Pancreatic fistula after a pancreaticoduodenectomy for ductal adenocarcinoma and its association with morbidity: a multicentre study of the French Surgical Association

Abstract

Backgrounds

A pancreatic fistula (PF) is the most relevant complication after a pancreaticoduodenectomy (PD). This retrospective multicentric study attempts to elucidate the risk factors and complications of a PF in a large cohort of patients undergoing a PD for ductal adenocarcinoma.

Methods

Using a survey tool, clinical data of 1325 patients undergoing a PD for ductal adenocarcinoma at 37 institutions, between January 2004 and December 2009, were collected. Peri-operative risk factors associated with PF and its association with morbidity and mortality were assessed. Morbidity and PF were graded according to the ISGPF (International Study group for pancreatic fistula) definition and the Dindo–Clavien classification.

Results

Overall PF, mortality, morbidity and relaparotomy rates were 14.3%, 3.8%, 54.4% and 11.7%, respectively. PF occurred more frequently after a pancreaticojejunostomy (PJ) compared with a pancreaticogastrostomy (PG) (16.8% vs. 10.4%; P = 0.0012). Independent risk factors for PF by multivariate analysis were absence of pre-operative diabetes (P = 0.0014), PJ reconstruction (P = 0.0035), soft pancreatic parenchyma (P < 0.0001) and low-volume centre (P = 0.0286). Clinically relevant PF (grade B and C) and severe complications (Dindo–Clavien grade IIIB, IV, V) were significantly more frequent after PJ than PG (71.6% vs. 28.3%; P = 0.030 and 24.8% vs. 19.1%; P = 0.015, respectively). Overall mortality and relaparotomy rates were similar after PG and PJ.

Conclusions

A soft pancreatic parenchyma, the absence of pre-operative diabetes, PJ and low-volume centre are independent risk factors for PF after PD for ductal adenocarcinoma. A significantly higher incidence and clinical severity of PF are associated with PJ.

Introduction

In spite of more effective chemotherapy regimens, a pancreaticoduodenectomy (PD) remains the only curative option for patients presenting with resectable ductal adenocarcinoma of the pancreatic head. As a result of improvements in surgical technique and in peri-operative care, the 5-year survival rates for patients undergoing a R0 resection for localized pancreatic cancer have approached 25%.^1–5 However, in spite of improvements in long-term survival and mortality, post-operative morbidity still remains high with rates reported between 30% to 50% in large series.^6–8 Of these, a post-operative pancreatic fistula (PF) still is the clinically most relevant complication after PD with an incidence ranging from 9.9% to 28.5%.⁹ It is often associated with the development of life-threatening intra-abdominal complications such as sepsis, abscesses, early or delayed haemorrhage, the need for a relaparotomy and death.¹⁰ Known risk factors for PF are associated with patient, disease, procedure and surgeon-related factors. Among these obesity,¹¹ diabetes,^12,13 pre-operative jaundice,¹⁴ and malnutrition constitute patient-related factors whereas ampullary or duodenal disease,^15,16 high pancreatic juice output,¹⁷ soft pancreatic parenchyma,¹⁸ fatty pancreatic texture^11,19,20 and a small pancreatic duct diameter⁶ are disease-related factors. Increased intra-operative blood loss,²¹ longer operative times,¹² low volume surgeon^16,22–24 and type of pancreatico-enteric reconstruction^25–28 constitute procedure and surgeon-related factors.

Several previous studies have reported a decreased risk of PF in patients undergoing PD for ductal adenocarcinoma compared with other diseases citing the presence of obstructive jaundice, hard pancreatic parenchyma and a dilated pancreatic duct as favourable factors for better outcomes after pancreatic head resections.^{16,18,29–34} However, studies specifically focusing on post-operative outcome and risk factors for PF after PD for ductal adenocarcinoma are lacking. The aim of the present multi-centric study was to study the occurrence of PF after PD for ductal adenocarcinoma, its risk factors and association with morbidity.

