Defining Value for Pancreatic Surgery in Early-Stage Pancreatic Cancer

Key Points

Question

Is the regionalization of surgical care for resectable pancreatic cancer at high-volume surgical centers and National Cancer Institute–designated cancer centers associated with improved survival, health care costs, and/or value (ie, survival relative to costs)?

Findings

In this retrospective analysis of 2786 patients with stages I to II pancreatic adenocarcinoma who underwent pancreatic resection, although high-volume centers were associated with greater overall survival, high-volume and National Cancer Institute–designated centers were not associated with improved health care costs or value.

Meaning

Targeted measures are needed at high-volume and National Cancer Institute–designated centers to enhance pancreatic surgery value, especially considering they treat a significant fraction of patients with pancreatic cancer.

This cohort study identifies patient and hospital characteristics associated with improved overall survival, decreased costs, and greater value among patients with pancreatic cancer undergoing curative resection.

Abstract

Importance

Value-based care is increasingly important, with rising health care costs and advances in cancer treatment leading to greater survival for patients with cancer. Regionalization of surgical care for pancreatic cancer has been extensively studied as a strategy to improve perioperative outcomes, but investigation of long-term outcomes relative to health care costs (ie, value) is lacking.

Objective

To identify patient and hospital characteristics associated with improved overall survival, decreased costs, and greater value among patients with pancreatic cancer undergoing curative resection.

Design, Setting, and Participants

This retrospective cohort study identified 2786 patients with stages I to II pancreatic cancer who underwent pancreatic resection at 157 hospitals from January 1, 2004, through December 31, 2012. The study used the California Cancer Registry, which collects data from all California residents newly diagnosed with cancer, linked to the Office of Statewide Health Planning and Development database, which collects administrative data from all California licensed hospitals. Data were analyzed from November 11, 2017, through September 4, 2018.

Exposures

Pancreatic resection at high-volume and/or National Cancer Institute (NCI)–designated cancer centers.

Main Outcomes and Measures

The primary outcomes were overall survival, surgical hospitalization costs, and value. High value was defined as the fourth quintile or higher for survival and the second quintile or less for costs. Costs were calculated from charges using cost-charge ratios and adjusted for geographic variation and inflation. Multivariable regression models were used to determine factors associated with overall survival, costs, and high value.

Results

Among the 2786 patients included (1394 [50.0%] male; mean [SD] age, 67.0 [10.7] years), postoperative chemotherapy (adjusted hazard ratio [aHR], 0.71; 95% CI, 0.64-0.79; P < .001) and high-volume centers (aHR, 0.78; 95% CI, 0.61-0.99; P = .04) were associated with greater overall survival. Higher Elixhauser comorbidity index scores (estimate, 0.006; 95% CI, 0.003-0.008), complications (estimate, 0.22; 95% CI, 0.17-0.27), readmissions (estimate, 0.34; 95% CI, 0.29-0.39), and longer lengths of stay (estimate, 0.03; 95% CI, 0.03-0.04) were associated with higher costs (P < .001), whereas postoperative chemotherapy was associated with lower costs (estimate, −0.06; 95% CI, −0.11 to −0.02; P = .006). National Cancer Institute–designated and high-volume centers were not associated with costs. Although grades III and IV tumors (odds ratio [OR], 0.65; 95% CI, 0.39-0.91; P = .001), T3 category disease (OR, 0.71; 95% CI, 0.46-0.95; P = .005), complications (OR, 0.68; 95% CI, 0.49-0.86; P < .001), readmissions (OR, 0.64; 95% CI, 0.44-0.84; P < .001), and length of stay (OR, 0.82; 95% CI, 0.78-0.85; P < .001) were inversely associated with high-value care, NCI designation (OR, 1.07; 95% CI, 0.66-1.49; P = .74) and high-volume centers (OR, 1.08; 95% CI, 0.54-1.61; P = .07) were not.

Conclusions and Relevance

In this study, high-value care was associated with important patient characteristics and postoperative outcomes. However, NCI-designated and high-volume centers were not associated with greater value. These data suggest that targeted measures to enhance value may be needed in these centers.

Introduction

Value in cancer care has become increasingly important in the age of ever-rising health care costs and increasing incidence of cancer. Cancer-related health care costs in the United States are expected to exceed $170 billion by 2020.¹ Meanwhile, the United States spends more than $400 billion annually to care for patients undergoing surgery, and this spending is projected to constitute 7% of the US gross domestic product by 2025.^2,3,4 Unfortunately, the rising costs of cancer and surgical care coincide with the rising incidence of cancer in the population. Hence, studies regarding the value of health care are essential. Value is defined as quality relative to the cost of care.⁵ It is a useful metric in health care services and economics research because maintaining the quality of care provided while controlling cost is imperative.

Pancreatic cancer surgery, although rare, is often studied for this purpose owing to the complexity of pancreatic resection and relatively high morbidity and mortality. An estimated 55 440 adults in the United States will be diagnosed with pancreatic cancer this year, and an estimated 44 330 will die of the disease.⁶ Studies have demonstrated that as many as 40% of patients undergoing this type of procedure develop a postoperative complication.^7,8 For pancreatic resections in particular, a consistent and strong association between high surgical volume and improved perioperative outcomes has been demonstrated by numerous studies.^{9,10,11,12,13,14} More recent research has begun to delve into the nuances between volume and cost as well as the hospital-, economic-, and patient-based factors contributing to each.^15,16,17 However, to our knowledge, no research to date has examined value in pancreatic cancer surgery.

Recognizing the complexity of this issue and the intricate associations at play, we performed a population-based analysis of patients with early-stage pancreatic cancer who underwent curative resection. The purpose of this study was to identify patient and hospital characteristics associated with improved survival, decreased costs, and overall greater value. We hypothesized that high-volume pancreatic surgery centers and National Cancer Institute (NCI)–designated cancer centers would be associated with improved survival and potentially lower costs secondary to regionalization of care and center-specific processes, and thus greater value compared with low-volume pancreatic surgery centers and non–NCI-designated cancer centers.

