Diuretic resistance predicts mortality in patients with advanced heart failure

Diuretic resistance predicts mortality in patients with advanced heart failure Gerald W. Neuberg, MD,a Alan B. Miller, MD,b Chris M. O’Connor, MD,c Robert N. Belkin, MD,d Peter E. Carson, MD,e Anne B. Cropp, PharmD,f David J. Frid, MD,c Regina G. Nye, MPH,f Milton L. Pressler, MD,g John H. Wertheimer, MD,h and Milton Packer, MD,a for the PRAISE Investigators New York and Valhalla, NY, Jacksonville, Fla, Durham, NC, Washington, DC, Groton, Conn, Indianapolis, Ind, and Philadelphia, Pa

Background In patients with chronic heart failure (CHF), diuretic requirements increase as the disease progresses. Because diuretic resistance can be overcome with escalating doses, the evaluation of CHF severity and prognosis may be incomplete without considering the intensity of therapy.

Methods The prognostic importance of diuretic resistance (as evidenced by a high-dose requirement) was retrospectively evaluated in 1153 patients with advanced CHF who were enrolled in the Prospective Randomized Amlodipine Survival Evaluation (PRAISE). The relation of loop diuretic and angiotensin-converting enzyme inhibitor doses (defined by their median values) and other baseline characteristics to total and cause-specific mortality was determined by proportion hazards regression.

Results High diuretic doses were independently associated with mortality, sudden death, and pump failure death (adjusted hazard ratios [HRs] 1.37 [P ⫽ .004], 1.39 [P ⫽ .042], and 1.51 [P ⫽ .034], respectively). Use of metolazone was an independent predictor of total mortality (adjusted HR ⫽ 1.37, P ⫽ .016) but not of cause-specific mortality. Low angiotensin-converting enzyme inhibitor dose was an independent predictor of pump failure death (adjusted HR ⫽ 2.21, P ⫽ .0005). Unadjusted mortality risks of congestion and its treatment were additive and comparable to those of established risk factors.

Conclusions The independent association of high diuretic doses with mortality suggests that diuretic resistance should be considered an indicator of prognosis in patients with chronic CHF. These retrospective observations do not establish harm or rule out a long-term benefit of diuretics in CHF, because selection bias may entirely explain the relation of prescribed therapy to death. (Am Heart J 2002;144:31-8.) In patients with chronic congestive heart failure (CHF), the assessment of disease severity and estimation of mortality risk are important for clinical management and in triage for special interventions such as transplantation.1-5 The prognostic value of various clinical and laboratory parameters has been demonstrated in numerous CHF studies.6-8 In practice, congestive symptoms and signs,

From aColumbia University, College of Physicians and Surgeons, New York, NY, the b University of Florida Health Sciences Center, Jacksonville, Fla, cDuke University Clinical Research Institute, Durham, NC, dNew York Medical College, Valhalla, NY, eVA Medical Center, Washington, DC, fPfizer Central Research, Groton, Conn, gKrannert Institute of Cardiology, Indianapolis, Ind, and hAlbert Einstein Medical Center, Philadelphia, Pa. Supported by Pfizer Central Research, Groton, Conn. Guest Editor for this manuscript was James B. Young, MD, the Cleveland Clinic Foundation, Cleveland, Ohio. Submitted August 1, 2001; accepted January 14, 2002. Reprint requests: Gerald W. Neuberg, MD, Division of Circulatory Physiology, Columbia University, Milstein Pavilion, 5GN-435, 177 Ft Washington Avenue, New York, NY 10032. © 2002, Mosby, Inc. All rights reserved. 0002-8703/2002/$35.00 ⫹ 0 4/1/123144 doi:10.1067/mhj.2002.123144

left ventricular (LV) dysfunction, reduced exercise capacity, hypotension, azotemia, and arrhythmias are the indicators most commonly used to guide treatment.1-5 However, prognostic algorithms have not attempted to relate the severity of CHF to the intensity of therapy. Medical therapy for CHF must be intensified as the disease progresses, and this is particularly true of diuretics.1-3 Clinicians assume that patients who require higher doses of diuretics to prevent fluid retention and control symptoms are sicker than those who require less aggressive therapy, but the prognostic importance of such factors has not been examined or compared with that of established risk markers. Despite the subjective nature of clinical decision-making, we hypothesized that consideration of the degree of congestion together with its treatment would be prognostically relevant in patients with chronic CHF. Therefore, we retrospectively evaluated the relation of medication doses and other baseline variables to mortality and mode of death in patients with advanced CHF who were enrolled in the Prospective Randomized Amlodipine Survival Evaluation (PRAISE).