Patients and methods

Patients selection

A multicentric study of PF after PD was promoted by three experienced pancreatic surgeon members (J.R.D., F.P. and P.B.) of the French Association of Surgery (Association Française de Chirurgie) in 2009. Patients who underwent a PD between 1 January 2004 and 31 December 2009 were eligible for this retrospective study. Using a standard questionnaire, pre-operative, intra-operative and post-operative data (early and late outcomes) were collected from 37 responding institutions (France, 34; Belgium, 1; Monaco, 1; and Switzerland, 1). Only patients with pancreatic ductal adenocarcinoma diagnosed on the final pathology constituted the study population. To ascertain uniform interpretation of the survey data³⁵, all data sheets were reviewed by the main study authors (J.R.D., F.P. and P.B.) before transfer to the database.

Operative technique

Several technical variations of PD, including pylorus preservation, the use of a trans-anastomotic stent and a pancreaticogastrostomy (PG) compared with a pancreaticojejunostomy (PJ) were performed according to the surgeons' preferences. The prophylactic use of somatostatin analogues varied among centres according to local guidelines. Segmental or lateral vascular resections of the superior mesenteric vein, portal vein, superior mesenteric artery and/or hepatic artery and en-bloc resection of the colon, kidney, liver and stomach were included in the study population. The extent of a lymphadenectomy depended on the surgeon and centre preferences. A standard lymphadenectomy was defined as removal of N1 lymph nodes according to the TNM-UICC 2002 classification.³⁶ An extended lymphadenectomy was defined as standard plus removal of interaortico-caval lymph nodes with circumferential clearance of the origin of the celiac trunk and superior mesenteric artery.

Definitions

In-hospital mortality and morbidity were limited to occurrences within 60 days after surgery. The severity of complications was graded using the Dindo–Clavien classification system.³⁷ A severe complication was defined as grade IIIb or above. A PF was defined using the current classification provided by the International Study group (ISGPF: grade A, B, C). Therefore a PF was defined as any measurable drainage from an operatively placed drain on or after post-operative day 3 demonstrating amylase levels greater than three times the upper limit of the normal serum amylase level.⁹ Grade B and C PF were considered clinically significant. PF were assessed using the Dindo–Clavien's classification system.³⁸ Intra-operatively, the parenchymal pancreatic texture was defined by the operating surgeon as soft, normal or fibrous at the time of the operation. According to the North American standards, high-volume centres were defined as equal or greater than 20 PDs performed per year, medium-volume centres performing between 10 and 19 PDs per year and low-volume centers performing less than 10 PDs per year.^39,40

A diagnosis of PF was obtained clinically during reoperation, radiologically and by laboratory tests as described above. Intra-abdominal collections were diagnosed by computed tomography. A digestive tract haemorrhage was defined as blood from the nasogastric tube and confirmed by endoscopy or angiography. An intra-abdominal haemorrhage was defined by the presence of blood in abdominal drains, on angiography or both. Medical complications included septic shock, deep vein thrombosis, and cardiac, pleuropulmonary, and neurologic events.

End points

The main post-operative outcomes were the presence of a PF and its association with mortality and morbidity. Secondary endpoints consisted of risk factors for post-operative PF, mortality and morbidity. They included: patient age, gender, body mass index (BMI), the presence of diabetes, the use of pre-operative biliary drainage, American Society of Anesthesiologist (ASA) class, the use of pre-operative neo-adjuvant treatment, the need for a red blood cell (RBC) transfusion, pylorus preservation, the type of pancreaticoenteric reconstruction, size of the pancreatic duct, texture of the pancreatic parenchyma, associated vascular resection, the number of lymph nodes harvested, the use of a biological glue, the use of abdominal drains, the use of pancreatic ductal stenting, associated multivisceral resections and centre volume.

Statistical analysis

Values were expressed as mean ± standard deviation (SD). The Mann-Whitney U-test, chi-square test, Fisher's exact test or Student's t-tests were used as appropriate. All variables reaching a P < 0.15 in the univariate analysis were included in multivariate analysis using a logistic regression model. Results were presented as relative risk (RR) with 95% confidence intervals. Significance was established at P < 0.05. Statistics were performed using the SPSS 17.0.1 system (SPSS Inc., Chicago, IL, USA).

