Economic evaluation of MR cholangiopancreatography compared to diagnostic ERCP for the investigation of biliary tree obstruction

Transcription

1 International Journal of Surgery (2006) 4, 12e19 Economic evaluation of MR cholangiopancreatography compared to diagnostic ERCP for the investigation of biliary tree obstruction Yolanda Bravo Vergel a, *, Jim Chilcott b, Eva Kaltenthaler b, Stephen Walters c, Anthony Blakeborough d, Steven Thomas e a Department of Economics and Related Studies, Centre for Health Economics, University of York, Heslington, York YO10 5DD, UK b ScHARR Technology Assessment Group, School of Health and Related Research, University of Sheffield, S1 4DA, UK c Medical Statistics, School of Health and Related Research, University of Sheffield, S1 4DA, UK d Department of Radiology, Royal Hallamshire Hospital, Sheffield, S10 2JF, UK e Department of Radiology, Northern General Hospital, Sheffield, S5 7AU, UK KEYWORDS Magnetic resonance cholangiopancreatography; Endoscopy; Economic evaluation; Biliary obstruction Abstract Background: Use of magnetic resonance cholangiopancreatography (MRCP) for confirmation of presence of biliary obstruction is virtually risk-free. However, unlike diagnostic endoscopic retrograde cholangiopancreatography (ERCP), no therapeutic option can be offered simultaneously with MRCP. The aim of the study is to assess the cost-effectiveness of MRCP when compared with the conventional practice of diagnostic ERCP for the investigation of biliary obstruction in adults. Methods: Cost-effectiveness analysis from the perspective of the health care provider. Sensitivity analysis includes presentation of a family of cost effectiveness acceptability curves and the impact of different risks of common bile duct stones associated with ultrasound and liver function test results. The main outcome measure is cost per quality adjusted life year (QALY). Results: Baseline results, at 37% probability of common bile duct stones, show that MRCP is the dominant strategy, with expected savings of 149 ( 325 to ÿ 15) and expected QALY gain of (0e0.030) per case. The probability of avoiding unnecessary therapeutic ERCP is 30%. For patients at high risk of common bile duct stones (probability >60%) ERCP is the preferable strategy. * Corresponding author. Tel.: þ44 (0) address: yb3@york.ac.uk (Y. Bravo) /$ - see front matter ª 2006 Surgical Associates Ltd. Published by Elsevier Ltd. All rights reserved. doi: /j.ijsu

2 Economic evaluation of MRCP compared to diagnostic ERCP 13 Conclusions: ThebaselineestimateisthatMRCPwouldbebothcostsavingand would result in improved quality of life outcomes compared to diagnostic ERCP, but its potential sources of economic benefit are highly dependent on access to, and waiting lists for adequate MRI technology at hospital level. ª 2006 Surgical Associates Ltd. Published by Elsevier Ltd. All rights reserved. Introduction Use of magnetic resonance cholangiopancreatography (MRCP) for confirmation of presence of biliary obstruction is virtually risk-free and its fixed cost is only about half that of diagnostic endoscopic retrograde cholangiopancreatography (ERCP). However, no therapeutic option can be offered simultaneously with MRCP, unlike ERCP; there is thus a trade off between increased diagnostic costs in patients who ultimately require invasive treatment and cost and health benefits for patients who can avoid an unnecessary invasive diagnostic procedure. In the absence of direct economic evidence on this trade off a model based assessment is required. Our model incorporates the most frequent conditions affecting the biliary tree, such as common bile duct stones (CBDS), benign biliary strictures and peripancreatic cancer. 1 The UK NHS R&D Programme commissioned this assessment of the clinical and cost effectiveness of MRCP compared to the conventional practice of diagnostic ERCP for an adult UK population. A monograph in the Health Technology Assessment series gives further details of methods. 2 Almost all patients with symptoms suspected to be of biliary origin in the UK will be referred for ultrasound (US) either by the general practitioner or specialist. 3 Patients for whom MRCP is contraindicated (i.e. exclusions for MRI, such as claustrophobia and cardiac pacemakers) or ERCP (i.e. previous gastric surgery) are excluded from the scope of the model. Cholangiocarcinoma in the intrahepatic bile duct and primary sclerosing cholangitis (PSC) were both excluded because they are uncommon conditions normally associated with liver treatment. The economic and clinical impact of MRCP or ERCP is highly dependent on the incidence of CBDS in the patient groups being considered. Previous research has estimated the incidence of CBDS within populations demonstrating different US and liver function test (LFT) results. 