World J Surg (2014) 38:1184–1195 DOI 10.1007/s00268-013-2384-z

Neoadjuvant Therapy and Vascular Resection During Pancreaticoduodenectomy: Shifting the Survival Curve for Patients with Locally Advanced Pancreatic Cancer Irene Epelboym • J. DiNorcia • M. Winner M. K. Lee • J. A. Lee • B. A. Schrope • J. A. Chabot • J. D. Allendorf

•

Published online: 5 December 2013 Ó Socie´te´ Internationale de Chirurgie 2013

Abstract Background Neoadjuvant therapy and vascular resection may offer patients with locally advanced pancreatic cancer potential cure. Methods We reviewed medical records of patients with ductal adenocarcinoma who underwent pancreaticoduodenectomy (PD) from 1992 through 2011. We identified patients who received neoadjuvant therapy (NA?) or required vascular resection (VR?) for locally advanced disease and compared outcomes to those who did not. Results Of the 643 patients who were initially explored, 506 (143 NA? and 363 NA- patients) ultimately underwent PD. There were no significant differences in R0 resection or morbidity. Mortality was higher in the NA? versus NA- group (7.0 vs 3.0 %, p = 0.04). More NA? patients underwent PD VR? (p \ 0.001). Among VR? patients, neoadjuvant therapy resulted in significantly lower R1 resection. Among resected patients, survival of NA? patients was significantly longer than both NApatients (27.3 vs 19.7 months, p \ 0.05) and patients abandoned because of locally advanced disease. Age, tumor grade, lymph node ratio, and R1 resection were independent predictors of poor survival. Conclusions Neoadjuvant therapy and vascular resection offer patients with locally advanced pancreatic cancer the chance for cure with acceptable morbidity and mortality. This study was a plenary presentation at the Society for Surgery of the Alimentary Tract (SSAT) meeting in San Diego, May 18–22, 2012. I. Epelboym (&)  J. DiNorcia  M. Winner  M. K. Lee  J. A. Lee  B. A. Schrope  J. A. Chabot  J. D. Allendorf Department of Surgery, Columbia University, College of Physicians and Surgeons, New York, NY 10032, USA e-mail: [email protected]; [email protected]

123

These patients have improved survival over patients deemed locally inoperable by traditional criteria.

Introduction Pancreatic cancer is one of the most lethal gastrointestinal malignancies and the fourth leading cause of cancer-related deaths in the United States [1]. Resection offers the best chance for long-term survival, yet \20 % of patients diagnosed with pancreatic cancer have resectable disease at presentation [2]. Nearly half have metastatic disease, and the remaining patients have locally advanced (borderline resectable and locally unresectable) disease [3]. Metastatic disease still precludes surgery, but new approaches in chemoradiotherapy [3, 4] and surgical technique [5, 6] are broadening what is considered resectable to include many patients with locally advanced disease. While the optimal management of locally advanced pancreatic cancer remains a subject of debate [7–9], neoadjuvant therapy and pancreatectomy with vascular resection and reconstruction are two strategies that can help bring patients with locally advanced pancreatic cancer to the operating room for potential cure. We previously reported our experience with pancreatectomy following neoadjuvant therapy in patients with locally advanced disease, demonstrating feasibility, efficacy, and equivalent survival when compared to resectable patients [10]. The aim of the present study was to examine our specific approach to patients with locally advanced ductal adenocarcinoma in the head of the pancreas. We evaluated perioperative outcomes, histopathologic characteristics, and survival in patients who underwent pancreaticoduodenectomy (PD) after neoadjuvant therapy and compared them to patients who underwent upfront surgery.

World J Surg (2014) 38:1184–1195

We similarly evaluated patients who underwent PD with vascular resection and reconstruction and compared them to patients who underwent standard PD. These analyses suggest that neoadjuvant therapy and vascular resection can recruit patients with locally advanced pancreatic cancer to resection with equivalent morbidity and improved survival compared to patients with traditionally resectable disease.

Methods We performed a retrospective review of a prospective database of patients with pancreatic disease. The database is maintained by The Pancreas Center of Columbia University Medical Center (CUMC), New York, and includes the patients of four surgeons (J. D. Allendorf, J. A. Chabot, J. A. Lee, and B. A. Schrope). After approval by the institutional review board (IRB) and in compliance with the Health Insurance Portability and Accountability Act (HIPAA), we queried the database to identify all patients who underwent attempted PD for ductal adenocarcinoma in the pancreatic head from 1992 through 2011. Patients with cholangiocarcinoma, ampullary carcinoma, and carcinoma arising in the background of intraductal papillary mucinous neoplasm were excluded. Patients who underwent total pancreatectomy also were excluded. We stratified patients into those who received neoadjuvant therapy (NA?) and those who did not (NA-). We also grouped patients by those who underwent vascular resection and reconstruction (VR?) and those who did not (VR-). Need for neoadjuvant therapy and resectability were determined at a weekly multidisciplinary conference dedicated to pancreatic disease where CUMC surgeons, gastroenterologists, oncologists, and radiologists systematically review the preoperative work-up for each patient. Definitions of ‘‘resectable’’ have changed over time and vary in the literature [11–13]; therefore, we retrospectively reviewed all available radiologic imaging reports to classify tumors preoperatively as resectable, borderline resectable, or locally unresectable. These classifications were defined according to the guidelines set by the National Comprehensive Cancer Network [14]. Resectable tumors met the following criteria: no radiographic evidence of superior mesenteric vein (SMV) or portal vein (PV) involvement with clear fat planes around the hepatic artery (HA), superior mesenteric artery (SMA), and celiac axis. Borderline resectable tumors met the following criteria: any SMV or PV involvement with technical option for resection and reconstruction, HA abutment or short segment encasement, or \180° SMA abutment. Locally unresectable tumors met the following criteria: unreconstructible SMV or PV involvement, [180° SMA

