Establishment of a Model for Predicting the Recurrence of Recurrent Acute Pancreatitis in A Short Period of Time

Background Little is known about what factors affect recurrence of recurrent acute pancreatitis (RAP) in a short-term. We performed a retrospective study of patients with RAP, and used non-invasive and quantitative biomarkers to establish a predictive model, so as to provide a basis for effectively predicting and preventing recurrence of RAP. We performed a retrospective study of 261 RAP patients admitted to Dazhou Central Hospital from January 2008 through December 2018. The short-term group (n=138) and the long-term group (n=123) were compared. Eight clinical patient characteristics and past medical history were collected from admission records. The prediction model was developed by independent risk factors (cid:0) and the model to predict the effects were conrmed by the area under the ROC cure (cid:0) The study found that β2-microglobulin (β2-MG) (OR, 0.534; 95% CI, 0.329 to 0.866) and C-reactive protein (CRP) (OR, 1.01; 95% CI, 1.002 to 1.017) were the independent predictors of short-term recurrence. Based on these Independent risk factors (cid:0) a predicted model for short-term recurrence in RAP patients was constructed (cid:0) The model was L=β2-MG -0.016BCRP. The AUC for model was 0.651 and the optimal cut-off value was 0.7414, with a sensitivity of 65.9%and a specicity of 63.4%.


Abstract
Background Little is known about what factors affect recurrence of recurrent acute pancreatitis (RAP) in a short-term. We performed a retrospective study of patients with RAP, and used non-invasive and quantitative biomarkers to establish a predictive model, so as to provide a basis for effectively predicting and preventing recurrence of RAP.

Methods
We performed a retrospective study of 261 RAP patients admitted to Dazhou Central Hospital from January 2008 through December 2018. The short-term group (n=138) and the long-term group (n=123) were compared. Eight clinical patient characteristics and past medical history were collected from admission records. The prediction model was developed by independent risk factors and the model to predict the effects were con rmed by the area under the ROC cure

Results
The study found that β2-microglobulin (β2-MG) (OR, 0.534; 95% CI, 0.329 to 0.866) and C-reactive protein (CRP) (OR, 1.01; 95% CI, 1.002 to 1.017) were the independent predictors of short-term recurrence. Based on these Independent risk factors a predicted model for short-term recurrence in RAP patients was constructed The model was L=β2-MG -0.016BCRP. The AUC for model was 0.651 and the optimal cut-off value was 0.7414, with a sensitivity of 65.9%and a speci city of 63.4%.

Conclusions
A new biomarker-based model for predicting risk of short-term recurrence in patients with RAP was satisfactorily developed. The model utilizing β2-MG and CRP may assist physicians in patient education and decision-making when preventing from recurrent attacks with RAP.

Background
Recurrent acute pancreatitis(RAP)is a clinical condition characterized by experienced 2 or more episodes of documented acute pancreatitis, separated by at least 3 months [1],which usually occurs in the setting of normal morphofunctional gland with self-limited edematous changes. Clinical data from all over the world show that the recurrence rate of acute pancreatitis is 10-30% [2][3][4][5][6]. In addition to increasing the social burden and affecting the quality of life of the patient with RAP, current epidemiological and imaging data suggest that RAP has a higher risk of developing chronic pancreatitis(CP), and the more times of recurrence, the higher the risk of progression to CP [7,8]. However, most of the literatures around the world focus on the analysis of high risk factors and imaging changes of RAP [9,10] it has not been reported in the literature that there are high risk factors and predictive models related to short-term recurrence of RAP. Therefore, in this study, the clinical data of 261 patients with RAP were retrospectively analyzed to explore the clinical characteristics and risk factors of short-term recurrence of RAP, and to use of non-invasive and quantitative biomarkers to establish a predictive model, so as to provide a basis for effectively predicting and preventing of RAP recurrence.

Data collection
Eight clinical patient characteristics and past medical history, including gender, age, drinking and smoking status, gallstone status, length of hospital stay (indays), recurrence-free interval, diabetes, and hyperlipemia were collected from admission records. We obtained biochemical data at diagnosis and 12 hours after presentation.

Statistical analysis
Continuous variables are described as the means ± standard deviations for normally distributed continuous variables. Median, 25th and 75th percentiles were calculated for non-normally distributed continuous variables. Numbers and percentages (%) were used to describe categorical data. The characteristics of the study participants were analysed by an independent sample t-test or Mann-Whitney U test for continuous variables. A Chi-square test was used to analyze the association between categorical variables. Independent variables with p<0.05 in univariate analyses were included in a binary logistic regression model to select independent predictors of short-term recurrence. Results were presented as odds ratios (ORs) with 95% con dence intervals (CIs). Signi cant (p 0.05) predictors were subsequently entered into the nal multivariable prediction model. The area under the receiver operating characteristic (ROC) curve of the model to predict recurrence was calculated. The optimal cutoff value, sensitivity, and speci city were calculated on the basis of the ROC curve. All statistical analyses were performed using SPSS software, version 18.0 (SPSS Inc., Chicago, IL).

Patient characteristics
Finally, a total of 261 patients were entered, consisted of 148 males and 113 females with a median recurrence-free interval of 11 months (interquartile range [IQR], 5 m -23.5 m). All patients were divided into 2 groups according to interval of recurrence: the short-term group (n=138) was de ned as having an interval of recurrence less than the Median time, and the long-term group (n=123) greater than the Median time. Table 1 gives baseline characteristics of patient groups.
Multivariate analysis, including diabetes, hyperlipemia, achol, indays, PDW, GLU, ALT, GGT, PA, β2-MG and CRP, indicated that β2-MG and CRP were the independent predictor of short-term recurrence. (Table 2). Building independent predictive model. B values of β2-MG and CRP (p 0.05), derived from the binary logistic regression model, were constructed for the model function: L=β2-MG -0.016CRP ROC curve analysis was conducted and the area under the ROC curve (AUC) was used to assess the predictive accuracy of model function. (Figure 1) Finally, the AUC for model was 0.651 and the optimal cut-off value was 0.7414, with a sensitivity of 65.9%and a speci city of 63.4%.

