Most patients with acute pancreatitis (AP) present a mild course while 10%–20% of them develop severe disease with significant mortality [
It was noticed that severe acute pancreatitis can alter serum lipid levels [
On the other hand, these simple parameter variables demonstrate low sensitivity despite high negative predictive values for prediction of SAP [
Therefore, the primary aim of the study was to develop a multivariable model for prediction of SAP. The secondary one was to assess the ability of serum creatinine at 24 hours after admission and HDL-C as predictors of SAP in a large sample study.
A total of 647 patients suffering from acute pancreatitis that attended First Affiliated Hospital of Wenzhou Medical University between January 2013 and December 2015 were enrolled in the study. Exclusion criteria included [
The following information was collected for each patient: age, gender, Body Mass Index (BMI), etiology, hematocrit, High-Density Lipoprotein Cholesterol (HDL-C) at admission. Blood Urea Nitrogen (BUN) and serum creatinine (Scr) were registered at the time of admission and 24 hrs after hospitalization.
This study protocol was approved by the Ethic Committee of the First Affiliated Hospital of Wenzhou Medical University and it was performed according to the principles expressed in the Declaration of Helsinki. Informed consent was obtained by the subjects.
The diagnosis of AP requires two of the following three features in the revised Atlanta criteria [
The primary endpoint was to develop a logistic regression equation to predict severe acute pancreatitis. The secondary one was to assess the ability of serum creatinine at 24 hours after admission and validate HDL-C predictors of SAP at initial admission entering the hospital.
A Shapiro-Wilk test was used to evaluate whether the continuous data showed a normal distribution. According to its results, continuous values were expressed by mean ± SD or median and Interquartile Range (IQR) and compared using one-way analysis of variance or the Kruskal-Wallis nonparametric test. Categorical values were described by count and proportions and compared by the
Linear trend of categorical and continuous variables was tested by a Royston extension of the Cochran-Armitage test [
All the variables found to be different between patients with and without severe acute pancreatitis on univariate analysis were included as eligible factors in a forward-conditional stepwise logistic regression analysis. For this analysis, the conditional probabilities for stepwise entry and removal of a factor were 0.05 and 0.10, respectively [
The area under the receiver operating characteristic (ROC) curve, that is, AUC, was used to evaluate the performance of predictions. A variable with an AUC above 0.7 was considered useful, while an AUC between 0.8 and 0.9 indicated excellent diagnostic accuracy [
As described by Maksimow et al. [
Differences were considered to be statistically significant if the two-tailed
The median age of the 647 patients included in the study was 47 (IQR 37–63), of which 406 (62.8%) were male. Biliary disease was the most common cause of the AP (272/647, 42.0%). The median BISAP score at the time of hospital admission was 1. There were 491 (75.9%), 98 (15.2%), and 58 (8.9%) with mild, moderate/severe, and severe acute pancreatitis, respectively. Among the 58 patients who developed SAP, multiple organ failure was noted in 27 (46.6%) of them. Respiratory failure (84.5%) was the most frequent manifestation. The median length of the hospital stay was 10 days (IQR 7–14 days), with 15.5 days (IQR 10–28 days) for SAP patients. Ten patients (1.55%) died during hospitalization.
As shown in Table
Univariate analysis of predictive factors of acute pancreatitis in 647 patients.
Characteristic | Mild AP | Moderate AP | Severe AP |
|
---|---|---|---|---|
( |
( |
( |
||
Median age, years (IQR) | 47 (37–62) | 47.5 (40–63) | 51 (38–66) | 0.165 |
Male sex, |
309 (62.9) | 66 (67.4) | 31 (53.4) | 0.219 |
BMI | 23.4 (20.9–26.1) | 24.6 (21.5–26.0) | 24.2 (22.1–26.6) | 0.048 |
Etiology | 0.002 | |||
Biliary, |
222 (45.2) | 32 (32.7) | 18 (31.0) | |
Alcohol, |
63 (12.8) | 23 (23.5) | 4 (6.9) | |
Hypertriglyceridemia, |
22 (4.5) | 7 (7.1) | 7 (12.1) | |
Idiopathic, |
184 (37.5) | 36 (36.7) | 29 (50.0) | |
Laboratory findings | ||||
Hematocrit | 0.42 (0.38–0.45) | 0.43 (0.39–0.47) | 0.44 (0.40–0.47) | 0.001 |
HDL-C (mg/dl) | 41.3 (32.0–51.0) | 36.7 (25.5–51.7) | 22.4 (17.8–38.2) | <0.001 |
BUN, mg/dl (IQR) | 13.2 (10.4–16.5) | 12.2 (9.8–16.5) | 19.9 (15.1–31.9) | <0.001 |
BUN (24 h), mg/dl (IQR) | 12.9 (9.5–16.8) | 12.2 (9.5–19.0) | 26.0 (17.1–34.5) | <0.001 |
Creatinine, mg/dl (IQR) | 0.72 (0.61–0.86) | 0.72 (0.61–0.88) | 0.92 (0.66–1.82) | <0.001 |
Creatinine (24 h), mg/dl (IQR) | 0.72 (0.59–0.86) | 0.68 (0.55–0.87) | 1.04 (0.74–2.34) | <0.001 |
BISAP score | 1 (0-1) | 1 (1-2) | 2 (1–3) | <0.001 |
IQR = Interquartile Range;
A logistic regression function
LR model calibration plot. Patients were ranked by their predicted probability and divided into 10 equal groups. The red bars represent the mean predicted probabilities for each of the 10 groups and blue bars represent the observed probabilities with severe acute pancreatitis in each of these same groups. LR model = logistic regression model.
