Download PDF
Endoscopy 2007; 39(3): 222-228
DOI: 10.1055/s-2007-966253
Original article
© Georg Thieme Verlag KG Stuttgart · New York

The roles of magnetic resonance and endoscopic retrograde cholangiopancreatography (MRCP and ERCP) in the diagnosis of patients with suspected sclerosing cholangitis: a cost-effectiveness analysis

S.  Meagher1 , I.  Yusoff2 , W.  Kennedy1 , M.  Martel1 , V.  Adam1, 3, 4 , A.  Barkun1, 3, 4
  • 1Division of Gastroenterology, McGill University Health Centre, Montreal General Hospital Site, McGill University, Montreal, Québec, Canada
  • 2Division of Gastroenterology, New South Wales University, Perth, Australia
  • 3Division of Clinical Epidemiology, McGill University, Montreal, Québec, Canada
  • 4McGill University Health Centre, Montreal General Hospital Site, McGill University, Montreal, Québec, Canada
Further Information

A. N. Barkun, MD 

Division of Gastroenterology

The McGill University Health Centre - the Montreal General Hospital Site

1650 Cedar Avenue

Montréal

Québec, H3G 1A4

Canada

Email: alan.barkun@muhc.mcgill.ca

Publication History

submitted 31 March 2006

accepted after revision 29 September 2006

Publication Date:
26 March 2007 (online)

Also available at
 PDF Download Permissions and Reprints
Table of Contents

See also:

Commentaire de travail de S. Meagher et al., pp. 222Endoscopy 2007; 39(03): 265-265
DOI: 10.1055/s-0032-1306918

Background and study aims: The optimal approach for diagnosing sclerosing cholangitis remains unclear in the face of competing imaging technologies. We aimed to determine the most cost-effective strategy.

Patients and methods: A decision model compared three approaches in the work-up of patients with suspected sclerosing cholangitis; all included an initial test, with, if unsuccessful, performance of a second cholangiographic method. They were magnetic resonance cholangiopancreatography (MRCP) and endoscopic retrograde cholangiopancreatography (ERCP), termed ”MRCP_ERCP”, ERCP and MRCP (”ERCP_MRCP”), or ERCP and a repeat ERCP (”ERCP_ERCP”). The implications of true and false positive and negative results with regard to costs and procedural complications were considered, including that of a liver biopsy, if indicated as a result of a negative work-up in the face of persistent clinical suspicion. The unit of effectiveness adopted was that of a correct diagnosis. Probability assumptions were derived from published literature, while cost estimates were derived from time-motion microanalyses or a national database, and expressed in Canadian dollars at 2004 values. Sensitivity analyses, including clinically relevant threshold analyses, were carried out.

Results: The average cost-effectiveness ratios were $ 414 for MRCP_ERCP, $ 1101 for ERCP_MRCP and $ 1123 for ERCP_ERCP, per correct diagnosis. The ERCP_MRCP strategy was dominated (more expensive and less effective) by MRCP_ERCP, while ERCP_ERCP was more effective and more costly than MRCP_ERCP, at $ 289 292 per additional correct diagnosis. Sensitivity and threshold analyses confirmed the robustness of these findings.

Conclusions: Based on the model assumptions, a strategy of initial MRCP, followed, if negative, by ERCP is currently the most cost-effective approach to the work-up of patients with suspected sclerosing cholangitis.

#

Introduction

Primary sclerosing cholangitis (PSC) is an uncommon chronic cholestatic liver disease occurring alone or in association with pre-existing illnesses, often suspected clinically on the basis of liver test or ultrasound abnormalities [1], and that may lead to cholangiocarcinoma.

Endoscopic retrograde cholangiopancreatography (ERCP) is considered the diagnostic test of choice for the early detection of PSC, with a sensitivity and specificity nearing 100 % [1]. However, ERCP is an invasive therapeutic procedure that is associated with significant complications, including pancreatitis in 4 % - 9 % of patients [2].

In contrast, magnetic resonance cholangiopancreatography (MRCP) is a noninvasive biliary imaging method with excellent accuracy [3] when a 1.0-T magnet is used. There unfortunately exists great variability in the availability of and accessibility to high quality MRCP to allow for a confident diagnosis of PSC. Its exact role thus remains unclear and requires better characterization. Furthermore, the comparative costs of each approach have been poorly quantified, let alone compared.

