Performance of the revised Atlanta and determinant-based classifications for severity in acute pancreatitis

Transcription

1 Original article Performance of the revised Atlanta and determinant-based classifications for severity in acute pancreatitis S. S. Bansal 1, J. Hodson 2, R. S. Sutcliffe 1, R. Marudanayagam 1, P. Muiesan 1, D. F. Mirza 1, J. Isaac 1 and K. J. Roberts 1 Departments of 1 Pancreatic Surgery and 2 Statistics, University Hospitals Birmingham NHS Trust, Birmingham, UK Correspondence to: Mr K. J. Roberts, Department of Pancreatic Surgery, University Hospitals Birmingham NHS Trust, Birmingham B15 2TH, UK ( j.k.roberts@bham.ac.uk) Background: Severity classification systems aim to stratify patients with acute pancreatitis reliably into coherent risk groups. Recently, the Atlanta 1992 classification has been revised (Atlanta 212) and a novel determinant-based classification (DBC) system developed. This study assessed the ability of the three systems to stratify disease severity among patients with acute pancreatitis. Methods: This was an observational cohort study of patients with acute pancreatitis identified from an institutional database. Cohort characteristics, investigations, interventions and outcomes were identified. Systems were compared using receiver operating characteristic (ROC) analysis and Spearman s correlation coefficients. Results: The in-hospital mortality rate was 6 6 per cent (15 of 228 patients). All of the outcomes considered correlated significantly with the three systems, with the exception of the need for surgery in Atlanta Atlanta 212 and the DBC had higher area under the curve (AUC) values than Atlanta 1992 for all outcomes. The revised Atlanta and DBC systems both performed similarly with regard to ICU admission (AUC 927 and 917 respectively; both P < 1), need for percutaneous drainage (AUC 879 and 891; both P < 1), need for surgery (AUC 827 and 845; P = 6 and P = 4 respectively) and in-hospital mortality ( 955 and 931; both P < 1). However, the critical category in the DBC system identified patients with the most severe disease; seven of eight patients in this group died in hospital, compared with 15 of 34 with severe pancreatitis according to Atlanta 212. Conclusion: The Atlanta 212 and DBC perform equally well for classification of disease severity in acute pancreatitis. The addition of a critical category in the DBC identifies patients with the most severe disease. Paper accepted 2 November 215 Published online 25 January 216 in Wiley Online Library ( DOI: 1.12/bjs.188 Introduction Acute pancreatitis is a common disease and a leading cause of hospital admission among patients with abdominal pain 1,2. It presents as a wide spectrum of disease, from a short self-limiting episode to multiple organ failure and death 3,4. Stratifying patients risk of serious complications allows early identification of those to concentrate resources on and monitor closely, and provides prognostic information. For comparison between cohorts, it is essential that these classification systems yield reproducible results between institutions. The original Atlanta classification 5 defined patients with mild or severe acute pancreatitis, but has poor reliability 6 8, prompting two recent alternative classifications the revised Atlanta classification in 212 (Atlanta 212) 9 and the determinant-based classification (DBC) 1. Both new systems recognize an increasing number of categories of severity (3 and 4 for Atlanta 212 and DBC respectively), with the aim of further stratifying patients at highest risk of adverse outcome. Both classifications were derived following discussion among international groups of experts using a consensus building approach. The main difference between Atlanta 1992 and Atlanta 212 is that the duration of organ failure is taken into account in the latter system; transient organ failure lasting less than 48 h is the hallmark of moderately severe acute pancreatitis, whereas persistent organ failure lasting for 48 h or more defines severe acute pancreatitis. Although all three systems recognize local complications, the DBC goes further to define pancreatic necrosis as the local determinant of an adverse outcome, and similarly 216 BJS Society Ltd BJS 216; 13:

