Hypoalbuminemia, An Independent Risk Factor for Severity and Mortality Affects One Third of Patients in Acute Pancreatitis: Multicenter Prospective International Cohort Analysis

Introduction The incidence and medical costs of acute pancreatitis (AP) are on the rise, and severe cases still have a 30% mortality rate. We aimed to evaluate hypoalbuminemia as a risk factor and the prognostic value of human serum albumin in AP. Methods Data of 2461 patients were extracted from the international, prospective, multicenter AP registry of the Hungarian Pancreatic Study Group. Data of patients with albumin measurement in the rst 48 hours (n=1149) and anytime during hospitalization (n=1272) was analyzed. Multivariate binary logistic regression and Receiver Operator Characteristic curve analysis were used. Results The prevalence of hypoalbuminemia (<35g/L) was 19% on-admission and 35.7% during hospitalization. Hypoalbuminemia dose-dependently increased the risk of severity, mortality, local complications, and organ failure and is associated with longer hospital stay. The predictive value of hypoalbuminemia on-admission was poor for severity and mortality. Severe hypoalbuminemia (<25 g/L) was an independent risk factor for severity (OR: 48.761; CI:25.276-98.908) and mortality (OR:16.83; CI: 8.32-35.13). Albumin loss during AP was strongly associated with severity (p<0.001) and mortality (p=0.002). Conclusion Hypoalbuminemia is an independent risk factor of severity and mortality in AP, and it shows a dose-dependent relationship with local complications, organ failure, and length of stay. collection and respiratory failure (p < 0.001 and p = 0.051). The rate of pancreatic necrosis, pseudocyst, or heart failure did not differ signicantly between the groups.


