Langenbeck's Archives of Surgery (2024) 409:21 https://doi.org/10.1007/s00423-023-03200-z RESEARCH The role of clinically relevant intra‑abdominal collections after pancreaticoduodenectomy Clinical impact and predictors. A retrospective analysis from a European tertiary centre Pablo Lopez1,2 · Elizabeth Pando1,2 · Nuria Ortega‑Torrecilla2 · Noelia Puertolas2 · Montse Adell2 · Nair Fernandes2 · Daniel Herms2 · Marta Barros2 · Laia Blanco2 · Joaquim Balsells2 · Ramon Charco2 Received: 17 January 2023 / Accepted: 5 December 2023 © The Author(s) 2023 Abstract Background There is controversial evidence regarding the impact of clinically relevant postoperative intra-abdominal col- lections (CR-IC) on the clinical course after pancreaticoduodenectomy. C-reactive Protein (CRP) has been validated as a predictor of postoperative pancreatic fistula (POPF). Still, its role in predicting CR-IC has not been studied. Methods A retrospective analysis was conducted on patients who underwent PD at a tertiary hospital between October 2012 and October 2017. The incidence of CR-IC, clinically relevant POPF and other complications, as well as mortality and length of hospitalisation, was retrieved. The impact of CR-IR on mortality and major complications was analysed. The serum CRP levels were retrieved on the third and fifth postoperative days (POD3 and POD5), followed by an analysis of sensitivity, specificity, and area under the curve to predict CR-IC using CRP. Results One hundred forty patients were enrolled following inclusion and exclusion criteria. The mean age was 66.5 years (15–83). The incidence of CR-IC was 33.7% (47), and CR-POPF was 24.3%. Pancreatic duct diameter ≤ 4 mm was identi- fied as a risk factor related to CR-IC occurrence. The group of patients who developed CR-IC after PD exhibited a higher rate of complications Clavien-Dindo ≥ III compared to patients without CR-IC (40.4% vs 7.5%, p < 0.001), as well as other events such as admission to the intensive care unit (25.5% vs 4.3%, p < 0.001), the incidence of CR-POPF (66% vs 3.2%, p < 0.001), prolonged hospital stay (32 vs 13 days, p < 0.001), postoperative haemorrhage (23.4 vs 5.4%, p = 0.002), and delayed gastric empty (38.8% vs 11.8%, p < 0.001) respectively. Logistic regression analysis identified CR-IC related to POPF as a risk factor for Clavien-Dindo > III: OR = 10.6 (95% CI: 3.90–28.7). No differences in mortality were reported between the CR-IC group and non-CR-IC group. CRP at postoperative day 3 (POD3) > 17.55 mg/dl and CRP at postoperative day 5 (POD5) > 13.46 mg/dl were predictors of CR-IC (AUC: 0.731 and AUC:0.821, respectively). Conclusions CR-IC has a significant impact after pancreaticoduodenectomy and is associated with a higher incidence of Clavien-Dindo ≥ III complications. Additionally, CRP levels at POD3 and POD5 play a role in predicting CR-IC. Prospective studies are essential to explore strategies for mitigating the occurrence of CR-IC after PD. Keywords Postoperative collections · Pancreatectomy complications · Pancreatic fistula · Surgical complications This publication has been made as part of the Doctoral Program in Introduction Surgery and Morphologic Sciences of the Universitat Autònoma de Barcelona, Spain. Pancreaticoduodenectomy (PD) is one of the most complex surgical procedures in pancreatic surgery. Although mortal- * Elizabeth Pando elizabeth.pando@vallhebron.cat ity rates have decreased over the past decades, PD is still associated with higher rates of postoperative complications 1 Universitat Autónoma de Barcelona, Barcelona, Spain ranging from 25 to 65% [1–3]. 2 Department of Hepato‑Pancreato‑Biliary and Transplant Postoperative intra-abdominal collections (IC) are Surgery, Hospital Universitari Vall d’Hebron, Barcelona, a frequent entity and could be related to postoperative Spain 13 Vol.:(0123456789) 21 Page 2 of 10 Langenbeck's Archives of Surgery (2024) 409:21 complications such as anastomotic leaks (pancreatic fistula, or malignant disease were included between January 2012 bile leak, gastrojejunal leak), infected haematoma or unde- and December 2017. Patients were divided into the CR-IC termined intra-abdominal abscess. However, an ongoing group and non-CR-IC based on the presence or absence of debate exists regarding the clinical implications of postop- clinically relevant intra-abdominal collections (CR-IC). erative intra-abdominal collections, as studies have either associated them with non-relevant complications or found Surgical procedure no complications [4]. In the literature, ICs are not system- atically reported as a separate entity, making it difficult to Until December 2017, all PD procedures were performed determine the true incidence of ICs after PD. Some authors using an open approach. A pylorus-preserving PD was per- reported an IC incidence between 6 and 11% [3, 5–7]. Post- formed in all cases. The pancreatojejunal classical duct- operative intra-abdominal collections can exhibit a clinically to-mucosa anastomosis was carried out using single inter- relevant course (CR-IC), requiring antibiotics, interventional rupted PDS 5/0–6/0 sutures with externalised pancreatic drainage, or surgery, and can be associated with secondary duct catheter of 5Fr for cases with a small-calibre pancre- systemic complications like systemic inflammatory response atic duct (< 4 mm), or an 8Fr catheter for all other cases. syndrome (SIRS) and sepsis [5, 6, 8]. Despite their clini- Termino-lateral hepaticojejunostomy was performed for bil- cal relevance, there are few reports focused on exploring