The aim of this study is thus to compare the cost-effectiveness of ERCP and MRCP-based strategies in the diagnosis of PSC.

#

Methods

#

Model structure, patient population, and outcome

A decision tree was developed to evaluate the cost-effectiveness of three different strategies in investigating patients with a suspected diagnosis of PSC. Each strategy is represented by one main branch and includes a pair of cholangiographic methods. Each method can initially either technically succeed or fail, and in the latter case it is followed by the second imaging attempt using a different or the same method. These are: MRCP followed, if needed, by ERCP (termed ”MRCP_ERCP”), ERCP followed, if unsuccessful, by MRCP (”ERCP_MRCP”), or ERCP followed, if needed, by a repeat ERCP attempt (”ERCP_ERCP”). In the case of a successful procedure, the obtained diagnosis can either be a true positive or true negative, or a false positive or false negative; moreover, ERCP can be associated with a complication or not, whether successful or failed. The model also considered patient management up to and including the performance of a liver biopsy and its attendant risks and costs in cases where ERCP or MRCP or both were negative, so as to encompass the full burden of true and false positive and negative imaging test results. [Figure 1] shows the structure for the MRCP_ERCP branch. Similar structures were adopted for the other approaches.

Zoom Image

Figure 1 The decision tree structure displays three possible clinical decisions that can be taken by a physician when deciding on imaging, following initial laboratory data and ultrasound, in a patient with suspected primary sclerosing cholangitis. The initial node at the left is a decision node allowing one of three possible strategies, including ERCP followed (if negative or unsuccessful) by MRCP; MRCP followed by ERCP; or ERCP followed by repeat-ERCP. The MRCP_ERCP branch is shown in detail. The branches of the tree are used to schematize the main health states that will occur as a consequence of each of the three possible imaging strategies. These steps include the potential failure or success of the procedure, the yield of the test (positive or negative), the potential complications following the procedure, and the need for a subsequent liver biopsy with its attendant possible complications. The tree also models for true or false positives and negatives. Each branch is linked to an associated cost and an associated probability for each event that occurs in the tree. At the far right of the tree are all possible heath outcomes resulting from the possible scenarios, with their associated final cost and probability. The final cost is the sum of all the costs included at each step for a given imaging strategy. The final probability is the product of all probabilities that occur throughout the course of a given imaging strategy

The study population included patients eligible to undergo both MRCP and ERCP.

Possible states of health include a possible outcome of successful or failed imaging method, a true- or false-positive, or true- or false-negative diagnosis of PSC, and the occurrence of a complication or not. Each state of health is associated with a probability and a cost. No procedural complications were attributable to MRCP. All second attempts at diagnosis were considered successful and their resulting diagnosis considered accurate. The unit of effectiveness adopted for the analysis is that of a correct diagnosis.

The decision model was constructed using the Treeage Pro software (TreeAge Software Inc., Braintree, Massachusetts, USA).

#

Probability assumptions

All relevant studies were identified through a Medline search of the previous 15 years, up to August 2005, using as key words ”ERCP,” MRCP,“ ”sclerosing cholangitis,” and ”cost-effectiveness.” Studies were selected according to the appropriateness of the patient population assessed, the ability to extract the required proportions, and the choice of diagnostic method used. All probabilities used for the model were obtained from relevant literature, and were abstracted by two independent observers ([Table 1]). Pooling of the data was carried out by summing appropriate numerators and denominators, thus weighting according to the number of patients in each study. Data of interest pertained to the probabilities of technical failure and success, the post-ERCP complications requiring hospitalization (cholangitis, cholecystitis, pancreatitis, and perforation), as well as the major post-liver biopsy complications such as hemorrhage.