4 The economic impact of MRCP compared to ERCP within these patient groups together with higher risk groups is explored. Methods A probabilistic economic model was constructed in order to evaluate the relative cost-effectiveness of adopting MRCP scanning compared to diagnostic ERCP for the investigation of biliary obstruction in adults. The primary outcome measure for the economic evaluation was cost per quality adjusted life year (QALY). The decision problem is illustrated in the structure of the decision tree presented in Fig. 1. The structure of this decision analytic model includes the most frequent conditions affecting the biliary tree where MRCP can provide diagnostic information comparable to ERCP. 5e8 The decision tree structure and its underlying assumptions were developed in discussion with a consultant gastroenterologist, two consultant radiologists and a consultant biliary pancreatic surgeon. The model considers the costs from the perspective of the health care provider as it is the most relevant to the decision maker within the context of reorganising NHS resources. The time horizon for the analysis was 12 months, the main reason being that the relief of pain is experienced in the short term after the removal of the stone and that more than three-quarters of patients suffering from pancreatic head lesions die within a year of diagnosis. 9 In order to facilitate the modelling, the following main assumptions have been introduced. For MRCP all patients incur the cost of the MRCP test. Patients with a negative test result incur no additional cost and achieve no gain in utility, with false negative patients maintaining a utility associated with an untreated condition, for example CBDS. All patients with a positive MRCP incur the additional cost and utility decrement of a therapeutic ERCP, with endoscopic stent for malignant strictures and surgery for benign strictures. True positives achieve a gain in utility from appropriate treatment and false positives achieve no utility gain. For diagnostic ERCP the tree follows a similar structure with the associated additional utility decrements and mortality risks. See also Table 1 and the HTA monograph 2 for more details of the assumptions within the model.

3 14 Y. Bravo et al. CBD stone - ERCP MRCP T+ Malignant stricture Benign st.- Surgery Endoscopic stent Resection True-negative T- CBD stone False-negative Malignant stricture Patient with suspected biliary obstruction Benign stricture CBD stone - ERCP T+ Malignant stricture Endoscopic stent Resection Diagnostic ERCP Alive T- Benign st. -Surgery True-negative False-negative CBD stone False-negative Malignant stricture Benign stricture Dead Figure 1 Basic decision tree. Base case values and parameters The parameters used in the model and the sources from which these are derived are shown in Table 2. Estimates of sensitivity and specificity for MRCP and ERCP were based on the clinical effectiveness results reported in Kaltenthaler et al. 2 In line with Table 1 Further specific assumptions Repeated diagnostic ERCP after failed cannulation and repeated MRCP because of pitfalls in interpretation are modelled In most cases therapeutic ERCP is performed with sphincterotomy so the option ERCP without sphincterotomy is not modelled There is some risk associated with diagnostic and therapeutic ERCP, so a healthy patient who is tested or treated with ERCP has a lower outcome than a healthy patient who is not MRCP is generally considered risk-free, so the utility of a healthy patient who is tested with MRCP equals perfect health the assumption that diagnostic ERCP is considered as the gold-standard diagnostic test for biliary tree investigation, we assumed perfect sensitivity and specificity. Only accuracy of diagnostic imaging in patients with CBDS was used in order to allow the modelling: estimates for both CBDS and strictures were not possible to pool statistically. Due to statistically significant heterogeneity between the studies, the median values for the sensitivity and specificity of MRCP were used. Peripancreatic cancer causing bile duct obstruction most commonly occurs at the head of the pancreas, and the chances of resectability are around 10%. The probability that curative resection is appropriate is low with the majority of patients undergoing palliative treatment through endoscopic stent placement. 10 The most common cause of benign biliary stricture is iatrogenic bile trauma during cholecystectomy, 11 and major bile duct injuries usually need surgical reconstruction. 12 The model assumes that surgical reconstruction is the primary treatment for benign extrahepatic strictures.