1185

encasement, or any celiac abutment. ‘‘Borderline resectable’’ and ‘‘locally unresectable’’ collectively are referred to as ‘‘locally advanced.’’ The treatment algorithm for these patients has been described previously [10]. Briefly, following preoperative attempt at staging based on imaging and endoscopy, if there was any question regarding extent or resectability of disease, diagnostic laparoscopy was performed to further elaborate the stage and exclude the presence of extrapancreatic disease. Those confirmed not to have metastatic disease proceeded to pancreatectomy. Patients with locally advanced disease had tissue biopsy to confirm the diagnosis of pancreatic ductal adenocarcinoma prior to starting neoadjuvant therapy. Neoadjuvant therapy consisted of chemotherapy with or without radiation either at CUMC or at outside institutions under the care of referring oncologists. There was no standardized protocol, although most patients received gemcitabine-based chemotherapy. All patients treated at CUMC received gemcitabine-based chemotherapy, often in conjunction with docetaxel and capecitabine, a regimen known as GTX [10]. In our early experience, we gave three cycles of neoadjuvant therapy, and if no progression of disease was observed on follow-up imaging, patients were brought to resection. Following this, three more cycles of adjuvant therapy were given. As we observed few patients making progress on neoadjuvant protocol, we have now moved to administering all six cycles preoperatively. We believe this achieves better systemic control of a disease that at presentation is frequently micrometastatic, as well as has a highest chance of downstaging the tumor, allowing for a higher proportion of patients who can be brought to successful resection. Finally, in our experience, patients are better able to tolerate most of the cytotoxic therapy upfront because of better performance status preoperatively than following surgery. Patients were restaged radiographically after neoadjuvant therapy and then brought to the operating room for exploration if there was no evidence of disease progression. Select patients with borderline resectable tumors who had isolated SMV or PV involvement with a safe and feasible option for resection and reconstruction as determined by the operating surgeon were taken to the operating room without neoadjuvant therapy. These patients were unable to have neoadjuvant therapy because a pathologically confirmed diagnosis of adenocarcinoma could not be obtained preoperatively or because of medical, psychosocial, or insurance reasons. Demographic and descriptive data were collected by review of patients’ medical records. Intraoperative variables were obtained from nurse, anesthesiologist, and surgeon reports. Operative time was defined as the time between patient entry into and exit from the operating room. Histopathologic variables were determined from

123

1186

standardized final pathology reports. These variables included tumor grade, tumor diameter, lymphovascular, and perineural invasion, regional lymph node status, and margin status. The margins evaluated for each specimen included pancreatic, biliary, gastric/duodenal, and retroperitoneal margins. Resection status was designated as R0 (margin negative), R1 (microscopically positive), or R2 (macroscopically positive) based on both operative and pathology reports. Lymph node ratio (LNR) was calculated as the ratio of positive lymph nodes to total lymph nodes removed and then stratified into four groups: 0, [0–0.2, [0.2–0.4, and [0.4. Pathologists determined the pathologic TNM stage according to the American Joint Committee on Cancer (AJCC) Cancer Staging Manual, 7th Edition. Perioperative complications were gathered from daily progress notes and discharge summaries and were graded with the system proposed by DeOliveira et al. [15]. Overall morbidity was defined as any complication, and major morbidity was defined as complications grade III and greater. Pancreatic fistula was assessed and graded according to recommendations by the International Study Group on Pancreatic Fistula [16]. Reoperation rate was defined as return to the operating room within the same hospital admission. Length of stay (LOS) was calculated from date of operation to date of hospital discharge. Perioperative mortality was defined as death within 30 days of the operation or within the same hospital admission as the operation. Survival was calculated from the date of treatment through the date of last follow-up. For NA? patients, date of treatment was defined as the start date of neoadjuvant therapy; for NA- patients, date of treatment was defined as the date of operation. Continuous variables are presented as mean and standard deviation or median and interquartile range (IQR) based on assessments of normality. We compared patient demographic, operative, and histopathologic characteristics with Student’s t test or the Mann–Whitney U test for continuous variables, and the Chi square or Fisher’s exact test for categorical variables. We examined relationships between patient and treatment characteristics and major morbidity with univariate and multivariate logistic regression models. LOS was analyzed with Kaplan–Meier curves and the Wilcoxon rank-sum test. Survival probability was estimated with the Kaplan–Meier method and with univariate and multivariate Cox proportional hazards models. We evaluated overall survival among all patients stratified by neoadjuvant therapy and completed resection. In this analysis, we excluded 63 patients whose resection was abandoned because of metastatic disease (as opposed to those abandoned because of locally unresectable disease) and 21 patients who died in the perioperative period. Survival durations were determined from the Kaplan–Meier curves and compared with the log rank test. We used the

123

World J Surg (2014) 38:1184–1195

Bonferroni method to correct for multiple comparisons when appropriate. Adjusted p values are indicated by p0 . We assessed the relationship between patient, operative, and pathologic characteristics and overall survival among resected patients with univariate Cox proportional hazards models. Clinically and statistically significant factors were then included in a multivariate model of overall survival. In univariate analysis, we dichotomized age above 65 and categorized year of operation into three groups (1992–1999, 2000–2005, 2006–2011); in multivariate analysis, age was treated as a continuous variable. Proportional hazards were confirmed visually, and log rank or Wilcoxon rank-sum p values were reported based on visual inspection of Kaplan–Meier curves. We used SAS 9.2 (SAS Institute, Inc. Cary, NC, USA) for statistical analysis. All tests were two-sided, and we considered a p value of B0.05 to be statistically significant. Of note, the groups as they are presented reflect intention to treat. For example, if a patient received only one cycle of neoadjuvant therapy and moved to surgery early, this patient is still included in the neoadjuvant therapy group. The neoadjuvant therapy group includes only those patients who ultimately went to surgical therapy and not those who were, for instance, borderline resectable at the start of neoadjuvant therapy, with an expectation of ultimate surgical therapy, but never made it for reasons that might include a decline in medical fitness or progression of cancer while receiving therapy.

