Abstract
Background
The ability of drain fluid amylase (DFA) to predict pancreatic fistula (POPF) and safely guide early drain removal is unknown in high-risk pancreatoduodenectomy (HR-PD). The aim of this analysis was to develop a new risk-based algorithm for drain management after pancreatoduodenectomy.
Methods
All consecutive patients undergoing pancreatoduodenectomy (2019–2024) at two institutions were included. During surgery, patients were stratified into HR-PD, defined according to the International Study Group on Pancreatic Surgery score (soft gland and duct ≤ 3 mm), and non-HR-PD. Postoperative predictors (DFA, serum amylase/lipase, C-reactive protein) were retrieved on postoperative days (PODs) 1 to 5 and correlated to POPF development using a risk-based approach. Results were validated in a retrospective cohort (2015–2019) of pancreatoduodenectomies performed at the same institutions.
Results
Out of 825 screened patients, 788 were included. HR-PD (251, 32%) had higher POPF rates (50% versus 9%; P < 0.001) and worse outcomes. In HR-PD, DFA had a lower area under the curve for POPF for POD 1 (0.73 versus 0.85; P = 0.003) and POD 3 (0.74 versus 0.88; P = 0.002) compared with non-HR-PD, but similar for POD 5 (0.83 versus 0.86; P = 0.535), whereas serum amylase/lipase maintained a low area under the curve for POPF compared with non-HR-PD. The optimal criteria for POD 3 drain removal in non-HR-PD were POD 1 DFA < 200 UI/l, POD 3 DFA < 150 UI/l, and absence of elevated serum amylase/lipase on POD 1 (Youden’s index 0.674). In HR-PD, the criteria for POD 5 drain removal were POD 5 DFA < 150 UI/l and POD 5 C-reactive protein < 150 mg/ml (Youden’s index 0.593).
Conclusion
In patients with an intrinsic higher risk of POPF, early predictors are less reliable, and drain management protocols should be tailored. In non-HR-PD, an early removal policy can be recommended based on DFA and serum amylase/lipase. In HR-PD, drain removal may be safely deferred until POD 5 according to DFA and C-reactive protein.
Keywords: pancreatectomy, postoperative, POPF, pancreatic, amylase, stent
This multicentre study of 788 patients undergoing pancreatoduodenectomy (PD) evaluates the reliability of drain fluid amylase (DFA) as a predictor of postoperative pancreatic fistula among high-risk (HR) and non-HR-PD. We found that, whereas early drain removal guided by DFA and serum enzymes is safe in non-HR-PD, DFA accuracy is reduced in HR-PD until postoperative day 5, when combined DFA and C-reactive protein levels allow safe drain removal. These findings support a dynamic, risk-based drain management protocol tailored to pancreatic fistula risk.
Introduction
Early drain removal is a fundamental step in the enhanced recovery after surgery pathway in patients undergoing pancreatoduodenectomy (PD)1. Its implementation in postoperative care is associated with improved outcomes for postoperative pancreatic fistula (POPF), other pancreas-specific morbidity, intra-abdominal infections, and length of hospital stay (LOS)2–4. The timely identification of patients at the lowest risk of developing a POPF is essential for the safe implementation of early drain removal policies5.
Several protocols have been proposed to guide decision-making, most of which are based on assessing drain fluid amylase (DFA) during the early postoperative days (PODs)6–10. These protocols typically rely on definite DFA thresholds, usually associated to a high negative predictive value (NPV). However, a protocol based solely on a single DFA threshold is not suitable for all clinical scenarios. A ‘one-size-fits-all’ approach is inherently inadequate, given the significant effect of patient- and pancreas-specific risk factors as well as the dynamic nature of POPF prediction and diagnosis during the early postoperative course8,9,11–13. A risk-based strategy for POPF prediction and mitigation is currently the standard of care during PD, but there are no standardized or widely applied protocols for postoperative drain management specifically for patients with significantly increased POPF risk. Moreover, variables beyond DFA may be employed not only to predict POPF severity and guide appropriate management but also to inform drain removal14–16.