Introduction
Acute pancreatitis is a common gastroenterological disorder, with rising incidence and high medical costs. The commonly used revised Atlanta Classi cation distinguishes between mild, moderate, and severe disease by the development and duration of organ failure 1 . As the mortality rate can reach 30% in severe cases, identifying risk factors and potential therapeutic targets is of utmost importance.
Human serum albumin is the most abundant protein in human serum, with a very diverse role. Although this hypothesis was contradicted by recent data, declining albumin levels during in ammation for long prompted physicians to underestimate its contribution to maintaining homeostasis during in ammation.
However, it plays a pivotal role in maintaining the plasma redox state 2 , and its scavenging activity is likely to in uence vascular resistance through the regulation of nitric oxide levels 3 . Furthermore, low albumin levels result in dilution and increased drug clearance, ultimately causing sub-optimal treatment 4 .
Small retrospective cohort studies showed that hypoalbuminemia is an independent risk factor for severe AP and in-hospital mortality in adults and children 5,6 . Serum albumin was reported to be associated with persistent organ failure and prolonged hospital stay 7 . However, whether albumin is only a marker or there is a cause-effect relationship between hypoalbuminemia and disease severity and mortality should be further evaluated.
While comprehensive analyses are missing on AP patients with hypoalbuminemia and albumin loss in AP, we aimed to evaluate (1) on-admission and in-hospital hypoalbuminemia as a risk factor in AP, (2) the prognostic potential of human serum albumin, (3) whether there is a dose-dependent relationship between albumin level and disease outcomes and (4) the association of albumin loss with severity and mortality.
We found evidence that AP patients with < 25 g/L serum albumin anytime during hospitalization have a 16.8 times higher risk of death and 48.8 times higher risk of severe AP than patients with normal albumin levels. We also pointed out that albumin loss during AP is associated with severity and mortality. These data highlight the unmet need for randomized controlled trials focusing on albumin replacement.
Older age, lower body mass index, abdominal guarding, on physical examination and non-biliary etiology are associated with on admission hypoalbuminemia Hypalbuminemia was associated with older age (average 59.7 ± 18.0 and 56.0 ± 16.1 years; p = 0.005, Supplementary Fig. S3). Males were overrepresented in the analyzed cohort (57%) and all subgroups (Fig.  S3). Although biliary etiology was the most frequent in all subgroups, signi cantly fewer patients had biliary etiology (34.4% versus 42.2%; p = 0.042) in the low albumin group, and a tendency of more alcoholic episodes (24.3% and 19%; p = 0.096) was seen ( Supplementary Fig. S3).
Considering the signs and symptoms, fewer hypoalbuminemia patients presented with abdominal pain (94.9% and 99.2%; p < 0.001) and more with abdominal guarding (27.2% and 19.9%; p = 0.023) (Fig. S5). General signs, such as duration and intensity of abdominal pain, abdominal tenderness, nausea, and vomiting, did not signi cantly differ. Hypoalbuminemia was associated with a dose-dependent increase in heart rate and a decrease in systolic and diastolic blood pressures on admission ( Supplementary Fig.  S5).
The ful llment of diagnostic criteria differed signi cantly (p < 0.001) among the low and normal albumin groups on-admission. Low albumin patients were less likely to present with pancreatic enzyme elevation, abdominal pain, and characteristic imaging ndings at the same time (42.7% versus 58.4%) (Supplementary Table S2).
On-admission hypoalbuminemia is dose-dependently associated with elevated CRP and PCT levels in AP The low albumin group had signi cantly lower serum amylase (p < 0.001) and lipase (p = 0.002) levels on admission. A dose-dependent C-reactive protein (CRP) (p < 0.001) and procalcitonin (PCT) (p < 0.001) increase was observed in the lower albumin groups. White blood cell count (WBC) (p = 0.017) levels were also signi cantly elevated in the low albumin group (Fig. S6-7). Concerning laboratory markers of renal function, hypoalbuminemia patients had signi cantly higher blood urea nitrogen (BUN) (p = 0.002) and creatinine (p = 0.002) levels and lower estimated glomerular ltration rate (eGFR) (p < 0.001) (Supplementary Fig. S8-9). Liver enzymes and total bilirubin levels did not differ between the low and normal albumin groups, but hypoalbuminemia was associated with higher direct bilirubin levels (p = 0.005) and a higher international normalized ratio (INR) (p < 0.001) (Supplementary Fig. S10-13).
Hematological parameters, lipids, ions, and glucose levels are shown in Supplementary Figures S14-17.
On-admission hypoalbuminemia is dose-dependently associated with complications, severity, and mortality in AP Signi cantly more patients developed local complications, and organ failure in the low albumin group (p = 0.016 and p < 0.001, respectively) ( Fig. 1-2). Lower albumin levels correlated with a higher rate of peripancreatic uid collection and respiratory failure (p < 0.001 and p = 0.051). The rate of pancreatic necrosis, pseudocyst, or heart failure did not differ signi cantly between the groups.
All types of local complications were signi cantly more frequent in the low albumin group. A dosedependent increase was seen in the rate of local complications and peripancreatic uid collection in both cohorts and in pancreatic necrosis and pseudocyst in the lowest measured albumin cohort. P < 0.05 is considered signi cant. Patients with albumin levels < 35 g/L were included in the low albumin group (groups 5-7).
Signi cantly more patients developed organ failure in the low albumin group in both cohorts. A dosedependent increase was seen in the case of all analyses in the lowest measured albumin cohort. Heart failure was dose-dependently increased in the on-admission cohort as well. P < 0.005 is considered signi cant.
On-admission hypoalbuminemia is an independent risk factor of severity and mortality, with an odds ratio up to 5.3 for mortality in acute pancreatitis Age, hypertriglyceridemia-induced (with or without concomitant alcoholic etiology), and idiopathic AP were independently associated with mortality. Severe on-admission hypoalbuminemia proved to be an independent risk factor for mortality with an OR of 3.782 (CI: 1.313-9.462) in group 6 (< 30 g/L) and an OR of 5.256 (CI: 1.389-16.112) in group 7 (< 25 g/L) ( Table 1.) Albumin levels were examined with a 35 g/L cut-off in a separate analysis, which found an independent association between hypoalbuminemia and mortality (OR: 2.070; CI: 1.021-4.033; Supplementary Table S1). Age, hypertriglyceridemia-induced AP, and, among the multifactorial etiologies the combination of hypertriglyceridemia and alcohol were independent risk factors of disease severity. On-admission albumin levels < 25 g/L were independently associated with severe AP (OR: 3.620; CI: 1.128-9.978; Table 1). On-admission albumin levels alone have poor predictive value in AP On-admission albumin levels have an AUC of 0.615 (sensitivity: 57.6%, speci city: 61.1%) for the severity with a cut-off at 39.3 g/L (Fig. 4). The AUC for mortality was 0.660 (sensitivity: 72.1%, speci city: 53.7%) with a cut-off at 37.0 g/L.
These data prompt that albumin plays a crucial role in the pathophysiology and clinical outcome of AP, however cannot be used as a single biomarker for predicting severity and mortality. Next, we wanted to understand whether albumin loss during the course of AP has any association with the outcome of the disease; therefore, we regrouped our patients based on the lowest measured albumin levels.
One out of three patients suffer from hypalbuminemia in AP during hospitalization, which dosedependently correlates with disease severity and mortality in AP The proportion of patients with hypoalbuminemia anytime during hospitalization was 35.7% (454 patients). A signi cant, dose-dependent increase was seen in the low albumin groups (group 5-7) compared to the normal albumin group regarding the rate of all examined systemic and local complications ( Fig. 1-2). The lowest measured albumin levels throughout hospitalization (n = 1272) were signi cantly and dose-dependently associated with severity (p < 0.001), mortality (p < 0.001), length of stay (p < 0.001), and maximum CRP values (p < 0.001) (Fig. 3).
Moderate and severe AP and mortality is associated with signi cantly lower albumin levels and greater albumin loss Albumin loss was analyzed using data from patients with at least two albumin measurements (n = 335; Supplementary Fig. S18). Compared to mild cases, patients with moderate and severe AP showed a greater decrease in albumin levels (medians 5.4 vs. 9 and 15.25 g/L; p < 0.001 for both comparisons). The comparison of delta albumin between the moderate and severe groups also yielded signi cant results (p = 0.003). Patients who died also lost signi cantly more albumin during hospitalization (medians 6.7 vs. 15.75 g/L; p = 0.002). The median time to the lowest albumin levels from admission was 4 days (IQR: 3-7 days).
AP patients with less than 25 g/L serum albumin have a 16.8 times higher risk of death, and 48.8 times higher risk of severe AP compared to patients with normal albumin levels Age is an independent risk factor for severe AP and mortality, whereas hypertriglyceridemia-induced and idiopathic AP and the combination of alcoholic and biliary causes are independently associated with mortality (