Results From March 1992 through December 2011, 643 patients with ductal adenocarcinoma in the head of the pancreas underwent attempted PD. The mean age was 66.4 years. For initially locally advanced disease, 174 (27.1 %) patients received neoadjuvant therapy (NA?). The remaining 469 (72.9 %) patients were explored at presentation (NA-). NA? Patients Of the 174 NA? patients, 133 (76.4 %) were deemed to have locally advanced disease by preoperative radiologic studies alone and 41 (23.6 %) by prior laparotomy. Almost half of the NA? patients (48.8 %) who had prior laparotomy were explored at CUMC, and the remaining 51.2 % were explored at other institutions. One hundred thirty-four NA? (77 %) patients received neoadjuvant therapy for borderline resectable disease and 40 (23 %) for locally unresectable disease. Table 1 shows the breakdown of neoadjuvant regimens. One hundred fifty-nine (91.4 %) NA? patients received gemcitabine-based chemotherapy. One hundred three (59.2 %) NA? patients received neoadjuvant therapy at CUMC, and 71 (40.8 %) received

World J Surg (2014) 38:1184–1195

1187

Table 1 Neoadjuvant regimens Therapeutic regimen

n (%)

Chemotherapy

74 (42.5)

GTX

62

Gemcitabine ± other

7

Other

5

Chemotherapy ? radiation

100 (57.5)

GTX ? XRT

63

Gemcitabine ± other ? XRT

27

Other ? XRT

10

GTX gemcitabine, docetaxel, capecitabine, XRT abdominal radiation

treatment at other institutions. One hundred three NA? (59.2 %) patients received preoperative abdominal radiation as part of their neoadjuvant therapy. The median time from start of neoadjuvant therapy to surgery was 5.4 months (IQR 4.4–7.1). Of note, patients who were offered neoadjuvant therapy in an effort to attempt to downstage initially unresectable disease, but who did not respond (evidence of extrapancreatic disease or progression of disease on re-staging follow-up scan after neoadjuvant therapy) were not included in this analysis. After neoadjuvant therapy, 143 (82.2 %) patients ultimately underwent successful PD while 31 (17.8 %) were abandoned, 20 (11.5 %) because of locally advanced disease and 11 (6.3 %) because of metastatic disease. Of the NA? patients who underwent surgical resection, 114 (79.7 %) initially had borderline resectable disease and 29 (20.3 %) had locally unresectable disease. Ninety-two (64.3 %) NA? patients required vascular resection and reconstruction to extirpate the tumor; 67 (72.8 %) were borderline resectable and 25 (27.2 %) were locally unresectable prior to neoadjuvant therapy. Seventy-seven (53.8 %) NA? patients underwent venous resection only, 2 (1.4 %) underwent arterial resection only, and 13 (9.1 %) underwent concomitant venous and arterial resection. NA- patients Of the 469 NA- patients explored at presentation, 363 (77.4 %) underwent successful PD and 106 (22.6 %) were abandoned after exploration, 54 (11.5 %) because of locally advanced disease and 52 (11.1 %) because of metastatic disease. Sixty-six (18.2 %) NA- patients required vascular resection and reconstruction to extirpate the tumor, all of which were venous resections. Of these 66 patients, 20 (30.3 %) were borderline resectable on preoperative imaging with isolated SMV or PV involvement amenable to planned resection and reconstruction. The remaining 46 (69.7 %) patients had venous involvement that was not appreciated on preoperative imaging.

Table 2 Demographic, operative, and postoperative characteristics of 506 patients with completed pancreaticoduodenectomy for ductal adenocarcinoma stratified by neoadjuvant treatment NA? (n = 143)

NA(n = 363)

p value

Age, years (mean, SD)

63.5 (9.9)

67.8 (10.3)

\0.001

Gender, male (%)

71 (49.7)

186 (50.8)

0.833

Race, white (%)

114 (79.7)

256 (70.0)

0.023

108 (75.5)

363 (99.2)

\0.001

Demographics

Prior resection attempt None (%) Outside institution (%)

19 (13.3)

1 (0.3)

CUMC (%)

16 (11.2)

2 (0.6)

Operative characteristics \0.001

Vascular resection (%)

92 (64.3)

66 (18.2)

Arterial and venous resection

13 (9.3)

0

Arterial resection only

2 (1.4)

0

Venous resection only

77 (53.9)

66 (18.2)

EBL, L (median, IQR)

1.5 (1.0,3.0)

1.0 (0.5,1.5)

\0.001

Operative time, min (median, IQR)

524 (437,630)

412 (373,490)

\0.001

Morbidity, overall (%)

71 (49.7)

179 (48.9)

0.901

Morbidity, major (%)

35 (24.5)

91 (24.9)

0.941

Pancreatic fistula (%)

6 (4.2)

20 (5.5)

0.547

Grade A

1

3

Grade B Grade C

1 4

8 9

Postoperative outcomes

Reoperation (%)

17 (11.9)

28 (7.7)

0.137

LOS, days (mean, SD)

9 (7,13)

10 (7,15)

0.106

Mortality (%)

10 (7.0)

11 (3.0)

0.044

SD standard deviation, IQR interquartile range, EBL estimated blood loss, CUMC Columbia University Medical Center, LOS length of stay

PD in NA? vs NA- patients: operative and postoperative characteristics Table 2 shows the demographic, operative, and postoperative characteristics of 506 patients who underwent successful PD, stratified by neoadjuvant therapy. Among resected patients, the NA? patients were significantly younger than NA- patients (63.5 vs 67.8 years, p \ 0.001). NA? patients more commonly had undergone prior exploration (24.5 vs 0.8 %, p \ 0.001), and more frequently required major vascular resection and reconstruction (64.3 vs 18.9 %, p \ 0.001). Estimated blood loss was larger (1.5 vs 1 L, p \ 0.001) and operative time was longer (524 vs 412 min, p \ 0.001) in the NA? group than in the NA- group. Overall morbidity (49.7 vs 48.9 %, p = 0.901), major morbidity (24.5 vs 24.9 %, p = 0.941), pancreatic fistula (4.2 vs 5.5 %, p = 0.547), and reoperation rates (11.9 vs

123

1188

World J Surg (2014) 38:1184–1195

Table 3 Histopathologic characteristics of 506 patients with completed pancreaticoduodenectomy for ductal adenocarcinoma stratified by neoadjuvant treatment NA? (n = 143)

NA(n = 363)

p value

Well/moderate

67 (48.9)

171 (47.6)

0.491a

Poor/undifferentiated

64 (46.7)

188 (52.4)

No detectable tumor

6 (4.4)

0

T0

6 (4.2)

0

T1

27 (18.9)

9 (2.5)

T2

9 (6.3)

14 (3.9)

T3

100 (69.9)

338 (93.6)

T4

1 (0.7)

0

Grade (%)

pT stage (%) \0.001b

pN stage (%) \0.001

N0

74 (51.8)

101 (28.0)

N1

69 (48.2)

260 (72.0)

2.6 (2.0, 3.5)

3.0 (2.2, 4.0)

0.002

Tumor size, cm (median, IQR) Lymphovascular invasion

71 (58.7)

263 (84.8)

\0.001

Perineural invasion

96 (72.2)

288 (88.9)

\0.001

Positive nodes, n (median, IQR)

0 (0, 2)

2 (0, 5)

