Early Predictors of Acute Pancreatitis Related In-Hospital Mortality : How Practical Are They ?

Background/Purpose: Early assessment of the severity of acute pancreatitis (AP) is a highly challenge for a physicians’ practice to improve the management and decrease the mortality. We aimed to determine early prognostic factors for AP related in-hospital mortality. Methods: Upon hospital admission, predictors of AP related in-hospital mortality were prospectively assessed using regression analysis over 129 consecutive AP patients. Predictive abilities of these prognostic factors were compared using the area under receiver operating characteristic curve (AUC). Results: AP related in-hospital mortality was 10.9%. Red cell distribution (RDW), serum creatinine, glucose and albumin were associated with AP mortality. RDW had the highest AUC followed by serum creatinine and albumin (AUC: 914, 95% CI: 0.797 0.975; 0.797, 95% CI: 0.695 0.878; 0.798, 95% CI: 0.677 0.865 respectively). The cut-off with the best ability to predict in-hospital mortality was 14.2 for RDW. By coupling RDW and serum creatinine, AUC was improved to 0.940, 95% CI: 0.839 0.986. Conclusion: RDW, serum creatinine, albumin, and glucose even with borderline level changes may predict AP related in-hospital mortality, where, RDW has the highest prognostic accuracy. Coupling RDW and serum creatinine model significantly improves their predictive accuracy that may aid in further improvement of the quality of care of AP patients.

varied from mild and self-limiting acute pancreatitis (MAP) in approximately 80% that resolves without serious complications to severe form that carries major morbidity as necrosis or organ failure and increased mortality [1].
The overall mortality of AP is about 10% -15% and can reach up to 42% in severe AP [2] [3] [4].Because of nearly 50% of deaths occur early within the first week, early diagnosis with accurate assessment of the disease severity is a highly challenge in practice [5] [6].Death from multi-organ failure may occur early within the first 10 days after hospitalization due to massive inflammatory responses or late at the end of the second week from the disease onset that related to infected pancreatic necrosis [5] [6].Therefore, monitoring AP patients in the first 24 hours of hospital admission is of the great importance that gives a rational for studies to identify those at risk of multi-organ failure to institute an early intervention to decrease mortality.Several prognostic scoring systems have been proposed for that issue; however, it may take longer time for observation i.e. after 48 hours as Ranson's criteria [7], or its use is complex as The Acute Physiology And Chronic Health Evaluation II (APACHE II) scoring system [8].Whereas, laboratory indexes such as IL-6, CRP, serum lipase and amylase, although, easy feasibility, but their results have been inconsistent and unrelated to the disease severity [9].Therefore, the presence of sensitive, easy available and inexpensive laboratory markers for predicting the early AP mortality is recommended.
The aim of this study was to assess the clinical and the laboratory data of AP patients upon hospital admission to investigate the early predictors of AP related in-hospital mortality.

Study Design
This was a prospective observational study carried out at Assiut University Hospital (AUH), Egypt, from January 2016 to December 2017.The study was approved by the Ethics Committee of AUH and was conducted in accordance with the previsions of the Declaration of Helsinki.An informed consent was obtained from all the participants before enrollment.

Study Population
This study consecutively included 129 adult patients with AP admitted to Tropical Medicine and Gastroenterology and Internal Medicine Departments, Al-Rajhi Liver Center, Assiut University Hospital, Egypt.AP diagnosis was based on at least 2 of the following Atlanta criteria: 1) typical clinical symptoms; abdominal pain consistent with AP; 2) elevated serum amylase and/or lipase levels greater than 3 times the upper normal limit; 3) characteristic radiological findings by abdominal computerized tomography or ultrasonography.
According to the revised Atlanta classification [6], acute pancreatitis (AP) was categorized into mild acute pancreatitis (MAP with absence of organ failure and Open Journal of Gastroenterology local/systemic complications.), moderately severe acute pancreatitis (MSAP with transient organ failure/organ failure and/or local or systemic complications that resolved within 48 hours), and severe acute pancreatitis (SAP with persistent single or multiple organ failure >48 hours duration).For the purpose of analysis, the MSAP cases with the SAP cases were grouped together [10].
The etiology of AP was either 1) biliary: when imaging [as ultrasonography or computed tomography (CT)], or endoscopic retrograde cholangiopancreatography (ERCP) revealed gallstones in the gallbladder or in the common bile duct.2) alcohol-induced was diagnosed when alcohol intake before the onset of the symptoms could be established.3) Other: when attributed to ERCP, hyperlipedemia, trauma, or drugs.4) Idiopathic or unknown in other cases.Patients with the evidence of chronic pancreatitis, anemia, malignancy, known kidney or liver failure, severe infection within the last month, or organ transplantation were excluded.