Results

Patients characteristics

In total, 1325 patients underwent a PD for ductal adenocarcinoma in the period study (754 women and 571 men) with a mean age of 64.9 ± 10.1 years (range, 25–87) and a mean body mass index (BMI) of 24.3 ± 4.2 (range, 11.2–46.9). Among them, 428 (32.30%) were older than 70 years, 224 (16.90%) had diabetes mellitus, 972 (73.55%) had pre-operative jaundice and 445 (33.58%) had pre-operative biliary drainage. Pre-operatively, 142 (10.7%) patients underwent radiotherapy and/or chemotherapy with 39 patients undergoing chemotherapy only. A mesenterico-portal vein resection was performed in 360 patients (27.16%) of which 133 had a non-segmental lateral resection. An extended lymphadenectomy and pylorus-preservation was performed in 303 patients (22.86%) and 163 patients (12.3%), respectively. A pancreaticojejunostomy and PG was performed in 733 (55.3%) and 563 patients (42.5%), respectively. The type of reconstruction was not specified in 29 patients (2.2%). The pancreatic duct diameter was < 3 mm in 411 patients, 4 and 6 mm in 344 patients, 7 and 9 mm in 107 patients, >10 mm in 85 patients and not reported in 378 patients (28.5%). The pancreatic texture was reported as fibrous in 640 patients, normal in 150 patients, soft in 357 patients and not specified in 149 patients. The majority of patients underwent duct to mucosa anastomosis (n = 811) and a transanastomotic stent was used in 296 patients. In patients who underwent an associated SMV-PV resection, PJ and PG was performed in 195 patients (54.2%) and 160 patients (44.4%), respectively. In five patients (1.4%), the type of reconstruction was not specified. One or more abdominal drains were used in 1267 (95.6%) patients, none in 12 patients and not specified in 46 patients. Octreotide analogues and biological glue were used in 534 (40.3%) and 72 (5.4%) patients, respectively.

Comparison of the study population according to the pancreatic anastomosis and the texture of pancreatic parenchyma

An invaginating anastomosis rather than a duct-to-mucosa technique was used more often in PG compared with PJ (318 vs. 62, P < 0.001).The anastomotic stenting rate was similar for both PG and PJ. The rate of invagination and stenting was significantly higher in the presence of a normal or soft pancreas and a small diameter pancreatic duct (Table 1).

Table 1.

Correlation between pancreatic remnant characteristics and type of pancreaticoenteric reconstruction^a

	Invagination (n = 380)	P	Stenting (n = 296)	P
Type of anastomosis performed		0.001		NS
PG	318		128
PJ	62		168
Texture of the pancreatic remnant		0.001		0.0001
Fibrotic	185		113
Normal	114		115
Soft	45		36
Size of the pancreatic duct		<0.01		<0.05
≤3 mm	99		88
4–6 mm	75		77
7–9	29		33
≥10	28		13

^aValues are number of patients for which the data were available.

Early post-operative outcomes

Mortality

Overall mortality was 3.8% (51 patients). The main surgical causes of death were intra-abdominal bleeding (15 patients), PF (13 patients), sepsis related to an intra-abdominal collection (10 patients), digestive bleeding (7 patients) and a bilio-digestive leak (2 patients). There were no differences in mortality according to the type of pancreatico-enteric reconstruction (4.1% in PJ vs. 3.7% in PG; P > 0.7).

Morbidity

Overall morbidity was 54.4% (721 patients). A total of 1180 complications occurred in 721 patients: an abdominal abscess in 215 patients, PF in 189 patients, DGE in 172 patients, intra-abdominal bleeding in 89 patients, intestinal bleeding in 67 patients, biliary fistula in 41 patients, gastric or duodeno-jejunal fistula in 6 patients, pulmonary or cardiovascular complications in 119 patients and other complications in 282 patients. Four hundred sixteen patients had one complication and 305 patients experienced more than one complication. The overall morbidity rate did not differ according to the type of pancreatico-enteric reconstruction (52.5% after PG vs. 55.7% after PJ; P = 0.7). Severe complications, defined as grade IIIb, IV and V according to the Dindo–Clavien classification, were significantly more frequent after PJ than PG (24.8% vs. 19.1%, P = 0.0156; Table 2). This did not lead to increased readmission rates in patient undergoing PJ (PJ 11.3% vs. PG 13.0%; P = 0.35).

Table 2.

