Threshold Analysis of the Cost-Effectiveness of Endoscopic Ultrasound in Patients at High-Risk for Pancreatic Ductal Adenocarcinoma

Abstract

Objectives:

Data from the International Cancer of the Pancreas Screening Consortium studies have demonstrated that screening for pancreatic ductal adenocarcinoma can be effective, and that surveillance improves survival in high-risk individuals. Endoscopic ultrasound (EUS) and cross-sectional imaging are both used, although there is some suggestion that EUS is superior. Demonstration of the cost-effectiveness of screening is important in order to implement screening in high-risk groups.

Methods:

Results from centers with EUS-predominant screening were pooled to evaluate efficacy of index EUS in screening. A decision analysis model simulated the outcome of a high-risk patients who undergo screening, and evaluated the parameters that would make screening cost-effective at a $100,000/quality adjusted life-year willingness to pay.

Results:

One-time index EUS has a sensitivity of 71.25% and specificity of 99.82% to detection to detect high-risk lesions. Screening with index EUS was cost-effective, particularly at lifetime pancreatic cancer probabilities of greater than 10.8%, or at lower probabilities if life expectancy after resection of a lesion that was at least 16 years, and if missed lesion rates on index EUS are ≤5%.

Conclusions:

Pancreatic cancer screening can be cost-effective through index EUS, particularly for those individuals at high-lifetime risk of cancer.

INTRODUCTION

Pancreatic ductal adenocarcinoma (PDAC) is the fourth most common cancer in the United States, and amongst the most lethal cancers, with a five-year survival of only 8.5%.^1,2 The American Cancer Society suggests there is a 1.5% lifetime risk for the average American to develop PDAC.³ Most patients develop PDAC sporadically, a result of epistatic interactions between biologic, lifestyle and environmental factors.^4–6 Five to ten percent of PDAC cases have a strong hereditary basis (Table 1).^4,7–11 Patients with these syndromes have elevated lifetime PDAC risk, as do those with strong family history of PDAC, even without an identified mutation.¹² Patients are classified as having familial pancreatic cancer (FPC) if they come from a kindred with ≥2 first-degree relatives (FDRs) with PDAC, and their risk of developing pancreatic cancer increases with the number of affected relatives.^8,13 Individuals from FPC kindred are considered high-risk individuals (HRIs) are considered appropriate candidates for potential screening for early detection of neoplastia.^2,8,11

TABLE 1.

Lifetime Risks of Pancreatic Ductal Adenocarcinoma (PDAC)

	Lifetime Probability of PDAC, %
Average risk12	1.5
FPC with 1 FDR8	6.75
FPC with 2 FDR8	9.6
FPC with 3+ FDR8	48
BRCA2 38	5
Peutz-Jeghers syndrome8,38	36
Hereditary pancreatitis8,38	30–50
Familial atypical multiple mole melanoma syndrome38	10–15

FPC indicates familial pancreatic cancer; FDR, first degree relative.

The optimal modality of PDAC screening in HRIs is still under investigation, and there are ongoing screening trials in the United States, including the Cancer of the Pancreas Screening (CAPS) initiative that includes 8 academic medical centers. The 2012 CAPS consortium recommended screening HRIs, and initial screening was performed with endoscopic ultrasonography (EUS) and/or magnetic resonance imaging (MRI)/magnetic resonance cholangiopancreatography (MRCP).² Although there is some evidence that EUS is superior to MRI in detection of pancreatic lesions, this premise has not been substantiated longitudinally.^14–16

The CAPS studies demonstrate that screening is effective in detecting lesions and that surveillance is associated with improved survival.¹⁴ Questions remain regarding the optimal modality, frequency of screening, and management of abnormal pancreatic findings.² Equally importantly, screening needs to be cost-effective for universal adoption, though studies evaluating this have been limited.^16–18

As part of the CAPS5 trial (ClinicalTrials.gov Identifier: NCT02000089), three centers predominantly perform annual EUS for HRIs (Case Western University Hospital, University of Pittsburgh, University of Pennsylvania). Using data from CAPS5, we are able to model potential outcomes for screening HRIs with EUS. The purpose of this study is to evaluate which factors make index EUS as a potential cost-effective strategy in identifying lesions in patients at high-risk of developing PDAC.

MATERIALS AND METHODS

Model Overview

We developed a decision-analytic model using TreeAge Pro 2019 release 1 (TreeAge, Williamstown, Mass), simulating an HRI who undergoes screening (Fig. 1) to evaluate parameters that would make screening for PDAC in HRI’s by index EUS cost-effective at a willingness to pay (WTP) of 100,000 US dollars/quality adjusted life year (QALY) gained.

FIGURE 1.

Decision tree model simulating the outcome of a high-risk individual who undergoes screening for pancreatic cancer.

The HRI can be screened by index EUS versus not screened (current standard of care). Screening includes one index EUS at age 55. Outcomes include normal EUS, an abnormal EUS that prompts surgery, or indeterminate EUS requiring follow-up EUS at some time interval but not prompting surgery. Individuals found to have a normal index EUS (true negative) could develop PDAC in the future, as their underlying risk persists throughout their life.

The primary neoplasm of interest was PDAC, although we also included high-grade dysplasia in pancreatic intraepithelial neoplasia (PanIN) or intraductal papillary mucinous neoplasia (IPMN), given these are the major precursor lesions of PDAC.¹⁹ Pancreatic neuroendocrine tumors (NETs), with features also included.⁷ Since our study involves an in silico mathematical model, institutional review board approval was not necessary. Institutional review board approval was obtained previously for the CAPS5 study at each center.

Study Selection and Pooling

In addition to CAPS5 data from centers that screen with annual EUS, we included 8 previously published studies that used EUS as index screening of PDAC in HRIs (Table 2), and these studies were pooled with CAPS5 data and meta-analysis was performed (Supplemental Methods).²⁰ No complications from EUS were registered during any of the screening programs, and overall complications are minimal.^21,22

TABLE 2.

Studies Included in Cost-Effectiveness Analysis

Study, Year	Mean Age, y	Female, %	Total Screened by index EUS, n	Cohorts Included	Abnormal EUS, n	Surgery, n	Found Lesions That are Premalignant or Malignant, n
CAPS5	61	65	268	FPC, BRCA 1 & 2, FAMMM, PJS, Lynch syndrome, PALB2, ATM	106	3	3 PDAC (all FPC)
Rulyak and Brentnall, 200139*	-	-	35	FPC	11	11	11 with PanIN 2 and 3 (all FPC)
Kimmey, et al, 200240†	-	-	46	FPC	24	12	12 with “widespread dysplasia, no carcinoma” (all FPC)
Poley, et al, 200938	50	59	44	FPC, FAMMM, BRCA1/2, PJS, HP, TP53	10	3	3 PDAC (1 with BRCA2, 2 related patients with FAMMM)
Canto, et al, 200429	57	61	38	FPC, PJS	29	7	4:1 PDAC, FPC Borderline IPMN, diffuse PanIN 1–2, PJS Diffuse PanIN 1–2, ectopic spleen, 0.5-cm microcystic serous cystadenoma, FPC Diffuse, multiple, PanIN 1–3, pancreatic abscess, mild focal acute and chronic pancreatitis, FPC
Zubarik, et al, 201141	59	60	26	-	12	3	2: 1 PDAC, 1 PNET
Verna, et al, 201036	52	65	31	FPC, BRCA1/2	7	4	4: 2 PDAC, 1 IPMN-B with moderate dysplasia and multifocal PanIN2, 1 IPMN-B with moderate dysplasia, focal PanIN2
Sud et al, 201442	51	93	16	FDR, BRCA1/2, PJS, p16	3	3	3: 2 PDAC (BRCA1/2), 1 IPMN with low grade dysplasia (PJS)
Gangi, et al, 201843‡	60	60	58	FPC, BRCA2, PJS	19		0

^*1 patient developed findings in follow up that prompted surgery, also found to have PanIN 2 and 3

^†In 24 months of follow up, 3 additional patients were found to have concerning imaging, all underwent surgery, also found to have “widespread dysplasia”

^‡In 5 years of follow up, 5 additional patients were found to have abnormal EUS findings on surveillance study, 1 underwent surgery, surgical specimen with IPMN

FPC indicates familial pancreatic cancer; FDR, first degree relative; PJS, Peutz-Jeghers syndrome; HP, hereditary pancreatitis; FAMMM, familial atypical multiple mole melanoma syndrome; IPMN, intraductal papillary mucinous neoplasm; PNET, pancreatic neuroendocrine tumor.

Model Inputs and Assumptions for Unknown Inputs

Probabilities, costs, and utilities are shown in Table 3. Costs were based upon previous studies using Center for Medicare Services data, and adjusted for inflation to 2018.^23,24 For those inputs with minimal or imperfect published literature, extrapolation or expert opinion was used (B.W.K., A.K.R.), and was widely varied in sensitivity analyses. The probability of future PDAC after one-time screening is unknown. It is possible that the initial EUS is most likely to reveal a lesion, but the value of the mitigated risk is unclear. We posited a future lifetime risk of being diagnosed with PDAC after a baseline EUS without detection of PDAC of 3%, twice that of an average risk person. We further posited that if an individual had a second EUS without PDAC at some point in the future, that would further decrease their risk by 50%. These assumptions were varied in sensitivity analyses.

TABLE 3.

Utilities and Costs

Parameter	Base Case	Sensitivity Analysis, Range or SD	Distribution	Reference
Probabilities
Lifetime probability of PDAC in HRI	9.6% (equal to lifetime probability of PDAC in HRI with 2+ FDR with PDAC)	0.05–0.2	Triangular	(Table 2)
Normal index EUS in HRI	0.607	0.0607	Normal	16,24,44
Index EUS with findings that prompt surgery	0.0812	0.00812	Normal	16,24,44
Index EUS with findings that prompt surgery, and surgery finds premalignant/malignant lesion	0.998	0.9–1	Triangular	16,24,44
Missed lesion on index EUS	0.05	0.025*	Normal	(Expert opinion)
Probability second EUS (after indeterminate index EUS) finds a lesion	0.05	0.025*	Normal	(Expert opinion)
Probability of future PDAC after normal index EUS	0.03	0.015*	Normal	(Expert opinion)
Utilities
Normal exam	1	0.99–1	Normal	16,24,44
PDAC	0.5	0.05	Normal	16,24
Screen abnormal, but don’t have cancer or undergo surgery	0.99	0.99–1	Normal	44
Screen abnormal, undergo surgery and removal of high-grade lesion	0.88	0.088	Normal	16,24
Screen abnormal, undergo surgery, lesion identified is not high-grade	0.88	0.088	Normal	16,24
Surgery post op period	0.5 × 3 mo	-	-	16,24
Years
Life expectancy after EUS without cancer	23.6	21.24–25.96	Normal	CDC/NCHS
Life expectancy of PDAC	0.8	0.72–0.88	Normal	16
Life expectancy after resection of premalignant/malignant lesion	12.5	11.25–13.75	Normal	26
Life expectancy after pancreatectomy without premalignant/malignant lesion found	15	13.5–16.5	Normal	14,28,29
Costs45
EUS with 1-hour anesthesia	$984.88	$886.39–$1083.37	Normal	24
No screen	$0	—	—	—
Cost of care for distant cancer, total lifetime cost	$56,562.94	$50,906.65–$62,219.23	Normal	23
Cost of care for resectable cancer, total lifetime cost	$155,490.36	$139.941.32–$171,039.40	Normal	23
Cost of pancreatectomy without associated cost of cancer treatments, total lifetime cost	$19,935.56	$17,942.00–$21,929.12	Normal	23

^*For the purpose of the analysis, the base case probabilities were altered when the sum total of all probabilities exceeded “1”

CDC indicates Centers for Disease Control and Prevention; NCHS, National Center for Health Statistics.

The probability of EUS missing a lesion is unknown, since screening programs are too recent to truly estimate “missed cancers” and there is no comparable “gold standard”. We conducted the model with different rates of “missed” lesions, ranging from 3–10%. We hypothesized a repeat EUS exam after an “indeterminate” EUS. The included studies had different “next steps” after abnormal EUS, varying from ERCP to CT to MR.² Given current clinical practice, we posited a repeat EUS at some time interval after an indeterminate index exam. That EUS could be reassuring, or find a lesion 5% of the time and require surgery.

The lifetime expectancy of those who undergo surgery and are found to have a benign lesion is unclear. Current literature reveals 75–80% 5-year overall survival after pancreatectomy.^25–27 Lifetime expectancy of patients who were found to have a lesion in screening and who undergo surgery is similarly not established. Small studies suggest prognosis correlates with lesion size.^28,29 Based upon previously published CAPS data, patients who undergo surgery can survive over 12 years.¹⁴ Our base case had a value of 12.5 years.¹⁴

Model Outputs

The model was evaluated using a hypothetical HRI with at least 5% lifetime risk. Threshold analysis included other FPC cohorts (Table 1).^7,8,30 Outcome measures are reported in incremental cost-effectiveness ratios (ICER).¹⁷ They were compared to a WTP of 100,000 US dollars/QALY gained. By multiplying life expectancy and utility, QALY was obtained. We chose to report from a third-party payer perspective, and included direct healthcare costs, but not societal costs, such as lost productivity, burden on families. All costs were discounted at a rate of 3% per year.³¹ A supplemental analysis evaluated the impact of QALY when using previously published age and sex-stratified utilities for the general US public.³² Previous cost-effectiveness studies did not use these stratifications and considered a year of “perfect health” to have a QALY of 1. To allow comparison to previous studies we followed the same strategy for our primary analysis.^{16–18,33,34}

RESULTS

The model estimated the cumulative probability of a pancreas screening EUS being normal in 61% of patients undergoing surveillance, indeterminate 31% of the time, and discovering a lesion concerning enough to prompt surgery 8% of the time. Of those that prompted surgery, a lesion was uncovered in 30% (14/46). We identified a cumulative sensitivity of 71.25% and specificity of 99.82% of index EUS at detecting PDAC.

For base model inputs across all groups in deterministic analyses, the mean cost for screening was $19,448, versus $6221 without screening. Screening had increased mean effectiveness of 21.21 QALYs, versus 21.05 without screening, with ICER $82,669/QALY gained (Supplemental Table 1).

Sensitivity Analyses

Using stochastic analyses, cost-effectiveness by lifetime risk of PDAC is presented in Figure 2. At a probability of above 10.3% lifetime risk of PDAC, our model recommends screening, and above 10.8% lifetime risk of PDAC, meets a WTP threshold of $100,000/quality adjusted life years (QALY), with an ICER of $95,220. This leads to 140 additional QALYs if screening 1000 HRI’s.

FIGURE 2.

One-way sensitivity analysis evaluating the cost effectiveness of screening by lifetime probability of pancreatic cancer. Below 10.25%, no screening dominates screening.

Life expectancies and the probability that EUS prompts surgery, is normal, or is indeterminate significantly impact cost-effectiveness (Supplemental Fig. 1). While costs do not grossly impact our preset WTP, if the cost of care of resectable cancer was half its current estimate, screening for PDAC would meet a $50,000 WTP threshold.

Two-way sensitivity analyses demonstrated that the model was sensitive to life expectancy of those who underwent surgery with removal of a pre-malignant or malignant lesion (Supplemental Fig. 2). If patients with 5% lifetime probability of PDAC live 30 years, screening is cost-effective (ICER $64,248/QALY). If patients with 8% lifetime probability of PDAC live 22 years, screening is cost-effective (ICER $51,959/QALY). If patients with 9.6% lifetime probability of PDAC live 16 years after surgery with removal of a pre-malignant or malignant lesion, screening is cost-effective (ICER $90,791/QALY). If patients with removal of a pre-malignant or malignant lesion live at least 26 years, any lifetime probability above 5% meets a WTP of $100,000. The model was not sensitive to increasing years of life for those who underwent surgery for what turned out to be a benign lesion.

Table 4 displays ICERs after deterministic sensitivity analyses for variables. At a 5% lifetime probability of PDAC, there is no posited level of the varied variables that would make screening cost-effective. At a 6.75% lifetime probability of PDAC (one first degree relative with PDAC), screening could be recommended if those who have surgery survived as long as healthy individuals.

TABLE 4.

ICER in Deterministic Sensitivity Analyses

	Lifetime Risk Of Cancer
	6.75%	9.6%	11%
Future probability of PDAC
1.5%	—	$129,172	$29,352
3%	—	—	$83,175
5%	—	—	—
10%	—	—	—
Years surviving after surgery, found to have premalignant or malignant lesion
5	—	—	—
10	—	—	—
15	—	$290,750	$35,457
20	—	$29,436	$16,512
23.6	$126,902	$17,871	$11,924
Years surviving after surgery, found to have benign disease
5	—	—	$83,863
10	—	—	$83,517
15	—	—	$83,175
20	—	—	$82,835
23.6	—	—	$82,592
Missed lesion on index EUS
1%	—	$33,350	$16,867
3%	—	$124,430	$29,067
5%	—	—	$83,175
7%	—	—	—
10%	—	—	—
Second EUS after indeterminate finds an intervenable lesion
1%	—	—	$37,628
3%	—	—	$53,393
5%	—	—	$83,175
7%	—	—	$160,629
10%	—	—	—

Dash (—) screening dominated by no screening ICER

At a 9.6% lifetime probability of PDAC (two first-degree relatives with PDAC), screening is cost-effective if those who undergo surgery survived at least 20 years or the missed lesion rate on index EUS was 1%.

At an 11% lifetime probability, index screening was cost-effective if the future probability of PDAC after index EUS was 1.5% or 3%, if those who undergo surgery survived at least 15 years, if missed lesions happened at a rate ≤5%, or if the second EUS found an intervenable lesion ≤7%.

Supplemental Table 2 demonstrates how QALYs would change based on previously published age and sex-stratified utilities evaluating quality of life among the general, noninstitutionalized civilian population in the US.³² This study demonstrates a mean health-related quality of life of 0.861 among 50–59 year old males, and 0.837 among 50–59 year old females. We altered utilities according to these adjustments, and demonstrated that among females, screening would provide 17.71 QALYs (versus 17.64 in no screening), and among males screening would provide 18.21 QALYs (versus 18.14 in no screening). In these cases, ICER would meet a $200,000 WTP threshold (ICER $188,957/QALY gained).

DISCUSSION

Screening for PDAC is an evolving field, but CAPS studies have shown that screening can be efficacious through identification of early pancreatic lesions. We developed a novel model for simulating screening in HRI and identify cases in which screening can also be cost-effective, using data from the largest group of EUS screening for PDAC collected thus far.

We find that index EUS is normal 61% of the time, indeterminate 31% of the time, and finds a lesion concerning enough to prompt surgery in 8%. Of those that evolve to surgery, 30% undergo surgery. Screening is cost-effective with an ICER of $82,669/QALY. To note, “cost-effectiveness” is nuanced, and while we had pre-determined to measure against an ICER of $100,000 per QALY, there is no “standard” QALY barometer, and an ICER of over $100,000 per QALY has been suggested to be more appropriate. Thus, even when our findings suggest that screening is not cost-effective against our initial predetermined WTP threshold, screening may still be a reasonable measure.³⁵

We find that cost-effectiveness can vary significantly, and the model was very sensitive to lifetime risk of PDAC (with lifetime risks above 10.8% easily meeting WTP), probability of future PDAC after normal index EUS, and probability of a missed lesion. Length of survival after resection impacted cost-effectiveness as well. Cost of pancreatectomy and years surviving after what turns out to be benign disease did not affect the model.

For a patient with ≥2 FDRs with PDAC, the most common indication for screening HRI, screening is cost-effective if those who had resection survive more than 16 years after surgery or the missed lesion rate was ≤5%. These are unknowns at this time, and further screening programs will provide real-world inputs to confirm, but our results suggest that persons with ≥2 FDRs with PDAC should continue to be enrolled in screening programs. As expected, screening is more cost-effective at higher lifetime risk of PDAC. Screening previously included those individuals without FPC who have just 1 FDR with PDAC, but this has now been removed from screening recommendations, which is supported by our model. For those with a higher lifetime PDAC probability, screening is cost-effective as long as the index screening truly decreases future probability of PDAC, if those who undergo resection survive more than 10 years, if missed lesions on index EUS happened at a rate of ≤5%, or if the second EUS found an intervenable lesion at a rate of ≤5%.

Our study includes the largest amount of available data on EUS for screening, and is the major strength of our study.^16–18 Limitations of our study include certain unknown variables, including: the miss rate of EUS, the findings of a second EUS after an initial indeterminate EUS, the probability of future PDAC after one-time screening, and life expectancies of those who undergo surgery. This required us to rely upon expert opinion, although as future prospective studies define further these variables, future models can incorporate real-world data. Additionally, information is needed on the precursor lesions which portend future risk. For example, the detection of PanIN remains of unclear prognostic significance. As is standard practice, we varied unknown variables significantly in sensitivity analyses. As such, due to the ranges of our estimates, deterministic (point estimates) versus stochastic (random probability distribution) models varied slightly, and explains why our ICERs are so variable. Given these are mathematical models, unable to mimic all potential variability of inputs, as prospective studies provide more real-world data, future mathematical modeling will continue to be more accurate in its assumptions. We also chose to only include index screening given that this is most widely reported and that patients often are lost to follow-up in other recent screening programs, which would limit information regarding natural history. While a Markov may represent a more complete picture, our model reflects the currently available state of data for one-time EUS screening. A Markov would require extrapolating data from the index screening group to estimate surveillance strategies, however, neither CAPS5 data nor previously published data were robust enough to provide data for all potential strategies. It has been established in other cancers that screening and surveillance rates are disparate, and as such, using pooled screening data to estimate detection rates after index EUS would not be reflective. As CAPS5 matures, we hope to have these data, and future studies could then improve upon our model. However, we feel that our study expands upon prior studies by utilizing real world data, strengthened by meta-analysis, and more comprehensive (although we acknowledge not fully exhaustive) decision tree. There was also heterogeneity in the non- CAPS5 studies, particularly the manner in which FPC was defined, as some studies specified number of first-degree relatives, while others included “any degree relatives.” Given lack of data information from the pooled studies, we were unable to distinguish the sensitivity of index EUS by genetic mutation or syndrome. This would have been important since test diagnostic accuracy may vary by high-risk syndrome, and subsequently impact whether or not screening is cost-effective for that group. Additionally, one study reclassified genetic mutations into high-, moderate-, or low-risk, limiting our ability to pool by cohort.³⁶ While we included total pancreatectomy in our model, based upon prior studies, partial pancreatectomy may also be an option for some patients. In these cases, peri-operative utility and survival values may change, and future screening may be indicated. This was beyond the scope of our study but future studies could consider modeling based on the frequency of surgical subtypes as increasing numbers of patients undergo detections of lesions in screening programs. Further evaluation of patient age, co-morbidities, and type of surgery in relation to peri-operative mortality can also be included in future studies, to further define the cost-effectiveness of screening. As with all mathematical models, our model does not reflect all possible avenues for complications. For example, mortality associated with surgical resection was not expressly included as mortality associated with pancreatic surgery is low (1–2% in expert centers), although morbidity was included. Finally, administration of a health-related quality of life survey, such as the EuroQoL EQ-5D, to patients undergoing pancreatic cancer screening would allow for the most targeted and accurate utility values among disease states, rather than extrapolating from other diseases, as we were required to do due to dearth of data.³⁷ As our supplemental analysis demonstrates (Supplemental Table 2), effectiveness can greatly vary based on utility values. Cost-effectiveness studies routinely use a QALY of 1 to indicate a year of “perfect health”, but it is rare that persons routinely report a year of “perfect health”.³² As such, there is some limitation of cost-effectiveness studies as a whole to reflect realistic health states. Lastly, this does not include MRI/MRCP as part of the screening algorithm, which many institutions and trials are including, limiting generalizability of the model.

More information is needed in future studies regarding: 1) long-term follow up of those who undergo resection, 2) development of PDAC outside screening, 3) the risk to patients with mutations and a family history of PDAC, and 4) the survival benefit conferred by early detection. Future studies could also further evaluate the cost-effectiveness of combination surveillance by EUS and imaging (MRI) based upon real-world data and the utility of serial examinations. Importantly, EUS is an operator-dependent tool, and this is an important factor in the ability to implement widespread screening of PDAC, which at this time remains limited to tertiary referral centers.

Our study demonstrates how screening for PDAC can be a cost-effective measure, particularly with groups that have at least 10.8% lifetime risk of PDAC, or if continued studies demonstrate long life expectancies post-resection, low rates of missed lesions on index EUS, and decreased future probability of PDAC after one-time screening. Screening programs should continue to collect data and surveillance programs are critical to identifying the natural history of HRIs.

Abbreviations

CAPS

Cancer of the Pancreas Screening

EUS

endoscopic ultrasound

FDR

first-degree relatives

FPC

familial pancreatic cancer

HRI

high-risk individuals

ICER

incremental cost-effectiveness ratios

IPMN

intraductal papillary mucinous neoplasia

MRCP

magnetic resonance cholangiopancreatography

MRI

magnetic resonance imaging

NET

neuroendocrine tumor

PDAC

pancreatic ductal adenocarcinoma

QALY

quality adjusted life year

WTP

willingness to pay