The bilirubin level has been associated with worse outcomes, but it has not been studied as a predictor for the mode of death in patients with systolic heart failure. The Prospective Randomized Amlodipine Evaluation Study (PRAISE) cohort (including New York Heart Association class IIIB-IV patients with left ventricular ejection fraction <30%, n = 1,135) was analyzed, divided by bilirubin level: ≤0.6 mg/dl, group 1; >0.6 to 1.2 mg/dl, group 2; and >1.2 mg/dl, group 3. Multivariate Cox proportional hazards models were used to determine the association of bilirubin with the risk of sudden or pump failure death. Total bilirubin was entered as a base 2 log-transformed variable (log 2 bilirubin), indicating doubling of the bilirubin level corresponding to each increase in variable value. The higher bilirubin groups had a lower ejection fraction (range 19% to 21%), sodium (range 138 to 139 mmol/L), and systolic blood pressure (range 111 to 120 mm Hg), a greater heart rate (range 79 to 81 beats/min), and greater diuretic dosages (range 86 to 110 furosemide-equivalent total daily dose in mg). The overall survival rates declined with increasing bilirubin (24.3, 31.3, and 44.3 deaths per 100 person-years, respectively, for groups 1, 2, and 3). Although a positive relation was seen between log 2 bilirubin and both pump failure risk and sudden death risk, the relation in multivariate modeling was significant only for pump failure mortality (hazard ratio 1.47, 95% confidence interval 1.19 to 1.82, p = 0.0004), not for sudden death mortality (hazard ratio 1.21, 95% confidence interval 0.98 to 1.49, p = 0.08). In conclusion, an increasing bilirubin level was significantly associated with the risk of pump failure death but not for sudden death in patients with severe systolic heart failure.
Assessing the risk of a specific mode of death in patients with systolic heart failure (HF) could be useful to assist in clinical decision making regarding treatment options such as defibrillators or ventricular assist devices. The 2 most common modes of death in patients with systolic HF are death from progressive HF (pump failure) and sudden death. Although several variables have been found to be associated differentially with the mode of death, this relation has not been assessed with bilirubin, which might be a good discriminator for the risk of pump failure death versus sudden death. Elevations in bilirubin result from a combination of reduced forward flow from left-sided HF and hepatic congestion due to right-sided HF, which frequently results from left heart dysfunction and, therefore, generally indicates more advanced stages of HF. Thus, bilirubin likely correlates with the severity of ventricular dysfunction, and, in fact, elevated bilirubin has been found in some studies to be an independent predictor of overall mortality. To define this association in greater detail, we analyzed the Prospective Randomized Amlodipine Evaluation Study (PRAISE) cohort of patients to determine whether the serum bilirubin level predicts the mode of death (pump failure vs sudden death) in patients with severe systolic HF.
Methods
The details of the PRAISE have been previously published; 1,153 patients with advanced HF were enrolled in PRAISE from 1992 to 1994 and were followed up for a median of 14 months. The inclusion criteria were New York Heart Association class IIIB or IV and a left ventricular ejection fraction of <30% with maximum medical therapy, including digoxin, diuretics, and angiotensin-converting enzyme inhibitors. The patients were randomized to treatment with oral amlodipine or placebo. The mode of death was adjudicated by Endpoints Committee review. This secondary analysis included all patients who had had the total bilirubin measured at study entry and who had clinical data available (18 subjects were excluded because of missing bilirubin data). Liver failure or insufficiency was not a specific exclusion criterion for the PRAISE.
For the purposes of describing the baseline characteristics, the population was divided into 3 groups according to the total bilirubin level, with cutoffs of 0.6 mg/dl (median for the entire study group) and 1.2 mg/dl (upper normal limit for the laboratory test). The bilirubin level was ≤0.6 mg/dl for group 1, >0.6 to 1.2 mg/dl for group 2, and >1.2 mg/dl for group 3. The baseline characteristics were compared using chi-square or analysis of variance testing for categorical or continuous variables, respectively.
Multivariate Cox proportional hazards modeling was performed to examine the association between total bilirubin and the outcomes of pump failure death or sudden death. Death from other causes was treated as a censored event. Total bilirubin was entered as a base 2 log-transformed variable (log 2 bilirubin), indicating a doubling of the bilirubin level corresponding to each increase in the variable value. Because bilirubin was not a normally distributed variable, it was necessary to perform log transformation for entry into the models, and a log 2 format was thought to be more easily interpretable, particularly in a clinical context. The candidate variables were first tested in univariate models and were included in the multivariate model if p <0.05. For the purposes of the present analysis, doses of diuretics other than furosemide were converted to equivalent furosemide doses according to the following equivalencies, such as were used in the Seattle Heart Failure Model and previous studies: furosemide 80 mg equals bumetanide 3 mg equals ethacrynic acid 100 mg, and metolazone 2 mg equals hydrochlorothiazide 25 mg. For the ease of comparison of variable effects between models, the same variables were entered into both models. The variables significant on univariate testing for both pump failure and sudden death mortality were age, blood urea nitrogen, creatinine, sodium, systolic blood pressure, diastolic blood pressure, white blood cell count, furosemide-equivalent diuretic dose, New York Heart Association class IV versus IIIB, white race, and uric acid level. The variables significant on univariate testing for pump failure death only were albumin, body mass index, hematocrit, hemoglobin, left ventricular ejection fraction, sodium, and ischemic etiology. Those significant on univariate testing for sudden death only were sodium, allopurinol use, and male gender. For the variables that were closely correlated clinically (e.g., blood urea nitrogen and creatinine, or systolic and diastolic blood pressure), the variable with the more significant chi-square value was used. To assess the respective strengths of the association between bilirubin and the mode of death, the Cox regression analysis was first stratified by the mode of death (pump failure or sudden death). Next, a second model was performed, including all predictors that were significant for either cause of death and also an interaction term between log 2 bilirubin and the mode of death. All variables remained in the models.