iary reconstruction, and an ante-colic duodenojejunostomy postoperative intra-abdominal collections and their associa- with a single-layer suture was performed for alimentary tract tion with more severe complications (Clavien-Dindo ≥ III) reconstruction. Two abdominal silicone drains of 19Fr with after PD. This knowledge is particularly relevant when vacuum aspiration were placed at the end of the interven- dealing with postoperative pancreatic fistula (POPF). Intra- tion: one anterior and one posterior to the pancreaticojejunal abdominal collections represent a specific manifestation of anastomosis. POPF, resulting in fluid leakage containing active pancreatic enzymes and enteric content. This leakage can lead to severe Postoperative management complications such as sepsis and vascular erosion, ultimately causing life-threatening haemorrhage [6]. Serum C-reactive protein (CRP), complete blood counts, and The current classification’s definition of POPF severity biochemical laboratory tests were determined on the third fails to consider its association with intra-abdominal collec- and fifth postoperative day or when an abnormal clinical tions. We hypothesise that the presence of intra-abdominal course was observed after surgery. collections associated with POPF or other complications CRP levels were determined using a standardised scatter- after PD results in more severe outcomes than their absence. ing turbidimetric assay. The optimal cut-off value for CRP Consequently, pursuing strategies for predicting and miti- on the third and fifth postoperative days in predicting CR-IC gating intra-abdominal collections is justified within this in our cohort was analysed using Youden’s index calculation. context. Somatostatin analogues were not routinely administered; Furthermore, biochemical markers had not been assessed they were only used in cases where POPF occurred. Infected as predictors of CR-IC. Several studies suggest that the ele- POPF was diagnosed through microbiological culture. vation of the serum C-reactive protein (CRP) levels is cor- Computed tomography (CT) scans were not routinely related with the development of a POPF [9, 10]. Neverthe- performed but were carried out in the following situations: less, the role of CRP in predicting CR-IC after PD remains based on the clinician’s judgement, when POPF was estab- unexplored. lished and associated with persistent SIRS or a significant This study aims to explore the clinical impact of postoper- deviation from the expected clinical course, and when CR-IC ative clinically relevant intraabdominal collections (CR-IC) or another type of complication was suspected (duodenoje- associated with POPF and other postoperative complications junal or hepaticojejunal anastomotic leak, intraabdominal after PD. It also aims to evaluate the role of postoperative bleeding, delayed gastric empty associated with POPF). CRP levels in predicting CR-IC. The decision to remove the drains was based on the sur- geon’s judgment, considering factors such as drain output volume, drain characteristics, and the patient's overall condi- Methods tion. Most frequently, the decision to remove the drain was influenced by the presence of the following: clear drain fluid Study design (serous or serohematic fluid), in the absence of unusual out- put like intestinal content, blood, or bile), low volume out- A retrospective observational single cohort was conducted put, lower amylase drain levels, and good clinical evolution at a tertiary referral centre was performed. Adult patients (no signs of SIRS-systemic inflammatory response syndrome, who underwent open pancreaticoduodenectomy for benign abdominal pain or distension, nausea or vomiting). 13 Langenbeck's Archives of Surgery (2024) 409:21 Page 3 of 10 21 After confirming the absence of a pancreatic fistula, the The impact of CR‑IC on Clavien‑Dindo ≥ III, intensive externalised stent is closed in the early postoperative days care unit, and mortality and typically removed approximately 5–6 weeks postopera- tively at the outpatient clinic. An odds ratio analysis was performed to explore the impact of CR-IC and each individual complication on the develop- Postoperative complications and definitions ment of Clavien-Dindo ≥ grade III complications. We also conducted a logistic regression analysis adjusting The postoperative pancreatic fistula was defined according for (1) pre-existing conditions, (2) intraoperative variables, to the ISGPS classification [11]. Only grade B and C fistu- and (3) the type of complications in relation to the occur- lae were considered clinically relevant. Other complications, rence of Clavien-Dindo ≥ III complications. such as delayed gastric emptying and postoperative haemor- We decided not to include infected collections due to the rhage, were also defined according to ISGPS [12]. risk of bias caused by multicollinearity in the regression Clinically relevant intraoperative collection (CR-IC) model, as they are present in most POPF cases. Addition- was defined as the presence of an intra-abdominal collec- ally, this was to avoid overfitting the model because of our tion visualised on the CT scan, including any size or shape small sample size. of intra-abdominal liquid reported by the radiologist; that led to a change in routine clinical practice [13], including Statistical analysis any of the follows: antibiotics administration, percutane- ous, endoscopic or surgical drainage or other interventional The Kolmogorov–Smirnov test was applied to the quantita- procedure, maintenance of