Journal of Critical Care 51 (2019) 139–144

Contents lists available at ScienceDirect

Journal of Critical Care

journal homepage: www.journals.elsevier.com/journal-of-critical-care

Serum albumin as a risk factor for death in patients with prolonged

sepsis: An observational study
Ryosuke Takegawa a,⁎, Daijiro Kabata b, Kentaro Shimizu a, Saya Hisano b, Hiroshi Ogura a,
Ayumi Shintani b, Takeshi Shimazu a
a
Department of Traumatology and Acute Critical Medicine, Osaka University Graduate School of Medicine, 2-15 Yamada-oka, Suita, Osaka 565-0871, Japan
b
Department of Medical Statistics, Osaka City University Graduate School of Medicine, 1-4-3 Asahi-cho, Abeno-ku, Osaka city, Osaka 545-0051, Japan

a r t i c l e i n f o a b s t r a c t

Keywords: Purpose: The aim of this study was to evaluate an association between nutritional biomarkers and prognosis in
Sepsis septic patients.
Time-dependent Cox proportional hazard re- Methods: We retrospectively searched the association between nutritional biomarkers including serum albumin
gression analysis (Alb), total protein (TP), total cholesterol (T-chol), and cholinesterase (ChE), and prognosis for septic patients
Albumin
treated in the ICU for N7 days. We used time-dependent Cox proportional hazard regression analysis to resolve
Prognosis
Biomarkers
the difference of the statistical weight of each day's data for all 14 consecutive days among individual sepsis
patients. The covariates were based on the minimum moving values determined from 1 day, 3 days, 7 days,
and 14 days of serial data. The values of these covariates and ICU survival were considered as outcomes.
Results: We included 136 septic patients. The decreases in the values of Alb, TP, T-chol, and ChE were significantly
associated with the risk of death in the septic patients (p b .05). Especially, the daily changes of Alb were signif-
icantly associated with mortality during the ICU stay (p b .05).
Conclusions: We found that the changes in serial data of the nutritional markers of Alb, TP, T-chol, and ChE
reflected the higher risk of death in patients with prolonged sepsis.
© 2019 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction
According to the ESPEN (European Society for Clinical Nutrition and
Metabolism) guideline, serum protein markers are the reflection of an
acute phase response and do not accurately represent nutritional status
in the ICU setting [1]. These nutritional biomarkers are often low in
the acute phase of sepsis because of decreased protein synthesis and
dilution by the systemic inflammatory response. However, it is not
clear whether serial changes in nutritional markers such those as seen
in progressive hypoalbuminemia reflect prognosis in patients with sep-
sis during their ICU stay.
Preoperative serum albumin is prognostic factor for complications
after surgery and also associated with impaired nutritional status [2].
Abbreviations: Alb, albumin; APACHE, Acute Physiology and Chronic Health
Evaluation; APS, Acute Physiology Score; ChE, cholinesterase; ICU, intensive care unit;
IQR, interquartile range; SIRS, systemic inflammatory response syndrome; SOFA,
Sequential Organ Failure Assessment; T-chol, total cholesterol; TP, total protein.
⁎ Corresponding author.
E-mail addresses: r-takegawa@hp-emerg.med.osaka-u.ac.jp (R. Takegawa),
kabata.daijiro@med.osaka-cu.ac.jp (D. Kabata), shimiken@hp-emerg.med.osaka-u.ac.jp
(K. Shimizu), ogura@hp-emerg.med.osaka-u.ac.jp (H. Ogura),
shimazu@hp-emerg.med.osaka-u.ac.jp (T. Shimazu).
The major physiological functions of serum albumin are the regulation
of both plasma oncotic pressure and capillary membrane permeability,
and ligand binding and transport. Albumin acts as a depot and carrier
for many endogenous and exogenous compounds, is involved in free
radical scavenging, and has anti-oxidant and circulatory protective
properties. Moreover, serum albumin is widely used clinically to treat
several conditions including hypovolemia, shock, burns, surgical blood
loss, and trauma [3]. In addition, tumor necrosis factor alpha is reported
to induce phosphorylation of C/EBPβ, a transcription factor that inhibits
albumin gene transcription [4]. Thus, serum albumin is diluted and its
production is inhibited during inflammation. Cholesterol is a precursor
of the steroid hormones and bile acids and is an essential constituent
of cell membranes. Total protein (TP) consists of albumin and
γ-globulin, which act in humoral immunity. Cholinesterase (ChE) is an
enzyme capable of hydrolyzing a variety of choline esters and is
produced mainly by hepatocytes.
Biomarkers of sepsis patients on admission could reflect not only the
severity but also the accumulated influences from the day of disease
onset because the day of admission of patients with sepsis is not always
the onset day of injury, such as that caused by trauma, burns, or cardiac
arrest. Nonetheless, most previous studies on these biomarkers focused
on a specific point in time, and there are few reports on how serial data
can be integrated and used in clinical practice.