American Heart Journal July 2002

32 Neuberg et al

Table I. Baseline daily doses of diuretics in PRAISE

Furosemide Bumetanide Metolazone Thiazides K-sparing agents

N

Median

1038 86 184 51 50

80 mg 2 mg

Maximum 480 mg 12 mg 10 mg

Methods Patient population Between March 1992 and December 1994, PRAISE enrolled 1153 patients with LV ejection fraction (EF) of ⬍30% and New York Heart Association (NYHA) functional class IIIb/IV CHF despite mandatory background treatment with digoxin, diuretics, and angiotensin-converting enzyme (ACE) inhibitors. At study entry, patients were excluded if their serum potassium level was ⬍3.5 or ⬎5.5 mmol/L and if their serum creatinine level was ⬎3.0 mg/dL (270 ␮mol/L), and/or if they met other standard exclusion criteria. After baseline evaluation, patients were randomized to receive amlodipine 10 mg per day versus placebo, in addition to routine clinical management. The use of ␤-blockers was prohibited because none were approved for the treatment of CHF during this study.

Trial end points During a median follow-up of 14 months, 413 (35.8%) patients either died or underwent cardiac transplantation (193 on amlodipine, 220 on placebo). All deaths were reviewed by a blinded end points committee, which classified mode of death by majority vote as sudden death (n ⫽ 182), pump failure (n ⫽ 157), or other causes (n ⫽ 74) (including the 7 transplant recipients). In this trial, the sudden death category includes deaths that were preceded by minor or self-limited worsening of symptoms but still were considered unexpected. Further details of the study design and main results have been published.9,10

Group definitions We examined prognostically relevant patient characteristics available in the PRAISE database, including baseline demographics, physical findings, laboratory data, prescribed background treatment, and randomized treatment assignment. Exercise capacity, neurohormones, and ventricular arrhythmias were not measured in PRAISE. Continuous variables were converted to a categorical format by expressing them as above versus at or below the median (henceforth denoted as “high” or “low” groups). Daily doses of diuretics and ACE inhibitors were similarly classified. The high- and low-dose diuretic groups were defined according to the daily dose of loop diuretic (furosemide or bumetanide), whereas the few patients receiving other agents as their primary diuretic were arbitrarily assigned to the low-dose group, because of their generally lower potency.

Statistical analysis Baseline characteristics of patients in high- and low-dose diuretic and ACE inhibitor groups were compared with the

use of t tests. Relationships of medication doses and other baseline variables to total and cause-specific mortality were evaluated by ␹2 test (for unadjusted event rates), and multivariate analysis was performed with the use of proportional hazards regression (to determine adjusted hazard ratios [HRs]). Variables entered into the regression analysis were age, sex, race, CHF cause, NYHA class, LVEF, baseline doses of loop diuretics, ACE inhibitors, and digoxin, and the use of metolazone, thiazides, potassium-sparing agents, potassium chloride supplements, B vitamins, and amlodipine versus placebo. In addition, variables included systolic and diastolic blood pressure (BP) and heart rate, presence of jugular distention, rales, and edema, and cardiothoracic ratio (CTR), as well as levels of serum sodium, potassium, magnesium, creatinine, blood urea nitrogen (BUN), cholesterol, and bilirubin. P values of ⬍.05 were considered significant.

Results Baseline medication use At study baseline, the primary diuretic was furosemide (n ⫽ 1038) or bumetanide (n ⫽ 86) in 97% of PRAISE patients, with comparable median daily doses of 80 mg and 2 mg, respectively, and daily doses ranging up to 480 mg and 12 mg, respectively (Table I). The remaining 29 patients received a variety of agents alone or in combination, including amiloride, chlorothiazide, hydrochlorothiazide, indapamide, methazolamide, spironolactone, and triamterene. Among patients receiving furosemide or bumetanide, some were receiving combination therapy with additional diuretics, including metolazone, thiazides, or potassium-sparing diuretics. As their ACE inhibitor, almost all patients (96%) were receiving captopril, enalapril, or lisinopril, with comparable median daily doses of 75 mg, 10 mg, and 10 mg, respectively.

Group characteristics The characteristics of patients receiving high versus low doses of loop diuretics and ACE inhibitors at study entry are shown in Table II, which reveals significant differences among treatment groups. Mean daily doses of loop diuretics in the high- and low-dose groups were 175 ⫾ 3.3 mg versus 53 ⫾ 0.9 mg for furosemide (P ⬍ .001) and 4.7 ⫾ 0.3 mg versus 1.3 ⫾ 0.1 mg for bumetanide (P ⬍ .001). Patients receiving higher diuretic doses were more frequently male and in NYHA class IV and had rales and edema and received metolazone more often than low-dose patients. High-dose diuretic patients also had significantly lower mean values for age, LVEF, systolic BP, serum sodium, and potassium and total cholesterol and had a significantly higher CTR and BUN, serum creatinine, and total bilirubin levels than low-dose patients. In the high-dose ACE inhibitor group, mean daily doses were similar to those used in placebo-controlled trials11-13 and in the high-dose group of the ATLAS (Assessment of Treatment with Lisinopril And Survival)