intra-abdominal drainages, use tive variables before parametric tests such as the T-student or of somatostatin analogues, the need for parenteral nutrition ANOVA. In cases with a non-normal distribution, non-para- due to ileus secondary to intraabdominal collections, or the metric tests were applied, such as the U-Mann Whitney or development of spontaneous enterocutaneous fistula. When Kruskal–Wallis test. The Chi-square test and relative risk cal- a CR-IC was diagnosed, addressing the underlying cause, if culation were used to analyse qualitative data. A p value less identified, was mandatory. than or equal to 0.05 was considered statistically significant. Suspected infection of IC was assumed when fever (≥ 38 °C), Logistic regression analysis was performed to explore the role increased leukocytes, CRP, or systemic inflammatory response of potential risk factors for CR-IC and Clavien-Dindo ≥ III com- syndrome appeared. The management of suspected infected IC plications. The Youden index was calculated for CRP at POD3 followed a step-up approach, beginning with intravenous anti- and POD5. The area under the curve and the receiving operator biotics, followed by percutaneous or endoscopic drainage, and characteristics (AUC-ROC) were analysed to evaluate CRP in surgical drainage as the last treatment. predicting CR-IC. Postoperative complications were categorised according to the Clavien-Dindo classification. Clavien-Dindo ≥ grade III complications were considered "major" complication [14]. Ethical considerations Our study was approved by the Ethics Committee of our hos- Mortality pital centre (Code PR(AG)81/2019), following the principles of the Declaration of Helsinki for human research. Postoperative mortality was considered at any time during the hospitalisation or up to day 90 after surgery. Analysis of risk factors for CR‑IC development Results For the analysis of risk factors for CR-IC and CR-POPF, Between October 2012 and October 2017, a total of 140 the following variables were included in the model: (1) pre- patients were included in the study. CR-IC and CR-POPF existing conditions: age, sex, pre-existing high blood pres- occurred in forty-seven patients (33.5%) and thirty-four sure, diabetes mellitus, cardiovascular disease (arrhythmia, (24.3%), respectively. myocardial infarction, cardiac valve disease), dyslipidae- There were no significant differences in baseline character- mia, chronic respiratory disease, chronic renal disease, istics between the CR-IC and non-CR-IR groups, except for smoking habit, alcohol habit, chronic pancreatitis as a the incidence of chronic respiratory disease in CR-IC group protective factor, ampullary cancer, (2) intraoperative vari- (Table 1). ables: portal vein (PV) or superior mesenteric vein resec- Regarding intraoperative findings, the incidence of tion (SMV), intraoperative blood transfusions and small small pancreatic duct calibre was higher in the CR-IC pancreatic duct calibre. group compared to the non-CR-IC group (80.9% vs. 13 21 Page 4 of 10 Langenbeck's Archives of Surgery (2024) 409:21 Table 1  Basal characteristics All population Without CR- IC With CR- IC p-value of patients according to the presence of CR-ICs N = 140 n = 93 n = 47 Age, years (median, range) 66.5 (15–83) 64 (15–83) 69 (43–81) 0.074 Male sex, n(%) 77(55%) 49(52.7%) 28(59.6%) 0.439 Pathology, n(%) 0.772 Pancreatic adenocarcinoma 70(50%) 49(52.7%) 21(44.7%) Ampullary adenocarcinoma 16(11.4%) 10(10–8%) 6(12.8%) Neuroendocrine tumour 12(8.6%) 8(8.6%) 4(8.5%) Squamous carcinoma 6(4.3%) 2(2.2%) 4(8.5%) Chronic pancreatitis 9(6.4%) 8(8.6%) 1(2.1%) Cholangiocarcinoma 6(4.3%) 4(4.3%) 2(4.3%) Other benign pancreatic disease 10(7.1%) 6(6.5%) 4(8.5%) Other malignancy 5(3.6%) 3(3.2%) 2(4.3%) IPMN 2(1.4%) 1(1.1%) 1(2.1%) GIST tumour 2(1.4%) 1(1.1%) 1(2.1%) Duodenal cancer 2(1.4%) 1(1.1%) 1(2.1%) Tumoral staging in cancer 0.072 Stage Tis 3(2.7%) 1(1.4%) 2(5.3%) StageIA 9(8.0%) 2(2.7%) 7(18.4%) StageIB 14(12.5%) 9(12.2%) 5(13.2%) StagIIA 22(19.6%) 14(18.9%) 8(21.1%) StageIIB 51(45.5%) 38(18.9%) 13(34.2%) StageIII 8(7.1%) 6(8.1%) 2(5.3%) Stage IV 5(4.5%) 4(5.4%) 1(2.6%) Pre-existing conditions, n(%) Smoking, n(%) 33(23.6%) 21(22.6%) 12(25.5%) 0.698 Alcohol, n(%) 13(9.3%) 10(10.8%) 3(6.4%) 0.305* High blood pressure, n(%) 46(32.9%) 30(32.3%) 16(34%) 0.832 Diabetes mellitus, n(%) 21(15%) 13(14%) 8(17%) 0.634 Cardiovascular disease, n(%) 12(8.6%) 6(6.5%) 6(12.8%) 0.173* Renal chronic disease, n(%) 0 0 0 Respiratory chronic disease, n(%) 9(6.4%) 3(3.2%) 6(12.8%) 0.039* Dyslipemia, n(%) 40(28.6%) 24(25.8%) 16(34%) 0.308 Personal history of cancer (%) 24(17.1%) 17(18.3%) 7(14.9%) 0.616 IPMN: Intrapapillar mucinous neoplasm GIST: Gastrointestinal Stromal Tumour (*) Fisher Test results CR-IC: Clinically relevant intraabdominal collections 49.5%, respectively, p < 0.001). The incidence of PV/VMS Among patients with CR-POPF, thirty-one of them (91%) resection was similar between the groups (p = 0.241). had associated intra-abdominal collections (sixteen of them The incidence of complications was higher in the CR-IC presented Clavien-Dindo ≥ III complications). group compared to the non-CR-IC group, including a higher incidence of Clavien-Dindo ≥ III complications and Drain removal CR-POPF, among others. No differences in ICU admis- sion and mortality between groups were found. Regard- We assessed the timing of drain removal based on the pre- ing CR-IC management, an invasive procedure (percuta- viously explained criteria and the occurrence of CR-POPF. neous or surgical) was necessary for 13 