Methods

We performed a retrospective analysis of patients diagnosed with resectable pancreatic adenocarcinoma from the California Cancer Registry (CCR) and Office of Statewide Hospital Planning and Development (OSHPD)–linked database from January 1, 2004, through December 31, 2012. The research protocol was approved by the institutional review board of the University of California, Davis, and the California Health and Human Services Agency Committee for the Protection of Human Subjects with a waiver of consent granted for use of deidentified data. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Study Population

We identified 8723 patients with stages I to II (T1-T3 N0-N1 M0) pancreatic adenocarcinoma based on International Classification of Diseases for Oncology, Third Revision, codes for site and histology (eTable in the Supplement).¹⁸ Patients who did not undergo a pancreatic resection (5207 [59.7%]) or with missing costs data (ie, Kaiser Foundation facilities, 730 [8.4%]) were excluded. The final cohort consisted of 2786 patients treated at 157 hospitals.

We abstracted patient demographic, clinicopathologic, and treatment data from CCR, including patient age, sex, race, tumor grade, tumor and node categories, composite American Joint Committee on Cancer (seventh edition) stage, chemotherapy, and radiotherapy. Pancreatic surgical resections were identified from the linked OSHPD data files based on International Classification of Diseases, Ninth Revision (ICD-9) codes 525, 5251, 5252, 5253, 5259, 526, and 527. The Elixhauser comorbidity index was used to measure risk associated with comorbid health conditions and created using ICD-9 diagnostic codes (index scores range from −11 to 62, with higher scores indicating more comorbidities).^19,20,21 High-volume pancreatic surgery centers were defined as centers performing at least 20 pancreatic resections annually based on previous research showing improved mortality at this cutoff.^22,23,24 Of the 157 hospitals, only 8 were high-volume surgical centers.

Postoperative complications were identified from ICD-9 codes for the index hospitalization and subsequent hospital readmissions within 30 days of the date of surgery. Complications included pulmonary failure, pneumonia, myocardial infarction, cardiac arrest, pulmonary embolism, deep vein thrombosis, gastrointestinal tract hemorrhage, surgical site and organ space infections, systemic shock, acute kidney injury, delayed gastric emptying, gastrointestinal and/or enterocutaneous fistula, and bile leak.¹⁵ We were unable to include pancreatic fistula because ICD-9 has no specific code for this complication. We abstracted length of stay, 30-day readmissions, in-hospital death, and costs from the OSHPD-linked data files.

Costs were estimated from charges for the index surgical hospitalization and all readmissions within 30 days of discharge. Charges were multiplied by the ratio of hospital-specific cost to charge, adjusted for geographic variation using the wage index and adjusted for inflation using the Personal Health Care Price Index.²⁵ Survival was defined as months alive from the date of pancreatic cancer diagnosis. Only 464 patients (16.7%) were censored; these patients were included owing to their extended follow-up (median, 56 months; range, 23-131 months; interquartile range [IQR], 37-85 months). High-value care was defined as the fourth quintile or higher for survival (≥26 months) and second quintile or less for costs (≤$40 674). Low-value care was defined as the second quintile or less for survival (≤15 months) and the fourth quintile or higher for costs (≥$53 971).

Statistical Analysis

Data were analyzed from November 11, 2017, through September 4, 2018. Patient demographic, clinicopathologic, and treatment characteristics were presented as means with SDs, medians with IQRs, and frequencies with percentages, as appropriate. Patients were divided into quintiles based on survival and costs. We used χ² tests, analysis of variance, and Kruskal-Wallis tests to compare demographic, clinicopathologic, and treatment characteristics between survival and cost quintiles and among patients with a high-, intermediate-, and low-value care designation. Multivariable Cox proportional hazards models were used to identify factors associated with overall survival. Multivariable generalized linear models were used to determine characteristics associated with costs, assuming a gamma distribution and log link (determined appropriate using the modified Parks and link tests).^15,26 Logistic regression was used to identify characteristics associated with high-value care. Robust standard errors adjusted for clustering of patients at the same hospital were calculated for all multivariable analyses. All tests were 2 sided. Statistical significance was set at P < .05. Analyses were performed using SAS, version 9.4 (SAS Institute Inc), and Stata, version 13 (StataCorp).

Owing to potential collinearity for NCI-designated cancer centers and high-volume surgical centers, in addition to the comprehensive model with both variables included, we performed separate multivariable models for NCI-designated centers and high-volume hospitals. Because the results were similar for the individual and comprehensive models for overall survival, costs, and value, we selected to present the comprehensive model.

Results

Table 1 presents patient demographic, clinicopathologic, and treatment characteristics of the included 2786 patients with pancreatic cancer who underwent surgical resection. The mean (SD) age was 67.0 (10.7) years, 1394 patients (50.0%) were male, 1392 (50.0%) were female, and 2308 (82.8%) were white. Most tumors were moderately differentiated (1373 [49.3%]) or poorly differentiated or undifferentiated (949 [34.1%]). The T3 category (2147 [77.1%]) and presence of nodal disease (1725 [61.9%]) were common. A total of 1238 patients (44.4%) underwent postoperative chemotherapy, and 874 (31.4%) underwent radiotherapy. Most patients underwent resection of pancreatic head tumors (2207 [79.2%]). One thousand thirty-seven patients (37.2%) experienced at least 1 complication, and 565 (20.3%) were readmitted within 30 days of discharge. The median length of stay was 12 days (IQR, 9-18 days). The mean (SD) costs for the index surgical hospitalization and 30-day readmissions were $60 939 ($56 779), with a median of $46 697 (IQR, $32 815-$68 765). Median follow-up time with survival was 20 months (IQR, 10-35 months). Most patients died of pancreatic cancer (2020 of 2731 [74.0%]). A total of 1428 patients (51.3%) were treated at a teaching hospital, 1297 (46.6%) were treated at a high-volume pancreatic surgery center, and 1127 (40.5%) were treated at an NCI-designated cancer center.

Table 1. Demographic, Clinicopathologic, and Treatment Characteristics of Patients With Pancreatic Cancer Who Underwent Pancreatic Resection.

