Cost-Effectiveness Analysis of Screening for Pancreatic Cancer Among High-Risk Populations

Abstract

PURPOSE

We evaluated the potential cost-effectiveness of combined magnetic resonance imaging (MRI) and endoscopic ultrasound (EUS) screening for pancreatic ductal adenocarcinoma (PDAC) among populations at high risk for the disease.

METHODS

We used a microsimulation model of the natural history of PDAC to estimate the lifetime health benefits, costs, and cost-effectiveness of PDAC screening among populations with specific genetic risk factors for PDAC, including BRCA1 and BRCA2, PALB2, ATM, Lynch syndrome, TP53, CDKN2A, and STK11. For each high-risk population, we simulated 29 screening strategies, defined by starting age and frequency. Screening included MRI with follow-up EUS in a subset of patients. Costs of tests were based on Medicare reimbursement for MRI, EUS, fine-needle aspiration biopsy, and pancreatectomy. Cancer-related cost by stage of disease and phase of treatment was based on the literature. For each high-risk population, we performed an incremental cost-effectiveness analysis, assuming a willingness-to-pay (WTP) threshold of $100,000 US dollars (USD) per quality-adjusted life year (QALY) gained.

RESULTS

For men with relative risk (RR) 12.33 (CDKN2A) and RR 28 (STK11), annual screening was cost-effective, starting at age 55 and 40 years, respectively. For women, screening was only cost-effective for those with RR 28 (STK11), with annual screening starting at age 45 years.

CONCLUSION

Combined MRI/EUS screening may be a cost-effective approach for the highest-risk populations (among mutations considered, those with RR >12). However, for those with moderate risk (RR, 5-12), screening would only be cost-effective at higher WTP thresholds (eg, $200K USD/QALY) or with once-only screening.

BACKGROUND

Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal malignancy,¹ and there is significant interest in methods for early detection to increase survival. Germline genetic pathologic variants have been identified in about 10% of patients with PDAC,^2-8 and germline genetic testing is recommended for all patients with the disease.^9,10 As germline genetic testing expands, more high-risk family members are being identified.

CONTEXT

• Key Objective

• We want to understand if screening for pancreatic cancer is cost-effective among a variety of high-risk groups.

• Knowledge Generated

• We modeled eight high-risk groups, defined by specific germline genetic mutations, and applied 29 screening schedules for each group. Staged magnetic resonance imaging/endoscopic ultrasound screening was cost-effective for patient populations with a relative risk of pancreatic ductal adenocarcinoma (PDAC) of 12 or higher (eg, CDKN2A and STK11), but for patients with moderate risk of PDAC (eg, BRCA2, Lynch syndrome), there was clinical benefit but above the willingness-to-pay threshold of $100,000 US dollars/quality-adjusted life year gained.

• Relevance

• Screening for pancreatic cancer is important for high-risk groups, but current screening approaches are expensive and imperfect. Different high-risk groups benefit from different starting ages and frequency of screening. Improved staged screening approaches are needed to best serve these high-risk populations.

While screening for pancreatic cancer carries the promise of early detection, the mainstay screening modalities, magnetic resonance imaging (MRI) and endoscopic ultrasound (EUS), have disadvantages including modest test performance characteristics for cancer detection and high associated cost, particularly when repeated yearly.¹¹ The low incidence of PDAC, combined with the relatively high-burden interventions required for surveillance and risk mitigation, makes population-level screening currently impractical and expensive. Several investigators have reported their experiences in annual surveillance of high-risk groups with EUS, MRI, and/or computed tomography and have found low but clinically relevant detection rates in these patients.^12-14 However, given imperfect detection rates and a narrow window of opportunity before symptomatic disease progression, the optimal timing of initiation and frequency of screening are yet to be determined. Similarly, the cost-effectiveness of these relatively expensive screening modalities is not well-understood.

High-risk groups are defined by a strong family history, a proven germline pathogenic variant, or both. Family history moderately elevates PDAC risk relative to the general population (relative risk [RR], 1.7-4.6); the extent of such elevation depends on the number of affected family members.^15,16 Several known germline genetic pathologic variants that are associated with increased PDAC risk include Lynch syndrome (MLH1, MSH2, MSH6, PMS2, EPCAM),^17,18 hereditary breast and ovarian cancers (HBOC; BRCA1, BRCA2, PALB2),^3,19-22 and familial atypical multiple mole melanoma (FAMMM; CDKN2A).^{4,6,11,19,23-25} Current recommendations by the Cancer of the Pancreas Screening (CAPS) Consortium are to screen patients with an estimated lifetime risk of pancreatic cancer of >10%.²⁶ This includes patients with ATM, BRCA1, BRCA2, PALB2, or Lynch syndrome mutations who also have a first-degree family member with PDAC and patients with STK11 or CDKN2A mutations regardless of family history.²⁶

We have previously shown that screening for PDAC has the potential to improve life expectancy in some high-risk groups^27,28 and that the age of screening initiation and frequency of screening can be adjusted to optimize this benefit.²⁹ In this study, we use a microsimulation model that projects the life expectancy benefit and attendant costs of PDAC screening among a variety of high-risk groups to determine the cost-effectiveness of potential screening strategies.