American Heart Journal Volume 144, Number 1

Neuberg et al 33

Table II. Baseline characteristics by treatment group Diuretic dose

N Demographics Age (y) LVEF (%) Sex: male (%) Etiology ischemic (%) NYHA Class IV (%) Race: white (%) Diabetes (%) Treatment Furosemide dose (mg) Bumetanide dose (mg) Captopril dose (mg) Enalapril dose (mg) Lisinopril dose (mg) Metolazone use (%) Thiazide use (%) K-sparing agents (%) KCl supplementation (%) Amlodipine (%) Physical and laboratory findings Systolic BP (mm Hg) Diastolic BP (mm Hg) Heart rate (beats/min) JVD (%) Rales (%) Edema (%) Cardiothoracic ratio (%) Serum sodium (mmol/L) Serum K (mmol/L) Serum Mg (mg/dL) BUN (mg/dL) Creatinine (mg/dL) Bilirubin (mg/dL) Cholesterol (mg/dL)

ACE inhibitor dose

High

Low

High

Low

402

751

384

766

63.6 ⫾ 0.6 20.2 ⫾ 0.3 79.3 63.9 23.6 74.4 31.8

65.5 ⫾ 0.4* 21.0 ⫾ 0.2* 74.0* 63.5 16.9* 77.8 46.8*

63.4 ⫾ 0.6 20.8 ⫾ 0.3 73.7 53.4 20.1 67.5 43.5

65.6 ⫾ 0.4* 20.5 ⫾ 0.2 76.9 68.6* 18.9 81.2* 33.7*

175 ⫾ 3.3 4.7 ⫾ 0.3 87.6 ⫾ 4.2 16.5 ⫾ 1.1 17.5 ⫾ 2.0 24.9 3.5 5.5 78.6 47.5

53 ⫾ 0.9* 1.3 ⫾ 0.1* 71.1 ⫾ 3.1* 12.8 ⫾ 0.7* 15.3 ⫾ 1.1 11.2* 4.9 3.7 56.7 50.6

109 ⫾ 4.2 3.0 ⫾ 0.5 153 ⫾ 3.6 26 ⫾ 1.0 26 ⫾ 1.3 14.6 3.1 3.1 65.1 47.1

90 ⫾ 2.6* 2.7 ⫾ 0.3 46 ⫾ 1.2* 6.9 ⫾ 0.2* 7.8 ⫾ 0.3* 16.6 5.0 4.8 64.0 50.7

114 ⫾ 0.8 71.6 ⫾ 0.5 80.6 ⫾ 0.7 5.5 30.4 50.5 57.9 ⫾ 0.41 138 ⫾ 0.20 4.31 ⫾ 0.03 2.13 ⫾ 0.02 30.2 ⫾ 0.89 1.44 ⫾ 0.23 0.89 ⫾ 0.03 195 ⫾ 2.6

119 ⫾ 0.7* 72.7 ⫾ 0.4 79.3 ⫾ 0.5 3.7 20.3* 33.3* 56.3 ⫾ 0.33* 139 ⫾ 0.14* 4.43 ⫾ 0.02* 2.10 ⫾ 0.01 23.9 ⫾ 0.46* 1.34 ⫾ 0.02* 0.76 ⫾ 0.02 206 ⫾ 1.8*

121 ⫾ 0.9 74.6 ⫾ 0.5 80.0 ⫾ 0.6 3.7 22.9 40.9 57.6 ⫾ 0.5 140 ⫾ 0.2 4.38 ⫾ 0.02 2.08 ⫾ 0.02 26.3 ⫾ 0.8 1.38 ⫾ 0.02 0.81 ⫾ 0.03 203 ⫾ 2.5

116 ⫾ 0.6* 71.1 ⫾ 0.4* 79.6 ⫾ 0.5 4.7 24.2 38.4 56.6 ⫾ 0.3 139 ⫾ 0.1* 4.40 ⫾ 0.02 2.12 ⫾ 0.01 25.9 ⫾ 0.5 1.38 ⫾ 0.02 0.80 ⫾ 0.02 202 ⫾ 1.8

Data for continuous variables are means ⫾ SE. *P⬍.05 between groups.

trial14 and to those recommended in American Heart Association/American College of Cardiology guidelines.2 In the low-dose ACE inhibitor group, mean doses were similar to those in the low-dose group of ATLAS. Patients receiving low-dose ACE inhibitors were more likely to be white and to have ischemic heart disease than were patients receiving high-dose ACE inhibitors. Patients receiving low-dose ACE inhibitors also were significantly older and had significantly lower BP and serum sodium level, but there were no significant differences in on-treatment levels of serum potassium or renal function.

