The ISGPF developed the first definition and grading system of POPF after PD in 2005, which has been widely used in clinical practice and scientific research. As of December 2015, it was cited more than 1700 times and applied to clinical research of more than 320,000 patients[8]. Over time, the grading system has developed an increasing number of problems and deficiencies. In 2016, the ISGPS (formerly known as ISGPF) revised the definition and grading standard for POPF, changing the “Grade A pancreatic fistula” in the 2005 edition to “biochemical leakage,” and “biochemical leakage” is no longer considered to be a kind of actual pancreatic fistula. The diagnosis of Grade B pancreatic fistula needs to be clinically related and affect the postoperative process. On the basis of “biochemical leakage,” any of the following situations can be found, such as continuous drainage of abdominal cavity for more than 3 weeks, change of treatment measures for clinically related pancreatic fistula, percutaneous or endoscopic drainage, angiographic intervention in the treatment of bleeding, and infection signs without organ failure. Then Grade B fistula can be diagnosed (in new grading standard, at the same time, it shall be pointed out that if the patients discharged from the hospital with a tube are generally in good condition and take food by mouth; to be with a tube for more than 3 weeks for observation is also graded as biochemical leakage). If it is needed to conduct secondary operation for Grade B pancreatic fistula, and single or multiple organ failure or death occurs, it will be upgraded to Grade C. In the 2016 edition, the clinical relevance of pancreatic fistula is emphasized, the definition and the grading are clearer, and the clinical operability is stronger.
There are many studies about the risk factors for POPF after PD including preoperative factors, intraoperative factors, and postoperative factors. To improve the accuracy of pancreatic fistula risk prediction, many risk factors have been combined to establish a pancreatic fistula prediction system by domestic and foreign scholars. At present, the main pancreatic duct diameter and the pancreas texture are generally recognized as risk factors related to the pancreatic fistula[9–13]. Other factors include age, gender, main pancreatic duct index, body mass index, intra-abdominal thickness, pathological diagnosis, intraoperative blood loss, preoperative blood amylase, postoperative C-reactive protein and procalcitonin levels, and portal vein invasion, all of which have been reported to be related to pancreatic fistula[12,14−19]. According to the definition and grading standards for pancreatic fistula in the new and old editions, the risk factors for pancreatic fistula were screened out in the study, consistent with the literature. Among them, the main pancreatic duct index and pancreatic CT value were both independent risk factors for POPF after PD in the old and new editions. The main pancreatic duct index was the ratio of the main pancreatic duct diameter to the pancreatic thickness. It was proposed by Akamatsu et al.[16], and is the strongest independent predictor of POPF. This index can better predict the occurrence of pancreatic fistula than the pancreatic duct diameter can do alone[14,20−21]. There are two reasons why the pancreas CT value replaces the soft and hard texture of pancreas in this study. First, there is no universally recognized standard for the soft and hard texture of pancreas, which is mainly judged by the operator’s touch during operation, and the subjective factors are too strong to be quantified. Second, the literature has proven that the pancreatic texture is related to the pancreas CT value[7,22−24]; the higher the CT value of the pancreas, the higher the density of the pancreatic tissue, the more severe the degree of pancreatic fibrosis, and the lower the risk of pancreatic fistula. In this study, it was also believed that the pancreas CT value can reflect the pancreas texture. Among the 124 patients, the pancreas CT value of patients with pancreatic fistula was 36.36 ± 6.49 in the 2005 edition and 36.16 ± 7.29 in the 2016 edition. The pancreas CT value of non-pancreatic fistula patients was 41.10 ± 9.61 in the 2005 edition and 39.67 ± 8.78 in the 2016 edition. The difference was statistically significant.
In the past decade, scholars at home and abroad have established a prediction system for POPF after PD by combining multiple risk factors related to pancreatic fistula. In 2010, the German scholar, Wellner et al.[25] established a pancreatic fistula prediction system by combining five indicators, i.e., age, preoperative diagnosis of non-pancreatic cancer or chronic pancreatitis, smoking history, emaciation history, and acute pancreatitis. The total score could accurately predict the patients at low, medium, and high risk for pancreatic fistula. In 2011, the Japanese scholar, Yamamoto et al.[26] established a preoperative pancreatic fistula prediction system based on sex, pancreatic cancer diagnosis, main pancreatic duct index, portal vein invasion, and intra-abdominal thickness. The results showed that the prediction accuracy was high and verified by many domestic medical centers. In 2012, Ansorge et al.[27] from Sweden included 110 patients in a prospective study, from which a prediction system based on intraoperative evaluation of pancreatic texture and pancreatic duct diameter could accurately predict the risk of postoperative pancreatic fistula. In 2013, Callery et al.[28] from the United States established an FRS pancreatic fistula prediction scoring system with a score of 0–10 in combination with pancreatic texture, main pancreatic duct diameter, pathology, and intraoperative blood transfusion volume. The scoring system divides patients into four risk levels: negligible risk (0 point), low risk (1–2 points), medium risk (3–6 points), and high risk (7–10 points). In 2014, Roberts et al.[29] from the United Kingdom established a pancreatic fistula prediction system based on body mass index and main pancreatic duct diameter. The prediction accuracy was verified in the Center. In 2016, Chinese scholar Yinmo et al.[30] also established a pancreatic fistula risk prediction model based on body mass index and main pancreatic duct diameter, and it was proved that the model had good prediction accuracy.
Previous studies on the risk factors for POPF were based on the 2005 edition of the definition and grading standard of pancreatic fistula. In this study, patients with pancreatic fistula were included according to the old standard in the 2005 edition and the revised standard in the 2016 edition, and the influence of new and old editions on the risk factors for POPF after PD were compared. According to the new edition of the pancreatic fistula standard, the incidence of pancreatic fistula in this study decreased from 49.2–25.8%, in line with clinical practice. Univariate analysis showed that there were nine risk factors in the old edition of pancreatic fistula standard including main pancreatic duct diameter, main pancreatic duct index, portal vein invasion diagnosis, intra-abdominal thickness, preoperative biliary drainage, pancreatic cancer diagnosis, margin pancreatic thickness, pancreas CT value, and preoperative serum amylase level. The new edition of pancreatic fistula standard was reduced to five including main pancreatic duct diameter, main pancreatic duct index, intra-abdominal thickness, margin pancreatic thickness, and pancreas CT value. It was suggested that these five risk factors were more closely related to clinical pancreatic fistula. Multivariate analysis showed that there were three independent risk factors in the old edition of pancreatic fistula standard including the main pancreatic duct index, pancreatic cancer diagnose, and pancreas CT value. The new edition of pancreatic fistula standard was reduced to two, including the main pancreatic duct index and pancreas CT value. The main pancreatic duct index and pancreas CT value could be obtained before operation. Based on this, a mathematical model for predicting pancreatic fistula was established. The calculated result of the model was the POPF probability of patients. When the calculated value was more than 32%, the patient was considered high risk for POPF. The greater the calculated value, the higher the risk of POPF.
According to the new definition and grading standard of pancreatic fistula, the prediction model of POPF after PD was established in this study. The prediction parameters can be obtained by CT before operation. The clinical operation is simple, objective and quantitative, and the repeatability is strong. The model is of clinical value in predicting the risk of POPF before surgery. However, this study was a single-center, retrospective study. The relationship between preoperative CT parameters and pancreatic fistula, as well as prospective study with large samples verifies the prediction model is the future research direction.