METHODS

Model Overview

As previously described, we have developed a microsimulation model of the natural history of PDAC.^27-30 The model, written in C++, simulates cohorts of hypothetical patients from age 20 years until death. As they age, they are at risk of developing precursor lesions, either cysts or pancreatic intraepithelial neoplasia (Fig 1). These precursor lesions may progress in risk level (cystic pathway) or grade (solid pathway) and may ultimately transition to preclinical (ie, undetected) PDAC. PDAC may progress in stage and may be detected by symptoms, becoming a clinical case. Individuals may die of other causes at any point in time; individuals with clinically diagnosed PDAC may also die from the disease. The values of all unobservable probabilities of transitioning between health states are derived by calibration to observable targets obtained from the Surveillance, Epidemiology, and End Results (SEER) Program database and published literature.^1,23,31-33 The simulation model is recalibrated at each PDAC risk level to match corresponding likelihoods of PDAC development and death.

FIG 1.

Model schema. The health states emphasized in a bold outline (high-risk cyst and undetected stage I) are the targets of screening. The dashed arrows represent risk-reducing surgery and early cancer detection. PDAC, Pancreatic Ductal Adenocarcinoma; PanIN, Pancreatic Intraepithelial Neoplasia.

Screening Strategies

For this analysis, we started with individuals who were free of diagnosed cancer at age 40 years. We used the model to evaluate screening strategies with starting age from 40 to 70 years in 5-year increments (ie, age 40, 45, 50 years, etc). We assumed that screening would be discontinued at age 75 years. We tested four screening frequencies: one time, every 5 years, every 2 years, and annually. Screening consisted of MRI with referral to EUS in the case of any concerning imaging findings. As data were not available to separately characterize these tests, we assumed an overall 62% sensitivity and 96% specificity for the detection of precancerous cysts and early-stage cancer; for metastatic malignancy, we assumed perfect detection.^11,25,34-44 If a low-risk cyst was detected, the patient was assumed to undertake 10 years of annual surveillance, during which we assumed perfect detection of instances of progression to a high-risk cyst. In the event of detection of a high-risk cyst (true or false positive), the patient was assumed to undergo surgical resection with an associated 2% probability of surgical mortality.^45-50 After PDAC detection, either cystic or solid, PDAC-associated mortality was assigned on the basis of SEER data.¹ We assumed that simulated individuals could not die from their cancer during their lead time (ie, the time between detection by screening and detection by symptoms in the absence of screening), and therefore, we applied cancer-specific mortality on the basis of the stage at detection, but not until the age at which the simulated individual would have been detected in the absence of screening.

Genetic Pathologic Variants

As in our previous analysis,²⁹ we defined eight specific cohorts of patients with germline pathologic variants known to increase the risk of PDAC. The relative lifetime risk (RR) of PDAC for each variant was estimated from the literature. The variants are PALB2 (RR, 2.33),⁵¹ BRCA1 (RR, 2.58),^{3,19,20,51-54} Lynch syndrome (MLH1, MSH2, MSH6, PMS2, EPCAM, RR 3.55),¹⁸ ATM (RR 5.71),^3,35,55,56 BRCA2 (RR 6.2),^3,21,51-54 TP53 (RR 6.7),⁵⁷ CDKN2A (RR 12.33),^58-63 and STK11 (RR 28).⁶⁴

Cost and Utility Estimation

Costs were estimated from a health care sector perspective. Costs of MRI and EUS were estimated by 2019 Medicare payment rates, including both physician and hospital reimbursement. Physician reimbursements were estimated using facility pricing from the 2019 Medicare and Medicaid Physician Fee Schedule.⁶⁵ Reimbursements for hospital outpatient services were estimated using Addendum B from the 2019 Hospital Outpatient Prospective Payment System.⁶⁶ We assumed that MRI would be performed with and without contrast and included professional fees for interpretation. We assume a 15% referral rate to EUS³⁵ and that EUS would include moderate sedation, proceduralist fees, fine-needle aspiration, and associated fees for pathologist interpretation.⁶⁷ The resulting total cost of a single screening event was estimated at $840 US dollars (USD).