Table 1 Probabilities adopted in the model: point estimates and range for sensitivity analysis
Probability Point estimate Range, 95 % CI References
ERCP success 0.9658 0.96 - 0.97 Christensen et al. [12], Loperfido et al. [2], Masci et al. [13]
ERCP true-positive rate 1 1.00 - 1.00 Christensen et al. [12], Loperfido et al. [2], Masci et al. [13], Freeman [14], Pasanen [15]
ERCP true-negative rate 1 1.00 - 1.00 Christensen et al. [12], Loperfido et al. [2], Masci et al. [13], Freeman et al. [14], Pasanen et al. [15]
Complication post ERCP 0.0366 0.029 - 0.044 Christensen [12], Loperfido [2], Masci et al. [13], Freeman et al. [14]. Pasanen et al. [15]
MRCP success 0.9139 0.87 - 0.95 Textor et al. [16], Angulo et al. [17], Fulcher et al. [3], Ernst et al. [18]
MRCP true-positive rate 0.8586 0.77 - 0.92 Textor et al. [16], Angulo et al. [17], Fulcher et al. [3], Ernst et al. [18]
MRCP true-negative rate 0.9688 0.94 - 0.99 Textor et al. [16], Angulo et al. [17], Fulcher et al. [3], Ernst et al. [18]
Complication post liver biopsy 0.005 0.0037 - 0.0064 Janes et al. [19], Pasha et al. [20], Garcia-Tsao & Boyer [21], McGill et al. [22]
ERCP, endoscopic retrograde cholangiopancreatography; MRCP, magnetic resonance cholangiopancreatography

Success and failure probabilities for repeat ERCP or MRCP were assumed to be the same as the initial success rates estimated from the literature.

The diagnostic yields, with corresponding 95 % confidence intervals, were obtained from two-by-two contingency tables that were reconstructed for every pertinent source article.

#

Cost assumptions

The ERCP and MRCP costs were estimated based on a time-motion costing microanalysis carried out at the McGill University Health Center using validated methodology [4].

The biopsy cost data were based on the Day Procedure Group section of the national database, CIHI 2004, of the Canadian Institute for Health Information [5].

The treatment costs for post-ERCP complications were estimated using the case mix group (CMG) database from CIHI 2004, selected for the 18 - 69-year-old group, the third level of complexity, and the relevant diagnosis group, that is CMG 325 for pancreatitis, 285 for perforation, and 329 for both cholangitis and cholecystitis. This source was also used for determining the cost of post-biopsy complications, where the first complexity level and the weighted average (90 % - 10 %) of two CMGs (323 and 901) were selected.

All the CIHI resource intensity weights were transformed into costs using the cost per weighted case (2001, CIHI). They were also adjusted for the inflation rate of the Consumer Price Index for the health sector of the United States [7]. The professional fees were excluded from all costs. All costs were expressed in Canadian dollars at 2004 values.

For the sensitivity analysis, we varied probability estimates across their 95 % confidence interval range.

#

Presentation of results, and sensitivity and threshold analyses

Results of the cost-effectiveness analysis are reported as incremental cost-effectiveness ratios (ICERs). More costly yet less effective strategies were classified as ”dominated” in the base case analysis.

Sensitivity analyses were guided by Tornado diagrams that identified those variables for which a greater than 20 % range in effectiveness or costs results existed as the point estimate was varied across the adopted ranges for probabilities or costs. Out of clinical interest, we also performed a priori determined threshold analyses for the effectiveness of MRCP_ERCP, ERCP_MRCP and ERCP_ERCP, and their procedural costs.

#

Results

#

Probability estimates

Excluded studies. Reasons for rejecting source studies were the choice of detection method for both diagnostic and therapeutic purposes [8], and features of the studied patient population (such as the sole inclusion of patients with biliary strictures [9], suspected pancreatic disease [10], or advanced PSC [11]).

Included studies. The probability estimates are detailed in [Table 1]. MRCP sensitivity and specificity values obtained from Angulo et al. [17] were calculated using ERCP/percutaneous transhepatic cholangiography (PTC) as a reference standard rather than ERCP alone. For the Fulcher et al. [3] article, we included the sensitivity and specificity calculated as an average from both observers, as this value was the most conservative. The recorded post-biopsy complications requiring hospitalization differed across studies and included hemorrhage, hypotension, hemothorax and pneumothorax, as in McGill et al [22]. However, only fatal and nonfatal hemorrhages were considered as major complications by McGill et al., whereas Janes et al. [19] solely considered persistent pain, orthostatic hypotension, and lightheadedness. No details concerning the nature of the major complications were presented in the review by Garcia-Tsao & Boyer [21].