As shown in Figure
ROC curves for various predictors for severe acute pancreatitis. The AUCs for BMI at admission, hematocrit at admission, HDL-C at admission, BUN at admission, BUN after 24 hrs of admission, Scr at admission, Scr after 24 hrs of admission, BISAP score, and LR model for the prediction of SAP were 0.56 ± 0.04, 0.60 ± 0.04, 0.76 ± 0.04, 0.75 ± 0.04, 0.79 ± 0.04, 0.67 ± 0.05, 0.76 ± 0.04, 0.82 ± 0.03, and 0.84 ± 0.03, respectively. The ideal AUC was 1.00. The reference line represents AUC of 0.50, based on chance alone. ROC curve = receiver operating characteristic curve; AUC = area under the receiver operating characteristic curve; BMI = Body Mass Index; HDL-C = High-Density Lipoprotein Cholesterol; BUN = Blood Urea Nitrogen; Scr = serum creatinine; BISAP = Bedside Index for Severity in Acute pancreatitis; LR model = logistic regression model.
Based on the ROC curve analysis, the optimum cut-off values of HDL-C, BUN at admission, BUN at 24 hours, Scr after 24 hrs, BISAP score, and LR model were 22.4 (mg/dl), 22.7 (mg/dl), 21.8 (mg/dl), 1.02 (mg/dl), 2, and −1.86, respectively (Table
Diagnostic values of various predictors of severe acute pancreatitis.
Variable | Cut-off value | Sensitivity | Specificity | PPV | NPV |
---|---|---|---|---|---|
HDL-C | 22.4 mg/dl | 51.7 | 90.5 | 34.9 | 95 |
BUN | 22.7 mg/dl | 46.6 | 90.7 | 32.9 | 94.5 |
BUN (24 hrs) | 21.8 mg/dl | 56.9 | 90.2 | 36.3 | 95.5 |
Scr (24 hrs) | 1.02 mg/dl | 51.7 | 90.2 | 34.1 | 95 |
BISAP score | 2 | 65.5 | 83.4 | 27.9 | 96.1 |
LR model | −1.86 | 62.7 | 93.2 | 47.4 | 96.1 |
HDL-C = High-Density Lipoprotein Cholesterol; BUN = Blood Urea Nitrogen; Scr = Serum creatinine; BISAP = Bedside Index for Severity in Acute pancreatitis; LR model = logistic regression model; PPV = positive predictive value; NPV = negative predictive value.
Using a cut-off value of −1.86 and the prevalence of organ failure in acute pancreatitis (8.9% in this study) as the pretest probability, the Fagan plot (Figure
Fagan plot for LR model for prediction of severe acute pancreatitis. LR model = logistic regression model.
The results of the present study demonstrated the following: (i) LR model and BISAP score were excellent predictors of SAP, with an AUC of more than 0.8 (Figure
A rise in the BUN level reflects the disease status of initial intravascular volume depletion and prerenal azotemia in AP [
Like BUN, Scr is also a marker of renal function. A rise in Scr reflects the disease states of initial hypovolemia and renal dysfunction in SAP and it represents an important factor for the assessment of severity [
The mechanism of progression from a mild to a severe AP is induced by the release of proinflammatory cytokines (such as TNF-
As expected, the LR model that consists of the above three parameters markedly improved sensitivity. With a cut-off of −1.86, the LR model achieved an acceptable sensitivity of 62.7%, excellent specificity of 93.2%, PPV of 47.4%, and NPV of 96.1% (Table
The strength points of this study include such a large sample size able to give the study a strong statistical power. Both patients in ICU and in general ward were enrolled in this study, thus reducing selection bias. According to our opinion, this is the first study in literature fully assessing Scr at 24 hours and validating HDL-C as a predictor of SAP, respectively, as well as determining the best cut-off value of HDL-C for prediction of severe acute pancreatitis. The developed LR model showed a good predictive performance in terms of discrimination and calibration. The LR model with a high AUC may be helpful to guide triage and to manage patient with AP. The limitation of our study was that BUN and Scr were measured at 24 hours after admission, which may influence the early application of our LR model in a subgroup of AP patients who rapidly developed progressive multiple organ failure in the first few days following the onset of acute pancreatitis (also recognized as fulminated acute pancreatitis in previous literature) [
In conclusion, we have confirmed that HDL-C at admission and Scr at 24 hours may predict development of SAP. The LR model consisting of HDL-C at admission and BUN and Scr at 24 hours takes on a high diagnostic accuracy of prediction of development of SAP. It is an additional tool to stratify patients at risk of SAP and its application on admission may improve clinical care and strategies of management in acute pancreatitis.
This study protocol was approved by the Ethic Committee of the First Affiliated Hospital of Wenzhou Medical University. This study was performed according to the principles expressed in the Declaration of Helsinki.
Informed consent was obtained from the subjects.
The authors declare that they have no potential conflicts of interest.
Wandong Hong joined in the design of the study and carried out the studies; Wandong Hong, Suhan Lin, Wujun Geng, and Chunfang Xu participated in data collection. Wandong Hong conducted data analysis and drafted the manuscript. Maddalena Zippi, Simon Stock, Vincent Zimmer, and Mengtao Zhou helped to finalize the manuscript. All the authors read and approved the manuscript. Wandong Hong and Wujun Geng contributed equally to this work.