Characteristic	Data (n = 2786)a
Age, mean (SD), y	67.0 (10.7)
Sex
Male	1394 (50.0)
Female	1392 (50.0)
Race
White	2308 (82.8)
Black	152 (5.5)
Asian/Pacific Islander	313 (11.2)
Other/unknown	13 (0.5)
Elixhauser comorbidity index score, mean (SD)b	13.9 (8.9)
Tumor grade
Well differentiated	295 (10.6)
Moderately differentiated	1373 (49.3)
Poorly differentiated or undifferentiated	949 (34.1)
Unknown	169 (6.1)
T category
T1	204 (7.3)
T2	435 (15.6)
T3	2147 (77.1)
N category
N0	1054 (37.8)
N1	1725 (61.9)
NX	7 (0.3)
Stage
I	399 (14.3)
II	2387 (85.7)
Chemotherapy
None	1070 (38.4)
Preoperative	89 (3.2)
Postoperative	1238 (44.4)
Preoperative and postoperative	50 (1.8)
Unknown	339 (12.2)
Radiotherapy	874 (31.4)
Pancreatic resection
Pancreaticoduodenectomy	2207 (79.2)
Distal pancreatectomy	387 (13.9)
Total pancreatectomy	98 (3.5)
Other	94 (3.4)
Complication within 30 d	1037 (37.2)
Readmission within 30 d	565 (20.3)
Length of stay, median (IQR), d	12 (9-18)
Surgery hospitalization costs, mean (SD), $c	60 939 (56 779)
Follow-up, median (IQR), mo	20 (10-35)
Vital status
Death due to pancreatic cancer	2020 (72.5)
Death due to other cause	247 (8.9)
Death due to unknown cause	55 (2.0)
Alive	464 (16.7)
High-volume hospital	1297 (46.6)
NCI-designated cancer center	1127 (40.5)
Teaching hospital	1428 (51.3)

Abbreviations: IQR, interquartile range; NCI, National Cancer Institute.

^aUnless otherwise indicated, data are expressed as number (percentage) of patients. Percentages have been rounded and may not total 100.

^bScores range from −11 to 62, with higher scores indicating a greater number of comorbidities.

^cIncludes the index surgical hospitalization and readmissions within 30 days.

Median survival for high-volume centers was 23 months (IQR, 11-46 months) compared with 18 months (IQR, 10-35 months) for low-volume centers (P < .001). Patients were more frequently treated with chemotherapy at high-volume vs low-volume centers (879 [67.8%] vs 824 [55.3%]; P < .001) and NCI- vs non–NCI-designated cancer centers (760 [67.4%] vs 943 [56.8%]; P < .001). Among those who underwent chemotherapy with a known regimen (n = 1473), patients who underwent surgery at high-volume centers were more frequently treated with multiagent chemotherapy compared with patients treated at low-volume centers (355 [46.2%] vs 282 [40.1%]; P = .02), whereas there was no difference in the use of multiagent chemotherapy among patients who underwent surgery at NCI- and non–NCI-designated cancer centers (279 [42.3%] vs 358 [44.0%]; P = .50).

Table 2 and Table 3 present univariate analysis of patient demographics and clinicopathologic and treatment differences by survival and cost quintiles. Older age (mean [SD], 69.3 [10.3] years; P < .001), male sex (337 of 615 [54.8]; P = .03), higher Elixhauser comorbidity index score (mean [SD], 16.7 [10.1]; P < .001), and higher rates of T3 category disease (511 of 615 [83.1%]; P < .001), nodal disease (438 of 615 [71.2%]; P < .001), stage II disease (561 of 615 [91.2%]; P < .001), and poorly differentiated or undifferentiated tumors (278 of 615 [45.2%]; P < .001) were associated with shorter survival. Postoperative chemotherapy (292 of 534 [54.7%]; P < .001), operations performed at high-volume surgical centers (277 of 534 [51.9%]; P < .001), and lower complication (177 of 534 [33.1%]; P < .001) and readmission (99 of 534 [18.5%]; P < .001) rates were associated with longer survival. Similarly, older age (mean [SD], 68.0 [10.0] years; P < .001), higher Elixhauser comorbidity index score (mean [SD], 17.9 [10.9]; P < .001), and higher rates of T3 disease (457 of 557 [82.0%]; P < .001) were associated with higher costs, whereas distal pancreatectomies (197 of 557 [35.4%]; P < .001) were associated with lower costs. In addition, higher rates of complications (416 of 557 [74.7%]; P < .001), readmission (226 of 557 [40.6%]; P < .001), and in-hospital death (63 of 557 [11.3%]; P < .001) were associated with higher costs.

Table 2. Patient Demographic, Clinicopathologic, and Treatment Characteristics by Survival Quintile.