In the case of high-risk PD (HR-PD), typically defined by a soft gland and not dilated pancreatic duct17, early drain removal based solely on DFA could be even detrimental. Yet, no alternative reliable predictors are currently available to guide decision-making in this critical subset of patients. Other factors, such as postoperative serum hyperamylasaemia/hyperlipasaemia18–20 or C-reactive protein (C-RP), may help improve patient stratification for early drain removal. Moreover, a dynamic reassessment of these parameters during the first PODs may be required to improve prognostic performance.
The aim of this study was to monitor dynamically the values of DFA and other potential predictors of POPF during the early postoperative period, correlate them with the onset of POPF, and establish a new risk-adapted algorithm protocol for postoperative drain management in patients undergoing PD.
Methods
Study design
The study was designed as a retrospective analysis of prospectively collected data. All patients undergoing PD from 1 January 2019 to 31 December 2024 at two high-volume hepato-pancreato-biliary surgery institutions were considered eligible. The study period was the same for both institutions. The follow-up was conducted over a 90-day recovery period. The study was performed according to the STROBE guidelines21 (see Supplementary material). Validation was performed in a retrospective cohort of patients who underwent PD at the two institutions before the study period (2015–2019).
Procedures
PD was performed according to institutional standards. During surgery, the presumed preoperative pathology, estimated blood loss, intraoperative pancreatic texture, and main pancreatic duct (MPD) size were recorded22. Fistula risk was assessed using the International Study Group on Pancreatic Surgery (ISGPS) four-tier classification17, which incorporates MPD size and gland texture. HR-PD was defined as ISGPS ‘class D’: soft gland texture with MPD size ≤ 3 mm.
All patients underwent pancreatico-jejunostomy, either single- or double-layered with duct-to-mucosa, or, less frequently, pancreatico-gastrostomy. An externalized trans-anastomotic stent (ETS) (PankreaPlus™ polyvinyl catheter, Munich, Germany) was employed as a specific mitigation strategy for POPF prevention in HR-PD and in general in case of MPD ≤ 3 mm, with a previously described surgical technique23. Patients where an internalized trans-anastomotic stent was employed were excluded from the present analysis.
Two open, passive drains were placed in the proximity of the pancreatic and biliary anastomoses in all patients. Early drain removal was defined as removal on PODs 1–3. According to institutional standards, early removal policies were followed in non-HR-PD patients, based on the drain fluid aspect and POD 1 DFA3,13,24. In all HR-PD patients, drains were left in place until POD 5, except in patients with very low output (≤ 20 cc). From POD 5 onward, in both non-HR-PD patients without early removal criteria and in all HR-PD patients, drains were periodically re-evaluated based on their appearance, output volume, amylase levels, and clinical parameters, and were removed at the surgeon’s discretion in the absence of predefined protocols.
Postoperative POPF indicators
Prospective collection of postoperative biochemical data was maintained during the entire duration of the study period and included DFA and serum C-RP (normal range 0–5 mg/ml) on PODs 1, 3, and 5, and serum pancreatic amylase (SA) (normal range 13–53 U/l) or serum lipase (SL) (normal range 5–60 U/l) on PODs 1, 2, and 3. Patients with missing DFA activity on both POD 1 and 3 were excluded from the analysis. The study flowchart is displayed in Fig. 1. For the retrospective validation cohort, DFA values on PODs 1, 3, and (if available) 5 were retrieved.
Fig. 1.
Study flowchart with intra- and postoperative outcomes
POD, postoperative day; DFA, drain fluid amylase; ISGPS, International Study Group on Pancreatic Surgery; HR-PD, high-risk pancreatoduodenectomy, ETS, externalized trans-anastomotic stent; POPF, postoperative pancreatic fistula; PPAP, post-pancreatectomy acute pancreatitis; PPH, post-pancreatectomy haemorrhage.
Outcomes of interest
The primary outcome was the incidence of POPF, defined according to the ISGPS25. Secondary outcomes included the severity of POPF (grade B or C, per ISGPS25). All postoperative outcomes were registered prospectively, including post-pancreatectomy haemorrhage (PPH)26, delayed gastric emptying (DGE)27, postoperative hyperamylasaemia, and post-pancreatectomy acute pancreatitis (PPAP)28, 90-day hospital mortality and readmission, major morbidity (defined as Clavien–Dindo ≥ 3)29, and LOS.
