| INTRODUCTIONPancreatic acute necrotic collections(ANC)is one of the most important local complications of acute pancreatitis(AP).With the development of treatment strategy,minimally invasive step-up approach has been well acknowledged by experts around the world and is recommended as the standard management procedure for pancreatic necrosis.While there still remains controversies about the timing for intervention of pancreatic necrosis,hence,in the current study,we aim to investigate the effects of intervention timing on the clinical outcomes of patients with ANC,hoping to shed some light on the personalized intervention timing for AP patients with varying disease severities.PART1Early versus Delayed Intervention in Necrotizing Acute Pancreatitis Complicated by Persistent Organ FailureBACKGROUND: Current guidelines recommend delayed invasive intervention for pancreatic necrosis to 4 or more weeks after disease onset to allow necrotic collections to become “walled-off”.However,for patients with clinical deterioration,especially those with persistent organ failure(POF),it is controversial whether this delayed approach should always be adopted.AIM: In this study,we aimed to assess the impact of early(<4 weeks)or delayed(≥4 weeks)intervention on clinical outcomes in a group of necrotizing acute pancreatitis(NAP)patients complicated by POF.METHODS: All NAP patients with POF admitted to our hospital undergoing at least one drainage procedure from 2013 to 2017 were screened for potential inclusion.They were divided into two groups based on the timing of initial intervention(early group and delayed group).The baseline data,organ failure prior to intervention,and main clinical outcomes were collected and compared between groups.The primary outcome was in-hospital mortality,and secondary outcomes included new-onset organ failure after intervention,length of ICU and hospital stay,total hospitalization cost,and major complications.RESULTS: Overall,131 NAP patients with POF were included for analysis.Among them,100(76.3%)patients were assigned in the early group and 31(23.7%)in the delayed group.Suspected infection was the most common indication for intervention in both of the two groups.As for organ failure prior to intervention,the incidence of renal failure,cardiovascular failure as well as multiple organ failure showed a tendency to be higher in the early group but not statistically different(0.05<P<0.1).For clinical outcomes,there was no significant difference with respect to mortality between the two groups.In addition,the incidence of new-onset organ failure after intervention,major complications,length of ICU and hospital stay were comparable between the early and delayed groups.CONCLUSION: In our study,early intervention did not worsen the clinical outcomes in NAP patients complicated by POF,despite that patients intervened within 4 weeks tended to be more severely ill.Early intervention should be considered when there is a strong indication like suspected infection of pancreatic necrosis,especially in patients with POF.PART2The prediction of minimally invasive intervention risk by early-stage organ function evaluation based on machine learning modelBACKGROUND: Acute necrotizing pancreatitis(ANP)is a common and complicated disease,and during the disease course,organ failure is shown to be one of the most important determinants for patients’ outcomes,especially,persistent organ failure(POF),which may cause substantial morbidity and even mortality to AP patients.Furthermore,the concomitant organ failure also plays a great role in the optimization for intervention of pancreatic necrosis,for instance,the timing and indication selection.In recent decades,with the bursting of artificial intelligence and medical big data,machine learning model has been widely studied in various research fields.Hence,in this part,machine learning model is adopted to predict the risks of invasive intervention for patients with pancreatic necrosis.AIM: In this study,we aimed to predict the risk of minimally invasive intervention in patients with pancreatic necrosis according to early-stage organ function evaluation based on machine learning model.METHODS: All NAP patients admitted to our hospital within one week from disease onset undergoing at least one percutaneous catheter drainage(PCD)procedure from 2010 to 2018 were screened in this study.The baseline data,organ failure scores within two weeks after admission,PCD procedure associated metrics,as well as the clinical outcomes were collected for each study patients.By feeding different variables,including intervention timing,organ failure scores and clinical outcomes into the machine learning algorithms,a XGBoost model was developed to predict the risk of PCD procedures by differing intervention timing,to further clarify the effects of intervention timing on clinical outcomes in patients with ANP.RESULTS: We finally constructed a XGBoost model,mainly according to the organ function score of patients.The predicted mortality of patients who underwent the first PCD within 28 days from onset or not was obtained.The AUC of the model was about 74.4%.CONCLUSION: The timing of PCD is one of the factors of the mortality.Doctors can determine the timing of PCD according to the organ function of patients,which may be expected to reduce the mortality of patients.However,a huge number of cases are still needed to complete the model construction and verification. |