Characteristic	Survival Quintilea					P Value
	Low (n = 615)	Low to Intermediate (n = 511)	Intermediate (n = 556)	Intermediate to High (n = 534)	High (n = 570)
Age, mean (SD)	69.3 (10.3)	67.2 (10.1)	65.8 (10.9)	66.9 (10.0)	65.4 (11.6)	<.001b
Male	337 (54.8)	249 (48.7)	282 (50.7)	267 (50.0)	259 (45.4)	.03c
Race
White	511 (83.1)	414 (81.0)	472 (84.9)	434 (81.3)	477 (83.7)	.61d
Black	35 (5.7)	25 (4.9)	24 (4.3)	37 (6.9)	31 (5.4)
Asian/Pacific Islander	68 (11.1)	69 (13.5)	58 (10.4)	59 (11.0)	59 (10.4)
Other/unknown	1 (0.2)	3 (0.6)	2 (0.4)	4 (0.7)	3 (0.5)
Socioeconomic status
Low (first quintile)	137 (22.3)	88 (17.2)	116 (20.9)	91 (17.0)	99 (17.4)	.32c
Low to middle (second quintile)	120 (19.5)	104 (20.4)	110 (19.8)	97 (18.2)	96 (16.8)
Middle (third quintile)	121 (19.7)	105 (20.5)	110 (19.8)	100 (18.7)	109 (19.1)
Middle to high (fourth quintile)	113 (18.4)	95 (18.6)	101 (18.2)	116 (21.7)	120 (21.1)
High (fifth quintile)	124 (20.2)	119 (23.3)	119 (21.4)	130 (24.3)	146 (25.6)
Elixhauser comorbidity index score, mean (SD)e	16.7 (10.1)	14.4 (9.1)	13.6 (8.4)	12.6 (8.1)	12.2 (7.9)	<.001b
T category
T1	24 (3.9)	18 (3.5)	28 (5.0)	59 (11.0)	75 (13.2)	<.001c
T2	80 (13.0)	70 (13.7)	80 (14.4)	88 (16.5)	117 (20.5)
T3	511 (83.1)	423 (82.8)	448 (80.6)	387 (72.5)	378 (66.3)
N category
N0	176 (28.6)	150 (29.4)	196 (35.3)	219 (41.0)	313 (54.9)	<.001d
N1	438 (71.2)	361 (70.6)	358 (64.4)	314 (58.8)	254 (44.6)
NX	1 (0.2)	0	2 (0.4)	1 (0.2)	5 (0.5)
Stage
I	54 (8.8)	49 (9.6)	53 (9.5)	92 (17.2)	151 (26.5)	<.001c
II	561 (91.2)	462 (90.4)	503 (90.5)	442 (82.8)	419 (73.5)
Grade
Well differentiated	40 (6.5)	29 (5.7)	58 (10.4)	71 (13.3)	97 (17.0)	<.001c
Moderately differentiated	274 (44.6)	246 (48.1)	294 (52.9)	296 (55.4)	263 (46.1)
Poorly differentiated or undifferentiated	278 (45.2)	207 (40.5)	183 (32.9)	129 (24.2)	152 (26.7)
Unknown	23 (3.7)	29 (5.7)	21 (3.8)	38 (7.1)	58 (10.2)
Pancreatic resection
Pancreaticoduodenectomy	491 (79.8)	402 (78.7)	456 (82.0)	412 (77.2)	446 (78.2)	.06c
Distal pancreatectomy	80 (13.0)	72 (14.1)	61 (11.0)	81 (15.2)	93 (16.3)
Total pancreatectomy	28 (4.6)	23 (4.5)	20 (3.6)	18 (3.4)	9 (1.6)
Other pancreatectomy	16 (2.6)	14 (2.7)	19 (3.4)	23 (4.3)	22 (3.9)
Chemotherapy
None	378 (61.5)	184 (36.0)	153 (27.5)	151 (28.3)	204 (35.8)	<.001 c
Preoperative	9 (1.5)	20 (3.9)	20 (3.6)	20 (3.7)	20 (3.5)
Postoperative	175 (28.5)	226 (44.2)	292 (52.5)	292 (54.7)	253 (44.4)
Sequence unknown/other	53 (8.6)	81 (15.9)	91 (16.4)	71 (13.3)	93 (16.3)
Radiotherapy	106 (17.2)	188 (36.8)	225 (40.5)	185 (34.6)	170 (29.8)	<.001c
NCI-designated cancer center	218 (35.4)	191 (37.4)	234 (42.1)	242 (45.3)	242 (42.5)	.004c
Teaching hospital	286 (46.5)	244 (47.7)	290 (52.2)	297 (55.6)	311 (54.6)	.005c
High-volume centerf	244 (39.7)	221 (43.2)	255 (45.9)	277 (51.9)	300 (52.6)	<.001c
Costs, mean (SD), $g	72 516 (69 025)	56 042 (52 301)	50 306 (29 487)	50 491 (40 577)	49 356 (49 384)	<.001h
Length of stay, median (IQR), d	15 (10-23)	13 (9-18)	12 (9-17)	11 (9-16)	11 (9-16)	<.001h
Complication within 30 d	324 (52.7)	196 (38.4)	175 (31.5)	177 (33.1)	165 (28.9)	<.001c
Readmission within 30 d	162 (26.3)	101 (19.8)	110 (19.8)	99 (18.5)	93 (16.3)	<.001c

Abbreviations: IQR, interquartile range; NCI, National Cancer Institute.

^aUnless otherwise indicated, data are expressed as number (percentage) of patients. Percentages have been rounded and may not total 100. Low indicates 0 to 9 months; low to intermediate, 10 to 15 months; intermediate, 16 to 25 months; intermediate to high, 26 to 41 months; and high, 42 or more months.

^bCalculated using analysis of variance.

^cCalculated using χ² test.

^dCalculated using Fisher exact text.

^eScores range from −11 to 62, with higher scores indicating a greater number of comorbidities.

^fIndicates 20 or more cases annually.

^gIncludes the index surgical hospitalization and readmissions within 30 days.

^hCalculated using Kruskal-Wallis test.

Table 3. Patient Demographic, Clinicopathologic, and Treatment Characteristics by Surgical Hospital Cost Quintile.