Statistical analysis
Continuous variables were expressed as median (interquartile range (i.q.r.)). Differences in variables between or among groups were tested using Student’s t-test for continuous variables, and the χ2 test or Fisher’s exact test for categorical variables. Receiver operating characteristic (ROC) curves were used to identify the optimal cut-off values of DFA with the highest performance for ruling out the occurrence of POPF, and the areas under the curve (AUC) were compared using the DeLong test for paired ROC curves. Missing-at-random variables were considered acceptable for exclusion without imputation with a threshold of ≤ 10% in the complete-case analysis. To evaluate the missing-at-random assumption, baseline characteristics of participants with complete data were compared with those of participants with missing values. Youden’s index (Youden) was used to select the most appropriate cut-off, along with the cut-off’s clinical relevance (for example, the number of patients for whom the cut-off can be applied and the POPF rate in the population selected by the cut-off). For each cut-off, sensitivity (SENS), specificity (SPEC), positive predictive value (PPV), NPV, and prevalence of patients with values above the threshold in the examined population (Prev+) were calculated. Multivariable analysis was performed using a logistic regression model, with results expressed as odds ratios (ORs) and 95% confidence intervals (c.i.). P values less than 0.05 were considered statistically significant. Statistical analyses were performed using STATA14 for Windows. Figures were created with BioRender.com.
Results
A total of 825 patients were enrolled (Fig. 1). Of these, 33 patients who had an internalized trans-anastomotic stent were excluded. An additional four patients were excluded due to unavailable DFA measurements on both PODs 1 and 3, resulting in a final study cohort of 788 patients.
Among the included patients, 251 (32%) were considered HR-PD and 537 (68%) non-HR-PD (Table 1 and Table S1). Compared with non-HR-PD, HR-PD patients had higher median body mass index (25 (i.q.r. 22–27) versus 24 (i.q.r. 22–27); P = 0.018), were less likely to have diabetes (17% versus 32%; P < 0.001), underwent jaundice palliation (46% versus 60%; P = 0.001) or neoadjuvant therapy less frequently (13% versus 35%; P < 0.001), and had a presumed diagnosis of pancreatic ductal adenocarcinoma (35% versus 73%; P < 0.001) or the need for vascular resection less frequently (9% versus 16%; P = 0.008).
Table 1.
Intraoperative and postoperative profile of included patients, according to POPF risk
| Total (n = 788) | HR-PD | P | ||
|---|---|---|---|---|
| No (n = 537, 68%) | Yes (n = 251, 32%) | |||
| Intraoperative | ||||
| PD type | 0.106 | |||
| Pylorus-preserving | 677 (86%) | 454 (85%) | 223 (89%) | |
| Whipple | 111 (14%) | 83 (15%) | 28 (11%) | |
| Surgical approach | 0.057 | |||
| Open | 694 (88%) | 481 (90%) | 213 (85%) | |
| Robotic | 94 (12%) | 56 (10%) | 38 (15%) | |
| Pancreatic anastomosis | 0.462 | |||
| Pancreatico-jejunostomy | 776 (98%) | 530 (99%) | 246 (98%) | |
| Pancreatico-gastrostomy | 12 (2%) | 7 (1%) | 5 (2%) | |
| External trans-anastomotic stent | < 0.001 | |||
| No | 479 (61%) | 442 (82%) | 37 (15%) | |
| Yes | 309 (39%) | 95 (18%) | 214 (85%) | |