https://doi.org/10.1016/j.jcrc.2019.02.004
0883-9441/© 2019 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
140 R. Takegawa et al. / Journal of Critical Care 51 (2019) 139–144

Table 1 enteral nutrition if the patient's intestines are functional, and if not func-
Patient characteristics. tional (i.e., such as with continuous vomiting or severe diarrhea), we ad-
Variable Patients minister parenteral nutrition. During the study period, we used
Total patients, n 136
25–30 kcal/IBW (ideal body weight) as the goal for the administered
Survivor, n (%) 111 (81.6) nutritional dose. This study was carried out according to the principles
Male sex, n (%) 89 (65) of the Declaration of Helsinki and was approved by the institutional
Age, yrs. (median, IQR) 70 (56.3–81) review board at Osaka University Hospital (approval no. 14242). The
Lactate value on hospital arrival (mg/dl) (median, IQR) 21 (12–36.8)
board waived the need for informed consent because this was a retro-
APS on admission (median, IQR) 10.5 (6–16)
APACHE II score on admission (median, IQR) 16 (11−23) spective study using clinical data.
SOFA score on admission (median, IQR) 4 (2–7)
Length of ICU stay, d (median, IQR) 17 (10.3–29.0)
Main diagnosis, n (%)
2.2. Methods
Pulmonary infection 40 (29)
Abdominal infection 38 (28) Data collection included that of each of the nutritional markers
Soft tissue infection 27 (20) generated from the liver (serum TP, albumin, T-chol, and ChE), patient
Urinary infection 13 (10)
characteristics, severity scores (Acute Physiology and Chronic Health
CNS infection 9 (7)
Mediastinal infection 3 (2) Evaluation II [APACHE II] score, Sequential Organ Failure Assessment
Lumbar infection 2 (1) [SOFA] score, and Acute Physiology Score [APS]) determined on admis-
Cardiac infection 1 (1) sion, and outcome.
Others 3 (2)

IQR: interquartile range, APS: Acute Physiology Score, APACHE: Acute Physiology and
Chronic Health Evaluation, SOFA: Sequential Organ Failure Assessment, ICU: intensive
care unit, CNS: central nervous system.
To resolve these problems, we analyzed the serial data of biomarkers
of sepsis patients who were treated for N7 days in the ICU and con-
ducted a time-dependent Cox proportional hazard regression analysis
for which minimum values for each of the prior days of measurement
were reset as explanatory variables we defined as “moving minimum”
values. We chose serial moving minimum values determined on days
1, 3, 7, and 14 to evaluate whether nutritional markers including albu-
min were associated with patient prognosis.
2. Methods
2.1. Patients
This was a retrospective cohort study that enrolled 136 patients with
sepsis according to SEPSIS-2 criteria [5] who were admitted to our ICU
for N7 days during April 2011 to July 2015. We usually administer
2.3. Statistical methods
To summarize variables indicating patients' baseline demographic
and clinical characteristics, the median and interquartile range (IQR)
were used for continuous variables whereas counts and percentages
were used for categorical variables. The Kaplan-Meier product limit
estimator was used to compute the cumulative survival probability of
the patients with sepsis.
A time-dependent Cox proportional hazard regression analysis was
performed to estimate the effect of each clinical nutritional marker
(albumin, TP, T-chol and ChE) on the mortality of sepsis patients with
adjustment for baseline values of the SOFA and APACHE II scores. In
the model, each marker value was quantified as a “moving minimum”
instead of the actual value of the nutritional markers on the i th day.
The moving “k-day minimum” value on the i th day was defined as
the minimum value of the corresponding variable over k days including
day i, i-1, … i-(k-1), where k was chosen as 3, 7, and 14 days. For exam-
ple, the 3-day minimum value of serum albumin on the 6th day was the
minimum value of albumin measured on days 4, 5, and 6. Furthermore,

