C-reactive protein identifies patients at risk of postpancreatectomy hemorrhage

Patients, who underwent PD between January 2003 and June 2018 at the Sahlgrenska University Hospital, Gothenburg, Sweden, a regional center for pancreatic surgery since 2009 and national center for advanced pancreatic surgery since 2017, were identified using the ICD-10 classification in the medical record. A total of 560 patients were classified as PD (Whipple’s operation); 73 patients were excluded due to contemporary multivisceral operations or transplantations (n:29), arterial resections (n:11), multitrauma operations (n:3), colon resections (n:12), splenectomies (n:10), or other including default encoding (n:8). We chose to exclude major concurrent procedures in order to reduce the number of confounding factors. Both multivisceral and arterial resections are associated with increased mortality and morbidity [20, 21].

Another 64 patients classified as being treated with a total pancreatectomy were re-analyzed. Twenty-three of them were found to be incorrectly classified and were re-classified as PDs. Thus, a total of 510 patients were retrospectively included in the analysis.

The indication for surgery was known or suspected periampullary cancer or precancerous conditions, such as intraductal papillary mucinous neoplasm (IPMN), mucinous cystic neoplasm (MCN), or secondary malignancies. By using our ERAS (enhanced recovery after surgery) register, established in January 2015, patients operated between January 2015 and June 2018 with Clavien-Dindo Grade III or worse were identified. These patients’ medical records were then studied in detail to identify patients with PPH C. To recognize patients with PPH C operated between January 2003 and December 2014, the medical records from discharge and return visits to the surgical outpatient clinic 4 to 6 weeks after the surgery were studied. In total 23 patients with PPH C were identified (Table 3). Five patients with late intraluminal PPH and according to the ISGPS definition severe (more than 3 units packed cells and relaparotomy) hemorrhage were excluded from the PPH C group. In these five patients with intraluminal PPH, the bleeding was controlled by a suture ligature of a bleeding vessel on the pancreatic surface or in the gastroenteroanasomis during a relaparotomy, and the patients recovered quickly after the bleeding had stopped. The five patients were indeed PPH C, but the purpose of this study was to address more complex life-threatening bleedings, extraluminal, with potential underlying anastomosis leakage.

The remaining 487 PD patients were matched against the PPH C patients according to gender, year of operation (± 1 year), and age (± 2 years). Four matched controls were randomly selected for each PPH C patient, a total of 92 patients.

The PD was performed as a Whipple procedure in all included patients. The head of the pancreas, the distal part of the stomach, the duodenum, the gallbladder, and the bile duct were removed. The dissection was performed as a standardized procedure at the clinic with electrical scissors and Ligasure™ Impact or Ligasure™ Maryland (Medtronic). The gastroduodenal artery was closed by nonabsorbable suture ligature. The pancreatic remnant reconstruction was performed as a pancreaticojejunostomy (PJ) with duct-to-mucosa until 2010/2011, when there was a gradual shift in the anastomosis technique to pancreaticogastrostomy (PG) with invagination technique. When PG was not possible, as in patients with prior gastric bypass surgery, PJ was performed instead. Perioperatively, two gravidity drains were placed in the abdominal cavity, one towards the hepaticojejunostomy and one towards the pancreatic anastomosis. All patients received preoperative antibiotic prophylaxis. Proton pump inhibitors were given during the first postoperative month. Low molecular weight heparin (LMWH) were routinely administrated to all patients starting the day before surgery and continued until discharge from the hospital. Octreotide was used neither as prophylaxis nor as treatment. A nasogastric tube was routinely inserted, and fasting with gastrointestinal decompression was continued in 3–7 days.

An ERAS program was implemented in January 2015, after which both drains and nasogastric tubes were removed at an earlier time-point. Drains were withdrawn on day 3 instead of day 5 if the drain amylase level was decreasing, < 2400U/L, and the drain fluid serous. The nasogastric tube was removed on day 3 instead of day 7 if the measured volume was < 400 ml.

All patients, even if domiciled in other regions, had a return visit to the Sahlgrenska University hospital at 4 to 6 weeks postoperatively to receive information on the pathology report and for a clinical follow-up.

Preoperative, perioperative, and postoperative data were collected retrospectively from medical records, and the data were saved without personal information for general anonymous analysis.