A p value <0.05 was considered statistically significant. Statistical analysis was performed using SAS, version 9.2 (SAS Institute, Cary, North Carolina).
Results
A total of 1,135 patients were included in the present analysis. The baseline characteristics are listed in Table 1 . The data are presented as the mean ± SD, unless otherwise indicated. The higher bilirubin groups had a greater proportion of men, patients with New York Heart Association class IV versus IIIB, and, on average, a lower left ventricular ejection fraction, greater hemoglobin and uric acid levels, lower sodium, lower blood pressure, greater heart rate, and greater total daily diuretic dosage. All patients were taking angiotensin-converting enzyme inhibitors and diuretics.
| Variable | Group 1 (Bilirubin ≤0.6 mg/dl) | Group 2 (Bilirubin >0.6–1.2 mg/dl) | Group 3 (Bilirubin >1.2 mg/dl) | p Value ∗ | Entire Group |
|---|---|---|---|---|---|
| Patients (n) | 581 | 396 | 158 | — | 1135 |
| Bilirubin (mg/dl) | — | ||||
| Mean ± SD | 0.5 ± 0.1 | 0.9 ± 0.2 | 1.9 ± 0.7 | 0.8 ± 0.5 | |
| Median | 0.5 | 0.8 | 1.6 | 0.6 | |
| Interquartile range | 0.4–0.6 | 0.7–1.0 | 1.4–2.0 | 0.5–0.9 | |
| Range | 0.1–0.6 | 0.7–1.2 | 1.3–6.3 | 0.1–6.3 | |
| Men | 402 (69.2%) | 320 (80.8%) | 142 (89.9%) | <0.0001 | 864 (76.1%) |
| Age (yrs) | 65.4 ± 10.5 | 65.0 ± 10.7 | 62.7 ± 13.1 | 0.02 | 64.9 ± 11.0 |
| Body mass index (kg/m 2 ) | 27.1 ± 5.8 | 26.8 ± 5.4 | 25.9 ± 5.4 | 0.047 | 26.8 ± 5.6 |
| White race | 477 (82.1%) | 308 (77.8%) | 116 (73.4%) | 0.04 | 901 (79.4%) |
| Ischemic etiology | 380 (65.4%) | 246 (62.1%) | 98 (62.0%) | 0.5 | 724 (63.8%) |
| Diabetes | 225 (38.7%) | 150 (37.9%) | 45 (28.5%) | 0.055 | 420 (37%) |
| New York Heart Association class | 0.02 | — | |||
| III | 484 (83.3%) | 312 (78.8%) | 117 (74.1%) | — | 913 (80.4%) |
| IV | 97 (16.7%) | 84 (21.2%) | 41 (26.0%) | — | 222 (19.6%) |
| Systolic blood pressure (mm Hg) | 120 ± 18 | 116 ± 17 | 111 ± 16 | <0.0001 | 118 ± 18 |
| Diastolic blood pressure (mm Hg) | 73 ± 10 | 73 ± 10 | 72 ± 11 | 0.6 | 72 ± 10 |
| Heart rate (beats/min) | 79 ± 12 | 81 ± 14 | 81 ± 14 | 0.009 | 80 ± 13 |
| Left ventricular ejection fraction (%) | 21 ± 6 | 20 ± 6 | 19 ± 6 | <0.0001 | 21 ± 6 |
| Smoking | 86 (14.8%) | 69 (17.4%) | 27 (17.1%) | 0.5 | 182 (16%) |
| Angiotensin-converting enzyme inhibitor | 581 (100%) | 396 (100%) | 158 (100%) | — | 1135 (100%) |
| Allopurinol | 45 (7.8%) | 47 (11.9%) | 19 (12.0%) | 0.06 | 111 (9.8%) |
| Diuretic | 581 (100%) | 396 (100%) | 158 (100%) | — | 1135 (100%) |
| Potassium-sparing diuretic | 18 (3.1%) | 7 (1.8%) | 4 (2.5%) | 0.4 | 29 (2.6%) |
| Furosemide | 558 (96.0%) | 385 (97.2%) | 155 (98.1%) | 0.3 | 1098 (96.7%) |
| Furosemide-equivalent total daily dose (mg) | 86 ± 70 | 104 ± 82 | 110 ±67 | <0.0001 | 96 ± 74 |
| Metolazone | 101 (17.4%) | 86 (21.7%) | 48 (30.4%) | 0.0005 | 23 (20.7%) |
| Albumin (g/dl) | 4.1 ± 0.4 | 4.2 ± 0.4 | 4.0 ± 0.4 | <0.0001 | 4.1 ± 0.4 |
| Blood urea nitrogen (mg/dl) | 25 ± 15 | 26 ± 15 | 28 ± 14 | 0.08 | 26 ± 15 |
| Creatinine (mg/dl) | 1.4 ± 0.5 | 1.4 ± 0.4 | 1.4 ± 0.4 | 0.8 | 1.4 ± 0.5 |
| Hemoglobin (g/dl) | 13.6 ± 1.6 | 14.2 ± 1.7 | 14.3 ± 1.8 | <0.0001 | 13.9 ± 1.7 |
| Hematocrit (%) | 40.8 ± 4.8 | 42.7 ± 5.1 | 43.4 ± 5.3 | <0.0001 | 41.8 ± 5.1 |
| Sodium (mmol/L) | 139 ± 4 | 139 ± 4 | 138 ± 4 | 0.0005 | 139 ± 4 |