Unadjusted mortality rates The univariate relation of baseline treatment to mortality is shown in Table III. High diuretic dose, use of

metolazone, and low ACE inhibitor dose were associated with a significant excess of total and cause-specific mortality in PRAISE. Unadjusted HRs for total mortality by treatment group were, in descending order, 1.62 for metolazone use (P ⬍ .01), 1.45 for high diuretic dose (P ⬍ .01), and 1.27 for low ACE inhibitor dose (P ⫽ .073). These ratios are modest, but few variables had higher univariate HRs, including 1.88 for high BUN (⬎22 mg/dL, P ⬍ .001), 1.61 for white race (P ⬍ .001), 1.53 for NYHA class IV (P ⬍ .001), and 1.52 for low systolic BP (⬍118 mm Hg, P ⬍ .001). Other variables were significant univariate predictors of total mortality in PRAISE, but their unadjusted HRs did not exceed those for aggressive diuretic therapy, including 1.41 for both ischemic etiology (P ⬍ .001) and high CTR (⬎57%, P ⬍ .001), 1.39 for low serum

American Heart Journal July 2002

34 Neuberg et al

Table III. Unadjusted mortality and mode of death in treatment groups Total mortality

Diuretic High Low ACE inhibitor High Low Metolazone Yes No

Sudden death

Pump failure death

%

HR

%

HR

%

HR

44.8 31.0

1.45*

20.4 13.3

1.53*

17.9 11.3

1.58

30.2 38.5

0.78*

14.8 16.3

0.91

7.0 16.7

0.42*

52.7 32.6

1.62*

21.7 14.7

1.48†

23.4 11.8

1.99†

HR, Hazard ratio (unadjusted). Statistics by ␹2. *P ⬍ .01. †P ⬍ .05.

Table IV. Unadjusted mortality (%) in subgroups Dose of diuretic

NYHA class 4 (n ⫽ 222) 3 (n ⫽ 931) Rales Yes (n ⫽ 274) No (n ⫽ 879) Edema Yes (n ⫽ 453) No (n ⫽ 700) Metolazone Yes (n ⫽ 184) No (n ⫽ 969) ACE inhibitor dose Low (n ⫽ 766) High (n ⫽ 384) LVEF ⱕ21% ⬎21% Systolic BP ⱕ118 ⬎118

NYHA class IV

III

Max HR

High

Low

61.0 39.7

40.9 30.0

2.03*

56.6 39.6

34.9 30.1

1.88*

48.3 41.2

35.3 28.9

1.67*

52.0 39.1

41.7 29.7

1.75*

50.8 35.6

32.9 26.3

1.93*

57.0 36.3

35.5 29.4

1.94*

58.3 38.3

38.4 25.8

2.26*

Max HR, Maximum hazard ratio of highest versus lowest risk subgroup. Statistics by ␹2. *P ⬍ .001.

cholesterol (⬍200 mg/dL, P ⬍ .001), 1.36 for older age (⬎65 years, P ⬍ .001), 1.35 for high serum creatinine (⬎1.3 mg/dL, P ⫽ .001), 1.34 for pulmonary rales (P ⫽ .001), 1.32 for low LVEF (⬍21%, P ⬍ .001), 1.27 for high heart rate (⬎80 beats/min, P ⫽ .003) and peripheral edema (P ⫽ .003), 1.23 for male sex (P ⫽ .036), and 1.21 for high bilirubin (⬎0.6 mg/dL, P ⫽ .017). The mortality risk associated with high diuretic doses reflected an excess of both of the common

modes of death, with similar unadjusted HRs of 1.53 (P ⬍ .01) for sudden death and 1.58 (P ⬍ .01) for pump failure death. The risk associated with metolazone use also reflected an excess of both sudden death (HR ⫽ 1.48, P ⬍ .05) and pump failure death (HR ⫽ 1.99, P ⬍ .05). In contrast, the risk associated with low ACE inhibitor doses was limited to an excess of pump failure death with an HR of 1.84 (P ⬍ .01). Other causes of death were relatively infrequent and were not clearly related to any treatment group.

American Heart Journal Volume 144, Number 1

Neuberg et al 35

Table V. Independent predictors of mortality and mode of death Adjusted hazard ratios (P) Cardiothoracic ratio ⬎57% BUN ⬎22 mg/dL Systolic BP ⱕ118 mm Hg Heart rate ⬎80 beats/min Ischemic etiology High diuretic dose Metolazone use Age ⬎65 years Low ACE inhibitor dose Cholesterol ⱕ200 mg/dL Serum sodium ⱕ139 mmol/L Female sex Peripheral edema

Total mortality

Sudden death

Pump failure death

1.75 (.0001) 1.65 (.0001) 1.64 (.0001) 1.51 (.0009) 1.49 (.0002) 1.37 (.004) 1.37 (.018) 1.33 (.025) – – – – –

1.75 (.002) 1.70 (.0002) – 1.45 (.017) – 1.39 (.042) – – – – – 1.70 (.012) 1.44 (.005)

2.16 (.0001) 2.06 (.0001) 1.97 (.0001) 1.97 (.0001) 2.09 (.0001) 1.51 (.034) – 1.67 (.019) 2.21 (.0005) 1.82 (.0005) 1.48 (.028) – –

Multivariate analysis by proportional hazards regression, using all variables in Table II (except diabetes). Continuous variables grouped by median. HRs of P ⱖ .05 not shown.