The potential risk factors for PPH C were divided into three categories: (1) preoperative factors (age, gender, year of surgery, body mass index (BMI), cardiovascular disease, history of abdominal upper gastrointestinal surgery or pancreatitis, preoperative biliary stent, C-reactive protein (CRP) levels, American Society of Anesthesiologists (ASA) score), (2) perioperative factors (amount of bleeding in ml, type of pancreatic anastomosis, operation time), and (3) postoperative factors (BL/POPF before and after bleeding, biliary fistula, and CRP levels). Additional information regarding sentinel bleeding, postoperative day of bleeding and site, pathological examination, angiography, re-operation, 90-day mortality rate, and whether POPF grade C was known before PPH C were also recorded.

Cardiovascular diseases were evaluated by the WHO definition. Biliary fistula was defined as the presence of bile or bile-stained fluid from surgical drains on or after postoperative day 3 [7]. The definition of sentinel bleeding was a minor blood loss via abdominal drains, wound, or the gastrointestinal tract, with an asymptomatic interval until development of hemorrhagic shock [22].

CRP was calculated as the median on postoperative day 2 or 3 (CRP POD 2/3) or postoperative day 5 or 6 (CRP POD 5/6). The number of patients with CRP levels > 180 mg/L was also recorded. Prior to 2015, the days when CRP was measured in the clinic were not standardize, which is why CRP levels were calculated both POD 2/3 and POD 5/6 in this study. After the introduction of the ERAS program in 2015, the CRP levels were always measured on days 1, 3, and 5. The cut-off of 180 mg/L was chosen based on a previous study of pancreas-specific complications after pancreaticoduodenectomy [23].

Preoperative biliary stents were placed by either endoscopic retrograde cholangiopancreatography (ERCP) or percutaneous transhepatic cholangiography (PTC). The median perioperative bleeding volume was measured, and the number of patients with bleeding > 500 ml was calculated separately. The operation time was recorded as the number of patients operated on in < or > 416 min. In the Swedish National Pancreatic Register, the median bleeding at the Sahlgrenska University Hospital Gothenburg was 500 ml in 2017 and 2018; the median operation time was 418 and 416 min, respectively.

The present study is a retrospective matched control study. The statistic method “Power analysis for matched case–control studies” was used to calculate the number of controls to achieve a power of 80%. The controls were matched against the cases according to gender, year of operation (± 1 year), and age (± 2 years). Four matched controls were randomly selected for each case.

Conditional logistic regression was used to calculate odds ratios (ORs) and p-values in univariable analyses, and the significant data were then analyzed in a multivariable analysis. Statistical tests were two-sided and p < 0.05 considered significant. Data are presented as numbers and percentages or medians and interquartile ranges (IQRs). Statistical analyses were carried out in Stata/IC 15.1 (StataCorp. 2019. Stata: Release 16. Statistical Software. College Station, TX: Stata Corp LLC).

The Regional Ethics Review Board of Gothenburg approved the study (nr: 2019–02,435).

We identified 23 patients with PPH C: 7 females (30%) and 16 males (70%). The matched control group of 92 patients comprised 28 females (30%) and 64 males (70%). The median age at time of surgery was 68 years in both groups. The pathologists’ examination confirmed twenty-two specimens in the PPH C group as pancreatic and periampullary malignancies and one benign inflammation, pancreatitis (Table 2). In the control group, 77 patients (84%) had pancreatic and periampullary malignancies, 7 (8%) had neuroendocrine or other malignant tumors, and 8 (9%) had benign specimens. There were no significant differences considering the number of malignancies between the groups. Venous resections were performed in two patients (9%) in the PPH C group and 15 patients (16%) in the matched control group and did not constitute a risk factor for PPH C in the subgroup analysis (Table 3).

PJ was performed in 7 patients and PG in 16 patients in the PPH C group, and in the control group, the anastomosis was a PJ in 34 patients and a PG in 58 patients. There was no significant difference in the frequencies of the two anastomosis types between the PPH C group and the control group (Table 3). Regarding the number of PG and PJ with confirmed PPH C compared to the total number of patients with PG and PJ in the whole material, the difference was not significant either.

In the PPH C group, 6 patients (26%) had no drain amylase leakage postoperatively. Twelve patients (52%) had BL, but all drains were removed during the first postoperative week. Five of 23 patients (22%) had known pancreatic fistulas, grade B POPF, at the time of the PPH C bleed, and they all had the drains left in place since the primary operation.

