A simple scoring system based on clinical factors related to pancreatic texture predicts postoperative pancreatic fistula preoperatively

Abstract

Background:

Postoperative pancreatic fistula (POPF) is regarded as the most serious complication of pancreatic surgery. The preoperative risk stratification of patients by simple means is of interest in perioperative clinical management.

Methods:

Based on prospective data, we performed a risk factor analysis for POPF after pancreatoduodenectomy in 62 patients operated between 2006 and 2008 with special focus on clinical parameters that might serve to predict POPF. A predictive score was developed and validated in an independent second dataset of 279 patients operated between 2001 and 2010.

Results:

Several pre- and intraoperative factors, as well as underlying pathology, showed significant univariate correlation with rate of POPF. Multivariate analysis (binary logistic regression) disclosed soft pancreatic texture (odds ratio [OR] 10.80, 95% confidence interval [CI] 1.80–62.20) and history of weight loss (OR 0.15, 95% CI 0.04–0.66) to be the only independent preoperative clinical factors influencing POPF rate. The subjective assessment of pancreatic hardness by the surgeon correlated highly with objective assessment of pancreatic fibrosis by the pathologist (r = −0.68, P < 0.001, two-tailed Spearman's rank correlation). A simple risk score based on preoperatively available clinical parameters was able to stratify patients correctly into three risk groups and was independently validated.

Conclusions:

Preoperative stratification of patients regarding risk for POPF by simple clinical parameters is feasible. Pancreatic texture, as evaluated intraoperatively by the surgeon, is the strongest single predictive factor of POPF. The findings of the study may have important implications for perioperative risk assessment and patient care, as well as for the choice of anastomotic techniques.

Introduction

Postoperative pancreatic fistula (POPF) is generally regarded as the most relevant complication after pancreatoduodenectomy (PD) because it potentially leads to deleterious secondary complications, increased health care costs and prolonged hospital stay.^1–3

Recent studies using the strict definition applied by the International Study Group of Pancreatic Surgery (ISGPS)³ report POPF rates of around 30%.^1–4 Many risk factors for the development of POPF have been identified and anastomotic techniques have been refined in multiple ways in an effort to minimize POPF rates. This sometimes leads to the use of special anastomotic techniques or perioperative precautions.^4–8 As multiple studies have not elucidated which factor has the highest predictive value, it remains of particular importance to evaluate pre- and intraoperative parameters which can predict the rate of POPF in order to identify patients at risk for POPF formation. This problem is of particular relevance when studies are designed to evaluate varying operative reconstructive techniques after pancreatic resection or the use of therapies that pharmacologically aim to reduce pancreatic fistula formation, such as sandostatin.

This study was designed as an objective, prospective evaluation of the importance of the hardness and fibrosis of the pancreatic remnant for the development of pancreatic fistula after PD using histological analysis of the fibrosis grade of the pancreas. Furthermore, we evaluated whether subjective evaluation of pancreatic hardness by an experienced pancreatic surgeon is as accurate as postoperative histopathological assessment by a pathologist in predicting POPF rate. The findings of the study may have important implications for perioperative risk assessment, postoperative patient care and, potentially, choice of anastomotic technique.

Materials and methods

Patients, data and statistics

We sourced our material from a prospectively maintained database of patients undergoing pancreatic surgery from 1994 at the Department of General and Visceral Surgery, University Hospital Freiburg, Freiburg, Germany, with the approved of the hospital ethics committee. A total of 62 consecutive patients were included in the study for risk factor analysis and the development of a preoperative risk score. To validate the risk score, we used preoperative and POPF data for an independent dataset of 279 patients operated at the Department of General and Visceral Surgery between 2001 and 2010. The risk score was calculated retrospectively after the review of patient records. Collection of data and statistics were performed using spss Version 17.0 (SPSS, Inc., Chicago, IL, USA). For statistical analysis, descriptive measures, Spearman's rank correlation, Fisher's exact test, Mann–Whitney U-test and binary logistic regression with stepwise backward elimination were employed as indicated at a significance level of 0.05.