Figure 1

In quartiles of loop diuretic dose, total mortality increased progressively without a clear risk threshold, more than doubling from the lowest-dose group to the highest-dose group (P ⫽ .001). Unadjusted mortality rates were 20.7% (n ⫽ 152), 30.7% (n ⫽ 313), 36.8% (n ⫽ 304), and 44.8% (n ⫽ 384) for increasing dose of furosemide (⬍40 mg, 40 to ⬍80 mg, 80 to ⬍120 mg, and ⱖ120 mg daily) or bumetanide (⬍1 mg, 1 to ⬍2 mg, 2 to ⬍3 mg, and ⱖ3 mg daily), respectively.

Unadjusted mortality in subgroups Mortality rates in subgroups relating congestion to its treatment are shown in Table IV. Patients with class IV symptoms, rales, or edema despite high diuretic doses had a significantly worse outcome than did those with lesser symptoms or signs on lower doses (P ⫽ .001). Between the highest- and lowest-risk subgroups of diuretic dose and NYHA class, differences (“maximum hazard ratios”) were 2.03 for mortality, 2.19 for sudden death, and 2.57 for pump failure death. For diuretic dose and rales, maximum HRs were 1.88 for mortality, 1.80 for sudden death, and 2.50 for pump failure death. For diuretic dose and edema, maximum HRs were 1.67 for mortality, 2.04 for sudden death, and 1.85 for pump failure death. Also, patients for whom metolazone was added to high-dose diuretics had 1.75-fold excess mortality over those receiving low-dose loop agents alone, reflecting HRs of 1.85 for sudden death and 2.84 for pump failure death. Patients receiving low diuretic doses and high ACE inhibitor doses had the best survival of any treatment group. HRs of high diuretic/low ACE patients versus low diuretic/high ACE patients were 1.93 for total mortality, 1.86 for sudden death, and 3.52 for pump failure death. Mortality differences in these subgroups were comparable to those for more conventional risk fac-

Cumulative mortality rates in diuretic/angiotensin-converting enzyme inhibitor subgroups, unadjusted for baseline characteristics. Kaplan-Meier curves are compared with ␹2 test.

tors. For example, in subgroups of NYHA class and LVEF, the maximum mortality difference was 1.94, whereas in subgroups of NYHA class and systolic BP, the maximum HR was 2.26. Cumulative mortality in diuretic/ACE inhibitor subgroups is shown in Figure 1.

American Heart Journal July 2002

36 Neuberg et al

Independent predictors of mortality To correct for baseline differences between treatment groups, multivariate analysis was performed (Table V). By proportional hazard regression, high diuretic dose was an independent predictor of total mortality (adjusted HR ⫽ 1.37, P ⫽ .004), sudden death (adjusted HR ⫽ 1.39, P ⫽ .042), and pump failure death (adjusted HR ⫽ 1.51, P ⫽ .034). Metolazone use also was an independent predictor of total mortality (adjusted HR ⫽ 1.37, P ⫽ .018) but not of cause-specific mortality. Low ACE inhibitor dose was an independent predictor of pump failure death (adjusted HR ⫽ 2.21, P ⫽ .0005) but not of sudden death or total mortality. Other independent predictors of total mortality in PRAISE were high CTR (adjusted HR ⫽ 1.75), high BUN (adjusted HR ⫽ 1.65), low systolic BP (adjusted HR ⫽ 1.64), high heart rate (adjusted HR ⫽ 1.51), ischemic etiology (adjusted HR ⫽ 1.49), and older age (adjusted HR ⫽ 1.33). These variables also were independent predictors of cause-specific mortality. In addition, low cholesterol and low serum sodium levels were independently associated with pump failure death (but not total mortality or sudden death), whereas female sex and peripheral edema appeared as independent predictors of sudden death (but not total mortality or pump failure death). Use of potassiumsparing diuretics was not independently predictive of outcome in PRAISE, but the numbers were very small (n ⫽ 50).

Discussion In patients with chronic CHF, diuretic dose requirements tend to increase as the disease progresses, because of worsening fluid retention and/or diuretic resistance.1-3 These phenomena are related by overlapping mechanisms, including decreased renal blood flow resulting from low cardiac output, reninangiotensin system activation and vasoconstriction, excess sodium and fluid intake, and use of nonsteroidal anti-inflammatory drugs.15,16 Other mechanisms of diuretic resistance described in patients with CHF include decreased drug absorption and renal adaptation to chronic therapy.16 Because diuretic resistance can be overcome by escalating doses and drug combinations,1-3,16 it follows that assessment of CHF severity may be inadequate without consideration of the intensity of therapy. Although it appears obvious that patients requiring higher diuretic doses to prevent fluid retention and control symptoms are sicker than are others, the prognostic importance of such factors has not been evaluated. In this study, we retrospectively examined the relation of CHF treatment and other baseline variables to mortality and mode of death in 1153 patients followed for a median of 14 months in the PRAISE trial.

