longitudinal data have been published to support or refute our findings, and cross-sectional data are not widely available. Data collected from 1994 to 1996 suggest that as many as 75% of patients with CHF with moderately or severely decreased systolic function attending an urban academic medical center in the Midwest were prescribed ACE inhibitors during the outpatient visit.¹¹ Considering the exclusion of normal systolic function, this observation compares favorably with our data for 1994, which show that 65% of all incident cases with low EF were prescribed ACE inhibitors at the time of CHF diagnosis. (This percentage is based on only 6 observations, so data were not reported in the "Results" section of this article.) More generalizable data on office-based physician treatment of CHF come from the National Ambulatory Medical Care Survey, which estimated the 1991-1992 prevalence of ACE inhibitor use to be 30% among elderly Americans, similar to our findings at that same time.¹² Our reported prevalence of ACE inhibitor use is also similar to findings from data collected during 1992 from 2 community hospitals in upstate New York.¹³ Results showed that 39% of hospitalized patients with CHF who had a previous history of CHF reported ACE inhibitor use on hospital admission, and 51% of all patients with CHF were prescribed ACE inhibitors at discharge. These data suggest an increase in use since 1988, when data from the SOLVD international registry of patients with CHF or left ventricular dysfunction (EF<0.45) showed that just over 30% of patients used ACE inhibitors according to medical records, collected primarily from hospital discharges.¹⁴

A particularly alerting finding was that more than 40% of incident CHF cases prescribed ACE inhibitors after diagnosis discontinued therapy by the next clinical examination, an average of 5.7 months afterward. Discontinuation rates in clinical trials demonstrating ACE inhibitors' efficacy at reducing morbidity and mortality have been markedly lower: 17% at 6.2 months' follow-up¹ and 33% at 41.4 months' follow-up.² Similar low treatment discontinuation rates were seen in low-EF primary prevention¹⁵ and post–myocardial infarction secondary prevention¹⁶ trials with ACE inhibitors. Community-based data on discontinuation of pharmacological treatment for hyperlipidemia treatment show high rates comparable with our findings.¹⁷ Although our data indicate that participants who discontinued ACE inhibitors tended to be nonhypertensive and to have higher EFs, we do not have substantive information to adequately describe the clinical and nonclinical reasons why therapy was terminated.

Prescribed mean doses of enalapril for prevalent and incident cases were below the doses shown to be effective in both CONSENSUS and SOLVD. Captopril, the most commonly used ACE inhibitor to treat CHF in this population despite a lack of evidence demonstrating its effectiveness in long-term clinical trials, was also prescribed at doses below the 150-mg daily dose shown to be clinically effective at improving CHF symptoms.^18- 19 Our findings corroborate data from hypothetical clinical encounters that suggest that physicians tend to prescribe low doses of ACE inhibitors to patients with CHF.²⁰ These findings are similar to results published on the treatment of survivors of myocardial infarction with β-adrenergic blockers, where mean doses used in clinical practice were below those shown to be effective at preventing death in clinical trials.²¹ It is unclear why the doses used in clinical practice are lower than the doses shown to be effective in the clinical trials. Side effects may be one explanation, but ACE inhibitors are well tolerated, especially at lower doses.^22- 26

As expected, participants who had a history of hypertension or who had a low EF were more likely to be using ACE inhibitors within the first year after diagnosis. Although age failed to reach conventional levels of statistical significance in the multivariate model, the data suggest that those aged 75 years and older may be less likely to receive ACE inhibitor treatment for CHF. Similar findings were reported in a 2-site study¹³ in which age was inversely associated with both (1) preadmission use of ACE inhibitors among patients with a history of CHF and (2) at-discharge use among all patients with CHF in bivariate analyses; the associations were not significant in multivariate analyses. These findings are similar to data published on the treatment of patients after myocardial infarction, where clinically beneficial therapies are often underused, especially among the elderly and those with more severe disease.^21,27- 28 Whether a person had medical insurance supplemental to their Medicare policy was not related to use of ACE inhibitors. Our data suggest that receiving usual medical care at a physician's office is associated with increased use of ACE inhibitors for CHF treatment compared with other sites of usual medical care, such as hospital clinics, public health departments, and emergency departments.

Angiotensin-converting enzyme inhibitors are indicated for hypertension as well as CHF treatment. Increased use of ACE inhibitors among hypertensive participants has been demonstrated in these data in earlier research,²⁹ and the trend toward increased use has continued. In 1989-1990, 16% of hypertensive participants were using ACE inhibitors, and this climbed to 27% in 1994-1995. With 62% of prevalent CHF cases in 1993-1994 also reporting hypertension, the increased use among CHF cases may be, in part, the artifact of increased use for the hypertension indication. Since indication data are not available, reasons can only be speculated.

