Low serum albumin: A significant predictor of reduced survival in patients with chronic heart failure

Background Low serum albumin is common in patients with chronic heart failure (HF). Hypothesis Albumin may have an impact on clinical outcome in HF. We evaluated the effect of albumin levels on clinical outcome in a real‐world cohort of patients with HF. Methods All patients with HF at a health maintenance organization were followed for cardiac‐related hospitalizations and death. Results A total of 5779 HF patients were included in the study; mean follow‐up was 576 days; median serum albumin was 4.0 g/dL (interquartile range 3.7‐4.2), and 12% of the patients had hypoalbuminemia (albumin<3.5 g/dL). Low albumin was associated with increasing age, higher urea and C‐reactive protein, lower sodium, hemoglobin, iron, less treatment with angiotensin‐converting enzyme inhibitor or angiotensin receptor blocker, reduced right ventricular function, and pulmonary hypertension. Cox regression analysis after adjustment for significant predictors demonstrated that decreasing quartiles of albumin was significantly associated with mortality: Lowest quartile compared to highest: hazard ratio (HR) 5.74, 95% confidence interval (CI) 4.08 to 8.07, P < 0.001. Cox regression analysis of albumin as a continuous parameter using restricted cubic splines after adjustment for significant parameters demonstrated that reduced albumin levels were directly associated with increased mortality (P < 0.001 for the adjusted model). Decreasing quartiles of albumin were also a significant predictor of increased cardiac‐related hospitalizations. A decrease in albumin on follow‐up was an independent predictor of increased mortality by Cox regression analysis: HR 2.58, 95% CI 2.12 to 3.14, P < 0.001. Conclusions Low albumin provides important information regarding several detrimental processes in HF and is a significant predictor of a worse outcome in these patients.


| BACKGROUND
Heart failure (HF) has emerged as a major epidemic and is a significant public health burden. It is associated with considerable morbidity and mortality. 1 There are numerous clinical parameters that predict clinical outcome. Albumin, a standard clinical parameter that is associated with multiple parameters affecting outcome including nutritional, inflammatory, and volume status of HF patients, should have a significant impact on clinical outcome in these patients. Despite this, there is limited and conflicting evidence in the literature regarding this parameter as an important predictor of outcome in chronic HF. [2][3][4] We hypothesized that this parameter would have a significant impact on clinical outcome in patients with chronic HF. We evaluated the impact of serum albumin levels on clinical outcome in a large real-world cohort of patients with chronic HF.

| METHODS
Clalit Health Services is the largest health maintenance organization (HMO) in Israel. It has a central computerized database in which all members have a complete digital record. The database includes demographics, comprehensive clinical data, diagnoses, and all laboratory data undertaken in a single centralized laboratory of the HMO. We identified and retrieved electronically from the computerized database all members with a diagnosis of HF as coded by the database in Jerusalem using the International Classification of Diseases, Ninth Revision (ICD-9) code 428. About 6946 patients had a diagnosis of HF. Validation of the diagnosis of HF was performed on a randomly computer-generated 5% of the diagnosed HF patients (N = 338) as previously described. 5 Clinical parameters in this group of patients were statistically comparable with the whole HF cohort. We reviewed all available data from medical records and hospital admissions. In this group, 99% fulfilled the European Society of Cardiology (ESC) criteria for the diagnosis of HF 6 based on typical symptoms, signs and structural or functional abnormalities of the heart per echocardiography.
Only 1% had equivocal clinical data for the diagnosis of HF. Natriuretic peptides are not routinely performed in Israel and were not available for analysis. Serum albumin levels are routinely performed as part of a periodic work-up in all members of the HMO. All patients in this cohort with serum albumin levels measured within 4 months of the time the database was established were included.
About 84% of the patients (N = 5779) had an albumin level available for analysis. We also collected an additional albumin measurement, closest to the end of the study. Echocardiography data were available in digital form for analysis in 26% of the cohort (N = 1489). The clinical characteristics of these patients were very similar to the whole cohort. All echocardiography data including left ventricular ejection fraction, categorized into preserved (EF ≥ 50%) and reduced (EF < 50%) as well as measurements of dimensions were performed according to standard recommendations of the American Society of Echocardiography (ASE) and were acquired and verified by qualified personnel. All hospitalizations in cardiac and internal medicine departments including cardiac and internal intensive care units were retrieved and analyzed. Data on mortality were retrieved from the

