Chapter  50:  Endoscopic Retrograde Cholangiopancreatography: Evidence Report/Technology Assessment Number 50

Search Strategy for the Identification of Articles

The National Library of Medicine (NLM) staff conducted a comprehensive literature search for journal articles on ERCP from the PubMed^®/MEDLINE^®, BIOSIS, EMBASE, and SciSearch^® databases with a publication date from 1980 through August 13, 2001. Articles that had been indexed to the NLM Medical Subject Heading (MeSH^®) "cholangiopancreatography, endoscopic retrograde," as well as those containing the following list of ERCP synonyms and textword combinations, were retrieved:

Endoscopic retrograde cholangiopancreatogr?
Endoscopic retrograde cholangio-pancreatogr?
Endoscopic retrograde pancreatocholangiogr?
Endoscopic retrograde pancreato-cholangiogr?
ERCP
ERCPs
Endoscopic retrograde cholangiogr?
ERC and endoscop?
ERC and cholangiogr?
Endoscopic cholangiogr?
Endoscopic retrograde pancreatogr?
ERP and endoscop?
ERP and pancreatogr?
Endoscopic pancreatogr?
Endoscopic cholangiopancreatogr?
Endoscopic cholangio-pancreatogr?
ECP and endosc?
ECP and cholangiogr?
Endoscopic pancreatocholangiogr?
Endoscopic pancreato-cholangiogr?
EPC and endoscop?
EPC and pancreatogr?

The "?" is a truncation symbol used to permit retrieval for variant word endings, such as cholangiopancreatography, cholangiopancreatographic, etc.

Excluded from the search results were articles that:

• Were written in a foreign language.

• Did not have abstracts as a part of the online record in any of the databases searched.

• Did not include human subjects.

• Contained reports of only a single case.

The literature search for Topic 1c on prediction of common bile duct stones and for additional studies selected by the secondary selection criteria for Topics 3 and 4 used a streamlined search process to identify key articles addressing the clinical issue of interest. Reference lists from these articles were reviewed, focused MEDLINE searches were performed, and related articles were identified.

The Technical Advisory Group and peer reviewers for this project were asked to inform the project team of any studies relevant to the key questions addressed in this evidence report that were not retrieved by either of the search strategies.

Search Results

The online searches of the PubMed, EMBASE, BIOSIS, and SciSearch databases in conjunction with additional citations identified through manual searching yielded a total of 5,698 titles and abstracts for review. Based on review of abstracts, 789 articles were selected for review in full text. Approximately 117 of these articles were excluded as review articles. Primary and secondary selection criteria were applied to articles identified as potential clinical trial reports. This process yielded a total of 149 included studies for the review of evidence.

Primary Selection Criteria

The selection criteria for all topics in this report were:

1. Full-length report in peer-reviewed medical journals.

2. Published in English.

3. Reported outcomes relevant to this systematic review.

4. Where there were multiple reports of a single study, only the report judged to be most recent and complete, based on number of included patients and length of followup, was included. If additional relevant outcomes were included in the duplicate reports, these data were abstracted and added to the data from the primary report with citation to the supplementary articles.

5. Prospective in design, or if retrospective, enrolled consecutive patients or used appropriate sampling methods (e.g., case-control sampling method).

In order to keep readers informed of ongoing studies, studies published only in abstract form since 1999 and judged to be important are noted in this systematic review; but data were not abstracted into the evidence tables.

Studies of diagnostic performance met the following additional selection criteria:

1. Compared ERCP and at least one of the relevant diagnostic alternatives or compared two ERCP alternatives.

2. Subjected at least 90 percent of participants to both ERCP and the relevant diagnostic alternative.

3. Addressed a relevant patient population.

4. Included at least 25 subjects.

5. Reported sufficient information to be able to calculate 2x2 contingency tables of diagnostic performance.

Studies of therapeutic outcomes met the following additional selection criteria:

1. Compared ERCP strategies with at least one of the relevant therapeutic alternatives.

2. Addressed a relevant patient population.

3. Included at least 25 subjects in each treatment group being analyzed separately.

4. Reported on at least one relevant outcome measure.

5. Were contemporaneous comparison studies. If not contemporaneous, the populations and treatment setting were comparable.

Studies of predictors of ERCP complications met the following additional selection criteria:

1. Included a multivariable analysis of the relationship between patient, procedure, or operator factors and ERCP complications.

2. Enrolled at least 100 patients if a cohort study or at least 25 cases if a case-control study.

3. Addressed potential confounding variables in either the selection of subjects or analysis.

Studies on the prediction of common bile duct stones met the following additional selection criteria:

1. Reported the association of either (a) specific risk factors of interest and the presence of a common bile duct stone (specific risk factors of interest were jaundice, liver function test results, and ultrasound finding of a dilated common bile duct) or (b) a prediction rule or model predicting likelihood of having a common bile duct stone and the presence of a common bile duct stone.

2. Enrolled at least 100 patients.

3. Reported sufficient information to be able to calculate 2x2 contingency tables of diagnostic performance in the prediction of presence or absence of a common bile duct stone.

Secondary Selection Criteria

There was a paucity of literature that met the primary selection criteria for questions on ERCP treatment of chronic pancreatitis (Topic 3b) and ERCP treatment of chronic abdominal pain of possible pancreaticobiliary origin (Topic 4b). In order to examine these questions, the original study selection criteria were relaxed for these topics to include:

1. Randomized controlled trials or otherwise concurrently controlled studies of an ERCP intervention compared to a relevant therapeutic alternative, regardless of sample size for pancreatitis.

2. Single arm pre-post-intervention studies that selected a well-defined population with a predictable natural history ascertained by baseline evaluation over 3 months. These studies must also have used an appropriate, well-designed outcome measure over at least 6 months of followup.

Topic 1: Patients with known or suspected common bile duct stones

Diagnostic performance of ERCP compared to alternatives:

• The search and selection process yielded 10 studies on MRCP (total n=834), 9 studies on EUS (total n=601), and 6 studies with 7 sets of findings on CTC (total n=266), but reference standards were not consistent among studies.

• Individual studies were relatively small and unlikely to have adequate power to detect a statistically significant difference, and no studies reported tests of statistical significance. Thus, it is not possible to determine with confidence whether the diagnostic performance is similar or poorer than ERCP or to accurately quantify any difference.

• The evidence comparing EUS to ERCP employs a reference standard that permits inferences regarding comparative performance. The evidence suggests that EUS is similar to ERCP in detecting common bile duct stones.

• MRCP has a degree of concordance with ERCP that results in sensitivities and specificities greater than 90 percent in most studies. Concordance of CTC with ERCP appears to be lower, with sensitivities as low as 80 percent in some studies.

• The role of alternative tests in the management of patients with suspected common bile duct stones cannot be determined strictly by diagnostic performance. The costs and risks of the tests, and the costs and risks of actions based on test results, along with the pretest probability of stones must be all be considered to determine the optimal management strategy.

ERCP treatment strategies compared to surgical or medical management:

• In order to evaluate ERCP treatment strategies, studies must account for patients through the diagnostic and treatment process, including additional procedures needed when initial treatment fails and total morbidity of the alternative strategies. Overall, the literature is very thin and spread out over many different comparisons of interest, preventing strong conclusions about any specific comparison of treatment strategies.

• The limited evidence available suggests that: laparoscopic common bile duct exploration may be better than ERCP strategies to manage cholecystectomy patients with the least resource use, definitive surgery with cholecystectomy prevents long-term complications at acceptable short-term morbidity when compared to sphincterotomy alone in high-risk surgical patients with suspected common bile duct stones, and endoscopic treatment of acute cholangitis reduces short-term mortality when compared to emergency surgery.

• Limited evidence suggests that the following techniques have similar stone removal rates and short-term complications: intracorporeal and extracorporeal lithotripsy methods for removing large common bile duct stones, balloon dilation and sphincterotomy, needle-knife fistulotomy and needle-knife precut papillotomy.

Diagnostic value of specific risk factors or predictive models for assessing the likelihood of having a common bile duct stone:

• The probability of a common duct stone is one important factor in determining diagnostic and treatment strategies. When preoperative probability is high, ERCP may be preferred. When probability is low, expectant management is preferred. Additional diagnostic tests may be used to discriminate among patients in the middle range of probability. The exact probability cutoffs depend on the risks and benefits of the diagnostic and treatment alternatives. The risk factor or prediction model with the best receiver-operating characteristics (ROC) would make the best decision rule if the cutoff threshold were set correctly.

• Thirteen studies (total n=7,409) reported multiple findings of sensitivities and specificities of a single or combination of risk factors to predict the presence of common bile duct stones. The single risk factors most commonly assessed were: clinical jaundice or elevated bilirubin, liver function tests, and ultrasound findings of a dilated common bile duct. All have significant associations with the presence of common duct stones, but none have both high sensitivity and specificity. Of the four studies testing prediction rules based on combinations of risk factors, only one study was a validation of an independently developed prediction rule. Multivariable prediction rules appear to have superior ROCs compared to individual risk factors.

• The absence of any risk factors for stones (or a discriminant function indicating absence of stones) is a very strong predictor of the absence of stones. Absence of any risk factor produces probabilities of stones that are in the same range as a negative ERCP exam in a patient with risk factors for stones (0 percent to 17 percent).

Topic 2: Patients with known or suspected pancreaticobiliary malignancy

Diagnostic performance of ERCP tissue sampling techniques in establishing a tissue biopsy diagnosis of pancreaticobiliary malignancy in comparison to each other and compared to alternative nonsurgical tissue sampling techniques:

• Twelve studies comparing at least two tissue sampling techniques were identified in this systematic review. The available studies are limited by small size and do not consistently compare techniques in the same group of patients. Most studies do not report statistical tests, so it is not possible to determine with confidence whether reported differences in sensitivity are significantly different. While available evidence is suggestive, larger studies are needed to draw conclusions on relative performance of tissue sampling techniques.

• The available evidence suggests that sensitivity for detecting malignancy is similar or higher for brush cytology vs. bile aspiration cytology, similar for FNA cytology vs. brush cytology, and similar or higher for forceps biopsy vs. brush cytology. Using combinations of two or more sampling techniques may increase overall sensitivity. No comparative studies evaluated whether incremental improvement could also be achieved by repeated sampling using the same technique.

• In the absence of comparative studies of EUS-FNA and ERCP-FNA, indirect comparison of single-arm studies was attempted. Results from 10 studies including at least 400 subjects with pancreatic mass suggest a range of sensitivity in detecting pancreatic malignancy of 60-94 percent with a specificity of 100 percent. Two studies of ERCP-FNA including 164 subjects with various pancreatobiliary tumors reported sensitivities ranging from 25 percent to 62 percent. While sensitivity reported in these studies appears to be lower than that for EUS-FNA, such a comparison is not valid due to differences in study populations, cytology techniques, and study settings.

Diagnostic performance of ERCP compared to alternatives in detecting malignant pancreaticobiliary obstruction:

• The available evidence directly comparing ERCP with either MRCP or EUS is modest in size and of varying methodological quality. The evidence comparing ERCP with MRCP is somewhat stronger than that comparing ERCP with EUS.

• Individual studies do not demonstrate statistically significant differences in diagnostic performance for ERCP vs. MRCP or for ERCP versus EUS for characterizing malignant strictures. In sum, the available studies suggest that both MRCP and EUS provide similar diagnostic performance as ERCP in detecting pancreaticobiliary malignant obstruction.

Treatment outcomes using ERCP strategies to treat malignant pancreaticobiliary obstruction compared to using surgical or interventional radiology treatment:

• Five studies compared endoscopic stent drainage with surgical bypass for palliation of malignant obstructive jaundice, and a randomized controlled trial of 204 patients provided the most robust evidence. There were no significant differences in overall survival, relief of jaundice, technical success, total hospitalization days, or perioperative mortality. Major complications were more frequent in the surgery group (11 percent vs. 29 percent, p=0.02), and stent replacement was required in 37 percent of patients treated with ERCP stents.

• Two randomized controlled trials (total n=206) and one nonrandomized trial (n=165) compare metal to plastic stents placed by ERCP for palliation of biliary obstruction due to malignancy. Both types of stents offer initial relief of jaundice, and the available evidence does not conclusively show any difference in perioperative adverse events. Overall patient survival is not significantly different when stent occlusions are treated with stent exchange as needed. Total resource utilization, including need for repeat ERCP, total hospital days, and costs, was reported to be lower with metal stents compared with plastic stents.

• Six studies (total n=782) addressed preoperative stenting compared to no stenting prior to surgery for malignant pancreaticobiliary obstruction. The available evidence is of poor methodologic quality and fails to demonstrate that preoperative stenting improves health outcomes. Few studies report overall complications including both those related to the preoperative stent and the surgery, and these suggest that when complications of preoperative endoscopic stenting are considered along with the perioperative complications of surgery, preoperative stenting is associated with more complications. Preoperative stenting does appear to significantly improve elevated bilirubin and liver function tests, but the available evidence does not suggest that surgical outcomes are improved as a result.

Topic 3: Patients with pancreatitis

Diagnostic performance of ERCP compared to alternatives to detect underlying causes or complications of pancreatitis that are amenable to treatment:

• Three studies (total n=190) were found that met selection criteria. Each study addresses a different potential cause or complication of pancreatitis amenable to treatment. The available evidence is insufficient to compare ERCP and other diagnostic modalities for the identification of treatable causes or complications of pancreatitis.

Treatment outcomes of ERCP strategies compared to surgical or medical therapy:

• For treatment of acute pancreatitis, three randomized controlled trials (total n=554) compared early ERCP to delayed or selective ERCP. The available evidence suggests that early ERCP reduces complications in patient populations with acute pancreatitis and signs and symptoms suggesting biliary obstruction. In patients with low likelihood of biliary obstruction, delayed or selective ERCP permits many patients to avoid the procedure, and may result in lower complication rates. In addition, one retrospective associational study of a Veterans Administration database of patients with acute pancreatitis (n=2,075) suggests that outcomes of ERCP treatment are similar to those of surgery.

• For ERCP treatment in patients with acute recurrent or chronic pancreatitis, study selection criteria were relaxed as described above. Although the available evidence is sparse and largely uncontrolled, it suggests that ERCP treatment reduces emergency room visits and hospitalization in patients with pancreas divisum and acute recurrent pancreatitis. Evidence on ERCP drainage of pseudocysts is also sparse and poorly controlled but suggests that pain relief with ERCP is similar to results of surgery.

Topic 4: In patients with abdominal pain of possible pancreaticobiliary origin

Diagnostic performance of ERCP with sphincter of Oddi manometry compared with alternatives to identify a pancreaticobiliary origin of pain:

• The available evidence is not sufficient to permit conclusions on the diagnostic performance of biliary scintigraphy for sphincter of Oddi dysfunction. The body of evidence consists of three studies that included only 54 patients with sphincter of Oddi dysfunction; results of these studies cannot be synthesized due to differences in populations and methodology. There was substantial variability in the reported performance characteristics of biliary scintigraphy.

Treatment outcomes of ERCP strategies compared to surgical or medical therapy:

• Two randomized controlled trials (total n=128) show that endoscopic sphincterotomy relieves pain in patients with pancreaticobiliary pain, sphincter of Oddi dysfunction, and elevated basal sphincter of Oddi pressure on manometry (greater than 40 mm Hg). The results of five single arm studies (total n=183) corroborate these data and suggest that patients with a dilated common bile duct and/or delayed contrast emptying may also benefit from endoscopic sphincterotomy.

• There is insufficient evidence to determine whether endoscopic sphincterotomy improves outcomes in patients with normal manometry findings. For this group, the small studies included in this review do not report significant improvements in pain with endoscopic sphincterotomy.

Topic 5: What patient, procedure, or operator factors are determinants of complications of ERCP?

• Thirteen studies reported on multivariable logistic regression analyses of factors associated with complications of ERCP. The four largest studies each included more than 1,800 patients, and the total number of complications observed in these studies ranged from 98 to 229. Overall, the methodologic quality of the available analyses is limited by over-fitting, i.e., testing an excessive number of factors relative to the number of complications observed. Consequently, this literature is exploratory in nature. Reported magnitudes of association are not reliable, significant independent variables may have been overlooked, and some significant associations may be misleading. Moreover, the existing studies do not use common, standardized definitions for the complications and factors of interest. Thus, caution should be used in drawing inferences for clinical practice from these studies.

• Patient, procedure, and operator factors were identified that were found to be significantly associated with complications in several of the more robust studies. Younger age (using various cutoffs, but generally 60 years or less) was significantly associated with total complications and with pancreatitis; as was suspected sphincter of Oddi dysfunction. Precut endoscopic sphincterotomy was the procedure-related factor most commonly associated with total complications or pancreatitis; a significant association with difficulty in cannulation was also reported, but less frequently. Multiple pancreatic contrast injections were associated with pancreatitis. For hemorrhage, the clearest association was patient factors related to coagulopathy. Case volume was the only operator-related factor found to be significantly associated with complications. These studies used various cutoffs to define lower volume centers: one or fewer procedures per endoscopist per week; fewer than 40 endoscopic sphincterotomies per endoscopist per year; and fewer than 150 procedures per year.

Epidemiology and Public Health Impact of Pancreaticobiliary Disease

Diseases of the pancreas and biliary tree are common in the United States population with various anatomic or acquired conditions resulting in a variety of obstructive, inflammatory, neoplastic, or functional conditions. An estimated 6 per 100,000 people are afflicted with common bile duct stones, representing only a small fraction of those with gallstones (WebMD/Lycos, 1999). Malignancy of the pancreas, gallbladder, or extrahepatic biliary tract represents approximately 57,400 newly diagnosed cases in the United States each year (Greenlee, Hill-Harmon, Murray, et al., 2001), and the associated prognosis is usually poor. Pancreatitis can occur in an acute, acute recurrent, or chronic pattern and may be associated with a variety of causes, with common etiologic factors including alcohol consumption and choledocholithiasis (Greenberger, Toskes, and Isselbacher, 1994).

In patients with persistent abdominal pain of suspected pancreaticobiliary origin, where no structural abnormality has been identified, functional disorders including sphincter of Oddi dysfunction may be present. Finally, complications of ERCP, such as pancreatitis, hemorrhage, infection, or intestinal rupture, occur in approximately 8% of patients undergoing ERCP depending on the case mix of diagnostic and therapeutic ERCP (Cotton, Lehman, Vennes, et al., 1991). Improving the understanding of risk factors for ERCP-related complications may improve patient selection or lead to improved methods of preventing complications in those at highest risk.

Endoscopic Retrograde Pancreatography (ERCP)

Patients with suspected pancreaticobiliary pathology require diagnostic assessment of the pancreaticobiliary tract to establish the correct diagnosis. Diagnostic assessment frequently includes imaging to detect the presence of dilation or narrowing of the ducts and to determine the cause of such morphologic changes.

Endoscopic retrograde pancreatography was first introduced for diagnostic evaluation of the pancreatic and biliary tree in the late 1960s. Using an endoscope inserted orally into the duodenum, a catheter can be placed into the biliary and/or pancreatic ducts for direct injection of radiographic contrast to provide X-ray images of the pancreaticobiliary ducts. Direct cholangiopancreatography can also be accomplished via a percutaneous transhepatic insertion of a needle or catheter with injection of radiographic contrast.

Noninvasive or less-invasive alternatives for imaging the pancreaticobiliary tree have been developed using magnetic resonance imaging, so-called magnetic resonance cholangiopancreatography (MRCP), ultrasound through an orally placed endoscope, so-called endoscopic ultrasonography (EUS), computed X-ray tomography often using specific biliary contrast agents, so-called computed tomography cholangiography (CTC), and nuclear medicine imaging with radiotracers specific to the biliary system, so-called biliary scintigraphy.

The endoscope used for ERCP can also be used selectively place catheters into the pancreaticobiliary ducts to obtain samples of pancreaticobiliary fluid or to deploy specialized tissue sampling devices (e.g., brush, fine-needle aspiration, forceps) to obtain cellular material for cytologic or histologic assessment. Alternative techniques for obtaining tissue samples for diagnosis include surgical biopsy, percutaneous fine-needle aspiration using imaging guidance, or endoscopic ultrasound guided fine-needle aspiration (EUS-FNA).

Once an accurate diagnosis has been established, surgical and nonsurgical treatment alternatives are frequently available. The ERCP scope permits access to the biliary tree to deliver endoscopic therapeutic interventions. Such interventions frequently include sphincterotomy of the sphincter of Oddi, which involves using an electrocautery device to cut and enlarge the opening of the pancreaticobiliary tract into the duodenum. Additional devices such as balloon catheters and specially designed wire baskets may be used to facilitate removal of duct stones, and specialized catheter insertion systems permit endoscopic placement of a variety of stents into the biliary or pancreatic ducts.

Secondary Search Strategy

The literature search for the supplemental question (Topic 1c), for the indirect comparison of single arm studies of for ERCP-guided fine needle aspiration (FNA) and EUS-guided FNA for Topic 2, and for additional studies selected by the secondary selection criteria for Topics 3 and 4, did not follow the same search process. The literature review process for these supplemental questions was based on a focused identification and selection of key articles addressing the clinical issue of interest. Reference lists from these articles, were then reviewed, focused MEDLINE searches were performed, and related articles identified. It was thought that this approach led to retrieval of the important studies addressing the questions of interest.

