Toward improving outcomes of ERCP There are few medical procedures with as much variability in outcome as ERCP. An ambulatory patient may leave a therapeutic ERCP with successful resolution of his or her presenting problem, no abdominal discomfort, be discharged home in 2 or 3 hours, eat dinner that night, and return to work the next day. Under different circumstances, the same patient might develop a fulminant complication such as pancreatitis which requires extended intensive care and surgery and may eventually result in permanent disability or even death. The potential impact of ERCP complications is great: If 500,000 ERCPs are performed annually in the United States with a complication rate of 10% and a resultant mortality rate of 0.5%, then ERCP accounts for 50,000 complications and 2500 deaths annually. How, then, do we know which patients, procedures, and operator characteristics contribute to the likelihood of one of these complications? And what can be done to reduce their frequency? Advances in our understanding of ERCP complications have recently occurred in several major areas: first, standardized outcome-based definitions have provided a uniform yardstick and are now widely used. 1,2 Second, large-scale multicenter cohort studies, using multivariate analyses, have allowed clearer identification of patient, technique, and operator-related risk factors for complications; these in turn can lead to better understanding of patterns and mechanisms of injury. Finally, there is increasing recognition of the entire spectrum of negative (as well as positive) outcomes beyond complications, including technical failures (which usually require repeat or alternative procedures), ineffectiveness of the procedure in resolving the presenting complaint, costs, extended hospitalization, missed work or usual activities, patient (dis)satisfaction, and long-term sequelae. 3 ,4 Dr. Loperfido and his colleagues are to be congratulated for shedding some light on these areas in a landmark study of complications of ERCP, published in this issue of Gastrointestinal Endoscopy.5 They report a prospective, multicenter study of complications of 3356 consecutive ERCP performed on Copyright © 1998 by the American Society for Gastrointestinal Endoscopy 0016-5107/98/$5.00 + 0 37/70/89932 96
GASTROINTESTINAL ENDOSCOPY
2769 patients at nine centers in Northeast Italy, including two university-affiliated centers and a number of private centers of varying activity. This paper is important for several reasons. It provides much-needed information on outcomes ofERCP outside tertiary-referral centers and thus examines the effectiveness, rather the efficacy, of the procedure. It examines the effect of endoscopists' experience on procedural outcomes. This study also helps to define some of the risk factors for individual and overall complications. In these regards, the study by Loperfido et a1. 5 is similar to ours 6 and a number of other recent studies of complications of ERCP and sphincterotomy. I will attempt to highlight the similarities and differences in the design and findings of these and other related studies. ENDOSCOPIST'S EXPERIENCE: DOES IT MAKE A DIFFERENCE?
The effect of endoscopic expertise on outcome of ERCP has been difficult to evaluate. Many suspect that results are not as good in community practice as in specialized centers. Few data have been available to evaluate this question, as there is little motivation for busy private practitioners, especially those without a special interest in ERCP, to conduct studies. Even if available, simple comparisons of complication rates of ERCP among centers can be misleading because the case mix, intended diagnostic and therapeutic goals, and success rates at achieving those goals vary widely. The major finding of the Italian study was that overall complications were closely associated with low case volume on the part of the endoscopists. Performance at a small center (defined as less than 200 ERCP per year) was one of only two variables that independently predicted complications in a multivariate analysis; low case volume was also independently associated with hemorrhage, cholangitis, and (by univariate analysis) with pancreatitis. The magnitude of the differences in outcome were significant and striking: at lower volume centers, overall complications were more than three times more common than in high-volume centers (7.1 versus 2.0%), as were procedure-related deaths (0.75% versus 0.18 %). Similarly, in our American studY,6 endoscopists who performed not more than one sphincterotomy per week had higher rates of overall complications than more experienced endoscopists (11.1 % versus 8.4%) and particularly of severe complications (2.3% versus 0.9%). In the Italian study,5 multivariate analysis suggested that experience was an independent predictor of complications, but our American study6 found case volume to be significant by univariate analysis VOLUME 48, NO. 1, 1998