We found that high doses of loop diuretics, use of metolazone, and low ACE inhibitor doses were independent predictors of total mortality and/or cause-specific mortality in patients with advanced CHF.

Diuretic dose as a risk marker These results suggest that diuretic resistance (defined clinically as a high-dose requirement) should be considered a determinant of prognosis in patients with chronic CHF. The prescribed dose of diuretics is a somewhat subjective measure of disease severity but can be partially objectified by considering the degree of congestion together with its treatment. The mortality risk of these factors was additive in PRAISE, and perhaps the assessment could be further refined by using a more precise fluid retention score as suggested by Cody,17 by documenting dietary sodium intake, and by examining changes in therapy over time. The association of high diuretic doses and metolazone use with mortality was modest in this study, with adjusted HRs of only 1.37. However, the risk of other CHF markers also was small, with adjusted HRs never exceeding 1.75 for any single variable. This probably relates to the homogenizing effect of a median analysis, as well as the severity of illness in the PRAISE population, in which no subgroup had a good outcome. Previous studies demonstrating a larger influence of risk factors on CHF mortality evaluated a broader spectrum of patients. For example, the Vasodilator Heart Failure Trial (V-HeFT) reported an 8-fold difference in mortality in subgroups defined by LVEF and exercise capacity.7 However, their analysis used multiple data cutpoints and included patients with LVEF up to 45% and all grades of CHF, whereas PRAISE was restricted to patients with LVEF of ⬍30% and advanced symptoms.

Diuretic treatment effects It may be tempting to view the association of high diuretic doses with mortality as possible evidence of an adverse treatment effect. Despite short-term clinical and hemodynamic benefits seen in uncontrolled studies of loop diuretics in CHF,18-21 outcomes theoretically could be worsened by known diuretic side effects, including neurohormonal activation, electrolyte depletion, and renal insufficiency.22-26 Unfortunately, the long-term effects of diuretics in CHF have not been evaluated in large randomized trials.27 In hypertension trials, low doses of thiazides improved outcomes,28,29 but this does not ensure the safety of more aggressive diuretic regimens used in patients with CHF. Furthermore, excess sudden death was found in an overview of thiazide trials,30 a risk that was absent in studies combining thiazides with a potassium-sparing diuretic.31 Such a hazard would be more serious in CHF,

American Heart Journal Volume 144, Number 1

because of its higher mortality rate and 30% to 50% prevalence of sudden death.32 Similar questions are raised by the benefit of low-dose spironolactone, which prevented hypokalemia and improved survival in patients with advanced CHF who were on loop diuretics and ACE inhibitors.33,34 Despite these concerns, the study data do not prove that diuretics caused harm in PRAISE. When interpreting the outcome of prescribed treatment retrospectively, patient selection must be viewed as an inherent source of bias, tending to exaggerate hazards and to obscure benefits.35-37 Predictably, patients who received higher diuretic doses were sicker than others, as evidenced by numerous prognostic characteristics. Multivariate analysis adjusts for such differences, yet residual confounding can persist, because clinical databases are incomplete and because a patient’s condition on treatment tends to underestimate previous disease severity.37 Furthermore, without an untreated control group, there is no way to address the possibility that both high- and low-dose groups might have fared worse with less treatment.38 Indeed, many patients had edema at baseline, suggesting that perhaps they could have benefited from even higher diuretic doses. The classification of cause-specific mortality remains subjective and imprecise in patients with CHF,39-41 but the nonspecific excess of sudden and nonsudden death in the high-dose group also suggests an indirect prognostic influence, rather than a direct hazard.42 The retrospective report by Cooper et al43 of an association between diuretic use and arrhythmic death in the Studies Of Left Ventricular Dysfunction (SOLVD) is misleading because other modes of death were ignored. Unadjusted event rates in their Table III reveal a nonspecific association because, in addition to an 80% excess of arrhythmic death, diuretic use in SOLVD was associated with a 132% excess of nonarrhythmic death and a 137% excess of nonarrhythmic cardiovascular death (usually implying pump failure). Their retrospective analysis is further confounded by the pooling of 2 distinct populations (SOLVD Treatment and Prevention patients) that were specifically defined by the presence or absence of drug treatment for CHF and prestratified with a clear expectation of differential outcome.

ACE inhibitors and mortality The relationship of ACE inhibitor dosing to mortality in PRAISE also is subject to confounding. These agents are more difficult to uptitrate in sicker patients, producing a selection bias opposite from that affecting diuretics.44 Such confounding may explain why this study suggested an advantage of high-dose ACE inhibition that was not seen in the ATLAS trial,14 in which high-dose versus low-dose lisinopril reduced hospitalization by 24% (P ⫽ .002) but reduced mortality by

Neuberg et al 37

only 8% (P ⫽ NS). On the other hand, the benefit of high ACE inhibitor doses in a sicker population, like that of PRAISE, could have been underestimated in ATLAS because of the inclusion of patients with relatively mild and stable CHF, as well as a 20% crossover rate. In placebo-controlled trials, the mortality benefit of high-dose ACE inhibition also was greater in advanced CHF11 than in mild-moderate CHF.12,13 Interestingly, the mortality risk associated with high diuretic and low ACE inhibitor doses was additive in PRAISE, and their combined risk exceeded that of any single factor examined. This management pattern thus may be a useful marker of poor prognosis related to a combination of avid fluid retention and relative hypotension.