| Clinical parameters
The study cohort included 5779 HF patients. Supporting information Figure SS1 presents the distribution of serum albumin levels in the HF cohort. Mean albumin levels were 3.9 ± 0.43 g/dL; Median 4.0 g/dL (interquartile range [IQR] 3.7-4.2 g/dL). About 12% of the patients (N = 704) had hypoalbuminemia (albumin<3.5 g/dL). The patient cohort was divided in quartiles according to albumin levels. Table 1 presents the demographics and clinical parameters of the patients stratified according to these quartiles. Low albumin was associated with increasing age, women gender, hypertension, peripheral vascular disease, prior stroke, dementia, and increased Charlson Comorbidity Index but less with ischemic heart disease. Low albumin was associated with lower body mass index, higher creatinine, and urea but lower sodium, hemoglobin, glucose, cholesterol, triglycerides, calcium, and iron. Low albumin was associated with increased Creactive protein. Patients with low albumin were treated less with angiotensin-converting enzyme inhibitors (ACE-Is) or angiotensin receptor blockers (ARBs), beta blockers and thiazides and more with furosemide. Low albumin was associated with preserved LV EF, smaller dimensions of the left ventricle, reduced RV systolic function, severe tricuspid regurgitation, and pulmonary hypertension. Multivariable linear regression demonstrated that older age, non-ischemic heart disease, increased urea, low hemoglobin and sodium, nontreatment with ACE-I/ARB and thiazide, low iron and increased CRP were predictive of low serum albumin (R 2 = 0.278, P < 0.0001, Table SS1).

| Albumin levels and clinical outcome
The median follow-up period was 576 days. The overall mortality rate during this period was 14%. Survival rate by Kaplan-Meier analysis demonstrated that decreasing albumin levels were directly associated with reduced survival (decreasing albumin quartiles: 95.7 ± 0.6% vs 92.9 ± 0.8% vs 88.3 ± 0.8% vs 67.2 ± 1.2%, P < 0.001). Decreasing albumin levels were also directly associated with decreased event-free survival from death or cardiovascular-hospitalizations (decreasing    albumin quartiles: 58.6 ± 1.3% vs 52.0 ± 1.5% vs 42.9 ± 1.2% vs 23.4 ± 1.1%, P < 0.001). Multivariable Cox regression analysis after adjustment for significant predictors demonstrated that decreasing albumin levels were a significant predictor of mortality (Table 2 and Figure 1A). After adjustment for other significant predictors (see Table 2 for significant clinical predictors included), low albumin levels (<3.8 g/dL) was associated with a 5-fold increase in mortality compared to the highest albumin quartile (>4.2 g/dL), with a hazard ratio of 5.74, 95% confidence interval (CI) 4.08 to 8.07, P < 0.001. Inclusion of HF medications (Table 3) demonstrated a very similar result with a direct relation between reduced albumin levels and mortality. We performed a further sensitivity analysis by analyzing albumin levels as a continuous parameter using restricted cubic splines. Cox regression analysis after adjustment for significant parameters included in Table 2 demonstrated a direct inverse relationship between albumin and mortality (Figure 2A). This analysis demonstrated that any decrease in albumin was a predictor of mortality with a continuous increase in the risk with lower albumin levels ( Figure 2A (Table 3 and Figure 1B).
After adjustment for other significant predictors, low albumin levels (<3.8 g/dL) was associated with a 2-fold increase in death and cardiovascular hospitalization compared with the highest albumin quartile (>4.2 g/dL), with a hazard ratio of 2.19, 95% CI 1.93 to 2.48, P < 0.001. Inclusion of HF medications (Table 3)    in albumin levels as a continuous parameter using restricted cubic splines was performed. Cox regression analysis after adjustment for significant parameters including baseline albumin demonstrated a direct relationship between the decrease in albumin and mortality ( Figure 2B). This analysis demonstrated that any decrease in albumin over time was a predictor of mortality with a continuous increase in the risk with decreasing albumin levels on follow-up ( Figure 2B),  Data are presented as hazard ratio (95% confidence interval), P value. Parameters that were included in the multivariable analysis model were age, gender, ischemic heart disease, diabetes, hyperlipidemia, hypertension, atrial fibrillation, log-transformed body mass index, log-transformed pulse, log-transformed serum urea levels, square root-transformed estimated glomerular filtration rate, hemoglobin, and serum sodium. Parameters that were included in the multivariable and drugs analysis included the above parameters and the drug treatment with angiotensin-converting enzyme inhibitor/angiotensin receptor blocker, beta blocker, furosemide, spironolactone, thiazide, and digoxin. osmotic pressure and microvascular integrity, ligand-binding and transport of substances, antioxidant and antithrombotic functions, and enzymatic activities. 10 Low albumin in HF promotes and aggravates congestion due to reduced intravascular colloid osmotic pressures, 11 increases oxidative stress, 12 inflammation, 10 and the susceptibility to infections. Therefore, low albumin is a summation of numerous deleterious factors in HF patients and would be expected to give important prognostic information in HF. Despite this, there is sparse data in the literature regarding this parameter as a prognostic factor in chronic HF.