Post-operative complications after a pancreaticoduodenectomy (PD) according to the type of pancreaticoenterico recostruction performed^a

	Clavien	Pancreatic fistula		Sepsis/ collections		Intrabdominal haemorrhage		Cardio/pulmonary		Digestive haemorrhage		DGE		Others
PJ (n = 733)	Mild (I, II,IIIA)	123	85	128	95	54	19	64	39	36	21	109	105	149	117
	Severe (IIIB,IV,V)		38		33		35		25		15		4		32
PG (n = 563)	Mild (I, II,IIIA)	59	48	84	64	33	6	50	34	30	17	63	62	130	110
	Severe (IIIB,IV,V)		11		20		27		16		13		1		20
Total (n = 1296)	Mild (I, II,IIIA)	182 (14%)	133	212 (16%)	159	87 (7%)	25	114 (9%)	73	66 (5%)	38	172 (13%)	167	279 (22%)	227
	Severe (IIIB,IV,V)		49		53		62		41		28		5		52

^aValues are the number of patients for which the data were available.

Pancreatic fistula and related morbidity

A total of 189 patients experienced a PF (14.3%) of which 41 patients had a grade A fistula, 97 had a grade B fistula and 51 had a grade C fistula.⁹ The PF rate was significantly higher (123, 16.8% vs. 59, 10.4%; P < 0.001) and more severe after PJ compared with PG (71.6% vs. 28.3%; P < 0.03), respectively (Table 3). Compared with patients who did not develop PF, patients with PF had a significantly higher rate of septic complications (41.5% vs. 13.1%; P < 0.0001), DGE (20.3% vs. 12.0; P < 0.002); intrabdominal bleeding (20.3% vs. 5.0%; i < 0.0001), digestive bleeding (12.6 vs. 3.9%; P < 0.001), cardiopulmonary complications (13.8 vs. 8.5%; P < 0.02), rehospitalization (17.2% vs. 11.3%; P < 0.02) and overall mortality (13.3% vs. 2.3%; P < 0.0001).

Table 3.

Post-operative pancreatic fistula grading according to the type of pancreaticoenteric reconstruction performed

	Grade
	A	B	C
PJ (n = 733)	22	63	38
PG (n = 563)	19	29	11
Total PF (n = 189)a	41 (21.7%)	92 (48.7%)	49 (25.9%)

^aData of pancreaticoenterico reconstruction in seven patients presenting with PF was not reported.

PJ, pancreaticojejunostomy; PG, pancreaticogastrostomy; PF, pancreatic fistula.

Relaparotomy

During the first 30 post-operative days, 155 patients (11.7%) underwent a relaparotomy. Data for the cause of relaparotomy were available for 143 patients. Thirty patients had two relaparotomies, 30 had three relaparotomies and 2 had four relaparotomies in the post-operative period. A PF was the second most common cause for a relaparotomy after a haemorrhage. The relaparotomy rate for PF was lower after PG (n = 4) compared with PJ (n = 21). During reoperation for PF, a completion pancreatectomy was performed in 11 patients, redo PJ anastomosis in one, conversion from PJ to PG in one and drainage of peri-anastomotic fluid collections in 15 patients.

Analysis of risk factors

Univariate and multivariate analysis of risk factors for morbidity, mortality, PF, digestive and abdominal bleeding and DGE are shown in Tables 4 and 5. Univariate analysis showed that higher post-operative mortality was associated with female gender, ASA status III/ IV, intra-operative RBC transfusion > 2, use of a transanastomotic drain and the occurrence of a post-operative PF. Multivariate analysis showed that the occurrence of a PF, a high ASA status and intra-operative RBC transfusion > 2 units were independent risk factors for overall mortality.

Table 4.

