Accepted Manuscript. S (18)35279-X DOI: Reference: YGAST 62268

Transcription

1 Accepted Manuscript An Endoscopic Transluminal Approach, Compared to Minimally Invasive Surgery, Reduces Complications and Costs for Patients With Necrotizing Pancreatitis Ji Young Bang, MD, MPH, Juan Pablo Arnoletti, MD, Bronte A. Holt, MBBS, FRACP, Bryce Sutton, PhD, Muhammad K. Hasan, MD, Udayakumar Navaneethan, MD, Nicholas Feranec, MD, C.Mel Wilcox, MD, Benjamin Tharian, MD, Robert H. Hawes, MD, Shyam Varadarajulu, MD PII: S (18)35279-X DOI: Reference: YGAST To appear in: Gastroenterology Accepted Date: 12 November 2018 Please cite this article as: Bang JY, Arnoletti JP, Holt BA, Sutton B, Hasan MK, Navaneethan U, Feranec N, Wilcox CM, Tharian B, Hawes RH, Varadarajulu S, An Endoscopic Transluminal Approach, Compared to Minimally Invasive Surgery, Reduces Complications and Costs for Patients With Necrotizing Pancreatitis, Gastroenterology (2018), doi: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

2 An Endoscopic Transluminal Approach, Compared to Minimally Invasive Surgery, Reduces Complications and Costs for Patients With Necrotizing Pancreatitis Ji Young Bang MD, MPH 1, Juan Pablo Arnoletti MD 2, Bronte A. Holt MBBS, FRACP 1, Bryce Sutton PhD 1, Muhammad K. Hasan MD 1, Udayakumar Navaneethan MD 1, Nicholas Feranec MD 3, C. Mel Wilcox MD 4, Benjamin Tharian MD 1, Robert H. Hawes MD 1, Shyam Varadarajulu, MD 1 1. Center for Interventional Endoscopy, Florida Hospital, Orlando, Florida, USA 2. Center for Specialized Surgery, Florida Hospital, Orlando, Florida, USA 3. Department of Radiology, Florida Hospital, Orlando, Florida, USA 4. Division of Gastroenterology-Hepatology, University of Alabama at Birmingham, Birmingham, Alabama, USA Correspondence Shyam Varadarajulu MD Medical Director Center for Interventional Endoscopy, Florida Hospital 601 East Rollins Street, Orlando, FL Tel: Fax: svaradarajulu@yahoo.com Word count: 6,847 words (for abstract, manuscript text, references and legends) ClinicalTrials.gov identifier: NCT Funding: None 1

3 Disclosures JYB, JPA, BH, BS, MH, UN, NF, BT, CMW: No conflicts of interest. SV: Consultant for Olympus America Inc. and Boston Scientific Corporation. RH: Consultant for Olympus America Inc. and Boston Scientific Corporation. Author Contributions JYB: Study design, acquisition of data, analysis and interpretation of data, statistical analysis, drafting of manuscript, critical revision of manuscript. SV: Study concept and design, endoscopist performing procedures in the trial, interpretation of data, drafting of manuscript, critical revision of manuscript. BH: Study design, acquisition of data, endoscopist performing procedures in the trial, critical revision of manuscript. UN, RH, BT: Critical revision of manuscript. JPA: Surgeon performing procedures in the trial, drafting of manuscript, critical revision of manuscript. CMW: Study concept, critical revision of manuscript. BS: Cost data analysis, drafting of manuscript, critical revision of manuscript. NF: Radiologist interpreting cross-sectional imaging in the trial, critical revision of manuscript. MH: Endoscopist performing procedures in the trial. 2

4 ABSTRACT Background & Aims: Infected necrotizing pancreatitis is a highly morbid disease with poor outcomes. Intervention strategies have progressed from open necrosectomy to minimally invasive approaches. We compared outcomes of minimally invasive surgery vs endoscopic approaches for patients with infected necrotizing pancreatitis. Methods: We performed a single-center, randomized trial of 66 patients with confirmed or suspected infected necrotizing pancreatitis who required intervention from May 12, 2014 through March 24, Patients were randomly assigned to groups that received minimally invasive surgery (laparoscopic or video-assisted retroperitoneal debridement, depending on location of collection, n=32) or an endoscopic step-up approach (transluminal drainage with or without necrosectomy, n=34). The primary endpoint was a composite of major complications (new-onset multiple organ failure, new-onset systemic dysfunction, enteral or pancreatic-cutaneous fistula, bleeding and perforation of a visceral organ) or death during 6 months of follow up. Results: The primary endpoint occurred in 11.8% of patients who received the endoscopic procedure and 40.6% of patients who received the minimally invasive surgery (risk ratio, 0.29; 95% CI, ; P=.007). Although there was no significant difference in mortality (endoscopy 8.8% vs surgery 6.3%; P=.999), none of the patients assigned to the endoscopic approach developed enteral or pancreaticcutaneous fistulae compared to 28.1% of the patients who underwent surgery (P=.001). The mean number of major complications per patient was significantly higher in the surgery group (0.69±1.03) compared to the endoscopy group (0.15±0.44) (P=.007). The physical health scores for quality of life at 3 months was better with the endoscopic approach (P=.039) and mean total cost was lower ($75,830) compared to $117,492 for surgery (P=.039). Conclusions: In a randomized trial of 66 patients, an endoscopic transluminal approach for infected necrotizing pancreatitis, compared to minimally invasive surgery, significantly reduced major complications, lowered costs, and increased quality of life. Clinicaltrials.gov no: NCT

5 KEY WORDS: endoscopy; surgery; necrotizing pancreatitis; minimally invasive surgery versus endoscopy randomized (MISER) trial 4

6 BACKGROUND Acute pancreatitis is the third most common gastrointestinal disorder in the United States requiring more than 275,000 hospitalizations annually. 1 Necrotizing pancreatitis occurs in 20% of patients who have acute pancreatitis and is associated with mortality rate of 8-39%. 2 Secondary infection of necrotic tissue is a dreaded complication that may result in sepsis and organ failure. 3 While the traditional treatment approach to infected necrotizing pancreatitis has historically been open surgical necrosectomy, the technique is associated with high rates of adverse events (34-95%) and death (11-39%) Recent evidence suggests that minimally invasive techniques that incorporate a step-up approach of percutaneous catheter placement with subsequent minimally invasive surgical necrosectomy is superior to open surgical necrosectomy with lower rates of postoperative adverse events (40%) and long-term morbidity. 12 As an alternative to surgery, endoscopic approaches have gained increasing acceptance for the treatment of necrotic collections with an adverse event rate between 10.4 and 24.5%, and a mortality rate under 10% Endoscopic techniques have undergone iterative improvements over the past two decades to include endoscopic ultrasound (EUS)-guided transluminal drainage by creation of single or multiple tracts with placement of plastic or metal stents, concomitant placement of percutaneous drainage catheters as well as mechanical debridement of the necrotic tissue via transluminal and/or percutaneous tracts using an endoscope These techniques, when structured to the size and extent of the necrotic collection, appear to yield better treatment outcomes. 18 It is unclear which minimally invasive approach, surgical or endoscopic, is better for the treatment of necrotizing pancreatitis in terms of clinical outcomes, quality of life and costs. Therefore, we conducted the Minimally Invasive Surgery versus Endoscopy Randomized (MISER) Trial in patients with necrotizing pancreatitis. METHODS STUDY DESIGN 5