| Changes in albumin and impact on outcome
In the setting of acute HF, albumin has shown to have an important impact on survival and several papers demonstrate that albumin is an important prognostic parameter of clinical outcome. 8,[13][14][15][16][17][18] Interestingly, albumin was not helpful in guiding diuretic therapy or predicting short-term outcome in acute HF diuretic therapy trials. 19 In the setting of chronic HF, data are much less abundant and less con- The prevalence of hypoalbuminemia in HF patients in this study was 12%. This is lower than previously published data, with a prevalence of 18% or higher. [2][3][4] This could be due the fact that the present cohort was a real-world community-based study which represents a broader spectrum of stable HF patients and thus perhaps represents the real-world prevalence of hypoalbuminemia in HF.
In the present study, low albumin was associated with increasing age, lower kidney function, lower hemoglobin, sodium and iron but less with ACE-I/ARB treatment. It was associated with increased Creactive protein but not with body mass index. This suggests that the nutritional state is perhaps not the dominant reason for reduced albumin in this HF cohort and other reasons such as inflammation may be more important. This was also seen in a previous study. 2 The minimal association of albumin with body mass index could be related to the fact that reduction in albumin occurs only at a more advanced stage of nutritional deficiency and reduction in weight and is largely affected by other factors such as inflammation. However, it is possible that body mass that is derived from body weight is not accurate enough in HF to estimate dry body (muscle) weight, which is the best   16,22 and provide prognostic information. An increase in albumin levels during hospitalization predicted a better long-term outcome, while a decrease predicted a worse outcome. In the acute setting, these changes are related to the acute event severity, sequel and its progression. In the chronic setting, these changes probably reflect the progression of the chronic disease with albumin reflecting the detrimental biological processes associated with low albumin and its significance to prognosis. Limitations of this study: The present study was an observational study. Data regarding clinical parameters and drug therapy was based on a digitized database. Although this database was validated and found to be highly accurate, not all data could be verified. While we tried to adjust for clinical relevant parameters, not all clinical parameters were available and it is impossible to adjust for all variables that may affect outcome. In particular, data on functional capacity and natriuretic peptide levels were not available. In addition, the cohort was a community-based cohort and the findings may not be applicable in more advanced or hospital-based HF cohorts.
In conclusion, low serum albumin provides important information regarding several detrimental processes in HF and is a significant predictor of a worse outcome in these patients. A decline in albumin was also a predictor of reduced survival.

CONFLICTS OF INTEREST
The authors have no conflicts of interest to report.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.