The Technical Advisory Group and individuals and individuals providing peer review also were asked to inform the project team of any studies relevant to the key questions addressed in this evidence report that were not retrieved by either of the search strategies.

Search Results

The online searches of the PubMed, EMBASE, BIOSIS, and SciSEARCH databases in conjunction with additional citations identified through manual searching yielded a total of 5,698 titles and abstracts for review. During application of Phase I of the selection process, 789 articles were selected for review in full text. Approximately 117 of these articles were identified as review articles. Primary and secondary selection criteria were applied to articles identified as potential clinical trial reports. This process yielded a total of 149 included studies for the review of evidence. Citations for the excluded articles and the reason(s) for exclusion are listed in Appendix A.

Primary Selection Criteria

The criteria which applied to all topic areas in this report were:

1. Full-length report in peer-reviewed medical journals.

2. Published in the English language.

3. Study reported outcomes relevant to this systematic review.

4. Where there were multiple reports of a single study, only the report judged to be most recent and complete, based on number of included patients and length of follow-up, was included. If additional relevant outcomes were included in the duplicate reports, these data were abstracted and added to the data from the primary report with citation to the supplementary articles.

5. Was prospective in design, or if retrospective, enrolled consecutive patients or with appropriate sampling methods (i.e. case-control sampling method).

For diagnostic performance topic areas, studies were included if the study:

1. Compared ERCP and at least one of the relevant diagnostic alternatives or compared two ERCP alternatives. Relevant diagnostic alternatives included endoscopic ultrasound, MRCP, intraoperative cholangiography, or other diagnostic tests as advised by the TAG. Studies reporting only non-breath hold MRCP imaging techniques were not included in this review as these do not represent the current state-of-the-art MRCP techniques.

2. Subjected all participants to both ERCP and the relevant diagnostic alternative;

3. Addressed a relevant patient population;

4. Included at least 25 subjects;

5. Reported sufficient information to be able to calculate 2x2 contingency tables of diagnostic performance.

For therapeutic outcome topic areas, studies were included if they:

1. Compared ERCP strategies with at least one of the relevant therapeutic alternatives. Relevant therapeutic alternatives included surgical methods to remove common ducts stones, surgical methods of bypassing malignant biliary obstructions, and surgical and medical methods of treating pancreatitis and pancreatitis-associated conditions.

2. Addressed a relevant patient population;

3. Included at least 25 subjects in each treatment group being analyzed separately; however, this criterion was relaxed to require 25 subjects in the trial for pancreaticobiliary malignancy and abdominal pain of possible pancreaticobiliary origin.

4. Reported on at least one relevant outcome measure;

5. Was a contemporaneous comparison study or if it was a noncontemporaneous study, the populations and treatment setting were comparable;

For Part V, a study was included if it:

1. Included an analysis of the relationship between patient, procedure, or operator factors and ERCP complications;

2. Enrolled at least 100 patients if a cohort study, or at least 25 cases if a case-control study;

3. Addressed potential confounding variables in either the selection of subjects or analysis.

For Part I, Section 3, a study was included if it:

1. Reported the association of individual risk factors of interest and the presence of a common bile duct stone. Based on a consensus from the TAG, these individual risk factors were jaundice, liver function test results, and an ultrasound finding of a dilated common bile duct.

2. Reported the association of a prediction rule or model predicting likelihood of having a common bile duct stone and the presence of a common bile duct stone;

3. Enrolled at least 100 patients;

4. Reported sufficient information to be able to calculate 2x2 contingency tables of diagnostic performance in the prediction of presence or absence of a common bile duct stone.

Secondary Selection Criteria

Due to a paucity of literature which met the primary selection criteria for Part III, Section 2 and Part IV, Section 2, additional selection criteria were created so that these questions could be examined. There was a lack of literature which provided comparative data on the value of ERCP treatment for these conditions. Thus studies were included from the primary search strategy and sought out using the secondary search strategy if the study was:

1. a randomized controlled trial or otherwise concurrently controlled study of an ERCP intervention compared to a relevant therapeutic alternative, regardless of sample size;

2. a single arm observational study (subject serves as own control) of ERCP intervention in treatment of chronic pancreatitis or chronic abdominal pain of possible pancreaticobiliary origin with a minimum size of 25 subjects; where the studies selected a well-defined population with a predictable natural history absent intervention based on thorough baseline evaluation; and where the study used an appropriate well-designed outcome measure. Baseline evaluation had to be obtained over a sufficient time period (approx. 3 months) and follow-up data needed be obtained over at least 6 months. Studies reporting exploration of subgroup differences in observed results were also included.

3. A single arm observational study of an ERCP intervention on pancreas divisum, subject to the above conditions in #2, but regardless of sample size.

In addition, there was an absence of direct comparative data for ERCP-guided fine needle aspiration (FNA) and EUS-guided FNA. Thus, an indirect comparison of single-arm studies was attempted. Studies of EUS-FNA that included at least 25 subjects for the evaluation of suspected pancreaticobiliary malignancy were identified and included.

Article Selection

Selection of articles was a two-stage process. All abstracts retrieved by the two search strategies were reviewed. First, titles and abstracts were reviewed using the primary and secondary study selection criteria. A single reviewer marked each citation as either: (1) eligible for review as full-text articles; (2) ineligible for full-text review; or (3) uncertain. Studies were excluded at this stage only if information revealed in the abstract showed that the study did not meet selection criteria. A second reviewer reviewed all citations marked as uncertain by the first reviewer, and a consensus decision was reached.

Using the primary and secondary study selection criteria, a single reviewer then reviewed the full-text article and determined whether selection criteria were met. The reviewer marked each full-text article as either (1) included in systematic review; (2) excluded from systematic review; or (3) uncertain. A second reviewer reviewed all articles marked as uncertain by the first reviewer, and a consensus decision was reached.

Records of the results of this evaluation were kept for each full-text paper retrieved including the reason for exclusion of each excluded study. Any disagreement about the inclusion or exclusion of a particular article was resolved by consultation with the Program Director or one or more members of the Technical Advisory Group.

Data Abstraction

Prior to the start of data abstraction, data elements were defined for abstraction from each selected article in consultation with the Technical Advisory Group. However, since some of the therapeutic key questions were not fully defined before articles were selected, many elements had to be defined based on the articles that ultimately met selection criteria. These data elements were abstracted from the articles that met final selection criteria. The data elements addressed:

1. Critical features of the study design (for example, patient inclusion/exclusion criteria, controlled or uncontrolled studies, randomized or non-randomized trials, number of subjects, or blinding, reference standard for diagnostic studies);

2. Treatment protocols;

3. The specified key outcomes.

For key questions assessing diagnosis, sensitivity, specificity, positive and negative predictive values, and prevalence of condition were all abstracted, including statistical analysis when available. Studies were grouped for presentation by categories according to diagnostic test, reference standard, clinically relevant patient subgroup, or other category of interest. For key questions assessing therapy, all outcomes that corresponded to the outcome categories that were specified in the protocol were abstracted, and studies were grouped by treatment alternative, clinically relevant patient subgroup, or other category of interest. Templates for evidence tables were then created in Microsoft Word.

Due to the anticipated heterogeneity in reported outcome measures, data were not abstracted into an electronic database. One reviewer performed primary data abstraction of all data elements into the evidence tables, and a second reviewer performed accuracy checks on the evidence tables. Disagreements were resolved between the two reviewers, or if necessary, consultation with the Program Director or relevant members of the Technical Advisory Group. If small differences occurred in quantitative estimates of data from published figures, the values abstracted independently by the two reviewers were averaged.

Good

Meets all criteria: Comparable groups are assembled initially and maintained throughout the study (follow-up at least 80 percent); reliable and valid measurement instruments are used and applied equally to the groups; interventions are spelled out clearly; all important outcomes are considered; and appropriate attention to confounders in analysis. In addition, for randomized controlled trials, intention to treat analysis is used.

Fair

Generally comparable groups are assembled initially but some question remains whether some (although not major) differences occurred with follow-up; measurement instruments are acceptable (although not the best) and generally applied equally; some but not all important outcomes are considered; and some but not all potential confounders are accounted for. Intention to treat analysis is done for randomized controlled trials.

Poor

Groups assembled initially are not close to being comparable or maintained throughout the study; unreliable or invalid measurement instruments are used or not applied at all equally among groups; and key confounders are given little or no attention. For randomized controlled trials, intention to treat analysis is lacking.

Quality criteria for diagnostic accuracy studies:

1. Enrollment of representative subjects. Appropriate spectrum of patients, unbiased enrollment, few eligible patients not enrolled, appropriate accounting of all potentially eligible subjects.

2. ERCP interpreted independently of diagnostic alternative.

3. Diagnostic alternative interpreted independently of ERCP.

Issues regarding the suitability and interpretation of different reference standards were not abstracted as quality measures but are discussed in each section of the report as needed. Study selection criteria required use of a reference standard in order to construct a 2 X 2 contingency table for diagnostic performance operating characteristics.

Summary ratings of diagnostic accuracy studies based on above criteria:

Good

Excellent documentation of prospective enrollment, identification and accounting of eligible and enrolled patients, few exclusions. Both ERCP and diagnostic alternative interpreted without knowledge of other test.

Fair

Had fair enrollment of patients, not too many exclusions, interprets reference standard independent of diagnostic test; and a good spectrum of patients, though reported details may have been incomplete.

Poor

Studies that had fatal flaws (e.g., Uses inappropriate reference standard; diagnostic test improperly administered; biased ascertainment of reference standard; very small sample size or very narrow selected spectrum of patients) were not eligible for inclusion in this systematic review. Thus, no included studies were assigned a Poor rating.

Quality Ratings for Multivariable Logistic Regression Analysis Studies

The most relevant criteria that provided discrimination of quality differences between studies were the degree of overfitting present in the multivariable models, the nature of statistical reporting, and the use of procedures to establish internal validity. Degree of overfitting was assessed using the ratio of the number of endpoints divided by the number of candidate variables in the model. Studies were classified as: Satisfactory, ratio > 10; Mild, ratio = 7 to <10; Moderate, ratio = 4 to <7; Severe, ratio <4. The nature of statistical reporting was considered satisfactory when the study reported both magnitude of effect estimates as well as associated confidence intervals or p-value for statistically significant findings. If either of these elements was not reported, studies were considered unsatisfactory. The degree of internal validity was evaluated by the use of procedures (e.g., test-validation split samples or bootstrapping) to guard against overfitting the model and spurious results.

Summary ratings of multivariable logistic regression analysis studies based on above criteria:

Good

Studies use procedures to guard against overfitting the model and spurious results; degree of overfitting is not severe for at least one analysis, and statistical reporting is satisfactory.

Fair

degree of overfitting is not severe for at least one analysis, and statistical reporting is satisfactory, but no use of procedures to guard against overfitting the model and spurious results.

Fair Minus

severe degree of overfitting for all analyses

Part I, Section 1: Diagnostic Performance of ERCP In Detecting Common Bile Duct Stones -- Comparison With Alternatives

Evidence Base

Table 1. Quality Assessment

Table 1. Quality Assessment

Study Author, Year	Patient Enrollment	Diagnostic performance of ERCP determined without knowledge of other test results	Diagnostic Performance of other test(s) determined without knowledge of ERCP results	Summary Evaluation
MRCP
Demartines, Eisner, Schnabel et al., 2000	Prospective (n=70) Uncertain enrollment of consecutive patients	Yes	Yes	Good
Guibaud, Bret, Reinhold, et al., 1995	Prospective (n=126) Some exclusions because of no ERCP confirmation	Uncertain	Yes	Fair
Holzknecht, Gauger, Sackmann et al., 1998	Prospective (n=61) 61 of 66 eligible patients enrolled, all exclusions accounted for	Yes	Yes	Good
Lomas, Bearcroft, and Gimson 1999	Prospective (n=69) Consecutive patients enrolled, all exclusions accounted for	Yes	Yes	Good
Soto, Barish, Alvarez et al., 2000	Prospective (n=49) Consecutive patients enrolled, all exclusions accounted for	Yes	Yes	Good
Stiris, Tennoe, Aadland et al., 2000	Prospective (n=50) Consecutive patients enrolled	Yes	Yes	Good
Varghese, Farrell, Courtney et al., 1999	Prospective (n=100) Consecutive patients enrolled, all exclusions accounted for	Yes	Yes	Good
Sugiyama, Atomi, and Hachiya 1998	Prospective (n=97) Nonconsecutive enrollment, but stated to be arbitrary without known selection bias	Uncertain	Yes	Fair
Varghese, Liddell, Farrell et al., 2000	Prospective (n=191) 191 of out 256 consecutive patients enrolled, all exclusions accounted for	Yes	Yes	Good
Burtin, Palazzo, Canard et al., 1997	Prospective (n=68) Consecutive patients enrolled	Yes	Yes	Fair -- unorthodox reporting of data, uncertain of data
Endoscopic Ultrasound
Canto, Chak, Stellato et al., 1998	Prospective (n=64) 64 out of 70 consecutive patients enrolled, 6 refusals	Yes	Yes	Good
Dancygier and Nattermann 1994	Prospective (n=41) Unstated whether consecutive	Uncertain	Yes	Fair
Norton and Alderson 1997	Prospective (n=46) Unstated whether consecutive	Yes	Yes	Fair
Prat, Amouyal, Amouyal et al., 1996	Prospective (n=119) Consecutive patients recruited, exclusions and refusals accounted for	Yes	Yes	Good
Sugiyama and Atomi 1997	Prospective (n=142) Consecutive patients enrolled	Uncertain	Yes	Fair
Sugiyama and Atomi 1998	Prospective (n=35) Consecutive patients enrolled	Uncertain	Uncertain	Fair
Chak, Hawes, Cooper et al., 1999	Prospective (n=36) Consecutive patients enrolled	Yes	Yes	Good
CTC
Ishikawa, Tagami, Toyota et al., 2000	Prospective (n=45) Unstated whether enrollment truly consecutive, not full accounting of exclusions	Uncertain	Uncertain	Fair
Polkowski, Palucki, Regula et al., 1999	Prospective (n=52) Full accounting of enrolled and excluded consecutive patients	Uncertain	Yes	Fair
Soto, Velez, and Guzman 1999	Prospective (n=29) Uncertain consecutive enrollment	Yes	Uncertain	Fair
Jimenez Cuenca, del Olmo Martinez, Perez Homs et al., 2001	Prospective (n=40) 40 of 60 consecutive patients enrolled, 20 excluded due to scheduling	Yes	Yes	Good
Neitlich, Topazian, Smith et al., 1997	Prospective (n=51) 51 of 96 consecutive patients enrolled, all exclusions accounted for	Yes	Yes	Good
Soto, Alvarez, Munera et al., 2000	Prospective (n=51) 51 of 56 eligible consecutive patients enrolled, all exclusions accounted for	Yes	Yes	Good

The search and selection process yielded 10 studies on MRCP (total n=834), 9 studies on EUS (total n=601), and 6 studies with 7 sets of findings on CTC (total n=266). In addition to these studies reporting diagnostic performance specific to common duct stones, 2 studies on MRCP which reported only on overall detection of obstructive abnormalities (total n=121) are also presented here. Study quality assessment is outlined in Table 1.

Review of Evidence: MRCP Performance

Table 2. Studies of MRCP, choledocholithiasis outcome, (more...)

Table 2. Studies of MRCP, choledocholithiasis outcome, ERCP used as reference standard for all studies except Sugiyama, Atomi and Hachiya (1998)

Study	N	Population	Diagnostic test	Prev (%)	Sens (%)	Spec (%)	PPV (%)	NPV (%)	Comments
Demartines, Eisner, Schnabel et al., 2000	40	Patients with suspected CBD stones referred for ERCP	MRCP	48	100	90	90	100
Guibaud, Bret, Reinhold, et al., 1995	126	Patients with suspected CBD obstruction referred for ERCP	MRCP	25	81	98	93	94	10 patients with other methods for gold standard
Holzknecht, Gauger, Sackmann et al., 1998	61	Patients referred for ERCP	MRCP (on-site reading) MRCP (off-site independent reading)	21	92 85	96 93	86 79	98 96
Lomas, Bearcroft, and Gimson 1999	69	Patients with suspected CBD stones or stricture referred for ERCP	MRCP	13	100	97	100	97
Soto, Alvarez, Munera et al. 2000	51	Patients with suspected CBD stones referred for ERCP	MRCP	51	96	100	100	96	1 false-negative ERCP considered positive after stone found at sphincterotomy
Soto, Barish, Alvarez et al., 2000	49	Patients with suspected CBD stones referred for ERCP	MRCP fast Spin Echo     Reviewer 1     Reviewer 2 Single Section half-Fourier RARE     Reviewer 1     Reviewer 2 Multisection half-Fourier RARE     Reviewer 1     Reviewer 2	49	96 92 100 92 92 96	96 100 96 96 92 92	96 100 96 96 92 92	96 93 100 92 92 96
Stiris, Tennoe, Aadland et al., 2000	50	Patients with suspected CBD stones referred for ERCP	MRCP	68	88	94	97	81
Varghese, Farrell, Courtney et al. 1999	100	Patients with CBD obstruction referred for ERCP	MRCP	30	93	99	97	97	12 patients with gold standard of IOC or PTC included in analyses
Varghese, Liddell, Farrell et al., 2000	191	Patients with CBD obstruction referred for ERCP	MRCP	18	91	98	91	98	5 patients with gold standard of IOC or PTC included in analyses
ERCP findings confirmed
Sugiyama, Atomi, and Hachiya 1998	97	Patients with suspected CBD stones referred for ERCP	MRCP ERCP (ERCP findings confirmed)	35	91 100	100 100	100 100	95 100	Positive ERCP confirmed by sphincterotomy, negative ERCP not confirmed

Ten studies studying a total of 834 patients were selected which examined the performance of MRCP compared to ERCP for the diagnosis of common bile duct stones (Table 2). Nine of the studies used ERCP as the reference standard, and thus measure the concordance of the two techniques rather than the relative performance. Only one study (Sugiyama, Atomi, and Hachiya, 1998) confirmed positive tests and allowed a comparison between the two tests. All the studies were rated as good quality with the exception of Guibaud, Bret, Reinhold, et al. (1995) and Sugiyama, Atomi, and Hachiya (1998).

Seven of the 9 studies which use ERCP as a reference standard show high concordance between the two tests with both sensitivity and specificity being greater than 90 percent. Two studies showed lesser degrees of concordance (Guibaud, Bret, Reinhold, et al., [1995], sensitivity 81 percent specificity 98 percent, and Stiris, Tennoe, Aadland et al. [2000], sensitivity 88 percent and specificity 94 percent).

Sugiyama, Atomi, and Hachiya (1998) did the only study that confirms positive ERCP tests and allows a comparison between the two tests. In that study of 97 patients, ERCP had 100 percent sensitivity, and MRCP had 91 percent sensitivity. Specificity for both tests was 100 percent. This was the only study that analyzed sensitivity by subgroups of stone diameter. Sensitivity was 100 percent for stone diameters from 11-27 mm, 89 percent for stone diameter from 6-10 mm, and 71 percent for stone diameter between 3-5 mm.

Table 3. Studies of MRCP, mixed outcome including CBD (more...)

Table 3. Studies of MRCP, mixed outcome including CBD stones, stratified by reference standard

Study	N	Population	Diagnostic test	outcome	Prev (%)	Sens (%)	Spec (%)	PPV (%)	NPV (%)	Comments
ERCP findings confirmed
Adamek, Albert, Weitz et al., 1998	60	Referrals for ERCP with suspected CBD obstruction	MRCP ERCP	Any abnormality	78	89 91	92 92	98 98	71 75	Uncertain method of ascertaining reference standard
ERCP used as reference standard
Holzknecht, Gauger, Sackmann et al., 1998	61	Patients referred for ERCP	MRCP (on-site reading) MRCP (off-site reading)	Any abnormality	75	91 94	80 80	93 94	75 80

Two studies reporting on a total number of patients of 121 had a mixed category of outcomes that included common duct stones (Table 3). In the study by Adamek, Albert, Weitz et al. (1998), the abnormalities included benign and malignant strictures, cholangiocarcinoma and choledochal cyst in addition to common duct stones. MRCP had a sensitivity and specificity for detecting any abnormality of 89 percent and 92 percent, whereas ERCP had a sensitivity of 91 percent and 92 percent.

In the study by Holzknecht, Gauger, Sackmann et al. (1998), the abnormalities detected included common bile duct dilatation and stenosis, in addition to common duct stones. Only the concordance with ERCP was evaluated. According to an image interpretation performed on-site, the sensitivity was 91 percent and the specificity was 80 percent. An off-site interpretation showed similar results.

In conclusion, most of the evidence on MRCP allows only conclusions as to whether MRCP and ERCP are concordant, rather than which test is superior. Most studies show fairly good concordance, with sensitivities and specificities both higher than 90 percent. Evidence limited to one study may indicate that ERCP is slightly better than MRCP.