Characteristic	Cost Quintilea					P Value
	Low (n = 557)	Low to Intermediate (n = 558)	Intermediate (n = 557)	Intermediate to High (n = 557)	High (n = 557)
Age, mean (SD), y	65.7 (11.3)	66.4 (10.7)	66.6 (10.4)	68.1 (10.7)	68.0 (10.0)	<.001b
Male	262 (47.0)	287 (51.4)	257 (46.1)	293 (52.6)	295 (53.0)	.06c
Race
White	468 (84.0)	459 (82.3)	452 (81.1)	472 (84.7)	457 (82.0)	.12d
Black	17 (3.1)	33 (5.9)	29 (5.2)	36 (6.5)	37 (6.6)
Asian/Pacific Islander	70 (12.6)	63 (11.3)	72 (12.9)	46 (8.3)	62 (11.1)
Other/unknown	2 (0.4)	3 (0.5)	4 (0.7)	3 (0.5)	1 (0.2)
Socioeconomic status
Low (first quintile)	84 (15.1)	100 (17.9)	116 (20.8)	104 (18.7)	127 (22.8)	.001c
Low to middle (second quintile)	79 (14.2)	111 (19.9)	112 (20.1)	116 (20.8)	109 (19.6)
Middle (third quintile)	123 (22.1)	100 (17.9)	110 (19.7)	112 (20.1)	100 (18.0)
Middle to high (fourth quintile)	112 (20.1)	104 (18.6)	103 (18.5)	109 (19.6)	117 (21.0)
High (fifth quintile)	159 (28.5)	143 (25.6)	116 (20.8)	116 (20.8)	104 (18.7)
Elixhauser comorbidity index, mean (SD)e	10.5 (6.0)	12.4 (7.6)	13.7 (8.1)	15.2 (9.5)	17.9 (10.9)	<.001b
T category
T1	61 (11.0)	29 (5.2)	39 (7.0)	41 (7.4)	34 (6.1)	<.001c
T2	101 (18.1)	108 (19.4)	79 (14.2)	81 (14.5)	66 (11.8)
T3	395 (70.9)	421 (75.4)	439 (78.8)	435 (78.1)	457 (82.0)
N category
N0	236 (42.4)	203 (36.4)	218 (39.1)	195 (35.0)	202 (36.3)	.04d
N1	319 (57.3)	354 (63.4)	339 (60.9)	358 (64.3)	355 (63.7)
NX	2 (0.4)	1 (0.2)	0	4 (0.7)	0
Stage
I	118 (21.2)	79 (14.2)	67 (12.0)	70 (12.6)	65 (11.7)	<.001c
II	439 (78.8)	479 (85.8)	490 (88.0)	487 (87.4)	492 (88.3)
Grade
Well differentiated	52 (9.3)	60 (10.8)	66 (11.8)	69 (12.4)	48 (8.6)	.41c
Moderately differentiated	284 (51.0)	275 (49.3)	276 (49.6)	264 (47.4)	274 (49.2)
Poorly differentiated or undifferentiated	186 (33.4)	197 (35.3)	189 (33.9)	189 (33.9)	188 (33.8)
Unknown	35 (6.3)	26 (4.7)	26 (4.7)	35 (6.3)	47 (8.4)
Pancreatic resection
Pancreaticoduodenectomy	310 (55.7)	459 (82.3)	480 (86.2)	474 (85.1)	484 (86.9)	<.001c
Distal pancreatectomy	197 (35.4)	68 (12.2)	35 (6.3)	49 (8.8)	38 (6.8)
Total pancreatectomy	17 (3.1)	7 (1.3)	26 (4.7)	24 (4.3)	24 (4.3)
Other pancreatectomy	33 (5.9)	24 (4.3)	16 (2.9)	10 (1.8)	11 (2.0)
Chemotherapy
None	204 (36.6)	165 (29.6)	193 (34.6)	205 (36.8)	303 (54.4)	<.001c
Preoperative	14 (2.5)	15 (2.7)	16 (2.9)	25 (4.5)	19 (3.4)
Postoperative	259 (46.5)	286 (51.3)	274 (49.2)	247 (44.3)	172 (30.9)
Sequence unknown/other	80 (14.4)	92 (16.5)	74 (13.3)	80 (14.4)	63 (11.3)
Radiotherapy	176 (31.6)	201 (36.0)	209 (37.5)	168 (30.2)	120 (21.5)	<.001c
NCI-designated cancer center	247 (44.3)	235 (42.1)	210 (37.7)	198 (35.5)	237 (42.5)	.01c
Teaching hospital	250 (44.9)	291 (52.2)	291 (52.2)	286 (51.3)	310 (55.7)	.008c
High-volume centerf	245 (44.0)	252 (45.2)	255 (45.8)	265 (47.6)	280 (50.3)	.25c
Survival, median (IQR), mo	26 (12-61)	24 (12-42)	22 (12-43)	17 (10-33)	13 (6-31)	<.001g
Length of stay, median (IQR), d	8 (7-10)	10 (9-13)	13 (10-15)	16 (11-20)	24 (18-34)	<.001g
Complication within 30 d	62 (11.1)	111 (19.9)	182 (32.7)	266 (47.8)	416 (74.7)	<.001c
Readmission within 30 d	36 (6.5)	54 (9.7)	103 (18.5)	146 (26.2)	226 (40.6)	<.001c
In-hospital death	8 (1.4)	7 (1.3)	7 (1.3)	19 (3.4)	63 (11.3)	<.001c

Abbreviations: IQR, interquartile range; NCI, National Cancer Institute.

^aUnless otherwise indicated, data are expressed as number (percentage) of patients. Percentages have been rounded and may not total 100. Low indicates less than $30 007; low to intermediate, $30 007 to $40 674; intermediate, $40 675 to $53 970; intermediate to high, $53 971 to $76 695; and high, greater than $76 995.

^bCalculated using analysis of variance.

^cCalculated using χ² test.

^dCalculated using Fisher exact text.

^eScores range from −11 to 62, with higher scores indicating a greater number of comorbidities.

^fIndicates 20 or more cases annually.

^gCalculated using Kruskal-Wallis test.

Univariate analysis of patient demographic, clinicopathologic, and treatment differences by value are presented in Table 4. Younger age (mean [SD], 65.7 [11.2] years; P < .001), female sex (284 of 524 [54.2%]; P = .008), lower Elixhauser comorbidity index score (mean [SD], 10.4 [5.9]; P < .001), lower rates of T3 category (331 of 524 [63.2%]; P < .001) and stage II (379 of 524 [72.3%]; P < .001) disease, absence of nodal disease (265 of 524 [50.6%]), lower rates of poorly differentiated or undifferentiated tumors (131 of 524 [25.0%]; P < .001), higher rates of distal pancreatectomy (137 of 524 [26.1%]; P < .001), and postoperative chemotherapy (262 of 524 [50.0%]; P < .001) were associated with high-value care. National Cancer Institute–designated cancer centers (247 of 524 [47.1%]; P = .001) and high-volume centers (272 of 524 [51.9%]; P = .02) were also associated with high-value care in this univariate analysis. Longer lengths of stay (median, 19 days [IQR, 14-28 days]; P < .001) and higher rates of complication (372 of 568 [65.5%]; P < .001), readmission (199 of 568 [35.0%]; P < .001), and in-hospital death (82 of 568 [14.4%]; P < .001) were associated with low-value care.

Table 4. Patient Demographic, Clinicopathologic, and Treatment Characteristics by Value.