Univariate analysis of risk factors for post-operative mortality, morbidity and pancreas-related complications^a

	Mortality (n = 51)		Morbidity (n = 721)		Pancreatic fistula (n = 189)		Intra-abdominal haemorrhage (n = 89)		Digestive haemorrhage (n = 67)		DGE (n = 172)
	Yes	P	Yes	P	Yes	P	Yes	P	Yes	P	Yes	P
Age		0.6		0.0919		0.2656		0.1132		0.6769		0.0080
≤70	32		468		120		66		46		357
>70	18		248		68		22		20		71
Gender		0.0215		0.0416		0.0669		0.2354		0.0462		0.7261
Male	14		329		93		33		21		72
Female	37		392		96		56		46		100
Centre volume		0.9778		0.0065		<0.0028		0.3691		0.2874		0.0001
≤10	19		305		97		37		27		99
10–20	22		293		32		40		32		35
>20	10		123		60		12		8		38
Body mass index		0.4156		0.6436		0.0895		0.0844		0.0651		0.3602
≤25	23		371		91		35		42		97
>25	18		220		71		32		15		51
Pre-operative diabetes		0.1671		0.6377		0.0072		0.0998		0.0776		0.5356
Absent	41		536		156		74		54		133
Present	5		119		20		10		6		27
ASA score		0.0136		0.1785		0.0605		0.0807		0.1709
I	7		126		42		20		12		38	0.9275
II	23		378		103		36		36		103
III	11		115		15		13		9		30
IV	1		2		1		1		1		1
Pre-operative biliary drainage		0.8521		0.1802		0.2087		0.5289		0.0075		0.6822
No	34		473		130		61		54		107
Yes	17		232		57		28		13		60
Neoadjuvant chemotherapy		0.4976		0.0065		0.0654		0.3678		0.0902		0.3643
No	47		659		176		82		64		157
Yes	2		62		13		7		3		15
Pylorus-preserving PD		0.6888		0.1284		0.8070		0.8169		0.0111		0.0388
No	42		618		161		76		52		159
Yes	7		98		24		10		15		13
Type of pancreato-enteric anastomosis		0.7019		0.2483		0.0012		0.2830		0.7348		0.0529
PJ	29		409		123		54		36		109
PG	20		296		59		33		30		63
Mesenterico-portal vein resection		0.0707		0.8338		0.0194		0.3270		0.1441		0.8236
No	30		517		148		67		53		122
Yes	19		194		38		20		13		48
Arterial resection		0.1503		0.9553		0.9177		0.1363		0.3099		0.0816
No	49		709		185		86		65		172
Yes	2		11		3		3		2		0
Associated multi-visceral resection procedure		0.7853		0.0663		0.0015		0.1173		0.7391		0.5673
No	47		650		162		82		64		163
Yes	3		43		18		7		3		9
Number of lymph nodes harvested		0.7052		0.3881		0.4345		0.7145		0.0307		0.0116
≤20	28		445		118		57		48		120
>20	19		258		64		32		16		49
RBC transfusion		0.0276		0.0208		0.5102		0.3142		0.0657		0.3859
≤2	10		127		153		75		52		148
>2	15		100		19		14		13		24
Texture of pancreatic remnant		0.3569		<0.0001		<0.0001		<0.0001		0.1607		0.0035
Soft	24		322		122		56		30		97
Fibrous	23		307		47		23		26		75
Pancreatic drain		0.0351		0.0326		0.4744		0.0387		0.8855		0.0097
No	3		151		32		12		11		25
Yes	14		176		50		29		14		55
Pancreatic fistula		<0.0001		<0.0001		NA		<0.0001		0.0002		0.0074
No	27		0		NA		56		47		136
Yes	24		189		NA		33		20		36
Use of biological glue		0.6654		0.7229		0.0724		0.0873		0.9337		0.0001
No	5		174		35		14		15		32
Yes	2		47		16		8		4		24
Prophylactic octreotide administration		0.1697		0.2033		<0.0001		0.4760		0.5629		0.6501
No	17		306		62		32		27		80
Yes	24		303		103		35		29		69
Use of abdominal drainage		0.5851		0.1363		0.8241		0.0693		0.6170		0.1783
No	1		20		6		0		3		3
Yes	47		692		180		85		62		169
Type of gasytro-jejunal anastomosis		0.0522		0.1441		0.0490		0.0216		0.2004		0.0017
Antecolic	12		242		52		22		17		41
Transmesocolic	38		445		119		64		40		114

^aValues are the number of patients for which the data were available.

Statistical significance indicated in bold.

DGE, delayed gastric emptying; ASA, American Society of Anesthesiologist; PD, pancreaticoduodenectomy; RBC, red blood cells;

NA, not applicable.

Table 5.

