World Journal of Surgery

, Volume 38, Issue 8, pp 2132–2137

Is Perineural Invasion More Accurate Than Other Factors to Predict Early Recurrence after Pancreatoduodenectomy for Pancreatic Head Adenocarcinoma?

Authors

  • Thibaut Fouquet
    • Department of HPB SurgeryUniversity Hospital Nancy
  • Adeline Germain
    • Department of HPB SurgeryUniversity Hospital Nancy
  • Laurent Brunaud
    • Department of HPB SurgeryUniversity Hospital Nancy
  • Laurent Bresler
    • Department of HPB SurgeryUniversity Hospital Nancy
    • Department of HPB SurgeryUniversity Hospital Nancy
Article

DOI: 10.1007/s00268-014-2465-7

Cite this article as:
Fouquet, T., Germain, A., Brunaud, L. et al. World J Surg (2014) 38: 2132. doi:10.1007/s00268-014-2465-7

Abstract

Background

Some patients operated by pancreaticoduodenectomy for resectable pancreatic head adenocarcinoma will present with a recurrence during the first year (early recurrence).

Objective

The aim of this study was to determine prognostic factors associated with early recurrence in a large retrospective study.

Methods

From January 1995 to November 2010, all patients operated by pancreaticoduodenectomy for pancreatic head adenocarcinoma in our institution were retrospectively included. Univariate and multivariate analyses were performed to determine factors associated with early recurrence.

Results

A total of 166 patients were included; 57 patients (34 %) developed early recurrence. In univariate analysis, factors associated with early recurrence were perineural invasion (p = 0.0002), preoperative bilirubin (p = 0.01), lymph node ratio (LNR) ≥0.2 (p = 0.009), and T stage (p = 0.02). In multivariate analysis, perineural invasion (odds ratio [OR] 3.31; 95 % confidence interval [CI] 1.42–7.72; p = 0.005), LNR ≥0.2 (OR 2.55; 95 % CI 1.17–5.52; p = 0.02), and preoperative bilirubin (OR 1.04; 95 % CI 1.01–1.07; p = 0.03) were independent factors associated with early recurrence. Perineural invasion was also associated with poor overall survival (p = 0.001) and poor disease-free survival (p = 0.07).

Conclusion

In our study, perineural invasion (OR 3.31) is more accurate than T stage and lymph node status (OR 2.55) to predict early recurrence after pancreatoduodenectomy for pancreatic head adenocarcinoma.

Introduction

Although several advances have been made in the treatment of pancreatic head adenocarcinoma, the prognosis remains dismal, with a 5-year survival rate of <5 % for all stages [1]. For resected head adenocarcinoma, the 5-year survival rate can reach 10–29 % [13]. This rate increased from 8 % in the 1980s to 25 % more recently [4, 5]. Surgery remains the only curative treatment but neo-adjuvant or adjuvant therapies are known to increase survival [68].

Some of these resected patients will present with early recurrence during the first year (early recurrence) as described by Zhang et al. [9]. This early recurrence is not known to impact survival of these patients for the time being, but we assume that survival is probably affected in cases of early recurrence. Furthermore, predictive factors for early recurrence are not known. Many authors have focused on prognostic factors, including biological values, histological characteristics, using extended lymphadenectomy, use of neo-adjuvant or adjuvant therapies, or postoperative morbidity to explain recurrence. The implications of those different factors are not clearly identified.

The main aim of this study was to evaluate factors associated with an early recurrence in a large cohort and the second aim was to evaluate factors associated with poor disease-free and overall survival.

Materials and methods

Patient database

From January 1995 to November 2010, all patients operated on by pancreaticoduodenectomy for pancreatic head adenocarcinoma were included retrospectively in our database. Patients with distant metastasis, peritoneal carcinomatosis, or arterial invasion (hepatic artery, celiac axis, or superior mesenteric artery) were excluded.