Costs of noncancer surgery (ie, for patients with high-risk cysts or those with false-positive screening results) were estimated by 2019 Medicare payment rates, including physician and hospital reimbursement for Whipple's procedure. Physician reimbursements were estimated using facility pricing from the 2019 Medicare and Medicaid Physician Fee Schedule.⁶⁵ Hospital reimbursements were estimated using the 2019 Final Rule and Correction Notice Tables from the Hospital Acute Inpatient Prospective Payment System for Diagnosis Related Group (DRG) codes 405-407.⁶⁸ The resulting total cost of noncancer surgery was estimated at $26,100 USD.

Costs of cancer care were estimated by 2019 Medicare payment rates, including patient copayments, coinsurance, deductibles, and payments from other insurers, and were assigned on the basis of available literature.⁶⁹ We mapped historical SEER stages to American Joint Committee on Cancer eighth edition staging⁷⁰ and designated three phases of costs: initial (12 months from diagnosis), terminal (12 months before death), and continuing (months between initial and terminal). For patients who lived <24 months, priority of assignment was given to the terminal phase, then initial, and then continuing.

All costs were expressed in 2019 USD, with future costs discounted at an annual 3% rate.⁷¹ Details of cost estimates are presented in Table 1.

TABLE 1.

Key Input Assumptions: Costs (in 2019 USD) and Utilities

Category of Cost	Base Case	Sensitivity Analysis (low; high)	References
Screening costs, USD
Abdominal MRI (HCPCS codes 74183, 99213)	$551		65,66
Endoscopic ultrasound-guided FNA, pathology included (HCPCS codes 43238, 99152, 99153, 80502, 99213)	$1,928		65,66
Total screeninga	$840	$420; $1,260	35,65,66
Surgery-related costs, USD
Whipple's procedure (HCPCS code 48150)	$26,107	$13,053; $39,160	65,68
Cancer careb costs by AJCC stage and phase of carec (per month), USD
Stage I and stage II
Initial	$6,900	$3,450; $10,350	69
Continuing	$1,175	$588; $1,763	69
Terminal (died of PDAC)	$9,033	$4,517; $13,550	69
Terminal (died of other causes)	$6,508	$3,254; $9,762	69
Stage III
Initial	$10,542	$5,271; $15,813	69
Continuing	$1,842	$921; $2,763	69
Terminal (died of PDAC)	$10,442	5,221; $15,663	69
Terminal (died of other causes)	$6,508	$3,254; $9,762	69
Stage IV
Initial	$9,175	$4,588; $13,763	69
Continuing	$3,383	$1,692; $5,075	69
Terminal (died of PDAC)	$11,450	$5,725; $17,175	69
Terminal (died of other causes)	$6,508	$3,254; $9,762	69
Additional cancer-related costs, USD
Undetected cancer	$11,450	$5,725; $17,175	69
Age-based utilities, by sex, years
Men
40-49	0.887	0.880; 0.894	72
50-59	0.861	0.853; 0.870	72
60-69	0.840	0.827; 0.852	72
70-79	0.802	0.788; 0.816	72
80+	0.782	0.757; 0.807	72
Women
40-49	0.863	0.855; 0.871	72
50-59	0.837	0.829; 0.846	72
60-69	0.811	0.800; 0.822	72
70-79	0.771	0.758; 0.784	72
80+	0.724	0.701; 0.747	72
Discount rate	0.03	0.00; 0.05	71

Abbreviations: AJCC, American Joint Committee on Cancer; FNA, fine-needle aspiration; MRI, magnetic resonance imaging; PDAC, pancreatic ductal adenocarcinoma; USD, US dollars.

^a We weighted the total reimbursement payment rates for MRI and endoscopic ultrasound-guided FNA using the study by Barnes et al³⁵ in which 10 of 65 patients (15%) were referred to an endoscopic ultrasound-guided FNA ($551 + $1,928 × 0.15).

^b We converted net annual 2019 US Medicare payments from the study by Mariotto et al⁶⁹ to net monthly payments. Because our model uses AJCC definitions of stage, but Mariotto et al reported costs by SEER historic stages, we mapped the historic stage of localized to stage I and II pancreatic cancer, regional to stage III pancreatic cancer, and distant to stage IV pancreatic cancer.

^c Initial phase is the 12 months after diagnosis. Terminal phase is the 12 months before death. Continuing phase is all the months between the initial and terminal phases, if any. Priority of assignment is given to the terminal phase, then the initial phase, and finally the continuing phase.

We estimated the quality-adjusted life years (QALYs) associated with each screening strategy by incorporating age- and sex-specific health-related utility weights for all individuals.⁷² We used only age- and sex-stratified utilities as utility weights specific to pancreatic cancer are sparse and variable.^73-77 As with costs, future QALYs were subject to discounting at an annual rate of 3%. Details of utility estimates are provided in Table 1.