2 428 S. S. Bansal, J. Hodson, R. S. Sutcliffe, R. Marudanayagam, P. Muiesan, D. F. Mirza et al. recognizes transient or persistent organ failure. The DBC defines moderate acute pancreatitis as sterile necrosis and/or transient organ failure, severe acute pancreatitis as infected necrosis or persistent organ failure, and critical acute pancreatitis as both infected necrosis and persistent organ failure. The performance of these systems has been assessed recently and it appears that both Atlanta 212 and DBC outperform Atlanta 1992, with little difference between the two more recent systems 11,12. In a North American study 12, the DBC had better ability to demonstrate the need for interventions, whereas Atlanta 212 stratified patients more effectively by duration of hospital stay. In a cohort of Spanish patients there were no significant differences between the Atlanta 212 and DBC 11 ; the value of the critical category of severity within the DBC was questioned as it was applied to just 6 per cent of episodes of acute pancreatitis. Validation studies are required to identify the optimal classification as performance may vary between measured outcomes, population cohorts or institutions. Defining disease severity and complications of acute pancreatitis facilitates rigorous review of the efficacy of interventions among patients with potentially diverse clinical outcomes The aim of this study was to compare the ability of these systems to stratify disease severity with regard to important clinical outcomes in patients with acute pancreatitis in a tertiary institution. Methods The study was conducted and the manuscript prepared in line with guidance set out in the STROBE statement 16. Data collection This was an analysis of consecutive patients with a diagnosis of acute pancreatitis treated by the pancreatic surgical team at the Queen Elizabeth Hospital, Birmingham, UK, between January 21 and July 214. No patients were excluded. This surgical service is provided by six dedicated hepatobiliary and pancreatic surgeons, and provides support to referring hospitals with a population exceeding 4 million inhabitants. Interventional radiology support is available at all hours. Patients were identified from a prospectively maintained departmental database. Acute pancreatitis was defined using established criteria as two of the following: serum amylase level at least three times the upper limit of normal, abdominal pain in keeping with acute pancreatitis, or CT/MRI images in keeping with acute pancreatitis 9.Any missing data were obtained from electronic records, case Table 1 Classification systems for severity of acute pancreatitis Mild Moderate* Severe Critical Atlanta 1992 Local complications No Yes and/or Organ failure No Yes and/or APACHE II 8 No Yes Atlanta 212 Local complications No Yes and/or Organ failure No Transient Persistent Determinant-based classification Peripancreatic No Sterile Infected Infected necrosis and and/or or and Organ failure No Transient Persistent Persistent *Termed moderately severe in Atlanta 212 and moderate in determinant-based classification. Acute fluid collection, pancreatic necrosis, pseudocyst, pancreatic abscess. Cardiovascular: systolic BP below 9 mmhg; respiratory: arterial partial pressure of oxygen 6 mmhg or less; renal: creatinine than greater than 2 mg/dl; gastrointestinal: bleeding exceeding 5 ml per 24 h. Acute fluid collection, pseudocyst, acute necrotic collection, walled-off necrosis. APACHE, Acute Physiology And Chronic Health Evaluation. notes or discussion with the patient s general practitioner. The severity of acute pancreatitis was defined by the three systems (Table 1). Outcomes The systems were compared for their ability to stratify patients in accordance with admission to the intensive care unit (ICU), need for surgical treatment of pancreatic necrosis or surgical complications, need for percutaneous drainage of pancreatic or peripancreatic necrosis, duration of ICU stay, overall duration of hospital stay, and death during the acute hospital admission. Statistical analysis Cohort characteristics were compared using non-parametric tests (Mann Whitney U test for continuous data and Fisher s exact test for categorical data). The performance of the systems was assessed using receiver operating characteristic (ROC) curve analysis for categorical data and by Spearman s ρ correlation coefficient for continuous data. The areas under the ROC curves (AUROCs) were then compared between the systems, using the method described by Hanley and McNeil 17 ;this paired analysis accounted for the fact that the ROC curves were derived from the same set of patients. The degree of correlation between the curves was calculated by splitting 216 BJS Society Ltd BJS 216; 13:

3 Comparison of performance of acute pancreatitis classifications 429 Table 2 Characteristics of the cohort Table 3 Prediction of outcomes by classification system No. of patients* (n = 228) Pancreatitis Oedematous 17 (74 6) Necrotizing 58 (25 4) Infected necrosis 18 (7 9) Local complications 89 (39 ) Acute fluid collection 28 (12 3) Pseudocyst 11 (4 8) Acute necrotic collection 24 (1 5) Walled-off necrosis 26 (11 4) Organ failure Single 9 (3 9) Multiple 28 (12 3) Persistent 3 (13 2) Transient 7 (3 1) Aetiology Gallstones 138 (6 5) Alcoholic 6 (26 3) Idiopathic 19 (8 3) ERCP 7 (3 1) Other 4 (1 8) Outcomes ICU admission 39 (17 1) Duration of ICU stay (days) 11 (4 36) Duration of hospital stay (days) 9 (5 17) In-hospital death 15 (6 6) *With percentages in parentheses unless indicated otherwise; values are median (i.q.r.). ERCP, endoscopic retrograde cholangiopancreatography; ICU, intensive care unit. the patients into those with and without the outcome being considered, and calculating correlation coefficients between the systems within these subgroups. The average correlation for the two systems being compared was then used to adjust the standard error, in order to account for the paired nature of the analysis. A z-score was calculated, and the associated P value reported. Comparison of the correlation coefficients for the continuous outcomes was performed using the method described by Meng and colleagues 18. This compared the correlation coefficients between two systems and the outcome, while accounting for the paired nature of these statistics by using the correlation between the systems themselves. The P value was based on a χ 2 test of the heterogeneity of the correlation matrix of these three coefficients. All tests were two-tailed and P < 5 was considered statistically significant. A medical statistician provided advice on the study design and performed the analyses using SPSS version 22 (IBM, Armonk, New York, USA). Results A total of 228 patients were identified, with a median age of 56 (i.q.r ) years; there were 118 men (51 8 per cent). Atlanta 212 DBC Atlanta 1992 ICU admission AUROC P (versus 5)* < 1 < 1 < 1 P (versus Atlanta 212) Surgery for necrosis AUROC P (versus 5)* P (versus Atlanta 212) Percutaneous drainage AUROC P (versus 5)* < 1 < 1 < 1 P (versus Atlanta 212) 72 n.c. In-hospital death AUROC P (versus 5)* < 1 < 1 < 1 P (versus Atlanta 212) n.c. n.c. Duration of ICU stay (days) Spearman s ρ P (versus )* < 1 < 1 < 1 P (versus Atlanta 212) 228 < 1 Duration of hospital stay (days) Spearman s ρ P (versus )* < 1 < 1 < 1 P (versus Atlanta 212) 77 1 *These P values test whether there is any significant association between the classification system and the outcome being considered. The comparisons against Atlanta 212 were performed using the method proposed by Hanley and McNeil 17 for receiver operating characteristic (ROC) curves; the method of Meng and colleagues 18 was used to compare correlations. DBS, determinant-based classification; ICU, intensive care unit; AUROC, area under the ROC curve; n.c., not calculable, because one of the variables for one of the scores had the same value for each patient so correlation could not be performed. Some 58 patients (25 4 per cent) developed necrotizing acute pancreatitis, and 37 (16 2 per cent) developed organ failure, which was transient in seven (3 1 per cent) and persistent in 3 (13 2 per cent) (Table 2). Fifteen patients (6 6 per cent) died during the acute hospital admission. At last follow-up a further six patients had died, none from acute pancreatitis or its complications. According to the Atlanta 1992 system, 127 patients (55 7 per cent) were considered to have mild and 11 (44 3 per cent) severe acute pancreatitis. Based on Atlanta 212, 129 (56 6 per cent), 65 (28 5 per cent) and 34 (14 9 per cent) patients had mild, moderately severe and severe acute pancreatitis respectively. In the DBC, 155 (68 per cent), 37 (16 2 per cent), 28 (12 3 per cent) andeight (3 5 per cent) patients were classified as having mild, moderate, severe and critical acute pancreatitis respectively. Some 34 patients (14 9 per cent) were transferred to the authors unit from other hospitals and the remaining patients were admitted directly from the emergency department. The severity of acute pancreatitis classified 216 BJS Society Ltd BJS 216; 13:

4 43 S. S. Bansal, J. Hodson, R. S. Sutcliffe, R. Marudanayagam, P. Muiesan, D. F. Mirza et al. 1 8 Atlanta 212 DBC Atlanta Mild Moderate Severe Critical a ICU admission 1 b Surgery for necrosis c Percutaneous drainage d In-hospital mortality 15 8 Hospital stay (days) 1 5 ICU stay (days) e Duration of hospital stay f Duration of ICU stay Fig. 1 Key outcomes stratified by disease severity for the three classification systems: a intensive care unit (ICU) admission, b surgery for treatment of necrotizing pancreatitis or its complications, c percutaneous drainage, d in-hospital mortality, e duration of hospital stay and f duration of ICU stay. Median values (bold horizontal line), i.q.r. (box), and range (error bars) excluding outliers ( ) are shown in e and f. DBC, determinant-based classification according to the Atlanta 212 system was associated with the likelihood of transfer: two (1 6 per cent)of 129patients with mild acute pancreatitis, 18 (28 per cent) of 65 with moderately severe acute pancreatitis and 14 (41 per cent) of 34 with severe acute pancreatitis were transferred to the unit (P < 1). The severity of acute pancreatitis as classified by the DBC system was also associated with the likelihood of transfer: eight (5 2 per cent) of 155 patients with mild acute pancreatitis, 12 (32 per cent) of 37 with moderate acute pancreatitis, nine (32 per cent) of 28 with severe 216 BJS Society Ltd BJS 216; 13:

5 Comparison of performance of acute pancreatitis classifications 431 acute pancreatitis and five (63 per cent) of eight with critical acute pancreatitis were transferred to the unit (P < 1). Comparison between classification systems All three classification systems correlated significantly with all of the outcomes considered, with the exception of Atlanta 1992 and the need for open surgery (P = 92); details are provided in Table 3. The Atlanta 212 classification was considered as the standard, as it was developed from the original Atlanta system, and was therefore compared with the other two systems (Table 3). Atlanta 1992 performed significantly worse than Atlanta 212 in ability to stratify patients by the need for ICU admission, need for surgery, and duration of ICU and hospital stay. It was not possible to compare the AUROC for percutaneous drainage or in-hospital death because all patients in whom these outcomes occurred were in the same risk category for each classification system, making correlations incalculable. However, the AUROCs were lower for the Atlanta 1992 system. Comparisons between the DBC and Atlanta 212 revealed no significant differences for any of the outcomes considered. The comparison was not possible for in-hospital deaths, but the AUROCs for the two systems were similar. A review of the proportion of patients who required ICU admission, open surgery, percutaneous drainage and who died in hospital, together with assessments of the duration of ICU and hospital stay, is shown for each system in Fig. 1 and Table S1 (supporting information). The critical category in the DBC seemed to identify patients with the highest risk of adverse outcome. There was no difference when the DBC critical and Atlanta 212 severe categories were compared with respect to the need for ICU admission (7 of 8 and 3 of 34 patients respectively), but the need for necrosectomy (2 of 8 versus 5 of 34) and percutaneous drainage (7 of 8 versus 22 of 34), in-hospital mortality (7 of 8 versus 15 of 34), length of ICU stay (median 3 versus 14 days) and duration of hospital stay (median 75 versus 62 days) were all greater among patients identified as critical by the DBC as opposed to having severe disease according to Atlanta 212. Of six patients with infected necrosis and no persistent organ failure (classified as severe pancreatitis by the DBC), five were considered to have moderately severe and one mild pancreatitis by Atlanta 212. Patients with infected necrosis were reviewed separately. Pancreatitis in eight and ten patients was classified as moderately severe and severe respectively by Atlanta 212. In the moderately severe group, no patient developed organ failure, there were no deaths and no patient required ICU admission; the median duration of hospital stay was 29 days. In the severe group, all had persistent organ failure and nine died during the hospital admission; the median length of hospital stay was 75 days. Discussion Disease severity systems are important to understand the efficacy of established and novel therapies among patients with homogeneous pathology. This is particularly important in patients with acute pancreatitis given the wide variation in outcomes as reported in recent publications The two new classification systems (Atlanta and DBC 1 ) have significantly greater ability to stratify patients by disease severity than the original Atlanta system. The revised Atlanta system provides updated terminology to define pathological and radiological features of acute pancreatitis and its local complications accurately. The addition of the fourth critical category in the DBC identifies patients with an extreme risk of death and need for intervention. The original Atlanta score was revised 9 to reflect evolution in understanding of the pathology of the disease 19 21, with descriptions of local complications on CT, and recognition that the duration of and severity of organ failure strongly affected