7 Prior to protocol finalization, pancreatic surgeons, endoscopists and radiologists from three tertiary care institutions deliberated the study methodology, particularly procedural techniques, to keep with current surgical and endoscopic practice standards in the United States ,19-22 Indications for intervention were consistent with recent guidelines. 23 Adult patients with confirmed or suspected infected pancreatic or peripancreatic necrosis or both, warranting intervention and whose collections were amenable to both endoscopic and surgical necrosectomy were eligible for randomization. Infected necrosis was defined by the presence of gas in the necrotic collection on computed tomography (CT) or positive culture of necrotic tissue obtained preprocedure or at first intervention. Infected necrosis was also suspected when sepsis was persistent or in the presence of ongoing clinical deterioration. Exclusion criteria were prior surgical or endoscopic drainage or necrosectomy, pancreatitis secondary to trauma or surgical intervention, presence of indwelling percutaneous catheters prior to randomization, chronic pancreatitis or pregnancy. All authors had access to the study data and have reviewed and approved the final manuscript. The trial was conducted at Florida Hospital in Orlando. Patients were randomized to undergo treatment by either the minimally invasive surgical or endoscopic treatment approaches. Computergenerated block randomization assignments were placed in opaque sealed envelopes. Concealed allocation was performed to minimize selection bias in this randomized trial. Block sizes of 4 were used for preparation of the randomization sequence. Randomization was performed by a study coordinator when patients met inclusion criteria. STUDY OVERSIGHT Patients or their legal representatives provided written informed consent prior to randomization. The study was approved by the Institutional Review Board. PREPROCEDURE EVALUATION 6

8 All patients were evaluated by surgical, interventional endoscopy, critical care and infectious diseases teams. An interdisciplinary expert panel consisting of a gastroenterologist, gastrointestinal surgeon and radiologist reviewed all cross-sectional images. Whenever possible, a dedicated magnetic resonance imaging (MRI) of the abdomen was obtained to quantify the degree of necrosis. If required, randomization was postponed to facilitate better demarcation of the necrotic collection so that an intervention could be performed. 23,24 Patients who could not wait and underwent interval placement of percutaneous catheters were excluded. All surgical and endoscopic interventions were undertaken within 72 hours of randomization. In order to facilitate direct comparison between treatment strategies and minimize the risk of fistula formation, placement of percutaneous catheters was permitted after randomization if patients met either one of these two criteria: the planned intervention was video-assisted retroperitoneal debridement (VARD) or to drain necrotic pockets not in communication with the dominant collection to be intervened. However, to ensure inclusion of only critically ill patients, any subjects who improved clinically within 72 hours of percutaneous catheter placement were excluded from the study. In patients who were alert and communicative, health-related quality of life (HRQoL) was assessed by administering the Medical Outcomes Study 36-Item Short-Form General Health Survey (SF-36) questionnaire. MINIMALLY INVASIVE SURGICAL APPROACH Laparoscopic cystogastrostomy with internal debridement was performed if necrosis was confined to the lesser sac. 19 If necrosis was more extensive and retroperitoneal access was available, then VARD with postoperative lavage was performed. 25,26 Further details of the surgical approach can be seen in Supplementary Figure 1. Technique of laparoscopic cystogastrostomy with pancreatic necrosectomy With the patient in the supine position and under general anesthesia, pneumoperitoneum was obtained by insufflating carbon dioxide via a 12mm trocar inserted below the umbilicus. Two additional 7

9 5mm trocars were inserted in each upper quadrant and another 12mm trocar in the right upper quadrant. An anterior longitudinal gastrostomy was performed with an ultrasonic power device (Harmonic ACER+7 Shears, Ethicon, Cincinnati, OH) and the area of maximum bulge was identified in the posterior gastric wall. The presence of the necrotic collection was confirmed with a laparoscopic ultrasound probe. The area of necrosis was entered directly with the ultrasonic power device and its liquid contents were aspirated. Cultures were obtained and submitted to microbiology. A cystogastrostomy was then created by firing a laparoscopic 60mm linear powered stapler (Echelon FlexTM, Cincinnati, OH). Necrotic debris was removed utilizing laparoscopic graspers and 10mm suction cannula. After completion of necrosectomy, a nasogastric tube was placed though the cystogastrostomy and into the area of necrosectomy for irrigation. The anterior gastrostomy was closed by firing the linear stapler and a 15Fr drain was placed adjacent to the closed gastrostomy site through one of the 5mm trocar sites. Technique of video-assisted retroperitoneal debridement (VARD) A percutaneous pigtail catheter was initially inserted by an interventional radiologist under imaging guidance. In the operating room and under general anesthesia, a 15mm skin incision was made at the pigtail catheter site and the catheter was followed into the necrotic collection. The tract was gently dilated and a 15mm laparoscopy trocar was inserted into the area of necrotic collection under direct vision. Gasless laparoscopy was then performed with a 5mm, 0 o camera. Necrotic debris was removed with 5mm laparoscopic graspers and contents were irrigated and aspirated with a 5mm suction cannula. Cultures were obtained and submitted to microbiology. A 16Fr rubber catheter (for future irrigation) together with a 0.5-inch drain was placed through the incision site and secured to the skin with sutures. An ostomy bag was then placed over the VARD site to allow drainage. ENDOSCOPIC APPROACH Drainage procedures 8

10 All procedures were performed under general anesthesia or monitored anesthesia care using propofol administered by anesthesia providers. Adopting a structured approach, unilocular collections ( 60mm but <80mm) were treated by single tract transmural cystogastrostomy/duodenostomy (singlegate technique). If the collections were 80mm in size or extended to the flanks, drainage was performed by creation of multiple transmural tracts (multi-gate technique). 13,18 Linear-array endoscopic ultrasound was used to identify the collection. After puncturing the collection using a 19G fine needle aspiration (FNA) needle (Expect, Boston Scientific Corp., Marlborough, MA) and coiling a inch guidewire (VisiGlide, Olympus America Inc., Center Valley, PA) under fluoroscopic guidance, the transmural tract was created using a taper-tip 4.5Fr cannula (Proforma, Conmed, Utica, NY) or a 10Fr cautery device (Cystotome, Cook Endoscopy, Winston-Salem, NC). Further dilation to 15mm was performed using a radial expansion balloon (CRE, Boston Scientific Corp.) and two 7Fr, 4cm double pigtail plastic stents (Cook Endoscopy) were placed. Lumen apposing metal stents (LAMS) (Hot AXIOS, Boston Scientific Corp.) were available in the United States from September 2015 after Federal Drug Administration (FDA) approval. LAMS were placed in lieu of plastic stents in patients with unilocular collections in the tail of the pancreas who were at low risk for disconnected pancreatic duct syndrome (DPDS). 27 In patients undergoing multi-gate drainage in whom DPDS was suspected, one tract was created using LAMS and another tract (towards the pancreatic body/tail) using plastic stents. This strategy was adopted because indwelling plastic stents were left in situ indefinitely in patients with DPDS (to minimize pancreatic fluid collection recurrence) and it was oftentimes difficult to exchange the LAMS for plastic stents after necrosis resolution LAMS used in this study were 15mm in diameter and 10mm in length. After puncturing the necrotic collection using the electrocautery tip, the delivery catheter was advanced into the collection and the distal flange was first deployed under EUS guidance. The proximal flange was then released under EUS-guidance or endoscopic view. Further details of the endoscopic approach can be seen in Supplementary Figure 1. 9