Operations and standard postoperative care

The standard operation was a pylorus-preserving PD with a reconstruction by pancreatogastrostomy (PG) or pancreatojejunostomy (PJ). Two soft silicon drains were placed near the pancreatic anastomosis and routinely removed on day 5 when no POPF was evident from the drain effluate. Two additional drains were placed next to the hepaticojejunostomy. Patients were treated on the intermediate care unit for at least 2 days before transfer to a normal ward. To detect every grade A POPF according to ISGPS definitions, amylase levels in the drain fluid were measured daily for at least the first 3 postoperative days. Amylase levels were also evaluated on every occasion fluid was obtained by puncture or drainage of intra-abdominal collections. Sandostatin was administered only if drain amylase activity exceeded 1000 U/l on postoperative day 3. The use of sandostatin was documented and included in the analysis.

Definition of POPF

Postoperative pancreatic fistula was defined according to the ISGPS definition: a POPF is present if amylase level on or after postoperative day 3 exceeds three times the upper serum limit. Briefly, grading of POPF is dependent on the clinical course: grade A POPF does not need specific treatment; grade B POPF requires prolonged drainage or specific medical treatment, and grade C POPF requires invasive therapy.³

Assessment of risk factors for POPF

Demographic data and co-morbidity were retrieved from the database, as well as data concerning known risk factors for the development of POPF.^9–22 Weight loss (≥3 kg over the previous 6 months) was assessed by patient interview for the 6 months before the operation. The following risk factors were assessed during the operation: pancreatic texture; duct diameter (measured with a scale at the cut surface); mobilization of the pancreatic remnant, and involvement of the uncinate process by the tumour or inflammatory mass. Two methods were used to quantify pancreatic hardness. Firstly, the pancreatic remnant was evaluated subjectively during the operation. The degree of pancreatic fibrosis (soft/hard) in the pancreatic remnant was evaluated subjectively by one experienced pancreatic surgeon (TK) during the operation as ‘hard’ or ‘soft’. Secondly, haematoxylin and eosin-stained tissue sections from the pancreatic cut surface were evaluated retrospectively by an experienced pathologist (GK) blinded to the intraoperative evaluation and the postoperative clinical course. Histological evaluation of pancreatic parenchyma at the resection margin was performed using the following four-stage scoring system: 0 = normal pancreas parenchyma, no fibrotic changes; 1 = mild fibrosis with thickening of periductal fibrous tissue; 2 = moderate fibrosis with marked sclerosis of interlobular septa, no evidence of architectural changes, and 3 = severe fibrosis with detection of architectural destruction. For univariate and multivariate risk factor analysis, POPF (grade A, B or C) was defined as the endpoint.

Results

A total of 62 consecutive patients who underwent partial PD between 2006 and 2008 were included in the study. Patient characteristics, operative findings and procedures and final histology are shown in Table 1. Factors showing a significant correlation with the development of POPF are also depicted in Table 1. Higher age was a risk factor for the development of POPF, whereas history of weight loss and history of pancreatitis constituted protective factors. Intraoperative risk factors for POPF were soft pancreatic texture as evaluated by the surgeon, more extensive mobilization of the pancreatic remnant and small duct diameter. Involvement of the uncinate process by the underlying disease was associated with less POPF. Regarding histopathological diagnoses, patients suffering from pancreatic carcinoma and chronic pancreatitis had lower rates of POPF compared with patients with other pathologies.

Table 1.