Discussion
To date, this is the most comprehensive evaluation of AP patients with hypoalbuminemia, using the largest, prospectively collected, high-quality dataset.
We found that almost one-fth of patients had hypoalbuminemia on admission (19%), and a further 25% developed hypoalbuminemia during hospitalization, meaning that every third patient was affected.
In our analysis, hypoalbuminemia under 25 g/L anytime during hospitalization was independently associated with a more than 47 times higher chance for severe AP and a more than 16 times higher chance for mortality.
Our ndings are in line with results regarding hypoalbuminemia in other diseases. Hypoalbuminemia was a prominent risk factor in community-acquired bloodstream infection with severe sepsis and septic shock 8 . A retrospective analysis from more than 20.000 emergency medical patients' data from Ireland revealed that hypoalbuminemia is independently associated with 30-day in-hospital mortality, with a nonlinear relationship between mortality and on-admission albumin levels 9 . Moreover, in a secondary analysis of a prospective cohort, AP patients with multiorgan failure (MOF; n = 18) demonstrated a sharper decline in serum albumin (P < 0.001) compared to non-MOF patients (n = 39) 10 .
We did not only prove that hypoalbuminemia is a risk factor but showed the dose-dependent association between low albumin levels and severity, mortality, number of patients with any local complications, number of patients developing organ failure, and maximum CRP levels in both analyses (on-admission and lowest measured albumin levels).
These associations can be explained by the numerous physiological functions of human serum albumin.
For long, albumin was considered a negative acute-phase protein, with decreasing production giving way to in ammatory cytokines in in ammation 11 . Serum albumin levels undoubtedly decrease in in ammatory states, which may be due to a shorter half-life and a larger interstitial pool which causes the dilution of albumin 12-14 . Capillary leak consequential to in ammatory processes plays a role in the decline of serum albumin, but it is argued that the escape of albumin to the tissues may be bene cial because of its antioxidant and scavenging activity 15 . However, a more than two times higher production rate was observed in critically ill ICU patients; this increased production can still not balance the higher demand. This can be considered as a relative synthetic insu ciency of hepatic function 16 .
In our analysis, albumin loss was signi cantly associated with severity and mortality. However, only 51.7% of patients had albumin measurements at least once, and 13.6% at least twice during their hospitalization in the HPSG database. This highlights how neglected albumin measurements are in AP.
From the clinician's point of view, the decline of serum albumin levels -regardless of on-admission albumin levels -signals clinical worsening and may help identify high-risk AP patients. However, clinicians mostly miss the opportunity to pre-emptively and frequently measure serum albumin, delaying timely intervention.
To date, no clinical trial examined therapeutic albumin administration in AP. As we know, albumin is similarly associated with outcomes in sepsis and septic shock; randomized controlled trials on this eld could be a starting block 15,17 . The controversial results of studies and meta-analyses on this eld may be explained by heterogeneous patient populations and the time sensitivity of this treatment 18 .
To exploit the potential in therapeutic albumin administration in AP, further, more detailed clinical studies are needed to identify the patient subpopulations bene ting the most from this therapeutical option.