M Freeman
Editorials
only. How do we explain the apparent difference? The Italian study examined relatively few technical details related to the ERCP, which probably are the final common pathway of injury to the patient. In contrast, our study included numerous technical details. Ultimately, however, it is the endoscopist driving the endoscope who influences those events. When we ran a multivariate model using only information available before ERCP and excluding procedural details related to cannulation difficulty, we also found that lower endoscopist procedure volume was one of only three variables (along sphincter of Oddi dysfunction and presence of cirrhosis) that predicted complications of sphincterotomy. Thus, both studies support the concept that the endoscopist's expertise seems to have an independent and pivotal role in determining complications. The available data probably underestimate the influence of operator experience on complications. Participants in multicenter outcome studies are voluntary, representing a minority of centers who were approached, and thus still may not represent the true spectrum of outcomes. It is quite possible that endoscopists who are willing to take the time to enter studies, and are willing to have their outcomes objectively scrutinized by outside investigators, have better outcomes than those who do not. Also, because high-volume endoscopists generally attempt more high-risk cases, and more importantly, seldom abandon attempts at therapeutics because of initially unsuccessful cannulation, they may achieve more successful results at the expense of somewhat more complications; in fact, Davis et al. 7 found that the principal difference in outcome of sphincterotomy performed by highly experienced endoscopists was in higher success rates, with only a trend toward lower complication rates. In the Italian study,5 the higher complication rates for low-volume endoscopists occurred in spite (or perhaps because of) a significantly lower success rate at ductal cannulation and greater need for repeat ERCP; this is a very important observation that was notably absent from ours and other studies addressing only sphincterotomy. The inclusion of all attempted ERCPs, not just therapeutic cases, is one ofthe very important features of the Loperfido study and probably contributes to their ability to clearly demonstrate for the first time a dramatic impact of endoscopic expertise on outcomes of ERCP. PRECUT SPHINCTEROTOMY
There are few issues in therapeutic endoscopy as controversial as use of precut, or "access papillotomy," to gain entry into the bile duct. Principle techVOLUME 48, NO. 1, 1998
niques include needle-knife sphincterotomy,S with some advocating prior insertion of a pancreatic stent9 and others favoring the Erlangen pull-type papillotome. 10 Opinions on the appropriate use of precut techniques vary as widely as the frequency with which it is used; in tertiary centers, some avoid precutting almost entirely, whereas others use it for as many as 40% of all sphincterotomies and even for diagnostic ERCP.l0 In a spectrum of practices, the Italian multicenter study5 reported use of precut techniques for fully 26.5% (419 of 1583) biliary sphincterotomies, whereas in our American multicenter study,6 precutting was used in only 4.7% (111 of 2347). Complication rates of precut sphincterotomy have been reported to vary from lows of under 5%10,11 to highs of 20% to 30%.5,12 Proponents of precut sphincterotomy have argued that any increase in complications with these techniques can be attributed to excessive preceding attempts at bile duct cannulation,S whereas opponents warn of potentially disastrous results, especially when this technique is performed by inexperienced endoscopists. 13 Even after adjusting for confounding variables related to difficult cannulation, both the Italian 5 and American 6 multicenter studies found precut sphincterotomy to independently increase the risk of overall complications; in the Italian analysis, precutting added primarily to the risk of perforation, whereas in the American study the excess risk was primarily for pancreatitis. Complications of precutting also vary with the indication for the procedure: when performed for suspected sphincter of Oddi dysfunction by a variety of endoscopists, usually without prior manometry or pancreatic stenting, precut sphincterotomy resulted in an astounding complication rate of 35%, with 24% severe complications and almost one half of the severe cases of pancreatitis occurring in the study.6 Anatomic