Patient and surgery-based parameters and univariate and multivariate analysis of risk factors for the occurrence of postoperative pancreatic fistula (grades A–C)

Parameter description		n or median (range) in groups		Univariate analysis		Multivariate analysis
Parameter	n (%) or median (range)	Without POPF (n = 43)	With POPF (n = 19)	Correlation coefficient	P-value	Odds ratio (95% CI)	P-value
Patient characteristics
Age, years	66 (39–84)	65 (39–84)	71 (40–84)	+0.27	0.038	1.01 (0.94–1.09)	0.739
Sex, male : female	30:32	21:22	9:10	+0.14	0.917
BMI	23 (16–34)	23 (16–30)	24 (19–34)	+0.14	0.296
Preoperative ERD	27 (43.5%)	20	7	−0.09	0.487
Preoperative PTD	4 (6.5%)	4	0	−0.18	0.175
Preoperative bilirubin, mg/dl	1.4 (0.2–37.4)	1.6 (0.2–37.4)	0.9 (0.3–24.7)	−0.11	0.410
Diabetes mellitus	9 (14.5%)	6	3	+0.02	0.853
Cardiac disease	13 (21.0%)	7	6	+0.17	0.178
Renal disease	5 (8.1%)	2	3	+0.19	0.142
Preoperative creatinine, mg/dl	0.8 (0.5–2.2)	0.8 (0.5–2.2)	0.9 (0.5–2.1)	+0.13	0.313
Pulmonary disease	7 (11.3%)	4	3	+0.10	0.465
History of weight loss, yes	38 (61.3%)	32	6	−0.41	0.001	0.15 (0.03–0.69)	0.015
History of acute pancreatitis	9 (14.5%)	9	0	−0.27	0.031	0.00 (0–∞)	0.999
Reported active smoking	18 (29.0%)	14	4	−0.12	0.366
Reported pack-years	0 (0–80)	0 (0–80)	0 (0–46)	−0.11	0.411
Reported alcohol abuse	2 (3.2%)	2	0	−0.12	0.347
Operations and intraoperative parameters
Classical Whipple procedure	4 (6.5%)	2	2	+0.11	0.394
Portal venous resection	11 (17.7%)	8	3	−0.34	0.793
Operation time, h	7.6 (4.7–11.4)	7.6 (5.1–11.4)	7.6 (4.7–11.2)	−0.03	0.809
Blood loss, l	0.5 (0.2–2)	0.5 (0.2–2)	0.5 (0.2–1.5)	+0.57	0.663
Pancreatic texture soft (evaluation by surgeon)	33 (53.2%)	16	17	+0.48	0.000	10.18 (1.26–82.39)	0.030
Mobilization of pancreatic remnant, cm	3 (2–5)	2.5 (2–5)	3 (2–4)	+0.31	0.014	1.60 (0.62–4.08)	0.329
Pancreatic duct diameter, mm	3 (2–10)	4 (2–10)	3 (2–8)	−0.30	0.018	0.94 (0.58–1.54)	0.809
Involvement of the uncinate process	21 (33.8%)	18	3	−0.25	0.046	0.46 (0.07–2.84)	0.402
Histopathology
Pancreatic carcinoma	27 (43.5%)	24	3	−0.37	0.003
Chronic pancreatitis	7 (11.3%)	6	1	−0.13	0.327
Ampullary carcinoma	11 (17.7%)	6	5	+0.15	0.247
Cholangiocellular carcinoma	4 (6.5%)	1	3	+0.25	0.048
Cystic neoplasia of the pancreas	4 (6.5%)	2	2	+0.11	0.394
Duodenal carcinoma	3 (4.8%)	2	1	+0.01	0.919
Neuroendocrine tumour	2 (3.2%)	1	1	+0.08	0.554
Other	4 (6.5%)	1	3	+0.25	0.048

Univariate analysis: two-tailed Spearman's rank correlation; negative values in correlation coefficient indicate protective factors and positive values indicate risk factors

Multivariate analysis: binary logistic regression

POPF, postoperative pancreatic fistula (International Study Group for Pancreatic Surgery definition); 95% CI, 95% confidence interval; ERD, endoscopic retrograde drainage; PTD, percutaneous transhepatic drainage; BMI, body mass index; CNP, cystic neoplasia of the pancreas; NET, neuroendocrine tumour

All preoperatively or intraoperatively available parameters that correlated significantly with the POPF rate (Table 1) were included in a multivariate binary logistic regression model. The only independent factors identified by this approach were pancreatic texture and history of weight loss prior to the operation. Soft pancreatic texture, as assessed by the surgeon intraoperatively, was the strongest predictor for POPF (Tables 1 and 2). There was also a strong correlation of soft pancreatic texture with lower histopathological grade of pancreatic fibrosis at the cut surface (Table 2).

