Our study demonstrated that human albumin administration was associated with higher risk-adjusted in-hospital mortality as well as longer hospital and ICU LOS, while had no association with output at 24-hour in HF patients with serum albumin level ≤ 2.9g/dl than non-albumin administration for the first time.
Serum albumin synthesized by liver is provide a variety of physiological functions. It accounts for 80% total plasma colloid oncotic pressure in the plasma, as well as maintains capillary membrane integrity. Therefore, it keep the balance across the capillary wall[5]. Previous studies demonstrated that albumin promote the effects of anticoagulant and antiplatelet [9–12]. It is also have suppressive effects on inflammation response by proinflammatory mediators[13]. As an anti-inflammatory agent, it can selectively inhibit cytokine stimulate endothelial activation, which is known to be a critical initiating pathway in atherosclerosis[14]. The concentration of serum albumin can be attributed to synthesis rates, degradation, and distribution inside and outside of the vascular compartments. Serum albumin concentration associated with a reliable prognostic value for cardiovascular disease(CVD) since a serendipitous result in 1989[15]. HF is a complex systemic clinical syndrome characterized by cardiocirculatory injuries, complex interplay of neurohormonal and biochemical disorders and organ failure. Albumin synthesis is reduced in patients with heart failure due to intestinal wall edema, liver congestion, or even congestive cirrhosis. Hypoalbuminemia is common occurs in HF patients and represents a powerful independent predictor of adverse outcomes and worse survival of the patients[16]. In addition, low albumin level is associated with incidence of preserved ejection fraction HF (HFpEF)[17, 18]. In our study, we analyzed the relationship between initial serum albumin levels and in-hospital mortality using ROC curves. In line with previous studies, we found hypoalbuminaemia serves as predictor of all-cause mortality in HF patients in hospital. The optimal cut-off value was 2.9g/dl.
Since serum albumin is so important to the body, is there any benefit to infusing albumin in patients with hypoproteinemia? Albumin administration may improve organ function in hypoalbuminemic critically ill patients[19]. Caraceni et al. demonstrate that albumin infusion group overall 18-month survival was significantly higher in decompensated cirrhosis patients [20]. But, the latest randomized controlled trial shows that, albumin administration increase the albumin level to a target of 30 g per liter or more had no good for hospitalized patients with decompensated cirrhosis[21]. Although previous studies presented conflicting results concerning the impact of albumin administration on hypoproteinemia, human albumin has been widely used for hypoproteinemia therapy. We analyzed the data from MIMIC-III and found albumin administration was associated with decreased mortality by univariate logistic regression analyses, but had adverse result by multivariate logistic regression analyses in HF with albumin level ≤ 2.9g/dl. After adjustment for confounding factors, albumin administration in HF plus hypoalbuminemic patients were relevant to higher in-hospital mortality and tended to longer hospital and ICU LOS than non-albumin administration subgroups. Furthermore, we investigated rough relationship between the amount of human albumin infusion at 24 hours and hospital mortality, and found that the elevated amount of albumin infusion is related to higher rates of hospital mortality.
Volume overload, a common problem in HF patients, treated with volume restriction and diuretic usage. However, furosemide resistance is a big problem in HF. Decreased amount of drug transferred to the target site is one of the vital mechanisms result in diuretic resistance[22]. Albumin as a biovehicle, can transported drugs to target organ[3, 23]. Hypoalbuminemia can reduce diuretic agent secretion to tubular lumen[24, 25]. Previous clinical study revealed that intravenous injection of furosemide bound to albumin can rapidly increase the urine volume of hypoproteinemia patients who present obvious resistance to this diuretic[26]. Furosemide plus albumin for acute lung injury/acute respiratory distress syndrome patients with hypoproteinemic therapy significantly improved oxygenation, net negative fluid balance and hemodynamic stability than furosemide therapy[27]. Chalasani et al. demonstrated that the parameters, including urinary, sodium excretion and urinary furosemide excretion, are similar between the furosemide group and furosemide plus albumin group in cirrhotic patients with ascites[28]. Currently, there is no direct evidence about benefits of albumin infusion in heart failure. But some physicians use human albumin infusion to overcome furosemide resistance in HF inpatients especially with hypoalbuminemic. In the present study, we found the volume of 24-hour output had no relevance to albumin injection. Besides, albumin as a colloidal solution also increases the volume load and may further aggravate the symptom of HF. Because of retrospective study with limitation data sets, we abandoned central venous pressure (CVP) measurement. Whether the CVP was decreased need to be further explored.
Our trial was the first study exploring albumin therapy in HF with hypoalbuminemic and did not found clinically important effect in hospitalized patients. On the contrary, we demonstrated that albumin administration in HF with hypoalbuminemic was relevant to increased hospitalized mortality, longer hospital and ICU LOS. There are several interpretations for the results. Firstly, albumin as a colloidal solution, can increase the volume load and further aggravate adverse outcomes of HF. While albumin may reduce resistance to diuretics and increase urine production, the increased volume load overweight increased urine amount. Secondly, albumin physiological functions concerning suppressive effects on inflammation response, anticoagulant and antiplatelet takes longer time to kick in. The maximum duration of our study was 90-day and we found in-hospital mortality is increased and 90-day mortality have not changed after albumin infusion. The mortality might be decreased at longer follow-up time. Patients who would benefit from albumin infusion in the long-term need be further investigated. Lastly, albumin modifications contain reversible oxidized albumin form human nonmercaptalbumin1(HNA1) and irreversible oxidized albumin form HNA2[29]. We could pay more attention to the ratio of HNA1 to HNA2 than the amount of serum albumin alone and focus on the conversion of HNA1 into human mercaptalbumin[30].
In our study, comorbidities including hypertension, diabetes and coronary disease were associated with decreased in-hospital mortality by logistic regression analyses. This might be because physicians payed more attention to the exacerbation of heart failure in these patients and therefore took more notice to medication and volume restriction.
Our results simply reflect the true effect of albumin administration in real-world clinical practice and did not show benefit effects in HF patients with hypoalbuminemia. These findings argue against the clinical utilization of albumin therapy strategy. The results can be probably extrapolated to other hypoproteinemia, such as cirrhotic patients with ascites. Clinicians should not waive human albumin therapy but consider how to utilize human albumin in an appropriate way.
Several limitations in our study should be considered. First, the amount of N-terminal pro-brain-type natriuretic peptide (NT-proBNP) data extracted from MIMIC-III was about 2000. Due to too many missing values, we had to discard the variable of BNP. Hence, the definition of HF was based on ICD-9 and admitting diagnosis. In addition, we could not get a cardiac function grade from this database. Second, our study was a retrospective single-center cohort study based on MIMIC-III database which ranged from 2001 to 2015, and the results might have only reflected local practices during the period. Third, some variables, such as 24-hour input amount, body mass index (BMI) and the use of diuretics in our study, which had over 50% missing values, were eliminated. Some other unmeasured confounders such as different haemodynamic monitoring techniques usage, including central venous pressure measurement and transthoracic echocardiography were uncertain. These confounding factors and variables could affect our results. Forth, there were multiple factors that influence the initial CVP levels, including the rate of albumin infusion, as well as other solution, were difficult to adjust in a retrospective observational study. Besides, we did not investigate the effect of protein concentration on the results in the further research. Finally, the causal relationship between human albumin administration and mortality was not explored thoroughly, and the LOS longer in HF cohort in our study might have nothing to do with albumin administration, as LOS was related with complex clinical practice. A randomized study comparing the effect of human albumin administration and non-human albumin administration is needed in the future.