Investigations examining the use of magnetic resonance cholangiography (MRC) for the diagnosis of primary sclerosing cholangitis (PSC) have described comparable accuracy when compared to endoscopic retrograde cholangiopancreatography (ERCP). The effectiveness of MRC based on overall cost, however, remains unknown. Our aim was to determine the average cost per correct diagnosis using MRC or ERCP as the initial testing strategy for the diagnosis of PSC. A decision analysis model was constructed employing diagnostic test parameters prospectively determined among 73 patients with clinically suspected biliary disease. ERCP was performed within 24 hours after MRC. Cost data were derived from average Medicare reimbursement fee schedules. The prevalence of PSC in the study cohort was 32%. The sensitivity and specificity of MRC for the diagnosis of PSC were 82% and 98%, respectively. The average cost per correct diagnosis of PSC was $724.00 for initial MRC (including the cost of ERCP following a negative MRC examination) versus $793.17 for initial ERCP. In the absence of biliary obstruction, the average cost per correct diagnosis of PSC was $549.64 with MRC versus $623.25 for ERCP. The average cost of managing post-ERCP–related complications among patients with PSC was $2902.20 (range, $1915.40-$5031.54). For ERCP to be the optimal initial test strategy, a prevalence rate of PSC greater than 45%, MRC specificity less than 85%, or reduction in the average cost per diagnosis to $538.30 would be required. In conclusion, MRC has comparable accuracy to ERCP and results in cost savings when used as the initial test strategy for diagnosing PSC. (HEPATOLOGY 2004;40:39–45.)
Endoscopic retrograde cholangiopancreatography (ERCP) has been considered the diagnostic test of choice for the evaluation of clinically suspected biliary tract disease.1, 2 Although considered a safe and effective procedure when performed by skilled operators, procedure-related complications from ERCP can occur in 3% to 8% of patients and include abdominal pain, pancreatitis, bleeding, common bile duct (CBD) perforation, biliary sepsis, and death.3, 4
Magnetic resonance cholangiography (MRC) for the study of biliary tract disease has only recently become available as an alternative to ERCP. Based on the difference in intensities between bile and surrounding tissues, complete visualization of the biliary tree can now be achieved. In comparison to ERCP, MRC is not invasive and has no significant morbidity when performed in appropriately selected individuals. The potential adverse effects of radiation exposure and contrast media associated with ERCP are also avoided with MRC.5
A number of investigations comparing MRC with ERCP suggest that both have similar diagnostic-test characteristics in the detection of biliary tract disorders,6–9 including primary sclerosing cholangitis (PSC).10, 11 However, the effectiveness of MRC for detecting PSC, described in the context of important economic outcomes such as resource utilization, has not been reported. Therefore, the aim of our investigation was to perform a cost-minimization analysis of MRC versus ERCP as the initial test strategy for the diagnosis of PSC.
ERCP, endoscopic retrograde cholangiopancreatography; CBD, common bile duct; MRC, magnetic resonance cholangiography; PSC, primary sclerosing cholangitis; EUS, endoscopic ultrasound.
Patients and Methods
The cost-minimization analysis is based on results from a prospective investigation evaluating the accuracy of MRC compared to ERCP in 73 patients with suspected biliary tract disease. Details of this study have been previously reported.10 All patients were observed to have clinical and/or biochemical evidence of unexplained cholestasis. Because of technical constraints, a total of 2 patients required conversion to percutaneous transhepatic cholangiography. For the purposes of this analysis, information from the 68 patients with quality diagnostic images on ERCP was used. In all instances, MRC was attempted prior to ERCP performance and successfully completed.
Magnetic Resonance Cholangiography Technique.
All ERCP procedures were completed within 24 hours of the preceding MRC. The examinations were performed using a General Electric Signa 1.5 Tesla MR scanner (Milwaukee, WI). Image acquisition was performed using a phase array torso multicoil. Two acquisitions each were performed in the axial and coronal planes using a single-shot fast-spin echo pulse sequence with a repetition time of infinity, average echo time equal to 90 milliseconds, matrix size equal to 256/256, and slice thickness of 5 mm with 0-mm skip. Each image slice required 2 seconds. Total breath-hold time averaged 20 seconds per patient. Multiple breath-hold scans were also performed to image the entire hepatic parenchyma if required. Three additional acquisitions were performed in the coronal plane and in planes approximately 45° oblique to the porta hepatis. A slice thickness of 5 cm, requiring a single 2-second breath hold, was used for the additional acquisitions.
MRC and ERCP Image Interpretation.