Table 2.

Correlation of soft pancreatic texture with occurrence of postoperative pancreatic fistula and low histopathological fibrosis grade

Pancreatic texture	Occurrence of postoperative pancreatic fistula (ISGPS definition)
	No POPF	POPF grade A	POPF grade B	POPF C
Soft	16	11	2	4
Hard	27	1	1	0
	Correlation coefficient = 0.47, P < 0.001a

Pancreatic texture	Histopathological fibrosis grade
0	1	2	3
Soft	6	13	10	2
Hard	0	0	13	14
	Correlation coefficient = −0.68,P < 0.001a

Pancreatic texture as assessed by the operating surgeon

Numbers are actual numbers of patients

^aCorrelation coefficient and P-value for two-tailed Spearman's rank correlation

POPF, postoperative pancreatic fistula; ISGPS, International Study Group for Pancreatic Surgery

Univariate correlation analysis was performed to identify the factors associated with hard pancreatic texture (Table 3). There was significant association between hard pancreatic texture and large duct diameter, involvement of the uncinate process, pancreatic carcinoma and chronic pancreatitis, whereas ampullary carcinoma was generally associated with a soft pancreas. Younger age, history of acute pancreatitis, active smoking and increasing number of pack-years also showed significant correlations with hard pancreatic texture.

Table 3.

Conditions correlating to hard pancreatic texture

Parameter	Correlation coefficient	P-value
Age, years	−0.34	0.007
History of acute pancreatitis	+0.26	0.045
Active smoking	+0.26	0.045
Pack-years	+0.33	0.009
Pancreatic duct diameter, mm	+0.38	0.002
Involvement of the uncinate process	+0.29	0.025
Pancreatic carcinoma	+0.48	0.000
Chronic pancreatitis	+0.28	0.028
Ampullary carcinoma	−0.44	0.000

Two-tailed Spearman's rank correlation

Preoperative evaluation of risk for POPF by clinical parameters

On the basis of the aforementioned findings, a preoperative risk score for use in the clinic was developed. Factors available preoperatively that had been found to correlate with risk for POPF (Table 1) and parameters that showed significant correlation with pancreatic texture (Table 3) were included. For simplicity, the score was calculated by adding 1 point for a risk factor and subtracting 1 point for a protective factor (Table 4). Preoperative diagnosis was presumed to be pancreatic carcinoma or chronic pancreatitis unless specific preoperative findings (history; imaging by computed tomography, magnetic resonance imaging [MRI] or endoscopic retrograde cholangiopancreatography; preoperative histology) supported another diagnosis such as ampullary carcinoma, bile duct cancer, cystic or neuroendocrine tumours of the pancreas or metastasis to the pancreas. As shown in Table 4, the risk score could distinguish significantly between a low-risk group of patients who developed no POPF, a medium-risk group with a 27% POPF rate, and a high-risk group with a 61% POPF rate (correlation coefficient = 0.47, P < 0.001 for two-tailed Spearman's rank correlation).

Table 4.

Calculation of the preoperative risk score and correlation to the occurrence of postoperative pancreatic fistula

Factor				Value
Age >66 years				+1
Preoperative diagnosis other than pancreatic carcinoma or chronic pancreatitis				+1
History of smoking				−1
History of weight loss				−1
History of acute pancreatitis				−1
Sum score				High risk: sum of 1 or 2
				Medium risk: sum of −1 or 0
				Low risk: sum of −3 or −2

