Serum Albumin as an Early Predictor of Severity in Patients with Acute Pancreatitis

Abstract

Acute pancreatitis (AP) is one of the gastrointestinal pathologies that most frequently requires hospital admission; about half of all deaths occur within the first two weeks and are caused by multi-organ failure. Predicting the degree of severity of AP before 48 h is a challenge. Background/Objectives: Having an early marker, before 48 h after admission, could be useful to avoid or diagnose early complications such as organ failure (OF). A few sentences could place the question addressed in a broader context and highlight the purpose of the study. Methods: A retrospective study conducted in a third-level hospital, during the period from August 2019 to June 2021. Patients aged >18 years, with a diagnosis of PA, who had a complete clinical history and complete biochemical and imaging data were included. The scores of the APACHE II, BISAP, revised Atlanta classification, and modified Marshall scales were recorded. Results: Of the 103 patients included, 60% were women, the mean age was 47.76 years, and the hospital stay was 8 days (IQR 6–12); the most frequent etiology was biliary in 46 (44.7%) patients; the most frequent BMI was overweight with 34 (33%) patients; and 38 (36.9%) patients had a systemic inflammatory response at admission. Hypoalbuminemia was observed in 34 (33%) of the 103 patients at admission; of these, 42 (40.8%) had an APACHE II score > 8 points, 17 (16.3%) a BISAP score > 2, 57 (54.8%) patients were classified as moderate AP according to the revised Atlanta classification, and 54 patients had a score according to the modified Marshall score > 2. A statistically significant difference in the development of death was observed between patients with hypoalbuminemia versus those with normal serum albumin levels. Conclusions: In this study, we show the usefulness of hipoalbuminemia (<3.5 g/dL) at hospital admission in patients with AP, as a severity and mortality indicator. With the results obtained, we conclude that low albumin levels are a good predictor of severity and are useful for establishing timely treatment and close follow-up.

1. Introduction

Acute pancreatitis (AP) is one of the gastrointestinal pathologies that most frequently require hospital admission; it is an inflammatory process of the pancreatic gland, due to the early activation of digestive enzymes secreted by the exocrine pancreas within the acinar cells, causing their self-digestion as well as the stimulation of macrophages leading to the production of proinflammatory cytokines, leading to damage in various tissues and organs [1]. Reported mortality is approximately 2% and increases between the fifth and sixth decades of life [2]; in Mexico, previous studies report a mortality of 0.5% of all deaths in the country [3,4]. Predicting the degree of severity of AP before 48 h is a challenge, and the purpose of having an early marker, before 48 h after admission, could be useful in preventing and minimizing organ dysfunction and complications. Patients with AP may develop OF, which may be temporary if such failure can be reversed no later than 48 h after admission; and persistent OF is defined as OF lasting more than 48 h [5,6].

Predictive Factors of Severity in AP

Determining the severity of AP as early as possible is a crucial step in optimizing treatment and preventing or minimizing complications or organ dysfunction. The following have been described as severity factors at hospital admission: body mass index (BMI), age, elevated hematocrit and blood urea nitrogen (BUN), as well as the presence of pleural effusion [7]. The persistence of systemic inflammatory response syndrome (SIRS) favors the development of multiple organ failure, and with it a mortality of 36% to 50% [5,6,8,9].

Brown et al. identified that a hematocrit higher than 44% at the time of hospital admission and its persistence at 24 h are predictors of severity, with a negative predictive value at 24 h of 96% for necrotizing pancreatitis and 97% for OF [10]. Wu bu et al. confirmed that the presence of an elevated BUN level on admission and failure to decrease within 24 h of starting treatment are independent factors for mortality in PA [11].