4 Economic evaluation of MRCP compared to diagnostic ERCP 15 Table 2 Base case probabilities, costs and utility estimates Diagnostic ERCP Mean value SD Model distribution MRCP Mean value SD Source Probabilities True positive test P[Tþ r S] 1 e Fixed; for Systematic review MRCP Beta True negative test P[Tÿ r H] 1 e Fixed; for Systematic review MRCP Beta CDBS 0.37 e Beta HES 2000 Malignant stricture 0.17 e Beta HES 2000 Benign stricture 0.9 e Beta HES 2000 Death after diagnostic ERCP Beta e e Cotton et al Overall complications after Normal e e Cotton et al diagnostic ERCP Resection for malignant strictures 0.10 e Fixed Clinical judgement Endoscopic stent for malignant strictures 0.90 e Fixed Clinical judgement Costs MRI - Medical Gastroenterology a e e Normal NHS Reference Costs 2000/01 CT scan - Medical Normal NHS Reference Costs 2000/01 Gastroenterology a Diagnostic ERCP examination Normal e e NHS Reference Costs 2000/01 bile duct w/o cc Diagnostic ERCP examination Normal e e NHS Reference Costs 2000/01 bile duct with cc Therapeutic ERCP - extraction Normal e e NHS Reference Costs 2000/01 CBDS Surgery malignant neoplasm Normal NHS Reference Costs 2000/01 extrahepatic bile ducts (>69 y or with cc) Memotherm biliary stent (palliative) Bilioplasty balloon catheter Normal Normal Northern General Hospital, Bradford Royal Infirmary Utilities Health-related QoL general 1 e Fixed Convention population MRCP examination e e Fixed 1 e Clinical judgement Chest and back pain relating to CBDS and strictures in extrahepatic bile ducts b Normal Cotton et al Diagnostic ERCP procedure e Beta e e Gregor et al only (w/o cc) c ERCP with cc c Normal Gregor et al Therapeutic ERCP w/o cc c,d 0.95 e Beta Gregor et al. 1996, 21 Bass et al Billiary stricture surgery c Normal Bass et al Extrahepatic malignant stricture Normal Luman et al Post-intervention for extrahepatic malignant stricture Normal Luman et al a Elective procedures, relating to hepato-biliary and pancreatic system HRGs section. b Approximated same discomfort relating to gallblader, according to clinical judgement. c Calculated as utility decrements from perfect health state. d Given that therapeutic ERCP costs are common to both intervention and comparator, complication costs are already included in this estimate.

5 16 Y. Bravo et al. Table 3 Key economic and clinical results Mean 95% CI Lower Upper Key clinical results True negative MRCP (%) (i.e. avoiding unnecessary diagnostic ERCP) True positive MRCP (%) (i.e. necessary therapeutic ERCP) Death under diag. ERCP (%) Key economic results Incremental QALYs Incremental costs ÿ 149 ÿ INB ( 20K per QALY) Cost effectiveness Dominant Dominant MRCP vs. ERCP Probability cost e e saving Probability incremental QALYs positive Probability cost effectiveness over 20K per QALY e e The costs of healthcare resources were estimated using the 2000/2001 NHS Reference Costs, for Elective Inpatients. 13 Complication costs are also shown in Table 2. A computed tomography (CT) scan is always performed before surgery for suspected benign or malignant strictures, so the CT fixed cost is added in both branches. Utility scores for the different health states were obtained from the Harvard CUA database, 14 with the exception of utilities related to biliary tree malignant neoplasms, which required an additional literature search. Whilst, of necessity, utility scores were taken from the international literature, all the health state valuations incorporated in the model used the EQ-5D generic measure. Those utilities reported by clinicians were preferred to those reported by patients because of the peculiarities of MRCP in terms of patient satisfaction: in spite of being a non-invasive procedure MRCP scores lower than expected by clinicians because of noise and claustrophobic related reasons. 2 Since we only considered a 12 months time horizon discounting of costs and outcomes was not carried out. The base case results are reported for a population at 37% risk of CBDS, 17% risk of malignancies and 15% risk of benign strictures, according to Finished Clinical Episode (FCE) records. Probabilities of events are proportions so the beta distribution has been used to estimate the probability of having a CBDS, a malignant stricture or a benign stricture, adding up to 69% risk of having an obstruction. Multivariate sensitivity analysis, using Monte Carlo simulation, is presented for the base case population, together with a range of populations at different risks of CBDS. Results The key clinical and economic results for patients with a baseline 31% likelihood of normal ducts, or 200 Incremental Costs (MRCP- ERCP) Figure 2 Incremental QALY (MRCP - ERCP) Cost Effectiveness Plane for MRCP vs Diagnostic ERCP.

