Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

DOI: 10.1148/radiol.2211001656
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by McMahon, P. M.
Right arrow Articles by Gazelle, G. S.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by McMahon, P. M.
Right arrow Articles by Gazelle, G. S.
(Radiology. 2001;221:93-106.)
© RSNA, 2001

Health Policy and Practice

Pancreatic Cancer: Cost-Effectiveness of Imaging Technologies for Assessing Resectability1

Pamela M. McMahon, BS, Elkan F. Halpern, PhD, Carlos Fernandez–del Castillo, MD, Jeffrey W. Clark, MD and G. Scott Gazelle, MD, PhD

1 From the Decision Analysis and Technology Assessment Group, Department of Radiology (P.M.M., E.F.H., G.S.G.), Department of Surgery (C.F.d.C.), and Department of Oncology (J.C.), Massachusetts General Hospital, Zero Emerson Pl, Suite 2H, Boston, MA 02114; PhD Program in Health Policy, Harvard University, Cambridge, Mass (P.M.M.); and Department of Health Policy and Management, Harvard School of Public Health, Boston, Mass (G.S.G.). Received October 16, 2000; revision requested December 15; revision received February 8, 2001; accepted April 16. Address correspondence to G.S.G. (e-mail: scott@the-data-group.org).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 APPENDIX: MODELING DETAILS
 REFERENCES
 
PURPOSE: To evaluate the cost-effectiveness of imaging strategies for the assessment of resectability in patients with pancreatic cancer.

MATERIALS AND METHODS: A decision model was developed to calculate costs and benefits (survival) accruing to hypothetical cohorts of patients with known or suspected pancreatic cancer. Results are presented as cost per life-year gained under various scenarios and assumptions of diagnostic test characteristics, surgical mortality, disease characteristics, and costs.

RESULTS: With best estimates for all data inputs, the strategy of computed tomography (CT) followed by laparoscopy and laparoscopic ultrasonography (US) had an incremental cost-effectiveness ratio of $87,502 per life-year gained, compared with best supportive care. This strategy was significantly more cost-effective than CT followed by magnetic resonance (MR) imaging and was significantly less expensive than other imaging strategies while providing a statistically and clinically insignificant difference in life-year gains. A strategy involving no imaging (immediate surgery) was more expensive but less effective than all imaging strategies. A hypothetical perfect test with cost equal to that of CT followed by MR had an incremental cost-effectiveness ratio of $64,401 per life-year gained, compared to best supportive care.

CONCLUSION: Most available imaging tests for assessing resectability of pancreatic cancer do not differ in effectiveness, but a strategy of CT, laparoscopy, and laparoscopic US would consistently result in significantly lower costs than other imaging tests under a wide range of scenarios.

Index terms: Cost-effectiveness • Pancreas, neoplasms, 770.321 • Radiology and radiologists, socioeconomic issues


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
 APPENDIX: MODELING DETAILS
 REFERENCES
 
There will be an estimated 28,300 new cases of pancreatic cancer (PC) in the United States in 2000 (1). This is perhaps a small number compared with those of prostate or breast cancer; however, the dismal 5-year survival of only 4.1% (2) makes PC the fourth leading cause of cancer death in both men and women (2). The annual cost in the United States of care for PC has been estimated at $2.9 billion (in 1999 dollars) (3). Smoking and advancing age are the only firmly established exogenous risk factors (1,4). Genetic research has led to a progression model for the development of carcinoma of the pancreas (5), but genetic tests for the early detection of PC are not yet available (6).

At presentation, most patients report symptoms of abdominal pain, weight loss, or jaundice (7) due to tumor blockage of the common bile duct. Only about 10%–20% of patients are diagnosed with early-stage tumors localized to the pancreas (8,9); these patients are eligible for surgical resection. Known generally as a Whipple procedure (10), the surgical treatment for localized pancreatic head cancer involves removal of the pancreas and duodenum (11,12). This invasive and major surgical procedure has historically been associated with mortality rates over 25% (8), but the mortality rates at large regional centers have dropped to less than 4% in recent years (8,13–16), due at least partially to experience of the surgical center (16–19).

In roughly 40% of patients with PC, the tumor is local but advanced and unresectable at presentation, owing to involvement of the adjacent vasculature (9). For these patients, the therapies typically offered are chemoradiation (20) and transhepatic or endoscopic stent placement in the bile duct (21,22), if necessary. Surgical bypass of the bile duct can be performed in cases where the tumor is diagnosed as resectable, but the patient is found to have locally advanced disease at laparotomy. Results of a recent study demonstrating chemoradiation-induced remission in five cases of locally unresectable PC appear promising but preliminary (23).

In 40%–55% of patients with PC, the cancer has metastasized at the time of presentation; the liver and peritoneum are the most common sites of distant metastasis (24). These patients are typically offered chemotherapy (20), but effects on survival are limited (25).

To offer the potentially curative surgical option to all patients who would benefit and to spare those already at advanced stages the morbidity caused by unnecessary surgery, it is critical to have accurate imaging tests. The relative strengths and limitations of available technologies depend on the aspect of PC being imaged (eg, intrahepatic metastases, vascular invasion), leading to recommendations for multimodality staging regimens (26–29).


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 APPENDIX: MODELING DETAILS
 REFERENCES
 
General
The cost-effectiveness analysis that uses the best point estimates of all parameters and probability estimates is referred to as the base case and follows, to the extent possible with available data, the recommendations of the Panel on Cost-effectiveness in Health and Medicine (30). In sensitivity analyses, the model is re-analyzed by using varying estimates of critical parameters and probabilities for which imprecise or multiple estimates appear in literature sources. The present study was conducted from a societal perspective, incorporating all costs and benefits, regardless of who incurs them. This study reports benefits in dollars per life-year saved, with some adjustment for patient utility due to surgical morbidity. All costs and survival times are discounted at an annual rate of 3% (30).

The costs and benefits of all relevant diagnostic strategies were calculated by using a decision tree and were compared. Strategies were then ordered by increasing effectiveness. A strategy that cost more than an alternative but offered the same or fewer life-years was considered dominated and removed from the list of alternatives. Incremental cost-effectiveness ratios (ICERs) were calculated for the remaining strategies. In an ICER, changes in resource use, as compared with resource use for the next best strategy, are included in the numerator, and additional health benefits, again compared with those of the next best strategy, are included in the denominator. Strategies that had lower effectiveness but a higher ICER than another alternative were referred to as weakly dominated and were removed from the list of alternatives; incremental ratios were then recalculated. The resulting list of alternatives was ranked in order of increasing ICERs.

Model Structure
We constructed a decision-analytic model (P.M.M., G.S.G.) to compare alternative strategies for assessing resectability of patients with known or suspected PC. The model was developed with DATA 3.5.6 (TreeAge Software, Williamstown, Mass). Costs and benefits (survival) were calculated for hypothetical patients undergoing strategies ranging from palliative therapy alone to multimodality diagnostic imaging. The base case was assumed to occur at a high-volume specialized center where the 30-day mortality rate after resection is less than 4% (18), and all imaging modalities modeled were available. Monte Carlo simulations were performed for the base case and for each scenario described in the sensitivity analysis section (discussed later); average costs and benefits for 96,000 trials of each strategy are reported. The Appendix lists all model inputs, data sources, and modeling details.