outcomes The DBC similarly places emphasis on both local and systemic determinants of disease severity 1. However, it goes further to stratify patients specifically by the presence of necrosis and whether this is infected, following a meta-analysis 25 demonstrating the lethality of persistent organ failure in the setting of infected necrosis. In combination with the duration of organ failure, this system yields the greatest number of categories of severity four. The main finding of the study is that both new classification systems outperform the original Atlanta definition for most outcomes reviewed. There was no significant difference between the performance of Atlanta 212 or the DBC with regard to the outcomes measured. However, addition of the critical category within the DBC identified patients at highest risk of adverse outcomes. Among patients in the critical DBC category, the in-hospital mortality rate was doubled, ICU stay was more than twice as long, and the need for surgery or percutaneous drainage was greater than in patients with severe pancreatitis according to the Atlanta 212 system. In a Spanish cohort 11, only three of 543 episodes of acute pancreatitis were defined as critical using the DBC and the benefit of a category that applied to so few patients was questioned. The authors view was that their study 216 BJS Society Ltd BJS 216; 13:

6 432 S. S. Bansal, J. Hodson, R. S. Sutcliffe, R. Marudanayagam, P. Muiesan, D. F. Mirza et al. included all patients presenting with acute pancreatitis and, as they did not accept transfer of patients with complicated acute pancreatitis, those with mild acute pancreatitis overwhelmingly predominated. In contrast, the present cohort comprised 14 9 per cent of patients who were transferred to the unit. Transferred patients tend to have more severe acute pancreatitis and consequently skew the data. Two other published cohorts 12,26 had higher proportions of patients classified as critical by the DBC. These three groups reflect specialist pancreatic surgery teams accepting patients with complicated acute pancreatitis from within and outwith their own institutions. Among the Indian 26, American 12 and present cohorts some 5, 7 and 3 5 per cent of patients respectively were categorized as critical, compared with 5 per cent in the Spanish study. It has been suggested that an increasing number of categories of severity is unhelpful as the number of patients in clinical subgroups needs to be greater to observe significant effects 11. However, in clinical practice identification of patients classified as critical according to the DBC has definite implications for prognosis, and may influence management by ICU and surgical teams. There are implications for primary and secondary care systems, however, because healthcare providers working in these environments rarely manage patients with critically severe pancreatitis as classified by the DBC. Creation of the DBC and Atlanta 212 systems differed significantly. Although both employed international web-based consultations of experts in pancreatology, the DBC process began with a review of published evidence 27. A systematic approach to identifying determinants was taken, and the system was discussed through an international, multidisciplinary forum 1. Conversely, the Atlanta 212 guidelines were developed through a series of web-based, iterative edits by interested parties over a 7-year interval. Both groups finally reviewed their proposed systems with further discussion among experts at an international conference (DBC) and through further web-based dialogue (Atlanta ). The great emphasis placed on the occurrence of infected necrosis in DBC may occasionally exaggerate disease severity compared with the Atlanta 212 system, which includes an assessment of organ failure. This is evident from the different outcomes for patients with infected necrosis, stratified as moderately severe or severe by the Atlanta 212 system. There were no deaths among those in the moderately severe group, but nine of ten patients with disease classified as severe died during the hospital admission. Limitations of the study include the small number of patients with critical or severe acute pancreatitis, and that this study reflects practices and outcomes in a single institution. A multicentre approach would provide more patients for review and reduce potential confounding of results owing to differing practices of managing various aspects of acute pancreatitis, such as techniques for draining infected necrosis. Disclosure The authors declare no conflict of interest. References 1 Johnson CD, Besselink MG, Carter R. Acute pancreatitis. BMJ 214; 349: g Peery AF, Dellon ES, Lund J, Crockett SD, McGowan CE, Bulsiewicz WJ et al. Burden of gastrointestinal disease in the United States: 212 update. Gastroenterology 212; 143: Working Party of the British Society of Gastroenterology; Association of Surgeons of Great Britain and Ireland; Pancreatic Society of Great Britain and Ireland; Association of Upper GI Surgeons of Great Britain and Ireland. UK guidelines for the management of acute pancreatitis. Gut 25; 54(Suppl 3): iii1 iii9. 