Multivariate analysis of risk factors for mortality, morbidity and pancreas-related complications

	Relative risk (95% CI)	P-value
Mortality
Pancreatic fistula (yes vs. no)	5.43 (2.70–10.89)	<0.0001
ASA status (III/IV vs. I/II)	3.33 (1.61–6.90)	0.001
Number of RBC (>2 vs. <2)	4.45 (2.15–9.19)	<0.0001
Morbidity
Gender (male vs female)	1.50 (1.19–1.90)	0.0009
Number of RBC (>2 vs. <2)	1.47 (1.03–2.11)	0.034
Centre volume (<20 vs. >20)	1.39 (1.09–1.78)	0.0077
Texture of pancreatic parenchyma (normal vs. fibrous)	1.71 (1.34–2.18)	<0.0001
Pancreatic fistula
Diabetes mellitus (absent vs. present)	2.64 (1.43–4.78)	0.0014
Pancreatic anastomosis (PJ vs. PG)	1.77 (1.20–2.61)	0.0035
Texture of parenchyma (normal vs. fibrous)	4.43 (3.00–6.55)	<0.0001
Centre volume (<10 vs. >10)	1.92 (1.07–3.45)	0.0286
Octreotide administration (yes vs. no)	1.76 (1.22–2.55)	0.0023
Intrabdominal haemorrhage
Pancreatic fistula (yes vs. no)	6.36 (2.81–14.41)	0.0001
Texture of parenchyma (normal vs. fibrous)	4.14 (1.45–11.78)	0.0076
Type of GJ anastomosis (antecolic vs. transmesocolic)	3.31 (1.042–10.52)	0.0423
Digestive haemorrhage
Pancreatic fistula (yes vs. no)	3.40 (1.98–5.83)	<0.0001
Pre-operative biliary drainage (non vs. yes)	1.80 (1.00–3.22)	0.0467
Number of lymph node harvested (<20 vs. >20)	1.79 (1.00–3.20)	0.0491
Pylorus preservation (yes vs. no)	2.29 (1.25–4.19)	0.0073
Delayed gastric emptying
Centre volume (<10 vs. >10)	1.70 (1.13–2.55)	0.0107
Age (>70 vs. <70 years)	1.62 (1.17–2.23)	0.0031
Pylorus preservation (no vs. yes)	1.90 (1.06–3.42)	0.0297
Pancreatic fistula (yes vs. no)	1.80 (1.21–2.67)	0.0034

CI, confidence interval; RBC, red blood cells.

Univariate analysis showed that higher post-operative overall morbidity was related to male gender, PDs performed in a low volume centre, the use of pre-operative neo-adjuvant treatment, a intra-operative RBC transfusion > 2 units, normal pancreatic texture and the use of a pancreatic ductal stent. At multivariate analysis male gender, a RBC transfusion > 2 units, normal pancreatic parenchyma and a low-volume centre were independent risk factors for overall morbidity.

Univariate analysis showed that a higher risk of PF was conferred by the absence of pre-operative diabetes, PJ reconstruction, a low-volume centre,-associated resection of the mesenterico-portal axis or a multivisceral resection and soft texture of the pancreatic parenchyma. There was a trend towards a higher PF rate in men, in patients with a BMI > 25 and in patients receiving pre-operative neo-adjuvant treatment. At multivariate analysis, the presence of a soft pancreatic parenchyma, PJ reconstruction, the absence of pre-operative diabetes, administration of somatostatin analogues and a low-volume centre were independent risk factors for PF.

Multivariate analysis showed that PF was associated with a higher risk of mortality, morbidity, abdominal and intestinal bleeding and DGE. (Table 5)

Discussion

This multicentric retrospective study in 1325 patients revealed the key independent risk factors for the development of a PF after PD for adenocarcinoma of the pancreas. The occurrence of a PF after PJ was associated with a greater clinical severity and a higher rate of major post-operative morbidity compared with PG. PD continues to be a challenging and substantially morbid procedure. The overall mortality and morbidity rates observed in this study are in the range of what is reported in several single centre and multicentre studies.^4,41–43

A PF remained the leading cause of lethal complications in this study. The majority of deaths were directly attributable to PF associated with intra-abdominal, intestinal bleeding and sepsis. Patient selection and surgical techniques devoted to minimize intra-operative bleeding remain important factors in minimizing the peri-operative mortality from this operation. Our multivariate analysis for mortality found that PF, a high ASA class (III–IV) and increased transfusion rate were independent risk factors. These findings suggest that patient selection is key in balancing the risk and benefits of PD in patients with adenocarcinoma.