11 Endoscopic Necrosectomy Endoscopic necrosectomy was performed at the index session only if there was no drainage of liquid debris after transluminal stent placement. Necrosectomy was undertaken using a cap-fitted, singlechannel therapeutic gastroscope (outer diameter 11mm; Olympus America Inc.). The technique of necrosectomy was tailored to the size of the necrotic cavity and the degree of adherence of debris and involved the following three steps: debridement, extraction of necrotic debris and irrigation. Debridement: 20-30mm polypectomy snares were used if the necrotic collection was larger than 80mm in size and smaller (<20mm) snares were used if the necrotic collection was less than 80mm in size. If the debris was non-adherent, 15-30mm oval snares (Acusnare, Cook Endoscopy) were used. If the necrotic debris was adherent, removal was accomplished using 15-25mm round, braided-wire snares (Captivator II, Boston Scientific Corp.). Only cold snare technique was adopted for debridement unless the necrotic debris was adherent to the cavity. In such instances, electrocautery-assisted debridement (ERBE USA Inc., Marietta, GA) was performed using the following settings: Endo Cut Q, Effect 3, Cutting duration 1, Cutting interval 6. The polypectomy snare was placed around the base of the necrotic debris, closed tightly and then lifted gently to ensure absence of entangled vasculature. Electrocautery was administered and the necrotic material was peeled away from the walls of the necrotic cavity. In patients with extensive collateral vasculature within the necrotic cavity, to minimize the risk of snaring a vessel, debridement was performed using a 19.5mm wide-jaw, rat tooth forceps (Rat Tooth Alligator Jaw Grasping Forceps, Olympus America Inc.). Intra-procedural bleeding was managed using hemostatic forceps (Coagrasper Hemostatic Forceps, Olympus America Inc.). Extraction of debris: After necrosectomy, liquefied debris was suctioned using the gastroscope. Solid debris was extracted by suctioning chunks into the cap with the aid of wide-jaw, rat tooth forceps or polypectomy snares. Smaller ( 14.9mm) rat tooth forceps were used for extraction of debris in patients with lumen-apposing metal stents (LAMS) to prevent entanglement of the distal flange and accidental dislodgement of the stent. As the final step, a retrieval net was used to remove any residual debris. 10

12 Irrigation: Normal saline was used intermittently for irrigation of the necrotic cavity during the procedure mL of half-strength hydrogen peroxide mixed in equal volume of normal saline was used for irrigation towards the completion of each necrosectomy session with the intent of sterilizing the necrotic area. Hydrogen peroxide was used only towards the end of the procedure because the emanating effervescence precluded adequate visualization of the necrotic cavity. Post-necrosectomy treatment: Two 7Fr, 4cm double pigtail plastic stents were left in situ in patients undergoing transgastric necrosectomy to maintain patency of the tract. An 18-24Fr nasocystic catheter was placed to facilitate lavage of the necrotic cavity by flushing 250 ml of normal saline every 4-6 hours, with frequent repositioning of the patient to facilitate irrigation and drainage of the necrotic contents. ADDITIONAL INTERVENTIONS If oral nutrition was not tolerated, enteral feeding via laparoscopic jejunostomy (surgical cohort) or endoscopically via percutaneous transgastric jejunal feeding tube (endoscopy cohort) was initiated. In patients with abdominal distension or those suspected to be at high-risk for developing compartment syndrome, the feeding tube was secured using T-fasteners to prevent dislodgement. Based on interdisciplinary consensus, reinterventions were undertaken at predetermined intervals if there was no clinical improvement (Supplementary Figure 2). In the endoscopic cohort, necrosectomy was performed if the contents of the cavity were predominantly (>50%) solid and additional transmural (multi-gate) tracts were created for drainage if the contents were predominantly liquid. 18 For collections inaccessible by endoscopy, percutaneous catheters were placed for drainage. Likewise, in the surgical arm, percutaneous catheters were placed and VARD was performed as required until clinical improvement was observed. POST-DISCHARGE FOLLOW-UP 11

13 Patients were discharged following clinical resolution of symptoms. An abdominopelvic CT and outpatient follow-up was scheduled at 6-weeks. The enteral feeding tube and percutaneous catheters were removed if patients were clinically well, CT scan demonstrated resolution of the necrotic collection and the output was less than 10mL/day. For the endoscopic cohort, as in routine clinical practice, a magnetic resonance cholangiopancreatography (MRCP) was performed at the initial outpatient clinic visit to assess the status of the main pancreatic duct. An ERCP was undertaken in lieu of an MRCP when the latter was contraindicated. If the main pancreatic duct was found to be intact, then the transmural stents were removed. In patients originally treated with LAMS but diagnosed with DPDS at follow-up, the LAMS were exchanged for double pigtail plastic stents. The plastic stents were left in situ in patients with DPDS to minimize the risk of recurrent collections. 28 If a pancreatic duct leak was identified on pancreatography, a transpapillary stent was placed to bridge the leak. Patients with persistent symptoms and necrotic collection on follow-up CT scan underwent further interventions as determined by the interdisciplinary panel. Further details on patient follow-up and data collection are included in Supplementary Table 1. ENDPOINTS The primary endpoint was a composite of major complications comprising new onset multiple organ failure or systemic dysfunction, enteral or pancreatic-cutaneous fistula, intraabdominal bleeding, visceral perforation or death during admission and until 6 months after discharge. In addition to analyzing the individual primary endpoint components as secondary endpoints, other analyzed variables included mean number of major complications per patient, presence of systemic inflammatory response syndrome (SIRS) at 72 hours after index intervention, surgical site infection, incisional hernia, pancreatic exocrine and endocrine status, procedure and disease-related adverse events, post-procedure length of intensive care unit (ICU) and hospital stay, health-related quality of life (HRQoL) and total costs (Table 1). DATA COLLECTION 12

14 All variables were collected in individual patients by two dedicated research coordinators over the 6-month period (Supplementary Table 1). All pre-intervention MRI and CT scans were prospectively evaluated by a dedicated radiologist blinded to study group assignments. During hospitalization, data were collected daily to include all interventions performed, presence of SIRS, complications, organ failure and systemic dysfunction. All primary and secondary endpoints were assessed by a research coordinator at discharge. Additionally, study participants were evaluated at 6 weeks (initial outpatient follow-up), 3 months and 6 months post-discharge for hospital readmissions, indications for readmissions, reinterventions and complications. HRQoL was assessed pre-intervention, at hospital discharge, 3 and 6 months post-discharge by research personnel blinded to the treatment group assignment. Hospital costs were evaluated in individual patients on completion of the 6-month follow-up. A research manager audited the data on a monthly basis and discrepancies if any were resolved by cross-referencing case record forms against onsite source data in consultation with an independent physician who was unaware of the randomization assignments and not involved in the conduct of the clinical trial. A quarterly audit was conducted jointly by the director of research, research manager and both study coordinators. Additionally, the study was audited biannually by the Institutional Review Board. Outcomes assessment was performed by an adjudication committee consisting of a gastroenterologist, interventional radiologist and an expert statistician who reviewed all data. A safety committee comprising a surgeon, a nurse practitioner and a gastroenterologist, who were not involved in the study, was appointed to oversee the clinical trial and examine safety endpoints after inclusion of the first 30 patients. STATISTICAL ANALYSIS A two-sided sample size calculation was performed based on the rate of significant morbidity and mortality, which was estimated at 12.5% for endoscopy 13,15,16,18,28,31,32 and 50% for surgery. 12,16,33 A sample size calculation performed using Stata 14 statistical software (StataCorp LP, College Station, TX) at 90% 13