The interpretation of images from MRC and ERCP was performed by two radiologists with experience in hepatic and biliary tract imaging. These physicians were blinded to the clinical and laboratory information, procedure indication, and results of previous imaging studies in each subject. Differences in image interpretation were resolved by consensus between the two radiologists. MRC interpretation was performed prior to ERCP interpretation in all cases. All MRC sequences were evaluated for image quality, which was classified as “satisfactory” or “unsatisfactory” based on the ability to visualize biliary tree structures and the presence of artifacts. Intrahepatic bile ducts were examined for the extent of visualization and morphological features. Common hepatic and common bile ducts were examined for visualization, the presence of dilatation, and obstruction. Visualization of the gallbladder, cystic duct, and presence or absence of cholelithiasis was also recorded.
Identical criteria were applied for the evaluation of ERCP images using the known diameter of the duodenoscope as a correction reference for magnification. Characterization of the ampulla as well as the diagnosis of strictures, dilatation, filling defects, etiology of obstruction (cholelithiasis, inflammation, or malignancy), and PSC were made using standard endoscopic and radiological criteria. Final diagnoses in each subject were based on all available clinical information including medical history and physical examination, biochemical testing, imaging studies (including ultrasound, CT scan, and ERCP), and clinical follow-up.
Decision Analysis Model.
A decision analysis model using DATA version 3.5 (TreeAge Software, Inc.) software was developed to compare the expected average cost (1999 US dollars) per correct diagnosis of the initial use of MRC or ERCP for detecting PSC (Fig. 1). Bayes theorem was used to determine the probability of a false-positive and false-negative result from both MRC and ERCP.12 The potential for procedure-related complications with ERCP (abdominal pain, pancreatitis, and CBD perforation) was accounted for in the model. Therapeutic maneuvers required at ERCP in the diagnostic and confirmatory setting, including any combination of balloon dilatation, endoscopic sphincterotomy, bile duct biopsies, stone extraction, and stent placement/removal, were also examined (Fig. 2). A discounting of costs was not performed given the brief time horizon associated with diagnostic testing.
Data and Assumptions.
Key assumptions and estimates for the model are outlined in Table 1. The prevalence of PSC was calculated as a proportion of the total cohort at risk. Sensitivity and specificity parameters for both MRC and ERCP were used in determining the likelihood of having PSC. Ranges of probability estimates were derived from the literature or assumed when published information was not available. The following assumptions were incorporated into the model:
1Values of the diagnostic parameters (sensitivity, specificity) were the same for ERCP whether used as an initial or a confirmatory test.
2The occurrence of a negative test result by MRC was subsequently followed by ERCP with the assumption that a clinical suspicion for PSC remained present after MRC.
3Probabilities of post-ERCP complication were the same for ERCP whether used as an initial or a confirmatory test.
4The management of post-ERCP abdominal pain was assumed to include a 24-hour hospitalization in a semiprivate room with intravenous hydration and pain medications.
5The management of post-ERCP acute pancreatitis was assumed to include a 3-day hospitalization in a semiprivate room, intravenous hydration and pain medications, and the performance of an abdominal CT scan with oral and intravenous contrast.
6The management of post-ERCP bile duct perforation was assumed to include a 5-day hospitalization in a semiprivate room, intravenous antibiotics and hydration, pain medications, and an abdominal CT scan with oral and intravenous contrast. No surgical intervention to repair the perforation was assumed for the purposes of this analysis.
Table 1. Estimates and Probabilities for Model Parameters
Direct costs for ERCP included consideration of the following components: supplies (including medications and IV equipment, occlusion balloon catheter for contrast injection, and contrast material), recovery room fees, and physician fees. Direct costs for MRC included consideration of film and processing fees, technician fees, and physician fees (Table 1). MRI scanner, side-viewing duodenoscopes and computer processing units, and ERCP procedure-room costs were considered fixed and excluded from the analysis. Facility fee charges were excluded from the cost analysis. Indirect costs such as wages lost from missed work were not included. Hospitalization costs were assumed to include the components for each post-ERCP complication as described in Materials and Methods.
All procedure costs (technical and professional) were estimated using average Medicare reimbursement schedules assigned by CPT procedure code. This included the costs associated with additional therapeutic maneuvers required at ERCP when indicated. Supply costs were estimated by multiplying institutional charges by the Medicare cost-to-charge ratio for fiscal year 1999 (0.5562). All costs were inflation-adjusted to the 1999 US dollar.
To determine the extent to which results affected model assumptions, sensitivity analyses were performed on individual parameter estimates to assess their impact on robustness of model outcomes. The following variables were varied: prevalence of PSC (1%–100%), sensitivity and specificity of MRC (50%–100%), probabilities of ERCP-related complications (50%–200% of baseline), and total costs of MRC, ERCP, and hospitalization (50%–200% of baseline). Threshold analysis was conducted within the context of the decision model to estimate the effect of parameter values beyond which the optimal test strategy changed.
