Pancreatic necrosis is a complication of both acute and chronic pancreatitis. Its clinical presentation can range from severe, life-threatening sepsis to oligosymptomatic or even asymptomatic patients. Septic patients with infected pancreatic necrosis on one end of the clinical spectrum of severity are critically ill and require immediate treatment. The range of interventional treatment modalities in these patients includes surgery [1], transgastric endoscopic drainage combined with necrosectomy [2, 3], placement of percutaneous drainages and irrigation [4] or combinations of the above modalities. Recent evidence suggests that minimal invasive or endoscopic approaches are superior to open surgery [5–7]. However, even infected necrosis can be managed conservatively in a proportion of patients [8]. Thus, even in this life-threatening condition, the selection of an appropriate treatment modality is an as yet unsolved problem [9].

Given the increased awareness and the improved imaging options of the pancreas such as endoscopic ultrasound, computed tomography and magnetic resonance imaging, more oligosymptomatic or asymptomatic pancreatic lesions including necroses are being identified and represent an additional challenge regarding their adequate management strategy. In contrast to infected necroses, the majority of these patients may be managed conservatively as suggested by recent guidelines [10, 11]. However, the technical advances of endoscopic and minimally invasive treatment modalities may lower the threshold of intervention in this patient group. In choosing an interventional strategy, the procedural risk has to be weighed against the likely natural course of the pancreatic necrosis. Data regarding the procedural risk of endoscopic necrosectomy and the recurrence rate in non-infected patients are scarce. Seifert et al. reported a 26% complication rate and a 7.5% mortality rate in 93 patients for endoscopic necrosectomy [7]. This series included 54% infected necroses – therefore, a significant proportion of this morbidity and mortality may be related to the active infection at the time of intervention. On the other hand, persistent or secondary infection of pancreatic necrosis in conservatively treated patients is of particular concern because potentially delayed treatment may lead to intractable sepsis with resultant mortality. The proportion of non-infected necroses that progress to infection depends on the clinical context. In acute pancreatitis necroses are most likely to become infected three to four weeks after the onset of acute pancreatitis [12]. In contrast to studies on acute treatment, data on secondary infection rates after this time frame are rare. Rau et al. reported a mortality rate of 6.2% in non-surgically treated patients with sterile necrosis [13].

The knowledge of factors predicting long-term success of conservative treatment would be critical to make a risk adjusted initial management decision for a newly diagnosed pancreatic necrosis. This study therefore sought to define morphologic endoscopic ultrasound criteria for the prognosis and need of intervention in a series of patients with pancreatic necroses that were initially managed conservatively.

The two significant predictors in the single variable analysis, namely the liquid score and the diameter of necrosis were correlated. Figure 2 shows the correlation of the liquid score and the diameter of the pancreatic necrotic cavity as scatterplot and the fitted linear model line (r² = 0.55, F-statistic p = 1.59 × 10^-6). In order to define the single most relevant parameter for the prediction of outcome, multiple logistic regression analysis using all potential predictor variables in forward likelihood based inclusion was performed. Here, the liquid score was identified as the single independent predictor (Wald statistic p = 0.009, deviance based statistic p_D = 5.75 × 10^-4). Models constructed by inclusion of the liquid score (deviance statistic _D = 5.75 × 10^-4), the diameter of necrosis (p_D = 0.16) and an interaction term (p_D = 0.99) confirmed these results.

In fact, none of the ten patients without documented fluid within their pancreatic necrosis on EUS needed any intervention during follow-up (median 133 days, range 60–548). On the basis of the logistic regression model, the risk for reaching the endpoint in patients with solid necrosis was estimated at 3%. In contrast, patients with liquid scores of one (low amounts of liquid present) and two (predominantly liquid content) were estimated 18% and 64%, respectively in the logistic regression model. The direct empirical estimates calculated from the data were very similar, namely 62% (8 out of 13) for liquid necrosis, 25% (2 out of 8) for necrosis with low amounts of liquid and 0% (10 out of 10) for solid necrosis.

The armamentarium for interventional and surgical treatment of pancreatic necrosis and cysts has expanded significantly over the last years [2, 4–7]. Although the technical feasibility regarding endoscopic necrosectomy is established and the training level of endoscopists is increasing, a clear definition of indications for intervention is still needed, because a significant proportion of patients with pancreatic necrosis can be managed conservatively [8]. On the other hand, an unduly delayed therapy of the pancreatic necrosis could transform an elective procedure in a stable patient into a high risk intervention in the setting of sepsis. Thus, this study aimed to define EUS criteria to guide the threshold of intervention in pancreatic necrosis that does not require immediate intervention because of sepsis, gastric obstruction or other complications.

In this study, the single most relevant predictor for the need of intervention proved to be the liquid content of the necrotic cavity. In the single predictor analysis, the size of the pancreatic necrosis was also significantly associated with outcome, but proved to be inferior to the liquid score on multivariate logistic regression analysis. Both parameters were highly correlated, which corresponds to the clinical reality – i.e. large necrosis cavities tend to have a predominantly liquid content. Biologically, the presence of undrained liquid may correlate with the chance of a secondary infectious event, since the distance from the blood circulation increases the difficulty to control bacteria by the immune system as well as by systemically given antibiotics while providing a medium for rapid spread of infection through the fluid pool. Indeed, such an infectious event led to intervention in 6 of our 10 patients with an endpoint event. Additionally, fluid could be an indicator of chronic minimal leakage fed by an injured pancreatic duct, leading to problems such as pain or increasing size.

Interestingly, increased echogenicity of fluid was not predictive for treatment necessity during follow up. This corresponds to clinical experience in endoscopically treated patients, that an increase of echogenicity is seen both in putrid fluid as well as in clear dark fluid without evidence of infection. Thus, the morphology of the necrosis as defined by liquid content and diameter defines the clinical course in the long term.

The initial decision for patients with EUS documented pancreatic necrosis for conservative treatment is based on lack of clinical indicators for immediate interventions, i.e. pain, uncontrolled infection, inability of oral nutrition, or suspicion for malignancy. In many patients, the decision based on the presence of these indicative symptoms is not clear-cut but based on a spectrum of clinical variables and personal experience. The non-randomized assignment of patients to conservative treatment is a potential weakness of this study. Being aware of this limitation, the long-term clinical follow-up provides a hard clinical outcome measure of the parameters obtained during the initial clinical evaluation. We thus hope, that the morphological parameters provided here, can help guide the therapeutic management resulting in a judicious use of endoscopic necrosectomy in the future.

Patients with necrotic pancreatic cavities of predominantly liquid content should be monitored closely and may be considered for earlier, elective interventional treatment, because this patient group has a significant risk for severe complications in their clinical course. Conversely, solid necrosis may be relatively safely managed conservatively. Further, preferably randomized studies on this subject are needed.