| Uric acid (mg/dl) | 8.5 ± 2.5 | 9.1 ± 2.4 | 9.8 ± 2.8 | <0.0001 | 8.9 ± 2.6 |
| White blood cell count (K/mm 3 ) | 7.8 ± 2.3 | 7.6 ± 2.3 | 7.4 ± 2.3 | 0.1 | 7.7 ± 2.3 |
| Follow-up period (days) † | 488 ± 256 | 426 ± 258 | 391 ± 270 | <0.0001 | 453 ± 261 |
∗ For comparison among bilirubin groups.
Of the entire study group, 409 (36%) died (29 deaths per 100 person-years of follow-up). Of these deaths, 164 (14.5%) were pump failure deaths and 183 (16.1%) were sudden deaths (11.6 and 13.0 deaths per 100 person-years of follow-up, respectively). The event rates per 100 person-years of follow-up are listed in Table 2 . An increasing number of events was seen with increasing bilirubin level group. A trend was found toward a greater number of sudden and pump failure deaths in group 2 versus group 1 that was not statistically significant; however, the increase in the number of events in group 3 versus group 1 was significant, more so for pump failure death than for sudden death. The Kaplan-Meier survival curves for pump failure mortality and sudden death mortality, stratified by the bilirubin group, are shown in Figures 1 and 2 , respectively.
| Per 100 Person-yrs | Group 1 (Bilirubin ≤0.6 mg/dl) | Group 2 (Bilirubin >0.6–1.2 mg/dl) | p Value ∗ | Group 3 (Bilirubin >1.2 mg/dl) | p Value ∗ |
|---|---|---|---|---|---|
| All-cause mortality | 24.3 | 31.3 | 0.02 | 44.3 | <0.0001 |
| Sudden cardiac death | 10.9 | 14.5 | 0.09 | 18.3 | 0.01 |
| Pump failure death | 9.1 | 12.3 | 0.09 | 21.2 | <0.0001 |
| Other death | 4.2 | 4.5 | 0.8 | 4.7 | 0.8 |
The results of multivariate Cox regression models for the outcomes of pump failure death and sudden death are listed in Tables 3 and 4 , respectively. Although log 2 bilirubin was significantly associated with the risk of pump failure death in the adjusted analysis (hazard ratio [HR] 1.47, 95% confidence interval [CI] 1.19 to 1.82, p = 0.0004), it was not significantly associated with the risk of sudden death in the adjusted analysis (HR 1.21, 95% CI 0.98 to 1.49, p = 0.08). In the multivariate model that included the interaction term between bilirubin and the mode of death, the interaction term was not significant (p = 0.1), confirming the absence of a significant difference in the overall direction of the relation between bilirubin and the risk of either mode of death (i.e., a positive relation was found between the log 2 bilirubin value and the risk of death for both modes of death). Other variables significantly associated with an increased risk of pump failure mortality included a greater uric acid level, diuretic dose, and white blood cell count; a lower systolic blood pressure, serum sodium, body mass index, and hemoglobin; an ischemic etiology; allopurinol use; and white race ( Table 3 ). An increased risk of sudden death was associated with a greater diuretic dose, lower systolic blood pressure, white race, and allopurinol use ( Table 4 ). At an α of 0.05, the study had a power of >0.999 for log 2 bilirubin, with a HR of 1.47 (95% CI 1.19 to 1.82), to detect a significant difference in pump failure death, and a power of 0.83 for log 2 bilirubin, with a HR of 1.21 (95% CI 0.98 to 1.49) to detect a significant difference in sudden death.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