Conclusions In summary, we found that high doses of loop diuretics (furosemide ⬎80 mg or bumetanide ⬎2 mg daily), use of metolazone, and low ACE inhibitor doses were independently associated with mortality in patients with advanced CHF. When the degree of congestion was considered together with its treatment, the associated risks were additive, suggesting that diuretic resistance should be considered an indicator of prognosis in chronic CHF. In this study, the prognostic influence of treatment intensity was modest but comparable to that of more conventional risk factors and should be further explored in a broader spectrum of patients with CHF. Finally, these retrospective findings do not establish harm, nor do they rule out a longterm benefit of diuretic therapy in advanced CHF because, despite multivariate analysis, residual and uncorrectable selection bias may explain the relation of prescribed therapy with mortality. Direct effects of diuretics on long-term CHF outcome must be determined prospectively.

References 1. Packer M, Cohn JN, et al. Consensus recommendations for the management of congestive heart failure. Am J Cardiol 1999;83:138A. 2. Williams JF Jr, Bristow MR, Fowler MB, et al. Guidelines for the evaluation and management of heart failure: report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee on Evaluation and Management of Heart Failure). J Am Coll Cardiol 1995;26:1376-98. 3. Stevenson LW, Massie BM, Francis GS, et al. Optimizing therapy for complex or refractory heart failure: a management algorithm. Am Heart J 1998;135(2 Suppl):S293-309. 4. Mudge GH, Goldstein S, Addonizio LJ, et al. 24th Bethesda Conference: cardiac transplantation. Task Force 3: recipient guidelines/prioritization. J Am Coll Cardiol 1993;22:21-31. 5. Kao W, Costanza MR. Prognosis determination in patients with advanced heart failure. J Heart Lung Transplant 1997;16:52-6. 6. Myers J, Gullestad L, Vagelos R, et al. Clinical, hemodynamic and

38 Neuberg et al

7.

8.

9.

10.

11.

12.

13.

14.

15. 16. 17.

18.

19.

20.

21.

22.

23.

24.

cardiopulmonary exercise determinants of survival in patients referred for evaluation of heart failure. Ann Intern Med 1998;129: 286-93. Cohn JN, Johnson GR, Shabetai R, et al. Ejection fraction, peak exercise oxygen consumption, cardiothoracic ratio, ventricular arrhythmias, and plasma norepinephrine as determinants of prognosis in heart failure. Circulation 1993;87(6 Suppl):5-16. Gottlieb SS, Kukin ML, Bernstein JL, et al. Predictors of mortality change as the disease progresses in patients with chronic heart failure [abstract]. J Am Coll Cardiol 1989;13:39A. Packer M, O’Connor CM, Ghali JK, et al. Effect of amlodipine on morbidity and mortality in severe chronic heart failure. N Engl J Med 1996;335:1107-14. O’Connor CM, Carson PE, Miller AB, et al. Calcium channel blockade with amlodipine in advanced heart failure: effects on mode of death in the PRAISE trial. Am J Cardiol 1998;82:881-7. The CONSENSUS Trial Study Group. Effects of enalapril on mortality in severe congestive heart failure: results of the Cooperative North Scandinavian Enalapril Survival Study (CONSENSUS). N Engl J Med 1987;316:1429-35. The SOLVD Investigators. Effect of enalapril on survival in patients with reduced left ventricular ejection fractions and congestive heart failure. N Engl J Med 1991;325:293-302. Cohn JN, Johnson G, Ziesche S, et al. A comparison of enalapril with hydralazine-isosorbide dinitrate in the treatment of chronic congestive heart failure. N Engl J Med 1991;325:303-10. Packer M, Poole-Wilson PA, Armstrong PW, et al. Comparative effects of low and high doses of the angiotensin-converting enzyme inhibitor, lisinopril, on morbidity and mortality in chronic heart failure. Circulation 1999;100:2312-8. Kramer BK, Schweda F, Reigger GAJ. Diuretic treatment and diuretic resistance in heart failure. Am J Med 1999;106:90-6. Ellison DH. The physiologic basis of diuretic synergism: its role in treating diuretic resistance. Ann Intern Med 1991;114:886-94. Cody RJ. Clinical trials of diuretic therapy in heart failure: research directions and clinical considerations. J Am Coll Cardiol 1993;22(A Suppl):165-71. Gottlieb SS, Khatta M, Wentworth D, et al. The effects of diuresis on the pharmacokinetics of the loop diuretics furosemide and torsemide in patients with heart failure. Am J Med 1998;104:533-8. Chomsky DB, Lang CC, Rayos G, et al. Treatment of subclinical fluid retention in patients with symptomatic heart failure: effect on exercise performance. J Heart Lung Transplant 1997;16:846-53. Fonarow GC, Stevenson LW, Walden NA, et al. Impact of a comprehensive heart failure management program on hospital readmission and functional status of patients with advanced heart failure. J Am Coll Cardiol 1997;30:725-32. Grinstead WC, Francis MJ, Marks GF, et al. Discontinuation of chronic diuretic therapy in stable congestive heart failure secondary to coronary artery disease or to idiopathic dilated cardiomyopathy. Am J Cardiol 1994;73:881-6. Bayliss J, Norell M, Canepa-Anson R, et al. Untreated heart failure: clinical and neuroendocrine effects of introducing diuretics. Br Heart J 1987;57:17-22. Francis GS, Benedict C, Johnstone DE, et al. Comparison of neuroendocrine activation in patients with left ventricular dysfunction with and without congestive heart failure. Circulation 1990;82:1724-9. Packer M. The neurohormonal hypothesis: a theory to explain the mechanisms of disease progression in heart failure. J Am Coll Cardiol 1992;20:248-54.

