Preoperative CT anthropometric measurements and pancreatic pathology increase risk for postoperative pancreatic fistula in patients following pancreaticoduodenectomy

Postoperative pancreatic fistula (POPF) is a common complication following pancreaticoduodenectomy (PD). However, risk factors for this complication remain controversial. We conducted a retrospective analysis of 107 patients who underwent PD. POPF was diagnosed in strict accordance with the definition of the 2016 update of pancreatic fistula from the International Study Group on Pancreatic Fistula (ISGPF). Univariate and multivariate logistic regression analyses were performed to identify independent risk factors for POPF. A total of 19 (17.8%) subjects of pancreatic fistula occurred after PD, including 15 (14.1%) with grade B POPF and 4 (3.7%) with grade C POPF. There were 33 (30.8%) patients with biochemical leak. Risk factors for POPF (grade B and C) were larger area of visceral fat (odds ratio [OR], 1.40; p = 0.040) and pathology other than pancreatic adenocarcinoma or pancreatitis (OR, 12.45; p = 0.017) in the multivariate regression analysis. This result could assist the surgeon to identify patients at a high risk of developing POPF.

Among the reported complications, postoperative pancreatic fistula (POPF) is the most common complication following PD and is associated with delayed gastric emptying, intraabdominal abscess and hemorrhage, and superimposed infection and sepsis, consequently increasing the length of stay and even leading to reoperation in some cases [9][10][11]. Therefore, reducing the rate of POPF after PD is a serious challenge for clinicians. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Various risk factors for POPF have been suggested, including sex, body mass index (BMI), pancreatic duct size, pancreatic texture, blood transfusion, intraoperative blood loss, operation time, and visceral fat area [9,12,13]. However, definite causal factors for developing POPF after PD remain controversial. Recently, some studies have suggested a relationship between surgical outcomes and anthropometric measurements, such as the core muscle mass and body fat area. Depleted skeletal muscle mass and visceral obesity increased postoperative complications after total gastrectomy, major hepatectomy, and colorectal cancer [14][15][16]. With respect to PD, patients with sarcopenia and visceral obesity showed decreased survival and increased morbidity [12,17,18]. As the POPF criteria were updated in 2016, biochemical leak (POPF A) is no longer considered true POPF [19]. A few studies had reported on the association of anthropometric measurement with POPF according to the revised criteria [20,21].
In this study, we performed a retrospective study to evaluate the association and predictive value of anthropometric measurements and other pre-and perioperative variables for POPF.

Subjects
This retrospective study was approved by the institutional review board (IRB) of Hanyang University Hospital. All experiments were performed in accordance with the relevant guidelines and regulations.
One-hundred and twenty-four consecutive subjects who underwent PD between October 2007 and October 2017 were enrolled into this study. The exclusion criteria were as follows: lack of clinical information (hardness of pancreas [n = 9], BMI [n = 1]); interval between preoperative computed tomography (CT) and surgery >40 days (n = 3); unavailability of preoperative CT (n = 3); and difficulty evaluating POPF (n = 1). Finally, 107 subjects (male: female = 64:43; mean age, 65.9 years; range, 35-82 years) who underwent PD were included. Detailed information was obtained from electronic medical records. For each subject, the following data were collected: (1) subjects' demographic and clinical features, including age, sex, BMI, preoperative albumin, and total bilirubin; (2) operative details, including pancreatic texture, operation time, performance of intraoperative blood transfusion, and the use of a pancreatic stent; and (3) final diagnosis of the tumor. The pancreatic texture of all subjects was examined by the surgeon during the operation and classified as either soft or hard. A pancreatic duct stent was occasionally used during reconstruction following PD according to the surgeon's judgment. External drains were inserted in proximity to the pancreatic anastomosis for each surgery, and the amylase level of drain fluid was routinely measured during the inserted period.