Cost-Effectiveness Analysis

For each cohort, defined by genetic or familial risk level, we evaluated a total of 29 screening strategies. Strategies varied by starting age and frequency; a no-screening strategy was also included. For each strategy, we calculated the associated life expectancy gain and QALY gained relative to the no-screening strategy. We also calculated the number of cancers that could be averted through each screening strategy relative to no screening for that cohort.

For each risk-defined cohort, we conducted an incremental cost-effectiveness analysis. Any strategy that was more costly and less effective than another strategy was deemed inefficient and eliminated from consideration, as was any strategy that was less costly and less effective than another strategy but provided additional QALYs gained at a higher additional cost. All remaining strategies were considered efficient options. We evaluated the relative performance of each efficient strategy by calculating its incremental cost-effectiveness ratio (ICER), defined as the additional cost of the screening strategy divided by its additional QALYs gained, relative to the next best option. We identified the cost-effective strategy from among the efficient options assuming a willingness-to-pay (WTP) threshold of $100,000 USD per QALY gained; the strategy with ICER closest to but less than the WTP threshold was deemed the cost-effective option. Our results are reported as per the standards of the Second Panel on Cost-Effectiveness in Health and Medicine, and the CHEERS 2002 checklist is available in Appendix Table A1 (online only).^71,78

Sensitivity Analysis

We conducted deterministic (one-way) sensitivity analysis to investigate how changes in model parameters could affect our results for each cohort. We specifically tested parameter ranges around the sensitivity and specificity of MRI/EUS screening, surgical mortality rate, utilities, discount rate, and costs of screening, surgery, and cancer care. In addition, we tested a higher WTP threshold of $200,000 USD per QALY gained.

RESULTS

Overview

We found that in men (Table 2), screening was only cost-effective at a WTP threshold of $100K USD/QALY gained for cohorts with RR 12.33 and RR 28 among those genetic variants considered. For men with RR 28 (STK11), the cost-effective strategy was annual screening starting at age 40 years, and for men with RR 12.33 (CDKN2A), it was annual screening starting at age 55 years. For women (Table 3), screening was cost-effective only for the cohort with RR 28 (STK11); this cost-effective strategy was annual screening starting at age 45 years. Further cohort-specific analysis is described below.

TABLE 2.

Efficient Screening Strategies for Each Cohort (men)

Population	RR PDAC	Strategy	aCancer Averted per 1,000 (% reduction)	aIncremental Cost ($), USD	aIncremental QALY Gained, Days	ICER (Δ Cost/Δ QALY gained, $)
PALB2	2.33	No screening	—	—	—	—
		Once at 70 years	1.4 (4)	810	0.25	1,187,000
		Every 5 years starting at 70 years	2.0 (6)	1,300	0.32	2,614,000
		Every 5 years starting at 65 years	2.9 (8)	2,000	0.35	11,217,000
BRCA1	2.58	No screening	—	—	—	—
		Once at 70 years	1.6 (4)	840	0.35	878,000
		Once at 65 years	1.8 (5)	960	0.39	966,000
		Every 5 years starting at 65 years	3.2 (8)	2,100	0.60	2,005,000
Lynch	3.55	No screening	—	—	—	—
		Once at 65 years	2.5 (5)	1,100	1.0	404,000
		Every 5 years starting at 65 years	4.3 (8)	2,200	1.7	623,000
		Every 5 years starting at 60 years	5.7 (11)	3,200	2.1	833,000
		Every 2 years starting at 65 years	6.4 (12)	3,600	2.2	1,483,000
		Every 2 years starting at 60 years	8.9 (17)	5,400	2.4	2,956,000
ATM	5.71	No screening	—	—	—	—
		Once at 60 years	4.0 (5)	1,400	2.7	191,000
		Every 2 years starting at 65 years	9.8 (12)	3,800	5.9	274,000
		Every 2 years starting at 60 years	14 (16)	5,700	8.2	292,000
		Annual starting at age 60 years	21 (25)	8,600	11	399,000
		Annual starting at age 55 years	27 (32)	12,200	13	619,000
BRCA2	6.2	No screening	—	—	—	—
		Once at 60 years	4.3 (5)	1,400	3.0	174,000
		Annual starting at age 65 years	16 (17)	5,300	8.9	239,000
		Every 2 years starting at 60 years	15 (16)	5,700	9.5	270,000
		Annual starting at age 60 years	23 (25)	8,600	13	307,000
		Annual starting at age 55 years	28 (31)	12,100	16	442,000
		Annual starting at age 50 years	33 (36)	16,600	16	10,214,000
TP53	6.7	No screening	—	—	—	—
		Once at age 60 years	4.7 (5)	1,500	3.4	160,000
		Annual starting at age 65 years	17 (17)	5,300	10	204,000
		Annual starting at age 60 years	24 (25)	8,500	15	252,000
		Annual starting at age 55 years	30 (31)	12,000	19	341,000
		Annual starting at age 50 years	35 (36)	16,500	20	1,648,000
CDKN2A	12.33	No screening	—	—	—	—
		Annual starting at age 65 years	27 (15)	5,100	23	80,000
		Annual starting at age 55 years	50 (28)	11,200	50	82,000
		Annual starting at age 50 years	59 (33)	15,400	62	137,000
		Annual starting at age 45 years	64 (36)	20,200	67	313,000
STK11	28	No screening	—	—	—	—
		Annual starting at age 45 years	110 (27)	17,400	220	29,000
		Annual starting at age 40 years	120 (29)	22,200	240	69,000