Preoperative data included age, past medical history, weight, tumor size, need for biliary stenting, bilirubin level (the day before surgery), and CA 19-9 value. Operative data included operative time, blood loss, the need for portal vein resection, texture of the pancreas, and type of pancreato-enteric anastomosis.

Preoperative staging

All patients included in this study underwent a computed tomography (CT) scan for diagnosis and staging before surgery.

Operative procedures

All patients underwent non-pylorus-preserving pancreaticoduodenectomy. Locoregional lymphadenectomy was performed routinely. The type of pancreato-enteric anastomosis was decided during the procedure by the surgeon. Patients who had a pancreas with soft texture and narrow pancreatic duct underwent a pancreato-gastrostomy. In the other cases, duct to mucosa pancreatojejunostomy in which a suture was placed between the main pancreatic duct and the jejunum, or a pancreatojejunostomy in which the entire pancreas was anastomosed to the jejunum was performed. An abdominal drain was routinely placed during the procedure next to the pancreatico-enteric anastomosis.

Histological characteristics

Histological data included tumor size, tumor differentiation, total number of lymph nodes, number of involved lymph nodes, margin invasion, perineural invasion, and vascular invasion. The lymph node ratio (LNR) is the ratio of the lymph nodes involved to the total number of lymph nodes. According to the literature, the cut-off used for this study was 0.2 [10].

All specimens were analyzed in the same histopathological department in our institution by two senior pathologists during the study period. Every specimen was colored using hematoxylin–eosin–safran, and immunohistochemistry was carried out (cytokeratin AE1–AE3 antibody). Perineural invasion was defined by the presence of tumor cells in the perineural space as shown in Fig. 1. Since 2009, criteria for R0 resections have been modified according to the Verbeke and Menon definition [11], and all specimens were systematically analyzed.
https://static-content.springer.com/image/art%3A10.1007%2Fs00268-014-2465-7/MediaObjects/268_2014_2465_Fig1_HTML.jpg
Fig. 1

Perineural invasion in immunohistochemistry (cytokeratin AE1–AE3 antibody)

Postoperative data

Postoperative morbidity was classified using the Clavien–Dindo classification [12] and was regrouped as minor (grades 1 and 2) and major (grades 3 and 4) morbidity. Postoperative fistula was classified using the International Study Group of Pancreatic Fistula classification [13]. Delayed gastric emptying was diagnosed using the International Study Group of Pancreatic Surgery classification [14].

Follow-up

All patients underwent clinical examination, blood test for CA 19-9 value, and CT scan every 6 months or earlier according to symptoms. Date of recurrence was the date when the CT scan diagnosed abnormal nodes, local recurrence, or distant metastasis.

Statistical analysis

Patients were divided into two groups. Group 1 included patients who developed recurrence within 1 year. Group 2 included patients without early recurrence. Consequently, patients who developed recurrence after 1 year and patients without recurrence were included in group 2. The two groups were compared using univariate analysis by the χ2 test for qualitative variables and the Wilcoxon test for quantitative variables. Variables considered as significantly different (p < 0.2) between the two groups were included in a stepwise multivariate analysis to identify independent predictive factors of early recurrence by Forward’s variable selections method.

Data were analysed using the SAS v9.2 program (SAS Institute, Inc., Cary, NC, USA).

Results

Clinical analysis

We included 166 patients in our study. Clinicopathological characteristics of the patients and data concerning histology and postoperative morbidity are shown in Table 1.
Table 1

Study population

Characteristics

  

Patient characteristics

 Number of patients

166

 

 Female/male

66/100

39.8/60.2 %

 Mean age

63

19–86

 Preoperative diabetes

30

18.1 %

 Preoperative bilirubin blood level (μmol/L)

103.8

17–631

 Preoperative biliary stenting

44

26.5 %

Intraoperative data

 Mean operative time (min)

267

132–570

 Portal vein resection

14

8.4 %

 Pancreaticogastrostomy

30

18.1 %

 Pancreaticojejunostomy

55

33.1 %

 Wirsungo jejunostomy

81

48.8 %

Histological characteristics

 Mean tumoral size (cm)