Methods
At the time of hospital admission, a thorough medical history and physical examination were taken; for example demographic data, etiology of pancreatitis, organ failure, and severity of AP were noted.Blood samples were also collected for laboratory investigations, and included: complete blood count including white blood cells (WBC), hemoglobin (Hb), platelets (PLT) and red cell distribution width (RDW), liver function tests, kidney function tests including serum creatinine and blood urea nitrogen (BUN), serum sodium, serum calcium, serum potassium, serum glucose (only for non-diabetic patients), C-reactive protein (CRP) and lactate dehydrogenase (LDH), and arterial blood gas analysis.The primary end point was in-hospital mortality.

Statistical Analysis
All statistical analyses were carried out using SPSS for Windows version 16 (SPSS, Chicago, IL, USA) and the MedCalc program.Quantitative data are expressed as mean ± standard deviation or median and the interquartile range and were compared using Student's t or Mann-Whitney U-tests for normally or abnormally distributed data, respectively.Categorical variables were expressed as percentage and compared using chi-squared (χ 2 ) or Fisher's exact probability test.All significant factors on univariate analysis were considered for inclusion in multiple regression analysis to predict AP-related in-hospital mortality.The receiver operating characteristic (ROC) curves were plotted to measure and compare the performance of different parameters to determine death risk and to select the best cut-off point at which sensitivity, specificity, positive (PPV) and negative (NPV) predictive value, positive and negative likelihood ratio (+LR, −LR) were calculated.The Kaplan-Meier method was used to estimate the overall survival rates.All tests were two-tailed and statistical significance was assessed at <0.05.

Characteristics of the Studied Patients
Total 129 patients with AP (70 males and 59 females; mean age 53.6 ± 10.7) were enrolled in the present study.Based on the Atlanta classification, 72 patients (55.8%) had MAP, and MSAP and SAP were detected in 57 (44.2%).The etiology of AP was biliary in 89 patients (69%), idiopathic in 25 patients (19.4%) and other causes in 15 patients (11.6%).None of patients were alcoholic.Fourteen patients died during the hospital stay (10.9%) from: septic and toxic shock (3 cases), multiple organ dysfunction syndrome (6 cases), disseminated intravascular coagulation (1 case) and acute renal insufficiency (4 cases).Out of 115 survivors, 25 patients had local pancreatic complications when they were discharged from the hospital, including 12 patients had pancreatic pseudocyst, 11 patients developed necrotizing pancreatitis, and 2 with pancreatic abscess.The mean hospital stay was 10.6 ± 5.9 days.Demographic, clinical, and laboratory characteristics of survivors and non-survivors were summarized in Table 1.All the non-survivors had SAP where the levels of serum bilirubin, urea nitrogen, creatinine, blood glucose, and RDW were significantly increased, while the levels of albumin, calcium, and hemoglobin were significantly decreased (P < 0.05).

Risk Factors Analysis for AP Related In-Hospital Mortality
To identify the factors that could independently predict AP related in-hospital mortality, multiple regression analysis was applied using the significant variables from the univariate analysis (P < 0.05) (Table 1).It was found that increased RDW, serum glucose and creatinine and decreased serum albumin were independent predictors of AP mortality (P < 0.05) (Table 2).Moreover, using multiple regression analysis with forward LR method to construct a model with higher accuracy for prediction of AP related in-hospital mortality, a combined use of serum creatinine and RDW offered an accuracy of 92.3% (Table 3).

Diagnostic Performance of Parameters for Prediction of AP Related In-Hospital Mortality
We examined the effectiveness of RDW, serum glucose, creatinine and albumin in prediction of AP related in-hospital mortality using ROC analysis.It was found that the area under curve (AUC) values of these variables were statistically significant to predict AP mortality (Figure 1).RDW had the highest AUC (0.914; 95% CI 0.797 -0.975), followed by serum creatinine, albumin and glucose (AUC = 0.797, 0.789 and 0.746 respectively).Moreover, at a cut-off value of RDW >14.2%, the highest sensitivity (78%), specificity (96%), PPV (70.3%), and NPV (97.3%) were achieved for prediction of AP related in-hospital mortalityin approximately 77.8% of the cases (Table 4).
Further, we examined the effectiveness of RDW-serum creatinine model in prediction of AP related in-hospital mortality using ROC analysis where AUC was significantly improved (0.940, 95% CI 0.839 -0.986, P < 0.001) (Figure 2).Open Journal of Gastroenterology