NOTE. For clarity, QALYs in this table are presented in days.

Abbreviations: ICER, incremental cost-effectiveness ratio; PDAC, pancreatic ductal adenocarcinoma; QALYs, quality-adjusted life years; RR, relative risk; USD, US dollars.

^a Compared with no screening.

TABLE 3.

Efficient Screening Strategies for Each Cohort (women)

Population	RR PDAC	Strategy	aCancer Averted per 1,000 (% reduction)	aIncremental Cost ($), USD	aIncremental QALY, days	ICER (Δ Cost/Δ QALY Gained, $)
PALB2	2.33	No screening	—	—	—	—
BRCA1	2.58	No screening	—	—	—	—
Lynch	3.55	No screening	—	—	—	—
		Once at 70 years	2.1 (5)	1,000	0.17	2,156,000
		Every 5 years starting at 70 years	3.0 (7)	1,600	0.21	5,697,000
		Every 5 years starting at 65 years	4.2 (10)	2,400	0.22	48,507,000
ATM	5.71	No screening	—	—	—	—
		Once at age 60 years	3.4 (5)	1,400	1.1	458,000
		Every 2 years starting at 65 years	9.9 (15)	4,000	2.7	615,000
		Annual starting at age 65 years	14 (21)	5,700	3.3	1,034,000
		Annual starting at age 60 years	20 (30)	9,200	3.7	3,084,000
BRCA2	6.2	No screening	—	—	—	—
		Once at age 60 years	3.7 (5)	1,500	1.4	387,000
		Every 2 years starting at 65 years	11 (15)	4,100	3.3	494,000
		Annual starting at age 65 years	16 (21)	5,700	4.2	631,000
		Annual starting at age 60 years	22 (30)	9,200	5.2	1,298,000
TP53	6.7	No screening	—	—	—	—
		Once at age 60 years	4.0 (5)	1,500	1.7	336,000
		Every 2 years starting at 65 years	12 (14)	4,100	4.0	411,000
		Annual starting at age 65 years	17 (21)	5,600	5.2	439,000
		Annual starting at age 60 years	23 (29)	9,100	6.8	806,000
		Annual starting at age 55 years	28 (36)	12,700	7.2	3,191,000
CDKN2A	12.33	No screening	—	—	—	—
		Annual starting at age 65 years	28 (19)	5,200	16	120,000
		Annual starting at age 55 years	49 (33)	11,500	32	145,000
		Annual starting at age 50 years	56 (38)	15,900	37	331,000
		Annual starting at age 45 years	60 (41)	20,900	37	10,452,000
STK11	28	No screening	—	—	—	—
		Annual starting at age 55 years	91 (28)	9,900	95	38,000
		Annual starting at age 50 years	110 (32)	13,100	130	38,000
		Annual starting at age 45 years	110 (35)	16,200	150	45,000
		Annual starting at age 40 years	120 (36)	21,300	160	152,000

NOTE. For clarity, QALYs in this table are presented in days.

Abbreviations: ICER, incremental cost-effectiveness ratio; PDAC, pancreatic ductal adenocarcinoma; QALYs, quality-adjusted life years; RR, relative risk; USD, US dollars.

^a Compared with no screening.

Highest-Risk Cohort (RR 28, STK11)

The ICERs for the STK11 cohort by screening strategy are shown in Figure 2 and Appendix Table A1. For men, the cost-effective strategy at a WTP threshold of $100K USD/QALY gained was annual screening starting at age 40 years. With this screening approach, men with RR 28 were estimated to achieve a QALY benefit of 240 days through prevention of 120 cancers per 1,000 patients (a 29% reduction). The incremental cost per patient was $22,200 USD for an ICER of $69,000 USD per QALY gained relative to the next most efficient strategy. Strategies involving less frequent screening (one time, every 5 years, or every 2 years) were dominated by the annual strategies.

FIG 2.