out of 47 (27.6%) We identified the date of drain removal date in 117 clini- patients (Table 2). cal records. The rate of CR-POPF was 5.1% in patients in 13 Langenbeck's Archives of Surgery (2024) 409:21 Page 5 of 10 21 Table 2  Postoperative outcomes according to the presence of CR-ICs All population Without CR- CI With CR-CI p-value N = 140 n = 93 n = 47 Pancreatic fistula (POPF) < 0.001 No pancreatic fistula, n(%) 100(71.4%) 84(90.3%) 16(34%) Biochemical leak, n(%) 6(4.3%) 6(6.5%) 0 Grade B POPF, n(%) 23(16.4%) 3(3.2%) 20(42.6%) Grade C POPF, n(%) 11(7.9%) 0 11(23.4%) Clinically relevant POPF (B-C), n(%) 34(24.3%) 3(3.2%) 31(66%) < 0.001 Delayed gastric empty, n(%) 29(20.7%) 11(11.8%) 18(38.3%) < 0.001 Bile leak, n(%) 10(7.1%) 3(3.2%) 7(14.9%) 0.017* Enterocutaneous fistula, n(%) 3(2.1%) 0 3(6.4%) 0.036* Post-operative haemorrhage, n(%) 16(11.4%) 5(5.4%) 11(23.4%) 0.002 Intraoperative transfusions, median ± SD 0 (0–1) 0 (1–1) 0 (0–1) 0.759 Postoperative transfusions, median ± SD 0 (0–22) 0 (0–6) 1 (0–22) < 0.001 Collections requiring percutaneous drainage, n(%) 4(2.9%) 0 7(14.9%) < 0.001* Collections requiring surgical debridement, n(%) 7(5.0%) 0 4(8.5%) < 0.001* Positive cultures in drainages, n(%) 37(26.4%) 4(4.3%) 33(70.2%) < 0.001 Clavien-Dindo grade of complications, n(%) < 0.001 No complications 48(34.3%) 48(51–6%) 0 I 7(5%) 7(7.5%) 0 II 59(42.1%) 31(33.3%) 28(59.6%) III 12(8.6%) 4(4.3%) 8(17%) IV 10(7.1%) 1(1.1%) 9(19.1%) V 4(2.9%) 2(2.2%) 2(4.3%) Clavien-Dindo ≥ grade III, n(%) 26(18.6%) 7(7.5%) 19(40.4%) < 0.001 Hospital length stay, days (median, range) 17 (8–189) 13 (8–189) 32 (10–184) < 0.001 ICU admission, n(%) 16(11.4%) 4(4.3%) 12(25.5%) < 0.001 Mortality, n(%) 4(2.9%) 2(2.2%) 2(4.3%) 0.412* POPF postoperative pancreatic fistula, CR-IC clinically relevant intraabdominal collections, ICU intensive care unit * Fisher test, otherwise: Chi-square test whom the drains were removed within the first ten days After logistic regression analysis, we found that a pancre- after surgery, compared to 51.3% CR-POPF in patients in atic duct calibre ≤ 4 mm was the only risk factor related to whom the drains were removed beyond the tenth postop- CR-CI development (OR: 4.3, CI 95%: 1.9–9.9). erative day (p < 0.001). For CR-POPF occurrence, pancreatic duct calibre < 4 mm When considering patients whose drains were removed and smoking habit were identified as risk factors (OR: 7.7, within the first five postoperative days, the rate of CR- CI95%: 2.5–24.3 and OR 3.8, CI95%:1.2–7.7 respectively). POPF was 0%. This suggests that the clinical criteria for drain removal were accurate in most cases. Clavien‑Dindo ≥ III complications Complications related to the occurrence of Clavien-Dindo ≥ III Risk factors related to CR‑CI and CR‑POPF complications included CR-IC in general (OR: 8.3), CR-POPF (OR: 8.5), CR-POPF associated with CR-IC (OR: 10.6), postop- Patients with a small pancreatic duct calibre (≤ 4 mm) erative haemorrhage and bile leak. The odds ratio probability for had a higher incidence of CR-CI and CR-POPF compared Clavien-Dindo ≥ III and ICU admission increased when compli- with patients with a pancreatic duct calibre > 4 mm (45.2% cations were associated with CR-IC (Table 3). vs 19.1%, p < 0.001 and 35.7% vs 11.8%, p =  < 0.001 Mortality occurred in four patients of the entire cohort respectively). (2.9%); from these, two cases presented CR-IC secondary 13 21 Page 6 of 10 Langenbeck's Archives of Surgery (2024) 409:21 Table 3  The impact of CR-IC on major outcomes Complications related to Clavien-Dindo III complica- ICU admission tions No (n = 114) Yes (n = 26) p-value OR (CI95%) No (n = 114) Yes (n = 26) p-value OR (CI95%) Events CR-IC in general 28(24.6%) 19(73.1%) < 0.001 8.3 (3.2–21.9) 35(28.2%) 12(75%) < 0.001 7.6 (2.3–25.3) CR-POPF (grade B-C) 18(15.8%) 16(61.5%) < 0.001 8.5 (3.3–21.8) 23(18.5%) 11(68.8%) < 0.001 9.7 (3.0–30.5) CR-POPF + IC 15(48.4%) 16(61.5%) < 0.001 10.6 (4.0–27.5) 20(16.1%) 11(68.8%) < 0.001 11.4 (3.6–36.5) Delayed gastric emptying 20(17.5%) 9(34.6%) 0.053 – 24(19.4%) 5(31.3%) 0.213 – Delayed gastric empty- 11(9.6%) 6(23.1%) 0.660 – 13(10.5%) 4(25%) 0,107 – ing + IC Hemorraghe in general 8(7%) 8(30.8%) 0.002 5.9 (2.0–17.7) 11(8.9%) 5(31.3%) 0.021 4.7 (1.4–15.9) Hemorraghe + IC 3(2.6%) 8(30.8%) < 0.001 16.4 (4.0–67.8) 6(4.8%) 5(31.3%) 0,003 8.9 (2.3–34) Bile leak 2(1.8%) 8(30.8%) < 0.001 24.9 (4.9–126.7) 5(4%) 5(31.3%) 0.002 10.8 (2.7–43.2) Bile leak + IC 1(0.9%) 6(23.1%) < 0.001 33.9 (3.8–296.8) 3(2.4%) 4(25.0%) 0,003 13.4 (2.7–62.7) POPF postoperative pancreatic fistula, CR-IC clinically relevant intraabdominal collections Table 4  Logistic regression of risk factors related to Clavien- CR‑IC prediction sub‑analysis Dindo ≥ III complications Model 1 (*) OR (CI95%) p CRP levels at POD3 were available in 103 cases, CRP Clinically relevant intraab- 10.6 (3.90–28.7) < 0.001 levels at POD5 were available in 123 cases. dominal collections-POPF A sub-analysis of the potential predictors of CR-IC was related performed on 95 patients who completed the CRP deter- Model 2 (*) OR (CI95%) p mination on POD3 and POD5. Cardiovascular disease 5.4 (1.06–27.18) 0.042 Patients were divided into the positive or negative CRP Clinically relevant intraab- 10.6 (3.36–33.65) < 0.001 group according to the Youden's index cut-off, calculated dominal collections-POPF for CRP and CR-IC occurrence, which were 17.55 mg/dl related and 13.46 mg/dl at POD3 and POD5, respectively. Bile leak 21.8 (3.53–134.61) < 0.001 No significant differences in the baseline characteris- * Model 1 