In the original prospective study,10 the 73 patients (33 male, 40 female) had a mean age of 56 years (range, 19–94 years). Twenty-two individuals (30%) were found to have normal biliary tree anatomy by ERCP. Forty-two patients (58%) had benign biliary tract disease, and 9 individuals (12%) were diagnosed with malignant biliary tract disease. Twenty-three of 73 subjects (32%) were diagnosed with PSC.
All patients with PSC had both intra- and extrahepatic biliary system involvement confirmed by ERCP. Therapeutic interventions with ERCP were performed in 16 of 22 subjects (73%) with PSC. These included endoscopic sphincterotomy, balloon dilatation, stone extraction, biliary stent placement and/or removal, and histological sampling. Six patients (8.8%) who underwent ERCP developed postprocedure complications of abdominal pain (n = 3), acute pancreatitis (n = 2), and CBD perforation (n = 1) not requiring surgical management.
Diagnostic test performance characteristics for MRC included sensitivity and specificity values of 82% and 98%, respectively, for the detection of PSC. The positive and negative predictive values were 95% and 92%, respectively. Overall accuracy of MRC for the diagnosis of PSC was 93%.
Base Case Analysis.
Among the entire subgroup of patients diagnosed with PSC (32%), the average cost per correct diagnosis using the test strategy of initial MRC was $724.00 versus $793.17 for initial ERCP. For MRC, the cost included expenses related to performing confirmatory ERCP when MRC results were negative and ERCP-related complications (Fig. 2). In the absence of biliary obstruction, the average cost per correct diagnosis for initial MRC was $549.64 versus $623.25 for ERCP. In patients with PSC and pretest evidence of biliary obstruction, the average cost per correct diagnosis for initial MRC was increased to $1,499.64 versus $950.00 for initial ERCP alone.
The average cost of managing post-ERCP–related complications in patients with PSC was $2,902.20. Values ranged from $1,915.40 (for abdominal pain after balloon dilation) and $5,031.54 (for perforation after balloon dilatation and stone extraction) when initial MRCP demonstrated evidence of biliary obstruction.
Sensitivity analyses were performed to determine the robustness of base case results following variations in the model's probability estimates. One-way sensitivity analyses revealed that a prevalence rate above 45% for PSC would change the preferred test strategy from initial MRC to ERCP (Fig. 3). When the specificity of MRC was less than 85%, initial ERCP was the optimal test strategy (Fig. 4). No change in test strategy was observed by varying the sensitivity of MRC for detecting PSC. Variation in the costs of initial MRC, ERCP, therapeutic maneuvers at ERCP, additional imaging by CT scan, or items related to hospitalization did not influence the choice of MRC as the preferred initial test strategy. With a study cohort prevalence rate of 32% for PSC, threshold analysis demonstrated that an average cost per correct diagnosis of $538.25 (including procedure-related complications) would be required for initial ERCP to be the preferred strategy over MRC.
The invasiveness and risk for complications with ERCP has created a need to identify alternate methods for confirming the presence of biliary tract disease. MRC appears comparable to ERCP for detecting bile duct abnormalities including PSC. From this cost-minimization analysis, the test strategy of initial MRC was associated with a reduced average cost per correct diagnosis compared to initial ERCP for evaluating patients with suspected PSC. Importantly, the subsequent use of ERCP following a negative MRC examination—intended to reflect clinical practice—was accounted for in these results. In patients with PSC and no evidence of significant biliary obstruction requiring therapeutic ERCP, the cost difference was even greater. From sensitivity analyses, a prevalence rate for PSC greater than 45% or MRC specificity less than 85% would be necessary for initial ERCP to become the preferred test strategy for detecting PSC. The initial use of ERCP would also be favored if the average cost per correct diagnosis (including the cost of postprocedure complications) was reduced to at least $538.30.