Review of Evidence: Endoscopic Ultrasound Performance

Table 4. Studies comparing ERCP to endoscopic ultrasonography, (more...)

Table 4. Studies comparing ERCP to endoscopic ultrasonography, ERCP findings confirmed except for one study (Sugiyama and Atomi, 1998)

Study	N	Population	Diagnostic test	Prevalence (%)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Comments
Prat, Amouyal, Amouyal et al., 1996	119	High suspicion of CBD stones, sphincterotomy candidates	EUS ERCP	66	94 90	98 100	99 100	89 84	Sphincterotomy and endoscopic exploration on all patients. Numbers differ from published report due to rounding errors in published report
Burtin, Palazzo, Canard et al., 1997	68	Patients with suspected CBD obstruction referred for ERCP	EUS ERCP	50	97 91	97 97	97 97	97 92	Unorthodox presentation of data in report, test characteristics calculated from text descriptions, technical failures counted as neg tests
Canto, Chak, Stellato et al., 1998	64	Patients with suspected CBD stones referred for ERCP	EUS ERCP	31	84 95	98 98	94 no report	93 no report	Actual numbers not reported, all values quoted from study. Positive ERCP confirmed with stone extraction, negatives with 12 mo clinical follow up
Norton and Alderson 1997	46	Patients with suspected CBD stones referred for ERCP	EUS ERCP	52	88 79	96 92	95 90	89 83	Positive ERCP and EUS confirmed by sphincterotomy, no confirmation of negative ERCP and EUS
Dancygier and Nattermann 1994	41	Patients with obstructive jaundice, referred for ERCP	EUS ERCP	39	94 100	100 100	100 100	96 100	Positive ERCP confirmed by sphincterotomy, no apparent confirmation of negative ERCP
Polkowski, Palucki, Regula et al., 1999	50	Patients referred for ERCP for suspected CBD stones	EUS ERCP	68	91 91	100 100	100 100	84 84	Positive ERCP confirmed by sphincterotomy, selective confirmation of negative ERCP
Sugiyama and Atomi 1997	142	Patients referred for ERCP for suspected CBD stones	EUS ERCP	36	96 100	100 100	100 100	98 100	Positive ERCP confirmed by sphincterotomy, no apparent confirmation of negative ERCP
Chak, Hawes, Cooper et al., 1999	36	Patients with suspected acute biliary pancreatitis	EUS ERCP	33	91 92	100 87	100 79	95 94	Positives for either test confirmed with sphincterotomy, negatives not confirmed
ERCP + sphincterotomy as ref standard
Sugiyama and Atomi 1998	35	Patients with suspected acute biliary pancreatitis	EUS	43	100	100	100	100	ERCP reference standard, but positive ERCP confirmed with stone removal

There are 9 studies (total n=601) reporting on the capability of endoscopic ultrasound to diagnose common duct stones compared to ERCP (Table 4). In all the studies except 1 (Sugiyama and Atomi, 1998), positive tests of either method were confirmed with sphincterotomy, allowing for inferences regarding comparative performance. The study by Prat, Amouyal, Amouyal et al. (1996) stands out in this regard by subjecting all patients to sphincterotomy and endoscopic exploration, and thus is the only study in this whole section examining common bile duct stones with a truly independent reference standard. Chak, Hawes, Cooper et al. (1999) and Canto, Chak, Stellato et al. (1998) were also rated as "good" quality studies.

Given the small differences in performance noted in most of the studies, none of the studies is likely to detect statistically significant differences in test performance. In three of the studies, the sensitivity of EUS was higher than ERCP (Prat, Amouyal, Amouyal et al., 1996, Norton and Alderson 1997; Burtin, Palazzo, Canard et al., 1997). In three studies, the sensitivity of ERCP was higher than EUS (Canto, Chak, Stellato et al., 1998; Dancygier and Nattermann 1994, Sugiyama and Atomi, 1997) and in the two other studies the sensitivities were within 1 percent (Polkowski, Palucki, Regula et al., 1999; Chak, Hawes, Cooper et al., 1999). The specificities were very close in all studies except Chak, Hawes, Cooper et al. (EUS 100 percent, ERCP 87 percent).

Although most of the studies are small, within the limits of the evidence available, it appears that EUS is similar to ERCP in the detection of common bile duct stones.

Review of Evidence: CTC Performance

Table 5. Studies comparing CTC to ERCP, stratified (more...)

Table 5. Studies comparing CTC to ERCP, stratified by reference standard and presence and by type of contrast

Study	N	Population	Diagnostic test	Prev (%)	Sens (%)	Spec (%)	PPV (%)	NPV (%)	Comments
ERCP used as reference standard (No biliary contrast)
Soto, Alvarez, Munera et al., 2000	51	Patients referred for ERCP for suspected CBD stones	CTC	51	65	84	81	70
ERCP used as reference standard (Oral biliary contrast)
Soto, Alvarez, Munera et al., 2000	51	Patients referred for ERCP for suspected CBD stones	CTC with oral biliary contrast	51	92	92	92	92
Soto, Velez, Guzman et al. 1999	29	Patients referred for ERCP for suspected CBD stones	CTC with oral biliary contrast Observer 1 Observer 2	48	93 86	100 100	100 100	94 88
ERCP findings confirmed (independent reference standard) IV biliary contrast
Ishikawa, Tagami, Toyota et al., 2000	45	Laparoscopic patients undergoing routine preoperative ERCP	CTC with IV biliary contrast ERCP	16	71 100	95 100	71 100	95 100	Positive ERCP apparently confirmed during cholecystectomy, negative ERCP unlikely to be confirmed
Polkowski, Palucki, Regula et al., 1999	50	Patients referred for ERCP for suspected CBD stones	CTC with IV biliary contrast ERCP	68	85 91	88 100	94 100	74 84	Positive ERCP confirmed by sphincterotomy, selective confirmation of negative ERCP
No biliary contrast, ERCP + sphincterotomy findings used as reference standard
Jimenez Cuenca, del Olmo Martinez, Perez Homs et al., 2001	40	Patients referred for ERCP for suspected CBD stones	CTC	50	80	100	100	83	ERCP reference standard based on image and/or sphincterotomy findings, not only images
Neitlich, Topazian, Smith et al., 1997	51	Patients referred for ERCP for suspected CBD stones	CTC	33	88	97	94	94	ERCP reference standard based on image and/or sphincterotomy findings, not only images

Seven sets of findings report the diagnostic characteristics of CTC compared to ERCP for the diagnosis of common bile duct stones (Table 5). The studies varied considerably in the reference standard used. Three studies used ERCP as a reference standard, 2 studies used an independent reference standard, and 2 studies used ERCP and sphincterotomy findings as a reference standard. Three variations of CTC were used -- no biliary contrast (3 studies, total n=142), intravenous biliary contrast (2 studies, total n=95) and oral contrast (2 studies, total n=80). This results in a body of literature in which, at most, 2 studies share the same CT technique and reference standard. The studies by Jimenez Cuenca, del Olmo Martinez, Perez Homs et al. (2001), Neitlich, Topazian, Smith et al. (1997), and Soto, Alvarez, Munera et al. (2000) were rated as "good" quality.

Three sets of findings from 2 studies, all from the same principal author (Soto, Velez, Guzman et al., 1999 and Soto, Alvarez, Munera et al., 2000), used ERCP images as the reference standard. Soto, Alvarez, Munera et al. (2000, n=51), which used no biliary contrast, showed poor concordance with ERCP (sensitivity 65 percent and 84 percent specificity). The other two sets of findings (Soto, Velez, Guzman et al., 1999, n=29 and Soto, Alvarez, Munera et al., 2000, n=51), found higher concordance with ERCP when using oral biliary contrast (sensitivities and specificities both greater than 90 percent).

Two studies (Ishikawa, Tagami, Toyota et al., 2000, n=45 and Polkowski, Palucki, Regula et al., 1999, n=50) examined CTC with IV biliary contrast, and both studies used methods where ERCP findings were confirmed. In both studies ERCP was more sensitive and specific than CTC (Ishikawa, Tagami, Toyota et al., 2000, ERCP 100 percent sensitivity, 100 percent specificity, CTC 71 percent sensitivity, 95 percent specificity; Polkowski, Palucki, Regula et al., 1999, ERCP 91 percent sensitivity, 100 percent specificity, CTC 85 percent sensitivity, 88 percent specificity).

Finally, the two studies that use ERCP sphincterotomy results as the reference standard (Jimenez Cuenca, del Olmo Martinez, Perez Homs et al., 2001, n=40 and Neitlich, Topazian, Smith et al., 1997, n=51) showed sensitivities of 80 percent and 88 percent, respectively, and specificities of 100 percent and 97 percent. A direct comparison to ERCP cannot be done with these data, but these sensitivities are lower than generally has been shown for ERCP.

In conclusion, most studies show a fair concordance with ERCP diagnosis of common bile duct stones, but in studies which allow a determination of which test is superior ERCP seems to have better sensitivity and specificity. However, no estimate of the magnitude of this superiority can be made from this evidence.

Part I, Section 2: Outcomes of Treatment Using ERCP for Common Bile Duct Stones - Comparison of Strategies Using ERCP, Surgery, or Medical Management

Evidence Base

Table 6. Quality Assessment

Table 6. Quality Assessment

Study Author, Year	Comparable Initial Groups?	Comparable Groups Maintained?	Comparable Performance of Intervention?	Comparable Measurement of Outcomes?	Appropriate Analysis	Summary Evaluation
Cuschieri, Lezoche, Morino et al., 1999	RCT (n=300) Good comparability - computerized randomization - comparable characteristics	31 patients not treated according to random allocation, reported separately	Adequate for comparison	Adequate outcome measures used.	Those treated to assigned treatment reported as principal findings. Patients not treated by assigned treatment reported separately.	good
Rhodes, Sussman, Cohen et al., 1998	RCT (n=80) Uncertain comparability - randomization technique unknown - limited data on comparability	All patients retained for analysis	Adequate for comparison	Outcomes were not assessed blindly Uncertain how morbidity rates determined	All retained patients analyzed	Good
Chang, Lo, Stabile et al., 2000	RCT (n=59) Good comparability - sealed envelope randomization - comparable characteristics	All patients retained for analysis	Adequate for comparison	Outcomes were not assessed blindly Definition of morbidity not provided	All retained patients analyzed	Good
Targarona, Ayuso, Bordas et al., 1996	RCT (n=98) Good comparability - stratified randomization with sealed envelopes - patient characteristics comparable	2 out of 100 patients excluded because of incorrect randomization	Adequate for comparison	Outcomes were not assessed blindly Short-term morbidity rates do not capture difference in invasiveness between treatments	All patients retained for short-term outcomes analysis 89/93 surviving patients retained for long term outcomes analysis	Good
Trias, Targarona, Ros et al., 1997	Prospective study with historical control group (n=110) Good comparability Patient characteristics comparable	All patients prospectively identified as eligible enrolled	Surgical arm may include endoscopic sphincterotomy, more intensive treatment	Outcomes were not assessed blindly Short-term morbidity rates do not capture difference in invasiveness between treatments	All patients retained for short-term outcomes analysis 99/105 surviving patients retained for long term outcomes analysis	Fair
Hammarstom, Holmin, Stridbeck et al., 1995	RCT (n=80) Good comparability -- random numbers -- patient characteristics comparable	All potential patients accounted for, few refusals	Adequate for comparison	Outcomes not systematically defined or enumerated	Adequate follow up	Poor, most results could not be tabulated
Lai, Mok, Tan et al., 1992	RCT (n=82) Good comparability -- randomized by consecutive envelopes -- patient characteristics comparable	82 of 96 patients with severe acute cholangitis enrolled	Adequate for comparison	Outcomes were not assessed blindly Complication rates do not capture difference in invasiveness between treatments	All patients retained for analysis	Good
Leese, Neoptolemos, Baker et al., 1986	Retrospective observational study (n=82) Not very comparable Patients undergoing ERCP older, greater numbers of risk factors	Not applicable-retrospective study	Adequate for comparison	Outcomes were not assessed blindly	Analysis does not take into account difference in risk factors	Poor
Adamek, Maier, Jakobs et al., 1996	Retrospective observational study (n=145) Fair comparability Patients comparable on all measured characteristics	Not applicable-retrospective study	Adequate for comparison	Outcomes were not assessed blindly	Simple unadjusted comparisons	Fair/poor
Neuhaus, Zillinger, Born et al., 1998	RCT (n=60) Good comparability -- randomization technique unknown -- patients comparable on all measured characteristics	All patients retained for analysis	Adequate for comparison	Outcomes were not assessed blindly	All patients retained for analysis	Good
Bergman, Rauws, Fockens et al., 1997	RCT (n=202) Good comparability -- blinded computer-generated randomization -- patients comparable on all measured characteristics	16 out of 218 excluded after randomization because of ineligibility	Adequate for comparison	Outcomes were not assessed blindly	All patients retained for analysis	Good
Ochi, Mukawa, Kiyosawa et al., 1999	RCT (n=110) Good comparability -- randomization not described -- patients comparable on all measured characteristics	All patients retained for analysis	Adequate for comparison	Outcomes were not assessed blindly	All patients retained for short-term outcome analysis 105/110 patients retained for long-term outcome analysis	Good
Mavrogiannis, Liatsos, Romanos et al., 1999	RCT (n=153) Good comparability - randomization by sealed envelopes - Baseline characteristics similar for age, gender, presence of GB and gallstones	No cross-overs, drop outs reported.	Adequate for comparison.	Adequate outcome measures used. Outcomes were not assessed blindly.	Intention to treat analysis used.	Good
Chopra, Peters, O'Toole et al., 1996	RCT (n=86) Good comparability -- Randomization by sealed envelopes -- patients comparable on all measured characteristics	All patients retained for analysis	Adequate for comparison	Outcomes not blindly assessed Adequate for comparison	All patients analyzed for short term outcomes, 82/86 followed for long term outcomes	good

For the purposes of this evidence review, the literature remaining after selection criteria were applied was very thin and spread out over many different research questions. Generally, there was only one or at most, two, studies on a specific comparison of interest. Study quality assessment is outlined in Table 6.

Review of Evidence: ERCP with Laparoscopic Cholecystectomy to Remove Common Bile Duct Stones

Table 7. Preoperative versus Postoperative ERCP in (more...)

Table 7. Preoperative versus Postoperative ERCP in Cholecystectomy: Randomized Trials

Study	N	Population and Interventions	Outcomes	P	Adverse effects, complications	P	Resource utilization	P
Chang, Lo, Stabile et al., 2000	59	59 patients with mild to moderate gallstone pancreatitis, undergoing cholecystectomy after acute pancreatitis Mandatory preoperative ERCP (n=30) vs. selective postoperative ERCP (n=29) based on IOC findings	Stone Removal, successful ERCP/ERCP with stones: Preop ERCP: 12/12, 100% Postop ERCP: 7/7, 100%		Morbidity rates (not defined) Preop ERCP: 10% Postop ERCP: 10%	n.s.	Hospital stay: mean, median days Preop ERCP: 11.7,9.5 Post op ERCP: 9.0, 8 ICU days: mean, median Preop ERCP: 1.7, 1 Post op ERCP: 1.9, 1 Total Costs: Preop ERCP: $10,210 Postop ERCP: $8,586	.04 n.s. .049

Table 8. Preoperative ERCP versus Intraoperative cholangiogram (more...)

Table 8. Preoperative ERCP versus Intraoperative cholangiogram and laparoscopic common bile duct exploration in patients undergoing laparoscopic cholecystectomy in patients with suspected common bile duct stones, randomized trials

Study	N	Population and Interventions	Outcomes	P	Adverse effects, complications	P	Resource utilization	P
Cuschieri, Lezoche, Morino et al., 1999	269	Patients with suspected CBD stones needing cholecystectomy Preoperative ERCP (n=136) versus IOC and laparoscopic CBD exploration (n=133) as initial strategies for removing stones	Stone clearance: Preop ERCP: 84% IOC, LCBDE: 84%	n.s.	Conversion to open cholecystectomy: Preop ERCP: 6% IOC, LCBDE: 13% Overall morbidity: Preop ERCP: 12.8% IOC, LCBDE: 15.8% Mortality: Preop ERCP: 1.5% IOC, LCBDE: 0.75%	.08 n.s. n.s.	Hospital stay, mean days: Preop ERCP: 9 IOC, LCBDE: 6	<.05

Table 9. Postoperative ERCP versus laparoscopic exploration (more...)

Table 9. Postoperative ERCP versus laparoscopic exploration of common bile duct in patients with common duct stones found on intraoperative cholangiography, randomized trials

Study	N	Population and Interventions	Outcomes	P	Adverse effects, complications	P	Resource utilization	P
Rhodes, Sussman, Cohen et al., 1998	80	80 patients with CBD stones found on cholangiography during cholecystectomy Laparoscopic CBD exploration (LCBDE) (n=40) versus postoperative ERCP (n=40)	Initial clearance of CBD stones: LCBDE: 75% Postop ERCP: 75% Final clearance of CBD stones: LCBDE: 100% Postop ERCP: 93%	n.s. n.s.	Overall Morbidity: LCBDE: 18% Postop ERCP: 15%	n.s.	Hospital stay, median days: LCBDE: 1 Postop ERCP: 3.5	<.01

Three randomized controlled trials enrolling a total of 289 patients compared alternative strategies for removal of common bile duct stones in patients undergoing laparoscopic cholecystectomy (Tables 7-9). Although all 3 trials were judged to be of good quality, the evidence is limited because there is only a single study addressing each comparison of interest. Each trial reported on a different comparison, with respect to both the procedures compared and the patient population selected.

Overall, both arms in each of these 3 studies reported similar rates of stone clearance and morbidity, although morbidity was not well defined in two of these trials (Chang, Lo, Stabile et al., 2000; Rhodes, Sussman, Cohen et al., 1998). Thus, the main outcome of interest is relative resource utilization for each pair of alternative strategies for stone removal.

Mandatory Preoperative ERCP versus Selective Postoperative ERCP

Chang, Lo, Stabile et al. (2000) randomized 59 patients undergoing cholecystectomy during recovery from acute gallstone pancreatitis. Selective postoperative ERCP was based on findings from intraoperative cholangiogram. Resource utilization was lower in the selective postoperative ERCP group as measured by mean total hospital stay (9.0 vs. 11.7 days, p=0.04), and total costs ($8,586 vs. $10,210, p=0.049)

Review of Evidence: ERCP Sphincterotomy alone versus Definitive Surgery for suspected common duct stones

Patients at High Surgical Risk

Table 10. Endoscopic sphincterotomy alone versus open (more...)

Table 10. Endoscopic sphincterotomy alone versus open cholecystectomy in high risk surgical patients as primary treatment for common bile duct stones, randomized trials

Study	N	Population and Interventions	Outcomes	P	Adverse effects, complications	P	Resource utilization	P
Targarona, Ayuso, Bordas et al., 1996	98	Surgical high risk patients presenting with symptoms consistent with CBD stones Endoscopic sphincterotomy only (n=50) versus open cholecystectomy and CBD exploration if necessary (n=48)	Initial failure of treatment: ES: 12% Surgery: 6% Immediate mortality: ES: 6% Surgery: 4%	0.3 .5	Immediate morbidity: ES: 16% Surgery: 23% LONG TERM Biliary complications: ES (n=46): 21% Surgery(n=43): 6% Readmissions: ES: 23% Surgery: 4% Cholecystectomy: ES: 15% Surgery: 0% Need for sphincterotomy: ES: 2% Surgery: 4%	0.4 .04 .01 .01 .9	Post-treatment length of stay, mean days: ES: 5 Surgery: 11	.001

Table 11. Endoscopic sphincterotomy alone versus laparoscopic (more...)

Table 11. Endoscopic sphincterotomy alone versus laparoscopic cholecystectomy (with or without preoperative ERCP) in high risk surgical patients as primary treatment for common bile duct stones, observational studies

Study	N	Population and Interventions	Outcomes	P	Adverse effects, complications	P	Resource utilization	P
Trias, Targarona, Ros et al., 1997	110	Surgical high risk patients presenting with symptoms consistent with CBD stones Endoscopic sphincterotomy only (n=50) versus laparoscopic cholecystectomy and with preoperative ERCP if necessary (n=60)	Initial failure of treatment: ES: 12% Surgery: 11% Immediate mortality: ES: 6% Surgery: 3%	n.s. 0.5	Immediate morbidity: ES: 16% Surgery: 18% LONG TERM Biliary complications: ES (n=46): 21% Surgery(n=53): 4% P Readmissions: ES: 23% Surgery: 2% P Need for reoperation: ES: 15% Surgery: 2%	n.s. <.04 <.01 <.01	Post-treatment length of stay, mean days: ES: 5 Surgery: 4.4	n.s.