Measurement of anthropometric measurements
We measured subjects' abdominal circumference, visceral and subcutaneous fat, and total abdominal muscle area on preoperative CT scans at the level of the third lumbar vertebra (L3). Distinction among the muscle, fat, and different tissues was based on Hounsfield units (HU) using AquariusNET Server (TaraRecon, Foster City, CA, USA). A threshold range of -29 to 150 HU was used to define muscle and a range of -190 to -30 HU was used to define fat. Hand adjustment of the selected area was performed (Figs 1 and 2). Skeletal muscle mass was normalized for the subjects' heights to calculate the skeletal muscle mass index (SMI, cm 2 /m 2 ). The ratio of visceral fat to SMI (VF/SMI) was also calculated. Sarcopenia was defined as SMI �52.4 cm 2 /m 2 for men and �38.5 cm 2 /m 2 for women based on a study by Prado et al. These cutoff values are accepted by an international consensus group on the diagnostic criteria for cachexia associated with cancer [22,23]. Visceral obesity was defined as a visceral fat area �100 cm 2 in both sexes. This value is widely used as a cutoff to define sarcopenic obesity in Asian populations and is equivalent to that used for the diagnosis of metabolic syndrome in Japan [24,25].

Definition of postoperative pancreatic fistula
Pancreatic fistula was defined according to the revised 2016 International Study Group on Pancreatic Fistula (ISGPF) classification and grading [19]. The previous 'grade A' POPF is newly classified as 'biochemical leak', which refers to a transient and asymptomatic biochemical fistula. Grade B and C POPF are clinically relevant fistulae, which require a change in postoperative management. If peripancreatic drainage persists for more than 3 weeks or is repositioned through interventional procedures, it is classified as Grade B. Grade C POPF refers to those with POPF-related organ failure, reoperation, or death.

Statistical analysis
Normally distributed numerical variables are presented as mean and standard deviation (SD) and were compared using the independent t-test, and non-normally distributed numerical variables are presented as median (the first quartile-the third quartile) and were compared using the Mann-Whitney test. Categorical variables are presented as frequencies with percentage and were tested using the chi-squared test or Fisher's exact test. Univariate logistic analysis and backward stepwise multivariate logistic regression analysis were performed to identify the independent risk factors for POPF after PD. Variables with P-values of <0.05 on univariate analyses were entered into the final multivariate model to reveal risk factors for POPF. For each parameter, an odds ratio (OR) for POPF was provided with a 95% confidence interval (CI). All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC). P-values of <0.05 were considered statistically significant. Table 1. The median time interval between preoperative CT and surgery was 14 days (mean, 15 days; range, 3-40 days). A total of 19 (17.8%) patients developed a pancreatic fistula after PD, including 15 (14.1%) with grade B POPF, and 4 (3.7%) with grade C POPF. In this study, all patients with POPF C underwent reoperation. There were 33 (30.8%) patients with biochemical leak.

Demographic and clinical characteristics are shown in
The area of visceral fat, VF/SMI ratio, pancreas hardness and pathology were significantly different between the two groups ( Table 2). Subjects with POPF had a significantly larger area of visceral fat (159.6 cm 2 vs. 120.3 cm 2 , p = 0.022), higher VF/SMI ratio (3.30 vs. 2.54, p = 0.030), more frequently had a soft pancreas (78.9% vs. 53.4%, p = 0.045), and more frequently had pathology other than pancreatic adenocarcinoma or pancreatitis (97.4% vs. 62.5%, p = 0.045) than patients without POPF. However, sex, BMI, pancreatic duct diameter, operation time, intraoperative blood transfusion and use of a pancreatic stent were not significantly different between the two groups. Univariate regression analysis showed significant correlation between POPF and the following factors ( In the multivariate regression analysis, higher visceral fat (OR: 1.40; 95% CI: 1.07-15.43, p = 0.040) and pathology other than pancreatic adenocarcinoma or pancreatitis (OR: 12.45; 95% CI: 1.59-99.28; p = 0.017) were identified as independent risk factors for POPF (Table 4).