Prevalence and Disease Characteristics
The initial cohort entering the model was composed of patients who presented with typical PC symptoms, including weight loss, pain, and jaundice (24,31). Prior routine abdominal computed tomographic (CT) or ultrasonographic (US) results indicated a high probability of PC. We designed our analysis to compare available tests for assessing resectability and confirming the (pathologic) diagnosis. For each patient, the model tracks both the true underlying disease state and the diagnosis based on imaging and biopsy results. Our base case estimate was that 90% of patients presenting actually had cancer, while the remaining 10% of patients had another disease that mimics PC (32). We assumed the following diseases constituted the conditions in the 10% of patients in the other disease category: pancreatitis (4%), ampullary cancer (4%), endocrine pancreatic tumors (1%–2%), and choledocholithiasis and lymphoma (<1% each).

Of the patients with PC, we estimated that 15% had localized disease, 40% had localized but unresectable disease, and the remaining 45% had distant metastases (9,21). Of the patients with metastases, we assumed that 80% had local unresectable disease, and the remaining 20% had a technically resectable local tumor (33). Our base case estimates for the types of metastases present were 80% liver metastases and 40% peritoneal dissemination, with 20% having both types (33). For the purposes of modeling, we defined peritoneal metastases to include metastases on the peritoneal surfaces of the liver; liver metastases were defined as intrahepatic (hematogenous) metastases only. Local nodal metastases (in the absence of liver or peritoneal metastases) are not considered a contraindication for resection (9) and thus were not explicitly modeled.

These characteristics defined the four categories used to classify patients for treatment: other disease, localized PC, localized cancer but unresectable due to vessel involvement, and cancer with distant metastases to the liver or peritoneal space.

Strategies Modeled
Table 1 lists the strategies compared in the model. The reference strategy is the minimal strategy (essentially the "do nothing strategy") to which all other strategies are compared. The reference strategy encompasses best supportive care, or palliation only, and addresses the principal symptoms of PC patients: pain, jaundice, and symptoms consistent with duodenal obstruction (24). Best supportive care is modeled as a weighted cost (see Costs section and Appendix for details) assigned to all patients and is assumed to provide essentially no survival gains; survival depends on disease state only.


View this table:
[in this window]
[in a new window]

  		TABLE 1. Comparison of Strategies

 
The strategy of immediate laparotomy (no imaging) is based on arguments that resection is the only chance for a cure of PC, so it ought to be offered to all patients (for critical review, see reference 34). For the base case, we assumed no patients underwent preoperative biliary stent placement prior to the scheduling of pancreaticoduodenectomy, since there is evidence that this practice increases complication and morbidity rates (35– 37). A schematic of the immediate laparotomy strategy is shown in Figure 1.



View larger version (41K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Schematic shows surgery subtree from decision model. All patients undergoing surgery for planned resection enter at the left and proceed through each node as described in the text.

 
Also modeled were strategies that involve preoperative diagnostic imaging, which can help select appropriate patients for surgery and possibly help plan the surgical approach (ie, by demonstrating vascular anomalies). Spiral or helical CT is the most commonly performed imaging test for PC; we assumed that it always included use of oral and intravenous contrast material and three-dimensional reconstruction of imaging data to produce CT angiograms (38,39). Similarly, we assumed magnetic resonance (MR) imaging always included MR angiography, as well as MR cholangiopancreatography (40,41). In addition to CT alone and MR alone, we modeled five two-test strategies, where patients whose cancer was diagnosed as resectable or equivocal on the basis of the first test result proceed to the second test: CT followed by laparoscopy, MR imaging followed by laparoscopy, CT followed by endoscopic US, MR followed by endoscopic US, and CT followed by MR (42). Finally, we modeled two three-test strategies: CT or MR imaging followed by laparoscopy followed by laparoscopic US (38). The base case used a standard laparoscopy technique (43).

All patients diagnosed as having distant metastases or local but unresectable disease on the basis of an imaging test result are assumed to undergo CT-guided pancreatic needle biopsy to confirm the diagnosis prior to beginning chemotherapy or radiation. Our base case estimate of the sensitivity of biopsy is 83% for cancer (44). We assumed that the biopsy would be repeated in the event of a negative test result. Patients with two consecutive negative biopsy results (independence of tests assumed) proceed to laparotomy. Patients diagnosed as candidates for resection do not undergo needle biopsy (21) and proceed to surgery or to the next imaging test, depending on the strategy modeled.

Angiography was not included since it has been supplanted in practice by CT or MR techniques (28). Other tests excluded were 2-[fluorine-18]fluoro-2-deoxy-D-glucose positron emission tomography, since limited sensitivity and specificity data are available (45), and US, based on its relatively poor performance (38). Endoscopic retrograde cholangiopancreatography was assumed to be performed for therapy only, since it provides little information about the extent of cancer (41).

Diagnostic Test Characteristics
Diagnostic tests were modeled to perform according to their sensitivity and specificity; these characteristics were defined separately for each of the following disease states: liver metastases, peritoneal metastases, and locally extended cancer (primarily vessel involvement). We assumed independence of tests for all strategies. Note that we did not use published estimates of sensitivity or specificity for overall prediction of resectability.

The sensitivity of helical CT was estimated at 0.73 for liver metastases (46), 0.54 for peritoneal metastases (47), and 0.80 for local extension (48). The specificity was estimated at 0.84 for liver metastases (46), 0.91 for peritoneal metastases (47), and 0.84 for local extension (46).

The sensitivity of fast MR imaging was estimated at 0.82 for liver metastases (49), 0.84 for peritoneal metastases (47), and 0.62 for local extension (48). The specificity was estimated at 0.92 for liver metastases (50), 0.87 for peritoneal metastases (47), and 0.67 for local extension (48).

Endoscopic US, either radial or linear in type (51), was modeled as performed only after a diagnosis of localized PC based on CT or MR findings. We estimated the sensitivity of endoscopic US for local unresectable cancer at 0.89 (52), with a specificity of 1.00 (51).

The base case laparoscopy modeled was a simple laparoscopic procedure (43) with no capacity to detect vessel involvement or intrahepatic metastases. The procedure includes a biopsy and has an estimated sensitivity for peritoneal metastases of 0.97, with a specificity of 1.00 (26).

Laparoscopic US has an estimated sensitivity of 0.89 for local extension (53) and 1.00 for liver metastases (53,54). The specificity was estimated at 1.00 for both liver metastases and local extension (53,54).

An assumption was made in the construction of the model that the true disease state would be determined at laparotomy (ie, sensitivity and specificity of 1.00 each for both local disease and distant metastases). It is true that in a small percentage of cancers, liver metastases will be overlooked, but these cases will be rare and common to all imaging strategies and so will not affect ICERs.

Survival and Effectiveness
Patients were triaged to therapy on the basis of test results. Survival time depended on the underlying true disease state and the treatment provided. Survival times were modeled as exponential survival distributions based on estimates from the literature. See the Appendix for details.