	Value Levela			P Value
	Low (n = 568)	Intermediate (n = 1694)	High (n = 524)
Age, mean (SD), y	69.4 (10.1)	66.5 (10.6)	65.7 (11.2)	<.001b
Male	313 (55.1)	841 (49.6)	240 (45.8)	.008c
Race
White	464 (81.7)	1410 (83.2)	434 (82.8)	.52d
Black	37 (6.5)	88 (5.2)	27 (5.2)
Asian/Pacific Islander	67 (11.8)	186 (11.0)	60 (11.5)
Other/unknown	0	10 (0.6)	3 (0.6)
Socioeconomic status
Low (first quintile)	125 (22.0)	331 (19.5)	75 (14.3)	.001c
Low to middle (second quintile)	119 (21.0)	323 (19.1)	85 (16.2)
Middle (third quintile)	111 (19.5)	333 (19.7)	101 (19.3)
Middle to high (fourth quintile)	108 (19.0)	323 (19.1)	114 (21.8)
High (fifth quintile)	105 (18.5)	384 (22.7)	149 (28.4)
Elixhauser comorbidity index, mean (SD)e	17.7 (10.9)	13.8 (8.5)	10.4 (5.9)	<.001b
T category
T1	25 (4.4)	111 (6.6)	68 (13.0)	<.001c
T2	78 (13.7)	232 (13.7)	125 (23.9)
T3	465 (81.9)	1351 (79.8)	331 (63.2)
N category
N0	171 (30.1)	618 (36.5)	265 (50.6)	<.001d
N1	396 (69.7)	1072 (63.3)	257 (49.1)
NX	1 (0.2)	4 (0.2)	2 (0.4)
Stage
I	57 (10.0)	197 (11.6)	145 (27.7)	<.001c
II	511 (90.0)	1497 (88.4)	379 (72.3)
Grade
Well differentiated	41 (7.2)	180 (10.6)	74 (14.1)	<.001c
Moderately differentiated	275 (48.4)	820 (48.4)	278 (53.1)
Poorly differentiated or undifferentiated	218 (38.4)	600 (35.4)	131 (25.0)
Unknown	34 (6.0)	94 (5.5)	41 (7.8)
Pancreatic resection
Pancreaticoduodenectomy	473 (83.3)	1386 (81.8)	348 (66.4)	<.001c
Distal pancreatectomy	56 (9.9)	194 (11.5)	137 (26.1)
Total pancreatectomy	29 (5.1)	62 (3.7)	7 (1.3)
Other pancreatectomy	10 (1.8)	52 (3.1)	32 (6.1)
Chemotherapy
None	335 (59.0)	570 (33.6)	165 (31.5)	<.001c
Preoperative	20 (3.5)	53 (3.1)	16 (3.1)
Postoperative	154 (27.1)	822 (48.5)	262 (50.0)
Sequence unknown/other	59 (10.4)	249 (14.7)	81 (15.5)
Radiotherapy	115 (20.2)	596 (35.2)	163 (31.1)	<.001c
NCI-designated cancer center	209 (36.8)	671 (39.6)	247 (47.1)	.001c
Teaching hospital	290 (51.1)	853 (50.4)	285 (54.4)	.27c
High-volume centerf	265 (46.7)	760 (44.9)	272 (51.9)	.02c
Survival, median (IQR), mo	8 (4-11)	20 (12-34)	51 (33-105)	<.001g
Costs, mean (SD), $h	106 613 (79 218)	55 568 (46 219)	28 792 (7371)	<.001g
Length of stay, median (IQR), d	19 (14-28)	12 (9-17)	9 (8-11)	<.001g
Complication within 30 d	372 (65.5)	601 (35.5)	64 (12.2)	<.001c
Readmission within 30 d	199 (35.0)	321 (18.9)	45 (8.6)	<.001c
In-hospital death	82 (14.4)	22 (1.3)	0	<.001c

Abbreviations: IQR, interquartile range; NCI, National Cancer Institute.

^aUnless otherwise indicated, data are expressed as number (percentage) of patients. Percentages have been rounded and may not total 100. High value was defined as the fourth quintile or higher for survival (≥26 months) and second quintile or less for costs (≤$40 674). Low value was defined as the second quintile or lower for survival (≤15 months) and fourth quintile or higher for costs (≥$53 971).

^bCalculated using analysis of variance.

^cCalculated using χ² test.

^dCalculated using Fisher exact text.

^eScores range from −11 to 62, with higher scores indicating a greater number of comorbidities.

^fIndicates 20 or more cases annually.

^gCalculated using Kruskal-Wallis test.

^hIncludes the index surgical hospitalization and readmissions within 30 days.

Multivariable models for survival, costs, and high-value care are presented in Table 5. Older age (adjusted hazard ratio [aHR], 1.01; 95% CI, 1.01-1.02; P < .001), male sex (aHR, 0.91; 95% CI, 0.83-0.99; P = .02), higher Elixhauser comorbidity index score (aHR, 1.01; 95% CI, 1.01-1.02; P < .001), higher-grade tumors (aHR, 1.84; 95% CI, 1.57-2.16; P < .001), higher T (aHR, 1.45; 95% CI, 1.16-1.80; P < .001) and N (aHR, 1.47; 95% CI, 1.32-1.63; P < .001) categories, stage II disease (aHR, 1.27; 95% CI, 1.02-1.57; P = .03), total pancreatectomy (HR, 1.47; 95% CI, 1.18-1.82; P < .001), and postoperative complications (aHR, 1.18; 95% CI, 1.07-1.30; P = .001) were associated with shorter survival. Postoperative chemotherapy (aHR, 0.71; 95% CI, 0.64-0.79; P < .001) and high-volume surgical centers (aHR, 0.78; 95% CI, 0.61-0.99; P = .04) were associated with greater survival. Higher Elixhauser comorbidity index score (estimate, 0.006; 95% CI, 0.003-0.008; P < .001), postoperative complications (estimate, 0.22; 95% CI, 0.17-0.27; P < .001), readmissions (estimate, 0.34; 95% CI, 0.29-0.39; P < .001), and longer lengths of stay (estimate, 0.03; 95% CI, 0.03-0.04; P < .001) were associated with higher costs, whereas distal pancreatectomies (estimate, −0.21; 95% CI, −0.26 to −0.17; P < .001) and postoperative chemotherapy (estimate, −0.06; 95% CI, −0.11 to −0.02; P = .006) were associated with lower costs. Operations performed at NCI-designated cancer centers and high-volume centers were not associated with cost differences.

Table 5. Multivariable Models for Overall Survival, Surgical Hospitalization Costs, and High Value.