The liver is responsible for synthesizing albumin; blood albumin levels are decreased in inflammatory processes, and it has been shown that these levels are related to the severity of the inflammatory state, as well as to prognosis and mortality [12]. Likewise, albumin is a protein that remains in the bloodstream and contributes fundamentally to the maintenance of oncotic pressure, so that its exit from the intravascular territory to the interstitial tissue involves the transport of substances from cell catabolism; it has an important antioxidant capacity, constituting the most powerful extracellular mechanism capable of binding free radicals, and once oxidized it is rapidly eliminated from the circulation [13]. Due to the relationship between inflammation and malnutrition, low albumin levels or hypoalbuminemia (<3.5 g/dL) are expected to be found in elderly hospitalized patients and in patients with chronic liver disease, sepsis, heart failure, and cancer [14].

In the study by Kaplan et al., the C-reactive protein/albumin ratio was evaluated to assess prognosis and mortality in AP. The conclusion was that this ratio was higher in patients who died than in survivors; it also correlated positively with the Ranson score, the Atlanta classification, and length of hospital stay [12]. Hong et al. observed that serum albumin levels 24 h after admission in patients with AP correlated with clinical outcomes; as serum albumin levels decreased, the risk of persistent OF and the severity of the AP event significantly increased; it was also associated with longer hospital stay and higher mortality [14].

Li et al. studied the usefulness of serum albumin and PTa results upon admission for AP, finding them useful for assessing severity during an AP event, becoming a good predictor of severity (AUC 0.82, IC 95% 0.78–0.87) [15].

Eighty percent of all AP cases will have a mild, self-limited form of the disease with a short hospital stay; however, 5–10% will have an intense systemic inflammatory response and intense local damage leading to necrosis, with the potential risk of developing pancreatic insufficiency and systemic failure with a fatal outcome [16,17,18].

In the present study, we evaluated whether low serum albumin levels (hypoalbuminemia), at admission and 48 h after hospitalization, were associated with higher scores on the APACHE II, BISAP, modified Marshall, and revised Atlanta classification for AP and with SIRS, with the intention of predicting severity in a timely manner in these patients.

2. Materials and Methods

This is a cross-sectional, descriptive, observational, and retrospective study conducted at the Hospital de Especialidades “Dr. Bernardo Sepúlveda Gutiérrez”, Centro Medico Nacional Siglo XXI, IMSS, Gastroenterology and General Surgery services in the period from January 2015 to June 2021. The Research Committee under number R-2021-2601-057 authorized the protocol. Patients with a diagnosis of AP who met the inclusion criteria were evaluated; clinical information was obtained from clinical records from January 2015 to June 2021. Inclusion criteria: Patients aged >18 years, who presented with a diagnosis of AP on admission to the emergency department and who were hospitalized by the Gastroenterology or General Surgery service, and who had biochemical and imaging studies for the purposes of this study. Non-inclusion criteria: Patients who do not meet the diagnostic criteria for AP; patients who have no record in terms of clinical history; patients referred to this unit already with a diagnosis of AP and more than 24 h of in-hospital treatment in another unit; patients with a diagnosis of chronic pancreatitis, necrotizing pancreatitis, or pancreatic cancer; patients with previous pancreatic surgery; patients with hypoalbuminemia secondary to diseases other than acute pancreatitis; patients who do not have serum albumin measurement at admission; patients who do not have clinical and biochemical measurements to integrate the presence of systemic inflammatory response syndrome or in whom it is not possible to calculate prognostic scales.

Statistic Analysis

The prevalence of AP from January 2015 to June 2021 was determined; it was calculated from the filter of the statistical census of the Gastroenterology and General Surgery services. According to the normality parameters, the summary measure was determined. Baseline characteristics in the different groups were compared according to serum albumin level by Mann–Whitney U test for categorical or ordinal qualitative variables and by X² for dichotomous qualitative variables. The quantitative data were recorded at admission and at 48 h. Scores of the APACHE II, BISAP, modified Atlanta, modified Marshall, and SIRS scales were calculated 24 and 48 h after hospital admission and compared between the group with normal serum albumin level (>3.5 g/dL) and hypoalbuminemia (<3.5 g/dL).

The disease was classified according to Atlanta in three groups: mild, moderately severe, and severe, while each quantitative biochemical parameter, each severity score (BISAP, APACHE II, modified Marshall), and number of days of hospital stay were included and subsequently compared between both groups (normal albumin and hypoalbuminemia); for this purpose, the Mann–Whitney U and X² statistical tests were used.