Patients diagnosed with localized PC proceeded to further staging tests or to laparotomy, depending on the strategy. (See Table 1.) For the base case analysis, patients awaiting attempted resection did not undergo endoscopic stent placement, as stated previously. The 30-day mortality rate after surgical bypass was estimated at 1.2%, or half of the mortality rate of 2.4% after a Whipple procedure (11). In this patient group with advanced cancer, the 30-day mortality rate after laparotomy was estimated at 0.6%, or half that after surgical bypass. We assumed that for 75% of patients who underwent successful resection, adjuvant chemoradiation therapy was administered (12). Median survival after successful resection and adjuvant therapy was estimated at 20 months, versus 16 months for patients who did not undergo adjuvant therapy (12). Median survival for misdiagnosed (unresected) patients with localized PC was assumed to be 11 months, intermediate between that of patients with localized PC with resection but no adjuvant therapy and that of patients with localized unresectable PC.

Median survival for patients undergoing laparotomy (with or without surgical biliary bypass) was modeled as equal to the survival of patients undergoing palliative therapies (described later), since the underlying disease state is the same.

Patients diagnosed with localized unresectable PC were treated with chemoradiation (5-fluorouracil based; see Appendix for details); 85% of patients underwent endoscopic placement of a biliary stent. This was based on the observation that 80% of patients receive endoscopically placed stents (22), and we estimated that patients in the category of localized unresectable PC would be more likely to experience biliary obstruction than those in the distant metastases category. Median survival for all patients (regardless of treatment offered) with a true disease state of localized unresectable PC was estimated at 10 months (12).

Patients diagnosed as having distant metastases (with cancer confirmed by biopsy) were treated with chemotherapy (gemcitabine regimen; see Costs section for details); median survival in patients with a correct diagnosis of distant metastases was estimated at 5.7 months (55). Mean survival for all patients with a true disease state of distant metastases but offered other therapies was estimated at 4 months (21). We estimated that 75% of patients with a diagnosis of distant metastases undergo endoscopic biliary stent placement.

We modeled no surgical biliary bypasses in patients who did not undergo laparotomy, since survival times after both surgical and percutaneous procedures are equivalent (56,57) and percutaneous procedures are less invasive and less expensive.

The majority of cases of other disease are detected post hoc (at surgery) and not on the basis of the results of any of the imaging tests modeled (58). Pancreatitis severe enough to mimic PC, periampullary cancers, endocrine pancreatic tumors, and lymphomas are all effectively treated with resection (59–61). There is some mortality associated with ampullary and endocrine cancers and lymphomas, but because of the relatively small numbers of these patients the effect on the overall analysis is small. For the base case, patients with other disease were assumed to have life expectancy equal to that of an age-matched general population, which was derived from the 1997 U.S. life tables (62). An alternative assumption was examined in a sensitivity analysis. The age of patients at the time of presentation was represented by a normal distribution with mean of 65.44 years (SD, 11.2 years) and a median of 68 years, based on data from our institution. (For comparison, the median age at diagnosis calculated by the National Cancer Institute was 72 years [2].)

The effect of surgery on patients’ quality of life was approximated by deducting a toll (in units of survival time) from patients recovering from any surgical procedure. At the time of surgery, patients are feeling the healthiest, and the conscious choice is made to temporarily decrease their quality of life in return for the chance of extending lifespan. With the assumption of a quality of life at only 70% of that of normal for 6 weeks after surgery, all patients surviving surgery are subjected to a toll of 0.3(6/52) = 0.03462 of a year, equal to roughly 121/2 days. Recovery after a Whipple procedure has been documented as 4–6 weeks until normal functioning (12), and patients typically require a similar recovery after a laparotomy, although it is less invasive than a Whipple procedure.

We did not subtract tolls for patients undergoing chemotherapy or radiation therapy for two reasons: (a) nearly all patients at late stages of PC are in pain (63), due either to metastases or to local extension, and thus already have diminished quality of life, and (b) there are no data available on which to base a toll. Also, we modeled a gemcitabine chemotherapy regimen, which has been shown (55) to reduce disease-related symptoms (pain, functional status, and weight).

All patients with cancer were assumed to receive terminal care to address pain, duodenal obstruction, and jaundice. See the Costs section for details.

Costs
All costs were converted to 1998 dollars by using the Medical Care component of the U.S. Consumer Price Index (64). All costs and benefits (survival time) occurring in future years were discounted at 3% (30). See the Appendix for further details.

We assumed all patients had previously undergone a complete medical history and physical examination, including blood counts, biochemical screening profiles, chest radiography, and routine abdominal CT or US examinations. This protocol was common to all strategies, so the cost was not included (there would be no effect on incremental ratios).

The base case cost of the best supportive care strategy was estimated as a weighted average of the cost of each of the following component outpatient treatments: endoscopic biliary stent placement in 80% of patients (22) (repeated every 3 months until death), percutaneous chemical splanchnectomy with alcohol for pain management in 10% of patients, morphine for 100% of patients, and percutaneous repair of duodenal obstruction in 25% of patients (65).

For patients initially treated with other therapies (eg, chemotherapy, resection), the cost of terminal care was added to each treatment. This terminal care cost was identical to the cost of best supportive care, except that the cost of the initial stent placement was subtracted for patients who had previously undergone stent placement. Terminal care costs of localized PC patients after a successful Whipple procedure were discounted by the median survival of that group (1.68 years). Note that some percentage of patients who undergo resection will survive or be cured, so this value will be an overestimate.

Costs for procedures requiring an inpatient hospital stay (laparotomy with or without surgical bypass, pancreaticoduodenectomy) were estimated on the basis of resource use data from our institution, since Medicare diagnosis-related groups are broad and nonspecific. Actual total costs, excluding professional fees, are collected from an administrative database (Transition Systems, a subsidiary of Eclypsis, Delray Beach, Fla), which interfaces with the hospital cost accounting system. The cost of a pancreaticoduodenectomy (Whipple procedure) was based on results of a regression analysis (SAS; SAS Institute, Cary, NC) on cost and clinical data from 136 inpatient pancreaticoduodenectomies performed between 1995 and 1999. Complications were assigned according to the method Cameron et al (14) were and based on International Classification of Diseases (ICD-9) codes for secondary diagnoses (limited to 10 in the Transitions Systems database). For pancreaticoduodenectomies with no complications, the cost was estimated at $16,618 (standard error, $1,059). The cost increased (P = .0012) to $21,934 (standard error, $1,925) when complications occurred. (For comparison, Holbrook et al [66] estimated these costs as $15,926 and $24,635, respectively.) The additional cost of professional fees, $2,650, was derived from Medicare reimbursement rates and is assumed to be the same for all pancreaticoduodenectomies. The cost of a laparotomy is estimated as our institutional resource use cost of $9,027 (standard error of the mean, $1,436) plus a $741 professional fee. The cost of surgical bypass is assumed to be the same as that of laparotomy.

The costs of outpatient diagnostic procedures were derived from Medicare reimbursement rates (both technical and professional components). See the Appendix for details. The total cost of outpatient chemotherapy was based on a gemcitabine regimen (55) and corresponding Medicare reimbursements. The average duration of chemotherapy was fixed at 4 months, and patients were assumed to be equally divided between intravenous ondansetron hydrochloride with dexamethasone (67) and oral prochlorperazine, for a weighted cost for antiemetic therapies. Chemoradiation was modeled as a 5.5-week 5-fluorouracil regimen. See the Appendix for details.