Variable	Survivala		Costsb		High Valuec
	aHR (95% CI)	P Value	Estimate (95% CI)	P Value	aOR (95% CI)	P Value
Age	1.01 (1.01 to 1.02)	<.001	0.000 (−0.001 to 0.001)	.90	0.99 (0.98 to 1.00)	.06
Sex
Male	1 [Reference]	NA	1 [Reference]	NA	1 [Reference]	NA
Female	0.91 (0.83 to 0.99)	.02	−0.02 (−0.04 to 0.01)	.17	1.10 (1.01 to 1.19)	.048
Race
White	1 [Reference]	NA	1 [Reference]	NA	1 [Reference]	NA
Asian/Pacific Islander	1.01 (0.88 to 1.16)	.88	−0.04 (−0.10 to 0.01)	.10	0.92 (0.42 to 1.41)	.73
Black	1.08 (0.90 to 1.31)	.40	0.02 (−0.04 to 0.09)	.44	1.02 (0.48 to 1.56)	.94
Other/unknown	0.84 (0.44 to 1.58)	.58	−0.10 (−0.29 to 0.10)	.34	1.11 (0.00 to 2.36)	.87
Socioeconomic status
Low	1 [Reference]	NA	1 [Reference]	NA	1 [Reference]	NA
Low to middle	1.01 (0.88 to 1.15)	.90	−0.01 (−0.05 to 0.03)	.69	0.87 (0.65 to 1.09)	.22
Middle	0.95 (0.83 to 1.09)	.46	−0.01 (−0.05 to 0.03)	.62	0.94 (0.77 to 1.11)	.47
Middle to high	0.92 (0.80 to 1.05)	.22	0.02 (−0.03 to 0.07)	.39	1.13 (0.90 to 1.36)	.30
High	0.89 (0.77 to 1.02)	.08	−0.02 (−0.08 to 0.03)	.39	1.25 (1.06 to 1.44)	.02
Elixhauser comorbidity index	1.01 (1.01 to 1.02)	<.001	0.006 (0.003 to 0.008)	<.001	0.97 (0.95 to 0.98)	<.001
Grade
Well differentiated	1 [Reference]	NA	1 [Reference]	NA	1 [Reference]	NA
Moderately differentiated	1.49 (1.28 to 1.73)	<.001	−0.03 (−0.09 to 0.03)	.40	1.06 (0.89 to 1.23)	.49
Poorly differentiated or undifferentiated	1.84 (1.57 to 2.16)	<.001	−0.03 (−0.10 to 0.05)	.46	0.65 (0.39 to 0.91)	.001
T category
T1	1 [Reference]	NA	1 [Reference]	NA	1 [Reference]	NA
T2	1.26 (1.03 to 1.55)	.03	0.01 (−0.06 to 0.08)	.40	1.20 (0.99 to 1.41)	.10
T3	1.45 (1.16 to 1.80)	<.001	0.05 (−0.04 to 0.15)	.27	0.71 (0.46 to 0.95)	.005
N category
N0	1 [Reference]	NA	1 [Reference]	NA	1 [Reference]	NA
N1	1.47 (1.32 to 1.63)	<.001	0.00 (−0.04 to 0.03)	.86	0.70 (0.00 to 1.42)	.33
AJCC stage
I	1 [Reference]	NA	1 [Reference]	NA	1 [Reference]	NA
II	1.27 (1.02 to 1.57)	.03	0.03 (−0.05 to 0.11)	.42	0.87 (0.64 to 1.09)	.21
Pancreatic resection
Pancreaticoduodenectomy	1 [Reference]	NA	1 [Reference]	NA	1 [Reference]	NA
Distal pancreatectomy	1.03 (0.91 to 1.18)	.61	−0.21 (−0.26 to −0.17)	<.001	1.31 (0.95 to 1.67)	.14
Other pancreatectomy	0.93 (0.73 to 1.18)	.55	−0.13 (−0.23 to −0.02)	.02	1.72 (1.29 to 2.14)	.01
Total pancreatectomy	1.47 (1.18 to 1.82)	<.001	0.05 (−0.04 to 0.15)	.26	0.40 (0.00 to 1.06)	.007
Chemotherapy
None	1 [Reference]	NA	1 [Reference]	NA	1 [Reference]	NA
Postoperative	0.71 (0.64 to 0.79)	<.001	−0.06 (−0.11 to −0.02)	.006	1.22 (1.02 to 1.43)	.05
Preoperative	0.78 (0.61 to 1.01)	.06	0.07 (−0.01 to 0.15)	.10	0.85 (0.55 to 1.15)	.28
Radiotherapy	0.96 (0.87 to 1.06)	.40	0.00 (−0.04 to 0.04)	.90	0.89 (0.77 to 1.00)	.05
NCI-designated cancer center	1.02 (0.82 to 1.28)	.84	−0.03 (−0.24 to 0.19)	.80	1.07 (0.66 to 1.49)	.74
Teaching hospital	1.06 (0.88 to 1.29)	.55	0.08 (−0.10 to 0.26)	.38	0.84 (0.33 to 1.35)	.49
High-volume centerd (≥20)	0.78 (0.61 to 0.99)	.04	0.11 (−0.09 to 0.32)	.28	1.08 (0.54 to 1.61)	.07
Complication within 30 d	1.18 (1.07 to 1.30)	.001	0.22 (0.17 to 0.27)	<.001	0.68 (0.49 to 0.86)	<.001
Readmission within 30 d	1.01 (0.90 to 1.12)	.93	0.34 (0.29 to 0.39)	<.001	0.64 (0.44 to 0.84)	<.001
Length of stay	1.00 (1.00 to 1.00)	.94	0.03 (0.03 to 0.04)	<.001	0.82 (0.78 to 0.85)	<.001

Abbreviations: aHR, adjusted hazard ratio; AJCC, American Joint Committee on Cancer, seventh edition; aOR, adjusted odds ratio; NA, not applicable; NCI, National Cancer Institute.

^aBased on Cox proportional hazards model, adjusted for within-hospital clustering of patients.

^bIncludes the index surgical hospitalization and readmissions within 30 days. Cost estimates based on generalized linear model with gamma distribution and log link, adjusted for within-hospital clustering of patients.

^cDefined as the fourth quintile for survival (≥26 months) and second quintile or less for costs (≤$40 674). Odds ratios based on logistic regression model, adjusted for within-hospital clustering of patients.

^dIndicates 20 or more cases annually.

Female sex was associated with higher odds of high-value care (adjusted odds ratio [aOR], 1.10; 95% CI, 1.01-1.19; P = .048), whereas higher Elixhauser comorbidity index score (aOR, 0.97; 95% CI, 0.95-0.98; P < .001), poorly differentiated or undifferentiated tumors (aOR, 0.65; 95% CI, 0.39-0.91; P = .001), T3 category (aOR, 0.71; 95% CI, 0.46-0.95; P = .005), total pancreatectomy (aOR, 0.40; 95% CI, 0-1.06; P = .007), postoperative complications (aOR, 0.68; 95% CI, 0.49-0.86; P < .001), readmissions (aOR, 0.64; 95% CI, 0.44-0.84; P < .001), and longer lengths of stay (aOR, 0.82; 95% CI, 0.78-0.85; P < .001) were associated with lower odds of high-value care. National Cancer Institute–designated cancer centers and high-volume centers were not associated with high-value care in multivariable analysis. Because perioperative outcomes (eg, complications, readmissions, and length of stay) were associated with higher costs and decreased survival, we repeated the multivariable survival, cost, and value models after removing these confounding variables; our findings remained unchanged in that neither hospital volume nor NCI-designated cancer center status was associated with high-value care.