ERCP procedural estimates. Estimates for ERCP success and complication rates were derived from five different study populations totalling 8012 consecutive patients [2] [12] [13] [14] [15]. The probabilities of ERCP failure and success for the detection of PSC were derived from three of the five studies [2] [12] [13]. Probabilities of post-ERCP pancreatitis were obtained from all five [2] [12] [13] [14] [15]. Post-ERCP perforation and cholangitis probabilities were derived from all but one series [15]. Post-ERCP cholecystitis values were derived from Christensen et al. [12], Loperfido et al. [2], and Masci et al. [13]. The overall rate for post-ERCP complications was 0.037 (95 % CI 0.029 - 0.044). True-positive and true-negative rates of PSC diagnosis using ERCP were, by definition, both 1 (100 %) ([Table 1]).

MRCP procedural estimates. MRCP accuracy and effectiveness estimates were derived from four studies involving a total of 257 patients [3] [16] [17] [18]. Inclusion criteria were suspected biliary disease [17], and clinically supported or ERCP-diagnosed PSC [3]. Exclusion criteria were pregnancy, use of a cardiac pacemaker, cerebral aneurysm clip, morbid obesity [3], intraocular metallic foreign bodies, caval and intravascular implants, and claustrophobia [17]. Patients were evaluated for specific criteria related to physical examination or serum chemistry. High levels of alkaline phosphatase, aspartate transaminase, bilirubin [16], aminotransferase, and prothrombin [17] were included as signs of PSC. The majority of patients showed signs of liver disease such as hepatosplenomegaly (jaundice), skin excoriation (xanthelasma) [16] or cholestasis [17]. ERCP was the gold standard [3] [16] in all but one study, where ERCP/PTC was used [17].

Procedural accuracy and effectiveness. The probability of initial procedural success was 0.97 (95 % CI 0.96 - 0.97) for ERCP and 0.91 (95 % CI 0.87 - 0.95) for MRCP. The true-positive rate for diagnosis using MRCP was 0.86 (95 % CI 0.77 - 0.92). The true-negative rate for MRCP was 0.97 (95 % CI 0.94 - 0.99).

Complications of liver biopsy. Complications following liver biopsy were compiled from four studies involving a total of 10 936 patients [11] [12] [13] [20], giving a probability of 0.005 (95 %CI 0.0037 - 0.0064).

#

Cost estimates for procedures and their complications

The cost estimates for ERCP and MRCP were $ 1050 and $ 244, respectively. Costs of post-ERCP complications are listed in [Table 2]. The cost of performing a liver biopsy was $ 905, and the cost incurred because of attendant complications was $ 2789.

Table 2 Costs adopted in the model: point estimates and range for sensitivity analysis
Cost of procedure or complication Cost Range
MRCP 244 122 - 366
ERCP diagnostic 1050 525-1575
Pancreatitis 5031 2515-7547
Cholangitis 4604 2302-6906
Cholecystitis 4604 2302-6906
Perforation 5584 2792-8376
Liver biopsy 905 n. a.
Complications of a liver biopsy 2789 n. a.
All costs are expressed in Canadian dollars at 2004 values.
n. a., none available
Ranges are defined as the interval spanning - 50 % to + 50 % of the pointestimate value.
#

Cost-effectiveness of the different strategies

Analysis per correct diagnosis. [Table 3] lists the different strategies according to average costs per patient and incremental costs (to image one additional patient), effectiveness (the probability of making a correct diagnosis), incremental effectiveness, and the corresponding average and incremental cost-effectiveness ratios that allow us to draw conclusions about the cost-effectiveness impact of each strategy.

The proportion of correct diagnoses using the different approaches were 0.998 for MRCP_ERCP, 0.997 for ERCP_MRCP, and 1.0 for ERCP_ERCP ([Table 3]). Compared with MRCP_ERCP, the incremental proportions of correct diagnoses were - 0.00057 for ERCP_MRCP, and 0.0025 for ERCP_ERCP. The costs per correct diagnosis amounted to $ 413 for MRCP_ERCP, $ 1098 for ERCP_MRCP, and $ 1123 for ERCP_ERCP ([Table 3]). Compared with MRCP_ERCP, incremental costs per correct diagnosis were $ 685 for ERCP_MRCP, and $ 710 for ERCP_ERCP.