Two patients had a known biliary leak when PPH C appeared. One of these patients also had a contemporary grade C POPF detected at the time of bleeding. One patient had an undiagnosed biliary leak and no grade C POPF confirmed at the time of PPH C. These three patients with biliary leak had a BL postoperatively.

After the bleeding had emerged, twenty patients could be verified with a grade C POPF, two with leakage in the biliary anastomosis and one with an abdominal infection at the time of PPH C, which were regarded as underlying casual factors.

In the matched control group, 58 patients (63%) did not have any drain amylase leakage postoperatively, 24 (26%) had BL, 4 (4%) had a grade B POPF, and in 6 (7%) data were missing. The missing data were due to participation in a randomized controlled drain study in three of the patients and absence of registration in old medical records in the remaining three. Three patients (4%) had postoperative biliary leak: one was re-operated, and the leakage subsided in the other two after conservative treatment with drains.

The PPH C occurred between postoperative days 6 and 37 (median 16 days: Table 2). Sentinel bleeding, with blood from drains still in place, openings from drains removed, or gastrointestinal bleeding, appeared in 17 (74%) patients before the onset of PPH C. The sites of bleeding were the hepatic artery and branches in 13 patients (57%), the splenic artery in 3 patients (13%), and the branches of the superior mesenteric artery in 1 patient (4%). No definite bleeding site could be defined in the remaining 6 patients (26%).

The 90-day mortality rate in the PPH C group was 35% (8 patients). In the subgroups based on emergency re-operation, angiographic intervention, or a combination performed in attempt to receive bleeding control, the 90-day mortality rate was 42% (5 out of 12 patients), 14% (1 out of 7 patients), and 50% (2 out of 4 patients), respectively (Table 2).

Twelve patients had an emergency re-operation with the purpose to stop the PPH C. During the re-operation, a grade C POPF was verified in eleven of the cases and an abdominal infection in one case. In nine patients, the remnant pancreases were removed due to total detachment from the jejunum or stomach, continues bleeding, or because proper draining could not be accomplished. In two patients, the remnants were left untouched, and only abdominal drains were placed. In one patient, bleeding control was so poor that no further surgery was considered. Five of the twelve patients died, all within 48 h of the re-operation.

Angiographic interventions with embolization or stenting of the bleeding vessels were performed in 11 patients. In four of these patients, the angiographic intervention was performed before or after an emergency re-operation. During the emergency re-operations or shortly after in an additional re-operation, the remnant pancreas was removed in three out of four patients. In this subgroup treated with both operation and angiographic intervention, one patient died shortly after the re-operation and one within 90 days due to postoperative complications. All the patients treated with a combination of angiographic intervention and operations were observed to have grade C POPFs.

Among the seven patients with solely angiographic procedures to attempt to accomplish bleeding control, four were re-operated to remove the remnant pancreas. The purpose of these “rescue” operations was to avoid future bleeds due to pancreatic fistula leakage, when satisfying draining could not be achieved. A grade C POPF was verified perioperatively in all patients who removed the pancreatic remnant. In two of the patients treated with solely angiographic interventions, there were distinct signs that biliary leakage was the casual factor underlying the PPH C and no signs of a grade C POPF. The patients with biliary leakage were treated with percutaneous transhepatic cholangiography (PTC). One of the patients treated with primary angiographic intervention at a regional hospital re-bled and died shortly thereafter. The autopsy could verify a grade C POPF.

In the matched control group, 4 patients (4%) died within 90 days: two due to recurrent disease with massive liver metastasis and multi-organ failure, one due to pulmonary embolism, and one due to liver failure after an unsuccessful venous resection.

The incidence of PPH C among patients in this study was 4.5%.

The univariable analysis of the postoperative factors recording in both the PPH C group and the control group indicated a median CRP POD 2/3 of 210 g/L, CRP POD 2/3 > 180 mg/L, median CRP POD 5/6 of 140 g/L, CRP POD 5/6 > 180 mg/L, and BL and POPF-B to be significant (p < 0.05) for PPH C development (Table 3). None of the preoperative or perioperative parameters investigated were found to contribute to an increased risk of PPH C in the univariable analysis (Table 3). The distribution and values of the parameters were similar in the PPH C group and control group.

Of the significant parameters in the univariable analyses, only median CRP POD 5/6 of 140 mg/L was significant for PPH C in the multivariable analysis (Table 4).