\0.001

Total nodes, n (median, IQR)

12 (9, 17)

12 (8, 16)

0.609

0

75 (52.5)

98 (27.2)

\0.001c

[0–0.2

48 (33.6)

97 (26.9)

[0.2–0.4

15 (10.5)

85 (23.6)

[0.4

5 (3.5)

80 (22.2)

R0

117 (81.8)

282 (77.7)

R1

26 (18.2)

81 (22.3)

Lymph node ratio (%)

Resection (%)

0.305

a

Statistical analysis performed on well and moderately differentiated versus poor and undifferentiated tumors

b

Statistical analysis performed on T0–T2 versus T3–T4 tumors

c

Statistical analysis performed on lymph node ratio\0.2 versus[0.2

7.7 %, p = 0.137) were similar between the NA? and NAgroups. There was no difference in LOS (9 vs 10 days, p = 0.106). Perioperative mortality was significantly higher in NA? patients (7.0 vs 3.0 %, p = 0.044). All but one of the patients who died had undergone venous resection and reconstruction, and three of them had concomitant arterial resection and reconstruction. PD in NA? vs NA- patients: pathology Histopathologic characteristics of resected lesions stratified by neoadjuvant therapy are summarized in Table 3. Pathologic tumor diameter was significantly smaller in the

123

NA? compared to the NA- group (2.6 vs 3.0 cm, p = 0.002). There was no difference in tumor grade between the two groups. NA? patients had significantly less lymphovascular invasion (58.7 vs 84.8 %, p \ 0.001) and perineural invasion (72.2 vs 88.9 %, p \ 0.001) identified in the specimens. NA? patients had significantly lower pathologic T and N stages than NA- patients. Patients in both groups had similar numbers of total lymph nodes removed, though the NA? patients had fewer positive lymph nodes than the NA- patients, resulting in overall lower LNR in the NA? group. An R0 resection was achieved with equal frequency between the two groups (81.8 vs 77.7 %, p = 0.305). No patient in either group underwent an R2 resection. A complete pathologic response to neoadjuvant therapy, manifested as no detectable tumor, was observed in the specimens of six patients (4.3 %). Vascular resection and reconstruction: operative and postoperative characteristics Of the 506 patients ultimately resected, 158 (31.2 %) required major vascular resection and reconstruction to extirpate the tumor. Demographic, operative, and postoperative characteristics of patients who underwent PD with (VR?) and without (VR-) vascular resection and reconstruction are summarized in Table 4. The VR? patients tended to be younger (65.2 vs 67.2 years, p = 0.05) and more commonly had undergone prior exploration (13.9 vs 4.6 %, p \ 0.001) than VR- patients. More than half (58.2 %) of the VR? patients had undergone neoadjuvant therapy. Estimated blood loss was larger (2.2 vs 0.9 L, p \ 0.001) and operative time was longer (540 vs 405 min, p \ 0.001) in the VR? compared to VR- group. Overall morbidity (53.8 vs 47.1 %, p = 0.164), major morbidity (28.5 vs. 23.0 %, p = 0.184), pancreatic fistula (5.7 vs 4.9 %, p = 0.702), and reoperation rates (10.8 vs 8.1 %, p = 0.32) were similar between the VR? and VRgroups. There was no difference in LOS (10 vs 10 days, p = 0.208). Perioperative mortality also was similar between the two groups (6.3 vs 3.2 %, p = 0.098). Given that 30 % of the perioperative deaths in the NA? group were in patients who underwent arterial resection, we evaluated perioperative mortality among venous, arterial, and combined vascular resections. The perioperative mortality rate among patients who underwent any arterial resection was 20.0 % (n = 3/15), which was significantly higher than the mortality of both those who underwent venous resection only (4.9 %, n = 7/143; p = 0.022) and those who did not undergo any vascular resection (3.2 %, 11/348; p \ 0.001). The perioperative mortality rate was no different between those undergoing venous resection only and no vascular resection (p = 0.353). However, after

World J Surg (2014) 38:1184–1195

1189

Table 4 Demographic, operative, and postoperative characteristics of 506 patients with completed resection for ductal adenocarcinoma stratified by vascular resection VR? (n = 158)

VR(n = 348)

Table 5 Histopathologic characteristics of 506 patients with completed resection for ductal adenocarcinoma stratified by vascular resection

p value

Demographics

VR? (n = 158)

VR(n = 348)

p value

73 (47.1)

165 (48.4)

0.586a

Grade (%)

Age, years (mean, SD)

65.2 (9.8)

67.2 (10.6)

0.050

Well/moderate

Gender, male (%)

81 (51.3)

174 (50.0)

0.792

Poor/undifferentiated

79 (51.0)

173 (50.7)

Race, white (%)

126 (79.8)

241 (69.3)

0.014

No detectable tumor

3 (1.9)

3 (0.9)

T0

3 (1.9)

3 (0.9)

pT stage (%)

Prior resection attempt None (%)

136 (86.1)

332 (95.4)

\0.001

0.914b

Outside institution (%)

10 (6.3)

10 (2.9)

T1

10 (6.3)

26 (7.5)

CUMC (%)

12 (7.6)

6 (1.7)

T2

7 (4.4)

16 (4.6)

T3

137 (86.7)

301 (87.0)

T4

1 (0.6)

0

N0

72 (45.6)

103 (29.8)

N1

86 (54.4)

243 (70.2)

Tumor size, cm (median, IQR)

3.0 (2.3,3.9)

3.0 (2.1, 3.8)

0.283

Lymphovascular invasion (%)

99 (71.2)

235 (80.5)

0.032

Neoadjuvant treatment \0.001

None (%)

66 (41.8)

297 (85.3)

NA ± radiation (%)

92 (58.2)

51 (14.7)

EBL, liters (median, IQR)

2.2 (1.1,4.0)

0.9 (0.5,1.2)

\0.001

Operative time, min (median, IQR)

540 (446,630)

405 (355,470)

\0.001

pN stage (%)

Operative characteristics

Postoperative outcomes

\0.001

Reoperation (%)

17 (10.8)

28 (8.1)

0.320

Perineural invasion (%)

125 (82.8)

259 (84.7)

0.610

Overall morbidity (%)

85 (53.8)

164 (47.1)

0.164

1 (0,3)

2 (0, 4)

\0.001

Major morbidity (%)

45 (28.5)

80 (23.0)

0.184

Mortality (%)

10 (6.3)