The time cost for patients was estimated in the base case as equal to the income for U.S. individuals older than 65 years, or $50 per day (68). Patient times for outpatient procedures or therapies were estimated on the basis of expert opinion. Patient times for inpatient procedures were based on average length of stay at our institution or expert opinion. (Comparable estimates of length of stay appeared in a recent study [66].)

Model Validation
The authors of several large surgical series reported the percentage of patients with a diagnosis of localized PC (on the basis of CT, MR, or other test results) who actually have resectable disease at surgery. For example, on the basis of a CT or MR diagnosis, 54% of patients with a diagnosis of localized PC had resectable disease (69). Other authors have reported higher percentages when staging laparoscopy was used (26). For each strategy modeled, we compared the diagnosis and true disease state in 1,000 hypothetical patients and found reasonable agreement with published results. As shown in Table 2, percentages of patients confirmed as having localized PC at surgery increase as the testing strategies increase in complexity. This validation did not address surgical effectiveness.


View this table:
[in this window]
[in a new window]

  		TABLE 2. Model Validation

 
Sensitivity Analyses Performed
The Appendix lists all values used and the literature sources.

Test characteristics.—To examine the upper range of effectiveness possible through improvements in diagnostic accuracy, we modeled a hypothetical perfect test with sensitivity and specificity equal to 1.0. Patient time for diagnosis was assumed to be the same as that for CT, and the cost was estimated as equal to the most expensive combination of tests (CT and MR). The cost of biopsy was not included. Sensitivity analyses were performed on estimates of sensitivity and specificity of the diagnostic tests, shown in the Appendix.

Preoperative stent placement.—Based on results from a large retrospective study (35), we included a scenario in which 53% of patients scheduled for laparotomy or resection underwent preoperative stent placement. These patients had higher complication (1.5 times the base case) and mortality (2.7 times the base case) rates. Note that contradictory results were found in other studies (36,37).

Modified laparoscopy technique.—A modified laparoscopy technique described by Conlon et al (70) and reviewed by Schwarze and Rattner (71) allows for vascular assessment and thus has a sensitivity for localized unresectable PC greater than 0, but estimates were not available in the literature, to our knowledge. Also, the test has low sensitivity for intrahepatic metastases (by means of dual-instrument palpation through the ports). We modeled two scenarios to examine a range of possible test characteristics: (a) In the optimistic scenario, we estimated the sensitivities for both liver metastases and local extension at 0.50 and the specificities for both of these disease states at 1.0; (b) in the pessimistic scenario, the analogous sensitivities were estimated at 0.10 and the specificities at 0.85. In both scenarios, the patient time was increased to 2 days, and the cost of the laparoscopy procedure was increased to $1,500 to reflect the increased costs of the more invasive procedure, which typically requires an overnight inpatient stay.

Costs.—We modeled the annual discount rate at 0% and 5% (30). We also assumed no patient time costs. We modeled costs of therapy (including best supportive care and terminal care, as well as chemotherapy, chemoradiation, and all surgical interventions) at both 50% and 150% of the base case estimates. We assigned a lower cost (corresponding to US guidance) for biopsies. In addition, we included a scenario with a more extensive regimen for best supportive care, with a cost of $11,813 (72).

Effects.—To determine the potential influence on cost-effectiveness of imaging strategies in the event of improved therapies for PC, we postulated improved survival times for patients in whom a correct diagnosis is established and who are treated with either chemotherapy, chemoradiation, or resection. Increases in median survival gains were estimated at 110% and 200% of the base case scenario.

We also performed sensitivity analyses on the expected survival for the other disease category, the Whipple complication and mortality rates (which are higher at low-volume surgical centers), surgical mortality rates for laparotomy and bypass procedures, and the quality of life of surgical patients (see Appendix). We performed a sensitivity analysis on the age of the cohort, using a mean of 69.4 years (4 years older than the base case estimate, chosen because the literature median was 4 years older than our base case estimate).

Disease state.—We varied several estimates of disease characteristics of the presenting population. The proportions of liver versus peritoneal metastases within the distant metastases category was varied from "all liver" to "all peritoneal," as well as the reverse of the base case estimate: 80% peritoneal, 40% liver, with 20% of patients having both. The relative proportions of these types of metastases are difficult to ascertain from the literature, primarily due to inconsistencies in definitions of "liver metastases"; authors often include in this category metastases that occur on the surface of the liver (considered as peritoneal metastases in our analysis).

We modeled longer median survival times (13 and 16 months) for misdiagnosed (no resection) localized PC patients. The percentage of patients with distant metastases but locally resectable pancreatic tumors was increased to 50% and decreased to 0% (ie, all patients with distant metastases also had localized unresectable PC). We also increased the percentage of patients with localized PC at presentation (and no distant metastases) to 20%, at the high end of the range typically reported (9).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 APPENDIX: MODELING DETAILS
 REFERENCES
 
Base Case
Table 3 and Figure 2 show the results of the base case analysis. The imaging strategies cluster in one area of the cost-effectiveness graph (Fig 2a). The strategy of CT followed by MR was weakly dominated and excluded from further consideration. The differences in effectiveness between the remaining strategies, in most cases less than a week, do not reach statistical significance (the standard error of the mean for each strategy exceeds the marginal difference in effectiveness between strategies; see Fig 2b). An increase in the number of Monte Carlo trials to 192,000 did not reduce the standard error of the mean sufficiently (data not shown). Given the clinically unimportant differences in expected survival, we analyzed 96,000 trials per scenario. The strategy of CT, laparoscopy, and laparoscopic US was significantly (P = 0.012) less costly than the next most costly strategy, CT, endoscopic US. Because of the clustering in effectiveness and some spread in costs, we focused our analysis on costs. It follows from the clustering of effectiveness estimates that ICERs between strategies are not meaningful, so we report ICERs versus the do nothing, or best supportive care, strategy. The least expensive strategy in the base case analysis was the strategy using CT, laparoscopy, and laparoscopic US ("CT–laparoscopic US"), which had an ICER of $87,502 per life-year gained, compared with that of best supportive care alone. The strategy of immediate surgery (no imaging) was dominated by all the imaging strategies.


View this table:
[in this window]
[in a new window]

  		TABLE 3. Results of Base-Case Analysis

 


View larger version (23K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2a. (a) Graph shows base case results plus hypothetical perfect test results. The perfect test has an incremental cost-effectiveness of $64,401 per life-year (LY) saved, compared with that of best supportive care (BSC). EUS = endoscopic US, LUS = laparoscopy, laparoscopic US, scope = laparoscopy. (b) Graph of base case results with standard error of the mean shown for life-years (LY). The incremental cost of CT-endoscopic US (CT-EUS) compared with that of CT, laparoscopy, laparoscopic US (CT-LUS) is $52; the standard errors of the mean for the cost estimates are $14.77 and $14.57, respectively (Table 3). scope = laparoscopy.