included: Pre-existing conditions, intraoperative variables tics were found between groups (CRP-POD5 > 13.55 mg/ and clinically relevant intraabdominal collections-POPF related dl and CRP-POD5 > 17.55 mg/dl), except for a higher inci- ** Model 2 included: Pre-existing conditions, intraoperative variables dence of pre-existing high blood pressure in the group of and clinically relevant intraabdominal collections-POPF related, bile leak, enterocutaneous fistula, post operative haemorrhage, delayed positive CRP at POD3 (Supplementary Table 1). gastric emptying. POPF: post operative pancreatic fistula Intraoperative characteristics to complicated CR-POPF and secondary intra-abdominal haemorrhage; one due to cardiovascular arrhythmia, and A positive CRP at POD3 and POD5 was significantly asso- the last case developed a rupture/dissection of the common ciated with a small pancreatic duct calibre (≤ 4 mm) com- hepatic artery. pared with negative CRP groups (p = 0.012 and p = 0.001, We tested logistic regression models, initially including respectively) (Supplementary Table 2). pre-existing conditions and intraoperative variables and gradually introducing complications, starting with CR-IC and adding other complications. Both CR-IC and CR-POPF Analysis of C‑reactive protein in predicting CR‑IC behaved as related factors. To avoid multicollinearity, we and complications decided to group CR-IC as either related to CR-POPF or unrelated to CR-POPF, resulting in a significant odds ratio A positive CRP at POD3 and POD5 was related to a that remained stable across various model tests. The logistic higher incidence of CR-IC and other postoperative out- regression analysis confirmed that CR-IC related to POPF comes, such as CR-POPF, infected collections, positive was a risk factor for major complications in the models per- microbiological culture in drainages, Clavien-Dindo ≥ III formed (Table 4). complications, ICU stay, postoperative transfusions, and 13 Langenbeck's Archives of Surgery (2024) 409:21 Page 7 of 10 21 hospital length, compared to the negative groups (Sup- the postoperative clinical outcomes when it coexists with plementary Table 3). other classical complications (POPF, bile leak, haemor- Sensitivity and specificity values for CRP-POD3 in pre- rhage). This is one of the few studies reporting CR-IC as a dicting CR-IC was 77.1% (CI95%: 61–-87%) and 61.7% separate entity and describing its negative impact after PD. (CI95%:49–72.9%) and for CRP-POD5: 82.9% (CI95%: F u r t h e r m o r e , we fo u n d t h a t s e r u m C R P- 67.3–91.9%) and 73.3% (CI95%: 61–82.9%) respectively POD3 > 17.55 mg/dl and CRP-POD5 > 13.46 mg/dl were (Supplementary Table 4). The analysis of AUC-ROC in pre- associated with CR-IC development in patients after PD. dicting CR-IC showed an AUC of 0.731 (0.622–0.84) for In our cohort, CRP > 13.46 mg/dl at POD5 showed good CR-POD3 and an AUC 0.821 (0.736–0.906) for CRPOD5 performance as a predictor test of CR-IC and CR-POPF, (Fig. 1). There were no significant differences when compar- with good sensitivity, specificity, and AUC values (0.82). ing both AUCs using the DeLong test (p = 0.062). In the last two decades, postoperative abdominal col- lections have been systematically reported separately from CR‑IC evolution and CRP values classical complications such as CR-POPF, bile leak or infected haematoma. However, these last complications are Serum values of CRP at POD3 and POD5 were relatively frequently associated with intra-abdominal collections and higher when CR-IC was related to CR-POPF compared to consequently, the real incidence of "intra-abdominal collec- bile leak and haemorrhage (Supplementary Table 5). Simi- tions" remains unknown [3, 5, 7]. larly, serum values of CRP at POD3 and POD5 were higher Recently, some authors reported the incidence of intra- when an interventional procedure (surgical or percutaneous abdominal collections after PD to be around 50–60% (clini- drainage) was needed (Fig. 2). cally relevant or not). These studies found a relationship between intra-abdominal collections in general and the development of clinically relevant POPF [15, 16]. However, Discussion the clinical impact of IC remains controversial, found to be related only to “minor” complications (Clavien-Dindo ≤ IIIa) CR‑IC: an old entity playing a crucial role after PD [4], while some publications reported an increased rate of Clavien-Dindo ≥ III complications (55% vs 35%, p = 0.033), This study explores the clinical impact of CR-IC after pan- as well as the rate of CR-POPF when intra-abdominal col- creaticoduodenectomy (PD) and its prediction using serum lections were present (29% vs 11%, p = 0.020) [6]. CRP. Our results establish CR-IC as a significant event after The relevance of reporting CR-IC relies on the following PD, closely associated with Clavien-Dindo ≥ III, comparable hypothesis: a poorly drained collection secondary to a sur- to CR-POPF. Additionally, the presence of CR-IC aggravates gical complication (POPF, bile leak, haematoma, intestinal Fig. 1  Receiver operating char- acteristic (ROC) and area under the curve (AUC) of CRP-PO3 and CRP-PO5 for clinically rel- evant intrabdominal collections CRP-PO3 CRP-PO5 Sensitivity Specificity 13 21 Page 8 of 10 Langenbeck's Archives of Surgery (2024) 409:21 Fig. 2  CRP-PO3 and CRP-PO5 performance according to the type of treatment over CR-ICs fistulae) will worsen the clinical prognosis of the compli- occurs, we hypothesise that “well-positioned drains” in a cations itself, increasing the probability of interventional