11 (3.2)

0.098

12 (8,17)

12 (8,16)

0.922

Length of stay, days (median, IQR)

10 (8,15)

10 (7,15)

0.208

Positive nodes, n (median, IQR) Total nodes, n (median, IQR)

Pancreatic fistula (%)

9 (5.7)

17 (4.9)

0.702

0

73 (46.2)

100 (29.0)

\0.001c

0

4

[0–0.2

45 (28.5)

100 (29.0)

[0.2–0.4

21 (13.3)

79 (22.9)

[0.4

19 (12.0)

67 (19.1)

Grade A Grade B

1

8

Grade C

8

5

Lymph node ratio (%)

Resection (%)

accounting for multiple comparisons, the difference in mortality between arterial and venous resections was no longer significant (p’ [ 0.05). Vascular resection and reconstruction: pathology Histopathologic characteristics stratified by vascular resection are summarized in Table 5. There was no difference in pathologic tumor size or tumor grade between the two groups. Compared to VR- patients, VR? patients had significantly less lymphovascular invasion (71.2 vs 80.5 %, p = 0.032) and similar perineural invasion (82.8 vs 84.7 %, p = 0.61) identified in the specimens. VR? patients had similar pathologic T stage, but lower N stage than VRpatients. Patients in both groups had similar numbers of total lymph nodes removed, although the VR? patients had fewer positive lymph nodes, resulting in overall lower LNR in the VR? versus the VR- patients. R0 resection rates were similar between the two groups (74.1 vs 81.0 %,

R0

117 (74.1)

282 (81.0)

R1

41 (25.9)

66 (19.0)

0.075

a

Statistical analysis performed on well and moderately differentiated versus poor and undifferentiated tumors

b

Statistical analysis performed on T0–T2 versus T3–T4 tumors

c

Statistical analysis performed on lymph node ratio\0.2 versus[0.2

p = 0.075). Among VR? patients, however, those who received neoadjuvant therapy had a significantly lower rate of R1 resection compared to those who did not receive neoadjuvant therapy (19.6 vs 34.9 %, p = 0.031). Among NA- patients, those who underwent vascular resection had a significantly higher rate of R1 resection compared to those who had PD VR- (34.9 vs 19.5 %, p = 0.007). PD in NA? vs NA- patients: survival Overall survival in 559 patients who underwent attempted PD is shown in Fig. 1, stratified by neoadjuvant therapy and

123

1190

World J Surg (2014) 38:1184–1195

Fig. 1 Overall patient survival after 559 attempted pancreaticoduodenectomies for ductal adenocarcinoma of the head of the pancreas between 1992 and 2011 stratified by neoadjuvant therapy and completed resection. Sixty-three patients aborted for metastatic disease and 21 patients who died in the perioperative period were excluded from the analysis. The median survival among successfully resected patients after neoadjuvant therapy was 27.3 months,

significantly longer than the median survival of 19.7 months among non-neoadjuvant patients who were successfully resected (p0 \ 0.05) and significantly longer than the median survival of the neoadjuvant (15.1 months) and non-neoadjuvant (8.4 months) patients whose resections were aborted because of locally advanced disease (p0 \ 0.05)

completed resection. With a median follow-up of 15.3 months, the median survival for NA? patients whose disease was successfully resected was 27.3 months, which was significantly longer than the median survival of 19.7 months among NA- patients who underwent resection (p0 \ 0.05). The median survival in the NA? group also was significantly longer than the median survivals of both NA? (15.1 months) and NA- (8.4 months) patients whose operations were abandoned because of locally advanced disease (p0 \ 0.05). With a median follow-up of 16.3 months, there was no difference in median survival between NA? patients who underwent PD VR? versus NA? patients who underwent PD VR- (24.9 vs 29.8 months, p0 [ 0.05). After accounting for multiple comparisons, there were no differences in overall survival when comparing any group that had a completed resection, stratified by neoadjuvant therapy and vascular resection (p0 [ 0.05). Table 6 shows univariate and multivariate Cox proportional hazard models of overall survival in 470 patients who underwent completed resection. In the univariate analysis, older age and earlier year of operation were associated with poorer 1-year survival. Neoadjuvant therapy was associated with better 1-year survival, whereas vascular resection had no association. Worse tumor grade, presence of lymphovascular and perineural invasion, worse pathologic T stage, high LNR, and positive margin status were associated with worse 1-year survival.

In a multivariate Cox proportional hazards model controlling for age, year of operation, and poor prognostic histopathologic features, neoadjuvant therapy was not an independent predictor of better survival. Each 10-year increase in age and operation in earlier years predicted poorer survival. Worse tumor grade, high LNR, and positive margin status remained independent predictors of worse survival. High LNR was the strongest predictor of survival, increasing the hazard of death 2.5 times that of node-negative patients. Results did not change when vascular resection was added to the model, and vascular resection remained statistically unassociated with overall survival (data not shown). Figure 2 shows overall survival in 470 patients who underwent completed resection stratified by neoadjuvant therapy and node status. The median survival of NA?, node-negative patients was 32.5 months, which was similar to the median survival of 29.1 months among NA-, nodenegative patients (p0 [ 0.05) and significantly longer than the median survival of 18.7 months among NA-, nodepositive patients (p0 \ 0.05).

123

Discussion Pancreatic ductal adenocarcinoma is a challenging malignancy and the management is complex. Despite advances

World J Surg (2014) 38:1184–1195 Table 6 Univariate and multivariate Cox proportional hazards models of overall survival in 470 completed pancreaticoduodenectomies for ductal adenocarcinoma performed between 1992 and 2011

1191

Variables

1-year survival (%)

p value

Adjusted HR

95 % CI

p value

1.25

1.10–1.42

\0.001

\0.001

Age B65 years

83.5

[65 years

72.5

Each 10-year increase in age Year of operation

0.004

0.003

1992–1999

64.0

Ref.

2000–2005

69.4

0.80

0.55–1.16

2006–2011

84.3

0.55

0.37–0.81

Each year increase \0.001

Neoadjuvant treatment No Yes

73.0 88.8

Vascular resection

0.77–1.41

0.722

No

76.0

Venous and/or arterial resection

80.2

Perineural invasion

0.023

0.261

No

87.5

Ref.

Yes

74.8

1.22

0.84–1.77

Unknown

84.2

0.78

0.47–1.32

\0.001

Lymphovascular invasion

0.500

No

88.3

Ref.

Yes

73.6

1.15

0.78–1.69

Unknown

81.1

0.88

0.49–1.60

\0.001

Lymph node ratio

\0.001

0

86.2

Ref.

[0–0.2 [0.2–0.4

77.7 71.3

1.30 1.73

0.92–1.82 1.19–2.51

[0.4

66.1

2.30

1.55–3.43

Tumor stage

0.041

T0–T2

84.8

T3–T4

76.4

0.654 Ref. 0.90

0.58–1.41

\0.001

Grade

0.005

Well/moderately differentiated

84.3

Ref.

Poor/undifferentiated

70.6

1.42

1.11–1.81

1.51

1.15–1.98

\0.001

Margin status HR hazard ratio, CI confidence interval

0.800 Ref. 1.04

R0

78.7

R1

72.5

in treatment and research, overall 5-year survival for patients with the disease remains \5 % [17]. The best survival outcomes are realized when the tumor can be resected with negative margins; unfortunately, for every one patient who presents with a resectable tumor, two patients present with locally advanced disease. Strategies that enable resection in patients with locally advanced disease thus have the potential to triple the surgical impact on pancreatic cancer. Although high-level evidence is lacking, neoadjuvant treatment in the form of chemotherapy and/or chemoradiation is being used more frequently to

0.003

downstage tumors prior to resection [18]. Coupled with the ability to resect and reconstruct major vasculature safely during pancreatectomy [19], a new paradigm has emerged in the treatment of patients with locally advanced pancreatic cancer. In this study, we have reported outcomes among patients with locally advanced cancer in the pancreatic head who underwent neoadjuvant therapy prior to PD with or without vascular resection and reconstruction. Figure 3 demonstrates that in recent years about one-third of PDs at our institution are being performed after neoadjuvant therapy

123

1192

World J Surg (2014) 38:1184–1195

Fig. 2 Overall patient survival after 485 completed resections for ductal adenocarcinoma of the head of the pancreas performed between 1992 and 2011, stratified by lymph node status and neoadjuvant chemotherapy. The median survival of patients who received neoadjuvant therapy and were node negative was 32.5 months, which was similar to the median survival of patients who did not receive neoadjuvant therapy and were node negative (29.1 months, p0 [ 0.05) and significantly longer than the median survival of 18.7 months of patients who did not receive neoadjuvant therapy and were node positive (p0 \ 0.05)

Fig. 3 Pancreaticoduodenectomy performed over time during the study period. a Proportion of patients who received neoadjuvant therapy over time. b Proportion of patients who underwent vascular resection and reconstruction over time

and about one-third involve vascular resection. All of the NA? patients in this series had locally advanced disease, whereas the NA- patients largely included patients with resectable disease. Twenty patients in the NA- group had borderline resectable disease and normally would have undergone neoadjuvant therapy prior to resection, but they were unable to, either because of inability to confirm a diagnosis of adenocarcinoma preoperatively or because of medical, psychosocial, or insurance reasons. All twenty of these patients had isolated SMV or PV involvement amenable to reconstruction and hence were taken directly to the operating room for planned PD VR? with 30 % major morbidity and no perioperative mortality. In the literature, resection following neoadjuvant therapy is reportedly achievable in 50–75 % of borderline resectable tumors and about 20 % of locally unresectable tumors [20]. All of the patients in our series were explored surgically.

123

Thus, we do not have a denominator that includes those NA? patients who never were explored, and we cannot accurately calculate rate of resection after neoadjuvant therapy. Nonetheless, our data show that among the 174 NA? patients explored, 65.5 % of those with borderline resectable disease and 16.6 % of those with locally unresectable disease on pretreatment evaluation were ultimately resected, suggesting that neoadjuvant therapy can downstage a substantial percentage of locally advanced patients and allow resection. That being said, it deserves mention that approximately 18 % of patients were not able to complete the full course of neoadjuvant therapy as planned, either due to progression of disease while on treatment or because of treatment intolerance. The most frequent complications of the neoadjuvant therapy were gastrointestinal side effects (nausea, vomiting, intolerance of food intake), as well as hematological issues (such as pancytopenia).

World J Surg (2014) 38:1184–1195

In our analysis, we did not see differences in terms of treatment response, survival, or rates and nature of complications by chemotherapy type or by chemotherapy only versus chemotherapy plus radiation. This effect would be interesting to investigate; however, our study is not powered enough to detect such a difference, and therefore at this time we are unable to make a statement regarding efficacy of one modality versus another. In our series, NA? patients were younger than NApatients, with a mean age of 63.5 years. Younger patients usually are more willing and able to tolerate neoadjuvant therapy followed by complex pancreatectomy. We capitalize on their age and motivation because aggressive treatment offers them the best chance for prolonged survival. NA? patients more often had a prior attempt at resection, had larger estimated blood loss, and longer operative times. Not surprisingly, PD in NA? patients more frequently required vascular resection and reconstruction: 64 % of patients in the NA? group underwent PD VR?. These results reflect the difficulty inherent in reoperative and post-radiation PD, as well as the complexity inherent in PD with vascular resection and reconstruction for locally advanced disease. Despite more difficult and complex operations, the NA? patients who underwent PD had similar overall morbidity, major morbidity, and LOS as the NA- patients. The mortality was higher in the NA? group (7.0 %), though not prohibitively so. Notably, 90 % of the perioperative deaths in the NA? group were patients who had undergone venous resection, and 30 % had concomitant arterial resection. On univariate analysis, patients who had any arterial resection trended toward suffering major complications more commonly than others, although this association did not reach statistical significance (data not shown). In patients with acceptable preoperative performance status and comorbidity burden, this was undertaken after careful discussion of potential risks and benefits among the multidisciplinary team and with the patient. However, based on our observations in the present study, and consistent with published literature, these data underscore the need to be highly selective when considering patients for potential arterial resection and reconstruction. We are therefore increasingly hesitant to offer this option, reserving vascular resection to patients whose disease involves venous structures only [21, 22]. Like other studies in the literature, the histopathologic findings in our series support the efficacy of neoadjuvant therapy followed by resection [20]. NA? patients had smaller tumors with less lymphovascular and perineural invasion than NA? patients. Importantly, NA? patients had significant pathologic downstaging with lower T stage, fewer positive lymph nodes, and lower LNR than patients who underwent resection at presentation. Similar numbers