In addition to its noninvasive approach, the initial use of MRC for diagnosing PSC has several other advantages. Identifying high-grade biliary obstruction on MRC improves the ability to justify the risk for potential complications when therapeutic ERCP is required.5–10 Additionally, the decision to refer for percutaneous transhepatic cholangiography if high-grade distal obstruction would inevitably result in an incomplete ERCP examination can be facilitated.9–11 Global hepatic parenchymal assessment to detect the presence of occult cirrhosis and portal hypertension can also be performed.13–16 In terms of patient preference, the average time for MRC completion is 15 to 20 minutes17 and is well tolerated compared to ERCP.18
In PSC, the presence of widespread annular strictures with proximal bile duct dilatation are characteristic cholangiographic findings on MRC and ERCP.9–16 In a retrospective review of 22 patients with PSC referred for MRC examination,15 the presence of abnormal bile duct findings were observed for all examination. Typical cholangiographic features of PSC, including intrahepatic bile duct dilatation (77%), stenosis (64%), and beading (36%), were observed. Revelon et al.16 observed that bile duct dilatation correlated with histological changes of inflammation in 29 of 40 (72%) patients. More recently, a prospective case-control study comparing MRC results between 34 patients with PSC and 68 age-matched controls with other hepatobiliary disease was reported.11 For the detection of PSC by MRC, the sensitivity and specificity ranged from 85% to 88% and 83% to 97%, respectively. Similar results were observed among patients with PSC in our previous study.10
Several published investigations using decision analysis techniques have examined the roles of MRC and ERCP for evaluating biliary tract disease. In a study comparing endoscopic ultrasound (EUS) or MRC versus ERCP for suspected biliary obstruction,19 the proportion of patients correctly identified with a normal biliary tree was 95% for EUS and 65% for MRCP (P < .02). The cost per patient for obtaining a correct diagnosis with EUS was $34.00 less than for MRCP. A relationship between poor MRC test performance and low disease (biliary tract abnormality) prevalence was also observed. Although described as a cost-utility analysis, data on health state preference for using accepted techniques was not reported. In a separate investigation comparing MRC and ERCP among patients with suspected biliary disease, the results obtained from MRC were said to avert less than 3% of all ERCP examinations performed at one institution.20 The investigation, however, did not clearly report that strict blinding of radiologists to ERCP results was performed while interpreting MRC images. From a detailed cost-effectiveness analysis in patients with acute biliary pancreatitis,21 a strategy of EUS followed by ERCP was the most cost-effective alternative when prevalence rates for choledocholithiasis were between 7% and 45%. Initial ERCP was the preferred approach when the risk for choledocholithiasis was greater than 45%. Despite the use of cost-effectiveness analysis, data on health state preference or utility were not reported.
In the present investigation, the technique of cost-minimization analysis for comparing initial test strategies of MRC and ERCP for the diagnosis of PSC was used. Advantages of this design include (1) the systematic incorporation of diagnostic test parameters in calculating the probability of a true positive outcome, and (2) accounting for the effect of unfavorable outcomes associated with further testing because of false-negative results.12 Cost-minimization analysis also determines whether differences in resource utilization between MRC and ERCP exist based on the premise that similar test characteristics are observed for both diagnostic modalities. In contrast, a cost-effectiveness analysis describes the differences in cost relative to gains in quality-adjusted life years between 2 or more competing strategies.22 Individual health state values or preferences, which are used to measure quality-adjusted life years, are not widely available for patients with PSC or those undergoing diagnostic testing.
Defining the level of clinical suspicion for PSC or biliary obstruction in patients with cholestatic liver test abnormalities involves a multifactorial approach. For example, the presence of symptoms (jaundice, pruritus), serum hyperbilirubinemia, and history of inflammatory bowel disease would likely influence the decision to pursue ERCP after negative MRC examination given the high likelihood of PSC. When a low degree of suspicion for biliary tract disease is present (as with the finding of asymptomatic cholestatic hepatitis with normal total bilirubin), liver biopsy rather than diagnostic ERCP may be more appropriate given a negative MRC exam. In the present analysis, in which 30% of patients had normal biliary tree anatomy, the pursuit of liver biopsy rather than diagnostic ERCP would still designate initial MRC as the cost-minimizing strategy.
Additional questions were raised by this analysis. The underdistension of bile ducts when examined by MRC may account for false-negative results in the presence of a biliary stricture.10, 11, 17 This is especially relevant in PSC given the 10% to 15% risk for developing cholangiocarcinoma as a complication of disease.32 Technical failure to cannulate the bile duct was not specifically examined given the rarity of this event in our original study. When considered, an increase in complication rates and perhaps cost from the use of percutaneous transhepatic cholangiography based on negative MRC results may be observed. The accuracy of MRC for cholangiocarcinoma in PSC also remains unknown. In addition, the diagnostic test performance characteristics of MRC in patients with early-stage PSC have not been clearly defined.17 For cost analyses, the use of reimbursement fee schedules may have biased results in favor of MRC because facility and supply costs may be discounted over several clinical areas. The use of metrics such as resource utilization rather than cost should be considered in future studies based on differences in payer mix seen in managed care organizations compared to referral centers.
Further investigations have attempted to estimate the potential impact of MRC as a diagnostic modality. It has been proposed that if MRC is available with accurate test performance features, up to 40% of patients with clinically suspected biliary tract disease could avoid diagnostic ERCP and less than 10% would require both MRC and ERCP.23, 24 These findings may have significant implications in clinical practice settings in which the prevalence of PSC is lower than what is observed in referral-based centers. In addition, MRC could be very useful for determining the presence of PSC in patients with serum cholestatic liver test abnormalities and inflammatory bowel disease. This cost-minimization analysis demonstrates that a test strategy of initial MRC compared to ERCP in the evaluation of patients with clinically suspected PSC has comparable diagnostic accuracy and cost savings.