considerations are likely important as well, the highest risk probably occurring with small papillas and with small, non-obstructed ducts; it is not hard to visualize how slight straying of the needle-knife could lacerate the pancreatic sphincter or result in perforation. In striking contrast to the findings of these two multicenter studies, many series from tertiary centers have suggested no increase in complications with precutting as compared with conventional sphincterotomy.S,10,1l,14,15 How do we reconcile the conflicting results regarding the safety of precut sphincterotomy? The answer probably includes variation in endoscopic expertise, in patient selection, and in technical details such as use of pancreatic stents. 9 The American 6 and Italian5 multicenter GASTROINTESTINAL ENDOSCOPY
97
Editorials
M Freeman
studies included endoscopists of varying experience, whereas those reporting low complication rates from precutting are almost invariably from specialized, high-volume centers. 8,10,1l,14,15 One has to be careful interpreting direct comparison of precutting techniques with standard sphincterotomy: Such univariate analyses can be misleading because precutting is usually an option for patients with obstructive jaundice and dilated ducts, patients inherently at low risk of pancreatitis and other complications, whereas conventional pull-type sphincterotomy is performed only after successful biliary cannulation, and sometimes for less clear and more risky indications. Nonetheless, a recent multivariate analysis of ERCP-induced pancreatitis at tertiary-referral centers, which adjusted for confounding variables, found no independent risk from precutting. 16 Sherman et al. 9 have reported that prior placement of a pancreatic stent dramatically reduces risk of precut-induced pancreatitis, and now report that biliary precutting over a pancreatic stent actually appears to be safer than conventional pull-type biliary sphincterotomy for patients with sphincter of Oddi dysfunction (S. Sherman, personal communication, 1998). Should use of precutting be disseminated among practicing endoscopists or discouraged? It depends on who will be performing it, and under what circumstances. The needle-knife is only a tool, like a surgical scalpel, and is not of itself a "technique"; in skilled hands, and appropriate situations, it can be used to provide safe bile duct access and avoid potentially morbid alternative approaches, whereas in other circumstances it can result in devastating complications. The effectiveness at obtaining immediate bile duct access is equally important to avoidance of complications. In our study, bile duct cannulation was achieved during that same procedure after 90% of precuts performed by high-volume endoscopists, versus only 52% for lower volume endoscopists; it seems hard to justify a complication rate of about 25% for a success rate of only 50%. Ultimately, the most important outcome is an endoscopists' overall success and complication rate for all endoscopically attempted bile duct access, rather than the method by which access is achieved. However, the new multicenter data from Italy5 and America6 suggest that in average hands, precutting is a potentially dangerous technique and should remain for the most part "for experts only."13 THE COMPLICATION GAP
The incidence of reported complications varies widely, depending primarily on definitions used and thoroughness of follow up. The Loperfido study is 98
GASTROINTESTINAL ENDOSCOPY
noticeably different from recent prospective American studies of complications in their substantially lower absolute complication rate, especially of pancreatitis. The rate of pancreatitis for diagnostic ERCP was reported at 0.74% for diagnostic and 1.4% for biliary therapeutic ERCP. In contrast, in an exhaustive review of the recent prospective literature, Gottlieb and Sherman17 tallied a cumulative rate of pancreatitis of 5.2% (nearly 10-fold higher than the Italian study) for diagnostic ERCP and 4.1% (three-fold higher than the Italian study) for therapeutic biliary ERCP. For more unambiguous, drastic complications such as perforation and hemorrhage, the Italian study reported rates (0.8% for perforation and 1.1% for hemorrhage) very comparable with those reported in our American study (0.4% and 2.0%, respectively). How can we reconcile these differences in absolute incidence of complications, especially of pancreatitis in the Loperfido study? One possibility might be differences in definition. The authors emphasize that they are reporting primarily moderate to severe complications, rather than the mild sequelae requiring 3 days or less of conservative management. However, the stated methods describe definitions similar to consensus criteria used