POPF grade	Training dataset			Validation dataset
	Preoperative risk score			Preoperative risk score
	Low	Medium	High	Low	Medium	High
None	14	22	7	43	117	15
	100.0%	73.3%	38.9%	78.2%	70.1%	26.3%
A	0	5	7	8	27	20
	0.0%	16.7%	38.9%	14.5%	16.2%	35.1%
B	0	1	2	1	11	6
	0.0%	3.3%	11.1%	1.8%	6.6%	10.5%
C	0	2	2	3	12	16
	0.0%	6.7%	11.1%	5.5%	7.2%	28.1%
	Correlation coefficient = 0.47			Correlation coefficient = 0.35
	P < 0.001a			P < 0.001a

The score was developed in the ‘training’ dataset of 62 patients and validated in an independent dataset of 279 patients

Numbers given are absolute and percentages refer to the total number of patients in one column

^aCorrelation coefficient and P-value for two-tailed Spearman's rank correlation

POPF, postoperative pancreatic fistula, Grades A–C according to International Study Group for Pancreatic Surgery definitions

Validation of the preoperative risk score

An independent dataset comprising 279 patients was used to validate the newly developed risk score. As Table 4 shows, the score could correctly distinguish the three risk categories. These correlations were highly significant (correlation coefficient = 0.35, P < 0.001 for two-tailed Spearman's rank correlation).

Discussion

From the earliest days of pancreatic surgery to the most recent clinical series, pancreatic secretions and POPF have been recognized as the main determinants of morbidity.^2,17,23–25 Despite the long history of this ‘Achilles heel’ in pancreatic surgery, an international consensus definition of POPF has only recently been introduced by the ISGPS.³ The significant impact of POPF on health care costs has also been recently demonstrated.¹

In an attempt to predict the development of POPF, clinical studies have identified many risk factors, such as gender,⁹ cardiovascular disease,^9,10 diabetes mellitus,⁹ obesity,¹¹ leukocytosis,¹² low serum albumin,¹² impaired renal function,¹³ centre effect,¹⁴ underlying pathology,^9,15–18 long operative time,⁹ concomitant surgery or radical lymphadenectomy,^9,15 high intraoperative blood loss,^9,17,19 small pancreatic duct diameter^15,17,20 and ‘fatty’²⁶ or ‘soft’^{10,14,15,17,19–22} pancreas. The latter has been described as a risk factor for the development of POPF in several series of pancreatic surgery since the 1970s.^{21,24,25,27–31} Some prospective trials in pancreatic surgery have acknowledged this by stratifying patients for randomization or final analysis.^16,32–34

In this study, patient characteristics were prospectively evaluated as risk factors for POPF. Univariate analysis identified several risk or protective factors for POPF. Multivariate analysis disclosed that only preoperative weight loss and hard pancreatic texture were independent predictive factors. Prediction by the surgeon was the most reliable of all evaluated parameters. There was also a very strong correlation between the grade of hardness assessed subjectively by the surgeon and the grade of pancreatic fibrosis assessed objectively by the pathologist. Although every experienced pancreatic surgeon is familiar with this correlation, few studies have examined it scientifically.³⁵

Furthermore, other factors that may predict the development of pancreatic fistula formation after partial PD were assessed. Preoperative weight loss was found to be a protective factor in the current analysis, which, to the authors' knowledge, has not been identified in previous studies. This finding may be explained by a positive correlation between body mass index (BMI) and pancreatic fat in previous studies,^26,36 as increased BMI¹¹ and pancreatic fat²⁶ are risk factors for POPF. These earlier studies did not correlate pancreatic fat with tissue hardness and thus it remains unproven whether the so-called fatty pancreas would be judged as soft by the surgeon. By contrast, weight loss is a typical feature of pancreatic carcinoma, which is usually associated with a hard pancreas and safer anastomosis.

In addition, episodes of acute pancreatitis contributed to pancreatic hardness. The negative correlation of pancreatic hardness with age may be explained by increasing pancreatic fat with age.²⁶ Interestingly, this study is the first to identify an influence of smoking on pancreatic texture. Both active smoking and number of pack-years correlated with the development of a hard pancreas. This is consistent with the experimental finding that tobacco smoke induces inflammatory lesions in the pancreas of rats.³⁷ Among intraoperative factors, mobilization of the pancreatic remnant is noteworthy because it is the only factor that can be influenced by the surgeon. Extensive mobilization resulted in a higher rate of POPF.