15 power and α=0.05 resulted in sample size estimation of 60 patients (30 per cohort) and hence was set at 32 patients per treatment group to account for a 5% drop-out. Continuous data were summarized as means with standard deviation or medians with interquartile range and range, and were compared using Wilcoxon rank-sum test. Categorical data were summarized as frequencies with percentages and were compared using the Chi-square or Fisher s exact test (as indicated). Risk ratios with 95% confidence intervals (CI) were calculated for categorical variables. The primary endpoint was also evaluated using the Cox proportional hazards model to account for variations in time to outcome and to identify factors associated with the primary endpoint. All analyses were performed using both intention-to-treat and per-protocol methods. No interim analysis was performed; however, an independent biostatistician had access to the database and performed sequential monitoring of major adverse events to report findings to the safety committee on a quarterly basis. Two-sided p- values were reported for comparison of all outcome measures and no adjustments were made for multiple testing. Statistical significance was determined at p<0.05. All statistical analyses were performed using Stata 14 (StataCorp LP). QUALITY OF LIFE ASSESSMENT For quality of life assessment, the change in physical and mental health component scores from baseline through follow-up evaluation was analyzed using a repeated-measures analysis based on mixedeffects individual growth models with random intercept and slope. 34 Individual growth models permit the analysis of inter- and intra-individual change, estimating both average and individual trajectories of physical and mental component scores over time, at pre-intervention, hospital discharge, 3 and 6 months post-discharge. Given the variable follow-up times in the trial, mixed models allow time to be treated as a continuous or discrete variable in which subjects do not have to be measured at the same time. The models make use of all data across all subjects and time, which increase the statistical power, and allow for correlation among repeated observations. To avoid imposing structure on the trajectory of the component scores, time was treated as a categorical variable. 14

16 COST ANALYSIS All relevant costs pertaining to treatment were taken into consideration: procedure costs, inpatient hospital stay from the date of procedure to discharge, readmissions, medications, materials, anesthesia, pharmacy and imaging studies. Costs were based on Medicare reimbursement fee structure and expressed in 2017 U.S. dollars. Generalized linear models (GLM) were used to examine the differences in cost between study groups. Prior to fitting the GLM, skewness of the cost data was assessed and determined to be highly positively skewed. The Pregibon and modified Park tests indicated that the GLM should be fitted to the study data using a log link and gamma distribution for the mean-variance relationship of the response variable. 35,36 Adjusted mean cost estimates by cost categories and study group were calculated using the method of recycled predictions. 37 Quality-adjusted life years (QALY) take into account both the quantity and the quality of life generated by healthcare interventions. It is the arithmetic product of life expectancy and a measure of the quality of the remaining life years. QALYs were calculated as the product sum of SF-6D based health utilities at successive measurements during follow-up (3 to 6 months after randomization) and the lengths of times in between measurement and baseline. We calculated confidence intervals for between-group differences using bias-corrected and accelerated (BCa) bootstrapping, stratified by treatment group and drawing 1000 samples of the same size as the original sample separately for each group and with replacement. RESULTS PATIENTS Between May 12, 2014 and March 24, 2017, total of 168 patients with acute necrotizing pancreatitis were assessed for eligibility and 70 met criteria for randomization (Supplementary Figure 3). 15

17 Four patients were excluded post-randomization due to resolution of symptoms after percutaneous catheter placement (n=2) or protocol violation (n=2). Patient and disease characteristics of the 66 patients in the two groups were similar (Table 2). Similar results were observed between intention-to-treat and per-protocol analyses (Table 3, Supplementary Table 2). MINIMALLY INVASIVE SURGICAL APPROACH Of 32 patients assigned to surgery, laparoscopic cystogastrostomy with internal debridement was attempted in 26 patients and VARD in 6. Nine (28.1%) patients had an indwelling percutaneous catheter at time of surgery. One patient without retroperitoneal access for VARD also failed laparoscopic intervention due to extensive necrosis of the tissue planes and therefore underwent open necrosectomy. Surgical intervention was aborted in two patients (American Society of Anesthesiologists [ASA)] class IV) due to intraoperative cardiovascular shock and both were later treated by transgastric endoscopic necrosectomy. Both patients had large, non-communicating, necrotic collections extending to the lower abdomen/pelvis and did not have clinical improvement with pre-intervention percutaneous catheter placement. Both of these patients had successful clinical outcomes at 6-month follow-up. After the first 72 hours, 12 (37.5%) patients without clinical improvement were managed by VARD in five (index intervention was laparoscopic cystogastrostomy in three and VARD in two) and/or percutaneous catheter placement in five patients. Three patients (ASA class III in two and ASA class IV in one) without clinical improvement were deemed by the interdisciplinary panel to be high-risk for repeat surgical intervention and hence were treated by endoscopic necrosectomy (n=1) or endoscopic drainage adopting the multi-gate technique (n=2). All three of these patients had successful clinical outcomes at 6-month follow-up. Two patients with postoperative gastrointestinal bleeding developed intraabdominal hematoma due to local hemorrhage at the cystogastrostomy site that were treated endoscopically using cautery and another patient had bleeding from a fistulous tract that required surgery 203 days post-intervention. Additional procedure-related adverse events included bleeding at the cystogastrostomy site in three 16

18 patients and infection at the feeding jejunostomy site in one, which were all managed conservatively. Four patients had enterocutaneous fistulae that were colonic in two, gastric in one, and duodenal in one. While one patient with the gastric fistula died due to underlying disease, the colonic fistulae were treated by surgery in two patients at 173 and 203 days, respectively. The duodenal fistula resolved spontaneously at long-term follow-up with conservative measures. Pancreatic fistulae were observed in four patients who underwent VARD and in four others who underwent percutaneous catheter placement at laparoscopic debridement. Three of the patients with pancreatic fistulae also had enterocutaneous fistulae. ENDOSCOPIC APPROACH Eighteen of 34 index endoscopic interventions were performed under monitored anesthesia care using propofol and 16 under general anesthesia. Index endoscopic interventions comprised drainage using the single-gate technique in 12 (35.3%) and multi-gate technique in 22 (64.7%). One patient treated by single-gate technique also underwent endoscopic necrosectomy. 16 patients underwent placement of LAMS, 18 were treated with plastic stents and there was no significant difference in treatment outcomes between patients treated only with LAMS versus plastic stents in this study. Eight of 22 patients treated with multi-gate technique were suspected to have DPDS and hence underwent placement of both LAMS and plastic stents. 14 (41.2%) patients had an indwelling percutaneous catheter at time of intervention. At 72 hours, 15 patients (44.1%) did not have clinical improvement and were managed by endoscopic transluminal necrosectomy in 11, multi-gate technique in three and/or percutaneous catheter placement in six patients. Procedure-related adverse events were encountered in four patients that included small bowel obstruction due to stent migration, which was managed conservatively (n=1), post-procedure cardiopulmonary arrest, in which the patient was resuscitated successfully (n=1), bleeding during transmural tract dilation that resolved spontaneously (n=1) and migration of the percutaneous feeding tube bumper into the gastric wall layers that was treated endoscopically (n=1). 17

19 MRCP or ERCP identified DPDS in 30 (88.2%) patients, pancreatic duct leak in two (5.9%) and two (5.9%) patients refused to undergo both studies. Patients diagnosed with DPDS at follow-up (n=22) were treated with indwelling plastic endoprostheses. Both patients with pancreatic duct leak were successfully treated by transpapillary pancreatic duct stent placement. ENDPOINTS Comparison of the primary and secondary endpoints are detailed in Table 3. The composite primary endpoint of major complications or death occurred in four patients (11.8%) in the endoscopic and 13 patients (40.6%) in the surgical cohort (risk ratio 0.29; 95% confidence interval (CI) , p=0.007). The difference was mainly due to the occurrence of enteral or pancreatic-cutaneous fistulae in the surgical cohort (0 vs. 28.1%, p=0.001). Total of five patients (7.6%) died with no significant difference between the cohorts: three in the endoscopic (8.8%) and two in the surgical group (6.3%). The causes of death were multiple organ failure in two patients in both groups and cardiopulmonary arrest in one patient in the endoscopic cohort. The mean number of major complications per patient was significantly higher for surgery compared to endoscopy (0.69 [standard deviation (SD)=1.03] vs [SD=0.44], p=0.007). Also, although there was no significant difference in baseline characteristics between the two treatment groups (Table 2), on Cox proportional hazards analysis, surgical treatment and acute physiology scores were associated with the occurrence of the primary endpoint of major complications or death (Table 4). At 72 hours, lower rate of SIRS was seen in the endoscopic group compared to surgery (20.6 vs. 65.6%, p<0.001), with early resolution of preexisting SIRS (68.8 vs. 18.8%, p=0.004) and fewer patients with new-onset SIRS (5.6 vs. 56.3%, p=0.002) after index intervention (Figure 1). Additionally, the postprocedure length of ICU stay was significantly shorter for the endoscopic cohort (median days [range], 0 [0-38] vs. 0 [0-53], p=0.044). 18