Determination of the Survival Analysis
Kaplan-Meier-estimated survival curves were generated for patients who fell above and below the cut-off values identified by the means of the ROC curves for RDW and serum creatinine constituting this model to predict AP-related in-hospital mortality (Figure 3).These cut-offs clearly differentiated between patients with different survival times; patients who had scores higher than the cut-off value had a significantly shortened survival period compared with patients who had lower scores (long rank, P = 0.013; Breslow, P = 0.01 and Tarone-Ware, P = 0.006).

Discussion
This prospective study demonstrated the early prognostic factors in AP patients and their diagnostic performance to predict mortality.Elaborating reliable  factors and identification of high-risk hospitalized AP patients remain a challenge of increasing suspicions for early decisions of proper management and organ support.In this study, the AP related in-hospital mortality rate of 10.9% was within the range of the previous estimates (10% -15%) [3] [4].However, this lower percentage may be attributed to the early and proper management of gallstones as an etiology of most of AP cases.
We found that all the non-survivors had SAP that was in agreement with earlier studies, where they reported that SAP can lead to a higher risk of systemic complications, pancreatic necrosis, prolonged hospital stay and increased mortality rates of up to 50% [11] [12].found that AP non-survivors had increased levels of RDW and creatinine but decreased levels of albumin compared to survivors.Talamini et al. [14], identified increased serum glucose more than 250 mg/dl and serum creatinine more than 2mg/dl as risk factors for death in AP patients.Thus, RDW, glucose and creatinine may have a risk while albumin may act a protective role in AP patients.
In the present study, the diagnostic accuracy of RDW (AUC = 0.914) was significantly higher than the other studied risk factors with 78% sensitivity, 96% specificity, 70.3% PPV and 97.3% NPV at cut-off >14.2%.This was comparable to the findings of Şenol et al. [15], who reported RDW having AUC of 0.817 with 47.6% sensitivity, 96.3% specificity, 79.9% PPV and 87.6% NPV at cut off >14.8 in predicting fatal outcome in AP patients.
The pathophysiologic mechanisms for the association between the increased RDW levels and AP mortality may be explained as the strong association of some inflammatory cytokines which inhibit erythrocyte maturation and allow larger reticulocytes to enter the circulation [16].It may be related to inflammation that influences the bone marrow function and iron metabolism and modifies RBC membrane glycoproteins and ion channels, contributing to the change of RBC morphology [17].In addition, high oxidative stress can also lead to elevated RDW by decreasing red blood cell survival and increasing release of large premature RBCs into the circulation [17].
In consistence with previous studies [18] [19], elevated serum creatinine level at admission was significantly higher in AP patients who died during hospitalization in this study.Additionally, at a cut-off >123 μmol/l (1.4 mg/dl) of serum Open Journal of Gastroenterology creatinine, AUC was 0.782 with 77.8% sensitivity, 81.1% specificity and 96.8% NPV in predicting AP related in-hospital mortality.Renal impairment was an independent predictor for in-hospital mortality in patients with AP, and measurement of creatinine on the first day proved to be a good predictor of mortality [19] [20].
Increased creatinine levels in AP may reflect renal dysfunction because of severe intravascular hypovolemia that may cause reduction in the kidney blood flow [21].
Previous studies reported different cut off values of serum creatinine levels measured at admission time and after 48 hours; Avinash et al. [21], found that serum creatinine ≥133 μmol/l (1.5 mg/dl) on admission was an indicator of the progress towards acute renal failure (ARF).Whereas, Muddana et al. [18], reported that serum creatinine value >159 μmol/l (1.8 mg/dl) at 48 hours indicated a higher possibility of the occurrence of pancreatic necrosis (PN), which was a major contributing factor to morbidity and mortality of AP.
On the other hand, high NPV of serum creatinine (96.8%) in this study agreed with Lankisch et al. [22], who reported that the clinical utility of determining serum creatinine at admission indicates that normal values signify that necrotizing pancreatitis is unlikely, and contrast-enhanced CT is not indicated unless complications occur.Acute renal failure in AP is caused by the release of vasoactive compounds, enzymes and cytokines from the necrotic pancreatic tissue into the circulation.In addition, activated intravascular coagulation and infection may contribute to the development of ARF in these patients [21].
Serum glucose level upon hospital admission was significantly higher among AP non-survivors in this cohort.This was in agreement with earlier studies that revealed that serum glucose on admission had been considered as a satisfactory predictor of mortality in AP [23] [24].Our findings revealed that at a cutoff value >8.8 mmol/l (158.4 mg/dl); it had an AUC of 0.746, 66.7% sensitivity, 90.5% specificity, 46.1% PPV and 95.7% NPV in predicting AP-related in-hospital mortality.
Several previous studies [7] [25] demonstrated that at admission, blood glucose level >11.11 mmol/l (200 mg/dl) was added to the early grave Ranson prognostic criteria, and blood glucose >13.88 mmol/l (>250 mg/dl) was identified to be a strong predictor of mortality in AP patients [26].
This higher blood glucose levels in the AP non-survivors may be explained to the related probability of damage of the endocrine pancreas, as determined by the extent of the pancreatic necrosis.Similarly to serum creatinine, high NPV of serum glucose (95.7%) indicates that its normal values on admission usually signify that necrotizing pancreatitis is unlikely.These findings were matched with previous studies that reported that a blood glucose concentration of 6.9 mmol/l on admission had a high negative predictive value (92%) for pancreatic necrosis and also can serve as a predictor for severity [23] [24].
Serum albumin level on admission was significantly lower in deceased AP pa-Open Journal of Gastroenterology tients in the current results where serum albumin <31 g/d was a good predictor for AP-related in-hospital mortality.This matched with earlier studies that reported that the mortality significantly increased when the serum albumin level was lower than 25 -30 g/l [2] [15] [27].Albumin is a negative acute phase reactant and its level in the blood decreases with inflammation severity disease prognosis and mortality [28].Decreased serum albumin levels at the early stage of AP may be attributed to its effusion through the damaged vessel wall into the pancreas and its surrounding tissue that induced by Elastase secreted from pancreas and neutrophils, in addition to the reduction of serum albumin synthesis as a result of damage in liver function in AP [29].
Based on the data in this study, we found that a coupling model of serum creatinine and RDW had a superior accuracy (92.3%) for prediction of AP related in-hospital mortality than each parameter alone (AUC = 0.940, 95% CI 0.839 -0.986, P < 0.001).In addition, patients with higher values than RDW and serum creatinine cut off values had short survival rate as shown by Kaplan-Meier survival analysis.
Thus, this new model may achieve early (within few hours of admission), simple and easily applicable tool to predict adverse outcomes in AP patients before using more sophisticated and expensive techniques.One reason for this superior predictive ability of this coupling model is the inclusion RDW and serum creatinine which are well-known predictors of mortality in AP as shown inprevious studies [13] [15] [19] [20].This model has yet to be validated prospectively.As this study was a single-center study, further multicenter studies on larger scales are needed to help guide early resuscitation and treatment strategies in acute pancreatitis.