One randomized, controlled trial (Targarona, Ayuso, Bordas et al., 1996) and an observational study derived from the Targarona trial (Trias, Targarona, Ros et al., 1997) addressed whether removal of common duct stones with endoscopic sphincterotomy alone has lower morbidity and mortality than approaches which also remove the gall bladder during initial treatment (Table 10 and Table 11). The population of interest is patients at high surgical risk if subjected to cholecystectomy. For patients at high surgical risk, there may be advantages to a nonsurgical approach for removing common duct stones during acute symptomatic episodes. However, there may be differences in long term outcome if the gall bladder is not removed. Study quality was judged to be "Good" for the Targarona, Ayuso, Bordas et al. (1996) trial, and "Fair" for the Trias, Targarona, Ros et al. (1997) study.

The Targarona and Trias studies included high-risk surgical candidates based on age, cardiac risk, and pulmonary disease. The technique used in the Targarona, Ayuso, Bordas et al. (1996) study may not be representative of current surgical practice as the investigators performed open cholecystectomy for the definitive surgery arm; only the observational study by Trias, Targarona, Ros et al. (1997) used laparoscopic cholecystectomy.

Targarona, Ayuso, Bordas et al. (1996; n=98) found that both groups had similar short-term treatment failure, mortality, and morbidity, but initial postoperative length of stay favored endoscopic sphincterotomy alone (5 versus 11 days, p<0.001). However, over the longer term, the cholecystectomy patients had fewer biliary complications (6 percent versus 21 percent, p=0.04) and fewer readmissions (4 percent versus 23 percent, p<0.01). Eventually, 15 percent of patients in the sphincterotomy group underwent cholecystectomy.

Trias and colleagues performed laparoscopic cholecystectomy with preoperative ERCP as needed in 60 high-risk patients, and compared outcomes the to endoscopic sphincterotomy arm of the Targarona, Ayuso, Bordas et al. (1996) trial. Short-term and long-term results were similar to the Targarona trial, but initial hospital length of stay no longer favored the endoscopic sphincterotomy group when compared to laparoscopic, rather than open, cholecystectomy.

Patients Not at High Surgical Risk

Table 12. Endoscopic sphincterotomy alone versus open (more...)

Table 12. Endoscopic sphincterotomy alone versus open cholecystectomy and CBD exploration in non-high risk surgical patients as primary treatment for common bile duct stones, randomized trials

Study	N	Population and Interventions	Outcomes	P	Adverse effects, complications	P	Resource utilization	P
Hammarstrom, Holman, Stridbeck et al., 1995	80	Patients presenting with CBD stones on ERCP with intact gall bladder Endoscopic sphincterotomy only (n=39) versus open cholecystectomy and CBD exploration if necessary (n=41)	Biliary outcomes not coherently tabulated		Biliary complications not coherently tabulated Deaths from non-biliary related disease ES: 30% Surgery: 10% 13 patients in ES group required cholecystectomy on follow up	0.02	Total hospitalization days, median ES: 13 Surgery: 16	NS

One randomized controlled trial by Hammarstrom, Holmin, Stridbeck et al. (1995) enrolled 80 patients with intact gallbladders diagnosed with common bile duct stones on ERCP (Table 12). Patients either received sphincterotomy alone or open cholecystectomy and common bile duct exploration. Patients were followed for 5 years.

The study does not coherently define and compare outcomes between treatment groups for the most part; rather, various post-procedure events are unsystematically enumerated, making it difficult to tabulate any overall sense of outcomes. Total hospital stay (short term and follow up stays) was compared between the groups and was not statistically significantly different (median stay, 13 days sphincterotomy, 16 days surgery, p=ns). Of patients who received sphincterotomy, 13 were subsequently treated with cholecystectomy, 4 urgently because of acute cholecystitis. The authors also noted that the death rate from non-biliary related causes was higher in the endoscopic sphincterotomy group (30 percent vs. 10 percent, p=0.02). The authors conclude that the two alternatives are equally effective in the long term, but that due to the difference in heart disease mortality surgery might be the better option.

Summary

The very limited available evidence shows that definitive treatment prevents long term recurrence of biliary symptoms, hospitalization, and need for further treatment. In high-risk patients as defined in these studies, definitive treatment can be performed with acceptable short term morbidity and equivalent mortality as sphincterotomy alone. Not all patients develop recurrent problems, so the choice of definitive treatment versus sphincterotomy alone involves the weighing of short term morbidity of treatment, be it sphincterotomy alone, open or laparoscopic surgery, against the probability of recurrent biliary symptoms.

Review of Evidence: ERCP versus surgery for patients with acute cholangitis

Table 13. Endoscopic drainage for treatment of acute (more...)

Table 13. Endoscopic drainage for treatment of acute cholangitis due to common bile duct stones, randomized trials

Study	N	Population and Interventions	Outcomes	P	Adverse effects, complications	P	Resource utilization	P
Lai, Mok, Tan et al., 1992	82	82 patients with acute severe cholangitis due to CBD stones diagnosed with diagnostic ERCP Nasobiliary drainage placed by ERCP (n=41) versus open CBD exploration (n=41)	Hospital mortality rate: ERCP: 10% Surgery: 32%	<.03	Overall complication rate: ERCP: 34% Surgery: 66%	>.05

Table 14. Sphincterotomy for treatment of acute cholangitis (more...)

Table 14. Sphincterotomy for treatment of acute cholangitis due to common bile duct stones, observational studies

Study	N	Population and Interventions	Outcomes	P	Adverse effects, complications	P	Resource utilization	P
Leese, Neoptolemos, Baker et al., 1986	71	Retrospective review of patients with acute cholangitis due to CBD stones Early sphincterotomy (n=43) versus early surgery (n=28)	30 day mortality ERCP: 5% Surgery: 21%	<.02	Total % of patients with complications: ERCP: 28% Surgery: 57%	N/A	Hospital stay, median days: ERCP: 20 Surgery 23	n.s.

Patients receiving ERCP had greater baseline medical risk factors than patients having surgery (2 vs. 1, P<.05)

Two studies compared of ERCP treatment to open surgery for patients with acute cholangitis due to common bile duct stones (Table 13 and Table 14). Lai, Mok, Tan et al. (1992) randomized 82 patients diagnosed with common bile duct stones by ERCP to endoscopic nasobiliary drainage or open common bile duct exploration. This study is from Hong Kong, where oriental cholangiohepatitis is a common cause of common duct stones, and may not generalize to populations with a different spectrum of disease. Leese, Neoptolemos, Baker et al. (1986) conducted a retrospective review of 43 patients treated with endoscopic sphincterotomy to 28 contemporaneous patients undergoing surgical decompression for relief of cholangitis.

The Leese, Neoptolemos, Baker et al. (1986) study was judged to be of poor quality due to imbalance of patient characteristics between groups.

Acute severe cholangitis is a condition of very high mortality, thus the important outcome is to reduce the acute mortality rate. Both studies show that short-term mortality from acute cholangitis is lower in the ERCP-treated group compared to open surgery. Lai, Mok, Tan et al. (1992) reported lower hospital mortality (10 percent versus 32 percent, p<0.05) in the group treated with endoscopic nasobiliary drainage. Despite prognostic factors favoring the open surgery group, Leese, Neoptolemos, Baker et al. (1986) found that mortality at 30 days was lower in the endoscopic sphincterotomy group (5 percent versus 21 percent, p<0.02).

Review of Evidence: Endoscopic lithotripsy vs. extracorporeal shock wave lithotripsy (ESWL) in stones not removable with standard endoscopic techniques

Table 15. Intracorporeal vs. extracorporeal lithotripsy (more...)

Table 15. Intracorporeal vs. extracorporeal lithotripsy for common bile duct stones, randomized trials

Study	N	Population and Interventions	Outcomes	P	Adverse effects, complications	P	Resource utilization	P
Neuhaus, Zillinger, Born et al. 1998	60	Patients with stones not removable with ERCP techniques due to impacted stones or inaccessable bile duct. 33 patients with endoscope access, 27 patients with percutaneous access Extracorporeal shock wave lithotripsy (ESWL) (n=30) versus intracorporeal laser lithotripsy (ILL) (n=30)	Bile duct clearance: ESWL: 73% ILL: 97%	<.05	Not formally enumerated, appeared to be mild		Treatment sessions needed, mean: ESWL: 3.0 ILL: 1.2 Duration of treatment, mean days: ESWL: 3.9 ILL: 0.9	<.001 <.001

Table 16. Intracorporeal vs. extracorporeal lithotripsy (more...)

Table 16. Intracorporeal vs. extracorporeal lithotripsy for common bile duct stones, observational studies

Study	N	Population and Interventions	Outcomes	P	Adverse effects, complications	P	Resource utilization	P
Adamek, Maier, Jakobs et al., 1996	125	Patients with stones not removeable with ERCP techniques due to large stone size, impaction, biliary stricture, inaccessable bile duct Extracorporeal shock wave lithotripsy (ESWL) (n=79) versus intracorporeal electrohydraulic lithotripsy (EHL) (n=46)	Fragmentation of stones: ESWL: 97% EHL: 93% Bile duct clearance: ESWL: 79% EHL: 74%	n.s. n.s.	Not formally compared between treatments		Treatment sessions needed, mean: ESWL: 2.0 EHL: 1.1 Hospital stay, mean days: ESWL: 13 EHL: 11	N/A N/A

Characteristics of patients, stone size, number of stones, stone location not statistically significantly different between treatment groups.

Two studies compared endoscopic lithotripsy techniques to extracorporeal shock wave lithotripsy (ESWL) in removing common bile duct stones that cannot be removed with standard endoscopic techniques (which includes mechanical lithotripsy) (Neuhaus, Zillinger, Born et al., 1998 and Adamek, Maier, Jakobs et al., 1996; Table 15 and Table 16). In these studies, successful removal of stones is the important outcome.

Neuhaus, Zillinger, Born et al. (1998) randomized 60 patients to ESWL or intracorporeal laser lithotripsy. Adamek, Maier, Jakobs et al. (1996) performed an observational comparison between ESWL (n=79) and intracorporeal electrohydraulic lithotripsy (n=46).

Neuhaus, Zillinger, Born et al. (1998), found that intracorporeal laser lithotripsy was more successful than ESWL in clearing the bile duct of stones (97 percent versus 73 percent, p<0.05). Adamek, Maier, Jakobs et al. (1996) found no significant difference between ESWL and electrohydrolic lithotripsy.

Review of Evidence: Endoscopic balloon dilation versus endoscopic sphincterotomy

Table 17. Endoscopic balloon dilation versus endoscopic (more...)

Table 17. Endoscopic balloon dilation versus endoscopic sphincterotomy for removal of bile duct stones, randomized trials

Study	N	Population and Interventions	Outcomes	P	Adverse effects, complications	P	Resource utilization	P
Bergman, Rauws, Fockens et al., 1997	202	Patients referred for ERCP for removal of CBD stones, stones visualized Balloon dilation and stone removal versus sphincterotomy and stone removal	Stone removal in one session: Balloon: 89% Sphincterotomy: 91% *9 patients in Balloon group required sphincterotomy to remove stones	n.s.	Early complications: Balloon: 17% Sphincterotomy: 24% Follow-up complications: Balloon: 18% Sphincterotomy: 23%	n.s. n.s.
Ochi, Mukawa, Kiyosawa et al., 1999	110	Patients referred for ERCP for removal of CBD stones, stones visualized, < 15 mm and less than 10 stones Balloon dilation and stone removal versus sphincterotomy and stone removal	Stone removal, final: Balloon: 93% Sphincterotomy: 98% Stone removal after initial procedure (before lithotripsy): Balloon: 78% Sphincterotomy: 94%	.36 .02	Early complications: Balloon: 2% Sphincterotomy: 6% Late complications: Balloon: 4% Sphincterotomy: 15%	n.s. n/a

Two randomized controlled trials (Bergman, Rauws, Fockens et al., 1997 and Ochi, Mukawa, Kiyosawa et al., 1999) compared endoscopic balloon dilation to endoscopic sphincterotomy for removal of common bile duct stones in a total of 312 patients (Table 17). Study quality was judged as "Good" for both trials.

Concern about possible long term effects of sphincterotomy on biliary function, plus concern about hemorrhage induced by sphincterotomy have led to consideration of dilation of the biliary sphincter as an alternative method to remove common bile duct stones. Dilation would potentially preserve the function of the biliary sphincter. However, concern has been raised that pancreatitis may occur more often as a complication after balloon dilation.

However, neither study assesses long term outcomes, so the only outcomes that can be assessed are success in removing common bile duct stones and early complications. Both studies found that although balloon dilation ultimately produces equivalent stone removal rates (Bergman, Rauws, Fockens et al., 1997, balloon 89 percent success, sphincterotomy 91 percent success; Ochi, Mukawa, Kiyosawa et al., 1999, balloon 93 percent success, sphincterotomy 98 percent). Some patients in the balloon treatment arm must either cross over or be subject to additional procedures such as mechanical lithotripsy to compensate for the lower initial success rate. Early complications and follow-up complications were not statistically significantly different in the Bergman, Rauws, Fockens et al. (1997) study. In the Ochi, Mukawa, Kiyosawa et al. (1999) study, early complications were not statistically different. Late complications were reported (balloon 4 percent, sphincterotomy 15 percent), but statistical significance tests were not reported.

DiSario, Freeman, Bjorkman et al., (1998) also completed a randomized controlled trial comparing balloon dilation to sphincterotomy, but this trial had only been reported in abstract form in 1998. The results of this study are summarized here because it is commonly cited in reviews and the findings on post-procedure pancreatitis are striking. In this randomized controlled trial of 240 patients, stone clearance was achieved in 99 percent of patients. However, morbidity occurred in 15 percent of balloon dilation patients and 4 percent of sphincterotomy patients (p=0.014) Most of the morbidity in the dilation group was due to moderate or severe pancreatitis which occurred in 4 patients and resulted in 2 deaths.

Review of Evidence: Needle-knife fistulotomy versus needle-knife precut papillotomy for the treatment of common bile duct stones in patients with difficult cannulations

Table 18. Needle-knife fistulotomy versus needle-knife (more...)

Table 18. Needle-knife fistulotomy versus needle-knife precut papillotomy for the treatment of common bile duct stones

Study	N	Population and Interventions	Outcomes	P	Adverse effects, complications	P	Resource utilization	P
Mavrogiannis, Liatsos, Romanos et al., 1999	153	Consecutive patients who required treatment of suspected choledocholithiasis who had difficulty achieving selective CBD cannulation were randomized to either needle-knife fistulotomy (NKF, n=74) or needle-knife precut papillotomy (NKPP, n=79). All patients had biochemical cholestasis and one or more of the following: biliary pain, bile duct cannulation, and gallbladder stones.	Cannulation success rates (overall): NKF=90.5% NKPP=88.6% Successful stone extraction without lithotripsy NKF (40/48) = 83% NKPP (45/46) =98% Overall stone extraction NKF =100% NKPP =100%	n.s. .05 n.s.	Comp (%): NKF NKPP Bleeding 6.75 5.06 Perforation 2.7 2.53 Cholangitis 1.35 0 Pancreatitis 0 7.59 Total 10.81 15.18 Hyperamylasemia 2.7 17.72 Death 0 1.26	n.s. n.s. n.s .05 n.s. .01 n.s.

Mavrogiannis, Liatsos, Romanos et al. (1999) performed a randomized, controlled trial (n=153) comparing two precutting techniques for cannulating the common bile duct when difficulty is encountered when trying to cannulate the common bile duct. (Table 18). Needle-knife fistulotomy (NKF) has been proposed as a safer method of precutting than traditional needle-knife precut papillotomy (NKPP), with the potential disadvantage of a smaller opening into the bile duct which may prevent successful stone removal.

Overall success in cannulating the common bile duct (after second attempts) was equivalent between the two techniques (NKF 91 percent, NKPP 89 percent, p=n.s.) Stone removal without use of lithotripsy was greater for NKPP than for NKF (98 percent versus 83 percent), but final stone removal rates were 100 percent for both groups. Overall complications were not statistically significantly different (NKF 11 percent, NKPP 15 percent, p=n.s.), but NKPP had a greater pancreatitis rate (7.6 percent versus 0 percent, p<0.05) and a higher rate of hyperamylasemia (17.7 percent versus 2.7 percent, p<0.01). Both methods appear to be similar in the management of patients with common bile duct stones.

Part I, Section 3: Diagnostic Value of Individual Risk Factors or Predictive Models for Assessing the Likelihood of Having a Common Bile Duct Stone

Review of Evidence: Univariate Risk Factors for Common Bile Duct Stones

The single risk factors commonly examined in studies included clinical jaundice or elevated bilirubin, liver function tests, and ultrasound findings of a dilated common bile duct. Studies varied in the definitions and cutoff thresholds for the various tests

Table 19. Jaundice or elevated bilirubin as a risk (more...)

Table 19. Jaundice or elevated bilirubin as a risk factor for CBD stone

Study	Population	% prevalence of stone in population	n	Rule tested	Predictive Value	Sensitivity	Specificity	Comments
Alponat, Kum, Rajnakova et al., 1997	Patients with risk factors for CBD stones having ERCP	32	192	jaundice	67	56	87
Barkun, Barkun, Fried et al., 1994	Patients undergoing lap cholecystectomy who had ERCP	48	139	bilirubin>1.8	57	48	48
Bergamaschi, Tuech, Braconier et al., 1999	Patients undergoing lap cholecystectomy	15	990	jaundice	76	24	99
Hauer-Jensen, Karesen, Nygaard et al., 1985	Patients undergoing cholecystectomy	12	319	jaundice bilirubin>1.5	29 42	26 45	91 91
Kim, Kim, Lee et al., 1997a	Patients undergoing lap cholecystectomy	17	561	jaundice bilirubin >2	52 53	36 41	93 92
Koo and Traverso 1996	Patients undergoing lap cholecystectomy	12	420	bilirubin>1.2	47	31	93
Menezes, Marson, Debeaux et al. 2000	Patients undergoing lap cholecystectomy	33	233	bilirubin>nl bilirubin>2xnl	95 92	48 31	98 99
Santucci, Natalini, Sarpi et al., 1996	Patients undergoing lap cholecystectomy	9	697	bilirubin>3	83	56	82
Trondsen, Edwin, Reiertsen et al., 1995	Patients undergoing lap cholecystectomy	38	599	jaundice	86	46	95

Five studies (total n=2,661) reported on clinical jaundice as a risk factor (Table 19). Positive predictive values ranged from 29 percent to 86 percent, sensitivity from 24 percent to 56 percent, and specificity from 87 percent to 99 percent. Clinical jaundice does not have an exact threshold cutoff value, nor is the reliability of measurement certain. In general, though, sensitivities are low, specificities are higher, and in the situation of a low prevalence condition such as common bile duct stones, the high specificity drives the predictive values to be high.

Six studies (total n=2369) reported on bilirubin levels. At varying cutoff levels, positive predictive values ranged from 42 percent to 95 percent, sensitivity from 31 percent to 56 percent, and specificity from 48 percent to 99 percent. In general, sensitivities were low, specificities higher, and the resulting positive predictive values are reasonably high.

Table 20. Elevated liver function tests as a risk factor (more...)

Table 20. Elevated liver function tests as a risk factor for CBD stone

Study	Population	% prevalence of stone in population	n	Rule tested	Predictive Value	Sensitivity	Specificity	Comments
Alponat, Kum, Rajnakova et al., 1997	Patients with risk factors for CBD stones having ERCP	32	192	Any LFT>2xnl AST > 2xnl ALT > 2xnl Alk phos >2xnl GGT > 2xnl LDH > 2xnl	37 41 40 43 35 38	84 89 87 84 87 68	33 40 38 46 22 46	Numbers for any LFT do not make sense, cannot be less sensitive
Barkun, Barkun, Fried et al., 1994	Patients undergoing lap cholecystectomy who had ERCP	48	139	AST>120 Alk phos>300	49 53	81 79	25 35
Bergamaschi, Tuech, Braconier et al., 1999	Patients undergoing lap cholecystectomy	15	990	Alk phos >400 and GGT>200	87	58	99
Hauer-Jensen, Karesen, Nygaard et al., 1985	Patients undergoing cholecystectomy	12	319	Alk phos>250	37	58	87
Kim, Kim, Lee et al., 1997a	Patients undergoing lap cholecystectomy	17	561	SGOT>50 SGPT>50 Alk phos>160	43 39 50	65 67 75	82 79 85
Koo and Traverso 1996	Patients undergoing lap cholecystectomy	12	420	SGOT>44 Alk phos>140	48 48	40 31	94 93
Menezes, Marson, Debeaux et al. 2000	Patients undergoing lap cholecystectomy	33	233	SGOT>nl SGOT>2xnl Alkphos>nl Alkphos>2xnl	88 93 77 97	47 35 66 44	97 99 90 99
Santucci, Natalini, Sarpi et al., 1996	Patients undergoing lap cholecystectomy	9	697	ALT> 40 AST> 40 GGT>150 Alk phos>300	88 76 75 94	94 78 80 72	79 78 76 90	Cutoffs established by ROC analysis, maximize sensitivity and specificity

Eight studies (total n=3,551) reported on various liver function tests (Table 20). Some studies examined more than 1 cutoff level. There was a broad range of predictive values, sensitivities and specificities for all the different liver function tests examined. In general, the trade off between sensitivity and specificity can be noted in all the studies. The studies with cutoff values that produce high specificity tend to have low sensitivity, but this type of cutoff produces the highest positive predictive values.