by other investigators. Another possibility might be case-mix. Perhaps Italians, and Europeans in general, do not perform as many diagnostic and (hopefully) therapeutic ERCPs in patients with unexplained abdominal pain and non-dilated ducts (i.e. suspected sphincter of Oddi dysfunction). However, the most likely explanation for the low rate of pancreatitis in the Italian study lies in the methods of data collection and verification in this large, multicenter trial involving many physicians of varying experience with clinical trials. One wonders whether all the participants explicitly understood the definition of pancreatitis, whether all patients with post-procedure abdominal pain and extended hospitalizations had serum amylase checked, whether all patients were routinely contacted at 30 days to discover any delayed complications after hospital discharge, and whether there was any internal validation of the data. A number of methods have been developed to improve consistency of data collection with multicenter outcome studies. Ideally, the data are randomly validated by second investigators. However, more practical techniques include stating the definition of the complications clearly on the study form and internally cross-checking for any potential discrepancies between presence or absence of indicated complications and length of hospital stay or need for readmission. Patient questionnaires may also reveal unrecorded events. These techVOLUME 48, NO.1, 1998
M Freeman
Editorials
niques will often uncover a substantial number of complications and other events not initially recognized or recorded by the study participants. UNDERSTANDING RISK FACTORS ANALYSES FOR COMPLICATIONS
Many previous studies attempting to identify risk factors for complications ofERCP have analyzed one or two risk factors using univariate (bivariate) analysis, which may produce misleading results because of inability to sort out confounding variables. 18-20 A flurry of recent and ongoing studies has sought to refine our understanding or risk fac tors using multivariate (multivariate) analyses. Multivariate analysis is a useful statistical modeling tool to sort out the effect of confounding variables on an outcome and quantify the independent risk associated with each of a number of predictor variables. 21 First, potential risk factors are identified by simple univariate comparison. Then, logistic regression or linear regression models are run to identify independent predictors. In such multivariate analyses, a particular risk factor can appear to be significant in one run-through but not in another, depending on what other interrelated variables are put into the model. An analogy can be drawn to a structure of stacked building blocks-how much weight is born by anyone of several foundation blocks? This will change if one or more nearby foundation blocks are removed or added. This phenomenon may explain some of the differences in findings between recent studies. ERCP-INDUCED PANCREATITIS
The major lesson learned from recent multivariate analyses of ERCP-iIiduced pancreatitis 6,16 is that some of the most important determinants of risk are related to the patient as much as to the endoscopic technique. Most observant endoscopists have learned from experience that older patients with obstructive jaundice seldom develop pancreatitis, even after prolonged and difficult cannulations. In contrast, they have learned to fear, and hopefully those outside specialized centers have learned to avoid altogether, performing ERCP for recurrent post-cholecystectomy pain in younger women without dilated ducts or clear evidence of bile duct stones because these patients may develop significant pancreatitis despite minimal manipulation. In our multivariate model of biliary sphincterotomy complications,6 young age doubled the risk of pancreatitis, whereas the indication for the procedure of unexplained abdominal pain or suspected sphincter of Oddi dysfunction increased risk of pancreatitis fivefold, even after adjusting for other factors. Although VOLUME 48, NO.1, 1998
technical variables such as difficult cannulation or multiple pancreatic contrast injections also increased risk, their importance was somewhat less than is generally thought. These findings have largely been reproduced in a subsequent multicenter analysis. 16 The importance of patient-related risk factors in determining pancreatitis risk provides a clue as to why (in both the Italian and American multicenter studies 5,6) expert centers did not seem to have a substantially lower rate of pancreatitis; the effect of expertise is counterbalanced by higher-risk case-mix because tertiary centers, at least in the United States, perform a substantially higher proportion of ERCP for complex and previously failed cases, for unexplained abdominal pain and for pancreatic therapeutics. The Italian study 5 identified several risk factors for pancreatitis. These included two risk factors in agreement with two other multicenter multivariate analyses,6,16 that is, younger age and opacification of the pancreatic duct, which presumably serves as a surrogate marker for difficult cannulation and multiple pancreatic contrast injections. Small duct diameter, which has been previously identified as a univariate risk factor for pancreatitis, especially in patients with sphincter of Oddi dysfunction,18-20 was also found in the study by Loperfido et al. 5 to be a multivariate predictor of pancreatitis. The validity of this analysis is somewhat attenuated by the probable under-detection of pancreatitis and by the fact that certain key variables were not examined in the model. In other studies, when confounding variables such as sphincter of Oddi dysfunction and difficulty of cannulation have been taken into account, small duct diameter has not been found to be independently significant. 6,16 In fact, a multicenter study of sphincterotomy for bile duct stones found no difference in complications with small ducts,22 whereas a recent multivariate analysis found no effect whatsoever of small duct diameter on ERCP-induced pancreatitis. 16 Thus, although small ducts should be approached with special caution and pose technical difficulties, their presence should not of itself inspire undue fear of sphincterotomy for indications such as bile duct stones. It is difficult to understand the finding in the study by Loperfido et al. that sphincter of Oddi dysfunction, or "benign papillary stenosis," was not associated with an increased risk of pancreatitis or other complications. 5 This is really a syndrome, as much as a clearly defined disorder, of recurrent abdominal pain, typically suspected in young to middIe-aged women with recurrent abdominal pain after cholecystectomy.23 Most American analyses have shown that these patients are at extraordinary GASTROINTESTINAL ENDOSCOPY 99
M Freeman
risk of pancreatitis from pancreaticobiliary instrumentation of any kind including biliary therapeutics, sphincter of Oddi manometry, and from simple diagnostic ERCP as well. 6,16 In our study of biliary sphincterotomy, pancreatitis occurred in 19.1% of patients with suspected sphincter dysfunction (versus 3.6% for other indications) and accounted for 7 of the 9 episodes of severe pancreatitis 6; in another recent prospective study, pancreatitis occurred after 26% of ERCPs performed for suspected sphincter of Oddi dysfunction. 16 Likewise, three of nine deaths caused by ERCP in a recent nationwide Danish analysis were due to pancreatitis resulting from diagnostic or empirical therapeutic ERCP in women with recurrent unexplained abdominal pain and normal ducts. 24 Providing an important clue to the enigmatic danger of ERCP in these types of pa tients, Tarnasky et al. 25 have identified untreated pancreatic sphincter hypertension as the culprit and in a subsequent randomized, controlled trial have shown that pancreatic stenting dramatically reduced risk in those patients. 26 Why, then, is there such a discrepancy in the risk of ERCP in suspected sphincter of Oddi dysfunction between the Italian study and other recent studies? Many might think that it is because Italians did not perform sphincter of Oddi manometry. However, modern techniques of sphincter manometry using the aspirating catheter in the pancreas and wireguided techniques for the bile duct do not appear to add significant additional risk to ERCP.27 In our study, empirical biliary sphincterotomy for suspected sphincter disease had an equally high pancreatitis incidence (20%) to endoscopic sphincterotomy preceded by manometry (18%), and in the multivariate analysis, after adjusting for the indication for the procedure and other technical details, performance of manometry added no independent risk. 6 The most likely explanation for the low risk of "papillary stenosis" in the Italian study, however, is in their definition of this disorder. In America, many diagnostic and therapeutic ERCPs are performed in patients with recurrent unexplained abdominal pain and nondilated ducts, a group at extraordinary risk of ERCP-induced pancreatitis (20% to 30% in most recent studies). In Europe, rightly or wrongly, such patients may be more likely to be labeled as having "irritable bowel" and not subjected to ERCP,28 rather, reserving the suspicion of "papillary stenosis" for patients with dilated ducts and abnormal liver enzymes. Thus, the warning should remain clear that patients with recurrent abdominal pain and no clear evidence of bile duct stones or other anatomic pathology are at special risk of complications of ERCP. Coupled with the extremely low like100