Quality-adjusted life-years gained and lifetime PDAC-related costs compared with no screening for 29 PDAC screening strategies among men with a STK11 mutation (RR, 28). Each datapoint represents a screening strategy. Strategies that make up the outer envelope are efficient, and the line that connects them is the efficient frontier. The incremental cost-effectiveness ratios for the efficient strategies are noted. For this population, three efficient strategies were identified: no screening, annual screening starting at age 45 years, and annual screening starting at age 40 years. Annual screening starting at age 50 years is close to but just below this frontier. PDAC, pancreatic ductal adenocarcinoma; QALY, quality-adjusted life year; RR, relative risk; USD, US dollars.

The cost-effective strategy at a WTP threshold of $100K USD/QALY gained for women with RR 28 was annual screening starting at age 45 years. With this screening approach, women with RR 28 were estimated to achieve QALY benefit of 150 days through prevention of 110 cancers per 1,000 patients (a 35% reduction). The incremental cost per patient was $16,200 USD for an ICER of $45,000 USD per QALY gained relative to the next most efficient strategy.

CDKN2A Cohort (RR 12.33)

Among men with RR 12.33 (CDKN2A), annual screening starting at age 55 years was cost-effective. With this screening approach, men with RR 12.33 were estimated to achieve QALY benefit of 50 days through prevention of 50 cancers per 1,000 patients (a 28% reduction relative to no screening). The incremental cost per patient was $11,200 USD for an ICER of $82,000 USD/QALY gained relative to the next most efficient strategy.

Women with RR 12.33 (CDKN2A) achieved the lowest ICER ($120,000 USD per QALY gained) with annual screening starting at age 65 years. This was not below the WTP threshold.

Other Cohorts

Among mutations evaluated, for men with a RR of PDAC between 2.33 and 6.7, screening was not cost-effective at a WTP threshold of $100K USD/QALY gained. Men with the PDAC RR of 6.7 (TP53), 6.2 (BRCA2), and 5.71 (ATM) had the lowest ICER with one-time screening at age 60 years, ranging from $160,000 USD to $191,000 USD per QALY gained. Men with a PDAC RR of 3.55 (Lynch syndrome) had the lowest ICER with one-time screening at age 65 years, with $ 404,000 USD per QALY gained. Men with the PDAC RR of 2.58 (BRCA1) and 2.33 (PALB2) had the lowest ICER with one-time screening at age 70 years, with $878,000 USD and $1,187,000 USD per QALY gained, respectively. These results are summarized in Table 2. The strategies on the efficient frontier for each cohort are summarized in Appendix Figure A1.

Among the mutations evaluated, women with the RRs of PDAC of 2.33 (PALB2) and 2.58 (BRCA1) were estimated to have fewer QALYs with versus without screening. For women with a RR of PDAC between 3.55 and 12.33, screening was not cost-effective at a WTP threshold of $100K USD/QALY gained. Women with the PDAC RR of 6.7 (TP53), 6.2 (BRCA2), and 5.71 (ATM) had the lowest ICER with one-time screening at age 60 years, ranging from $336,000 USD to $458,000 USD per QALY gained. Women with a PDAC RR of 3.55 (Lynch syndrome) had the lowest ICER with one-time screening at age 70 years, with $2,156,000 USD per QALY gained. These results are summarized in Table 3. The strategies on the efficient frontier for each cohort are summarized in Appendix Figure A1.

Sensitivity Analysis

The results were sensitive to test performance characteristics and, in particular, the specificity of screening. Perfect specificity screening increased ICERs and resulted in strategies with one-time screening, rather than annual screening, being cost-effective. Lower specificity assumptions (eg, 95%) did not substantially change the cohorts for which screening was cost-effective. Variations in screening sensitivity, utility assessments, and the cost of cancer care led to a minimal change in results. These results are summarized in Figure A2.

The results were also sensitive to screening costs. Lowering the cost of screening resulted in a lower risk threshold at which cost-effective strategies were identified at any given WTP threshold. For example, with screening costs halved, annual screening for women with RR 12.33 (CDKN2A) would be cost-effective at a threshold of $100K USD/QALY gained. Variation in the discount rate from 0% to 5% also led to substantial changes in results. In an undiscounted analysis, screening would be cost-effective in men with RR ≥5.71 and in women with RR ≥12.33, at a WTP threshold of $100K USD/QALY. These results are summarized in Figures 3 and 4.

FIG 3.