6 Economic evaluation of MRCP compared to diagnostic ERCP 17 minor conditions that might require only expectant management, are shown in Table 3. Costeffectiveness results have been displayed on the cost-effectiveness plane in Fig. 2, this indicates the high probability that MRCP dominates ERCP, that is, MRCP results in an improved quality of life and saves money. The probability of avoiding unnecessary diagnostic ERCP, that is the probability of a true negative MRCP, is estimated at 30%. These patients could avoid the unnecessary risk of complications and death associated with diagnostic ERCP, with an overall expected QALY gain estimated at (95% CI 0.0e0.03). The probability that MRCP is cost saving is estimated at approximately 97%, with an overall expected saving of 149 (95% CI 325 to ÿ 15). From this result, we can infer that the cost savings in terms of avoided complications for healthy patients (i.e. true negatives) can compensate for the added diagnostic costs in those patients who need therapeutic intervention (i.e. true positives). The probability that imaging with MRCP has a cost effectiveness better than 20,000 per QALY is estimated at 99%, with an incremental net benefit (INB) of 364 over diagnostic ERCP. Sensitivity analyses The economic impact of MRCP within different risk groups is explored in Table 4 and the cost effectiveness acceptability curves (CEACs) are presented in Fig. 3. The probability of MRCP being cost-effective decreases as the probability of CBDS increases. At a risk of CBDS of 70%, the probability of MRCP being cost-effective compared to diagnostic ERCP at a threshold of 20,000 per QALY is below 50%, in fact diagnostic ERCP is the dominant strategy with an expected QALY gain of (95% CI ÿ0.005 to 0.016) and estimated cost savings of 81 (95% CI ÿ 96 to 222). At a risk of CBD stones around 60%, MRCP is estimated to be cost neutral when compared with ERCP. Discussion It is remarkable that although ERCP has been assumed to be a true gold standard, with perfect specificity and sensitivity, MRCP is still very likely to be considered economically attractive. The baseline estimate is that MRCP would be both cost saving and result in improved quality of life compared to diagnostic ERCP. Table 4 Key economic and clinical results at different risks of CBDS 9% (95% CI) 21% (95% CI) 32% (95% CI) 37% (95% CI) 50% (95% CI) 70% (95% CI) 56 (48e62) 44 (35e53) 34 (24e43) 30 (20e40) 17 (7e27) 0 (ÿ12e9) Key clinical results True negative MRCP (%) 8(5e13) 20 (13e27) 30 (22e37) 34 (26e42) 46 (38e55) 65 (55e74) True positive MRCP (%) Key economic results Incremental QALYs (0.005e0.039) (0.004e0.036) (0.002e0.032) (0.000e0.030) (ÿ0.001e0.022) (ÿ0.005e0.016) Incremental Costs ÿ 329 (ÿ 535eÿ 121) ÿ 249 (ÿ 422eÿ 75) ÿ 177 (ÿ 337eÿ 14) ÿ 149 (ÿ 325e 15) ÿ 60 (ÿ 233e 94) 81 (ÿ 96e 222) INB ( 20K per QALY) 645 ( 321e 1154) 521 ( 234e 978) 403 ( 156e 828) 364 ( 95e 786) 214 ( 2e 547) ÿ 1 (ÿ 170e 267) ICER Dominant Dominant Dominant Dominant Dominant 20, Probability cost saving According to Everett et al. 4 the risks associated with US and liver function test (LFT) results are: US and LFT normal (9% risk of CBDS); US normal and LFT abnormal (21%); US abnormal and LFT normal (32%); and US and LFT abnormal (37%). The risk of benign and malignant strictures is held constant. CBDS, common biliary duct stone; QALY, quality adjusted life year; INB, incremental net benefit; ICER, incremental cost effectiveness ratio.