high-risk pancreatic anastomosis (PA) would have an indo- or surgical procedures, sepsis, or organ failure, contrary to lent clinical course, compared to a “non” well-positioned a well-drained complication. Even in a scenario of distal drains around (PA), with the secondary development of an pancreatectomy, in which there is no risk of leakage from intra-abdominal abscess or diffuse peritonitis, as suggested enteric anastomosis, a recent report indicates that patients by the previous studies [18]. with CR-POPF collections had an increased incidence of Regarding whether to drain a pancreatic anastomosis cor- other pancreatic surgery-specific complications. rectly, a prospective study conducted by the Verona Group The clinical relevance of POPF associated with an intra- analysed the phenomenon of “drains dislocation” following abdominal collection (IC) varies according to the type of pancreatic surgical resections (both PD and DP) [19]. The treatment or the development of systemic complications. For study revealed that drain dislocation was a frequent occur- POPF-associating IC, a percutaneous (Clavien-Dindo IIIa) rence by POD3 and did not significantly impact postoperative or surgical drainage (Clavien-Dindo IIIb) could be neces- complications. In fact, it observed a lower incidence of POPF sary. In the presence of organ failure, a Clavien-Dindo IVa or and major complications. However, the study did not assess IVb will occur depending on single or multiple organ failure. the precise location of drains in relation to the pancreatic anas- On the other hand, a POPF without an IC will probably tomosis or pancreatic stump in the group of “non-dislocated” have an indolent clinical course, needing antibiotics or the drains. Therefore, further evidence is required to elucidate maintenance of surgical drains (also considered a grade B, the role of drain placement, the number of drains, and the but a Clavien-Dindo grade II). Considering that, it is still distance from the anastomosis after PD and their influence being debated whether grade B-POPF has to be redefined, on the development of clinically relevant POPF collections. approaching a most realistic classification to differentiate The use of routine intra-abdominal drains after pancrea- whether a grade B POPF will be more deleterious inside tectomy is still a debated topic. Previous studies concluded the same classification group [17]. Our study provides that routine intra-abdominal drainage strategies did not offer new data regarding the impact of CR-IC in aggravating a advantages over non-routine use; however, no stratification CR-POPF clinical outcome, increasing the risk of Clavien- by high-risk pancreas was performed, and the sample size Dindo ≥ III complications and ICU admission when both includes distal pancreatectomy in some studies [20–22]. entities coexist. Although those classical studies advocating a non-routine In that line, if relevant postoperative collections are the use of drainages did not perform a routine radiological com- key in the clinical evolution after PD, mainly when POPF puted tomography scan during the postoperative period, the 13 Langenbeck's Archives of Surgery (2024) 409:21 Page 9 of 10 21 impact of mispositioned drains in the “routine use” group Conclusions was not reported [20]. A recent meta-analysis found that leaving routine drains after a PD with a high pancreatic fis- This study has identified CR-IC as a significant complica- tula risk is beneficial [23, 24]. tion following pancreaticoduodenectomy, mainly when Additionally, well-designed randomised control trials associated with POPF. Additionally, CRP levels at POD3 have concluded that non-routine intra-abdominal drainage and POD5 play a role in predicting CR-IC and CR-POPF. after PD increased mortality (12% vs. 3%); mortality was Prospective studies are necessary to investigate methods to significantly higher in patients with risk factors for POPF reduce the occurrence of CR-IC after PD. (small pancreatic duct calibre, soft pancreas texture), con- firming that non-routine drains after PD would not be recom- Supplementary Information The online version contains supplemen- tary material available at https://​doi.​org/​10.​1007/​s00423-​023-​03200-z. mended in high-risk pancreatic anastomosis [25]. Our cohort had 64.2% of patients with small pancreatic duct calibre Authors’ contributions EP devised the project and the main conceptual (≤ 4 mm). While all patients received routine postoperative idea. PL and EP wrote the main manuscript text. PL, NO, NP, MA and drains, 45.9% developed CR-IC. One limitation of this ret- NF contribute to data collection and follow-up during the study period. DH, MB, and LB contributed to the final version of the manuscript. rospective study is that we did not assess the exact position JB and RCH participated as senior reviewers. All authors reviewed of drains and their relation to the development of CR-IC and the manuscript. other major complications. Funding Open Access Funding provided by Universitat Autonoma de Barcelona. CR‑IC prediction Data Availability The authors confirm the accessibility of the data sub- C-reactive protein has been validated as an anastomotic stantiating the results presented in this study, which can be accessed dehiscence and intra-abdominal infection predictor in fields in the main manuscript and its Supplementary Material by requesting the data at https://​datav​erse.​csuc.