 


View larger version (27K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2b. (a) Graph shows base case results plus hypothetical perfect test results. The perfect test has an incremental cost-effectiveness of $64,401 per life-year (LY) saved, compared with that of best supportive care (BSC). EUS = endoscopic US, LUS = laparoscopy, laparoscopic US, scope = laparoscopy. (b) Graph of base case results with standard error of the mean shown for life-years (LY). The incremental cost of CT-endoscopic US (CT-EUS) compared with that of CT, laparoscopy, laparoscopic US (CT-LUS) is $52; the standard errors of the mean for the cost estimates are $14.77 and $14.57, respectively (Table 3). scope = laparoscopy.

 
Other trends are apparent. The addition of laparoscopy to MR or CT resulted in no cost savings. However, the addition of endoscopic US resulted in lower costs, as compared with costs for CT or MR alone, since fewer unnecessary surgeries were performed (See Table 2). The addition of both laparoscopy and laparoscopic US reduced costs significantly (P < 0.001), as compared with costs for CT or MR alone or with laparoscopy.

Sensitivity Analyses
Figure 2b illustrates the result of the perfect test strategy, which dominates all available imaging strategies. Table 4 shows the results of the remaining sensitivity analyses performed. The strategy of CT followed by MR was excluded by weak dominance. Since the remaining strategies did not differ in effectiveness, Table 4 provides the ICER of the least expensive strategy, as compared with that of best supportive care.


View this table:
[in this window]
[in a new window]

  		TABLE 4. Results of Sensitivity Analyses

 
The results of most sensitivity analyses mirrored those of the base case; CT–laparoscopic US was typically the least expensive strategy, with an ICER compared with that of best supportive care of between $83,500 and $91,000 per life-year saved. Several scenarios resulted in large changes in the estimated ICER of the CT–laparoscopic US strategy. An increase in the cost of best supportive care reduced the ICER to $26,451 per life-year saved, since CT–laparoscopic US was compared with a more expensive reference strategy. An increase in the costs of all therapies by 150% over the base case value resulted in an ICER of $124,739 per life-year saved, while a reduction in those costs to 50% of the base case cost resulted in an ICER of $50,535 per life-year saved. A 10% increase over base case estimates of median survival for cancer patients provided with the correct therapy for their disease stage (due to hypothetical future advances in therapies) resulted in a decrease of the ICER of CT–laparoscopic US to $59,916 per life-year saved. A doubling of those survival times further decreased the ICER to $16,024 per life-year saved.

Five scenarios resulted in a least expensive strategy of MR imaging–laparoscopic US: one scenario in which the sensitivity of MR for liver metastases was increased to 1.00 (base case sensitivity, 0.82), two scenarios in which the proportion of intrahepatic metastases was lower than that of peritoneal metastases, and two scenarios with increased median survival of PC patients in whom a misdiagnosis of unresectable PC was rendered. The ICER for these MR imaging–laparoscopic US strategies varied between $85,693 and $117,584 per life-year saved, again compared with that of best supportive care.

MR imaging–endoscopic US was the least expensive strategy under the pessimistic Conlon laparoscopy scenario, with an ICER of $93,999 per life-year saved.

At a low-volume surgical center with a mortality rate of 16% after Whipple surgery, all strategies were less effective than best supportive care, while a doubling of the surgical mortality rate over the base case rate (to 4.8%) resulted in an ICER for CT–laparoscopic US of $108,010.


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
APPENDIX: MODELING DETAILS
 REFERENCES
 
Despite few statistically significant differences in effectiveness between strategies for assessing the resectability of PC, strategies that use laparoscopy and laparoscopic US were consistently and significantly less expensive than the other imaging strategies modeled, under a wide range of assumptions. Similarly, but with smaller cost savings, the addition of endoscopic US to either MR or CT decreased the overall cost of the strategy. However, the addition of laparoscopy alone did not appear to reduce the overall cost of either MR or CT in the base case. Whether the optimal (least expensive) strategy used CT or MR as the initial test depended on the scenario.

None of the imaging strategies compared (except for CT–MR imaging, which was significantly less effective than most others, in the base case and in most sensitivity analyses) were significantly different from each other in terms of effectiveness. They added only 0.15 year (1.8 months) to average life expectancy, compared with life expectancy of best supportive care. However, the hypothetical perfect test added only 0.21 year (roughly 21/2 months) to average life expectancy. Thus, unless perfect or near-perfect imaging tests are also very inexpensive, further improvements in test characteristics will not contribute substantially to the cost-effectiveness of imaging strategies in the assessment of resectability of PC.

Our results demonstrate that future improvements in therapies will increase the median survival of patients in whom diagnosis is successful and will greatly improve the cost-effectiveness of all the imaging strategies. Clearly, the results of this analysis were affected by treatment setting and the different surgical mortality rates at different centers.

The results of this analysis lend support to the practice of multimodality testing for assessment of resectability in PC patients. Multistep imaging strategies such as CT–laparoscopic US or CT–endoscopic US help reduce the number of unnecessary surgeries while offering a potentially curative surgery to more patients who might benefit (see Table 2). Thus, while the costs of these strategies, in terms of patient time costs and hospital costs, may seem high, they are offset by cost reductions from avoiding unnecessary surgeries. The estimated cost-effectiveness ratio of CT, laparoscopy, and laparoscopic US ($87,502 per life-year saved) is at the upper end of the range of commonly used interventions in the United States (73,74).

Limitations
This study was subject to the limitations common to all modeling exercises: the necessary oversimplification of the problem, heterogeneity in the patient population that might limit the generalizability of the results, and uncertainty surrounding parameter estimates. Additional limitations include the difficulty of modeling tests that improve over time; the "moving target" problem is partially addressed by the hypothetical perfect test. Also, it is difficult to model the subjective nature of resection criteria used by different surgeons. Assumptions used in this analysis may affect results in ways that are not easily quantifiable.

An important aspect of this analysis may be the quality of life during any added survival, but the necessary data on the quality of life in PC patients are not currently available. To adhere fully (ie, perform a reference case cost-effectiveness analysis) to the recommendations of the U.S. Panel on Cost-effectiveness in Health and Medicine (30), outcomes from cost-effectiveness analyses should be given in the metric of dollars per quality-adjusted life-year. In a recent review of quality of life studies in PC, Fitzsimmons and Johnson (75) noted that only nine of 78 studies analyzed used a previously validated quality of life questionnaire, and these were inconsistent in methodology. Published studies of the quality of life in cancer patients typically use health-state valuations or document the presence of clinical symptoms (eg, days until a return to normal eating). These measures are clinically important but are not preference based and thus cannot be the basis of quality-adjusted life-years (30). A collection of these preference-based weights using available utility instruments is beyond the scope of this study; future investigations to determine the societal utility weight for health states in PC are needed. In this study, we report outcomes in dollars per life-year saved, with some adjustment made for high morbidity effects of invasive surgery, but this is not directly comparable to studies that used the metric of dollars per quality-adjusted life-year.

Our surgical cost estimates were dependent on our institution’s overhead allocations.

An assumption was made that distant metastases are limited to peritoneum and liver, but as many as 5%–10% of distant metastases from pancreatic tumors may be to bone, lung, and brain (76). These are typically not detected until they are symptomatic; clinicians rarely look for lung metastases in practice. Our results were somewhat sensitive to the estimated median survival of patients with misdiagnosed localized cancer, an unobservable parameter.