Fig. 1. Serial changes of serum total protein, albumin, total cholesterol, and cholinesterase over 14 days. This figure shows the data of all biomarkers for the 14 days from hospital admission
in each group of survivors and non-survivors.
 R. Takegawa et al. / Journal of Critical Care 51 (2019) 139–144 141

a nonlinear restricted-cubic spline was used to assess the nonlinear
effect of the moving minimum of each marker in the model.
To assess the effect of the change in the daily difference of each nu-
tritional marker, a cross-product term between the daily change of each
nutritional marker and the actual value of the nutritional marker from
the day before was included in the time-dependent Cox proportional
hazard regression model along with the nonlinear restricted-cubic
spline for the change in the daily difference.
All statistical inferences were made with a two-sided significance
level of 5% except for the interaction analyses. Because of the under-
powered nature of the interaction analysis, a two-sided significance
level of 20% was used with statistical inferences for interaction analyses.
All statistical analyses were conducted using the “rms” package in R
software version 3.3.2 (https://www.r-project.org/foundation/).
3. Results
3.1. Patient characteristics
The patients' demographic and clinical characteristics are shown in
Table 1. The study group comprised 89 men and 47 women with a
median age of 70 years. The main diagnosis on ICU admission included
pulmonary infection, abdominal infection, and soft-tissue infection. The
median values of the APACHE II score, APS, and SOFA score on admission
were 16, 10.5, and 4, respectively. The median length of ICU stay was
17 days. The mortality rate was 18.4%.
3.2. Serial nutritional markers and Kaplan-Meier curve
The serial changes in each of the nutritional markers over 14 days
are shown in Fig. 1. All of the markers tended to decrease in the acute
phase. Supplemental Fig 1 shows the Kaplan-Meier curve for the cumu-
lative incidence of mortality from the day of admission for sepsis.
3.3. Time-dependent Cox proportional hazard regression analysis
The results of the time-dependent Cox proportional hazard regres-
sion analysis of the effect of the serum nutritional markers on mortality
are shown in Table 2, Fig. 2-a, and Fig. 2-b, and Supplemental Fig. 2. All
of the 1-, 3-, 7-, and 14-day minimum moving values of serum albumin,
TP, T-chol, and ChE were significantly associated with mortality