Conclusion
In conclusion, RDW, serum creatinine, albumin, and glucose even with borderline level changes can predict AP related in-hospital mortality, where, RDW has the highest prognostic accuracy.Coupling RDW and serum creatinine significantly improves their predictive accuracy for AP related in-hospital mortality that may aid in quality of care improvement and in the reduction of short-term mortality.Also, this model may offer a strategy to stratify high-risk patients on hospital admission who may need intensive treatment even without previous episodes of AP, which may require further studies.

Figure 1 .
Figure 1.Area under the receiver operating characteristic curve (AUC) of RDW, serum creatinine, albumin and glucose to predict AP-related in-hospital mortality.RDW had the highest AUC in predicting mortality (AUC = 0.914) followed by serum creatinine, albumin and glucose (AUC = 0.797, 0.789 and 0.746 respectively).AP: Acute pancreatitis; RDW: Red cell distribution width.

Table 1 .
Demographic and clinical characteristics of the study patients with acute pancreatitis (survivors and non-survivors).

Table 2 .
Multiple regression analysis of risk factors affected acute pancreatitis related in-hospital mortality in the studied sample.

Table 3 .
Diagnostic models for prediction of acute pancreatitis related in-hospital mortality.

Table 4 .
Diagnostic accuracy of prognostic parameters to predict acute pancreatitis related in-hospital mortality with the best predictive cut-offs.