Model-based screening for pancreatic cancer in Sweden

Figures & data

Figure 1. Survival of pancreatic cancer in Sweden stratified by stage I to IV [1].

Table 1. Variables used in the cost estimation.

Loss of household production			Loss of production
Sex, age	Male, 50–54 years	Female, 50–54 years	Sex, age	Male, 50–54 years	Female, 50–54 years
Incidence	2	2	Incidence	2	2
Total number of lost days	250	264	Employment-to-population ratio	0.879	0.814
Hours per day, decimal	3.03	3.08	Mean wage (gross), SEK	487,300	390,000
Total number of lost hours	758	813	Social fees	0.43285	0.43285
Mean wage (gross), SEK	487,300	390,000	Survival rate	0.916	0.946
Weeks	52	52	Lost production time, years	15	15
Mean numbers of work hours in a week	32.5	28.5	Correcting and discounting
Final tax, SEK	0.31	0.272	Growth factor	0.024
Disability grade	0.4	0.4	Discount rate	0.03
Survival rate	0.527	0.645

SEK: Swedish krona. Examples of variables used in the cost estimation. In this example male and females between 50 and 54 years with localized disease are used.

Table 2. Cost items included in diagnostics of false positive screening cases, number per patient and cost in SEK [15,31–33].

Cost item	N per patient	Cost per unit (SEK)
MDTM	0.88	4053
US abdomen	0.09	1854
CT abdomen	1	2163
CT pancreas	1	3090
CT thorax	1	2163
MRI	0.036	3090
ERCP	0.88	6114
FDG-PET	0.15	12,000
US-FNA	0.59	2575

MDTM: multidisciplinary team meeting; US: ultrasound; CT: computed tomography; MRI: magnetic resonance imaging; ERCP: endoscopic retrograde cholangiopancreatography; FDG-PET: fluorodeoxyglucose-positron emission tomography; US-FNA: ultrasound guided-fine needle aspiration.

Figure 2. Positive and negative predictive values (PPV and NPV) for a blood test used to detect pancreatic cancer among different populations. The solid lines represent PPV (y-axis to the left) and the dotted lines represent NPV (y-axis to the right) when the test has a specificity of 0.80, 0.85, 0.90, 0.95 and 0.99.

Figure 3. ICER of a blood test used to detect pancreatic cancer among different populations. The y-axis starts at EUR -50,000 as some ICERs were negative. The ICER for the general population and smokers continued over EUR 400,000 and are not visualized. Willingness to pay at EUR 100,000 is marked.

Table 3. Variables associated with the change in the specificity and sensitivity of the test.

General population, 50–79 years
Sensitivity and specificity	99 %	95 %	90 %	85%	80%
PPV (%)	3	0.6	0.3	0.2	0.1
NPV (%)	99.99	99.99	99.99	99.99	99.99
Accuracy (%)	99	95	90	85	80
FPO (N)	33,942	169,712	339,324	509,135	678,847
Extra Cost (EUR)	66,054,000	330,272,000	660,349,000	990,816,000	1,321,088,000
Smokers 50–79, years
Sensitivity and specificity	99 %	95 %	90 %	85%	80%
PPV (%)	3	0.6	0.3	0.2	0.1
NPV (%)	99.99	99.99	99.99	99.99	99.99
Accuracy (%)	99	95	90	85	80
FPO (N)	4169	20,843	41,686	62,530	88,584
Extra Cost (EUR)	8,112,000	40,562,000	81,125,000	121,687,000	172,390,000
Familial and hereditary pancreatic cancer, 40–79 years
Sensitivity and specificity	99 %	95 %	90 %	85%	80%
PPV (%)	63	25	13	9	6
NPV (%)	99.98	99.91	99.81	99.70	99.58
Accuracy (%)	99	95	90	85	80
FPO (N)	64	322	645	967	1290
Extra Cost (EUR)	125,000	627,000	1,255,000	1,882,000	2,509,000
New-onset diabetes, 50–79 years
Sensitivity and specificity	99 %	95 %	90 %	85%	80%
PPV (%)	40	11	6	4	3
NPV (%)	99.99	99.97	99.93	99.88	99.83
Accuracy (%)	99	95	90	85	80
FPO (N)	305	1524	3048	4572	6095
Extra Cost (EUR)	593,000	2,966,000	5,931,000	8,897,000	11,862,000
Early symptoms, 40–79 years
Sensitivity and specificity	99 %	95 %	90 %	85%	80%
PPV (%)	75	37	22	15	11
NPV (%)	99.97	99.84	99.66	99.46	99.23
Accuracy (%)	99	95	90	85	80
FPO (N)	325	1624	3248	4873	6497
Extra cost (EUR)	632,000	3,161,000	6,321,000	9,482,000	12,643,000

PPV: positive predictive value, in percent; NPV: negative predictive value, in percent; FPO: false positive outcome, in total numbers. The rows stand for test specificity and sensitivity.

Table 4. The pretest and posttest probability for a hypothetical test with sensitivity and specificity both set at 90% or 99%.

Incidence in screened population	Pretest probability	Posttest probability 90%	Posttest probability 99%
1:10,000^a	0.0001	0.0009	0.001
1:3000^b	0.0003	0.003	0.03
1:1250	0.0008	0.007	0.07
1:300	0.003	0.03	0.20
1:100^c	0.01	0.08	0.50
1:60^d	0.017	0.13	0.63
1:33^e	0.03	0.22	0.76
1:2^f	0.50	0.90	0.99

^aApproximated incidence of pancreatic cancer in the general population in Sweden. ^bApproximated incidence of pancreatic cancer among smokers between 50 and 79 years in Sweden. ^cApproximated incidence of pancreatic cancer in individuals with new-onset diabetes between 50 and 79 years in Sweden. ^dApproximated incidence of pancreatic cancer in individuals with familial/hereditary risk between 40 and 79 years in Sweden. ^eApproximated incidence of pancreatic cancer in persons with early symptoms suggestive of pancreatic cancer. ^fThe incidence of pancreatic cancer in persons with Peutz–Jegher syndrome.

Table 5. Screening of new-onset diabetes patients between 50 and 79 years with a hypothetical test with sensitivity and specificity both at 80% or 99%.

80%	Patients with pancreatic cancer	Patients without pancreatic cancer
Positive test	162	6095
Negative test	41	24,381
All	203	30,476
99%	Patients with pancreatic cancer	Patients without pancreatic cancer
Positive test	200	305
Negative test	2	30,172
All	202	30,477

Regionala cancercentrum i samverkan. Kvalitetsregister för tumörer i pankreas och periampullärt. Årsrapport nationellt kvalitetsregister. 2020. Available from: https://cancercentrum.se/samverkan/.

 Google Scholar

Tingstedt B, Andersson E, Flink A, et al. Pancreatic cancer, healthcare cost, and loss of productivity: a register-based approach. World J Surg. 2011;35(10):2298–2305.

 PubMed Web of Science ®Google Scholar