Conclusion
Hypoalbuminemia is remarkably common in AP (seen in 19% of patients on-admission and 35.7% during hospitalization) and an independent risk factor of severity, mortality. Importantly, albumin loss during hospitalization was also associated with severity and mortality, suggesting that routine monitoring of serum albumin is recommended, and albumin administration should be examined as a therapeutic intervention in AP.
Implications for research: Clinical trials assessing the potential bene t of albumin replacement in AP are needed.
Implication for practice: 1) Albumin levels should be measured for all AP patients, 2) Albumin levels should be controlled at least in those patients whose condition is worsening during AP, 3) Albumin administration should be considered at least in those patients with severe hypoalbuminemia (< 25 g/L).

Study design and de nitions
This analysis of an international, prospective, multicenter cohort was done using data from the Acute Pancreatitis Registry of the Hungarian Pancreatic Study Group (HPSG) 19 . Patient data were collected from registry establishment until 31 December 2019 on electronic case report forms and validated using a four-tiered data validation protocol. Contributing centers are shown in the supplementary material (Table and Fig

Participants
To answer a post-hoc clinical research question, analyses were performed on patients' data with albumin measurement anytime during hospitalization (lowest measured albumin cohort, n = 1272) and in the rst 48 hours of hospitalization (on-admission albumin cohort, n = 1149). The cut-off value between the low and normal albumin group was 35 g/L in both cases, based on the commonly used lower normal value. Subjects were further divided into 7 subgroups (group 1 to 7) using the lowest (n = 1272) or rst measured (n = 1149) albumin values.
In the analyses of albumin change, selected patients (n = 335) with at least two albumin measurements were included. Delta albumin was calculated as the difference between the rst and lowest measured albumin levels.

Statistical analysis
Descriptive statistics are presented as median with 25% and 75% percentiles (IQR) or mean with standard deviation (SD) for continuous variables and as numbers and proportions for categorical variables.
Chi-squared test or Fisher's exact test were used for the assessment of the relationship between categorical variables. Mann-Whitney U test or Kruskal-Wallis test followed by Dunnett's post hoc test was used to evaluate differences between groups in case of continuous variables.
Multivariate binary logistic regression analysis was performed to identify the risk factors independently associated with severe disease and mortality. Odds ratios (OR) with 95% con dence intervals (CI) were calculated.
Receiver Operator Characteristic (ROC) curve and Area Under the Curve (AUC) with 95% CI was used to identify the ability of albumin levels predicting mortality or severity of AP (AUC between 0.5-0.6 was considered as fail, between 0.6-0.7 as poor, between 0.7-0.8 as fair, between 0.8-0.9 as good and above 0.9 as excellent). Best cut-offs were calculated by using the Youden index 32 .

Representativity
The main characteristics of the analyzed cohorts are in accordance with literature data. However, they differed signi cantly from the entire cohort (n = 2461) in terms of severity, length of stay, and mortality ( Fig. S2).

Reporting
We report our results following The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement, using the provided checklist 35 .

Data availability
The full dataset is available upon reasonable request.

Strengths And Limitations
We conducted the most extensive, most comprehensive cohort study on the role of hypoalbuminemia in acute pancreatitis to date. We analyzed high-quality data from a prospective, international, multicentric registry. We identi ed hypoalbuminemia as an independent risk factor in AP, present in at least every third patient. We also found a dose-dependent relationship between albumin levels and main outcomes, which was previously not described.
Among the limitations, we must mention the arbitrary classi cation of albumin levels (except for the lownormal cut-off), the missing data on albumin levels and albumin administration during the hospital stay, and the limited number of albumin measurements during the hospital stay, which could introduce bias. Our analyzed cohorts differed from the total cohort in some aspects, which may signal performance bias, as albumin measurements are more frequently ordered for patients with expected hypoalbuminemia.
Declarations Figure 1 Association between albumin level and local complications de ned by the Revised Atlanta Criteria in acute pancreatitis All types of local complications were signi cantly more frequent in the low albumin group. A dose-dependent increase was seen in the rate of local complications and peripancreatic uid collection in both cohorts and in pancreatic necrosis and pseudocyst in the lowest measured albumin cohort. P<0.05 is considered signi cant. Patients with albumin levels <35 g/L were included in the low albumin group (groups 5-7).

Figure 2
Association between albumin level and organ failures, de ned by the Revised Atlanta Criteria in acute pancreatitis Signi cantly more patients developed organ failure in the low albumin group in both cohorts.
A dose-dependent increase was seen in the case of all analyses in the lowest measured albumin cohort.
Heart failure was dose-dependently increased in the on-admission cohort as well. P<0.005 is considered signi cant.

Figure 3
Association between albumin level and disease severity, mortality, length of stay and maximum Creactive protein level in acute pancreatitis Severity, mortality, length of stay and maximum C-reactive protein levels were signi cantly and dose-dependently associated with hypoalbuminemia in both cohorts.
P<0.05 is considered signi cant.