In concordance with previous studies,^9,15–18 pancreatic texture was hard in chronic pancreatitis and pancreatic carcinoma and soft in cases of smaller ampullary carcinoma. However, histopathological diagnosis cannot be used for prospective and preoperative risk assessment.

What makes a soft pancreas so dangerous? Many surgeons like to point out the technical difficulties of a pancreatoenteric anastomosis. Obviously, soft tissue cannot resist the tearing forces of sutures as reliably as hard tissue. By contrast, a soft parenchyma is, as in our study, often associated with a small, non-dilated pancreatic duct, which makes the popular duct-to-mucosa anastomosis difficult or even impossible. A sometimes neglected aspect is the degree of preserved exocrine pancreatic function in the individual patient, which cannot be influenced by the surgeon. A soft parenchyma and non-dilated pancreatic duct have been shown to be associated with well-preserved exocrine pancreatic function,^{22,30,31,38,39} thereby representing an important reason for the development of POPF.

Given the tremendous importance of pancreatic texture in pancreatic surgery, substantial effort has been devoted to finding ways of obtaining objective measurements of pancreatic hardness. Intraoperative use of a durometer has been proposed for the evaluation of pancreatic hardness.^35,40 Preoperative assessment of pancreatic texture is possible by means of dynamic MRI^8,18 and preoperative tests of exocrine pancreatic function.^30,41 However, radiodensity of the pancreas did not correlate with pancreatic hardness or fibrosis.³⁵ These reported methods are cumbersome or expensive and are unlikely to be applied in larger series. By contrast, careful evaluation of patient history and intraoperative risk factors represents a reliable way of risk prediction at no cost. This study has demonstrated a clinical risk score which can be easily calculated from the patient's history and which has been validated against a large independent dataset. This risk score was able to distinguish between three risk groups, showing rates of around 6% for clinically relevant POPF (grade B or C) in the low-risk group and almost 40% for relevant POPF in the high-risk group. The strongest single, yet only intraoperatively available, predictor remains pancreatic texture as evaluated by the experienced surgeon, which showed POPF rates of around 50% in cases of soft pancreas and only 6% in cases of hard pancreas. No grade C fistula was seen in patients with a hard pancreas.

In order to reduce the POPF rate in high-risk patients with a soft pancreas, various operative procedures have been described since the first large series of PD. Techniques that use some form of invagination in either PJ or PG or techniques using duct ligation and drainage have been reported to be especially beneficial.^42–44 Some surgeons have even suggested performing total pancreatectomy in cases of very soft pancreas,^25,35,45 an approach that we do not generally support, given the broad armamentarium of safe reconstructive techniques. A recent prospective study¹⁴ comparing two techniques of PJ demonstrated that pancreatic technique represents a strong bias when comparing anastomotic techniques. By contrast with the trial hypothesis, the authors found the invagination method to be superior in terms of POPF rate. This difference was only obvious after stratification for pancreatic texture. At our institution, a randomized trial is currently ongoing to test whether PG leads to a reduced POPF rate in comparison with PJ. With regard to non-operative measures, preoperative octreotide treatment has been shown to harden the pancreas^5,40 and may reduce POPF rates in high-risk patients.^6,7

In summary, pancreatic texture represents a highly significant determinant of pancreatoenteric anastomotic leakage rates and corresponds to the degree of pancreatic fibrosis and exocrine function. Careful evaluation of patient history and intraoperative assessment of pancreatic texture by the surgeon are the most simple yet very reliable methods of preoperative and intraoperative prediction of risk for POPF. The currently proposed clinical risk score needs to be prospectively evaluated. Measures that may reduce POPF rates in high-risk patients are choice of pancreatoenteric anastomoses and pre- or intraoperative octreotide treatment.

Conflicts of interest

None declared.