Discussion
RAP still represents a challenging disease, repeated episodes of which contribute substantially to the overall disease burden by prompting repeated hospitalization, adversely impacting quality of life as well as increasing the risk of progression to chronic pancreatitis and pancreatic tumor [2,[11][12][13][14]. RAP may run a milder clinical course than an index episode [15,16], but the fatality rate of which was signi cantly higher than that of patients with rst-episode AP [17]. Few reports have studied that the more frequent the onset of RAP, the more noticeable the decrease of the total pancreas volume and tail diameter, and the higher the incidence of CP [8,18]. Therefore, the prediction of short-term recurrence of RAP is worthwhile to the early intervention and treatment of CP.
Although Some causes were a well-established risk factors for RAP [12,[18][19][20][21]. Theoretically, if the causes of acute pancreatitis existed, they could make acute pancreatitis to relapse repeatedly. In our retrospective study of patients with RAP, we found that diabetes, hyperlipemia, drinking status, ALT, GGT were involved in increased risk of RAP recurrence. Yadav et al showed that about 1/4 of patients with recurrent acute pancreatitis of known or unknown causes have underlying chronic pancreatic disease, which may invalidate treatment to remove known possible causes of acute pancreatitis [14]. Sajith et al found that clinical features and complications were the same in the idiopathic group and those with known etiology [22]. So, in the present study, we only found that β2-MG and CRP were independent risk factors for short-term recurrence of RAP.
Although challenging to prove in humans, the changes in the pancreas linked to increased AP susceptibility likely involve epigenetic changes, microanatomic changes, and/or immunological changes such as populating the parenchyma widespread with tissue histiocytes (eg, macrophages, dendritic cell) [23], excessive in ammatory reaction of the body [24], and changes of in ammatory mediators and cytokines [25].Thus, AP results in persistent changes that strongly predisposes to RAP.β2-MG is a low molecular weight protein that is synthesized by all human nucleated cells and platelets; It is indirectly involved in the regulation of the immune system response [26]. Some authors have found that serum β2-MG levels are elevated in patients with chronic pancreatitis [26][27][28],RAP is the most important risk factor for progression to end-stage CP [7,14]. Recurrent episodes of acute pancreatitis may complicate the course of chronic subclinical pancreatitis, meaning they are the clinical expression of chronic pancreatitis diagnosed in an early phase, or otherwise they may themselves induce chronic lesions as a consequence of repeated damage. Therefore, in our prediction model, the accuracy of RAP recurrence prediction is positively correlated with β2-MG, why which is it negatively related to CRP? This may be associated with pancreatic brosis after repeating recurrence of RAP. There is further emerging evidence to suggest that pancreatic brosis reduces severity in acute-on-chronic pancreatitis [29]. As such, recurrent attacks of pancreatitis may lead to increasing burden of parenchymal brosis, which in turn may be protective against the in ammatory cascade. Another speculated explanation is that protective immunological mechanisms are up-regulated in recurrent pancreatitis that not activates in the initial bout.
However, as a retrospective study predicting the recurrence of RAP, there were several limitations to the present study. Firstly, the present study was a single-center study enrolling limited number of patients.
Secondly, our data lack dynamic follow-up during hospitalization and after discharging. Thirdly, CRP level is closely associated with the time of in ammation, but the measurement of CRP in our trial is the result of patients within 12 hours after admission, but the interval between onset and measurement is uniform. It may affect the results, or it may be one reason why the speci city and sensitivity of our prediction model are low. Eventually, this study only used biochemical detection technique, which may cause bias in the results. Further investigation using a multi-center and prospective study and involving radiomics technique is necessary in order to improve the accuracy of the results.

Page 6/11
A new biomarker-based model for predicting risk of short-term recurrence in patients with RAP was satisfactorily developed. The model utilizing β2-MG and CRP may assist physicians in patient education and decision-making when preventing from recurrent attacks with RAP. Meanwhile, we still have to make more efforts to improve the speci city and sensitivity of our model. Abbreviations RAP, Recurrent acute pancreatitis; WBC, white blood cell; NEU, neutrophil; HGB, hemoglobin; PLT, platelet; PDW, platelet distribution width; GLU, glucose; ALT,alanine aminotransferase;AST,glutamic oxaloacetic transaminase;ALP,alkaline phosphatase;GGT,γ-glutamyl transferase;PA,prealbumin;CREA,creatinine;UA, uric acid;β2-MG, β2-microglobulin;CYSC, cystatin C;CHO, cholesterol;TG, triglyceride;CRP, C-reactive protein; B, coe cient of regression; SE, standard error; df, free degree; OR, odds ratio; CI, con dence interval; ROC, operator characteristic; AUC, area under curve.

Competing interests
The authors declare that they have no competing interests.

Funding
None.

Authors' contributions
The study was designed by Bo Yang and Yijun Yu. Peiyun Li, Honglan Liu and Xuesong Bai collected clinical datas. Bo Yang carried out data analysis and wrote the paper.All authors read and approved the nal manuscript.   Figure 1 The ROC curve analyses of the predictive accuracy of model function.