An obvious relationship exists between PF and the occurrence of other complications such as DGE, bleeding and sepsis.^18,44,45 Our study supports this strong relationship and identified PF as an independent risk factor for the occurrence of post-operative intra-abdominal, intestinal bleeding and DGE.

In the present study, the investigators selected a homogenous group of patients exclusively operated on for ductal adenocarcinoma of the pancreatic head to better evaluate the risk factors and the association of PF after PD. The pancreatic remnant characteristics, which are directly correlated with underlying pancreatic disorder influences the early post-operative outcome, in particular the incidence and severity of PF formation. It has been shown that patients undergoing PD for ampullary, biliary or neuoroendocrine tumours have most commonly an undilated pancreatic duct, thin pancreatic body and a non-fibrotic pancreatic parenchyma, conditions that greatly increase the risk of PF.¹⁵ However, PF incidence and its association with morbidity and mortality in patients uniquely presenting with pancreatic adenocarcinoma for PD is less well understood. Lin et al.¹⁸ and Winter et al.⁴ reported a PF rate of 5% in patients undergoing PD for pancreatic adenocarcinoma compared with a 15.4% to 18.4% rate in patients operated on for distal cholangiocarcinoma, duodenal carcinoma and ampullary cancer. In these reports up to 75% of patients with adenocarcinoma had pre-operative jaundice.⁴ In the present study, the PF rate after PD for pancreatic ductal adenocarcinoma was 14% which is comparable to rates recently reported by large single institutions.^46,47 The association between the presence of pre-operative diabetes and PF is still under debate. In a retrospective study, Srivasatva¹² identified diabetes as an independent risk factor for PF after PD for benign and malignant periampullary lesions. Chu et al.⁴⁸ reported similar results in patients presenting with pancreatic ductal adenocarcinoma with a higher PF rate in diabetic patients compared with non-diabetics (10.3% vs. 3.7%). Pre-operative diabetes remains an independent risk factor for PF even when stratified to age, comorbidities, BMI, pre-operative albumin level, operation type, operative time and pancreatic parenchymal quality.

Lin reported that patients without a history of diabetes were at a significantly higher risk for post-operative PF compared with diabetics patients (12.0% vs. 7.7%)¹⁸ A possible explanation for these findings was reported more recently by Mathur et al.²⁰ Patients with PF had a lower rate of diabetes (13% vs. 33%) and a significantly increased amount of pancreatic fat with less fibrosis and a smaller pancreatic duct. They concluded that patients with diabetes therefore may have less pancreatic fat and a higher amount of pancreatic fibrosis, protecting them from PF formation.²⁰ Similarly, in this study the absence of pre-operative diabetes was an independent risk factor for PF. These results supports the hypothesis that patients with diabetes more often present with a firm pancreatic texture leading to a lower PF rate after PD.

Disease-related risk factors for an increased PF rate after PD mainly include a non-dilated pancreatic duct, a soft pancreatic parenchyma and a fatty pancreatic texture.^6,11,18–20 A soft pancreatic remnant often coexists with a small pancreatic duct increasing the technical difficulties to perform a duct-to-mucosa anastomosis. The PF rate after PD in patients presenting a soft pancreatic parenchyma is approximately 25% and is significantly higher than that observed in patients presenting with a firm pancreas.^18,32 In the present study, the PF rate of 23.4% in patients presenting with normal/soft pancreatic parenchyma was significantly higher than observed in patients with a fibrous/hard pancreatic remnant confirming that also in a patient with adenocarcinoma the pancreatic texture predicts the occurrence of PF. However, still one-third of the patients in this study had a normal or soft pancreas and a pancreatic duct less than 3 mm, which might explain the relatively high rates of invagination and ductal stenting used for the pancreatico-enteric anastomosis in our study.