Average cost-effectiveness ratios were $ 414 for MRCP_ERCP, $ 1101 for ERCP_MRCP and $ 1123 for ERCP_ERCP. Incremental cost-effectiveness ratios (ICER) per correct diagnosis are shown in [Table 3]. The ERCP_MRCP strategy was dominated, being less effective and more expensive than MRCP_ERCP, while ERCP_ERCP was more effective, but at a higher cost of $ 289 292 per additional correct diagnosis compared with MRCP_ERCP.

Table 3 Results of the cost-effectiveness analysis
Strategy
MRCP_ERCP ERCP_MRCP ERCP_ERCP
Cost*, average, $ 413.1 1098 1123.3
Incremental cost, $ 684.8 710.2
Effectiveness, probability of correct diagnosis 0.99754 0.99697 1.00000
Incremental effectiveness, probability of correct diagnosis - 0.00057 0.00246
Cost-effectiveness ratio, $/correct diagnosis 414 1101 1123
Incremental cost-effectiveness ratio (ICER), $/correct diagnosis (Dominated) 289 292
* All costs are in Canadian dollars in 2004
#

Sensitivity analysis

Probabilities. The probabilities of MRCP failure and of true-positive diagnosis had the greatest effect on costs. As the probability of MRCP failure increased from 0 to 0.53, costs rose for MRCP_ERCP from $ 320 to $ 878 per correct diagnosis. An increase in MRCP sensitivity from 0.47 to 1.0 led to a decrease in costs for MRCP_ERCP from $ 798 to $ 340 per correct diagnosis ([Figure 2]).

Zoom Image

Figure 2 Sensitivity analysis of change in cost-effectiveness according to the sensitivity of MRCP.

As the probability of ERCP failure increased from 0.0073 to 0.051, costs of ERCP_ERCP rose from $ 1096 to $ 1140 per correct diagnosis, and from $ 1091 to $ 1108 per correct diagnosis for ERCP_MRCP.

Costs. Raising ERCP costs from $ 520 to $ 1580 resulted in a cost increase from $ 570 to $ 1634 per correct diagnosis for ERCP_MRCP, and from $ 577 to $ 1673 for ERCP_ERCP, while that of MRCP_ERCP remained stable ([Figure 3 a]. Raising MRCP costs from $ 120 to $ 370, resulted in an increase from $ 290 to $ 541 in cost per correct diagnosis for MRCP_ERCP, while both ERCP_ERCP and ERCP_MRCP values remained stable ([Figure 3 b]). Variations in the costs of complications had no effect on the cost-effectiveness of any of the three strategies.

Zoom Image
Zoom Image

Figure 3 Sensitivity analyses of change in cost-effectiveness: a according to the cost of ERCP; b according to the cost of MRCP.

#

Threshold analysis

All threshold values that would alter the conclusions reflected very unlikely clinical scenarios. MRCP_ERCP would stop being dominant and the most cost-effective approach if MRCP failure rose above 0.70, if the MRCP sensitivity value dropped below 0.18, if the cost of ERCP dropped below $ 298, or if the cost of MRCP were to rise above $ 1126 (all of these were well outside our range of plausible probabilities or procedural costs). No similar threshold was identified by increasing the ERCP failure rate.

The cost-effectiveness plots of ERCP_MRCP and ERCP_ERCP were almost perfectly parallel for all sensitivity analyses, excluding any possible threshold values that could alter the final conclusions.

#

Discussion

When assessing patients with PSC, clinicians need to consider not only the test performance characteristics of available individual methods of cholangiography, but also the implications of false-positive or false-negative findings, with regard to possible subsequent complications and costs. We have tried to include these in the current cost-effectiveness analysis, considering the work-up, if negative, all the way up to and including the performance of a liver biopsy for a persistent clinical suspicion of PSC, if indicated: that is, if imaging is negative with, for example, continued cholestasis on liver tests in a patient with a known colitis.

The cost-effectiveness analysis demonstrates the superiority of a strategy where MRCP is the initial diagnostic test, assuming conservative estimates for the failure rate of MRCP and probability assumptions with favorable assumptions for ERCP in determining the presence of PSC. Because of the methodology of the trials that assessed the accuracy of MRCP, the 16 % disease prevalence that corresponds to our point estimates for true-negative and true-positive rates is difficult to judge clinically. Alteration of this estimate is inherent within the sensitivity analyses that vary the estimates of both true-positive and true-negative values for MRCP. Regardless, the sensitivity analysis and clinical irrelevance of the calculated threshold values demonstrate the robustness of our conclusions.