Discussion
Pancreatic fistula after PD remains a challenging problem, even in high-volume centers. Identification of patients at a high risk of developing pancreatic fistula helps in a more elaborate risk-benefit assessment before surgery and may allow clinicians to coordinate perioperative care. In our study, we observed POPF in 17.8% (19/107) of patients. Higher visceral fat and pathology other than pancreatic adenocarcinoma or pancreatitis were independent risk factors for developing POPF after PD.
Recently, studies regarding the effects of sarcopenia and visceral obesity on POPF have been conducted. To date, few studies have examined the impact of sarcopenic obesity on  In our study, a larger area of visceral fat was significantly associated with POPF in both univariate and multivariate analyses, consistent with previous studies. The frequency of visceral obesity was higher in patients with POPF (84.2%, 16/19) compared to those without POPF (61.4%, 54/88), although it was not statistically significant. Percorali et al. also reported that the visceral fat area was an independent predictor of pancreatic fistula in patients undergoing PD [12]. Generally, patients with greater subcutaneous and visceral fat accumulation provide greater technical difficulty for surgeons. The view of the surgical field is deeper and poorer in obese patients, which may increase the risk of pancreatic fistula [14,29,30]. Other than The skeletal muscle index and sarcopenia showed little association with POPF in this study. In a meta-analysis by Ratnayake et al., preoperative sarcopenia was not a significant negative predictive factor in postoperative morbidity, including POPF, following pancreatic resection [21]. On the contrary, there are several studies showing that sarcopenia is an independent risk factor for POPF [36,37]. Meanwhile, VF/SMI ratio was higher in POPF group than non-POPF group in our study, although it was not significant at multivariate analysis. High VF/SMI ratio could be referred to as sarcopenic obesity, a condition in which loss of muscle mass is accompanied by increase in fat accumulation [22]. Jang et al. also examined 284 patients who underwent PD between 2005 and 2016, concluded that sarcopenic obesity was the only predictor for POPF [20]. Although our study and study by Jang et al. used predefined cutoff for visceral obesity, sarcopenia, and sarcopenic obesity, an accurate definition of sarcopenic obesity has not yet been established. These various conclusions from existing studies suggest that the impact of sarcopenia and sarcopenic obesity in developing POPF following PD is still inconclusive and therefore further research is needed. Higher POPF rate was observed in pathology other than PDAC or chronic pancreatitis, which was categorized according to the previously established fistula risk score criteria [38]. The mechanism by which pathologies other than PDAC or chronic pancreatitis increases the risk of POPF is related to the effect of soft pancreas parenchyma. Pathologies with pancreatic adenocarcinoma and chronic pancreatitis are more likely to result in hard parenchyma, therefore there could be a benefit in reducing the incidence of POPF. Hu et al. retrospectively analyzed 536 cases and also found that a soft pancreas was an independent risk factor for developing pancreatic fistula, including biochemical leak and POPF B and C, after PD (OR: 3.05, p<0.001) [9]. Kawai et al. reviewed 1,239 patients from 11 Japanese medical centers and concluded that a soft pancreas was a significant predictive factor for pancreatic fistula (OR: 2.7, p = 0.001) [39]. In this study, among a total of 107 patients, a soft pancreas was more frequent in subjects with POPF than in those without POPF (78.9% vs. 53.4%), although the result was not significant at multivariable analysis. A soft pancreas is more prone to laceration when performing suturing and tying. Moreover, the soft texture of pancreatic remnants induces technical difficulties in performing pancreatoenteric and duct-to-mucosa anastomoses, which increase the risk of anastomotic leakage [40]. Although it is not possible to evaluate soft pancreas before surgery, the possibility of POPF can be assessed by guessing the pathology through preoperative CT.
Our study had several limitations. First, due to its retrospective nature, it may have been influenced by selection and information biases. Second, it is a single-center study, which only includes Koreans; therefore, the findings may not be applicable to other populations. Third, several surgical anastomotic techniques have been established in recent years, but the technique used in each surgery was not analyzed in this study. Also, the pancreatic texture during the operation was evaluated by the surgeon, which may be subjective. Studies related to quantitative measurement of pancreas texture have been reported, such as MR elastography or ultrasound elastography [41,42]. Further research regarding the objective assessment of the pancreas texture is warranted.
In conclusion, larger area of visceral fat and pathology other than pancreatic adenocarcinoma or pancreatitis are independent predictors of POPF after pancreaticoduodenectomy and may allow better preparation of the postoperative care for patients by identifying patients at a high risk of developing pancreatic fistulas.