Deterministic sensitivity analyses on the cost-effective strategy at a WTP threshold of $100K USD and $200K USD (men). This figure shows the cost-effective screening strategy for each cohort defined by the RR of PDAC (along the top), each uncertainty scenario (along the left), and each assumption about the WTP for a quality-adjusted year of life gained (along the bottom). Within each box, the symbols represent the cost-effective strategy (no screening, screening once, screening every 1, 2, or 5 years, with starting at age ranging from 40 to 75 years, in 5-year increments) for each of the nine cohorts. For example, with the low discount rate assumption, for RRs of 1-3.55, no screening is the cost-effective strategy. For RRs of 5.71-6.7, screening once at age 60 years is cost-effective. At RR 12.33, annual screening starting at age 50 years is cost-effective, and at RR 28, annual screening starting at age 40 years is cost-effective. The right-hand column shows the same analysis using a WTP threshold of $200K USD/QALY gained. PDAC, pancreatic ductal adenocarcinoma; QALY, quality-adjusted life year; RR, relative risk; USD, US dollars; WTP, willingness-to-pay.

FIG 4.

Deterministic sensitivity analyses on the cost-effective strategy at a WTP threshold of $100K USD and $200K USD (women). This figure shows the cost-effective screening strategy for each cohort defined by the RR of PDAC (along the top), each uncertainty scenario (along the left), and each assumption about the WTP for a quality-adjusted year of life gained (along the bottom). Within each box, the symbols represent the cost-effective strategy (no screening, screening once, screening every 1, 2, or 5 years, starting at age ranging from 40 to 75 years, in 5-year increments) for each of the nine cohorts. The right-hand column shows the same analysis using a WTP threshold of $200K USD/QALY gained. PDAC, pancreatic ductal adenocarcinoma; QALY, quality-adjusted life year; RR, relative risk; USD, US dollars; WTP, willingness-to-pay.

We also tested alternative WTP thresholds. At a threshold of $200K USD/QALY gained, once-only screening at age 60 years was cost-effective among men with RR between 5.71 and 6.7, but even at this higher threshold, women with RR < 12.33 did not achieve sufficient benefit to justify screening. These results are summarized in Figures 3 and 4.

DISCUSSION

We found that given currently available screening modalities, a staged MRI/EUS screening approach could be cost-effective for the populations at highest risk for PDAC. However, for populations at moderate risk (RR, 5-10, represented here by ATM, BRCA2, and TP53), screening was only cost-effective for men at higher WTP thresholds (eg, $200K USD/QALY) and with one-time screening. CAPS has proposed annual or every-other-year screening starting around age 50 years for these moderate-risk patients.²⁶ In our analysis, however, these approaches did not achieve the highest life expectancy gains, and they were generally not cost-effective.

Several groups have considered the cost-effectiveness of screening for pancreatic cancer with MRI and/or EUS approaches. A cost analysis of MRI and EUS screening found that screening could be afforded for populations with at least 8-fold risk of PDAC.⁷⁹ One modeling study found that MRI-based screening every 3 years starting at age 40 years could be cost-effective for individuals with 5-fold risk of PDAC, as could EUS-based screening on the same schedule for those with 20-fold risk (both as compared with no screening).⁸⁰ However, a similar analysis of EUS screening of patients with 20-fold risk of PDAC found that screening led to a reduction in QALY gained because of the mortality and morbidity of pancreatectomy for false-positive screening results.⁸¹ A real-world prospective screening program in Denmark for high-risk patients (CDKN2A, STK11, PRSS1, and familial pancreatic cancer with 6.4×-32× population risk) showed that annual EUS identified two cancers from a cohort of 40 first-degree relatives and computed a corresponding ICER of $47,867 USD/QALY.⁴⁰ Relative to these studies, our analysis differs by including a wider range of screening options, by assuming that EUS is used as a follow-up test for positive MRI findings, and by conducting a more detailed cost analysis. In addition, our estimates of imaging sensitivity and specificity are slightly less favorable than some previous analyses.

We included strategies with screening frequencies ranging from once to annually. Strategies with screening every 2 years or every 5 years were in some cases the most effective strategies, if considering only the number of cancers averted and QALY gained for the cohort, but the high cost of these strategies led to them not being cost-effective. Similarly, strategies that initiate screening at younger ages had lower QALY gained than those that started at older ages, partially because of the increased number of surgeries with attendant mortality risk and modest life expectancy gains (sensitivity analysis showing results for a surgical mortality set to 0% can be reviewed in Appendix Fig A3).