3.2

1–11

 Perineural invasion

133

81.1 %

 R0 resection

139

83.7 %

 Lymph node invasion (N+)

107

73 %

 LNR ≥0.2

76

45.8 %

Postoperative data

 Minor complication (Clavien 1 and 2)

39

23.5 %

 Major complication (Clavien 3 and 4)

30

18.1 %

 Postoperative fistula

38

22.9 %

 Delayed gastric emptying

52

31.3 %

 Mortality (30 days)

11

6.6 %

Major complications occurred in 30 patients (18.1 %) (Clavien 3b–4). Eleven patients (6.6 %) died within 30 days (Clavien 5): four (2.4 %) of medical causes and seven (4.2 %) of surgical causes. Reoperation was necessary during the same hospital stay for 31 patients (18.7 %): 18 (10.8 %) for pancreatic fistula, 9 (5.4 %) for intra-abdominal hemorrhage, 2 (1.2 %) for mesenteric ischemia, and 2 (1.2 %) for intestinal obstruction.

Early recurrence within 1 year occurred in 57 patients. In univariate analysis, factors associated with early recurrence were as follows: preoperative bilirubin blood level (p = 0.0126), perineural invasion (p = 0.0002), LNR ≥0.2 (p = 0.009), T stage (p = 0.021), and presence of metastatic lymph nodes (p = 0.007). Age of patients, need for preoperative biliary stenting, mean operative time, need for portal vein resection, type of pancreatic anastomosis, tumor size, tumor differentiation, postoperative morbidity, R0 versus R1/R2 resection, and the use of adjuvant therapies did not differ in these two groups (Table 2).
Table 2

Comparison of the two groups in univariate analysis

Characteristics

No recurrence during the first year (N = 109)

Early recurrence (N = 57) 34 %

p value

Men/women (%)

61.5/38.5

57.9/42.1

0.65

Age

62.5 ± 11.9

63 ± 10.3

0.79

Jaundice

76 (69.7)

47 (82.5)

0.07

Preoperative bilirubin blood level (μmol/L)

81.8 ± 102.6

146.8 ± 150.2

0.0126

Preoperative biliary stent

26 (23.9)

18 (31.6)

0.28

Mean operative time (min)

266.6 ± 93.9

268.1 ± 75.5

0.48

Portal vein resection

9 (8.3)

5 (8.8)

0.91

Diameter of the Wirsung

  

0.18

 <5 mm

53 (49.5)

22 (38.6)

 

 >5 mm

54 (50.5)

34 (61.4)

 

Anastomosis

  

0.31

 DTMPJ

49 (45)

31 (54.4)

 

 PJ

37 (33.9)

19 (33.3)

 

 PG

23 (21.1)

7 (12.3)

 

Tumoral size (cm)

3.3 ± 1.856

3.1 ± 1.1

0.82

Differentiation

  

0.08

 Poor

14 (14.6)

8 (14.3)

 

 Moderate

47 (49)

37 (66.1)

 

 Well

33 (36.5)

11 (19.6)

 

Angioinvasion

33 (56.9)

28 (77.8)

0.039

Perineural invasion

53 (48.6)

45 (78.9)

0.0002

R1–R2 resection

14 (13)

12 (21.1)

0.39

T score

  

0.021

 T1

12 (11.3)

1 (1.8)

 

 T2

27 (25.5)

9 (15.8)

 

 T3

50 (47.2)

40 (70.2)

 

 T4

17 (16)

7 (12.3)

 

Presence of metastatic lymph nodes

63 (65.6)

48 (85.7)

0.007

Lymph node ratio ≥0.2

42 (38.5)

34 (59.6)

0.009

Anastomotic leakage

28 (25.9)

9 (15.8)

0.27

Delayed gastric emptying

32 (30.2)

17 (29.8)

0.96

Postoperative hemorrhage

16 (14.7)

5 (8.8)