Table 21. Dilated CBD as a risk factor for CBD stone (more...)

Table 21. Dilated CBD as a risk factor for CBD stone

Study	Population	% prevalence of stone in population	n	Rule tested	Predictive Value	Sensitivity	Specificity	Comments
Alponat, Kum, Rajnakova et al., 1997	Patients with risk factors for CBD stones having ERCP	32	192	Dilated CBD with stone on ultrasound Dilated CBD without stone on ultrasound	72 36	42 31	92 74
Barkun, Barkun, Fried et al., 1994	Patients undergoing lap cholecystectomy who had ERCP	48	139	Dilated CBD, subjective	64	53	73
Bergamaschi, Tuech, Braconier et al., 1999	Patients undergoing lap cholecystectomy	15	990	CBD > 8mm	75	28	98
Hauer-Jensen, Karesen, Nygaard et al., 1985	Patients undergoing cholecystectomy	12	319	CBD >10 mm	34	63	92
Kim, Kim, Lee et al., 1997a	Patients undergoing lap cholecystectomy	17	561	CBD > 10 mm	61	94	88
Koo and Traverso 1996	Patients undergoing lap cholecystectomy	12	420	CBD> 5mm + 1 mm per decade over age 50	28	22	92
Menezes, Marson, Debeaux et al. 2000	Patients undergoing lap cholecystectomy	33	233	CBD dilated (not defined)	91	51	97
Santucci, Natalini, Sarpi et al., 1996	Patients undergoing lap cholecystectomy	9	697	CBD> 8 mm	74	59	72
Trondsen, Edwin, Reiertsen et al., 1998	Patients undergoing lap cholecystectomy	15	171	CBD > 6 mm	35	64	79
Trondsen, Edwin, Reiertsen et al., 1995	Patients undergoing lap cholecystectomy	38	599	CBD dilated (not defined)	85	31	96

Ten studies (total n=4,321) reported on the finding of a dilated common bile duct seen on ultrasound (Table 21). The threshold for a dilated duct varied from 5 to 10 mm, and was undefined in a few studies. Predictive values ranged from 28 percent to 91 percent, sensitivities from 28 percent to 94 percent, and specificities from 72 percent to 98 percent. Studies with high sensitivity tend to have low specificity, and vice versa.

In sum, although all the previously mentioned single risk factors for common duct stones have significant associations with the presence of stones, none of them have outstanding ROC characteristics. The presence of any of these factors certainly increases the probability of the presence of a common bile duct stone, possibly high enough to change clinical decision-making. However, changing the cutoff value to increase the positive predictive value (by increasing the specificity) usually results in poor sensitivity.

Review of Evidence: Multivariable Predictors for Common Bile Duct Stones

Table 22. Decision rules for prediction of stones

Table 22. Decision rules for prediction of stones

Study	population	% prevalence of stone in population	n	Rule tested	Predictive value	Sensitivity	Specificity	Comments
Hawasli, Lloyd, Pozios et al., 1993	Patients undergoing lap cholecystectomy	4	459	High suspicion combination	75	83	99
Menezes, Marson, Debeaux et al. 2000	Patients undergoing lap cholecystectomy	15	211	Score>= 2 Score>=3 Based on logistic regress	56 67	86 82	66 80
Trondsen, Edwin, Reiertsen et al., 1995	Patients undergoing lap cholecystectomy	38	599	Discriminant function	91	95	94	Rule applied to same data used to develop function
Trondsen, Edwin, Reiertsen et al., 1998	Patients undergoing lap cholecystectomy	17	192	Discriminant function	60	94	88	Same 2 by 2 data as Trondsen, Edwin, Reiertsen et al., 1995, above

Four studies (total n=1,461) examined the use of multiple risk factors for prediction of the presence of common bile duct stones (Table 22). Many studies that simply used the criterion of "any one risk factor" as a prediction rule were not included in this evidence review, as such a criterion has been used for many years to select patients for ERCP and has a known poor specificity and low positive predictive value.

The four studies varied in the analytic technique used to develop the prediction rule. Hawasli, Lloyd, Pozios et al. (1993) did not use any quantitative technique but defined combinations of risk factors to classify patients at high risk of stones. Menezes, Marson, Debeaux et al. (2000) developed a logistic model based on age, sex, jaundice, presence of cholangitis, liver function tests, and ultrasound examination of the common bile duct. Trondsen, Edwin, Reiertsen et al. (1995) used a discriminant analysis technique based on age, bilirubin, alanine aminotransferase, and gamma glutamyltransferase. In Trondsen, Edwin, Reiertsen et al. (1998), a new rule was not developed, but the previously developed discriminant analysis rule was prospectively validated in a new population of patients.

Thus, except for Trondsen, Edwin, Reiertsen et al. (1998), the findings of the three other studies should be viewed as optimistic estimates of stone prediction, since the performance of the rules was only evaluated on the set of patients used to develop the rule.

All the studies produced decision rules in which both the sensitivity and specificity were greater than 80 percent. However, these findings should be viewed cautiously, since there has been no independent validation. The prospective validation study by Trondsen, Edwin, Reiertsen et al. (1998) is a particularly strong finding, since the rule was derived from an independent population -- the sensitivity was 94 percent and the specificity was 88 percent in an independent set of patients. The discriminant function cutoff could be varied to increase sensitivity at the expense of specificity or vice-versa, but since both are high the actual discriminative capability of the rule compared to individual risk factors was far superior.

In conclusion, multivariable modeling of risk factors for prediction of common duct stones shows promise as a method of triage for determining appropriate treatments, given that they appear to have superior discriminatory power. These prediction models have yet to be integrated into clinical decision models to determine optimal cutoffs.

Review of Evidence: Absence of Any Risk Factor as A Predictor of Common Bile Duct Stone Absence

Table 23. Rules ruling out stones, absence of stone (more...)

Table 23. Rules ruling out stones, absence of stone is the outcome

Study	population	% prevalence of stones in population	n	Rule tested	Prevalence of stone in those ruled out by rule (1 - PPV)	Sensitivity--% of stone-free patients detected by rule	Specificity--% of patients with stones ruled out by rule	Comments
Carroll, Phillips, Rosenthal et al., 1996	Patients undergoing lap cholecystectomy	15	100	Normal LFTs, CBD, past history	4	61	87
Hawasli, Lloyd, and Cacucci 2000	Patients undergoing lap cholecystectomy	5	2834	Normal LFTs, CBD, past history	0.25	89	96	Hawasli, Lloyd, Pozios et al. 1993 results of this same question included in these data
Khaira, Ridings, and Gompertz 1999	Patients undergoing lap cholecystectomy	5	154	Normal LFTs, CBD, past history	1	60	88
Koo and Traverso 1996	Patients undergoing lap cholecystectomy	12	420	Normal LFTs, US, past history	7	78	60
Santucci, Natalini, Sarpi et al., 1996	Patients undergoing lap cholecystectomy	9	697	Normal LFTs, US, past history	1.4	98	86	Clinical followup to detect stones in patients with no indications
Trondsen, Edwin, Reiertsen et al., 1998	Patients undergoing lap cholecystectomy	17	192	Discriminant function value negative	1.4	88	94	Rule applied to validation set of patients
Trondsen, Edwin, Reiertsen et al., 1995	Patients undergoing lap cholecystectomy	38	599	Discriminant function value negative	3	94	95	Rule applied to same data used to develop function

Seven studies (total n=599) examined the prediction of absence of common duct stones (Table 23). Usually, the absence of any of the known risk factors (all the individual factors reviewed previously) was used as the indicator. Trondsen, Edwin, Reiertsen et al. (1995) and Trondsen, Edwin, Reiertsen et al. (1998) reviewed previously, are also included here because the discriminant function used to predict stones can also be used to predict the absence of stone.

If the prevalence of stone is low enough in some patients, then some clinicians might avoid use of any diagnostic test to diagnose common duct stones. Such a case would be very compelling if the probability of stone is in the same range or lower as it is in the case of a negative ERCP examination. Although ERCP is selectively performed on patients with higher risk of common duct stones, if physicians are willing to believe a negative ERCP, they should be willing to believe a prediction rule if the probabilities of stones are equally low.

The seven studies reported a probability of common duct stones in those predicted not to have stones between a range of 0.25 percent to 7 percent. In all studies, a reasonable sensitivity for stone-free patients was shown, from 60 percent to 98 percent, and reasonable specificity, 60 percent to 96 percent. Thus, the decision rules all can identify more than half of the patients that do not have stones.

The strongest finding is Trondsen, Edwin, Reiertsen et al. (1998), in which the same discriminant function which identifies stones can rule out stones with both high sensitivity (88 percent) and specificity (94 percent). This study is also a validation study of an independently developed discriminant function, which further increases its validity.

These probabilities of stones compare quite favorably to the probabilities of stones in patients having a negative ERCP. If the probability is calculated, using the equation "1-NPV" and some of the reported NPVs of the ERCP studies in the section of this report comparing ERCP to EUS, a range of stone probabilities is calculated from 0 percent to 17 percent.

In conclusion, the absence of any risk factors for stones (or a discriminant function indicating absence of stone) is a very strong predictor of the absence of stones, producing probabilities of stones that are in the same range as a negative ERCP exam in a patient with risk factors for stones.

Part II, Section 1: Diagnostic Performance of Nonsurgical Tissue Sampling Techniques in Pancreaticobiliary Malignancy -- Comparison of Strategies Using ERCP, EUS, or Percutaneous Approach

Evidence Base

Table 24. Quality Assessment

Table 24. Quality Assessment

Study Author, Year	Patient Enrollment	Diagnostic performance of ERCP determined without knowledge of other test results	Diagnostic Performance of other test(s) determined without knowledge of ERCP results	Summary Evaluation
Jaiwala, Fogel, Sherman et al., 2000	(n=133 pts) Prospective Study Enrollment of subjects stated to be selected and nonconsecutive and reasons for exclusion were stated.	No	No	Fair
Kurzawinski, Deery, Dooley et al., 1993	(n=46 pts) Prospective study of 37 of 46 consecutive pts w/ biliary tract stricture had ERCP and 9 had PTC cytology. Reasons for exclusions provided.	No	No	Fair
de Peralta-Venturina, Wong, Purslow et al., 1996	(n=74 pts; 104 spec) Retrospective review of all eligible cytology specimens during 1990 to mid 1994 in pts with verified diangosis.	Yes	Yes	Good
Foutch et al. 1991	(n=30 pts; 78 specimens) Prospective study 30 consecutive patients with bile duct stricture	Yes	Yes	Good
Mansfield et al. 1997	(n=43 pts; 54 procedures) Prospective study All pts with biliary stricture suspicious for malignancy	Yes	Yes	Good
Sugiyama, Atomi, Wada et al., 1996	(n= 43 pts) Prospective study 52 Consecutive pts with stricture (n=48) or filling defect (n=4) Papillary lesions excluded. Analysis includes 43 pts with all 3 techniques	No	No	Fair
Howell, Beveridge, Bosco et al., 1992	?Prospective 31 consecutive patients with malignant appearing strictures	No	No	Fair
Ferrari, Lichtenstein, Slivka et al., 1994	(n=74) Retrospective study of all pts who had ERCP with brush cytology of biliary or pancreatic duct stricture	No	No	Fair
Ponchon, Gagnon, Berger et al., 1995	(n=193) Prospective study Enrolled subjects meeting entry criteria. Complete explanation of enrollment process provided.	Yes	Yes	Good
Schoefl, Haefner, Wrba et al., 1997	119 consecutive pts (133 samples) ?retrospective	No	No	Fair
Pugliese, Antonelli, Vincenti et al., 1997	(n=52) Prospective enrollment of consecutive biliary strictures at ERCP Excluded strictures associated with bile duct stones, periampullary tumors, or postop stricture	Yes	Yes	Good
Gmelin and Weiss 1981	(n=32) 32 proven malignant or benign tumors in papillary region out of 36 consecutive cases.	Uncertain	Uncertain	Fair

Twelve studies comparing at least two tissue sampling techniques were identified in this systematic review. Quality ratings are displayed in Table 24. Five of these studies were rated as "Good" quality, signifying the use of blinded interpretation of test results. Only three studies include over 100 patients, and six studies include less than 50 subjects.

There is considerable variation in reported estimates of sensitivity for each tissue sampling technique, and comparison of results for the same technique across studies may be limited due to differences in populations with regard to distribution of tumor types as well as differences in tissue sampling technique and interpretation methods. To minimize this problem, this analysis will focus primarily on within-study comparisons of the relative sensitivity of alternative sampling techniques. However, this problem is not completely avoided because the selected comparative studies frequently reported diagnostic performance for individual sampling techniques being compared on a different number of patients and thus slight differences in the population characteristics may be present.

Given that the expected difference in diagnostic performance between tissue sampling techniques and the diagnostic alternatives reported here are frequently relatively small and the number of cases with the outcome of interest is generally small, these studies may have limited power to detect statistically significant differences in test performance. Only 4 of 12 studies (Jaiwala, Fogel, Sherman et al., 2000; Sugiyama, Atomi, Wada et al., 1996; Ponchon, Gagnon, Berger et al., 1995; Kurzawinski, Deery, Dooley et al., 1993) actually reported any statistical comparisons, and all of these only reported chi square comparisons of sensitivity.

The specificity estimates for cytology techniques reported in these studies were generally close to 100%, though Jaiwala, Fogel, Sherman et al. (2000; n=133) found that specificity fell to 90% when any atypia was considered equivalent to malignancy.

The nonsurgical tissue sampling techniques being evaluated in these studies are measured against a reference standard incorporating the best available information from surgical findings, surgical or nonsurgical pathology, autopsy, imaging follow-up, and clinical follow-up.

Review of Evidence: Diagnostic Performance

Bile Aspiration Cytology Compared to Brush Cytology

Table 25. Comparisons of Bile Cytology and Brush Cytology (more...)

Table 25. Comparisons of Bile Cytology and Brush Cytology

Study	N Pts	N Spec	Diagnostic test	Prevalence (%)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Adequate Specimens (%)	Quality Rating and Comments
Kurzawinski, Deery, Dooley et al., 1993	37 31	37 31	ERCP-Bile cytology ERCP-Brush cytology	81 77	33a 71b	100 100	100 100	26 50		Fair p< 0.05 a vs. b p< 0.01 c vs. d
	9 15	9 15	PTC-Bile cytology PTC-Brush cytology	?	0c 67d	n.r. n.r.
de Peralta-Venturina, Wong, Purslow et al., 1996	74	13 61	Bile cytology Brush cytology 10	? ?	50 100	100 95	100 95	40 100	69 98	Good Stratified results for bile vs. brushing not reported by ERCP vs. PTC technique
		55 19	ERCP PTC	? ?	100 43	95 100	96 100	100 57	98 79

Table 26. Comparisons of Bile Cytology, Brush Cytology, (more...)

Table 26. Comparisons of Bile Cytology, Brush Cytology, and Other Technique

Study	N Pt	N Sp	Diagnostic test	Prevalence (%)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Adequate Specimens (%)	Quality Rating and Comments
Foutch et al. 1991	30	31 31 16	Bile cytology Brush cytology 1 Stent cytology	58 58 69	6 33 36	100 100 100	100 100 100	43 52 42		Good
Mansfield et al. 1997	43	54 54 19 19 54	Bile cytology Brush cytology 2 Soehendra stent retriever screw head Stent Combined	96 96 ? ? ?	12 42 25 37 54	100 100 ? ? 100	100 100 ? ? 100	4 6 ? ? 8	44 96 70 84	Good Clearly malignant or suspicious cytology = (+)
Sugiyama, Atomi, Wada et al., 19963	43 43 43	43 43 43	Bile cytology Brush cytology 4 Forceps biopsy	72 72 72	32a 48b 81c	100 100 100	100 100 100	36 43 67	100 88 87	Fair p<0.01, a vs c; p<0.05, b vs. c; p = n.r., a vs b

1

Milrose Lab, 230 cm, 2.5-mm diameter

2

Combocath, Microvasive, Boston Scientific

3

Specifically excluded patients with papillary tumor.

4

BC-23Q cytology brush (outer diameter, 1.8 mm, Olympus, Tokyo, Japan)

Five studies (total n=approximately 178), including 3 with "Good" quality, (Kurzawinski, Deery, Dooley et al., 1993; de Peralta-Venturina, Wong, Purslow et al., 1996; Foutch et al. 1991; Mansfield et al. 1997; Sugiyama, Atomi, Wada et al., 1996) provided comparisons between bile cytology and brush cytology for biliary strictures (Table 25 and Table 26). In each comparison, brush cytology provided higher sensitivity than bile aspirate cytology, although only one study reported a statistical assessment. The absolute increase in sensitivity ranged from 16 to 50%. Reported range of bile cytology sensitivity was 6-50% and that for brush cytology was 33-100%.

Two studies reported comparative data for tissue sampling using an ERC approach versus a percutaneous transhepatic cholangiographic (PTC) approach. de Peralta-Venturina, Wong, Purslow et al. (1996) noted lower sensitivity with PTC compared with ERC, 43 versus 100%. Kurzawinski, Deery, Dooley et al. (1993) observed similar sensitivity for brush cytology techniques using either approach and possibly lower sensitivity for bile aspirates with PTC.

In sum, the available studies are relatively small and most are limited by lack of statistical analysis but do provide suggestive evidence that brush cytology is more sensitive than bile aspiration cytology.

Brush Cytology Compared to FNA Cytology

Table 27. Comparisons of Brush Cytology and Biopsy (more...)

Table 27. Comparisons of Brush Cytology and Biopsy Technique

Study	N Pt	N Sp	Diagnostic test	Prevalence (%)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Adequate Specimens (%)	Quality Rating and Comments
Howell, Beveridge, Bosco et al., 1992	31		Brush cytology10 FNA - ERCP Combined	84 84 84	8 62 65	100 100 100	100 100 100	17 33 36		Fair
Ferrari, Lichtenstein, Slivka et al., 1994	70 51 19 29		Brush cytology 10 - Overall - Biliary - Pancreatic FNA - percutaneous	76 ?	56 54 64 91	100 100 100 75	100 100 100 95	51 45 67 60	93	Fair
Ponchon, Gagnon, Berger et al., 1995	233	193 118 105	Brush cytology10 Forceps biopsy 5 Combination	66 69 70	35a 43b 63c	97 97 97	96 97 98	66 69 70	90 57	Good p= n.s. for a vs b p<0.001 for a vs c p<0.05 for b vs. c
Schoefl, Haefner, Wrba et al., 1997	59 106 48	65 119 51	Brush cytology 6 Forceps biopsy 7 Combination	?	47 65 70	100 100 100	100 100 100	62 69 71		Fair
Pugliese, Antonelli, Vincenti et al., 1997	52	52	Brush cytology 8 Forceps biopsy 9 Combination	69 69 69	53 53 61	100 100 100	100 100 100	48 48 53		GoodUncertain cytology was considered negative.
Gmelin and Weiss 1981	32	32 26 26	Papillary tumors Brush cytology Forceps biopsy	85 81	18 71 55 86	100 100 100 100	100 100 100 100	18 45 29 63		Fair Suspicious cells considered negative Suspicious cells considered positive

5

Either Biomed 31010 (Paris, France: 175 cm length, 2mm diameter, round and fenestrated jaw with 2mm diameter, flexible tip, no needle) or Olympus prototype (Scop Medecine; 180cm length, 2.2mm diameter, round and fenestrated jaw with 2mm diameter, teflon sheath, no needle)

6

Endo-Flex 42 22E-A

7

Olympus FB-19N for about 60% and FB26N for about 30% and FB-39Q for about 10%

8

Olympus mod. BC-19Q or Wilson-Cook Medical Inc., Winston-Salem, NC, Mod. GBC-200-3-3.5

9

Olympus FB-19K or FB-39Q

Table 28. Comparison of Brush Cytology, FNA cytology, (more...)