4 Whitcomb DC. Clinical practice. Acute pancreatitis. NEngl JMed26; 354: Bradley EL III. A clinically based classification system for acute pancreatitis. Summary of the International Symposium on Acute Pancreatitis, Atlanta, Ga, September 11 through 13, Arch Surg 1993; 128: Banks PA, Freeman ML; Practice Parameters Committee of the American College of Gastroenterology. Practice guidelines in acute pancreatitis. Am J Gastroenterology 26; 11: Pandol SJ, Saluja AK, Imrie CW, Banks PA. Acute pancreatitis: bench to the bedside. Gastroenterology 27; 132: Vege SS, Chari ST. Organ failure as an indicator of severity of acute pancreatitis: time to revisit the Atlanta classification. Gastroenterology 25; 128: Banks PA, Bollen TL, Dervenis C, Gooszen HG, Johnson CD, Sarr MG et al. Classification of acute pancreatitis 212: revision of the Atlanta classification and definitions by international consensus. Gut 213; 62: Dellinger EP, Forsmark CE, Layer P, Levy P, Maravi-Poma E, Petrov MS et al. Determinant-based classification of acute pancreatitis severity: an international multidisciplinary consultation. Ann Surg 212; 256: Acevedo-Piedra NG, Moya-Hoyo N, Rey-Riveiro M, Gil S, SempereL,MartinezJet al. Validation of the determinant-based classification and revision of the Atlanta classification systems for acute pancreatitis. Clin Gastroenterol Hepatol 214; 12: BJS Society Ltd BJS 216; 13:

7 Comparison of performance of acute pancreatitis classifications Nawaz H, Mounzer R, Yadav D, Yabes JG, Slivka A, Whitcomb DC et al. Revised Atlanta and determinant-based classification: application in a prospective cohort of acute pancreatitis patients. Am J Gastroenterol 213; 18: Bang JY, Holt BA, Hawes RH, Hasan MK, Arnoletti JP, Christein JD et al. Outcomes after implementing a tailored endoscopic step-up approach to walled-off necrosis in acute pancreatitis. Br J Surg 214; 11: da Costa DW, Boerma D, van Santvoort HC, Horvath KD, Werner J, Carter CR et al. Staged multidisciplinary step-up management for necrotizing pancreatitis. Br J Surg 214; 11: e65 e Sarathi PP, Das K, Bhattacharyya A, Ray S, Hembram J, Sanyal S et al. Natural resolution or intervention for fluid collections in acute severe pancreatitis. Br J Surg 214; 11: von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ 27; 335: Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983; 148: Meng XL, Rosenthal R, Rubin DB. Comparing correlated correlation coefficients. Psychol Bull 1992; 111: Bakker OJ, van Santvoort H, Besselink MG, Boermeester MA, van Eijck C, Dejong K et al. Extrapancreatic necrosis without pancreatic parenchymal necrosis: a separate entity in necrotising pancreatitis? Gut 213; 62: Sakorafas GH, Tsiotos GG, Sarr MG. Extrapancreatic necrotizing pancreatitis with viable pancreas: a previously under-appreciated entity. J Am Coll Surg 1999; 188: Singh VK, Bollen TL, Wu BU, Repas K, Maurer R, Yu S et al. An assessment of the severity of interstitial pancreatitis. Clin Gastroenterol Hepatol 211; 9: Buter A, Imrie CW, Carter CR, Evans S, McKay CJ. Dynamic nature of early organ dysfunction determines outcome in acute pancreatitis. Br J Surg 22; 89: Halonen KI, Pettilä V, Leppäniemi AK, Kemppainen EA, Puolakkainen PA, Haapiainen RK. Multiple organ dysfunction associated with severe acute pancreatitis. Crit Care Med 22; 3: Johnson CD, Abu-Hilal M. Persistent organ failure during the first week as a marker of fatal outcome in acute pancreatitis. Gut 24; 53: Petrov MS, Shanbhag S, Chakraborty M, Phillips AR, Windsor JA. Organ failure and infection of pancreatic necrosis as determinants of mortality in patients with acute pancreatitis. Gastroenterology 21; 139: Thandassery RB, Yadav TD, Dutta U, Appasani S, Singh K, Kochhar R. Prospective validation of 4-category classification of acute pancreatitis severity. Pancreas 213; 42: Petrov MS, Windsor JA. Classification of the severity of acute pancreatitis: how many categories make sense? Am J Gastroenterol 21; 15: Supporting information Additional supporting information may be found in the online version of this article: Table S1 Summary of key outcomes stratified by disease severity for the three classification systems (Word document) 216 BJS Society Ltd BJS 216; 13: