Endoscopy 2007; 39(9): 833-835
DOI: 10.1055/s-2007-966771
Editorial

© Georg Thieme Verlag KG Stuttgart · New York

30 years of ERCP and still the same problems?

N.  M.  Guda1 , M.  L.  Freeman2
  • 1Pancreatobiliary services, St. Luke’s Medical Center, Milwaukee, Wisconsin, USA
  • 2Department of Medicine, Hennepin County Medical Center, University of Minnesota, Minneapolis, Minnesota, USA
Further Information

Publication History

Publication Date:
17 August 2007 (online)

ERCP was first performed in the late 1960s, became a therapeutic modality when the first sphincterotomy was carried out in 1974 [1] [2], and has since evolved to become a predominantly therapeutic procedure. Although progress has been made, fear of serious complications still looms large in the minds of endoscopists and patients alike.

A number of prospective multicenter studies have helped us to understand patient- and procedure-related risk factors. One study from the USA and two studies from Italy have included large numbers of endoscopic retrograde cholangiopancreatography (ERCP) and/or sphincterotomy procedures that were carried out in a wide spectrum of practices [3] [4] [5]. Post-ERCP pancreatitis was specifically addressed in two other large studies [6] [7]. Risk factors were defined after both univariable and multivariable analyses ([Table 1]). The risk of overall complications is largely independent of patient comorbidity, but is dependent on certain patient-related risk factors such as presence of suspected sphincter of Oddi dysfunction (SOD), and technique-related factors such as difficulty of cannulation and use of precut sphincterotomy for biliary access (at least in mixed hands). Risk factors for post-ERCP pancreatitis ([Table 2]), also include younger age and female sex, injection of the pancreatic duct, and performance of pancreatic sphincterotomy [9].

Table 1 Risk factors for overall complications of ERCP in multivariate analyses; comparison of four large multicenter studies Risk factors Study and reference Freeman, 19 963 Loperfido, 19 984 Masci, 20 015 Williams, 20 078 ERCPs, nComplications, n (%)Severe complications, % 2347229 (9.8)1.6 182798 (5.4)? 2444121 (5.0)? 4561230 (5.0)1.8 Young age U n. s. S n. s. Comorbid illness burden n. s. ? ? n. s. Sex n. s. n. s. n. s. ? SOD S ? ? S Coagulopathy n. s. ? ? n. s. Cirrhosis S ? ? n.s Peri-ampullary diverticulum n. s. n. s. ? n.s Difficult cannulation S ? ? S Pancreatic duct injected U ? U ? Biliary ES All ? n. s. ? Precut ES S S S S PTC combined procedure S ? ? n. s. Failed drainage U ? S n. s. Case volume Ua S ? n. s. All, all patients had this variable; ERCP, endoscopic retrograde cholangiopancreatography; ES, endoscopic sphincterotomy; n. s., not significant by univariate analysis; PTC, percutaneous transhepatic cholangioscopy; S, significant by univariate and multivariate analysis; SOD, suspected sphincter of Oddi dysfunction; U, significant by univariate analysis only. a Significant by multivariate analysis if technique-dependent factors excluded.?, not examined.

Table 2 Risk factors for pancreatitis after ERCP in multivariate analyses Risk factors Study and reference Freeman, 19 963 Loperfido, 19984 Freeman, 20 017 Masci, 20 015 Cheng, 20 066 Williams, 20 078 ERCPs, nPancreatitis, n (%)Severe pancreatitis, % 2347127 (5.4)0.4 182729 (1.6)? 1963131 (6.7)0.3 244444 (1.8)0.1 1115168 (15.1)1.0 456174 (1.6)0.5 Young age S S U S S S Sex (F) U n. s. S n. s. U S No CBD stone ? ? U ? U n. s. Chronic pancreatitis absent ? ? S ? n. s. n. s. History of PEP U ? S ? S ? SOD S n. s. S U S U Difficult cannulation S ? S ? U S Pancreatic duct injected S S S U S U Acinarize pancreas U n. s. n. s. U n. s. n. s. Pancreatic ES None None S ? Sb U Precut ES S n. s. U S n. s. n. s. Balloon dilatation None None S ? n. s. n. s. Case volumea n. s. U U ? S S CBD, common bile duct; ERCP, endoscopic retrograde cholangiopancreatography; ES, endoscopic sphincterotomy; None, no patients had this variable; n. s., not significant by univariate analysis; PEP, post-ERCP pancreatitis; S, significant by univariate and multivariate analysis; SOD, suspected sphincter of Oddi dysfunction U, significant by univariate analysis only.?, not examined. a Defined variably as endoscopist case volume, fellow participation, center case volume, or district versus university hospital. b Minor papillotomy.

Although it has been hard to demonstrate, it is widely suspected that endoscopic expertise plays a major role in preventing complications. Earlier studies have shown that performance of fewer than one sphincterotomy per week per endoscopist, and fewer than 200 ERCP procedures a year per center were risk factors for overall complications, especially severe complications [3] [4]. Interestingly, the link between post-ERCP pancreatitis and endoscopist expertise has been harder to establish: one study found no association [3], whereas the other showed an association on univariate analysis only [4]. Only one single-center study has found a clear relationship between endoscopist experience and pancreatitis rates [10]. Another study found an independent association between trainee participation and increased risk of pancreatitis - finally providing hard evidence for a convenient excuse that many of us at training programs have used for years [6]! Any difference in rates of pancreatitis between low- and high-volume endoscopists is often blunted by a disparity in complexity and risk profile of cases attempted.

To further examine complications of ERCP, Williams et al. have reported a multicenter prospective study involving 66 hospitals in five English regions, at which 4561 patients underwent 5264 procedures [8]. The study was methodologically sound, including follow-up to 30 days after the procedure. Intention-to-treat analysis was used, including cases with failed cannulation at ERCP. Analysis was focused on 4561 patients undergoing their first recorded ERCP procedure. A strength of the study is that it included a wide spectrum of practices and endoscopists, overcoming the natural tendency for predominantly centers with the most expertise to participate in rigorous studies. The authors examined a variety of risk factors including endoscopist experience as well as type of hospital (both tertiary and nontertiary/rural centers) at which the procedures were carried out.

The majority (76 %) of the procedures were therapeutic, but the relatively high number of diagnostic procedures probably reflects inconsistent availability of magnetic resonance cholangiopancreatography (MRCP) and endoscopic ultrasound (EUS). The study group represents a self-selected group of individuals at low risk of complications based on data from earlier studies, in that mean age was older (65 years), most had biliary stones or malignancy, and SOD was present in < 1 %.

Complications occurred in 230 (5.0 %) of 4561 cases, with severe complications in 84 (1.8 %). On univariate analysis precut papillotomy (odds ratio [OR] 1.55), multiple cannulation attempts (OR 1.48) and SOD (OR 2.56) were significant. By multivariate analysis, only precut papillotomy demonstrated a significant risk (OR 2.56). The high risk of precut is similar to other studies in which a wide spectrum of endoscopists was included, but not at centers with advanced expertise [6]. Overall complications were neither dependent on operator volumes (hospital or endoscopist), nor the location of the hospital (rural or urban).

For post-ERCP pancreatitis, prior history of pancreatitis, endoscopist’s volume/year, and number of cannulation attempts were significant by univariate analysis, but only younger age, female sex, and greater than one cannulation attempt were significant by multivariate analysis. Similar to previous studies, risk of bleeding post sphincterotomy was found to be higher in hemodialysis patients [11], and aspirin use was not found to raise risk. Coagulopathy may not have been sufficiently prevalent to allow analysis. The authors did not analyze the most potent previously described risk factor, namely administration of anticoagulation shortly after sphincterotomy [3]. Pure cut current appeared to increase the risk of bleeding as well, and was lower with microprocessor-based current [12]. Overall, most findings of Williams and colleagues are similar to those of previous studies ([Tables 1] and [2]).

In the Williams study, severe complications were disproportionately common, occurring in 1.8 % (more than one in three) out of a total complications rate of 5.0 %. By comparison, a multicenter US study limited to biliary sphincterotomy found overall complications in 9.8 % (twice as many), but severe complications in 1.6 %. In the Williams study, post-ERCP pancreatitis occurred in only 1.6 %, of cases, but severe pancreatitis occurred in 0.5 %, or about one in three. By comparison, two US studies reported overall post-ERCP pancreatitis rates of 5.4 % and 6.7 % (much higher than the Williams study), with severe pancreatitis in only 0.4 % and 0.3 %, respectively (less than one in 10). In the Williams study, the lower rate of overall but not severe complications might be explained by the hypothesis that mild and moderate complications were under-detected. In a large registry-type study including a very large number of centers, mild or moderate pancreatitis requiring a few extra days of hospitalization could easily escape detection, if amylase was not measured or, simply, the follow-up was not uniform. In contrast, severe pancreatitis is quite drastic, seldom escapes detection, and thus perhaps serves as a better yardstick of comparison between studies. It is unlikely that the biology of post-ERCP pancreatitis is significantly different between studies, and in fact in American studies most cases of severe post-ERCP pancreatitis have been associated with SOD, a risk factor largely absent from this British study.

Although Williams and colleagues propose that the overfitting of multivariate models has limited validity of conclusions of previous studies, and that larger numbers of patients would overcome that limitation, their study suffers the same constraints. Overfitting of a model is dependent not on the number of patients so much as the number of outcomes. In their study, there were 230 complications and 74 cases of pancreatitis, which is comparable to or lower than numbers of adverse outcomes in most previous large studies ([Tables 1] and [2]).

The most interesting and novel aspect of the Williams study is its examination of the relationship between endoscopist experience and complications. Community hospitals had higher rates of post-ERCP pancreatitis than university hospitals. The authors point out that the essential role of ancillary services for complex procedures might be responsible. Somewhat counter to current thinking, neither the endoscopist’s personal case volume nor experience, or the hospital case volume resulted in significant differences in complications [13].

One of the most striking findings of the Williams study was the high rate of failed cannulation. In all, 14 % of initial procedures resulted in failed access, which is probably similar to rates in many community practices in the USA, but much higher than at specialized advanced centers, where bile duct access is almost always successful. It would have been of interest to analyze the relationship between cannulation success and case or center volume. The lower rates of failure and shorter hospital stays at higher-volume centers was recently confirmed in a US administrative database [13]. It seems likely that the major difference in outcomes of expert and not-so-expert endoscopists lies not so much in overall complication rates, but in success rates, especially at biliary access and therapy. From the patient and healthcare provider point of view, a successful procedure with a mild complication may be preferable to a failed procedure with no complication, as failure requires repeat or alternate invasive procedures with their own morbidity and cost [14].

Where are we in over 30 years since the advent of therapeutic ERCP? With the help of meticulous data abstraction and prospective studies we now understand that complications are linked to patient-related factors, procedure-related factors and to the endoscopist’s expertise. The current study of Williams and colleagues reinforces the notion that with careful patient selection one can reduce the risk of complications. ERCP is the riskiest in those who need the procedure the least. Diagnostic ERCPs should be avoided now that alternate strategies with minimal risk such as MRCP and EUS are widely available. High-risk patients, especially those with possible SOD and recurrent pancreatitis, are probably better served by larger centers with advanced expertise. Now that complications are fairly well understood, the next frontier for large scale outcomes research in ERCP is to refine strategies to minimize risk, and to define the overall relative importance of technical success at determining patient outcomes.

Competing interests: None

References

M. L. Freeman

Department of Medicine

Hennepin County Medical Center

University of Minnesota

Minneapolis

Minnesota 55415

USA

Fax: +1-612-904-4299

Email: freem020@umn.edu