Table 28. Comparison of Brush Cytology, FNA cytology, and Forceps biopsy in biliary strictures

Study	N Pts	N Spec	Diagnostic test	Prevalence (%)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Adequate Specimens (%)	Quality Rating and Comments
Jaiwala, Fogel, Sherman et al., 2000	133	133	Brush cytology 10 FNA cytology 11 Forceps biopsy 12 or 13 Brush + FNA Brush + Biopsy Biopsy + FNA Brush+Biopsy+FNA	78	48a 38b 54c 57d 71e 64f 77g	90 97 76 86 69 72 66	94 98 89 94 89 89 89	33 30 31 36 40 36 44	n.r. n.r. n.r. n.r. n.r. n.r. n.r.	Fair Any atypia on cytology was considered equivalent to cancer. P<0.05 for: a vs. e, f, g; b vs. c, d, e, f, g; c vs. e, f, g; d vs. e, g; f vs. g Only high-grade atypia considered equivalent to cancer. P<0.05 for: a vs. c, d, e, f, g; b vs. c, d, e, f, g; c vs. e, f, g; d vs. e, f, g All atypia on cytology considered negative. P<0.05 for: a vs. c, e, f, g; b vs. c, e, f, g; c vs. e, d, f; d vs. e, f, g.
			Brush cytology FNA cytology Forceps biopsy Brush + FNA Brush + Biopsy Biopsy + FNA Brush+Biopsy+FNA		30a 30b 43c 39d 55e 53f 62g	100 100 90 100 90 90 90	100 100 94 100 95 95 96	28 28 31 32 36 35 39
			Brush cytology FNA cytology Forceps biopsy Brush + FNA Brush + Biopsy Biopsy + FNA Brush+Biopsy+FNA		26a 25b 37c 34d 48e 46f 52g	100 100 100 100 100 100 100	100 100 100 100 100 100 100	27 27 31 30 35 34 37

10

Geenan brush system (Wilson-Cook Medical, Inc. Winston-Salem, N.C.)

11

Howell needle system (Wilson-Cook)

12

Malleable forceps (Olympus America, Inc., Melville, N.Y.)

13

Standard colonoscopic pinch forceps (Ballard Medical Products, Draper, Utah)

Three studies (total n=approximately 193), all rated "Fair" (Jaiwala, Fogel, Sherman et al., 2000; Howell, Beveridge, Bosco et al., 1992; Ferrari, Lichtenstein, Slivka et al., 1994) compare brush cytology with FNA cytology (Table 27 and Table 28). The first two studies use ERCP to obtain both the FNA specimen and the brush cytology specimens while Ferrari, Lichtenstein, Slivka et al. (1994) compares ERCP brush cytology with percutaneous CT-guided FNA. The largest study, (Jaiwala, Fogel, Sherman et al., 2000, n=133) reports similar sensitivity for FNA and for brush cytology and the combination of both techniques increased overall sensitivity by about 9%. This difference was not statistically significant in 2 of 3 comparisons and was found significant (p<0.05) only when high-grade atypia was considered equivalent to malignancy.

The study by Howell, Beveridge, Bosco et al. (1992, n=31) notes a higher sensitivity for FNA than for brush cytology (62% vs. 8%) but the combination of both techniques only yielded a slight increase to 65% sensitivity. Ferrari, Lichtenstein, Slivka et al. (1994, n=29 with FNA and 70 for brush cytology) found percutaneous CT-guided FNA to be more sensitive than brush cytology (91% versus 56%) but the large difference in sample sizes makes direct comparison limited. Furthermore, the small size and lack of statistical analysis of these two studies limits the interpretation of these findings.

Among these studies, the findings of Jaiwala, Fogel, Sherman et al. (2000) provide the more reliable information and suggest that brush cytology and ERCP-FNA may be similar in sensitivity. When used together, the available evidence does not demonstrate a statistically significant increase in sensitivity.

Forceps Biopsy Sampling Compared to Brush Cytology

Six studies (total n=approximately 437), including the 3 largest studies and 3 "Good" quality studies, compared forceps biopsy sampling to brush cytology (Tables 25-28). Gmelin and Weiss (1981) exclusively studied papillary tumors and found an increase in sensitivity of about 30% using forceps biopsy over brush cytology (86% versus 55%), but statistical analysis was not reported. Sugiyama, Atomi, Wada et al. (1996) specifically excluded papillary tumors and also found a large increase in sensitivity with forceps biopsy, 81% versus 48%, p<0.05. The remaining studies (Jaiwala, Fogel, Sherman et al., 2000; Ponchon, Gagnon, Berger et al., 1995; Schoefl, Haefner, Wrba et al., 1997; Pugliese, Antonelli, Vincenti et al., 1997) included a mixture of pancreaticobiliary malignancies. These studies reported generally similar sensitivity with forceps biopsy compared with brush cytology, though one study (Jaiwala, Fogel, Sherman et al., 2000) noted statistically significant increases for forceps biopsy over brush cytology when atypia was not interpreted as malignancy).

In addition, each of these studies reports that the combination of forceps biopsy and brush cytology increases the sensitivity in detecting malignancy between 5-20%. Jaiwala, Fogel, Sherman et al. (2000) and Ponchon, Gagnon, Berger et al. (1995) both reported the increase in sensitivity for the combination of forceps biopsy plus brush cytology compared to forceps biopsy alone to be statistically significant (p<0.05).

In sum, the available evidence suggests that forceps biopsy provides similar, or higher, sensitivity compared to brush cytology, and both tests used in combination may slightly increase sensitivity over that achieved with either technique alone.

Combination of Three Sampling Techniques

Jaiwala, Fogel, Sherman et al. (2000; n=133) also reports on the combination of brush cytology, FNA cytology, and forceps biopsy (Table 28). This study reports increases in overall sensitivity for detecting pancreaticobiliary malignancy as more sampling techniques are added together. The size of incremental the gains in sensitivity and statistically significance associated with adding the third sampling technique vary depending on the criteria used to interpret positive results on cytology. The largest gains are observed when forceps biopsy is being added as the third procedure (approximately 18-23% higher sensitivity, p<0.05), but smaller gains are still noted when one of the cytology techniques is added as the third procedure (approximately 4-13%).

Comparison of ERCP-FNA with EUS-FNA

Table 29. Supplemental Analysis: Single Arm Studies (more...)

Table 29. Supplemental Analysis: Single Arm Studies Reporting Diagnostic Operating Characteristics of EUS-FNA in Pancreatic Mass

Study	N Enr	N Res	Diagnostic test Population setting	Prev (%)	Sens (%)	Spec (%)	PPV (%)	NPV (%)	Adequate Specimens (%)	Comments
Wiersema, Vilmann, Giovannini et al., 1997 Multicenter - Including Indiana University and University of California	124	124	EUS-FNA Subgroup with pancreatic mass	74	89	100	100	76	97	Prospective 4 inadequate specimens excluded. Results in article are unclear regarding 5 cases of suspicious or atypical cytology.
Gress, Gottlieb, Sherman et al., 200114 Indiana University	102	94	EUS-FNA Suspected pancreatic ca after negative CT-FNA or ERCP cytology	64	88	100	100	92		Prospective 8 inconclusive or nondiagnostic results excluded
Gress, Hawes, Savides et al., 199714 Indiana University	121	121	EUS-FNA Pancreatic mass	42	80	100	100	88		Prospective
Brandwein, Farrell, Centano et al., 2001 Massachusetts General Hospital	96	93	EUS-FNA Suspected pancreatic ca underwent surgery	85 23 58	60 50 60	100 100 100	100 100 100	29 60 60		Retrospective Solid lesions (n=43) Cystic Lesions (n=26) Dilated duct (n=24)
Williams, Sahai, Aabakken et al., 1999 University of South Carolina	144	144	EUS-FNA All EUS-FNA referrals to single center	85	72 73 70	100 100 100	100 100 100	38 34 45		Retrospective All pancreatic masses Pancreatic mass > 3 cm Pancreatic mass < 3 cm
Bentz, Kochman, Faigel et al., 1998 University of Pennsylvania	45	38	EUS-FNA Pancreatic mass	82	94	100	100	78	84	Prospective
Chang, Nguyen, Erickson et al., 1997 University of California	44 pts 47 les	44	EUS-FNA Pancreatic mass	70	92	100	100	75	95	Retrospective

14

Both studies by Gress et al. are reported from the same institution, but patient selection criteria differ with the 2001 report choosing only the subset with persistently high clinical suspicion of pancreatic cancer following otherwise negative workup. The earlier study provides more generally selected patients.

In the absence of comparative studies directly comparing EUS-FNA and ERCP-FNA, an indirect comparison of single arm studies was attempted. Ten articles were identified, including one large multicenter report (Wiersema, Vilmann, Giovannini et al., 1997), three reports from Indiana University (Gress, Gottlieb, Sherman et al., 2001; Gress, Hawes, Savides et al., 1997; Wiersema, Kochman, Cramer et al., 1994), one report from Massachusetts General Hospital (Brandwein, Farrell, Centano et al., 2001), two reports from University of South Carolina (Williams, Sahai, Aabakken et al., 1999; Bhutani, Hawes, Baron et al., 1997), two reports from University of California (Chang, Nguyen, Erickson et al., 1997; Chang, Katz, Durbin et al., 1994), and one report from University of Pennsylvania (Bentz, Kochman, Faigel et al., 1998) (Table 29). Overlap of patient populations and data from separate reports from the same institution is difficult to assess due to limitations in reported detail. An attempt was made to minimize duplicate reporting of subjects. Earlier reports of studies from the same institution that were later published with more subjects have omitted from Table 29. However, some duplication of results likely remains between the multicenter report and separate reports from contributing institutions. The two reports by Gress et al. (Gress, Gottlieb, Sherman et al., 2001 and Gress, Hawes, Savides et al., 1997) address differently selected, but probably overlapping patient groups; however, both are included as they address slightly different questions.

All of these studies reported results separately for diagnosis of pancreatic mass. Additional results on lymph node evaluation and intestinal lesions were not relevant to this review. Despite uncertainties over the exact number of subjects included among the reports detailed in Table 29, the available studies include at least 400 subjects with pancreatic mass and report a range of sensitivity in detecting pancreatic malignancy of 60-94% with a specificity of 100%. Brandwein, Farrell, Centano et al. (2001; n=93) reported results separately for cystic versus solid pancreatic masses and found slightly lower sensitivity for cystic lesions, 50% versus 60%.

The sensitivity estimates for ERCP-FNA derived from the two studies identified in the systematic review (Jaiwala, Fogel, Sherman et al., 2000, n=133; Howell, Beveridge, Bosco et al. (1992, n=31) were obtained in subjects with a mixture of pancreaticobiliary malignancy and included subjects with pancreatic cancer, ampullary tumors, cholangiocarcinoma, and metastases. While the reported range of sensitivity of 25-62% for ERCP-FNA appears to be lower than that reported for EUS-FNA, direct comparisons do not seem appropriate due to differences in the case mix of tumors between studies. Further limitations secondary to relatively small numbers of subjects in ERCP-FNA studies and potential differences in cytology techniques and interpretations between studies preclude direct comparison of these estimated ranges of sensitivity.

Part II, Section 2: Diagnostic Performance of ERCP In Pancreaticobiliary Malignant Obstruction -- Comparison To Alternatives

Evidence Base

ERCP vs. MRCP

Table 30. Quality Assessment

Table 30. Quality Assessment

Study Author, Year	Patient Enrollment	Diagnostic performance of ERCP determined without knowledge of other test results	Diagnostic Performance of other test(s) determined without knowledge of ERCP results	Summary Evaluation
MRCP Studies
Varghese, Farrell, Courtney et al., 1999	Prospective (n=100) Complete explanation provided of 113 consecutive enrolled and 13 excluded subjects	Yes	Yes	Good
Adamek, Albert, Weitz et al., 1998	Prospective (n=60) 60 of 86 pts w/ suspected biliary obstruction Reasons for exclusions fully explained	Yes	Yes	Good
Arslan, Geitung, Viktil et al., 2000	Retrospective (n=135) 135 of 153 consecutive patients had diagnostic MRCP and ERCP Results reported in 78 patients with diagnostic quality MRCP and ERCP among of 85 patients with obstruction	Uncertain	Uncertain	Fair
Lee, Lee, Kim et al., 1997	? Retrospective (n=46) Complete explanation of 71 consecutive eligible patients and 25 exclusions	Yes	No	Fair MRCP results seem to factor into the reference standard determination
Holzknecht, Gauger, Sackmann et al., 1998	Prospective (n=61) Complete explanation provided of 66 consecutive enrolled patients and 5 excluded subjects	Yes	Yes	Good
Lomas, Bearcroft, and Gimson 1999	Prospective (n=69) Complete explanation provided of 76 enrolled and 7 excluded subjects	Yes	Uncertain	Fair
Adamek, Albert, Breer et al., 2000	Prospective (n=124) 124 of 141 pts w/ suspected pancreatic malignancy Reasons for exclusion fully explained	Yes	Yes	Good
Guibaud, Bret, Reinhold et al., 1995	Prospective (n=126) Some exclusions because of no ERCP confirmation	Uncertain	Yes	Fair
EUS Studies
Kaneko, Nakao, Inoue et al., 2001	Prospective (n=27) Consecutive patients with no reported exclusions	No	No	Fair
Glasbrenner, Schwarz, Pauls et al., 2000	Prospective (n=95) Consecutive patients referred for surgical resection of pancreatic mass	Yes	Yes	Good
Rosch, Schusdziarra, Born et al., 2000	Retrospective (n=184) Full explanation of 18 exclusions provided but selection based on having all 3 diagnostic tests creates a potential bias.	Yes	Yes	Fair
Cellier, Cuillerier, Palazzo et al., 1998	Retrospective (n=47) Consecutive patients with partial explanations for 17 excluded patients.	Uncertain	Yes	Fair
Burtin. Palazzo, Canard et al., 1997	Prospective (n=68) Consecutive patients enrolled	Yes	Yes	Fair -- unorthodox reporting of data, uncertain of data
Dancygier and Nattermann 1994	Prospective (n=41) Unstated whether consecutive	Uncertain	Yes	Fair
Snady, Cooperman, Siegel et al., 1992	Retrospective (n=60) Methods not well described other than pts were "diagnostically problematic"	No	No	Fair

Eight studies (total n=538) were identified that compared ERCP with MRCP and that used current MRCP technique. Five studies utilized an independent reference standard consisting of best available information derived from surgery, biopsy, imaging, and clinical follow-up to establish the final diagnosis, thus providing comparative data for ERCP and MRCP. The remaining three studies considered ERCP to be the reference standard against which MRCP was measured, yielding concordance of findings of MRCP with ERCP. Four studies were rated "Good" quality, signifying use of blinded interpretation of tests (Table 30). Four of these studies included over 100 subjects and the smallest study contained 46 subjects.

ERCP vs. EUS

Seven studies (total n=466) were identified that compared ERCP with EUS. Six of these employed an independent reference standard consisting of best available information derived from surgery, biopsy, imaging, and clinical follow-up to establish the final diagnosis, and therefore reported data for both EUS and ERCP. Only one study was rated "Good" (Glasbrenner, Schwarz, Pauls et al., 2000, n=90-91) (Table 30). Three studies addressed populations with obstructive jaundice, two studies addressed populations with suspected pancreatic cancer, and two studies addressed patients with either known or suspected intraductal papillary mucinous tumors of the pancreas.

Review of Evidence: Diagnostic Performance

Presence of Malignant Stricture/Lesion

ERCP vs. MRCP

Table 31. Comparison of MRCP and ERCP

Table 31. Comparison of MRCP and ERCP

Study	N Pt	N Res	Diag Test	Outcome	Prev (%)	Sens (%)	Spec (%)	PPV (%)	NPV (%)	Adeq Studies (%)	Comments
Independent Reference Standard15
Adamek, Albert, Weitz et al., 1998	86	60	MRCP ERCP	Presence of malignant stricture	45 45	81 93	100 94	100 93	87 94	97 79	Good, prospective p=n.r., but "equivalent"
Arslan, Geitung, Viktil et al., 2000	153	78	MRCP ERCP	Presence of malignant stricture		86 (74-94) 89 (77-96)	82 (67-93) 94 (82-99)			98.7 90	Fair, retrospective Kappa = 0.82
Lee, Lee, Kim et al., 199716	71	46	MRCP ERCP	Presence of malignant stricture	46 46	81 71	92 92	89 88	85 79	98 n.r.	Fair, ?retrospective McNemar p>0.05
Adamek, Albert, Breer et al., 2000	141	124	MRCP ERCP	Presence of pancreatic cancer	30 30	84 70	97 94	91 84	93 88	n.r. n.r.	Good, prospective McNemar p=0.059
Varghese, Farrell, Courtney et al., 199917	113 113	100 98 100 98	MRCP ERCP MRCP ERCP	Presence of stricture Level of stricture	28 28 28 28	100 100 100 100	100 100 100 100	100 100 100 100	100 100 100 100	97 89 97 89	Good, prospective No statistical analysis
ERCP Reference Standard
Guibaud, Bret, Reinhold et al., 1995	126	126	MRCP	Presence of malignant stricture	11	86 (67-100)	98 (96-100)	86	97	99	Fair, prospective
Lomas, Bearcroft, and Gimson 1999	76 76 76	69 69 69	MRCP	Presence of malignant stricture Presence of stricture Level of stricture	17 29 n.r.	92 100 100	100 98 (94-100) 100	100 95 (85-100) 100	98 100 100	97 97	Fair, prospective Kappa = 0.88
Holzknecht, Gauger, Sackmann et al., 1998	66	61	MRCP 18	Presence of stricture	59	89	84	89	84		Good, prospective No statistical analysis

15

Independent reference standards relied on best available information from surgery, biopsy, cytology, imaging, and clinical follow-up.

16

Reference standard also took into consideration MRCP and ERCP results as well as surgery

17

MRCP provided additional information over ERCP regarding cause of stricture in one case of 1.5 cm periampullary adenocarcinoma

18

This study performed MRCP using only "snapshot" techniques (RARE and half-Fourier RARE) in the coronal and angles sagittal planes. It is unclear whether axial images were routinely obtained.

Five studies including a total of 379 patients reported on diagnostic performance of MRCP in identifying and characterizing a malignant stricture (Table 31). In the two studies where ERCP was the reference standard (Guibaud, Bret, Reinhold et al., 1995; n=126; Lomas, Bearcroft, and Gimson 1999, n=69; both rated "Fair"), MRCP showed 86% and 92% sensitivity and 98 and 100% specificity. These data suggest good concordance between MRCP and ERCP results.

The three studies comparing MRCP and ERCP with an independent reference standard report slight differences in estimates of sensitivity and specificity, but none of these differences is statistically significant. The one study rated "Good" quality (Adamek, Albert, Weitz et al., 1998, n=60), reported slightly lower sensitivity (81% vs. 93%) and higher specificity (100% vs. 94%) for MRCP compared with ERCP, but both tests were considered equivalent. The largest study (Arslan, Geitung, Viktil et al., 2000, n=78) found similar sensitivity (86% vs. 89%) and reports lower specificity (82% vs. 94%) for MRCP, but 95% confidence intervals overlap significantly. Finally, Lee et al. (1998; n=46) reports higher sensitivity (81% vs. 71%) and similar specificity (92% vs. 92%) for MRCP, but overall accuracy was not statistically different.

ERCP vs. EUS

Table 32. Comparison of EUS and ERCP

Table 32. Comparison of EUS and ERCP

Study	N Pt	N Res	Diag test	Outcome	Prev (%)	Sens (%)	Spec (%)	PPV (%)	NPV (%)	Adeq Stud (%)	Comments
Population with obstructive jaundice
Independent Reference Standard15
Burtin. Palazzo, Canard et al., 1997	34	34	EUS ERCP	Presence of malignant lesion	36 36	89 89	96 92	89 80	96 96	97 97	Fair, prospective data not clearly reported p=n.s., diagnostic accuracy
Snady, Cooperman, Siegel et al., 1992	60	60 54	EUS ERCP+CT	Presence of malignant lesion	67 67	85 75	80 65	89 81	73 57		Fair, retrospective p=n.s.
ERCP Reference Standard
Dancygier and Nattermann 1994	41	41	EUS	Presence of malignant lesion	100	100	100	100	100		Fair, prospective No statistical analysis
	41	41	EUS	Level of stricture	100	100	100	100	100
Population with suspected pancreatic disease
Independent Reference Standard15
Glasbrenner, Schwarz, Pauls et al., 2000	95	90 91 90	EUS ERCP Combo	Presence of pancreatic cancer	54 53 53	78 81 92	93 88 86	93 89 88	78 80 90		Good, prospective p=n.s. for all comparisons
Rosch, Schusdziarra, Born et al., 2000	184	184 184	EUS ERCP Clinical	Presence of pancreatic cancer vs. chronic pancreatitis	42	86 81 81	87 85 85				Fair, retrospective p=n.s. p=n.s.
	184	184 184	EUS ERCP Clinical	Presence of pancreatic cancer vs. inflammatory tumor	42	86 81 81	72 61 72
Population with IPMT
Independent Reference Standard19
Kaneko, Nakao, Inoue et al., 2001	27	27 27	EUS ERP	Presence of mural nodules 20	81 81	59 50	100 100	100 100	36 31		Fair, prospective p=n.s.
Cellier, Cuillerier, Palazzo et al., 1998	47	21 29	EUS ERCP	Presence of invasive tumor 21	43 31	78 55	75 90	70 71	82 82		Fair, retrospective No statistical analysis

19

Reference standard consists of surgical specimen histology and/or pancreatography

20

Population of patients with suspected intraductal papillary mucinous tumors of the pancreas

21

population of patients with histologically proven diagnosis of intraductal papillary mucinous tumors of the pancreas

Three studies, all rated "Fair" quality and including a total of 129 patients with obstructive jaundice, reported on the diagnostic performance of EUS in identifying the presence of a malignant lesion/stricture (Table 32). One study (Burtin. Palazzo, Canard et al., 1997, n=34) reported similar diagnostic performance for ERCP and EUS, with both tests achieving 89% sensitivity and similar specificity (96% for EUS and 92% for ERCP). Dancygier and Nattermann (1994, n=41) reported complete concordance between EUS and ERCP. One study (Snady, Cooperman, Siegel et al., 1992, n=54-60) compared EUS with the combination of ERCP plus CT and reports both higher sensitivity and specificity for EUS, 85% vs. 75% sensitivity, and 80% vs. 65% specificity, respectively, but these differences were not statistically significant.

In summary, individual studies were relatively small and did not identify significant differences in diagnostic performance between ERCP and either MRCP or EUS. These data permit preliminary conclusions that MRCP and EUS provide similar diagnostic assessment as ERCP for detection of malignant pancreaticobiliary obstruction.

Diagnosis of Pancreatic Cancer

MRCP vs. ERCP

Diagnostic performance for demonstrating pancreatic cancer in 37 of 124 was reported by Adamek, Albert, Breer et al. (2000; Table 31). This study compares MRCP and ERCP and reported slightly higher sensitivity (84% vs. 70%) and similar specificity (97% vs. 94%) for MRCP and ERCP, respectively, but these differences did not reach statistical significance (McNemar p=0.059). This study was rated "Good" for quality.

EUS vs. ERCP

Diagnostic performance for pancreatic cancer was reported in two studies specifically addressing populations with suspected pancreatic disease (Table 32). Rosch, Schusdziarra, Born et al. (2000) retrospectively evaluated 184 patients who had ERCP, EUS, and CT and compared the diagnostic performance of clinical assessment with the various imaging tests. This study finds similar performance for clinical assessment, ERCP, or EUS in distinguishing pancreatic cancer from chronic pancreatitis and in distinguishing pancreatic cancer from inflammatory tumor. Interpretation of Rosch, Schusdziarra, Born et al. (2000) is somewhat limited by the retrospective selection of patients on the basis of having all three imaging tests, which might bias the study toward cases where findings were inconclusive. Glasbrenner, Schwarz, Pauls et al. (2000; n=95) noted ERCP and EUS to have similar sensitivity (81% vs. 78%, respectively) and specificity (88% vs. 93%, respectively), and the combination of the two tests yielded 92% sensitivity and 86% specificity, but these differences were not statistically significant.

Summary

In summary, there is little evidence directly comparing ERCP with either MRCP or EUS in diagnosing pancreatic cancer. The available evidence does not demonstrate statistically significant differences between ERCP and either MRCP or EUS.

Presence of Stricture

ERCP vs. MRCP

Three studies reported diagnostic performance in demonstrating the presence of stricture (either benign or malignant) (Table 31). One of the two studies rated as "Good" independently verified results and found 100% sensitivity and 100% specificity for both MRCP and ERCP (Varghese, Farrell, Courtney et al., 1999, n=98-100). The other (Holzknecht, Gauger, Sackmann et al., 1998, n=61) used ERCP as reference standard and reported 89% sensitivity and 85% specificity for MRCP relative to ERCP, though this study utilized only projection ("snapshot") MRCP techniques without additional multislice techniques which may limit its comparability. One additional study (Lomas, Bearcroft, and Gimson 1999, n=69) rated as "Fair" quality because of uncertainties with regard to complete blinding of interpretation, noted 100% concordance for MRCP with ERCP.

ERCP vs. EUS

No studies reported this specific analysis.

Summary

In summary, the evidence specifically evaluating MRCP in relation to ERCP for detecting strictures is sparse and suggests similar results for MRCP and ERCP in identifying the presence of a stricture. However, these studies do not report full statistical analysis. The relative performance of EUS and ERCP in this setting has not been reported.

Level of Stricture

ERCP vs. MRCP

One study comparing ERCP and MRCP (Varghese, Farrell, Courtney et al., 1999, n=98-100, "Good") specifically reported 100% sensitivity and specificity for both MRCP and ERCP in defining the level of the stricture (Table 31). Lomas, Bearcroft, and Gimson (1999, n=69, "Fair") also reported complete concordance for MRCP with ERCP in defining the level of malignant strictures.

ERCP vs. EUS

Only one study comparing ERCP and EUS (Dancygier and Nattermann 1994, n=41, "Fair") specifically reported sensitivity and specificity in defining the level of the stricture (Table 32). This study reports 100% sensitivity and specificity for both ERCP and EUS.

Summary

In summary, there is little evidence specifically reporting the diagnostic accuracy of MRCP or EUS relative to ERCP in defining the level of stricture, but the available studies suggest that all three tests provide highly accurate localization of pancreaticobiliary stricture.

Evaluation of Suspected Intraductal Papillary Mucinous Tumors (IPMT) of the Pancreas

ERCP vs. MRCP

No studies reported this specific analysis

ERCP vs. EUS

Two studies evaluated EUS in comparison with endoscopic retrograde pancreatography (ERP) in patients with either known or suspected IPMT of the pancreas (Table 32). Kaneko, Nakao, Inoue et al. (2001; n=27, "Fair") found that EUS and ERP were similarly sensitive (59% vs. 50%, respectively) in detecting mural nodules while both tests were 100% specific for this finding. Cellier, Cuillerier, Palazzo et al. (1998; n=47, "Fair") compared ERCP and EUS in defining the presence of invasive tumor and reported EUS to be more sensitive (78% vs. 55%) and less specific (75% vs. 90%), but no statistical analysis was reported.

These two small studies, reporting estimates of diagnostic performance relating to different diagnostic endpoints, suggest that EUS may provide a similar information to ERCP in patients with known or suspected intraductal papillary mucinous tumors of the pancreas, but confirmation of these findings would be helpful.

Part II, Section 3: Outcomes of Treatment Using ERCP and Endoscopic Sphincterotomy and Endoscopic Stent for Palliation of Pancreaticobiliary Malignancy -- Comparison of Strategies Using ERCP, Surgery, or Interventional Radiology

Part II, Section 3A. Comparison of ERCP Stent Versus Surgical Bypass

Body of Evidence

Table 33. Quality Assessment

Table 33. Quality Assessment

Study Author, Year	Comparable Initial Groups?	Comparable Groups Maintained?	Comparable Performance of Intervention?	Comparable Measurement of Outcomes?	Appropriate Analysis	Summary Evaluation
Smith, Dowsett, Russell et al., 1994	RCT (n=204) Good comparability -  Randomization by computer minimization on age, bilirubin, albumin, urea, and Hb conc. -  Patient characteristics not significantly different	Surgery: (n=103) 2 excluded due to benign disease 7 did not get surgery (2 technical failures, 1 elected crossover, 3 deteriorated clinically and got stents, 1 deteriorated and got no further rx) Stent: (n=101) 1 excluded due to benign disease 5 did not get stents (1 elected crossover, 3 technical failures got surgery, 1 technical failure got no further rx)	Adequate for comparison	Adequate outcome measures used. Outcomes were not assessed blindly.	Intention-to-treat analysis used	Good
Andersen, Sorensen, Kruse et al., 1989	RCT (n=50) Good comparability -  Sealed envelopes -  Patient characteristics not significantly different	Surgery: n=25 6 did not undergo surgery (2 wanted crossed over, 1 found inoperable at surgery, 2 psychological compromise, 1 surgeon not available) Endoprosthesis: n=25 None	Adequate for comparison	Adequate outcome measures used. Outcomes were not assessed blindly.	Intention-to-treat analysis used Results also analyzed by treatment received and findings were consistent.	Good
Shepherd, Royal, Ross et al., 1988	RCT (n=52) Fair comparability -  Randomization method not specified -  Patient characteristics mostly comparable	Surgical: n=27 4 total: 2 withdrawn (1 died pre-op and 1 had attempted curative surgery). 2 technical failures crossed over to endoprosthesis. Endoprosthesis: n=25 6 total: 1 had benign biopsies but later found to have cancer at surgery; 4 failed and crossed-over to surgery; 1 failed both stent and surgery	Adequate for comparison	Adequate outcome measures used. Outcomes were not assessed blindly.	Does not clearly state method of analysis	Fair
Raikar, Melin, Ress et al., 1996	Retrospective series (n=66)Fair to Poor comparability Baseline patient characteristics show no SSD but differences in performance status distribution noted with ERCP subjects having relatively higher percentages of good and poor PS while surgery had relatively higher midrange PS.	All subjects included in analysis	Adequate for comparison	Adequate outcome measures used. Outcomes were not assessed blindly.	Univariate analysis does not account for important confounders	Poor
Leung, Emergy, Cotton et al., 1983	Retrospective series (n=98) Poor comparability Baseline patient characteristics show differences in age and lesion location.	All subjects included in analysis	Adequate for comparison	Adequate outcome measures used. Outcomes were not assessed blindly.	Univariate analysis does not account for important confounders	Poor

Table 34. Overview of studies and reported outcomes (more...)

Table 34. Overview of studies and reported outcomes

Study	Population	Procedure	N ERCP Surg (treated)	Outcome Measures Reported									Study Quality
				Total Hospital Days	Initial Hospital Days	Readmissions	Need for Add'l Procedure	Survival	Jaundice Relief	Quality of Life	Perioperative Mortality	Perioperative Morbidity
Randomized Controlled Trials
Smith, Dowsett, Russell et al., 1994	Malignant distal CBD obstruction and jaundice Mean age 70	10 Fr stents 22 vs. Bypass Surgery	101 (100) 103 (101)	X			X	X		X	X	X	Good
Andersen, Sorensen, Kruse et al., 1989	Malignant distal CBD obstruction and jaundice Age>60y	7-10 Fr stents vs. Bypass Surgery	25 (19) 25 (30)	X			X	X		X	X	X	Good
Shepherd, Royal, Ross et al., 1988	Malignant distal CBD obstruction Mean age 73	10 Fr stents vs. Bypass Surgery	27 (23) 25	X	X	X	X	X	X		X	X	Fair
Retrospective Studies
Raikar, Melin, Ress et al., 1996	Unresectable pancreatic carcinoma	10-12 Fr stents vs. Bypass Surgery	34 32			X	X	X			X	X	Poor
Leung, Emergy, Cotton et al., 1983	Malignant obstructive jaundice (CBD location not specific)	8-10 Fr stents vs. Bypass Surgery	64 34			X	X	X			X		Poor

22

19 of 101 stent patients required combined ERCP and percutaneous transhepatic approach to place stent

Table 35. Treatment Outcomes

Table 35. Treatment Outcomes

Study	Study arm N Enrolled/(treated or results)	Survival (median) (*mean) (**Life Table Analysis)	P	Relief of Jaundice	p	Quality of Life	p
Randomized Controlled Trials
Smith, Dowsett, Russell et al., 1994	ERCP 23 101 (100)	21 weeks	ns	97%	ns
	Surgery 103 (101)	26 weeks		98%
Andersen, Sorensen, Kruse et al., 1989	ERCP 25 (19)	**84 days (3-498) 24	ns			57% survival time mean normal activity or limited, no aid	ns
	Surgery 25 (30)	**100 days (10-642)				51% survival time mean normal activity or limited, no aid
Shepherd, Royal, Ross et al., 1988	ERCP 27 (23)	**152 days (39-411)	ns	91	nr
	Surgery 25	**125 days (52-354)		92%
Retrospective Studies
Raikar, Melin, Ress et al., 1996	ERCP 34	*9.7 months (10d-35)	0.13
	Surgery 32	*7.3 month (7d-29)
Leung, Emergy, Cotton et al., 1983	ERCP 64	6 mos. approximate	Ns
	Surgery 34	6 mos. approximate

23

Stent placement was attempted first with ERCP approach. In 19 patients a combined transhepatic-endoscopic approach was required when initial ERCP failed.

24

No significant difference when analyzed by treatment received.

Table 36. Adverse Outcomes

Table 36. Adverse Outcomes

Study	Study arm N Enrolled/(treated or results)	Perioperative Mortality	P	Perioperative Complications	p
Randomized Controlled Trials
Andersen, Sorensen, Kruse et al., 1989	ERCP 25 (19)	5 (20%)	Nr	36% (total severe infection)	Ns
	Surgery 25 (30)	6 (24%)		20% (total severe infection)
Shepherd, Royal, Ross et al., 1988	ERCP 27 (23)	2 (9)%	Ns	7 procedure-related complication events	Ns
	Surgery 25	5 (20%)		14 procedure-related complication events
Smith, Dowsett, Russell et al., 1994	ERCP 25 101 (100)	8% 26	Ns	11% major complications	0.02
	Surgery 103 (101)2 (n)	15%		29% major complications
Retrospective Studies
Leung, Emergy, Cotton et al., 1983	ERCP 64	1 (3%)	Nr	21%	ns
	Surgery 34	1 (4%)		33%
Raikar, Melin, Ress et al., 1996	ERCP 34	10 (16%)	Nr
	Surgery 32	3 (9%)

25

Stent placement was attempted first with ERCP approach. In 19 patients a combined transhepatic-endoscopic approach was required when initial ERCP failed.

26

Procedure related mortality was significantly higher in the surgery group (14% vs. 3%, p=0.006). Also of note, 3 deaths in the surgical group were in patients who did not undergo surgery.

Table 37. Resource Utilization Outcomes

Table 37. Resource Utilization Outcomes

Study	Study arm N Enrolled/(Treated or Results)	Total Hospital Days median 27 (range)	p	Initial Hospital Days (median) (*mean)	p	Readmission to Hospital N (%)	p	Need for Additional Procedure	p
Randomized Controlled Trials
Smith, Dowsett, Russell et al., 1994	ERCP 28 101 (100)	19 (4-59)	ns					Recurrent obstructive jaundice requiring stent replacement in 36 (36%) Late gastric outlet obstruction requiring gastric bypass in 10 (10%)	ns ns
	Surgery 103 (101)	26 (8-85)						Recurrent obstructive jaundice in 2 (2%). One required stent. Late gastric outlet obstruction requiring gastric bypass in 5 (5%)
Andersen, Sorensen, Kruse et al., 1989	ERCP 25 (19)	26 (3-210)	ns 29					1 (4%) early failure requiring surgical bypass.	nr
	Surgery 25 (30)	27 (10-202)						3 (12%) early failure requiring stent placement.
Shepherd, Royal, Ross et al., 1988	ERCP 27 (23)	8 30 (2-30)	<0.01	5 (2-16)	<0.002	10 (43%)	nr	Gastric outlet obstruction developed in 2 (9%)	nr
	Surgery 25	13 (8-49)		13 (8-49)		3 (12%)		Gastric outlet obstruction developed in 1 (4%)

Study	Study arm N Enrolled/(Treated or Results)	Total Hospital Days median 31 (range)	p	Initial Hospital Days (median) (*mean)	p	Readmission to Hospital N (%)	p	Need for Additional Procedure	p
Retrospective Studies
Raikar, Melin, Ress et al., 1996	ERCP 34	$17,738	.05	7*	<0.001	12 (35%)	nr	Average of 1.7 stent replacements per patient One patient developed gastric outlet obstruction requiring surgical gastric bypass.	nr nr
	Surgery 32	$25,101		14*		8 (25%)		Two patients required stent placement for recurrent jaundice. No report of surgical patients developing gastric outlet obstruction.
Leung, Emergy, Cotton et al., 1983	ERCP 64			14* (4-30)	Nr	8 (13%)>32	nr	Recurrent jaundice developed in 3 (5%) Gastric outlet obstruction developed in 2 (3%)	nr nr
	Surgery 34			30* (14-79)		3 (9%)		Recurrent jaundice developed in 1 (3%) Gastric outlet obstruction developed in 2 (6%)

27

Results generally reported as median. Results reported as mean are demarcated by an asterisk (*)

28

Stent placement was attempted first with ERCP approach. In 19 patients a combined transhepatic-endoscopic approach was required when initial ERCP failed.

29

Comparison of hospital stay was not statistically significant when analyzed by treatment received.

30

Calculated only in patients who were alive 30 days post-op.

31

Results generally reported as median. Results reported as mean are demarcated by an asterisk (*)

32

Local complications included cholangitis, recurrent jaundice, duodenal obstruction, or chest wall metastasis

Five studies compared results of surgical bypass with endoscopic stent drainage for palliation of malignant obstructive jaundice. Quality assessments are described in Table 33. Results of these studies are detailed in the "Evidence Tables" section and summarized in Tables 34-37. Three randomized, controlled trials were identified comparing surgical biliary bypass with endoscopic biliary stent placement. Two of these (Smith, Dowsett, Russell et al., 1994, n=204; Andersen, Sorensen, Kruse et al., 1989, n=50) were rated as "Good" quality, and Shepherd, Royal, Ross et al. (1988, n=52) was rated as "Fair"). Two retrospective comparisons (Raikar, Melin, Ress et al., 1996, n=66; Leung, Emergy, Cotton et al., 1983, n=98) were both rated as "Poor."

Review of Evidence: Treatment Outcomes

All studies reported that there was no significant difference in overall patient survival between the ERCP and the surgery groups (Table 35). Two randomized controlled trials reported both treatments to have high rates for relief of jaundice but no statistically significant difference. A third study reported on quality of life, as measured by mean percentage of survival time with normal activity or limited activity with no aid; there were no significant differences.

Review of Evidence: Adverse Outcomes

There were no significant differences in perioperative mortality (Table 36). The randomized controlled trial by Smith, Dowsett, Russell et al. (1994) was designed to show a 5-20% decrease in 30-day mortality at 95% power with 115 patients entered into each arm. Accrual was stopped at 204 patients when interim analysis indicated that additional accrual would not change the outcome. While this trial did not show a statistically significant difference in perioperative (30-day) mortality, intent-to-treat analysis showed significantly greater procedure-related mortality in the surgery arm (14% vs. 3%, p=0.006). Smith, Dowsett, Russell et al., (1994) also found that major complications were significantly greater in the surgery group than in the ERCP group (29% vs. 11%, p=0.02). Andersen, Sorensen, Kruse et al. (1989) reported severe infections in 36% of ERCP patients compared to 20% of surgical patients, but the difference was not statistically significant. Shepherd, Royal, Ross et al. (1988) found twice the rate of complications in the surgical group, but again this was not statistically significant.

Review of Evidence: Resource Utilization

The two randomized controlled trials rated as good quality found no significant difference in total days of hospitalization, including the largest of trials in this group of studies (Smith, Dowsett, Russell et al., 1994, n=203) (Table 37). Three studies report on initial hospitalization; including 1 randomized controlled trial (Shepherd, Royal, Ross et al., 1988, n=52). All show fewer days of initial hospitalization with ERCP, and 2 report that the difference is statistically significant. Readmissions were more common with ERCP, but tests of statistical significance were not reported. The randomized controlled trial by Shepherd, Royal, Ross et al. (1988) reports significantly fewer initial and total hospitalization days with ERCP, despite a readmission rate twice that of surgery. However, this randomized controlled trial was judged of lesser quality ("fair"), largely due to lack of clarity in the method of analysis.

Stent replacement was reported in the Smith, Dowsett, Russell et al., (1994) study as necessary in 37% of patients, all but 1 case due to recurrence of obstructive jaundice. Raikar, Melin, Ress et al. (1996) reported an average of 1.7 stent replacements per patient.

Summary

The most robust evidence is provided in the randomized controlled trial by Smith, Dowsett, Russell et al. (1994). There were no significant differences in overall survival, relief of jaundice, technical success, total hospitalization days or perioperative mortality. Major complications were more frequent in the surgery group (11% vs. 29%, p=0.02), presumably reflecting the more invasive nature of surgical versus endoscopic treatment. Stent replacement was required in 37% of ERCP patients.

Part II, Section 3B. Comparison of Metal vs. Plastic Stents During ERCP

Evidence Base

Table 38. Study Quality Assessment

Table 38. Study Quality Assessment

Study Author, Year	Comparable Initial Groups?	Comparable Groups Maintained?	Comparable Performance of Intervention?	Comparable Measurement of Outcomes?	Appropriate Analysis	Summary Evaluation
Davids, Groen, Rauws et al., 1992	RCT (n=105) Good comparability - Randomization by computer generated random number - patient characteristics well-balanced	115 initially randomized and 105 included in analysis 10 patients excluded. 5 due to prior history of malignancy in past 10 years and 5 due to selection for surgical therapy. None lost to follow-up	Adequate for comparison.	Adequate outcome measures used. Outcomes were not assessed blindly.	Method of analysis not clearly stated.	Fair
Prat, Chapat, Ducot et al., 1998	RCT (n=101) Good comparability - Randomization by blocks of six and stratified for gender and investigation center - patient characteristics well-balanced	4 of 105 excluded Three for failed endoprosthesis insertion and one for not complying with required quarterly stent changes for group 2 Four lost to follow-up (3 moved away and 1 no follow-up information)	Adequate for comparison.	Adequate outcome measures used. Outcomes were not assessed blindly.	Method of analysis not clearly stated	Fair
Schmassmann, Von Gunten, Knuchel et al., 1996	Retrospective study (n=165) Fair comparability Baseline patient characteristics similar for age, gender, bilirubin, type of tumor and stage, location of stricture, or associated procedures	All subjects included in analysis	Adequate for comparison 87% of metal stent and 100% of plastic stent patients had sphincterotomy	Adequate outcome measures used. Outcomes were not assessed blindly.	Univariate analysis does not account for confounders	Poor

Table 39. Overview of studies and reported outcomes (more...)

Table 39. Overview of studies and reported outcomes

Study	Population	Procedure	N (treated) Metal Plastic	Outcome Measures Reported									STUDY QUALITY
				Total Hospital Days	Initial Hospital Days	Cost Utilization	Need for Add'l Procedure	Survival	Jaundice Relief	Stent Patency	Periop Mortality	Periop Morbidity
Randomized Controlled Trials
Davids, Groen, Rauws et al., 1992	Patients with irresectable distal bile-duct malignancy Pancreatic ca = 93 Papillary ca = 12	Metal stent 33 Straight 10 Fr polyethylene stent 34	49 56				X	X	X	X	X	X	Fair
Prat, Chapat, Ducot et al., 1998	Patients with malignant CBD strictures Not involving hilum Pancreatic ca = 65 Cholangioca = 21 Ampullary ca = 3 Metastatic = 12	Metal stent Polyethylene 11.5 Fr stent 35 w/ routine exchange Polyethylene 11.5 Fr stent w/ as needed exchange	34 33 34	X		X	X	X	X		X	X	Fair
Retrospective Studies
Schmassmann, Von Gunten, Knuchel et al., 1996	Consecutive patients with unresectable malignant biliary obstruction	Metal stent 33 Straight 12 Fr or 10 Fr polyethylene stent 36	95 70				X	X	X	X	X		Poor

33

Metal stents were of the Wallstent type (Schneider, Switzerland (Davids et al.; Schmassmann et al.)) or (Schneider-Howmedical, Lyons, France (Prat et al.)).

34

Polyethylene stents were made by PBN Medicals (Stenlose, Denmark)

35

Polyethylene stents were made by Wilson-Cook (Winston-Salen, N.C.)

36

Polyethylene stents 12 Fr were made by Olympus (Volketswil, Switzerland) and 10 Fr Huibregtse (Cook, Nottwil, Switzerland)

Table 40. Treatment Outcomes

Table 40. Treatment Outcomes

Study	Study arm N Enrolled/(treated or results)	Survival (median)	P	Relief of Jaundice N (%)	p	First Stent Patency (median)	p
Randomized Controlled Trials
Davids, Groen, Rauws et al., 1992	Metal 49	5.8 months 37	0.45	47/49 (96%)	n.r.	9.1 months 37	0.006
	Plastic 56	4.9 months 37		53/56 (95%)		4.2 months 37
Prat, Chapat, Ducot et al., 1998	Metal 34	4.5 months	n.s.	48h Decrease in bilirubin: 41%	n.s.	4.8 months	<0.05
	Plastic-routine 33	5.6 months		34.3%		Not reported separately
	Plastic-as needed 34	4.8 months		35.4%		3.2 months
Retrospective Studies
Schmassmann, Von Gunten, Knuchel et al. 1996	Metal 95	6.5 months 38	<0.05	95%	n.s.	10 months 39	<0.001
	Plastic 70	4 months		88%		4 months

37

Data were converted to months from reported days by dividing by 30.

38

When 29 subjects (8 metal stent, 21 plastic stent) who died related to untreated stent dysfunction were excluded from the analysis, the remaining 136 subjects had similar survival between the two groups.

39

Subgroup analysis did not show any significant difference between different locations (common bile duct vs. hilar or intrahepatic stricture) but numbers were small in the hilar and intrahepatic subgroups.

Table 41. Adverse Outcomes

Table 41. Adverse Outcomes

Study	Study arm N Enrolled/(treated or results)	Perioperative Mortality	P	Complications	p
Randomized Controlled Trials
Davids, Groen, Rauws et al., 1992	Metal 49	7 (14%) 40	0.047	6 (12%) 41	n.r.
	Plastic 56	2 (4%) 42		6 (11%)
Prat, Chapat, Ducot et al., 1998	Metal 34	Overall rate was 3.9% No significant difference between groups		Overall rate was 11.9% No significant difference between groups
	Plastic-routine 33
	Plastic-as needed 34
Retrospective Studies
Schmassmann, Von Gunten, Knuchel et al. 1996	Metal 95	2%	n.s.
	Plastic 70	3%

40

Causes of death were sepsis after recurrent cholangitis (1); cardiac failure (2); cachexia (4).

41

Complications in Davids et al. were measured in 7 days after procedure.

42

Causes of death were cachexia (2).

Table 42. Resource Utilization Outcomes

Table 42. Resource Utilization Outcomes

Study	Study arm N Enrolled/(Treated or Results)	Total Hospital Days median (range)	p	Resource Utilization Costs	p	Need for Additional Procedure	p
Randomized Controlled Trials
Davids, Groen, Rauws et al., 1992	Metal 49					1.3 per person	n.r.
	Plastic 56					1.8 per person
Prat, Chapat, Ducot et al., 1998	Metal 34	5.5 ± 1.4*	*0.01 others n.s.	Mean costs (95% CI) $4643 (4207-5079)	n.r.	1.2 ± 0.4 per patient	0.01 ANOVA
	Plastic-routine 33	10.6 ± 1.7*		$6770 (5394-8146)		2.5 ± 1.9 per patient
	Plastic-as needed 34	7.4 ± 1.5		$5547 (4082-7013)		1.7 ± 1.3 per patient
Retrospective Studies
Schmassmann, Von Gunten, Knuchel et al., 1996	Metal 95					1.2 per patient	<0.005
	Plastic 70					1.58 per patient

Three studies were identified comparing endoscopically placed metal or plastic stents for palliation of biliary obstruction due to malignancy. Quality ratings are described in Table 38. Results are detailed in the "Evidence Tables" chapter and summarized in Tables 39-42. Two randomized, controlled trials (total n=206) were identified. Davids, Groen, Rauws et al. (1992, n=105, "Fair" quality) compared metal versus plastic stents. Prat, Chapat, Ducot et al. (1998, n=101, "Fair" quality) randomized patients into 3 arms (either metal stents, plastic stents with exchange as needed for stent dysfunction, or plastic stents with routine exchange every 3 months). In addition, Schmassmann, Von Gunten, Knuchel et al. (1996, n=165, "Poor" quality) retrospectively compared results with metal versus plastic stents.

Review of Evidence: Treatment Outcomes

Metal stents showed statistically significantly longer patency rates compared with plastic stents in all three studies (Table 40). Two of the studies reported that median duration of patency with metal stents was twice as long as plastic stents (9.1-10 months versus 4-4.2 months, p<0.006), but one of the randomized trials showed a smaller benefit for metal stents (4.8 months versus 3.2 months, p<0.05).

The two randomized studies reported no significant difference in overall survival for patients treated with metal or plastic stents, with median survival ranging from 4.5-5.8 months. In contrast, the retrospective study found slightly longer median survival in the metal stent group (6.5 months versus 4 months, p<0.05), but related this observation to increased mortality in 18% of subjects (predominantly plastic stent group) who did not receive treatment for stent dysfunction.

All studies reported both treatments to have high rates for relief of jaundice with no statistically significant differences reported.

Review of Evidence: Adverse Outcomes

Two studies (Prat, Chapat, Ducot et al., 1998; Schmassmann, Von Gunten, Knuchel et al., 1996) reported no significant difference in perioperative mortality (Table 41). The randomized, controlled trial by Davids, Groen, Rauws et al. (1992) noted a higher perioperative mortality rate in the metal stent group (14% vs. 4%, p=0.047), but the causes of death in 6 of 7 cases were completely unrelated to biliary pathology. No significant differences were noted in complications in the two randomized studies and the retrospective study did not specifically report complications other than perioperative mortality.

Review of Evidence: Resource Utilization Outcomes

All studies examined the relative utilization of ERCP procedures and found patients receiving metal stents to require the fewest ERCP procedures (Table 42). Patients receiving metal stents required 1.2-1.3 ERCP procedures on average and those receiving plastic stents and undergoing stent exchange only when needed required 1.58-1.8 ERCP procedures. The study by Prat, Chapat, Ducot et al. (1998) examined the strategy of routine plastic stent exchange every 3 months which necessitated an average of 2.5 ERCP procedures per patient. The differences in ERCP utilization between metal and plastic stents were reported to be statistically significant in two studies and a statistical comparison was not reported in the third study.

Prat, Chapat, Ducot et al. (1998) also examined utilization of total hospital days and found the metal stent group averaged 5.5 days while the plastic stent groups required 7.4 to 10.6 days on average, depending on whether "as needed" or routine stent exchange was used, respectively. The difference between metal stents and routinely exchanged plastic stents was statistically significant (5.5 ± 1.4 versus 10.6 ± 1.7, p=0.01) while the differences between metal stents and plastic stents exchanged as needed were not statistically significant.

Prat, Chapat, Ducot et al. (1998) also reported lower average total costs for the metal stent group than costs associated with either of the plastic stent strategies, but statistical analysis was not reported for these results.

Summary

Three studies including a total of 371 subjects provide consistent evidence that metal stents remain patent longer than plastic stents. Both types of stents offer initial relief of jaundice and the available evidence does not conclusively show any difference in perioperative adverse events. Overall patient survival is not significantly different when stent occlusions are treated with stent exchange as needed. Total resource utilization including need for repeat ERCP, total hospital days, and costs was reported to be lower with metal stents compared with plastic stents.

Part II, Section 3C. Additional Comparisons of ERCP Strategies

Evidence Base

Table 43. Quality Assessment

Table 43. Quality Assessment

Study Author, Year Record Number	Comparable Initial Groups?	Comparable Groups Maintained?	Comparable Performance of Intervention?	Comparable Measurement of Outcomes?	Appropriate Analysis	Summary Evaluation
van Berkel, Boland, Redekop et al., 1998	RCT (n=84) Good comparability - Randomization by computer generated numbers in sealed envelopes - Patient characteristics similar	97 consecutive patients enrolled. 13 excluded for protocol violations (11 had surgical resection, 1 had PTH drainage, 1 refused treatment). Details about which treatment arm patients were assigned to were not provided. None lost to follow-up.	Adequate for comparison.	Adequate outcome measures used. Outcomes were not assessed blindly.	Method of analysis not stated but all 84 included in analysis.	Fair
Pedersen 1993	Prospective study (n=89) Fair comparability Differences in age noted with younger 7Fr group. No SSD in stenosis location, gender, or type of cancer.	All subjects included in analysis	Adequate for comparison. Adjunctive sphincterotomy was performed equally in 7Fr and 10Fr groups.	Adequate outcome measures used. Outcomes were not assessed blindly.	Univariate analysis does not account for important confounders	Poor
Speer, Cotton, MacRae et al., 1988	Retrospective study (n=79) Fair comparability Baseline patient characteristics similar for age and site of obstruction.	All subjects included in analysis	Limitations for comparison 8 Fr stents had pigtails whereas 10Fr stents were straight	Adequate outcome measures used. Outcomes were not assessed blindly.	Univariate analysis does not account for important confounders	Poor
Sung, Chung, Tsui et al., 1994	RCT (n=70) Good comparability - Sealed envelopes - Patient characteristics show no SSD	SH: (n=35) NSH: (n=35) 3 subjects dropped out before 4 week f/u and were excluded from analysis	Adequate for comparison	Adequate outcome measures used. Patient and follow-up physician were blinded to type of stent placed.	Method of analysis not reported but no crossover reported.	Good
Speer, Cotton, Russell et al., 1987	RCT (n=75) Good comparability - Computer generated random numbers and stratified by referring center - Patient characteristics similar for age, ASA 43 grade, duration of jaundice, bilirubin, albumin, creatinine, and Hb, but ERCP group had more proximal obstructions, more unrelated medical problems, and more elevated WBC. No statistical results reported.	ERCP: (n=39) No dropouts 4 failures Percutaneous: (n=36) No dropouts 8 failures	Percutaneous stents were initially 6Fr and exchanged 2-3 days later to 12 Fr while endoscopic stents were 10 Fr in size	Adequate outcome measures used. Outcomes were not assessed blindly.	Intention-to-treat analysis used. Results were also analyzed taking into account relevant confounders that were not balanced.	Good
Pedersen, Lassen, De Muckadell et al., 1998	RCT (n=34) Good comparability - Randomization by computer generated numbers and sealed numbered envelopes - Baseline characteristics similar for age, type of cancer, and no SSD for gender	Stent above SO (n=22) 22 randomized - 5 technical failures crossed over. Final n=17. No other dropouts. Stent across SO (n=19) 19 randomized - 2 withdrawn for curative surgery. Final n=17. No other dropouts.	Adequate for comparison.	Adequate outcome measures used. Outcomes were not assessed blindly.	Method of analysis primarily based on treatment received. Results for one outcome reported using intention-to-treat.	Fair
DePalma, Galloro, Iovino et al., 2001	RCT (n=157) Good comparability - Randomization by sealed opaque envelopes - Baseline characteristics similar	Unilateral stent (n=79) No dropouts Bilateral stent (n=78) No dropouts	Adequate for comparison.	Adequate outcome measures used. Outcomes were not assessed blindly.	Intention to treat used.	Good
Chang, Kortan, and Haber 1998	Retrospective study (n=141) Baseline patient characteristics were comparable for age, gender, and tumor type	All subjects included in analysis	Adequate for comparison.	Adequate outcome measures used. Outcomes were not assessed blindly.	Analysis made some attempts to stratify results by Bismuth type, but did not fully consider possible confounders.	Fair
Deviere, Baize, de Toeuf et al., 1988	Retrospective study (n=70) Baseline patient characteristics were not reported other than stricture type	All subjects included in analysis	Adequate for comparison.	Adequate outcome measures used. Outcomes were not assessed blindly.	Analysis made some attempts to stratify results by Bismuth type, but did not fully consider possible confounders.	Poor

43

American Society of Anesthesiology's performance status classification

The ERCP literature systematically reviewed for this report also included nine studies comparing various alternative ERCP treatment techniques. The comparisons reported in these studies were sufficiently dissimilar from the studies reviewed in preceding sections on palliative treatments of pancreaticobiliary malignancy that they are briefly summarized separately in this section. The quality assessments of these studies are detailed in Table 43 and the results of these studies are in Tables 44-46.

Review of Evidence: Stent Material and Design

Four studies, including two randomized controlled trials (one quality rated as "Good" and one as "Fair") and two nonrandomized studies (both rated "Poor" quality) compared different features of endoscopically placed stents for palliation of pancreaticobiliary malignancy (Tables 44-46.).

Table 44. Comparison of Plastic versus Teflon™ (more...)

Table 44. Comparison of Plastic versus Teflon™ stents

Study	N	Population and Interventions	Outcomes	Adverse Events	Comments
Randomized Controlled Trials
van Berkel, Boland, Redekop et al., 1998	84	Patients with distal malignant biliary stricture. No previous drainage procedure. Pancreas ca = 76 Papilla ca = 1 Bile duct ca = 5 Metastasis = 2 42 Teflon™ stents 42 polyethylene stents (Amsterdam-type) All stents 10Fr and 9cm Baseline characteristics comparable.	Median survival (days) Teflon™ 165   Poly 140 p=0.6 Successful biliary drainage Teflon™ 90%   Poly 92%   Median stent patency (days) Teflon™ 83   Poly 80 p=0.93 No significant differences found in: Mean weight gain for 26 removed stents	Perioperative mortality Teflon™ 14% Poly 14% Early procedure-related complications Teflon™ 4 (10%) Poly 4 (10%) Late complications   Stent Repeat #   dysfunc ERCP ERCP Teflon™ 28 24 79 Poly 29 25 75	Univariate analysis of factors associated with reduced stent patency was reported. Previous failure of cannulation (p=0.03) Previous CBD contrast injection without papillotomy (p=0.004) Previous papillotomy (p=0.08) Gender, age>75, jaundice> 14 days, bilirubin > 300 mmol/L not significant factors.

van Berkel, Boland, Redekop et al. (1998, n=84, "Fair") randomized patients to receive stents made of Teflon™ versus stents made of polyethylene and found no significant differences in efficacy or complications (Table 44). Median stent patency duration was 83 days for Teflon™ stents and 80 days for polyethylene stents (p=0.93).

Table 45. Comparison of different caliber stents

Table 45. Comparison of different caliber stents

Study	N	Population and Interventions	Outcomes	Adverse Events	Comments
Prospective observational studies
Pedersen 1993	89	Pts with malignant biliary strictures 31 Single 7 Fr (S7) 45 Single 10 Fr (S10) 13 Double 7Fr (D7) 85% of all patients also had sphincterotomy, evenly distributed between 7 and 10 Fr. 7 Fr stent chosen when no large bore ERCP scope available. Baseline patient characteristics were different for age (7Fr group younger than 10Fr group). No SSD in stenosis location, gender, or type of cancer.	Median Stent Patency (days) Median, 25%-75% range S7 67 (20-336) S10 144 (39-237) D7 110 (62-145) Total 110 (33-237) P=0.11, comparing 7Fr vs. 10Fr	Mortality (2-week) S7 (n=31) 4 (13%) S10 (n=45) 4 (9%) D7 (n=13) 2 (15%) p=0.84 Total Early Complications S7 (n=31) 13% S10 (n=45) 22.1% D7 (n=13) 23.1% p=n.s. Fever S7 (n=31) 9.7% S10 (n=45) 17.7% D7 (n=13) 23.1% p=n.r. Bleeding S7 (n=31) 6.5% S10 (n=45) 4.4% D7 (n=13) 0% p=n.r. Perforation S7 (n=31) 3.2% S10 (n=45) 0% D7 (n=13) 0% p=n.r.
Retrospective studies
Speer, Cotton, MacRae et al., 1988	79	All patients receiving stent palliation for malignant obstructive jaundice 28 8Fr pigtail stents 51 10Fr straight stents Baseline patient characteristics similar for age and site of obstruction.	Median Stent Patency (weeks) 8 Fr 12 10 Fr 32 p<0.001 Patency advantage of 10Fr stents primarily in first month.	Early complications (2 week) Cholangitis 8 Fr (n=28) 13 (34%) 10 Fr (n=51) 3 (5%) p<0.01 (text) Local perforation 8 Fr (n=28) 2 (5%) 10 Fr (n=51) 4 (5%) p=n.s. Stent migration 8 Fr (n=28) 3 (8%) 10 Fr (n=51) 2 (3%) p=n.s. Late complications Need for stent replacement 8 Fr 12 (43%) 10 Fr 13 (25%) p=n.r.

Pedersen (1993, n=89, "Poor") and Speer, Cotton, MacRae et al. (1988, n=79, "Poor") both compared outcomes using different caliber stents, but neither of these studies uses a randomized, controlled design (Table 45). Speer, Cotton, MacRae et al. (1988) found significantly longer median stent patency for 10Fr stents compared with 8Fr stents (32 weeks vs. 12 weeks, p<0.001). Complications reported included a lower rate of cholangitis with 10 Fr stents (5% vs. 34%, p<0.05), and similar rates of local perforation and stent migration. However, the 8Fr stents had pigtail-shaped ends compared with straight-shaped 10Fr catheters, a potential confounding factor in interpreting this study. Pedersen (1993) did not reveal a statistically significant difference in stent patency comparing 10Fr and 7 Fr, and did not show significant differences in total complication rates. However, this study also suffered from baseline differences in age, with younger patients receiving 7 Fr stents, increasing concerns over interpretation of findings.

Table 46. Comparison of stents with or without sideholes (more...)

Table 46. Comparison of stents with or without sideholes

Study	N	Population and Interventions	Outcomes	Adverse Events	Comments
Randomized Controlled Trials
Sung, Chung, Tsui et al., 1994	70	Most pts (93%) had malignant obstruction SH= side-hole stent (n=35) NSH = no side-hole (n=35) 10Fr stents Patient characteristics show no SSD for age, gender, diagnosis, location of stent, prior stent	Biochemical improvement at 4 weeks SH (n=35) 95%   NSH (n=32) 78% p>0.1 All stent patency (weeks), median (range) SH (n=35) 7.8 (2.6-28)   NSH (n=32) 7.9 (0.6-28) p>0.1 Initial stent patency (weeks), median (range) SH (n=35) 9.5 (6.3-28)   NSH (n=32) 8.0 (0.6-28) p>0.1 Second stent patency (weeks), median (range) SH (n=35) 6.6 (2.6-19.9)   NSH (n=32) 5.6 (0.9-23.3) p>0.1	Mortality SH (n=35) 8 (23%)   NSH (n=32) 8 (25%) p=n.r. Fever SH (n=35) 82%   NSH (n=32) 83%? p=n.r.

Sung, Chung, Tsui et al. (1994, n=70, "Good") randomized patients to receive 10Fr stents with or without sideholes (Table 46). No statistically significant differences were noted in stent patency and reported complications appeared similar, although statistical analysis was not reported.

None of these studies provides a sufficient basis for a conclusion regarding the relative efficacy the stent features being compared.