​cat/ other than pancreatic surgery, such as colorectal and oesoph- ageal surgery [26, 27]. Thereby, concomitant collections or Declarations infections may explain the release of acute-phase reactant, which could be related to an unfavourable clinical course. Competing interests The authors declare no competing interests. According to previous investigations, CRP values at POD3 or POD5 have helped predict clinically relevant POPF. Open Access This article is licensed under a Creative Commons Attri- (grade B-C according to ISGPS) [28–33]. Our study among bution 4.0 International License, which permits use, sharing, adapta- tion, distribution and reproduction in any medium or format, as long the few in the literature to demonstrate that CRP could also as you give appropriate credit to the original author(s) and the source, help identify a CR-IC, one of the most relevant complica- provide a link to the Creative Commons licence, and indicate if changes tions in our cohort. C-reactive protein at POD5 showed a were made. The images or other third party material in this article are slight superiority over CRP-POD3 in predicting CR-IC, but included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in no statistical differences were found after applying the De the article's Creative Commons licence and your intended use is not Long test. Interestingly, CRP-POD3 and CRP-POD5 values permitted by statutory regulation or exceeds the permitted use, you will tended to be higher when associated with the need for inter- need to obtain permission directly from the copyright holder. To view a ventional procedures or associated with the development of copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. POPF when an intra-abdominal collection occurred. C-reactive protein is a widely available and easily meas- ured method. The findings of this study could help guide deci- References sions to perform an early CT scan to identify CR-IC requiring drainage or to initiate empirical antibiotic treatment. 1. Gooiker GA, Lemmens VE, Besselink MG et al (2014) Impact of One of the biases of this study is its retrospective nature, centralization of pancreatic cancer surgery on resection rates and which resulted in the loss of some relevant data, such as survival. Br J Surg 101(8):1000–1005. https://​doi.​org/​10.​1002/​ bjs.​9468 pancreatic texture in the surgical report and intraoperative 2. Yeo CJ (2003) The Whipple operation: is a radical resection of blood loss. Another bias is the unknown incidence of all benefit? Adv Surg 37:1–27 intra-abdominal collections, including those not clinically 3. Winter JM, Cameron JL, Campbell KA et al (2006) 1423 pan- relevant. Additionally, the incidence of “well-positioned” creaticoduodenectomies for pancreatic cancer: a single-institution experience. J Gastrointest Surg 10(9):1199–210. https://​doi.​org/​ drainages and their relationship with outcomes is unknown. 10.​1016/j.​gassur.​2006.​08.​018. (discussion 1210-1) However, this study reflects usual clinical practice, consid- 4. Zhao N, Cui J, Yang Z et al (2019) Natural history and thera- ering only clinically relevant intra-abdominal collections, peutic strategies of post-pancreatoduodenectomy abdominal fluid where any deviation from the ordinary postoperative clinical collections: ten-year experience in a single institution. Medicine 98(22):e15792. https://​doi.​org/​10.​1097/​MD.​00000​00000​015792 course is present. 13 21 Page 10 of 10 Langenbeck's Archives of Surgery (2024) 409:21 5. Sierzega M, Kulig P, Kolodziejczyk P, Kulig J (2013) Natural trial of the value of intraperitoneal drainage after pancreatic resec- history of intra-abdominal fluid collections following pancreatic tion. Ann Surg 234(4):487–494. https://​doi.​org/​10.​1097/​00000​ surgery. J Gastrointest Surg 17(8):1406–1413. https://​doi.​org/​10.​ 658-​20011​0000-​00008 1007/​s11605-​013-​2234-1 21 Nickel F, Lang F, Kowalewski K et al (2022) Pancreatic surgery 6. Maehira H, Iidam H, Matsunaga T et al (2020) The location of with or without drainage: propensity score-matched study. Br J perianastomotic fluid collection predicts postoperative complica- Surg 109:739–745. https://​doi.​org/​10.​1093/​bjs/​znac1​23 tions after pancreaticoduodenectomy. Langenbeck’s Arch Surg 22. Witzigmann H, Diener MK, Kienkötter S et al (2016) No need for 405(3):325–336. https://​doi.​org/​10.​1007/​s00423-​020-​01880-5 routine drainage after pancreatic head resection: the dual-center, ran- 7. Fernández-del Castillo C, Morales-Oyarvide V, McGrath D et al domized, controlled PANDRA Trial (ISRCTN04937707). Ann Surg (2012) Evolution of the Whipple procedure at the Massachusetts 264(3):528–537. https://​doi.​org/​10.​1097/​SLA.​00000​00000​001859 General Hospital. Surgery 152(3 Suppl 1):S56–S63. https://​doi.​ 23 Brubaker LS, Casciani F, Fisher W et al (2022) A risk-adjusted org/​10.​1016/j.​surg.​2012.​05.​022 analysis of drain use in pancreaticoduodenectomy: Some is good, 8. Sohn T, Yeo C, Cameron J et al (2003) Pancreaticoduodenectomy: but more may not be better. Surgery 171(4):1058–1066. https://​ role of interventional radiologists in managing patients and com- doi.​org/​10.​1016/j.​surg.​2021.​07.​026 plications. J Gastroint Surg 7(2):209–219. https://d​ oi.o​ rg/1​ 0.1​ 016/​ 24. Liu X, Chen K, Chu X, Liu G, Yang Y, Tian X (2021) Prophylac- s1091-​255x(02)​00193-2 tic Intra-peritoneal drainage after pancreatic resection: an updated 9. Giardino A, Spolverato G, Regi P et al (2016) C-Reactive pro- meta-analysis. Front Oncol 11. https://d​ oi.o​ rg/1​ 0.3​ 389/f​ onc.2​ 021.