Fig. 2. Logarithmic hazard ratio of serum albumin, total protein, total cholesterol, and cholinesterase. This figure shows hazard ratios of biomarkers of 1-day (a) and 3-day (b) minimum
moving values. For example, 3-day minimum moving values are covariates of minimum values of a given day and the previous 2 days. The relative hazard ratio was not linearly associated
with each moving minimum value. The grey area indicates the 95% confidence interval. Alb: albumin, T-chol: total cholesterol, TP: total protein, ChE: cholinesterase, SOFA: Sequential
Organ Failure Assessment, APACHE II: Acute Physiology and Chronic Health Evaluation II.
142 R. Takegawa et al. / Journal of Critical Care 51 (2019) 139–144
Fig. 2 (continued).
adjusted for SOFA and APACHE II scores determined on admission day
(p b .05) except for the 1-day minimum moving values of T-chol.
3.4. Time-dependent Cox proportional hazard regression analysis with
change in daily difference of nutritional markers
Fig. 3 shows the time-dependent Cox proportional hazard regression
analysis of the change in the daily difference values of the nutritional
markers. When serum albumin values were 1.0 or 2.0, the increase in
the daily difference value of albumin decreased the logarithmic relative
hazard significantly (p b .05). An increase in the serum TP value tended
to decrease the hazard ratio. When the serum cholesterol was 50, the in-
crease in the daily difference values of T-chol tended to decrease the
logarithmic relative hazard ratio. However, when the serum cholesterol
was 100 or 150, no such tendency was observed. When the value
of serum ChE was 200, the increase in the daily difference values of
T-chol tended to decrease the logarithmic relative hazard ratio, but no
such tendency was observed when the serum ChE was 100 or 150.
There were no significant interaction differences in TP, T-chol, and ChE
(p = .239, 0.701, and 0.348, respectively).
We also conducted an analysis to determine the difference in
albumin level between a given day and that on the day of admission
(Supplemental Fig. 3). Changes in albumin levels were not significantly
different, but those of the other three markers were all significantly
different (p b .05).
4. Discussion
The results showed the association between the risk of death and the
minimum value of each nutritional marker over a short term. We found
that the values and changes occurring during shorter terms strongly
reflected the risk of death especially for albumin, TP, T-chol, and Ch-E.
In the intensive care setting, serum albumin has long been a predic-
tor of poor clinical and surgical outcomes although it is also a reflection
of the acute-phase response [1]. In 83 critically ill patients, decreased
plasma albumin concentrations paralleled increased pulmonary vascu-
lar permeability irrespective of the underlying disease and fluid status.
Albumin levels in these patients had a high sensitivity and negative pre-
dictive value for elevated pulmonary vascular permeability and acute
respiratory distress syndrome [6]. As a result, the addition of hypoalbu-
minemia (b 17.5 g/L) to the American European Consensus Conference
criteria or the lung injury score increased their predictive values for
elevated pulmonary vascular permeability. In a comprehensive meta-
analysis of 90 cohort studies comprising 291,433 acutely ill patients
evaluating hypoalbuminemia as an outcome predictor by multivariate
analysis and in nine prospective controlled trials on correcting
 R. Takegawa et al. / Journal of Critical Care 51 (2019) 139–144 143

Table 2 in the serum albumin concentration significantly raised the odds of

Time-dependent Cox proportional hazard regression analysis for 3-, 7-, and mortality by 137% [7]. More recent studies including that of 5894
14-day minimum moving values.
acutely ill adult medical patients have also shown hypoalbuminemia
Biomarkers Adjusted HR (95% CI) P at admission to be an independent marker of 30-day all-cause mortality
Alb 3-day min 0.429 (0.243–0.760) 0.002 [8]. However, there is little research focused only on sepsis patients. In
Alb 7-day min 0.448 (0.258–0.777) 0.007 the present study, we found an association of the serial change of both
Alb 14-day min 0.556 (0.321–0.963) 0.046 serum albumin and TP with prognosis in the patients with sepsis. Espe-
TP 3-day min 0.402 (0.165–0.980) 0.000
cially, the serial change occurring over the shorter term had a strong ef-
TP 7-day min 0.276 (0.077–0.994) 0.001
TP 14-day min 0.352 (0.050–2.464) 0.039 fect on prognosis. The changes of the 1- or 3-day minimum moving
T-chol 3-day min 0.510 (0.318–0.819) 0.009 values of serum albumin were more strongly associated with prognosis
T-chol 7-day min 0.574 (0.355–0.928) 0.017 than those of 7 and 14 days. These results may be reasonable because
T-chol 14-day min 0.700 (0.433–1.132) 0.011 the progressive decrease in serum albumin and TP caused by decreased
Ch E 3-day min 0.170 (0.054–0.537) 0.010
Ch E 7-day min 0.367 (0.161–0.837) 0.045
synthesis in the liver has an effect on colloid osmotic pressure and the
Ch E 14-day min 0.358 (0.165–0.775) 0.019 plasma carrier of endogenous and exogenous compounds and can
dynamically change vascular permeability, which might contribute to
HR: hazard ratio, CI: confidence interval, Alb: albumin, min: minimum, TP: total protein,
T-chol: total cholesterol, Ch E: cholinesterase. death. To improve hypoalbuminemia, several albumin replacement
studies were performed such as the SAFE, CRISTAL, and ALBIOS studies,
hypoalbuminemia that included 535 acutely ill patients, the association but they did not establish improved mortality in patients with
between hypoalbuminemia and poor outcome appeared to be indepen- sepsis [9-11]. Early nutrition therapies may contribute to improving
dent of both nutritional status and inflammation. Each 10-g/L decrease hypoalbuminemia.