GASTROINTESTINAL ENDOSCOPY
Editorials
lihood of finding anatomic or obstructive pathology in such patients, it is recommended that these patients should undergo ERCP primarily at specialized centers that are equipped to investigate functional sphincter disease with manometry and are adept at techniques to protect the pancreas. Magnetic resonance cholangiopancreatography (MRCP) may prove useful to identify abnormal pancreatobiliary anatomy in such patients before considering standard ERCP in a community setting. HEMORRHAGE, CHOLANGITIS, PERFORATION, AND OVERALL COMPLICATIONS
Significant hemorrhage after sphincterotomy was found to be closely related to inexperience on the part of the endoscopists in both the study of Loperfido et a1. 5 and our studies. 6 This probably reflects less precise control of the incision or less effective endoscopic control of any bleeding occurring during the procedure. In addition, our study found that coagulopathies and institution of anticoagulation therapy (not evaluated in the Italian study) were very important factors in delayed hemorrhage. 6 Neither study found length of sphincterotomy to be important. These observations and the fact that hemorrhage is typically delayed, suggest that hemorrhage is primarily the result of incomplete coagulum, rather than of transection of an aberrant retroduodenal artery. Cholangitis after ERCP probably results primarily from failure to achieve drainage in obstructed patients: this is reflected in the observation of Loperfido et al. 5 that cholangitis was associated with operator inexperience and jaundice and in our study with both failed drainage and malignant obstruction. 6 We also found that combined percutaneous procedures, which create potential fistulae between biliary and vascular structures, were associated with cholangitis. 6 Perforation is a dreaded complication whose infrequency precludes detailed statistical analysis; however, both studies suggest that it is more common in situations with less precise control of sphincterotomy incision (Le., precut and Billroth II). Although somewhat counterintuitive, it is quite clear from these and previous studies that old age, most comorbid conditions, and emergency procedures pose no increased risk; in fact, perhaps they provide some protection from specific ERCP-related complications such as pancreatitis. The exception is cirrhosis, which we found to be a univariate risk factor for hemorrhage and a multivariate risk factor for overall complications, whereas the Italian study did not. The likely explanation is that our American study included a substantial number of tertiary liver transplantation centers and more patients with VOLUME 48, NO.1, 1998
Editorials
M Freeman
advanced (Child's B and C) cirrhosis, who are substantially more fragile than those with well-compensated liver disease. SUMMARY
The most important lesson learned from this Italian study of ERCP complications is that the experience of the endoscopist is important in determining the outcome of ERCP. This is reflected not only in lower complication rates for experienced endoscopists, but in higher success rates as well. Higher technical success rates may be at least as important as avoidance of complications in determining outcome of ERCP. With failed attempts at therapeutic ERCP, complications are no less common, but extended hospital stay, need for repeat procedures, and additional costs are substantial.2 9 How much experience and training are required for an endoscopist to become effective at ERCP? A recent study by Jowell et al. 3o showed that trainees required 180 to 200 supervised procedures to achieve even the modest success rate of 70% at deep bile duct cannulation. It is not known, however, how much ongoing experience is required to maintain proficiency. In the Italian study, centers performing less than 200 ERCP per year had less satisfactory results than higher volume centers, but the case volume of the individual endoscopist may be of ultimate importance. In our study, a reduction in rate of complications and failed bile duct access was observed for endoscopists performing more than 50 biliary sphincterotomies a year, probably representing at least 100 to 150 ERCPs annually. Only a minority of endoscopists in the United Sates are likely to achieve this volume: an older survey of practicing gastroenterologists in the United States showed that the median number of ERCPs performed per year was 30 procedures, and 25% did five or fewer ERCPs per year, a figure which is incompatible with satisfactorily high success rates and low complication rates. 31 There is also substantial variation in performance between individual endoscopists. A recent survey in the United Kingdom showed surprisingly wide differences between endoscopists in success rates (from 76% to 95%) and serious complications (0% to 16%).32 This variation in practice and outcomes seems unacceptably high. Overall, the Italian data support the increasing evidence that outcomes will be optimal if fewer endoscopists perform more ERCP.28 This is likely to be an unpopular recommendation with many endoscopists, but perhaps not with patients or thirdparty payers. Does merely performing more ERCP guarantee better outcomes? Certainly not. First, there is very gradual and slow learning curve. VOLUME 48, NO. 1, 1998