20 At 6-month follow-up, significantly more patients in the surgical group had disease-related adverse events than the endoscopic group (43.8 vs. 5.9%, p<0.001). More details are available in Supplementary Table 3. QUALITY OF LIFE AND COSTS The SF-36 questionnaire showed significantly better physical component scores for the endoscopic treatment group at 3 months (p=0.039) (Table 5). The mean total of direct and indirect medical costs per patient during admission and up to 6-month follow-up were US$75,830 for endoscopic and $117,492 for surgical approach, for a mean absolute difference of $41,662 per patient (Table 6). The number of QALYs gained for the endoscopy group was (BCa 95% CI, ) compared with (BCa 95% CI, ) for surgery. The mean difference was ( to 0.025). For every gain in QALYs, the savings would be $41,662 or $20,831,000 for a single QALY. The probability of the endoscopic group being cost-effective is 0.86 at a societal willingness to pay level of $50,000 per QALY (Supplementary Table 4). DISCUSSION The current study shows that the endoscopic approach, as compared with minimally invasive surgery, reduced the rate of composite endpoint of major complications or death, mean number of major complications per patient, rate of disease-related adverse events, and costs, among patients with confirmed or suspected infected necrotizing pancreatitis. In this study, as in prior randomized trials, the primary endpoint was defined as a composite of major complications or death during 6-month follow-up, rather than resolution of sepsis or necrotic collection as this enables the most accurate comparison of the two treatment disciplines. Our findings are consistent with prior observations on minimally invasive surgical and endoscopic treatment approaches. 12,16 In a randomized trial, endoscopic necrosectomy was shown to 19

21 significantly reduce the proinflammatory response compared to surgical necrosectomy. 16 Likewise, in the present study, while the rate of SIRS decreased after endoscopic interventions, they increased after surgical necrosectomy and this difference remained significant even at 72 hours. General anesthesia is recognized to induce or prolong systemic inflammation in critically ill patients. 38 In addition to being more invasive, the longer procedural duration necessitates prolonged periods of general anesthesia for surgical interventions. Also, patients in the endoscopic treatment group did not develop enteral or pancreatic-cutaneous fistulae compared to more than one-fourth in surgical cohort. Although surgical step-up approach, as compared to open surgical necrosectomy, has been shown to reduce major complications such as organ failure, no significant difference has been observed in the incidence of enteral or pancreatic-cutaneous fistulae between these treatment approaches. 39 Surgical necrosectomy (open or minimally invasive) and local drainage (percutaneous or surgical) enable external extravasation of enzyme-rich inflammatory exudate and hence the potential for pancreatic-cutaneous fistula formation, with their persistence depending on the presence of proximal pancreatic ductal obstruction or a disconnected pancreatic segment. 40 Enteral fistulae may occur as a consequence of exocrine secretions dissecting throughout the retroperitoneal tissue planes causing small or large bowel necrosis due to vascular infarction. This risk may be compounded by iatrogenic insults such as repeated instrumentation or pressure necrosis on a segment of bowel by an adjacent drain. 41 All patients with an enterocutaneous fistula in our study required more than one surgical intervention or multiple port placements. Endoscopic approach on the other hand utilizes a natural orifice as the access route with drainage or debridement being performed internally. Given the minimally invasive nature of both treatment approaches, we did not observe any differences in the rates of new-onset diabetes or exocrine insufficiency between the groups. However, disease-related adverse events such as infection and abdominal pain were significantly higher in the surgical cohort. This is of obvious importance as these factors likely have considerable impact on quality of life, increase utilization of health care resources and therefore contribute to higher treatment costs. 20

22 There are several important differences between the MISER trial and other preceding randomized trials that have addressed a similar question. 16,42 First, the MISER study population comprised a more severely ill cohort with 95% being ASA class III or IV, compared to only 7 and 30%, respectively, in other studies. 16,42 Also, the median APACHE-II scores of 30 and 21 for the endoscopic and surgical groups, respectively, were higher than in prior studies. Therefore, given the severity of underlying illness, percutaneous catheter placement when applied as sole therapy in some of our subjects did not yield clinical improvement. Second, one-third to 51% of patients in the surgical step-up group underwent only percutaneous drainage as treatment in prior studies. 12, 42 The methodology in the present study was more rigorous because we excluded subjects who improved clinically with percutaneous catheters. While the placement of additional percutaneous catheters in patients without clinical improvement may have delayed or obviated the need for a surgical intervention, it would have precluded a meaningful direct comparison between endoscopic and surgical treatment approaches. Third, repeat interventions were undertaken by assessing clearly defined clinical outcomes at predetermined intervals. This was important given the lack of definitive guidelines or recommendations for the treatment of acute necrotizing pancreatitis. Additionally, rather than performing standard, single-tract, transmural drainage, we incorporated techniques such as the multi-gate technique using devices such as LAMS in the endoscopic cohort and adopted both laparoscopic and VARD techniques in the surgical arm, thereby tailoring interventions to the size and extent of the necrosis. Consequently, despite being sicker, the median length of hospital stay was shorter and the mortality rate was less than 10%, which are lower than in prior studies. 12,16,42 Fourth, in addition to enterocutaneous fistula, we also included pancreatic-cutaneous fistula as a major endpoint. This complication occurs primarily in percutaneous and surgery-based interventions leading to extended hospital stay, additional interventions such as pancreatic resection, readmissions, higher cost, and poor quality of life. 43 The difference in primary endpoint in this study was mainly due to the occurrence of enteral or pancreatic-cutaneous fistulae in the surgical cohort, which is an inherent limitation of the technique. Fifth, we compared HRQoL between treatment groups, which is important from a patient perspective. The physical health scores were significantly better in the endoscopy group at 21

23 3-month follow-up, which could be attributed to the shorter procedural duration, faster resolution of SIRS, fewer disease-related adverse events and shorter length of ICU stay. However, a limitation of the MISER trial is that it was a single center investigation conducted at a tertiary referral center and the LAMS was commercially available only when the study was in progress contributing to heterogeneity in the type of endoprosthesis used. In summary, our findings indicate that given the overall clinical, health-related and economic advantages, the endoscopic approach should be the preferred treatment strategy for very ill patients with infected necrotizing pancreatitis. Clinicaltrials.gov number: NCT