Discussion

In this analysis of patients with early-stage pancreatic cancer in California, pancreatic resections with higher value, defined as high-quality care (ie, longer survival) with low health care costs, were associated with important patient, tumor, and treatment characteristics. However, hospital-level factors, including NCI-designated cancer centers, teaching hospitals, and high-volume centers, were not associated with greater value in multivariable analyses. This finding is surprising in the context of existing literature supporting the association between high-volume centers and improved perioperative outcomes.^{9,10,11,12,13,14} These findings suggest that regionalization of pancreatic cancer surgical care may translate into potential improvements in patient long-term survival but not into lower health care costs or, importantly, greater value.

This study expands on previous research^27,28,29 suggesting that patients treated at NCI-designated cancer centers may experience better surgical outcomes for pulmonary, colorectal, and gastric cancer but not necessarily for pancreatic cancer. Similar to this study’s findings, Birkmeyer et al²⁷ found no significant differences in operative mortality or 5-year survival among Medicare patients who underwent pancreatic resection at NCI- and non–NCI-designated cancer centers. Interestingly, these findings contradict the study by Wolfson et al,²⁹ which showed improved survival among Los Angeles County patients who underwent pancreatic resections at NCI-designated comprehensive cancer centers. This may be secondary to the limited population sample in the study by Wolfson et al.²⁹ Regardless, in the present study, using a population-based sample of patients with early-stage pancreatic cancer who underwent surgical resection in 157 California hospitals, we did not find improvements in survival, costs, and, most importantly, the value of care provided at NCI-designated cancer centers.

Our findings suggest a continued need for the identification of value-enhancement measures in pancreatic surgery, especially in high-volume and NCI-designated cancer centers. National Cancer Institute–designated cancer centers and high-volume centers will need to perform surgery for higher-risk patients because higher-risk cancer is their area of expertise. In addition, significant advances in survival of resectable pancreatic cancer are likely to come from more effective chemotherapy rather than improvements in operative management. As such, improvements in value are more likely to be obtained through efforts to reduce cost of health care delivery. At the provider level, all efforts should be made to reduce complications, because complications have been clearly demonstrated to correlate with extended lengths of stay, increased costs, and lower value.^15,30,31 From a systems standpoint, implementation of standardized, empirically validated processes of care need to be implemented in an effort to reduce complications, lengths of stay, and readmissions. However, Ho et al³² found that implementation of processes of care in response to complications may increase hospital costs. A counterpoint by Ejaz et al³¹ noted significant variation among providers with respect to outcomes and cost driven by excess use of resources. Therefore, resource-conscious processes of care designed to reduce cost with consistent adherence by providers is needed in the current health care system.

The implications of our findings on regionalization of pancreatic surgical care, with respect to preferential referrals to high-volume pancreatic surgery centers, are substantial. Consistent with prior research, high-volume centers were associated with improved overall survival but were not associated with significant differences in health care costs.^15,16,17,33 The greater survival associated with high-volume centers is likely multifactorial, because it may be secondary to surgical therapy and more advanced multiagent chemotherapy regimens adopted at high-volume centers. Multiagent chemotherapy regimens, including gemcitabine hydrochloride–based combination therapies and, more recently, FOLFIRINOX (leucovorin calcium, fluorouracil, irinotecan hydrochloride, and oxaliplatin) have been associated with improved survival compared with single-agent gemcitabine in randomized clinical trials in resectable and metastatic pancreatic cancer.^{34,35,36,37,38} In the present study, multiagent chemotherapy was more frequently used in patients treated at high-volume surgical centers despite such multiagent regimens not yet being the standard of care for adjuvant chemotherapy in resectable pancreatic cancer during the time period. Future research is needed to evaluate the differences in multimodality treatment practices at high- and low-volume pancreatic centers and the effect of these treatment practices on patient prognosis as multiagent chemotherapy is more commonly used.

Limitations

This study must be considered in light of its limitations. Because this study uses retrospective observational data, it is subject to the inherent risk of confounding and selection bias. We acknowledge that we are limited by the data provided, and potentially unmeasured confounders, such as functional status, may influence our findings. For example, the finding that adjuvant chemotherapy was associated with lower costs for the surgical hospitalization is potentially secondary to patients receiving adjuvant chemotherapy having better perioperative performance status compared with patients who did not receive adjuvant chemotherapy, and better performance status is associated with lower surgical hospitalization costs. As with most administrative database research, we lack granular clinical detail, including the severity of comorbidities and complications. In addition, because we were only provided with the total costs of each hospitalization from OSHPD, we were not able to perform an analysis of itemized costs (ie, costs of medications, nursing care, operation, imaging, and laboratory studies). Future research should examine itemized costs to determine specific measures that may be used to reduce health care costs for pancreatic surgery. Additionally, because OSHPD does not report data from Kaiser Permanente facilities, we were unable to include these patients in our analysis.

Although value has been consistently described as quality relative to costs,⁵ methods of defining and classifying health care value is variable in present research. Soneji and Yang³⁹ defined value as a ratio of survival relative to costs (ie, value = survival/costs). However, when using this ratio to define high- and low-value care, a patient with poor survival but extremely low hospitalization cost may still be categorized as high value, despite poor survival being inconsistent with high-value care. In contrast, Lawson et al⁴⁰ defined high value as requiring high quality and low costs. We defined high- and low-value care consistent with Lawson et al⁴⁰ because we preferred to recognize those who met quality (ie, survival) and cost criteria as high or low value.

Conclusions

In this population-based analysis of survival, costs, and value of pancreatic surgery for early-stage cancer, high-value care was associated with important patient, tumor, and treatment characteristics. High-volume pancreatic surgery centers were associated with improved survival; however, high-volume centers and NCI-designated cancer centers were not associated with high-value care. These data suggest that although regionalization of pancreatic cancer surgical care may lead to improvements in patient survival, regionalization is not associated with decreased costs or high value. Therefore, targeted measures to enhance value are needed at high-volume and NCI-designated cancer centers, because these centers already treat a significant fraction of patients with pancreatic cancer and are likely to bear an increasing demand for the care of these patients.

Supplement.

eTable. Site and Histology Codes for Pancreatic Adenocarcinoma