0.28

Postoperative morbidity

  

0.49

 Minor

22 (22.4)

17 (30.4)

 

 Major

21 (21.4)

9 (16.1)

 

Adjuvant chemotherapy

35 (32.1)

24 (42.1)

0.2

Data are presented as N (%) or mean ± standard deviation unless otherwise indicated

In multivariate analysis (Table 3), factors associated with early recurrence were perineural invasion (odds ratio [OR] 3.31; 95 % confidence interval [CI] 1.42–7.72; p = 0.0047), LNR ≥0.2 (OR 2.55; 95 % CI 1.17–5.52; p = 0.021), and preoperative bilirubin blood level (OR 1.04; 95 % CI 1.01–1.07; p = 0.0258).
Table 3

Multivariate analysis: independent risk factors of early recurrence

 

Multivariate analysis

HR

95 % CI

p value

Perineural invasion

3.31

1.42–7.72

0.0047

LNR ≥0.2

2.55

1.17–5.52

0.021

Preoperative bilirubin blood level

1.04

1.01–1.07

0.0258

CI confidence interval, HR hazard rate

Survival

Mean follow-up was 22 months (range 1–125). Overall 1-, 3-, and 5-year survival rates were 75, 43, and 27 %, respectively. Median survival was 19 months (range 1–125).

In univariate analysis (Table 4), factors associated with poor overall survival were portal vein resection (p = 0.03), blood loss (p = 0.0007), postoperative major morbidity (p = 0.02), LNR ≥0.2 (p = 0.0003), and perineural invasion (p = 0.000005). In multivariate analysis, factors associated with poor overall survival were perineural invasion (OR 2.77; 95 % CI 1.4–5.26; p = 0.001), and LNR ≥0.2 (OR 2.38; 95 % CI 1.4–4.0; p = 0.001).
Table 4

Factors associated with poor overall survival

Variables

Univariate analysis

Multivariate analysis

OR

95 % CI

p value

Gender

NS

   

Age

NS

   

Operative time

NS

   

Portal vein resection

0.03

   

Blood loss

0.0007

   

Tumor size

NS

   

Differentiation

NS

   

Angioinvasion

NS

   

Perineural invasion

0.000005

2.77

1.4–5.26

0.001

LNR ≥0.2

0.0003

2.38

1.4–4.0

0.001

Postoperative major morbidity

0.02

   

CI confidence interval, NS not significant, OR odds ratio

The 1-, 3-, and 5-year disease-free survival rates were 73, 39, and 25 %, respectively.

In univariate analysis (Table 5), factors associated with poor disease-free survival were portal vein resection (p = 0.01), blood loss (p = 0.0001), angioinvasion (p = 0.04), LNR ≥0.2 (p = 0.00004), and perineural invasion (p = 0.000003). In multivariate analysis, the only factor associated with poor disease-free survival was perineural invasion (OR 2.77; 95 % CI 1.31–5.88; p = 0.007).
Table 5

Factors associated with poor disease-free survival

Variables

Univariate analysis

Multivariate analysis

OR

95 % CI

p value

Gender

NS

   

Age

NS

   

Operative time

NS

   

Portal vein resection

0.01

   

Blood loss

0.0001

   

Tumor size

NS

   

Differentiation

NS

   

Angioinvasion

0.04

   

Perineural invasion

0.000003

2.77

1.31–5.88

0.007

LNR ≥ 0.2

0.00004

   

Postoperative major morbidity

NS

   

CI confidence interval, LNR lymph node ratio, NS not significant, OR odds ratio

The 1-, 3-, and 5-year overall survival rates of patients with perineural invasion were 71, 24, and 18 %, respectively, which is worse than the 77, 70, and 49 %, respectively, for patients without perineural invasion (p < 0.0001). The 1-, 3-, and 5-year disease-free survival rates for patients with perineural invasion were 69, 20, and 11 %, respectively, while they were 77, 67, and 47 %, respectively, for patients without perineural invasion (p < 0.0001).