The benefit of PJ compared with PG after PD still remains a matter of debate. In this study, the incidence of PF was higher after PJ compared with PG. Over the past 20 years, there have been four prospective randomized trials comparing PJ and PG. While three of these studies have reported no difference in the PF rate, one demonstrated a significantly lower rate of PF after PG than PJ (4% vs. 18%).^16,49–51 However, a multitude of anastomotic techniques and definitions for PF were used.⁵⁰ Much of the actual evidence supporting PG over PJ comes from single institutional retrospective studies that use exclusively telescoping or invaginating techniques.^{25–28,33,52,53} Several non-randomized comparative studies have shown a significantly lower PF and relaparotomy rates after invaginated PG compared with PJ.^{25–28,33,52} A meta-analysis comparing PG and PJ showed a significant decrease in the incidence of PF and mortality for PG.⁵⁴ In our study, PJ was identified as an independent risk factor for PF after PD. While there was no correlation between a particular anastomotic technique for PG and the reported consistency of the pancreas, however, invagination was the most common technique used for PG. This may explain the observed lower rate of PF in patients undergoing PG in our study. We also observed that PF occurring after PG was associated with a significantly lower grade and significantly less severe complications than after PJ. This may be explained by the improved ability to manage PF after PG owing to the fact that nasogastric tube suction can be applied to reduce fistula output and the absence of enterokinase in the stomach, preventing activation of pancreatic enzymes. Aranha et al.³³ reported a shorter and more favourable evolution in PF after PG compared with PJ. Oussoultzoglou et al.²⁸ reported a significant reduction in relaparotomy and completion pancreatectomy rates with PG over PJ. Munoz-Bongrand et al.³⁴ reported that conservative management of PF after PG was successful in more than 75% of patients. It has been shown that PF after PJ increased the use of interventional or operative management compared with PJ.^28,33 In this study, the overall relaparotomy rate did not differ between PJ and PG. However, because the study is underpowered in this regard and there was a lack of uniformity in centre-related policies concerning relaparotomy for PF a firm conclusion cannot be obtained.

PD remains a challenging procedure characterized by a steep learning curve. Previous studies suggest that after 60 PDs surgeons improved their peri-operative outcomes in terms of blood loss, operative time, length of hospital stay and margin status.²⁴ Recently, Pecorelli et al.⁵⁵ reported that an individual surgeon's volume is directly correlated with an increase in the PF rate. However, low-volume surgeons in high-volume institutions have morbidity and mortality rates similar to high-volume surgeons, demonstrating the ability of experienced centres to anticipate and manage PF-related complications.⁵⁵ Although we did not evaluate individual surgeons volume in this study, we found that centres including less than 20 PD for adenocarcinoma per year experienced a higher overall morbidity after PF. In addition, low-volume centres had a statistically significant higher rate of PF and DGE. These findings are in concordance with previous reports demonstrating improvement in patient outcomes for high-volume centres with decreased rates of clinically relevant PF and mortality.^22,56

This study's limitations deserve commentary. First, its retrospective design limits the accuracy of our conclusions. Second, the type of surgical technique used and of the ability and strategies to manage patients' complications reflects the diversity in centre and surgeon expertise and volume. However, our data reflect the common practice in a large European country where pancreatic surgery is performed in both academic and community settings. It shows that the technique of PD has been well standardized and mortality and morbidity are improved over historical data. Third, the larger proportion of patients undergoing PJ and lack of randomization may have contributed to the increased number of complications detected in this subgroup of patients. Lastly, the inclusion of low-, medium- and high-volume centres might have introduced a bias towards higher overall and PF-related morbidity rates. The large cohort of patient and completeness of the collected data support the strength of this study. The results of the present analysis will hopefully lead to a prospective randomized study with the ultimate goal of a centralized national programme for pancreatic surgery.

Conclusions

The present French multicentric retrospective survey on PD for adenocarinoma found that a soft pancreatic parenchyma, the absence of pre-operative diabetes, PJ and low volume centres were independent risk factors for PF. The incidence, the clinical severity as well as post-operative major complications were significantly higher after PJ compared with PG. This study confirms the relationship between PF and the occurrence of major post-operative complications in this select group of patients and the association between low-volume centres and higher overall and PF-related morbidity.

Conflicts of interest

None declared.