Second, because of skill and good clinical judgment, there are no doubt some individuals who are able to maintain reasonably high success rates and low complication rates with ~ moderate volume of ERCP, whereas other individuals may have a propensity toward complications no matter how many cases they perform, so that increasing their case volume will simply result in more complications. We clearly need some kind of yardstick to measure performance of endoscopists and assess outcomes of endoscopic procedures such as ERCP. However, there currently is no mechanism to do this in a standardized fashion, and such efforts will be met with resistance by some and used for personal gain by others. As physician groups increasingly market their "outcomes" to health care organizations, the potential for subjective, biased reporting of such outcomes is obvious. Therefore rigorous, standardized assessments, such as those widely used for cardiac surgery, should be instituted in the gastrointestinal endoscopy community. These must include objective reporting of outcomes for individuals and practices performing these procedures, using standardized consensus criteria for complications and outcomes. In addition to complications, any evaluations must include careful adjustment for case mix and technical success rates. Electronic endoscopic databases may well provide the best foundation for these assessments. Only after careful assessment can the endoscopic community learn how effective and safe is ERCP at the present time, so that we can improve outcomes of this procedure in the future. Martin L. Freeman, MD A4inneapoH~
A4mnesota
REFERENCES 1. Cotton PB, Lehman G, Vennes JA, Geenen JE, Russell RCG, Meyers WC, et al. Endoscopic sphincterotomy complications and their management: an attempt at consensus. Gastrointest Endosc 1991;37:383-91. 2. Cotton PB. Complications, comparisons and confusion: commentary. In: Cotton PB, Tytgat GNJ, Williams CB, editors. Annual of gastrointestinal endoscopy. London: Current Science Ltd; 1990. p. 7-9. 3. Cotton PB. Outcomes of endoscopy procedures: struggling towards definitions. Gastrointest Endosc 1994;40:514-8. 4. Fleischer DE. Better definition of endoscopic complications and other negative outcomes. Gastrointest Endosc 1994;40: 511-14. 5. Loperfido S, Angelini G, Benedetti G, Chilovi F, Costan F, De Berardinis F, et al. Major early complications from diagnostic and therapeutic ERCP: a prospective multicenter study. Gastrointest Endosc 1998;48:1-10. 6. Freeman ML, Nelson DB, Sherman S, Haber GB, Herman ME, Dorsher PJ, et al. Complications of endoscopic biliary sphincterotomy. N Engl J Med 1996;335:909-18. 7. Davis WZ, Cotton PB, Arias R, Williams D, Onken JE. ERCP and sphincterotomy in the context of laparoscopic cholecys-
GASTROINTESTINAL ENDOSCOPY
101
MWiersema
tectomy: academic and community practice patterns and results. Am J GastroenteroI1997;92:597-601. 8. Slot WB, Schoeman MN, Disario JA, Wolters F, Tytgat GNJ, Huibregtse K. Needle-knife sphincterotomy as a pre-cut procedure-a retrospective evaluation of efficacy and complications. Endoscopy 1996;28:334-9. 9. Sherman S, Hawes R, Earle D, Baute P, Bucksot L, Lehman G. Does leaving a main pancreatic duct stent in place reduce the incidence of biliary precut-induced pancreatitis [abstract]? Gastrointest Endosc 1994;40:124. 10. Binmoeller KF, Seifert H, Gerke H, Seitz U, Portis M, Soehendra N. Papillary roof incision using the Erlangentype pre-cut papillotome to achieve bile duct cannulation. Gastrointest Endosc 1996;44:689-95. 11. Howell DA, Qaseem T, Hanson BL, Parsons WG, Elton E, Bosco JJ. Needle knife papillotomy (NKP) without stent insertion for the difficult sphincterotomy (ES): a standardized technique explained [abstract]. Gastrointest Endosc 1996;43:383. 12. Shakoor T, Geenen JE. Pre-cut papillotomy. Gastrointest Endosc 1992;38:623-7. 13. Cotton PB. Precut papillotomy: a risky technique for experts only. Gastrointest Endosc 1989;35:578-9. 14. Kasmin FE, Cohen D, Batra S, Cohen SA, Siegel JH. Needleknife sphincterotomy in a tertiary referral center: Efficacy and complications. Gastrointest Endosc 1996;44:48-53. 