We observed that in high-risk populations, life expectancy benefits may be incurred from false-positive screening results, because of reductions in cancer risk despite increased mortality and morbidity from surgical interventions. In our analysis as in real-world practice, if MRI/EUS screening flags a potential high-risk cyst, surgery is recommended. If that screening result is incorrect, the affected individual would still likely incur a risk mitigation benefit from resection, as less pancreatic tissue would be present for PDAC development (in our model, this is simulated by removal of the patient's unseen cancer precursors). This circumstance creates the seemingly paradoxical result that higher specificity of the screening tests can lower the life expectancy benefits by eliminating this avenue for risk mitigation. In our analysis, we did take into account surgical mortality, which offsets this effect to a degree; however, we did not specifically apply any postsurgical quality-of-life disutility. Clinically, postpancreatectomy patients have a higher risk of diabetes and other long-term complications^82,83; further research is needed to enable incorporation of these risks and corresponding disutilities into analyses such as ours. One previous modeling analysis⁸¹ proposed a higher surgical mortality rate (5%) and significant mortality and quality-of-life impacts from the development of diabetes mellitus; these assumptions significantly lowered the value of screening in their analysis.

Additional limitations of our analysis merit mention. Our results are based on heterogenous information sources, rather than from a specific clinical trial. We assume perfect screening availability and adherence. We do not capture the costs nor uncertainty of genetic testing because our simulation begins with the population's genetic predisposition already known. Costs of care are estimated on the basis of Medicare populations, which may not be applicable to younger populations. We have attempted to address some of the limitations of data availability through calibration and sensitivity analyses; as information changes and practice evolves, the results here will also need to evolve to reflect those changes. Finally, this analysis assumes a staged MRI/EUS screening approach. If alternative screening modalities become available, particularly a serologic biomarker, optimal strategies—from clinical effectiveness and cost-effectiveness standpoints—would likely change.

In conclusion, our analysis shows that MRI/EUS screening for pancreatic cancer can be cost-effective for high-risk populations. Patients with moderately increased risk of PDAC, such as those with ATM, BRCA2, or TP53, may experience clinical benefit, but with a high cost relative to this benefit. Investigation of more effective or lower-cost screening modalities may make screening more beneficial for these groups.

APPENDIX

TABLE A1.

Efficient Frontier Calculations for STK11 Men

Screening Strategy	Incremental Cost ($), USD	Incremental QALY Gained, Years	ICER
No screening	—	—	—
Once at 75 years	—	—	Weakly dominated
Once at 70 years	—	—	Weakly dominated
Every 5 years starting at 70 years	—	—	Strongly dominated
Once at 65 years	—	—	Weakly dominated
Every 2 years starting at 70 years	—	—	Strongly dominated
Once at 60 years	—	—	Weakly dominated
Annual starting at age 70 years	—	—	Strongly dominated
Every 5 years starting at 65 years	—	—	Strongly dominated
Once at 40 years	—	—	Weakly dominated
Once at 55 years	—	—	Weakly dominated
Once at 45 years	—	—	Weakly dominated
Once at 50 years	—	—	Weakly dominated
Every 2 years starting at 65 years	—	—	Weakly dominated
Every 5 years starting at 60 years	—	—	Strongly dominated
Annual starting at age 65 years	—	—	Weakly dominated
Every 5 years starting at 55 years	—	—	Strongly dominated
Every 2 years starting at 60 years	—	—	Weakly dominated
Every 5 years starting at 50 years	—	—	Strongly dominated
Annual starting at age 60 years	—	—	Weakly dominated
Every 2 years starting at 55 years	—	—	Weakly dominated
Every 5 years starting at 45 years	—	—	Strongly dominated
Every 5 years starting at 40 years	—	—	Strongly dominated
Every 2 years starting at 50 years	—	—	Strongly dominated
Annual starting at age 55 years	—	—	Weakly dominated
Every 2 years starting at 45 years	—	—	Strongly dominated
Every 2 years starting at 40 years	—	—	Strongly dominated
Annual starting at age 50 years	—	—	Weakly dominated
Annual starting at age 45 years	17,419	0.59	29,000
Annual starting at age 40 years	22,178	0.66	69,000

NOTE. A strategy that was both less effective and more expensive than another strategy was considered strongly dominated by the other strategy. A strategy that was either less effective or more expensive such that it was not on the efficient frontier was considered weakly dominated.

Abbreviations: ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life year; USD, US dollars.

FIG A1.

Efficient frontiers for all cohorts. QALY, quality-adjusted life year, US dollars.

FIG A2.

Deterministic sensitivity analysis. ICER, incremental cost-effectiveness ratio; RR, relative risk; USD, US dollars.

FIG A3.

Sensitivity analysis on surgical mortality. USD, US dollars; WTP, willingness-to-pay.

AUTHOR CONTRIBUTIONS

Conception and design: Mary Linton B. Peters, Andrew Eckel, Barak Davidi, Pari V. Pandharipande

Financial support: Mary Linton B. Peters

Collection and assembly of data: Mary Linton B. Peters, Andrew Eckel, Claudia L. Seguin, Barak Davidi, Pari V. Pandharipande

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Cost-Effectiveness Analysis of Screening for Pancreatic Cancer Among High-Risk Populations

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/op/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).