​ tein and procalcitonin as predictors of postoperative inflamma- 658829 tory complications after pancreatic surgery. J Gastrointest Surg 25 Van Buren G, Bloomston M, Hughes S et al (2014) A rand- 20(8):1482–1492. https://​doi.​org/​10.​1007/​s11605-​016-​3171-6 omized prospective multicenter trial of pancreaticoduodenec- 10. Smits F, Henry A, van Eijck C et al (2020) Care after pancreatic tomy with and without routine intraperitoneal drainage. Ann Surg resection according to an algorithm for early detection and mini- 259(4):605–612. https://d​ oi.o​ rg/1​ 0.1​ 097/S ​ LA.0​ 00000​ 00000​ 00460 mally invasive management of pancreatic fistula versus current 26 Cousin F, Ortega-Deballon P, Bourredjem A, Doussot A, Giac- practice (PORSCH-trial): design and rationale of a nationwide caglia V, Fournel I (2016) Diagnostic Accuracy of procalcitonin stepped-wedge cluster-randomized trial. Trials 21(1). https://​doi.​ and C-reactive Protein for the early diagnosis of intra-abdominal org/​10.​1186/​s13063-​020-​4167-9 infection after elective colorectal surgery: a meta-analysis. Ann 11. Bassi C, Marchegiani G, Dervenis C et al (2016) The 2016 update Surg 264(2):252–256. https://​doi.​org/​10.​1097/​SLA.​00000​00000​ of the International Study Group (ISGPS) definition and grading 001545 of postoperative pancreatic fistula: 11 years after. Surgery. https://​ 27 Warschkow R, Beutner U, Steffen T et al (2012) Safe and early doi.​org/​10.​1016/j.​surg.​2016.​11.​014 discharge after colorectal surgery due to C-reactive protein: a 12. Wente MN, Veit JA, Bassi C et al (2007) Postpancreatectomy diagnostic meta-analysis of 1832 patients. Ann Surg 256(2):245– hemorrhage (PPH): an International Study Group of Pancreatic 250. https://​doi.​org/​10.​1097/​SLA.​0b013​e3182​5b60f0 Surgery (ISGPS) definition. Surgery 142(1):20–25. https://​doi.​ 28. Guilbaud T, Birnbaum D, Lemoine C et al (2018) C-Reactive Pro- org/​10.​1016/j.​surg.​2007.​02.​001 tein on postoperative day 1 is a reliable predictor of pancreas-spe- 13. Strasberg SM, Linehan DC, Clavien P-A, Barkun JS (2007) Pro- cific complications after pancreaticoduodenectomy. J Gastrointest posal for definition and severity grading of pancreatic anastomosis Surg 22(5):818–830. https://​doi.​org/​10.​1007/​s11605-​017-​3658-9 failure and pancreatic occlusion failure. Surgery 141(4):420–426. 29. Mintziras I, Maurer E, Kanngiesser V, Bartsch D (2020) C-reac- https://​doi.​org/​10.​1016/j.​surg.​2006.​12.​001 tive protein and drain amylase accurately predict clinically rel- 14 Dindo D, Demartines N, Clavien P (2004) Classification of surgi- evant pancreatic fistula after partial pancreaticoduodenectomy. Int cal complications: a new proposal with evaluation in a cohort of J Surg (London, England) 2020:76. https://​doi.​org/​10.​1016/j.​ijsu.​ 6336 patients and results of a survey. Ann Surg 240(2):205–213. 2020.​02.​025 https://​doi.​org/​10.​1097/​01.​sla.​00001​33083.​54934.​ae 30 Pecorelli N, Guarneri G, Palucci M et al (2022) Early biochemi- 15 Bruno O, Brancatelli G, Sauvanet A, Vullierme M, Barrau V, cal predictors of clinically relevant pancreatic fistula after distal Vilgrain V (2009) Utility of CT in the diagnosis of pancreatic fis- pancreatectomy: a role for serum amylase and C-reactive pro- tula after pancreaticoduodenectomy in patients with soft pancreas. tein. Surg Endosc 36(7):5431–5441. https://​doi.​org/​10.​1007/​ AJR Am J Roentgenol 193(3):W175–W180. https://​doi.​org/​10.​ s00464-​021-​08883-3 2214/​AJR.​08.​1800 31 Chen G, Yi H, Zhang J (2021) Diagnostic value of C-reactive 16. Hashimoto M, Koga M, Ishiyama K et al (2007) CT features of protein and procalcitonin for postoperative pancreatic fistula fol- pancreatic fistula after pancreaticoduodenectomy. AJR Am J Roent- lowing pancreatoduodenectomy: a systematic review and meta- genol 188(4):W323–W327. https://​doi.​org/​10.​2214/​AJR.​05.​1174 analysis. Gland Surg 10(12):3252–3263. https://d​ oi.o​ rg/1​ 0.2​ 1037/​ 17. Maggino L, Malleo G, Bassi C, Allegrini V, McMillan MT, Borin gs-​21-​658 A, Chen B, Drebin JA, Ecker BL, Fraker DL, Lee MK, Paiella 32. Nakano Y, Endo Y, Kitago M et al (2023) Clinical characteristics S, Roses RE, Salvia R, Vollmer CM (2019) Decoding grade B and predictive factors of postoperative intra-abdominal abscess Pancreatic fistula: a clinical and economical analysis and subclas- after distal pancreatectomy. Langenbeck’s Arch Surg 408(1). sification proposal. Ann Surg 269(6):1146–1153. https://​doi.​org/​ https://​doi.​org/​10.​1007/​s00423-​023-​02914-4 10.​1097/​SLA.​00000​00000​002673 33. Palani Velu L, McKay C, Carter C, McMillan D, Jamieson N, 18 Blot S, Antonelli M, Arvaniti K, Blot K, Vogelaers D (2019) Epi- Dickson E (2016) Serum amylase and C-reactive protein in risk demiology of intra-abdominal infection and sepsis in critically ill stratification of pancreas-specific complications after pancreati- patients: “AbSeS”, a multinational observational cohort study and coduodenectomy. Br J Surg 103(5):553–563. https://​doi.​org/​10.​ ESICM Trials Group Project. Intensive care Med 45(12):1703– 1002/​bjs.​10098 1717. https://​doi.​org/​10.​1007/​s00134-​019-​05819-3 19 Marchegiani G, Ramera M, Viviani E et al (2019) Disloca- Publisher's Note Springer Nature remains neutral with regard to tion of intra-abdominal drains after pancreatic surgery: results jurisdictional claims in published maps and institutional affiliations. of a prospective observational study. Langenbeck’s Arch Surg 404(2):213–222. https://​doi.​org/​10.​1007/​s00423-​019-​01760-7 20. Conlon KC, Labow D, Leung D, Smith A, Jarnagin W, Coit DG, Merchant N, Brennan MF (2001) Prospective randomized clinical 13