Discussion

Our study included a large cohort of patients undergoing a pancreatoduodenectomy for pancreatic head adenocarcinoma. All patients had been treated in a tertiary referral center within a period of 15 years. Only patients with final diagnosis of primary pancreatic adenocarcinoma were included. We focused on early recurrence, overall survival, and disease-free survival. To our knowledge, this study is one of the largest studies evaluating these three endpoints.

T score is known to be an important prognostic factor for survival [15]. Some authors have demonstrated that TNM classification is not efficient enough as it does not take into account histological factors like lymphovascular invasion, perineural invasion, and extra pancreatic invasion [16]. In our study, perineural invasion appears to be more accurate than the T score classification to predict early recurrence and survival.

Perineural invasion is known to be very common in the pancreas head carcinoma [17]. This is due to acute neurotropism of pancreatic cells and the particular innervation of the gland. Some authors have demonstrated that perineural invasion is the first step before nodal invasion [18]. Others have described perineural invasion and lymphovascular invasion as two strong prognostic factors for survival [19]. Few authors have studied perineural invasion as a prognostic factor for early recurrence. Shimada et al. [20] concluded that perineural invasion was the main factor associated with disease-free survival.

To our knowledge, our study is the first in which perineural invasion is described as the main prognostic factor for early recurrence, disease-free survival, and overall survival. Since 2007, every patient who had a final pathological diagnosis of pancreatic head adenocarcinoma benefited from adjuvant chemotherapy with gemcitabine [6]. In the CONKO study published in 2007, the authors demonstrated that the use of an adjuvant chemotherapy with gemcitabine in resected pancreas head adenocarcinoma for all stages of cancer dramatically improved median free-survival compared with observation alone [6]. This adjuvant chemotherapy remains the standard of care today. Adjuvant chemotherapy had been described as an important factor for delaying recurrence after complete resection of pancreatic adenocarcinoma [21]. Our results have not changed our practice after this surgery for the time being. However, we can hypothesize that, in the future, patients with perineural invasion may benefit from more aggressive chemotherapy like folfirinox [22].

Lymph node invasion is known to affect survival [23]. LNR, in our study, is a stronger prognostic factor than the presence of metastatic lymph nodes. Slidell et al. [10] demonstrated that patients with fewer than 11 nodes analyzed had worse survival than patients with more than 11 nodes analyzed (16 vs. 23 months; p < 0.0001). More recently, the LNR was described as a stronger prognostic factor than the total number of lymph nodes involved. Two recent series described this ratio and evaluated different levels (0, 0.1, 0.2, and 0.3) [10, 23]. The most discriminate cut-off appears to be 0.2. In our study, we demonstrated that a LNR >0.2 is strongly correlated with early recurrence and overall survival.

This study suffers from bias. First, this is a retrospective study that includes changes in per-operative practices (portal vein resection, standardization of pancreatic anastomosis technic). Second, our study included patients over a long period (15 years) and only patients treated after 2007 benefited from systematic adjuvant chemotherapy with gemcitabine. This represents 36 % of our cohort. It could be the reason why we found no implication of adjuvant chemotherapy in early recurrence, disease-free, and overall survival. Third, we report a R0 rate of 83.7 %, which is probably overestimated. Since 2010, all specimens are inked in order to help pathologists, and R1 was defined by the presence of tumor cells at least 1 mm from the surgical margins, which is known to decrease R0 rate as reported by Verbeke and Menon [11]. We overestimated the rate of R0 resection, which can also probably explain why the R0 or R1/R2 status has no impact on early recurrence in our study.

In our study, perineural invasion is the main factor associated with recurrence within 1 year (early recurrence) and the main factor associated with poor disease-free and overall survival. This factor is more accurate than T score and lymph node status in predicting early recurrence, and disease-free and overall survival after pancreaticoduodenectomy for resected pancreas head adenocarcinoma.

Copyright information

© Société Internationale de Chirurgie 2014