24 REFERENCES 1. Peery AF, Crockett SD, Barritt AS, et al. Burden of Gastrointestinal, Liver, and Pancreatic Diseases in the United States. Gastroenterology 2015; 149: Tenner S, Baillie J, DeWitt, Vege SS. American College of Gastroenterology guideline: management of acute pancreatitis. Am J Gastroenterol. 2013; 108: Whitcomb DC. Acute pancreatitis. N Engl J Med 2006; 354: Rodriguez JR, Razo AO, Targarona J, et al. Debridement and closed packing for sterile or infected necrotizing pancreatitis: insights into indications and outcomes in 167 patients. Ann Surg 2008; 247: Connor S, Alexakis N, Raraty MG, et al. Early and late complications after pancreatic necrosectomy. Surgery 2005; 137: Freeman ML, Werner J, van Santvoort HC, et al. Interventions for necrotizing pancreatitis: summary of a multidisciplinary consensus conference. Pancreas 2012; 41: Rau B, Bothe A, Beger HG. Surgical treatment of necrotizing pancreatitis by necrosectomy and closed lavage: changing patient characteristics and outcomes in a 19-year, single-center series. Surgery 2005; 138: Buchler MW, Gloor B, Muller CA, Friess H, Seiler CA, Uhl W. Acute necrotizing pancreatitis: treatment strategy according to the status of infection. Ann Surg 2000; 232: Ashley SW, Perez A, Pierce EA, et al. Necrotizing pancreatitis: contemporary analysis of 99 consecutive cases. Ann Surg 2001; 234: Tsiotos GG, Luque-de Leon E, Sarr MG. Long-term outcome of necrotizing pancreatitis treated by necrosectomy. Br J Surg 1998; 85: Howard TJ, Patel JB, Zyromski N, et al. Declining morbidity and mortality rates in the surgical management of pancreatic necrosis. J Gastrointest Surg 2007; 11:

25 12. Van Santvoort HC, Besselink MC, Bakker OJ, et al. A step-up approach or open necrosectomy for necrotizing pancreatitis. N Engl J Med 2010; 362: Varadarajulu S, Phadnis MA, Christein JD, Wilcox CM. Multiple transluminal gateway technique for EUS-guided drainage of symptomatic walled-off necrosis. Gastrointes Endosc 2011; 74: Gluck M, Ross A, Irani S, et al. Dual modality drainage of symptomatic walled-off pancreatic necrosis reduces length of hospitalization, radiological procedures, and number of endoscopies compared to standard percutaneous drainage. J Gastrointest Surg 2012; 16: Gardner TB, Coelho-Prabhu N, Gordon SR, et al. Direct endoscopic necrosectomy for the treatment of walled-off pancreatic necrosis: results from a multi-center U.S. series. Gastroinest Endosc 2011; 73: Bakker OJ, van Santvoort HC, van Brunschot S, et al. Endoscopic transgastric vs. surgical necrosectomy for infected necrotizing pancreatitis: a randomized trial. JAMA 2012; 307: Sarkaria S, Sethi A, Rondon C, et al. Pancreatic necrosectomy using covered esophageal stents: a novel approach. J Clin Gastroenterol 2014; 48: Bang JY, Holt BA, Hawes RH, et al. Outcomes after implementing a tailored endoscopic step-up approach to walled-off necrosis in acute pancreatitis. Br J Surg 2014; 101: Zyromski N, Nakeeb A, House MG, Jester AL. Transgastric pancreatic necrosectomy: How I do it. J Gastrointest Surg 2016; 20: Harrison S, Kakade M, Varadarajula S, Parden J, Morgan D, Christein J. Characteristics and outcomes of patients undergoing debridement of pancreatic necrosis. J Gastrointest Surg 2010; 14: Parekh D. Laparoscopic-assisted pancreatic necrosectomy: A new surgical option for treatment of severe necrotizing pancreatitis. Arch Surg 2006; 141: Horvath K, Freeny P, Escallon J, et al. Safety and efficacy of video-assisted retroperitoneal debridement for infected pancreatic collections: a multicenter, prospective, single-arm phase 2 study. Arch Surg 2010; 145:

26 23. Working Group IAP/APA Acute Pancreatitis Guidelines. IAP/APA evidence-based guidelines for the management of acute pancreatitis. Pancreatology 2013; 13:e Besselink MG, Verwer TJ, Schoenmaeclers EJ, et al. Timing of surgical intervention in necrotizing pancreatitis. Arch Surg 2007; 142: Van Santvoort HC Besselink MG, Hovath KD, et al. Videoscopic assisted retroperitoneal debridement in infected necrotzing pancreatitis. HPB (Oxford) 2007; 9: Hovath KD, Kao LS, Ali A, Wherry KL, Sinanan MN. Laparoscopic-assisted percutaneous drainage of infected pancreatic necrosis. Surg Endosc 2001; 15: Bang JY, Navaneethan U, Hasan MK, et al. EUS correlates of disconnected pancreatic duct syndrome in walled-off necrosis. Endosc Int Open. 2016; 4:E Bang JY, Wilcox CM, Navaneethan U, et al. Impact of disconnected pancreatic duct syndrome on the endoscopic management of pancreatic fluid collections. Ann Surg 2018; 267: Bang JY, Hasan M, Navaneethan U, et al. Lumen-apposing metal stents (LAMS) for pancreatic fluid collection (PFC) drainage: may not be business as usual. Gut 2017; 66: Bang JY, Hasan MK, Navaneethan U, et al. Lumen-apposing metal stents for drainage of pancreatic fluid collections: When and for whom? Dig Endosc. 2017; 29: Varadarajulu S, Bang JY, Phadnis MA, Christein JD, Wilcox CM. Endoscopic transmural drainage of peripancreatic fluid collections: outcomes and predictors of treatment success in 211 consecutive patients. J Gastrointest Surg 2011; 15: Bang JY, Wilcox CM, Trevino J, et al. Factors impacting treatment outcomes in the endoscopic management of walled-off pancreatic necrosis. J Gastroenterol Hepatol 2013; 28: Raraty MG, Halloran CM, Dodd S, et al. Minimal access retroperitoneal pancreatic necrosectomy: improvement in morbidity and mortality with a less invasive approach. Ann Surg 2010; 251: Brown H, Prescott R. Normal mixed models. In: Brown H, Prescott R, eds. Applied mixed models in medicine. 2nd ed. Hoboken, NJ: John Wiley & Sons, 2006;

27 35. Manning WG, Mullahy J, O Hagan A, Thompson S. Estimating log models: to transform or not to transform? J Health Econ 2001; 20: Pregibon D. Goodness of link tests for generalized linear models. J R Stat Soc Ser C Appl Stat 1980; 29: Basu A, Rathouz PJ. Estimating marginal and incremental effects on health outcomes using flexible link and variance function models. Biostat Oxf Engl 2005; 6: Strom T, Martinussen T, Toft P. A protocol of no sedation for critically ill patients receiving mechanical ventilation. Lancet 2010; 375: Cirocchi R, Trastulli S, Desiderio J, et al. Minimally invasive necrosectomy versus conventional surgery in the treatment of infected pancreatic necrosis: A systematic review and a meta-analysis of comparative studies. Surg Laparosc Endosc Percutan Tech. 2013; 23: Tsiotos GG, Smith CD, Sarr MG. Incidence and management of pancreatic and enteric fistulas after surgical management of severe necrotizing pancreatitis. Arch Surg. 1995; 130: Doberneck RC. Intestinal fistula complicating necrotizing pancreatitis. Am J Surg. 1989; 158: Van Brunschot S, van Grinsven J, van Santvoort HC, et al. Endoscopic or surgical step-up approach for infected necrotizing pancreatitis: a multicenter randomized trial. Lancet 2018; 391: Voss M, Ali A, Eubanks WS, Pappas TN. Surgical management of pancreaticocutaneous fistula. J Gastrointest Surg 2003; 7: American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference: definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Crit Care Med 1992; 20: Banks PA, Bollen TL, Dervenis C et al. Classification of acute pancreatitis--2012: revision of the Atlanta classification and definitions by international consensus. Gut 2013; 62:

28 TABLES Table 1. Definitions of primary and secondary endpoints Table 2. Baseline characteristics Table 3. Primary and secondary endpoints (intention-to-treat analysis) Table 4. Cox proportional hazards model to determine factors associated with the occurrence of the primary endpoint of major complications or death Table 5. Longitudinal Analysis for SF-36 Mental and Physical Component Scores Table 6. Cost analysis FIGURE LEGENDS Figure 1: Diagram of mean number of SIRS criteria present over time post-intervention 27

29 Table 1. Definitions of primary and secondary endpoints (Adapted from van Santvoort et al. 13 with modifications) Endpoint Primary endpoint components (Major complications) New onset multiple organ failure Definitions New onset failure of 2 organs within 48 hours of intervention New onset multiple systemic dysfunction New onset worsening of any pre-existing medical conditions in 2 systems within 48 hours of intervention Organ failure Respiratory failure Cardiovascular failure Renal failure Gastrointestinal failure Hepatic failure Hematological failure Neurological failure Metabolic failure Systemic dysfunction Enterocutaneous fistula Pancreatic-cutaneous fistula Visceral perforation Intraabdominal bleeding Secondary endpoints Incisional hernia PaO 2 < 60mmHg on FiO 2 30% or requiring ventilatory support Systolic blood pressure < 90mmHg following sufficient intravascular resuscitation or administration of inotropes for pressure support Rise in serum creatinine by > three-fold or oliguria (urine output < 0.3mL/Kg/hr) or newly requiring hemofiltration/hemodialysis Blood loss from the gastrointestinal tract > 500mL or ischemia/infarction of the gastrointestinal tract At least three-fold elevation in AST/ALT with elevation in INR (> 1.5) and encephalopathy Disseminated intravascular coagulation defined as prolonged PT/PTT, platelet count < 100,000/μL, schistocytes and/or elevated fibrin degradation products Glasgow coma scale < 13 or encephalopathy Serum calcium < 7.5 mg/dl or metabolic acidosis/alkalosis Worsening of any pre-existing medical condition within 48 hours of intervention Formation of a fistula between the small bowel or colon and skin as indicated by feculent output from a percutaneous catheter or surgical wound site, confirmed on imaging or surgery Formation of a fistula between the pancreas and skin as indicated by amylase-rich fluid output from a percutaneous catheter or surgical wound site, persisting after 3 months, and requiring specific treatment that include parenteral or enteral nutrition, antibiotics, somatostatin analogues and minimally invasive drainage Perforation of a visceral organ requiring interventional procedure Bleeding within the intraabdominal cavity or gastrointestinal tract requiring interventional procedure Herniation of intraabdominal structures through an area of weakness in the abdominal fascia, requiring surgical intervention or persisting at 6-month follow-up Surgical site infection Infection in superficial or deep layers of the abdominal wall at the site of prior intervention, as evident by the presence of purulent drainage or abscess

30 on physical examination, imaging and/or requiring intervention Procedure-related adverse events Any adverse event resulting from endoscopic or surgical intervention, not meeting criteria of a major complication Disease-related adverse events New onset diabetes Pancreatic insufficiency Any adverse event resulting from necrotizing pancreatitis, not meeting criteria of a major complication New onset elevation in fasting plasma glucose 126 mg/dl, 2-hour plasma glucose 200 mg/dl after an oral glucose tolerance test or HbA1c 6.5% Fecal elastase level < 200μg/g in patients not previously taking pancreatic enzyme supplements Systemic Inflammatory Response >1 of the following criteria met: Syndrome (SIRS) 44 1) Body temperature > 38 C or < 36 C 2) Heart rate > 90 beats/minute 3) Respiratory rate > 20 breaths/minute or PaCO 2 < 32mmHg 4) White blood cell count > 12 or < 4 x10 9 /L or > 10% immature neutrophils Clinical improvement Resolution of SIRS/sepsis OR resolution of 1 or more organ failure in patients without SIRS/sepsis, AND > 25% decrease in the size of the necrotic collection on CT at 72 hours post-intervention Abbreviations: ALT, alanine transaminase; AST, aspartate transaminase; CT, computed tomography; FiO2, fraction of inspired oxygen; INR, international normalized ratio; PaO2, partial pressure of oxygen; PT, prothrombin time; PTT, partial thromboplastic time; SIRS, systemic inflammatory response syndrome

31 Table 2. Baseline characteristics Endoscopy Surgery (n=34) (n=32) p-value Age (years): Mean (SD) 55.6 (14.2) 52.9 (14.2) Median (IQR) 59 (46-66) 57.5 ( ) Gender: n (%) Female 12 (35.3) 11 (34.4) Male 22 (64.7) 21 (65.6) Race: n (%) Black 1 (2.9) 3 (9.4) Hispanic 5 (14.7) 6 (18.8) White 28 (82.4) 23 (71.9) Cause of pancreatitis: n (%) Gallstones 14 (41.2) 8 (25.0) Alcohol 6 (17.6) 11 (34.4) Idiopathic 14 (41.2) 11 (34.4) Other a 0 2 (6.3) Coexisting conditions: n (%) Cardiovascular disease 15 (44.1) 16 (50.0) Pulmonary disease 6 (17.6) 3 (9.4) Renal disease 4 (11.8) 1 (3.1) Diabetes mellitus 12 (35.3) 7 (21.9) ASA class: n (%) II 2 (5.9) 1 (3.1) III 26 (76.5) 23 (71.9) IV 6 (17.6) 8 (25.0) CT severity index: n (%) (2.9) 3 (9.4) (97.1) 29 (90.6) Type of necrotic collection: n (%) b Walled-off necrosis 21 (61.8) 23 (71.9) Acute necrotic collection 13 (38.2) 9 (28.1) Size of necrotic collection-ap axis (cm): Mean (SD) 10.0 (4.5) 10.0 (3.3) Median (IQR) 8.7 ( ) 9.2 ( ) Size of necrotic collection-transverse axis (cm): Mean (SD) 11.7 (4.1) 11.8 (3.7) Median (IQR) 11.0 ( ) 11.2 ( ) Percentage of necrosis: (%) Mean (SD) 42.2 (18.1) 40.8 (15.8) Median (IQR) 40 (25-60) 40 (30-55) Collection extending to lower abdomen/pelvis: n (%) 14 (41.2) 12 (37.5) Infected necrosis: n (%) c 31 (91.2) 30 (93.8) Disease severity: n (%) SIRS 16 (47.1) 16 (50.0) ICU/high acuity care 24 (70.6) 21 (65.6) Single-organ failure 2 (5.9) 3 (9.4) Multiple organ failure 7 (20.6) 7 (21.9) Acute physiology score (APS): Mean (SD) 26.9 (25.1) 25.1 (18.9) Median (IQR) 18 (10-37) 20 ( ) APACHE II score Mean (SD) 33.7 (13.5) 27.1 (20.3) Median (IQR) 30 (26-35) 21 (16-23) Time to intervention from onset of pancreatitis (weeks): < 4 weeks 9 (26.5) 7 (21.9) weeks 19 (55.9) 16 (50.0) > 6 weeks 6 (17.6) 9 (28.1) Nutritional support: n (%) Enteral feeding 10 (29.4) 16 (50.0) Parenteral feeding 5 (14.7) 4 (12.5)

32 Oral diet 16 (47.1) 12 (37.5) Nil per os 3 (8.8) 0 Percutaneous catheter in situ prior to intervention: n (%) 14 (41.2) 9 (28.1) Abbreviations: AP, anteroposterior; APACHE II; Acute Physiology And Chronic Health Evaluation II; ASA, American Society of Anesthesiologists; CT, computed tomography; ICU, intensive care unit; IQR, interquartile range; SD, standard deviation; SIRS, systemic inflammatory response syndrome a. Other causes of pancreatitis: etiology: Hypertriglyceridemia (n=1); Medication (n=1). b. Walled-off necrosis is a mature, encapsulated collection of pancreatic and/or peripancreatic necrosis that has developed a welldefined inflammatory wall. Acute necrotic collection is a collection containing variable amounts of both fluid and necrosis associated with necrotizing pancreatitis that is not fully encapsulated. 45 c. Infected necrosis was defined as the presence of gas on cross-sectional imaging and/or positive culture of necrotic tissue (positive culture n=49: Escherichia coli in 6, Enterobacter aerogenes/cloacae in 3, Enterococcus faecalis in 9, Klebsiella pneumoniae in 10, Streptococcus/Staphylococcus species in 5, candida albicans/tropicalis in 12 and others in 4). 14 patients had both gas on imaging and positive culture. The remaining patients (n=5) were suspected to have infected necrosis based on persistent sepsis or ongoing clinical deterioration.