The purpose of this study was not to address cost-effectiveness of chemotherapy or treatment protocols, but rather to assess that of imaging tests. Once an imaging strategy is determined to be cost-effective under a range of scenarios, the same model can be extended to examine the cost-effectiveness of different therapies.


   APPENDIX: MODELING DETAILS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 APPENDIX: MODELING DETAILS
REFERENCES
 
In DATA 3.5.6 (TreeAge Software), three cohorts of 32,000 Monte Carlo trials were run for each strategy under each scenario. Text reports were exported to SAS (SAS Institute) for statistical analysis (A1–A4).


View this table:
[in this window]
[in a new window]

  		TABLE A1. Costs (1998 U.S. Dollars) for Diagnostic Tests and Treatments

 

View this table:
[in this window]
[in a new window]

  		TABLE A2. Costs in Patient Time and Disease State Data

 

View this table:
[in this window]
[in a new window]

  		TABLE A3. Test Accuracy

 

View this table:
[in this window]
[in a new window]

  		TABLE A4. Therapies and Outcomes

 
Initially, a random draw from a uniform distribution (with equal probability of the selection of any number from 0 to 1, inclusive) is made to determine the true disease state (cancer or other disease) of each hypothetical patient. Of those with cancer, further draws were made to specify presence or absence of hepatic or peritoneal metastases or presence or absence of locally unresectable disease. Definition of true underlying disease state allowed survival to depend on true state and allowed comparison of diagnosis to true state for model validation.

In reality, the base case (simple) laparoscopy technique provides no information about local extension or liver metastases. For modeling purposes, the sensitivities of laparoscopy for local extension or liver metastases are estimated at 0, and specificities for local extension and liver metastases are estimated at 1.0 (since the test result is always negative).

Most estimates in the literature are median survival times. We converted to mean survival to allow modeling of survival as exponential distributions—reasonable approximations for severe disease—for each PC disease state and treatment category as follows: exponential, f(x) = {lambda}e-{lambda}x. At median survival, survival = 0.5, so 0.5 = 1-e(-median/mean) and {lambda} = 1/mean.

The cost of best supportive care was estimated as 10% of the cost of chemical splanchnectomy with alcohol (Current Procedural Terminology [CPT] code 64680; reimbursement, $173 [CODEMANAGER; American Medical Association, Chicago, Ill]), 100% of the cost of morphine sulphate tablets (60 days at 100 mg per day, $27 per 100 mg average wholesale price, or $1,620 [77]), 80% of the cost of percutaneous biliary stent placement (CPT 43268 + 74330; reimbursement, $766; repeated every 3 months until death; estimated as the median survival weighted by the proportion of patients in each disease state), plus 25% of the cost of percutaneous gastronomy tube placement (CPT 43750 + 74350; reimbursement, $527), plus the cost of two outpatient visits for care management (CPT 99214; reimbursement, $55 each). The weighted average of these therapies results in an estimate for best supportive care of $2,649 per patient.

The chemoradiation cost was estimated as the sum of $2,781 for 5.5 weeks of outpatient visits for radiation treatment (CPT code 77413), and $2,002, which includes the cost of installation of the venous access port, weekly refilling and maintenance of the pump, 2 months of pump rental, the average wholesale cost of 5-fluorouracil (77), and weekly outpatient physician visits (CPT codes 36533, 96520, E0781-RR, 99213).

The cost of laparoscopic US is estimated as equal to the cost of intraoperative US; there is no code for laparoscopic US.


   FOOTNOTES
 
Abbreviations: ICER = incremental cost-effectiveness ratio, PC = pancreatic cancer

Author contributions: Guarantors of integrity of entire study, P.M.M., G.S.G.; study concepts, P.M.M., G.S.G.; study design, P.M.M., G.S.G., E.F.H.; literature research, P.M.M.; data acquisition, P.M.M.; data analysis/interpretation, all authors; statistical analysis, P.M.M., E.F.H.; manuscript preparation, definition of intellectual content, editing, revision/review, and final version approval, all authors.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 APPENDIX: MODELING DETAILS
 REFERENCES
 