The cut-off point for albumin was determined by the ROC curve, and this cut-off point was compared with each of the prognostic scales (Figure 1). Logistic regression was used to determine the risk of developing severe AP comparing each of the prognostic scales, as well as the number of days of hospital stay with the presence or absence of hypoalbuminemia. SPSS program version 21.0 was used, and a p < 0.05 with its respective 95% confidence intervals was considered statistically significant.

3. Results

Of a total of 454 patients, 103 met the inclusion criteria.

Table 1. Demographic characteristics.

	N = 103
Gender
Men	43 (41.7)
Women	60 (58.3)
Etiology
Biliary	46 (44.7)
Hypertriglyceridemia	11 (10.7)
Alcohol	5 (4.9)
Post-ERCP	15 (14.6)
Idiopathic	21 (20.4)
Tumor	4 (3.9)
Drug	1 (1.0)
Body mass index
Low	5 (4.9)
Normal	19 (18.4)
Overweight	34 (33)
Obesity	31 (30)
Diabetes mellitus
No	80 (77.7)
Yes	23 (22.3)
Acute kidney injury
No	74 (71.8)
Yes	29 (28.2)
SIRS
No	65 (63.1)
Yes	38 (36.9)
APACHE score
<8	58 (56.3)
>8	42 (40.8)
BISAP score
≤2	86 (82.7)
>2	17 (16.3)
MARSHALL score
<2	46 (44.2)
>2	57 (54.8)
Age (years)	47.76 (±15.38) **
Number of days of hospital stay	8.0 (6–12) *
Kirby	301.37 (±75.81) **

Non-ND * ND **.

and

Table 2. Serum levels in the general population.

Albumin
<3.5 g/dL	34 (33)
>3.5 g/dL	69 (67)
Albumin (g/dL)	4.0 (3.3–4.40) *
Leukocytes (103)	13.52 (±6.28) **
Creatinine (mg/dL)	0.94 (0.75–1.35) *
BUN (mg/dL)	14 (11.2–28.0) *
Hematocrit (%)	43 (33.9–48.3) *
Amilasa (U/L)	786 (316–1521) *
Lipasa (U/L)	845 (458–2447) *

Non-ND * ND **.

describe the baseline characteristics and serum levels of the population, finding that 58.3 (60%) patients were women; there was a mean age of 47.76 (±15.38) years and hospital stay of 8.0 (6–12) days; the most frequent etiology was biliary in 46 (44.7%) patients, followed by idiopathic cause in 21 (20.4%) patients; the most frequent BMI was overweight, with 34 (33%) patients; and 38 (36.9%) patients presented data of systemic inflammatory response at admission.

Subsequently, the population was divided into two groups according to the maneuver under investigation: 69 (65%) patients with normal albumin and 34 (35%) patients with hypoalbuminemia; X² and Mann–Whitney U analyses were performed as appropriate with the baseline demographic data to determine if there was a statistically significant difference according to baseline albumin values, which revealed a significant difference in patients with hypoalbuminemia and the following variables: the presence of acute kidney injury (p = 0.001), an APACHE II score > 2 points (p = 0.002), and the revision of the Atlanta classification with moderately severe grade (p = 0.001) and BUN > 25% (p = 0.001) (

Table 3. Serum albumin level at hospital admission.