33 Table 3. Primary and secondary endpoints Endoscopy Surgery Risk ratio (n=34) (n=32) (95% CI) p-value Primary composite endpoint: n (%) a 4 (11.8) 13 (40.6) 0.29 ( ) Primary endpoint components: Death: n (%) 3 (8.8) 2 (6.3) 1.41 ( ) New onset multiple organ failure: n (%) 2 (5.9) 3 (9.4) 0.63 ( ) New onset multiple systemic dysfunction: n (%) 0 1 (3.1) Enteral-Pancreatic cutaneous fistula: n (%) 0 9 (28.1) Visceral perforation: n (%) Intraabdominal bleeding: n (%) 0 3 (9.4) No. of positive primary endpoints: Mean (SD) 0.15 (0.44) 0.69 (1.03) Median (IQR) 0 (0-0) 0 (0-1) Range Secondary endpoints: Procedure duration (mins): Mean (SD) 53.5 (34.0) (37.2) Median (IQR) 47.5 (22-80) 120 (90-135) <0.001 Range Other adverse events: Incisional hernia: n (%) 0 2 (6.3) b Surgical site infection: n (%) 0 2 (6.3) Procedure-related adverse events: n (%) 4 (11.8) 6 (18.8) 0.63 ( ) Disease-related adverse events: n (%) a 2 (5.9) 14 (43.8) 1.67 ( ) <0.001 Post-procedural SIRS: SIRS at 72 hours post-index procedure: n (%) 7 (20.6) 21 (65.6) 0.31 ( ) <0.001 No. of SIRS at 72 hours post-index procedure: Mean (SD) 0.71 (0.97) 2.03 (1.33) Median (IQR) 0 (0-1) 2 (1-3) <0.001 Range New onset SIRS at 72 hours post-index procedure: n (%) c 1 (5.6) 9 (56.3) 0.10 ( ) Resolution of SIRS at 72 hours post-index procedure: n (%) d 11 (68.8) 3 (18.8) 3.67 ( ) Health care resource utilization: Reinterventions due to lack of clinical improvement: n (%) 15 (44.1) 12 (37.5) 1.18 ( ) Total no. of interventions per patient: e Mean (SD) 1.7 (1.0) 1.5 (0.9) Median (IQR) 1 (1-2) 1 (1-2) Range Total no. of interventions per study group: Necrosectomy performed: n (%) 11 (32.4) 31 (96.9) f 0.33 ( ) <0.001 No. of necrosectomy procedures: 0 23 (67.6) 1 (3.1) < (23.5) 23 (71.9) 2 2 (5.9) 6 (18.8) 3 1 (2.9) 2 (6.3) No. of endoscopic drainage procedures: (84.4) < (70.6) 5 (15.6) g 2 8 (23.5) (5.9) 0 Post-intervention percutaneous catheter placement: n (%) 6 (17.6) 5 (15.6) 1.13 ( ) Length of hospital stay (days): h Mean (SD) 16.5 (12.2) 23.3 (17.5) Median (IQR) 14 (6-22) 18.5 ( ) 0.057

34 Range Length of ICU stay (days): h Mean (SD) 2.4 (7.1) 6.6 (12.1) Median (IQR) 0 (0-0) 0 (0-6.5) Range Readmissions: n (%) i 11 (32.4) 16 (50.0) 0.65 ( ) Total no. of readmissions: i Mean (SD) 0.47 (0.83) 1.47 (2.17) Median (IQR) 0 (0-1) 0.5 (0-2) Range Endocrine and exocrine function: New onset diabetes: n (%) j 6 (27.3) 9 (36.0) 0.76 ( ) New diagnosis of pancreatic insufficiency: n (%) 29 (85.3) 28 (87.5) 0.97 ( ) Fecal elastase level (μg/g): Mean (SD) (96.1) 88.1 (96.2) Median (IQR) 62 (30-128) 56.5 ( ) Range Mean total costs (US $): 75, , Quality of life at 3-month follow-up: MCS: β (95% CI) -0.22( ) PCS: β (95% CI) 5.29 ( ) Abbreviations: ICU, Intensive Care Unit; MCS, Mental Component Score; PCS, Physical Component Score; QOL, quality of life; SIRS, Systemic Inflammatory Response Syndrome a. Multiple events in a single patient were counted as a single event for comparison. b. Both patients developed small bowel obstruction from incisional hernia and were treated surgically. c. Denominator: n=18 for endoscopy, n=16 for surgery. d. Denominator: n=16 for endoscopy, n=16 for surgery. e. Only endoscopic and surgical treatments were counted as interventions. f. 1 patient had predominantly liquefied necrosis and did not require necrosectomy. g. 5 patients in the surgical cohort were crossed over to endoscopy. h. From index intervention to discharge. i. From discharge to 6-month follow-up. j. Denominator: n=22 for endoscopy, n=25 for surgery.

35 Table 4. Cox proportional hazards model to determine factors associated with the occurrence of the primary endpoint of major complications or death Predictor variable Hazards ratio (95% CI) p-value Treatment arm: Surgery vs. Endoscopy 4.38 ( ) Age (years) 1.01 ( ) Gender: Female vs. Male 1.05 ( ) Acute physiology score 1.03 ( ) Time to intervention from onset of pancreatitis (weeks) 1.10 ( ) Size of necrotic collection (cm) 1.03 ( ) Percentage of necrosis (%) 1.00 ( ) 0.839

36 Table 5. Longitudinal Analysis for SF-36 Mental and Physical Component Scores MCS a PCS a Factor β (95% CI) p-value β (95% CI) p-value Intercept (25.06 to 32.06) < (24.56 to 29.22) < Treatment group b 2.01 (-3.41 to 7.43) (-4.63 to 2.62) Time c Discharge 7.03 (2.16 to 11.89) (-3.45 to 2.88) month follow-up (9.01 to 22.36) < (6.37 to 12.68) < month follow-up (22.24 to 31.83) < (14.30 to 24.48) < Time treatment interaction Treatment x discharge -0.08(-6.26 to 6.10) (-0.68 to 7.30) Treatment x 3 months -0.22(-9.18 to 8.74) (0.27 to 10.31) Treatment x 6 months -2.37(-9.34 to 4.61) (-1.97 to 11.01) a. A high MCS and PCS represents a good quality of life. b. The reference group for treatment is surgery. c. The reference group for time is pre-discharge.

37 Table 6. Cost analysis Cost Endoscopy (US$) Surgery (US$) p-value Hospital 21, , Procedure 11, , Pharmacy 14, , Radiology 8, , Anesthesia 2, , Laboratory 6, , Other 10, , Total 75, ,

38