Fig. 3. Time-dependent Cox proportional hazard regression analysis. This figure shows time-dependent Cox proportional hazard regression analysis of changes in the daily difference
values of the nutritional markers of serum albumin, total protein, total cholesterol, and cholinesterase. The logarithm of the relative hazard ratio for the daily differences of each of the
markers is shown with three representative values. Shaded colors indicate the mean 95% confidential intervals. Alb: albumin, T-chol: total cholesterol, TP: total protein, ChE:
cholinesterase.
144 R. Takegawa et al. / Journal of Critical Care 51 (2019) 139–144
Hypocholesterolemia is crucial because of the regulatory roles of ste-
roids and signaling proteins [12]. Therefore, hypocholesterolemia may
be an indicator of the metabolic response in stress. In 638 consecutive
critically ill surgical patients, the frequency of hypocholesterolemia
(b 120 mg/dL) was 38%, and these patients had higher mortality than
those without hypocholesterolemia [13]. Yamano et al. reported that
serum cholesterol was associated with prognosis using the minimum
value during the first 14 days in patients with sepsis [14]. One mecha-
nism for this association is the ability of lipids and lipoproteins to bind
and neutralize bacterial endotoxins [15].
Serum ChE activity is reduced in liver dysfunction due to reduced
synthesis, in contrast to other serum enzymes associated with the clin-
ical assessment of liver function whose activities increase as a result of
increased release from their cellular sources following cell membrane
damage. In a study comparing 20 patients with overt liver dysfunction
with 20 with non-liver disease, serum ChE was considered to be a
marker distinguishing between overt liver disease and non-liver disease
when there might be aberrations of some liver function tests [16]. In 97
underweight outpatients with anorexia nervosa, the average ChE clearly
decreased by about 19% compared with controls comprised of 56
healthy female participants, independently of age, weight, or body
mass index. This finding was considered reflective of the effect of the
nutritional status on liver function [17]. In 465 patients with chronic
heart failure who underwent cardiopulmonary exercise testing, echo-
cardiography, and simultaneous blood examination including indicators
of nutrition, the patients with cardiac events had lower ChE levels than
those without events during the follow-up periods (p b .001) [18]. Jin
et al. reported that serum ChE within the first 24 h was an independent
risk factor for prognosis in 124 ICU patients with systemic inflammatory
response syndrome, with low serum ChE levels indicating poor progno-
sis [19]. Using minimum moving values assessed on days 1, 3, 7, and 14
after admission for sepsis, we could show an association between ChE
and mortality over the short term.
The information from this study is clinically important because we
could use progressive changes of nutritional biomarkers as prognostic
factors for mortality. In a future trial, we may consider multimodal
therapy including antibiotics therapy and nutritional support to inhibit
inflammation and the inflammatory response indicated by these
biomarkers.
This study has some limitations. First, it is a single-center, retrospec-
tive, observational study with small sample size. Second, there were
two patients with liver disease. However, their albumin values were
3.0 and 2.8 g/dL on admission, which were not extremely low values.
Further study is needed to find better nutritional therapy for patients
with hypoalbuminemia.
5. Conclusion
We found that changes in some nutritional markers, such as
albumin, TP, T-chol, and ChE, were associated to higher risk of death
in the short term in septic patients. Progressive changes in nutritional
biomarkers may be useful as prognostic markers.
Competing interests
The authors RT, KS, HO, and TS declare that they have no competing
interests. The institution of DK, AS, and SH received a consultant and sta-
tistical analysis fee from the corresponding author (RT).
Funding
This research did not receive any specific grant from funding agen-
cies in the public, commercial, or not-for-profit sectors.
The following are the supplementary data related to this article.
References
[1] Pierre S, Annika RB, Mette MB, et al. ESPEN guideline on clinical nutrition in the in-
tensive care unit. Clin Nutr 2018:1–32. https://doi.org/10.1016/j.clnu.2018.08.037.
[2] Weimann A, Braga M, Carli F, et al. ESPEN guideline: Clinical nutrition in surgery.
Clin Nutr 2017;36(3):623–50.
[3] Fanali G, di Masi A, Trezza V, Marino M, Fasano M, Ascenzi P. Human serum albumin:
from bench to bedside. Mol Asp Med 2012;33(3):209–90.
[4] Buck M, Zhang L, Halasz NA, Hunter T, Chojkier M. Nuclear export of phosphorylated
C/EBPbeta mediates the inhibition of albumin expression by TNF-alpha. EMBO J
2001;20(23):6712–23.
[5] Mitchell M, Mitchell PF, John CM, et al. 2001 SCCM/ESICM/ACCP/ATS/SIS Interna-
tional Sepsis Definitions Conference. Crit Care Med 2003;31:1250–6.
[6] Aman J, van der Heijden M, van Lingen A, et al. Plasma protein levels are markers of
pulmonary vascular permeability and degree of lung injury in critically ill patients
with or at risk for acute lung injury/acute respiratory distress syndrome. Crit Care
Med 2011;39(1):89–97.
[7] Vincent JL, Dubois MJ, Navickis RJ, Wilkes MM. Hypoalbuminemia in acute illness: is
there a rationale for intervention? A meta-analysis of cohort studies and controlled
trials. Ann Surg 2003;237(3):319–34.
[8] Jellinge ME, Henriksen DP, Hallas P, Brabrand M. Hypoalbuminemia is a strong pre-
dictor of 30-day all-cause mortality in acutely admitted medical patients: a prospec-
tive, observational, cohort study. PLoS One 2014;9(8):e105983.
[9] The SAFE. Study investigators; a comparison of albumin and saline for fluid resusci-
tation in the intensive care unit. N Engl J Med 2004;350:2247–56.
[10] Djillali A, Shidasp S, Samir J, et al. Effects of fluid resuscitation with colloids vs crys-
talloids on mortality in critically ill patients presenting with hypovolemic shock: the
CRISTAL randomized trial. JAMA 2013;310(17):1809–17.
[11] Pietro C, Gianni T, Serge M, et al. Albumin replacement in patients with severe sepsis
or septic shock. N Engl J Med 2014;370:1412–21.
[12] Porter JA, Young KE, Beachy PA. Cholesterol modification of hedgehog signaling pro-
teins in animal development. Science 1996;274(5285):255–9.
[13] Gui D, Spada PL, De Gaetano A, Pacelli F. Hypocholesterolemia and risk of death in
the critically ill surgical patient. Intensive Care Med 1996;22(8):790–4.
[14] Yamano S, Shimizu K, Ogura H, et al. Low total cholesterol and high total bilirubin
are associated with prognosis in patients with prolonged sepsis. J Crit Care 2016;
31(1):36–40.
[15] Ulevitch RJ, Johnston AR, Weinstein DB. New function for high density lipoproteins.
Isolation and characterization of a bacterial lipopolysaccharide-high density lipopro-
tein complex formed in rabbit plasma. J Clin Invest 1981;67(3):827–37.
[16] Ogunkeye OO, Roluga AI. Serum cholinesterase activity helps to distinguish between
liver disease and non-liver disease aberration in liver function tests. Pathophysiology
2006;13(2):91–3.
[17] Montagnese C, Scalfi L, Signorini A, De Filippo E, Pasanisi F, Contaldo F. Cholinester-
ase and other serum liver enzymes in underweight outpatients with eating disor-
ders. Int J Eat Disord 2007;40(8):746–50.
[18] Sato T, Yamauchi H, Suzuki S, et al. Serum cholinesterase is an important prognostic
factor in chronic heart failure. Heart Vessel 2015;30(2):204–10.
[19] Jin QH, He XJ, Li TL, Chen HH. Predictive value of serum cholinesterase for the prog-
nosis of aged patients with systemic inflammatory response syndrome. Chin Med J
2011;124(17):2692–5.