	Albúmina Normal N: 69 n (%)	Hipoalbuminemia N: 34 n (%)	p
Gender			0.934 *
Man	29 (67.4)	14 (32.6)
Woman	40 (66.7)	20 (33.3)
Etiology			0.058 *
Biliary	32 (69.6)	14 (30.4)
Hypertriglyceridemia	4 (36.4)	7 (63.6)
Alcohol	3 (60)	2 (40)
Post-CPRE	14 (93.3)	1 (6.7)
Idiopathic	14 (66.7)	7 (33.3)
Tumor	2 (50)	2 (50)
Drugs	0 (0)	1 (100)
Age			0.131 *
<55 years	50 (73.5)	20 (58.8)
>55 years	18 (26.5)	14 (41.2)
Body mass index			0.59 *
Low weight	2 (40)	3 (60)
Normal	13 (68.4)	6 (31.6)
Overweight	24 (70.6)	10 (29.4)
Obesity	20 (64.5)	11 (35.5)
Acute kidney injury			0.001 *
No	57 (77)	17 (23)
Yes	12 (41.4)	17 (58.6)
APACHE score			0.002 *
<8	47 (81)	11 (19)
>8	22 (52.4)	20 (47.6)
SIRS			0.053 *
No	48 (73.8)	17 (26.2)
Yes	21 (55.3)	17 (44.7)
ATLANTA (entry)			0.000 *
Mild	40 (87)	6 (13)
Moderately severe	29 (50.9)	28 (49.1)
Hematocrit < 44%	68 (57.9)	33 (36.7)	0.001 **
Creatinine	69 (47.14)	34 (61.85)	0.019 **
BUN < 25%	69 (47.18)	34 (61.7)	0.020 **
BUN			0.001 *
<25%	60 (53.6)	20 (26.4)
>25%	9 (15.4)	14 (7.6)
BISAP score (>2)	69 (44.75)	34 (66.7)	0.000 **

* X². ** U Mann–Whitney.

).

A logistic regression analysis was performed to predict whether the presence of hypoalbuminemia at hospital admission and 48 h later was associated with a higher score on prognostic scales for organ failure (

Table 4. a. Prognostic scales and hypoalbuminemia at hospital admission. b. Logistic regression. Prognostic scales at 48 h vs. hipoalbuminemia.

a.
	p	OR	IC 95%
APACHE ≥ 8	0.000	5.19	2.14–12.58
BISAP ≥ 2	0.001	6.98	2.21–22.05
Atlanta M-S	0.000	6.37	2.31–17.55
Modified Marshall score ≥ 2	0.000	4.41	1.74–11.19
SIRS ≥ 2	0.045	2.28	0.98–5.32
b.
	p	OR	IC 95%
APACHE score > 8	0.000	5.40	2.19–13.32
BISAP score > 2	0.006	5.85	1.65–20.72
Atlanta (FOP)	0.000	6.94	2.7–17.78
Modified Marshall score > 2	0.000	5.43	2.18–13.53
SIRS	0.055	2.43	0.98–6.02

a,b). Furthermore, it was observed that hypoalbuminemia conferred an OR = 1.66 for having a hospital stay longer than 7 days.

A value of 3.5 g/dL was determined as the cut-off point for albumin, with an AUC of 0.788, p 0.000, 95% CI: 0.665–0.910, a sensitivity of 74%, and a specificity of 71% (Figure 1).

When comparing the four scales with the albumin levels at hospital admission, we observed that they do not present good sensitivity; however, the specificity and PPV at hospital admission are good in each of these scales. According to the results in

Table 5. Se, S, PPV, LR, and PTP of each scores vs. albumin levels.

	APACHE Score		BISAP Score		Modified Atlanta		Modified Marshall
	Admission	48 h	Admission	48 h	Admission	48 h	Admission	48 h
Se	55	58	70	69	49	54	48	52
S	81	79	74	72	87	85	83	83
PPV	65	58	93	26	82	76	76	73
NPV	74	79	35	94	58	68	58	65
LR	2.89	2.82	2.76	2.49	3.77	3.72	2.77	3.13
PTP	65	59	35	26	82	76	76	74

Se: sensitivity; S: specificity; PPV: positive predictive value; LR: likelihood ratio, and PTP: post-test probability.

, presenting albumin levels > 3.5 g/dL mean it is unlikely that a patient admitted with a diagnosis of AP will develop persistent OF; likewise, each of the scales has a good PPV (BISAP score, revised Atlanta classification, and modified Marshall score) so that presenting hypoalbuminemia on hospital admission means that these patients are at greater risk of developing persistent organ failure.

Conclusion. In the present study, we demonstrate the usefulness of low serum albumin levels (<3.5 g/dL) as a poor prognostic factor in patients admitted with a diagnosis of AP. The hypoalbuminemia figures were compared with the APACHE II score, BISAP, revised Atlanta classification, and modified Marshall scale; we obtained the scores for each of them on admission and 48 h after hospital admission. With the results obtained, we conclude that low albumin levels are a good predictor of severity, which is useful for establishing timely treatment and close follow-up.

4. Discussion

A total of 103 patients were included, in which we found that the condition was more common in women (57%), with an average age of 47.7 years; the most frequently reported etiology was biliary (44.2%), followed by idiopathic (21%); 34 patients were overweight; and on admission, 34 patients had hypoalbuminemia (<3.5 g/dL). We determined whether they had acute kidney injury on admission, and of the 103 patients, 29 (28.2) had it. Prognostic scales at hospital admission were determined: 42 (40.8%) patients had an APACHE score > 8, 38 (36.9%) had SIRS, 17 (16.3%) patients had BISAP > 2 points, 46 (44.2%) patients were classified as mild Atlanta and 57 (54.8%) as moderate Atlanta, and 54 (51.9%) of the patients had a modified Marshall scale score > 2 points.

There are scores that allow one to predict the severity of the disease, with the disadvantage that they require at least 48 h to increase prognostic accuracy [3]. The APACHE II score is used to predict severity during the first 48 h and can distinguish between moderate and severe pancreatitis; it also predicts mortality [19]. The determination of this score at hospital admission has a sensitivity ranging from 34% to 70% and a specificity of 76% to 98%; at 48 h, the sensitivity remains at 50%, with a specificity of 90 to 100%; the disadvantage of this score is that we need 48 h to reach maximum specificity, with poor sensitivity in the first 48 h [20].

The BISAP score has the advantage of being easier to calculate, with only five variables; a score higher than three is associated with a 7 to 12 times greater probability of developing OF [21]; although the simplicity of this score stands out from others, it has not been shown to be more accurate than other scores [22].

Some studies have attempted to compare multiple scores to predict severity, in particular the study by Kumar et al. [23] where 50 patients were included and the optimal cut-off points and areas under the curve (AUC) of the CTSI, BISAP, APACHE II, and Ranson scores were compared; the AUC of the modified CTSI score was the highest of the four compared as a marker of severity in AP, in particular for pancreatic necrosis (0.993), the need for intensive care unit admission (0.919), and organ failure (0.893). In this study, the APACHE II score was comparable to the modified CTSI score in terms of the severity of acute pancreatitis. Similarly, the modified CTSI was superior to the BISAP and Ranson scores in predicting severity [23,24,25,26].

In our study, the comparison of hypoalbuminemia with the different prognostic scales showed statistical significance in predicting severity; sensitivity, specificity, positive predictive value (PPV), and likelihood ratio (LR) were determined, and it was observed that all of them have good specificity as well as PPV, which means that patients with AP and a serum albumin level within normal ranges at the time of hospital admission have between a 74% and 87% chance of not developing severity, and between a 76% and 82% chance of developing severity if hypoalbuminemia is detected at the time of admission.

Two groups were performed (normal serum albumin ≥ 3.5 g/dL and hypoalbuminemia < 3.5), finding that women more frequently presented hypoalbuminemia; however, there was no significant difference between men and women (p 0.934), nor was there a difference between the multiple etiologies whether or not they presented altered albumin levels (p 0.058). Neither age (p 0.131) nor body mass index (p 0.59) presented a statistically significant difference in the development or not of hypoalbuminemia; however, there was a significant difference (p 0.001) between those patients who were admitted with acute kidney injury and had hypoalbuminemia. Patients with hypoalbuminemia on admission had higher APACHE II scores (p 0.002) and, according to the modified Atlanta classification, were admitted with moderate activity (p 0.000) and a BISAP scale score > 2 points (p 0.000). Patients with normal albumin at admission were more likely (p 0.005) to have a hematocrit level below 44%; patients with BUN > 25% were more likely to have hypoalbuminemia at hospital admission (p 0.001).

We compared the severity of acute pancreatitis according to the revised Atlanta classification (mild, moderately severe, or severe) versus albumin levels on admission and observed that patients with hypoalbuminemia progressed toward the development of persistent organ failure (p 0.000), and mortality was higher compared to patients with normal albumin levels on hospital admission (p 0.038). To date, there is no scale that allows us to predict severity and mortality before 48 h in these patients; we have used a combination of these scales to try to predict clinical outcomes.

There are studies where the usefulness of the CPR/albumin index has been evaluated as a prognostic [13] in acute pancreatitis, and where it has been used to evaluate the incidence of acute kidney injury in AP [27]; in 2023, it was used to evaluate this index to predict severity and as a prognostic in these patients [28]. In the first study, 192 patients with AP were included, and it was determined that the cut-off point for the CPR/albumin index > 16.28 conferred higher mortality versus an index < 16.28. Something that is striking in this article is that 50 patients who reported with mild Atlanta had an index > 16.28, and a mortality of 19% was reported, a high percentage considering that most of them had a mild degree of the disease (66%); in our study it was 6.7%, which is similar to that reported worldwide. In the study of the year 2023, which included 284 patients, men predominated, unlike in our study and others where the gender that presented greater affection for this pathology was the female gender; something relevant to mention is that patients were included with a time of evolution equal to or greater than 48 h, the time in which the patient had or had not developed transient or persistent organ failure. Another point is that 35 patients had severe acute pancreatitis, but only 29 were in the group of those who developed organ failure, if by definition cataloging them as severe already made them carriers of organ failure [28]. In the 2024 study, 1514 patients were included, of whom 257 (16.9%) developed acute kidney injury, which increases mortality [27].

During AP, pancreatic lipase releases FFA, also called nonesterified fatty acids (NEFAs), from circulating triglycerides as well as from injured adipocytes. In ex vivo studies, no association between FFA levels and the development of necrosis or systemic complications has been demonstrated [29]; in the study by Paye et al., a slight increase in FFA concentrations was observed, but these remained within physiological ranges, so that systemic toxicity can hardly be attributed to this slight increase [29].

The study by Sztefko K et al. analyzed FFA levels in patients with AP in relation to the severity of the disease, including 36 patients with edematous AP and 29 with necrotizing AP (16 of them without complications and 13 with complications). In all patients, there was an increase in total serum FFA levels; this increase was accompanied by a decrease in albumin levels, and the concentration of albumin was inversely proportional to the severity of the disease. In this study, hypoalbuminemia and the high concentration of AG was observed in the group of patients with necrotizing AP and complications, due to the higher percentage of non-ionic cytotoxic AG bound to lipoproteins; however, the highest mortality was reported in these patients who had undergone surgical treatment, but the article does not explain the indication for surgery, how many procedures were performed, and how long after admission they underwent surgery—important points to consider, since it has been demonstrated that they increase morbidity and mortality [30].

Chung BH et al. performed an in vitro study, in which they observed that the decrease in serum albumin levels and the increase in the concentration of fatty acids condition their binding to plasma lipoproteins [31]. Approximately 50% of the FFA that bind to lipoproteins do so in a nonionic form, which confers hydrophobic properties that allow them to damage the cell membrane. Under normal conditions, FFA bind to albumin; this study showed that, when rapid and extensive lipolysis occurs, albumin binds to a small portion (14–35%) of the FFA generated by lipolysis; this was also noted by Bird DA et al., and as a consequence most of the FFA bind to lipoproteins generating lipotoxicity [32].

To date, no study has been carried out to evaluate the usefulness of serum albumin levels in the prediction of severity, or how it compares with prognostic scales. The weaknesses of this study are that it is a retrospective study; we had a few more than 300 patients in the database, but due to the inclusion criteria only 103 could be included since we did not have all the necessary parameters for the analysis.