15. Foutch PG. A prospective assessment of results for needleknife papillotomy and standard endoscopic sphincterotomy. Gastrointest Endosc 1995;41:25-32. 16. Sherman S, Lehman G, Freeman ML, Earle D, Watkins J, Barnett J, et a1. Risk factors for post-ERCP pancreatitis: a prospective multicenter study [abstract]. Am J Gastroenterol 1997;92:1639. 17. Gottlieb K, Sherman S. ERCP and endoscopic biliary sphincterotomy-induced pancreatitis. Gastrointest Endosc Clin N Am 1998;8:87-114. 18. Sherman S, Ruffolo TA, Hawes RH, Lehman GA. Complications of endoscopic sphincterotomy: a prospective series with emphasis on the increased risk associated with sphincter of Oddi dysfunction and nondilated bile ducts. Gastroenterology 1991;101:1068-75. . 19. Boender J, Nix GA, de Ridder MA. Endoscopic papillotomy for common bile duct stones: factors influencing the complication rate. Endoscopy 1994;26:209-16. 20. Chen YK, Foliente RL, Santoro MJ, Walter MH, Collen MJ. Endoscopic sphincterotomy-induced pancreatitis: increased risk associated with nondilated bile ducts and sphincter of Oddi dysfunction. Am J GastroenteroI1994;89:327-33. 21. Concato J, Feinstein AR, Holford TR. The risk of determining risk with multivariable models. Ann Intern Med 1993;118: 201-10. 22. Cotton PB, Geenen JE, Sherman S. Sphincterotomy for stone is safer than advertised, even in young patients with smaller ducts. A multicenter prospective study [abstract]. Gastrointest Endosc 1994;40:104. 23. Kozarek RA. Biliary dyskinesia: Are we any closer to defming the entity? Gastrointest Endosc Clin N Am 1993;3:167-78. 24. Adamsen S, Trap R, Hansen OH. ERCP complications compensated by public insurance [abstract]. Endoscopy 1997;29:E38. 25. Tarnasky P, Cunningham J, Cotton P, Hoffman B, Palesch Y, Freeman J, et a1. Pancreatic sphincter hypertension increases the risk of post-ERCP pancreatitis. Endoscopy 1997; 29:252-7. 26. Tarnasky P, Palesch Y, Cunningham J, Cotton P, Hawes RH. Pancreatic stenting prevents pancreatitis after biliary sphincterotomy in patients with sphincter of Oddi dysfunc-
102
GASTROINTESTINAL ENDOSCOPY
Editorials
tion: a prospective randomized trial [abstract]. Gastrointest Endosc 1997;45:AB150. 27. Sherman S, Troiano FP, Hawes RH, Lehman GA. Sphincter of Oddi manometry: decreased risk of clinical pancreatitis with use of a modified aspirating catheter. Gastrointest Endosc 1990; 36:462-6. 28. Huibregtse K. Complications of endoscopic sphincterotomy and their prevention. N Engl J Med 1996; 335:961-2. 29. Freeman ML, Nelson DB, Sna,dy HW, DiSario JA, Overby CS, Ryan ME, et a1. Failures and complications of ERCP: impact on procedural outcome and resource utilization [abstract]. Am J GastroenteroI1997;92:1634. 30. Jowell PS, Baillie J, Branch MS, Affronti J, Browning CL, Phillips B. Quantitative assessment of procedural competence: a prospective study of training in endoscopic retrograde cholangiopancreatography. Ann Intern Med 1996;125:983-9. 31. Wigton RS. Measuring procedural skills. Ann Intern Med 1996; 125:1003-4. 32. Tanner AR. ERCP: present practice in a single region. Eur J Gastroenterol HepatoI1996:8:145-8.
Diagnosing chronic pancreatitis: shades of gray When assessing a new test to diagnose chronic pancreatitis, investigators will typically evaluate patients with advanced disease. These results are then compared with healthy controls. The sensitivity and specificity figures which arise from such studies are usually quite promising. However, when the test is then applied to a more realistic mix of patients who have less severe forms of the disease, the sensitivity and specificity figures are often found to be less than previously predicted and the enthusiasm for the test wanes. In this issue two reports are presented for which we should commend the authors for doing the exact opposite. 1,2 They have performed endosonography (EUS) on a group of individuals that predominantly have no evidence for or only mild to moderate chronic pancreatitis. Both of these studies hypothesize that the enhanced imaging resolution permitted with EUS may afford some advantage and allow for the accurate diagnosis of chronic pancreatitis. However, taking on the task of examining the operating characteristics of EUS in the evaluation of chronic pancreatitis is fraught with difficulty. From the standpoint of a gold standard, for most cases, this does not exist. We do not routinely biopsy the pancreas to establish or refute the diagnosis of chronic pancreatitis. Therein lies the problem with the claimed superiority (or inferiority) of any test being scrutinized. In clinical Copyright © 1998 by the American Society for Gastrointestinal Endoscopy 0016-5107/98/$5.00 + 0 37/70/89933 VOLUME 48, NO.1, 1998