7 18 Y. Bravo et al. Prob. E[INB] > ,000 10,000 15,000 20,000 25,000 Threshold (MAICER) Baseline risk (37%) Risk CBDS 50% Risk CBDS 60% Risk CBDS 70% Figure 3 Family of cost effectiveness acceptability curves for MRCP, with four levels of risk of CBDS. Although the perspective of the economic evaluation was the NHS as recommended by the National Institute for Clinical Excellence, 15 a societal perspective would have included an account of patients time and personal costs and loss of production due to treatment and poor health. Due to the fact that there is likely to be little difference in the relative impact of societal valuation between the interventions and health outcomes, with potentially a small bias in favour of the less invasive intervention, MRCP, the results are likely to be robust to a change in perspective. One potentially negative consequence of using MRCP as the initial diagnostic technique in evaluating suspected biliary obstruction is that it may delay therapeutic intervention in cases where this is required. No quantitative evidence was available concerning the potential magnitude of this risk although to a large extent it will be dependent on logistic factors determining the delay between diagnostic and therapeutic procedures. A potential benefit of MRCP not captured in the analysis is the use of the magnetic resonance images for presurgical staging of perampullary cancers and additional information about the state of the pancreaticobiliary duct. 16,17 Quantifying the impact on improved final outcomes is not possible from current data sources. The economic impact of considering repeated diagnostic ERCP after failed cannulation or repeated MRCP, because of pitfalls in interpretation, was examined in an alternative model and these modifications did not affect the main economic results. Only one other economic analysis evaluating MRCP compared to diagnostic ERCP has been identified and this was limited to the management of acute biliary pancreatitis. 18 These results showed that the economics of MRCP in patients with acute biliary pancreatitis were highly dependent on the probability of CBDS. At probabilities of CBDS higher than 58%, ERCP was reported to be the least expensive strategy. Our analysis is consistent with these results. Many of the limitations in the technology assessment relate to the poor design and reporting of comparative studies available, in particular on diagnostic test accuracy compared to final diagnosis and impact on clinical practice. The fact that ERCP is not a perfect gold standard reference test means that the diagnostic accuracy of MRCP may be under or over estimated. The clinical impact of these issues is discussed by Kaltenthaler et al. 2 but this assumption also means that the economics of MRCP may well also be under or over estimated depending on the true accuracy of ERCP. Problems with comparative studies of alternative diagnostic strategies for pancreaticobiliary conditions have also been recently noted elsewhere. 19 Ideally, the decision analytic model would use a ROC curve analysis of sensitivity and specificity to capture interdependence and the impact of variation in implied diagnostic thresholds and would be based upon comparative studies of the interventions measuring their impact on final outcomes. Whilst MRCP might have economic benefits compared with diagnostic ERCP, one potentially constraining factor is the availability of

8 Economic evaluation of MRCP compared to diagnostic ERCP 19 adequate MRI technology at hospital level and length of waiting time. However, ERCP is also subject to variation in operator ability and skill and depends on the availability of skilled radiographers and radiologists to carry out and interpret the imaging. In order to understand the real opportunity costs associated with performing MRCP on a routine basis, studies are needed to assess the relative benefits from access to MRI services, so priorities can be addressed rationally. Acknowledgements Funding for this work was provided by the National Coordinating Centre for Health Technology Assessment. Dr W.E.G. Thomas (Consultant Biliary Pancreatic Surgeon and Clinical Director of Surgery for Sheffield Teaching Hospitals) and Dr Conrad Beckett (Consultant Gastroenterologist, Bradford Royal Infirmary) provided clinical advice. References 1. Brennan ME. Periampullary and pancreatic cancer. In: Blumgart LH, Fong Y, editors. Surgery of the liver and biliary Tract. London: W.B. Saunders; Kaltenthaler E, Bravo Vergel Y, Chilcott J, Thomas S, Blakeborough T, Walters S, et al. A systematic review and economic evaluation of magnetic resonance cholangiopancreatography compared with diagnostic endoscopic retrograde cholangiopancreatography. Health Technology Assessment 2004;8(10). 3. Blackbourne LH, Earnhardt RC, Sistrom CL, Abbitt P, Jones RS. The sensitivity and role of ultrasound in the evaluation of biliary obstruction. The American Surgeon 1994;60(9):683e90 [Review]. 4. Everett SM, Hamlin J, Beckett C, Bzeizi K, Kay C. MRCP e An important investigation in patient at low risk of pancreaticobiliary disease. Endoscopy 2002;34(Suppl. II):A Demartines N, Eisner L, Schnabel K, Fried R, Zuber M, Harder F. Evaluation of magnetic resonance cholangiography in the management of bile duct stones. Archives of Surgery 2000;135(2):148e Lomas DJ, Gimson A. Magnetic resonance cholangiopancreatography. Hospital Medicine (London) 2000;61(6): 395e9 [Review]. 7. Farrell RJ, Noonan N, Mahmud N, Morrin MM, Kelleher D, Keeling PW. Potential impact of magnetic resonance cholangiopancreatography on endoscopic retrograde cholangiopancreatography workload and complication rate in patients referred because of abdominal pain. Endoscopy 2001;33(8):668e Alcaraz MJ, De la Morena EJ, Polo A, Ramos A, De la Cal MA, Gonzalez MA. Comparative study of magnetic resonance cholangiography and direct cholangiography. Revista Espanola de Enfermedades Digestivas 2000; 92(7):427e NHS Executive. Guidance on commissioning cancer services. Improving outcomes in upper gastro-intestinal Cancers. The Manual. London; Luman W, Cull A, Palmer KR. Quality of life in patients stented for malignant biliary obstructions. European Journal of Gastroenterology and Hepatology 1997;9(5): 481e John TG, Greig JD, Carter DC, Garden OJ. Carcinoma of the pancreatic head and periampullary region. Tumor staging with laparoscopy and laparoscopic ultrasonography. Annals of Surgery 1995;221:156e Allema JH, Reinders ME, van Gulik TM, van Leeuwen DJ, Verbeek PC, de Wit LT, et al. Results of pancreaticoduodenectomy for ampullary carcinoma and analysis of prognosis factors for survival. Surgery 1995;17:247e Department of Health. National schedule of reference costs. London; Harvard CUA database < pdf/preferencescores.pdf> [accessed 15/03/03]. 15. National Institute of Clinical Excellence. Information for national collaborating centres and guidelines for developers. Available at: London; Lopez H, Amthauer H, Hosten N, Ricke J, Bohmig M, Langrehr J. Prospective evaluation of pancreatic tumors: accuracy of MR imaging with MR cholangiopancreatography and MR angiography. Radiology 2002;224:31e Freeny PC. Computed tomography in the diagnosis and staging of cholangiocarcinoma and pancreatic carcinoma. Annals of Oncology 1999;10(Suppl. 4):12e Arguedas MR, Dupont AW, Wilcox CM. Where do ERCP, endoscopic ultrasound, magnetic resonance cholangiopancreatography, and intraoperative cholangiography fit in the management of acute biliary pancreatitis? A decision analysis model. American Journal of Gastroenterology 2001; 96(10):2892e Flamm CR, Mark D, Aronson N. Evidence-based review of ERCP: introduction and description of systematic review methods. Gastrointestinal Endoscopy 2002;56(Suppl. 6): S161e Cotton PB, Jowell PS, Baillie J. Spectrum of complications after diagnostic ERCP and effects of comorbidities. Gastrointestinal Endoscopy 1994; Gregor J, Ponich T, Detsky A. Should ERCP be routine after an episode of idiopathic pancreatitis? A cost-utility analysis. Gastrointestinal Endoscopy 1996;44: 118e Bass E, Pitt HA, Lillemoe K. Cost-effectiveness of laparoscopic cholecystectomy versus open cholecystectomy. American Journal of Surgery 1993;165(4):466e71.