1193

of lymph nodes were retrieved in both groups, suggesting that neoadjuvant therapy does not compromise the adequacy of lymphadenectomy in PD. Remarkably, six NA? patients demonstrated a complete pathologic response to therapy with no detectable tumor on histopathologic examination of their specimens. Neoadjuvant therapy coupled with resection thus may offer a subset of patients with locally advanced pancreatic cancer the potential for cure, though long-term follow-up is necessary. Over 30 % of PDs during the study period involved vascular resection and reconstruction; therefore we performed a subset analysis to evaluate outcomes specific to that operation. Not surprisingly, blood loss was larger and operative times were longer for PD VR? than for PD VR. Morbidity, mortality, and LOS, however, were similar between the two groups. Patients in the VR? group tended to have more benign histopathology, although these findings likely reflect the effect of neoadjuvant therapy, which was more common in the VR? group. Survival among VR? and VR- patients stratified by neoadjuvant therapy was similar. These results are consistent across the literature and suggest that PD VR? is feasible and safe[23–25]. It has been established that margin-negative resection is the most effective treatment for pancreatic cancer, yet patients with locally advanced disease are at high risk for R1 resection [7, 26]. The strategy of neoadjuvant therapy followed by resection with or without vascular resection can maximize the potential for R0 resection. In our series, patients with locally advanced disease who underwent PD after neoadjuvant therapy had R0 resection rates similar to those for patients who underwent upfront surgery for resectable disease. Importantly, among patients requiring vascular resection, neoadjuvant therapy reduced the R1 resection rate from 35 to 20 %. Moreover, among patients who did not receive neoadjuvant therapy, those who underwent vascular resection had a significantly higher rate of R1 resection compared to those who underwent standard PD, providing further evidence that neoadjuvant therapy facilitates margin-negative resection in patients with locally advanced cancer in the pancreatic head. In a recent systematic review of the published literature on the contribution of neoadjuvant therapy to successful resection and survival of patients with primarily unresectable pancreatic adenocarcinoma, Morganti et al. [27] underscore the impact of this modality. They calculate a compounded rate of 8–64 % successful resection, with an 8–57 % R0 resection rate following administration of neoadjuvant chemoradiotherapy; however, it should be noted that the patients evaluated by Morganti and colleagues represent a very heterogeneous group and the authors attribute such a wide range of resectability to nonstandardized unresectability criteria across the studies, as well as temporal changes in our understanding of limits of

123

1194

resectability. In particular, one of the largest confounders is the need for vascular resection and reconstruction; in many (especially early) series, involvement of portal or splenic veins and the confluence (including post-neoadjuvant persistence of vascular involvement) was categorized as unresectable disease. We now consider these lesions operable in select cases, as presented in our study as well as in many other reports in the published literature that successfully attempted to surgically achieve local control of the lesions. Our results are encouraging in this context, and they provide an important framework for management of patients with pancreatic cancer. Survival analysis in our series was encouraging. Patients with locally advanced disease who underwent PD following neoadjuvant therapy had a median survival of 27.3 months, which was longer than the median survival of 19.7 months for NA- patients who underwent resection and significantly longer than the median survival of both NA? (15.1 months) and NA- (8.4 months) patients whose operations were abandoned because of locally advanced disease. As demonstrated by the Kaplan–Meier curves, neoadjuvant therapy shifted the survival curve of patients with locally advanced disease beyond that of patients who underwent resection at presentation, supporting the efficacy of this strategy. To better understand these survival curves, we performed multivariate analyses to model predictors of survival. Neither neoadjuvant therapy nor vascular resection was associated with survival, but this is perhaps because we did not have a sufficient sample size to power detection of statistically significant difference when other predictive variables were taken into the account. Age was an independent predictor of poor survival; we speculate that older patients less often can tolerate neoadjuvant therapy and complex pancreatectomy, perhaps secondary to comorbid conditions. Treatment received by these patients in earlier years also was an independent predictor of poor survival; it is not surprising that we have improved our outcomes with experience over time. These multivariate analyses also support what is known about the prognostic importance of tumor grade, regional lymph node disease, and negative resection margins; all three of which were independent predictors of poorer survival, with extent of lymph node disease being the strongest predictor. When we examined survival of surgical patients stratified by neoadjuvant therapy and nodal disease, those with node-negative disease tended to survive longer than those with node-positive disease. Again, neoadjuvant therapy tended to prolong survival, as reflected by a shift in the curves to the right. Interestingly, the median survival of NA? patients with node-negative disease (32.5 months) was significantly longer than that of NA- patients with nodepositive disease (18.7 months). Presumably, some patients with locally advanced disease had positive lymph nodes

123

World J Surg (2014) 38:1184–1195

prior to neoadjuvant therapy, and neoadjuvant therapy downstaged their nodal disease and improved survival. It thus makes sense to target tumor biology and locoregional spread with neoadjuvant therapy, downstage the tumor to make resection possible, and performing vascular resection if necessary to achieve negative margins. This study is limited by its retrospective design, nonstandardization of neoadjuvant therapy, and nonrandomization. As our entire experience with PD for ductal adenocarcinoma, the study encompasses a heterogeneous group of patients over two decades, during which time preoperative evaluation, neoadjuvant protocols, surgical technique, and postoperative care have changed considerably. Without randomized, prospective evaluation, selection bias is inherent in this group. By examining only those patients who were explored surgically, we have selected biologically responsive tumors. Nonetheless, these patients would not be offered resection and the chance for better survival without these strategies. These data support the safety, feasibility, and efficacy of using neoadjuvant therapy and vascular resection to treat patients with locally advanced pancreatic cancer. We focused on patients with ductal adenocarcinoma in the head of the pancreas to inform and guide perioperative decision making for future patients with similar tumors. Further work in prospective trials is necessary to evaluate specific neoadjuvant therapy regimens and to define clinical and histopathologic characteristics of locally advanced tumors that might be more responsive to neoadjuvant therapy and subsequent resection. At this time, given our findings, it would be reasonable to suggest that any patients with borderline resectable disease should be enrolled in a protocol of neoadjuvant chemotherapy ± radiation; we know pancreatic cancer is a systemic disease, often micrometastatic at diagnosis, and this approach gives the patient the best possible chance of a complete surgical resection with negative margins and nodes. At present we are conducting two phase II trials that have nearly completed enrollment. Both are examining the neoadjuvant protocol of GTX alone versus GTX in conjunction with abdominal radiation. We hope that the data from these studies will inform future prospective, randomized trials. The shifts in survival of neoadjuvant patients in the present series beg the question, should all patients be treated with neoadjuvant therapy prior to resection? A trial that randomizes patients with resectable tumors to either neoadjuvant therapy followed by surgery versus surgery followed by adjuvant treatment might provide the answer.