  1. American Cancer Society. Cancer Facts and Figures. 2000; Available at: www.cancer.org/statistics/cff2000/. Accessed August 2000..
  2. Ries LAG, Kosary CL, Hankey BF, Miller BA, Clegg LX, Edwards BK. SEER cancer statistics review, 1973–1996 Bethesda, Md: National Cancer Institute, 1999.
  3. Elixhauser A, Halpern MT. Economic evaluations of gastric and pancreatic cancer. Hepatogastroenterology 1999; 46:1206-1213.[Medline]
  4. Haddock G, Carter DC. Aetiology of pancreatic cancer. Br J Surg 1990; 77:1159-1166.[Medline]
  5. Hruban RH, Wilentz RE, Goggins M, Offerhaus GHA, Yeo CH, Kern SE. Pathology of incipient pancreatic cancer. Ann Oncol 1999; 10(suppl 4):9-11.[Free Full Text]
  6. Goggins M, Kern SE, Offerhaus JA, Hruban RH. Progress in cancer genetics: lessons from pancreatic cancer. Ann Oncol 1999; 10(suppl 4):4-8.[Free Full Text]
  7. Janes RH, Niederhuber JE, Chmiel JS, Winchester DP, Menck HR. National patterns of care for pancreatic cancer. Ann Surg 1996; 223:261-272.[CrossRef][Medline]
  8. Lillemoe KD, Cameron JL. Survival after surgical treatment of pancreatic cancer, 1172. In: Beger HG, Warshaw AL, Buchler MW, et al., eds. The pancreas. Malden, Mass: Blackwell Science, 1998; 125.
  9. Warshaw AL, Fernandez–del Castillo C. Pancreatic carcinoma. N Engl J Med 1992; 326:455-465.[Medline]
  10. Whipple A, Parsons W, Mullins C. Treatment of carcinoma of the ampulla of Vater. Ann Surg 1935; 102:763-779.[CrossRef][Medline]
  11. Di Carlo V, Zerbi A, Balzano G. Surgical treatment: Whipple-Kausch pancreatoduodenectomy with or without pylorus preservation. In: Beger HG, Warshaw AL, Buchler MW, et al., eds. The pancreas. Malden, Mass: Blackwell Science, 1998; 1028-1040.
  12. Pisters P, Lee J, Evans D. Standard forms of pancreatic resection. In: Reber H, eds. Pancreatic cancer: pathogenesis, diagnosis and treatment. Totowa, NJ: Humana, 1998; 181-199.
  13. Trede M, Schwall G, Saeger HD. Survival after pancreatoduodenectomy. Ann Surg 1990; 211:447-458.[Medline]
  14. Cameron JL, Pitt HA, Yeo CJ, Lillemoe KD, Kaufman HS, Coleman J. One hundred and forty-five consecutive pancreaticoduodenectomies without mortality. Ann Surg 1993; 217:430-438.[Medline]
  15. Fernandez–del Castillo F, Rattner D, Warshaw A. Standards for pancreatic resection in the 1990s. Arch Surg 1995; 130:295-300.[Abstract]
  16. Yeo CJ, Cameron JL, Sohn TA, et al. Six hundred fifty consecutive pancreaticoduodenectomies in the 1990s. Ann Surg 1997; 226:248-260.[CrossRef][Medline]
  17. Gordon TA, Burleyson GP, Tielsch JM, Cameron JL. The effects of regionalization on cost and outcome for one general high-risk surgical procedure. Ann Surg 1995; 221:43-49.[Medline]
  18. Birkmeyer JD, Finlayson SRG, Tosteson ANA, Sharp SM, Warshaw AL, Fisher ES. Effect of hospital volume on in-hospital mortality with pancreaticoduodenectomy. Surgery 1999; 1999:250-256.
  19. Birkmeyer HD, Warshaw AL, Finlayson SRG, Grove MR, Tosteson ANA. Relationship between hospital volume and late survival after pancreaticoduodenectomy. Surgery 1999; 126:178-183.[Medline]
  20. Tempero M. Chemotherapy of pancreatic cancer. In: Reber H, eds. Pancreatic cancer: pathogenesis, diagnosis and treatment. Totowa, NJ: Humana, 1998; 265.
  21. Fernandez–del Castillo C, Warshaw A. Pancreatic cancer: indications for resection. In: Beger HG, Warshaw AL, Buchler MW, et al., eds. The pancreas. Malden, Mass: Blackwell Science, 1998; 1021- 1027.
  22. Wong R, Carr-Locke D. Endoscopic stents for palliation in patients with pancreatic cancer. In: Reber H, eds. Pancreatic cancer: pathogenesis, diagnosis and treatment. Totowa, NJ: Humana, 1998; 235-252.
  23. DeNittis AS, Stambaugh MD, Lang P, et al. Complete remission of nonresectable pancreatic cancer after infusional colloidal phosphorus-32 brachytherapy, external beam radiation therapy, and 5-fluorouracil. Am J Clin Oncol 1999; 22:355-360.[CrossRef][Medline]
  24. Rosewicz S, Wiedenmann B. Pancreatic carcinoma. Lancet 1997; 349:485-489.[CrossRef][Medline]
  25. Bakkevold KE. Chemotherapy of unresectable pancreatic cancer, 1120. In: Beger HG, Warshaw AL, Buchler MW, et al., eds. The pancreas. Malden, Mass: Blackwell Science, 1998; 120.
  26. Jimenez RE, Warshaw AL, Rattner DW, Willett CG, McGrath D, Fernandez–del Castillo C. Impact of laparoscopic staging in the treatment of pancreatic cancer. Arch Surg 2000; 135:409-415.[Abstract/Free Full Text]
  27. Awad SS, Colletti L, Mulholland M, et al. Multimodality staging optimizes resectability in patients with pancreatic and ampullary cancer. Am Surg 1996; 63:634-638.
  28. Gloor B, Todd KE, Reber HA. Diagnostic workup of patients with suspected pancreatic carcinoma: the University of California-Los Angeles approach. Cancer 1997; 79:1780-1786.[CrossRef][Medline]
  29. Andren-Sandberg AA, Lindberg CG, Lundstedt C, Ihse I. Computed tomography and laparoscopy in the assessment of the patient with pancreatic cancer. J Am Coll Surg 1998; 186:35-40.[CrossRef][Medline]
  30. Gold MR, Siegel JE, Russell LB, Weinstein MC. Cost-effectiveness in health and medicine New York, NY: Oxford University Press, 1996.
  31. Barkin J, Goldstein J. Diagnostic approach to pancreatic cancer. Gastroenterol Clin North Am 1999; 28:709-721.[CrossRef][Medline]
  32. Nagahama H, Okada S, Okusaka T, et al. Clinicopathological features in misdiagnosed pancreatic carcinoma. Hepatogastroenterology 1999; 46:2983-2985.[Medline]
  33. Kayahara M, Nagakawa T, Ueno K, Ohta T, Takeda T, Miyazaki I. An evaluation of radical resection for pancreatic cancer based on the mode of recurrence as determined by autopsy and diagnostic imaging. Cancer 1993; 72:2118-2123.[CrossRef][Medline]
  34. Fernandez–del Castillo C, Warshaw AL. The role of laparoscopy and peritoneal cytology in the management of pancreatic cancer, 1004. In: Beger HG, Warshaw AL, Buchler MW, et al., eds. The pancreas. Malden, Mass: Blackwell Science, 1998; 107.
  35. Povoski SP, Karpeh MS, Jr, Conlon KC, Blumgart LH, Brennan MF. Association of preoperative biliary drainage with postoperative outcome following pancreaticoduodenectomy. Ann Surg 1999; 230:131-142.[CrossRef][Medline]
  36. Heslin M, Brooks A, Hochwald S, Harrison L, Blumgart L, Brennan M. A preoperative biliary stent is associated with increased complications after pancreatoduodenectomy. Arch Surg 1998; 133:149-154.[Abstract/Free Full Text]
  37. Pitt H, Gomes A, Lois J, Mann L, Deutsch L, Longmire W. Does preoperative percutaneous biliary drainage reduce operative risk or increase hospital cost?. Ann Surg 1985; 201:545-553.[Medline]
  38. Baker ME. Radiologic techniques for diagnosis and staging of pancreatic carcinoma Totowa, NJ: Humana, 1998.
  39. Novick S, Fishman E. Three-dimensional CT angiography of pancreatic carcinoma: role in staging extent of disease. AJR Am J Roentgenol 1998; 170:139-143.[Free Full Text]
  40. Vahldiek G, Broemel T, Klapdor R. MR-cholangiopancreaticography (MRCP) and MR-angiography: morphologic changes with magnetic resonance imaging. Anticancer Res 1999; 19:2451-2458.[Medline]