American Heart Journal July 2002

25. Dei Cas L, Metra M, Leier CV. Electrolyte disturbances in chronic heart failure: metabolic and clinical aspects. Clin Cardiol 1995; 18:370-6. 26. Seligmann H, Halkin H, Rauchfleisch S, et al. Thiamine deficiency in patients with congestive heart failure receiving long-term furosemide therapy. Am J Med 1991;91:151-5. 27. Gheorghiade M, Benatar D, Konstam MA, et al. Pharmacotherapy for systolic dysfunction. Am J Cardiol 1997;80:14-27H. 28. Psaty BM, Smith NL, Siscovick DS, et al. Health outcomes associated with anti-hypertensive therapies used as first-line agents: a systematic review and meta-analysis. JAMA 1997;277:739-45. 29. MacMahon S, Rodgers A. The effects of blood pressure reduction in older patients: an overview of five randomized controlled trials in elderly hypertensives. Clin Exp Hypertens 1993;15:967-78. 30. Hoes AW, Grobbee DE. Diuretics and risk of sudden death in hypertension: evidence and potential implications. Clin Exp Hypertens 1996;18:523-35. 31. Freis ED. The efficacy and safety of diuretics in treating hypertension. Ann Intern Med 1995;122:223-6. 32. Uretsky BF, Sheahan RG. Primary prevention of sudden cardiac death in heart failure: will the solution be shocking? J Am Coll Cardiol 1997;30:1589-97. 33. The RALES Investigators. Effectiveness of spironolactone added to an angiotensin-converting enzyme inhibitor and a loop diuretic for severe congestive heart failure (The Randomized Aldactone Evaluation Study [RALES]). Am J Cardiol 1996;78:902-7. 34. Pitt B, Zannad F, Remme WJ, et al. The effect of spironolactone on morbidity and mortality in patients with severe heart failure. N Engl J Med 1999;341:709-17. 35. Passamani E. Clinical trials: are they ethical? N Engl J Med 1991; 324:1589-92. 36. Naylor CD, Guyatt GH, for the Evidence-Based Medicine Working Group. Users’ Guides to the Medical Literature. X. How to use an article reporting variations in the outcomes of health services. JAMA 1996;275:554-8. 37. Michels KB, Rosner BA, Manson JE, et al. Prospective study of a calcium channel blocker use, cardiovascular disease, and total mortality among hypertensive women: the Nurses’ Health Study. Circulation 1998;97:1540-8. 38. Myerburg RJ, Mitriani R, Interian A, et al. Interpretation of outcomes of antiarrhythmic clinical trials: design features and population impact. Circulation 1998;97:1514-21. 39. Luu M, Stevenson WG, Stevenson LW, et al. Diverse mechanisms of unexpected cardiac arrest in advanced heart failure. Circulation 1989;80:1675-80. 40. Ziesche S, Rector TS, Cohn JN. Interobserver discordance in the classification of mechanisms of death in studies of heart failure. J Card Fail 1995;1:127-32. 41. Pratt CM, Greenway PS, Schoenfeld MH, et al. Exploration of the precision of classifying sudden death: implications for the interpretation of clinical trials. Circulation 1996;93:519-24. 42. Goldman S, Johnson G, Cohn JN, et al. Mechanism of death in heart failure: the Vasodilator-Heart Failure Trials. Circulation 1993;87(6 Suppl):24-31. 43. Cooper HA, Dries DL, Davis CE, et al. Diuretics and risk of arrhythmic death in patients with left ventricular dysfunction. Circulation 1999;100:1311-5. 44. Ferreira A, Bettencourt P, Cortez M, et al. Angiotensin-convertingenzyme inhibitors in heart failure: physicians’ prescribing behavior. J Card Fail 1997;3:295-302.