Conclusions Many patients with locally advanced pancreatic cancer can be brought to resection through neoadjuvant therapy and

World J Surg (2014) 38:1184–1195

vascular resection with acceptable morbidity and mortality. The risk of these strategies is balanced by the opportunity for significantly improved survival over patients deemed locally inoperable by traditional criteria. The strategy of neoadjuvant therapy followed by PD with or without vascular resection is a promising approach for patients with locally advanced cancer in the pancreatic head that warrants further investigation in prospective, randomized trials. Acknowledgments This work was supported with financial aid from the I. W. Foundation.

References 1. Jemal A, Siegel R, Xu J et al (2010) Cancer statistics. CA Cancer J Clin 60:277–300 2. Raimondi S, Masionneuve P, Lowenfels AB (2009) Epidemiology of pancreatic cancer: an overview. Nat Rev Gastroenterol Hepatol 6:699–708 3. Arvold ND, Ryan DP, Niemierko A et al (2011) Long-term outcomes of neoadjuvant chemotherapy before chemoradiation for locally advanced pancreatic cancer. Cancer 18:3026–3035 4. Wanebo HJ, Glicksman AS, Vezeridis MP et al (2000) Preoperative chemotherapy, radiotherapy, and surgical resection of locally advanced pancreatic cancer. Arch Surg 135:81–87 5. Chua TC, Saxena A (2010) Extended pancreaticoduodenectomy with vascular resection for pancreatic cancer: a systematic review. J Gastrointest Surg 14:1442–1452 6. Martin RCG, Scoggins CR, Egnatashvili V et al (2009) Arterial and venous resection for pancreatic adenocarcinoma: operative and long-term outcomes. Arch Surg 144:154–159 7. Jia Y, Wang TJC, Allendorf J et al (2012) Management of borderline resectable pancreatic adenocarcinoma: highlights from the 2012 ASCO gastrointestinal cancers symposium. JOP 13:147–150 8. Tucker ON, Rela M (2008) Controversies in the management of borderline resectable proximal pancreatic adenocarcinoma with vascular involvement. HPB Surg. doi:10.1155/2008/ 839503 9. Evans DB, Farnell MB, Lillemoe KD et al (2009) Surgical treatment of resectable and borderline resectable pancreas cancer: expert consensus statement. Ann Surg Oncol 16:1736–1744 10. Allendorf JD, Lauerman M, Bill A et al (2008) Neoadjuvant chemotherapy and radiation for patients with locally unresectable pancreatic adenocarcinoma: feasibility, efficacy, and survival. J Gastrointestinal Surg 12:91–100 11. Katz MHG, Pisters PWT, Evans DB et al (2008) Borderline resectable pancreatic cancer: the importance of this emerging stage of disease. J Am Coll Surg 206:833–848

1195 12. Tzeng CWD, Fleming JB, Lee JE et al (2012) Defined clinical classifications are associated with outcome of patients with anatomically resectable pancreatic adenocarcinoma treated with neoadjuvant therapy. Ann Surg Oncol 19:2045–2053 13. Chun YS, Milestone BN, Watson JC et al (2010) Defining venous involvement in borderline resectable pancreatic cancer. Ann Surg Oncol 17:2832–2838 14. Pancreatic adenocarcinoma version 2 (2012) NCCN Clinical Practice Guidelines in Oncology, 2012. http://www.nccn.org/pro fessionals/physician_gls/f_guidelines.asp#pancreatic. Accessed 15 Oct 2013 15. DeOliveira ML, Winter JM, Schafer M et al (2006) Assessment of complications after pancreatic surgery: a novel grading system applied to 633 patients undergoing pancreaticoduodenectomy. Ann Surg 244:931–939 16. Bassi C, Dervenis C, Butturini G et al (2005) Postoperative pancreatic fistula: an international study group (ISGPF) definition. Surgery 138:8–13 17. Schnelldorfer T, Ware AL, Sarr MG et al (2008) Long-term survival after pancreaticoduodenectomy for pancreatic adenocarcinoma: is cure possible? Ann Surg 247:456–462 18. Assifi MM, Lu X, Eibl G et al (2011) Neoadjuvant therapy in pancreatic adenocarcinoma: a meta-analysis of phase II trials. Surgery 150:466–473 19. Ramacciato G, Mercantini P, Petrucciani N et al (2009) Does portal-superior mesenteric vein invasion still indicate irresectability for pancreatic carcinoma? Ann Surg Oncol 16:817–825 20. Barugola G, Partelli S, Crippa S et al (2012) Outcomes after resection of locally advanced or borderline resectable pancreatic cancer after neoadjuvant therapy. Am J Surg 203:132–139 21. Bockhorn M, Burdelski C, Bogoevski D et al (2011) Arterial en bloc resection for pancreatic carcinoma. Br J Surg 98:86–92 22. Mollberg N, Rahbari NN, Koch M et al (2011) Arterial resection during pancreatectomy for pancreatic cancer: a systematic review and meta-analysis. Ann Surg 254:882–893 23. Tseng JF, Raut CP, Lee JE et al (2004) Pancreaticoduodenectomy with vascular resection: margin status and survival duration. J Gastrointest Surg 8:935–950 24. Yekebas EF, Bogoevski D, Cataldegirmen G et al (2008) En bloc vascular resection for locally advanced pancreatic malignancies infiltrating major blood vessels: perioperative outcome and longterm survival in 136 patients. Ann Surg 247:300–309 25. Zhou Y, Zhang Z, Liu Y et al (2012) Pancreatectomy combined with superior mesenteric vein-portal vein resection for pancreatic cancer: a meta-analysis. World J Surg 36:884–891. doi:10.1007/ s00268-012-1461-z 26. Kang CM, Chung YE, Park JY et al (2012) Potential contribution of preoperative neoadjuvant concurrent chemoradiation therapy on margin-negative resection in borderline resectable pancreatic cancer. J Gastrointest Surg 16:509–517 27. Morganti AG, Massaccesi M, La Torre G et al (2010) A systematic review of resectability and survival after concurrent chemoradiation in primarily unresectable pancreatic cancer. Ann Surg Oncol 17:194–205

123