  41. Georgopoulos SK, Schwartz LH, Jarnagin WR, et al. Comparison of magnetic resonance and endoscopic retrograde cholangiopancreatography in malignant pancreaticobiliary obstruction. Arch Surg 1999; 134:1002-1007.[Abstract/Free Full Text]
  42. Sheridan MB, Ward H, Guthrie HA, et al. Dynamic contrast-enhanced MR imaging and dual-phase helical CT in the preoperative assessment of suspected pancreatic cancer: a comparative study with receiver operating characteristic analysis. AJR Am J Roentgenol 1999; 173:253-590.
  43. Warshaw A, Tepper J, Shipley W. Laparoscopy in the staging and planning of therapy for pancreatic cancer. Am J Surg 1986; 151:76-80.[CrossRef][Medline]
  44. Niederau C, Grendell J. Diagnosis of pancreatic carcinoma: imaging techniques and tumor markers. Pancreas 1992; 7:66-86.[Medline]
  45. Friess H, Langhans J, Ebert M, et al. Diagnosis of pancreatic cancer by 2[18F]-fluoro-2-deoxy-D-glucose positron emission tomography. Gut 1995; 36:771-777.[Abstract/Free Full Text]
  46. Zeman RK, Cooper C, Zeiberg AS, et al. TNM staging of pancreatic carcinoma using helical CT. AJR Am J Roentgenol 1997; 169:459-464.[Abstract/Free Full Text]
  47. Low RN, Barone RM, Lacey CL, Sigeti JS, Alzate GD, Sebrechts CP. Peritoneal tumor: MR imaging with dilute oral barium and intravenous gadolinium-containing contrast agents compared with unenhanced MR imaging and CT. Radiology 1997; 204:513-520.[Abstract/Free Full Text]
  48. Nishiharu T, Yamashita Y, Abe Y, et al. Local extension of pancreatic carcinoma: assessment with thin-section helical CT versus with breath-hold fast MR imaging—ROC analysis. Radiology 1999; 212:445-452.[Abstract/Free Full Text]
  49. Gazelle GS. Cost-effectiveness of imaging and surgery in patients with colorectal cancer liver metastases. Thesis Cambridge, Mass: Harvard University, 1999.
  50. Trede M, Rumstadt B, Wendl K, et al. Ultrafast magnetic resonance imaging improves the staging of pancreatic tumors. Ann Surg 1997; 226:393-407.[CrossRef][Medline]
  51. Gress F, Savides T, Cummings O, et al. Radial scanning and linear array endosonography for staging pancreatic cancer: a prospective randomized comparison. Gastrointest Endosc 1997; 45:138-142.[CrossRef][Medline]
  52. Ahkahoshi HK, Chihjiiwa Y, Nakano I, et al. Diagnosis and staging of pancreatic cancer by endoscopic ultrasound. Br J Radiol 1998; 71:492-496.[Abstract]
  53. Minnard EA, Conlon KC, Hoos A, Dougherty EC, Hann LE, Brennan MF. Laparoscopic ultrasound enhances standard laparoscopy in the staging of pancreatic cancer. Ann Surg 1998; 228:182-187.[CrossRef][Medline]
  54. John TG, Greig JD, Carter DC, Garden OJ. Carcinoma of the pancreatic head and periampullary region: tumor staging with laparoscopy and laparoscopic ultrasonography. Ann Surg 1995; 221:156-164.[Medline]
  55. Burris H, Moore M, Andersen J, et al. Improvements in survival and clinical benefit with gemcitabine as first-line therapy for patients with advanced pancreas cancer: a randomized trial. J Clin Oncol 1997; 15:2403-2413.[Abstract/Free Full Text]
  56. Smith A, Dowsett J, Russell R, Hatfield A, Cotton P. Randomised trial of endoscopic stenting versus surgical bypass in malignant low bile duct obstruction. Lancet 1994; 344:1655-1660.[CrossRef][Medline]
  57. Brandabur J, Kozarek R, Ball T, et al. Nonoperative versus operative treatment of obstructive jaundice in pancreatic cancer: cost and survival analysis. Am J Gastroenterol 1988; 83:1132-1139.[Medline]
  58. Johnson P, Outwater E. Pancreatic carcinoma versus chronic pancreatitis: dynamic MR imaging. Radiology 1999; 212:213-218.[Abstract/Free Full Text]
  59. Jimenez R, Fernandez–del Castillo C, Rattner D, Chang Y, Warshaw A. Outcome of pancreaticoduodenectomy with pylorus preservation or with antrectomy in the treatment of chronic pancreatitis. Ann Surg 2000; 231:293-300.[CrossRef][Medline]
  60. Riker A, Bartlett D. Pancreaticoduodenectomy for chronic pancreatitis: a case report and literature review. Hepatogastroenterology 1999; 46:2005-2010.[Medline]
  61. Smith C, Behrns K, Van Heerden J, Sarr M. Radical pancreatoduodenectomy for misdiagnosed pancreatic mass. Br J Surg 1994; 81:585-589.[Medline]
  62. Centers for Disease Control and Prevention. National vital statistics report, 1997 Hyattsville, Md: National Center for Health Statistics, 1997.
  63. Caraceni A, Portenoy R. Pain management in patients with pancreatic carcinoma. Cancer 1996; 78:639-653.[Medline]
  64. Bureau of Labor Statistics. Consumer price index: medical care.; Available at: www.bls.gov/. Accessed August 2000..
  65. Singh SM, Longmire WP, Jr, Reber HA. Surgical palliation for pancreatic cancer: the UCLA experience. Ann Surg 1990; 212:132-139.[Medline]
  66. Holbrook RF, Hargrave H, Traverso W. A prospective cost analysis of pancreatoduodenectomy. Am J Surg 1996; 171:508-511.[CrossRef][Medline]
  67. Italian Group for Antiemetic Research. Dexamethasone alone or in combination with ondansetron for the prevention of delayed nausea and vomiting induced by chemotherapy. New Engl J Med 2000; 342:1554-1559.[Abstract/Free Full Text]
  68. U.S. Census Bureau. Historical income tables: people.; Available at: www.census .gov/. Accessed July 8, 1999.
  69. Callery M, Strasberg S, Doherty G, Soper N, Norton J. Staging laparoscopy with laparoscopic ultrasonography: optimizing resectability in hepatobiliary and pancreatic malignancy. J Am Coll Surg 1997; 185:33-39.[CrossRef][Medline]
  70. Conlon KC, Dougherty EMA, Klimstra DS, Coit DG, Turnbull ADM, Brennan MF. The value of minimal access surgery in the staging of patients with potentially resectable peripancreatic malignancy. Ann Surg 1996; 223:134-140.[CrossRef][Medline]
  71. Schwarze M, Rattner DW. Diagnosis and staging of pancreatic cancer: the role of laparoscopy. In: Reber H, eds. Pancreatic cancer: pathogenesis, diagnosis and treatment. Totowa, NJ: Humana, 1998; 171-178.
  72. Glimelius B, Hoffman K, Graf W, et al. Cost-effectiveness of palliative chemotherapy in advanced gastrointestinal cancer. Ann Oncol 1995; 6:267-274.[Abstract/Free Full Text]
  73. Graham JD, Corso PS, Morris JM, Segui-Gomez M, Weinstein MC. Evaluating the cost-effectiveness of clinical and public health measures. Annu Rev Public Health 1998; 19:125-152.[CrossRef][Medline]
  74. Garber AM, Phelps CE. Economic foundation of cost-effectiveness analysis. J Health Econ 1997; 16:1-31.[CrossRef][Medline]
  75. Fitzsimmons D, Johnson CD. Quality of life in pancreatic cancer. In: Reber H, eds. Pancreatic cancer: pathogenesis, diagnosis and treatment. Totowa, NJ: Humana, 1998; 319.
  76. Mao C, Domenico D, Kitai K, Hanson D, Howard J. Observations on the developmental patterns and the consequences of pancreatic exocrine adenocarcinoma: findings of 154 autopsies. Arch Surg 1995; 130:125-134.[Abstract]
  77. 1999 Drug topics red book Montvale, NJ: Medical Economics, 1999.
  78. Karlson B, Forsman C, Wilander E, et al. Efficiency of percutaneous core biopsy in pancreatic tumor diagnosis. Surgery 1996; 120:75-79.[CrossRef][Medline]
  79. Geer RJ, Brennan MF. Prognostic indicators for survival after resection of pancreatic adenocarcinoma. Am J Surg 1993; 165:68-73.[CrossRef][Medline]



This article has been cited by other articles:


Home page