Several limitations of this study deserve attention. First, the study was designed to obtain a general sample of community-dwelling adults in the United States by recruiting participants by means of Health Care Financing Administration files.⁴ Inasmuch as the 4 counties faithfully reflect national cardiovascular morbidity and medical care patterns, our results can be generalized. The Cardiovascular Health Study placed a strong emphasis on accurately identifying prevalent and incident cardiovascular events among the sample.^4- 6 Although a small degree of misclassification is unavoidable, the method used to assess CHF would have underestimated the prevalence of less severe CHF cases, especially for subjects not currently pharmaceutically treated with vasodilators or diuretics. Our estimates of ACE inhibitor use would therefore tend to overestimate its prevalence among community-dwelling elderly with CHF.

Data on medication were missing for an average of 16% of those with prevalent CHF during the 5 years of follow-up. Thirty percent of these participants had a telephone interview so that a medication inventory was not completed, and 20% were unable to attend the clinical visit for health-related reasons. The remaining 50% had non–health-related reasons for not attending. The percentage of ACE inhibitor use among these participants can only be speculated.

At baseline, 11% of the CHF cases were determined exclusively by self-report, since confirmatory data were not available from medical records or physicians. These participants were not likely to be treated for CHF. When these participants were excluded from analyses, the percentage changes were trivial.

Estimates of EF were based primarily on baseline measurements. The prevalence of low EF compares favorably with 18% and 28% prevalence estimates derived from other nonhospitalized population samples of adults with CHF.^30- 31 Updated EF data were available for an average of 11% of the prevalent cases and 14% of the incident cases of CHF during the follow-up interval. It is possible that for an unknown proportion of the participants, updated EF measures were available to physicians and not available for these analyses. This misclassification of EF status may have underestimated ACE inhibitor use among true low-EF cases.

The observed increase in use among prevalent cases cannot be solely attributable to prevalent cases initiating ACE inhibitor therapy. The addition of incident cases who use more ACE inhibitors to a pool of prevalent cases who use less ACE inhibitors necessarily increases the overall percentage of those who use ACE inhibitors. We concluded that this effect was small, since ACE inhibitor use increased similarly during the 5 years of follow-up among the 250 cases that were prevalent at baseline (from 26% to 31%) compared with the mixed incident-prevalent cases (from 29% to 35%) shown in Figure 1. We also know that this effect was further minimized by the high rate of ACE inhibitor discontinuation among incident cases.

Medication data collected at the annual examinations would not have included medications prescribed but later withdrawn by the physician or not used by the participant in the previous 2 weeks. Angiotensin-converting enzyme inhibitors are generally tolerated in the elderly, but side effects may be one explanation for nonuse. Serum creatinine measurements were limited to data collected at baseline and 3-year follow-up. Transient increases in serum levels that may have led to ACE inhibitor withdrawal were not available for these analyses.

Last, CHF can be precipitated by other cardiovascular events and interventions. Although we adjusted for a history of myocardial infarction inclusive of events that may have precipitated heart failure and we adjusted for procedure-precipitated events in multivariate analyses, residual confounding may still be present.

In summary, these findings suggest that while the medical management of CHF with ACE inhibitors has increased modestly over time among community-dwelling elderly with prevalent CHF, (1) these drugs may still be underused, especially among incident cases, and (2) prescribed daily doses appear to be below those shown to be effective in clinical trials.

Accepted for publication November 6, 1997.

This study was supported by contracts N01-HC-85079, N01-HC-85080, N01-HC-85081, N01-HC-85082, N01-HC-85083, N01-HC-85084, N01-HC-85085, and N01-HC-85086 from the National Heart, Lung, and Blood Institute, Rockville, Md, and grant R01-AG-09556 from the National Institute on Aging, Bethesda, Md.

Presented in part at the American Heart Association Epidemiology and Prevention Council Meeting, San Francisco, Calif, March 15, 1996.

We thank Linda M. Levenson for her help in assembling cardiac catheterization and hospital medication data.

Reprints: Nicholas L. Smith, PhD, MPH, Cardiovascular Health Research Unit, Metropolitan Park, East Tower, Suite 1360, 1730 Minor Ave, Seattle, WA 98101 (e-mail: nlsmith@u.washington.edu).