The cost-effectiveness of initial MRCP is attributable to its low relative cost compared with ERCP, coupled with the absence of procedure-related complications for the level of diagnostic accuracy it provides. Such an assumption could not, however, have been validated without the use of a decision model, in the absence of a formal head-to-head trial, especially considering the subsequent impacts of false test results. Talwalker et al. [23] performed a cost-minimization analysis comparing the costs of diagnosing PSC utilizing a single cholangiographic method, (magnetic resonance cholangiography [MRC] alone versus ERCP alone). Their assumptions were based on institutional data. The average cost per correct diagnosis of PSC was $ US 549.64 with MRC compared with $ US 623.25 for ERCP. The average cost of managing subsequent post-ERCP complications among patients with PSC was $ US 2902.20. For ERCP to be the optimal initial test strategy, a prevalence of PSC of greater than 45 %, an MRC specificity of less than 85 %, or a reduction in the average cost per diagnosis to $ US 538.30 would be required.

We chose to incorporate a more complex set of diagnostic techniques because of the clinically relevant possibility of procedure failure and the consequences of incorrect test results.

Additional studies are required to assess the ”real-life“ test performance of MRCP in diagnosing PSC, but based on current published data from expert centers, it is unlikely that the accuracy would be low enough to alter our conclusions. Perhaps a greater issue is that of accessibility to the MR scanner. The performance of emerging competing imaging technologies would also, in time, need to be studied from a cost-effectiveness perspective. This may be the case for endoscopic ultrasound which may be helpful in detecting extrahepatic PSC by determining the thickness of the common bile duct [24].

In conclusion, MRCP followed by ERCP, and ERCP followed by ERCP, were the two most cost-effective methods with the latter being very slightly more effective but much more expensive. For its part, ERCP followed by MRCP was both more costly and less effective than the previous combinations. Based on the model assumptions, a strategy of initial MRCP, followed, if negative, by ERCP is the most cost-effective approach to investigating patients with suspected PSC.

#

Acknowledgments

Dr. Barkun is a Research Scholar (Chercheur National) funded by the Fonds de la Recherche en Santé du Québec.

Competing interests: None

#

References

A. N. Barkun, MD 

Division of Gastroenterology

The McGill University Health Centre - the Montreal General Hospital Site

1650 Cedar Avenue

Montréal

Québec, H3G 1A4

Canada

Email: alan.barkun@muhc.mcgill.ca

#

References

A. N. Barkun, MD 

Division of Gastroenterology

The McGill University Health Centre - the Montreal General Hospital Site

1650 Cedar Avenue

Montréal

Québec, H3G 1A4

Canada

Email: alan.barkun@muhc.mcgill.ca

   Permissions and Reprints
Zoom Image

Figure 1 The decision tree structure displays three possible clinical decisions that can be taken by a physician when deciding on imaging, following initial laboratory data and ultrasound, in a patient with suspected primary sclerosing cholangitis. The initial node at the left is a decision node allowing one of three possible strategies, including ERCP followed (if negative or unsuccessful) by MRCP; MRCP followed by ERCP; or ERCP followed by repeat-ERCP. The MRCP_ERCP branch is shown in detail. The branches of the tree are used to schematize the main health states that will occur as a consequence of each of the three possible imaging strategies. These steps include the potential failure or success of the procedure, the yield of the test (positive or negative), the potential complications following the procedure, and the need for a subsequent liver biopsy with its attendant possible complications. The tree also models for true or false positives and negatives. Each branch is linked to an associated cost and an associated probability for each event that occurs in the tree. At the far right of the tree are all possible heath outcomes resulting from the possible scenarios, with their associated final cost and probability. The final cost is the sum of all the costs included at each step for a given imaging strategy. The final probability is the product of all probabilities that occur throughout the course of a given imaging strategy

Zoom Image

Figure 2 Sensitivity analysis of change in cost-effectiveness according to the sensitivity of MRCP.

Zoom Image
Zoom Image

Figure 3 Sensitivity analyses of change in cost-effectiveness: a according to the cost of ERCP; b according to the cost of MRCP.