| Vascular resection | 106 (13%) | 84 (16%) | 22 (9%) | 0.008 |
| Operative time (min), median (i.q.r.) | 405 (335–475) | 400 (330–470) | 420 (340–480) | 0.048 |
| Blood loss (ml), median (i.q.r.) | 300 (200–400) | 300 (200–400) | 300 (200–450) | 0.882 |
| MPD size (mm), median (i.q.r.) | 4 (3–5) | 5 (4–6) | 2 (2–3) | < 0.001 |
| MPD size ≤ 3 mm | 354 (45%) | 103 (19%) | 251 (100%) | < 0.001 |
| Pancreatic texture | < 0.001 | |||
| Hard | 430 (55%) | 430 (80%) | _ | |
| Soft | 358 (45%) | 107 (20%) | 251 (100%) | |
| ISGPS class | < 0.001 | |||
| A | 327 (42%) | 327 (61%) | _ | |
| B | 103 (13%) | 103 (19%) | _ | |
| C | 107 (13%) | 107 (20%) | _ | |
| D | 251 (32%) | _ | 251 (100%) | |
| Postoperative | ||||
| POPF | 173 (22%) | 48 (9%) | 125 (50%) | < 0.001 |
| POPF grade | < 0.001 | |||
| B | 143 (18%) | 36 (7%) | 107 (43%) | |
| C | 30 (4%) | 12 (2%) | 18 (7%) | |
| PPAP | 57 (7%) | 13 (2%) | 44 (18%) | < 0.001 |
| PPAP grade | < 0.001 | |||
| POH | 159 (20%) | 59 (11%) | 100 (40%) | |
| B | 39 (5%) | 8 (1%) | 31 (13%) | |
| C | 18 (2%) | 5 (1%) | 13 (5%) | |
| Intra-abdominal collection (deep surgical site infection) | 155 (20%) | 58 (11%) | 97 (39%) | < 0.001 |
| Chylous fistula | 38 (5%) | 32 (6%) | 6 (2%) | 0.029 |
| PPH | 105 (13%) | 59 (11%) | 46 (18%) | 0.005 |
| PPH grade | 0.025 | |||
| A | 16 (2%) | 11 (2%) | 5 (2%) | |
| B | 46 (6%) | 25 (5%) | 21 (8%) | |
| C | 43 (5%) | 23 (4%) | 20 (8%) | |
| DGE | 180 (23%) | 99 (18%) | 81 (32%) | < 0.001 |
| DGE grade | < 0.001 | |||
| A | 49 (6%) | 34 (6%) | 15 (6%) | |
| B | 73 (9%) | 37 (7%) | 36 (14%) | |
| C | 58 (7%) | 28 (5%) | 30 (12%) | |
| Sepsis | 75 (10%) | 35 (7%) | 40 (16%) | < 0.001 |
| Percutaneous drainage | 63 (8%) | 26 (5%) | 37 (15%) | < 0.001 |
| Relaparotomy | 75 (10%) | 37 (7%) | 38 (15%) | < 0.001 |
| Unplanned intensive care unit admission | 99 (13%) | 47 (9%) | 52 (21%) | < 0.001 |
| POD of drain removal (days), median (i.q.r.) | 5 (3–8) | 4 (3–6) | 10 (5–23) | < 0.001 |
| PODs 1–3 (early) drain removal | 226 (29%) | 217 (40%) | 9 (4%) | < 0.001 |
| PODs 4–5 drain removal | 186 (24%) | 142 (26%) | 44 (18%) | 0.006 |
| Discharged with drains | 94 (12%) | 34 (6%) | 60 (24%) | < 0.001 |
| LOS (days), median (i.q.r.) | 10 (7–17) | 9 (7–13) | 15 (9–25) | < 0.001 |
| Readmission (30 days) | 71 (9%) | 38 (7%) | 33 (13%) | 0.006 |
| Major morbidity | 168 (21%) | 82 (15%) | 86 (34%) | < 0.001 |
| 90-day mortality | 26 (3%) | 13 (2%) | 13 (5%) | 0.043 |
Values are n (%) unless otherwise stated. POPF, postoperative pancreatic fistula; HR-PD, high-risk pancreatoduodenectomy; min, minutes; i.q.r., interquartile range; MPD, main pancreatic duct; ISGPS; International Study Group on Pancreatic Surgery; PPAP, post-pancreatectomy acute pancreatitis; POH; postoperative hyperamylasaemia; PPH, post-pancreatectomy haemorrhage; DGE, delayed gastric emptying; POD, postoperative day; LOS, length of hospital stay.
Whereas, during surgery, all HR-PD patients (251, 32%) were classified as ISGPS D (soft gland/MPD ≤ 3 mm), non-HR-PD patients (537, 68%) were classified as 327 (42%) ISGPS A (hard gland/MPD ≥ 3 mm), 103 (13%) ISGPS B (hard gland/MPD ≤ 3 mm), and 107 (13%) ISGPS C (soft gland/MPD ≥ 3 mm). During surgery, an ETS was positioned in 214 (85%) HR-PD patients, compared with 95 (18%) non-HR-PD patients (P < 0.001).
The overall POPF rate was 22% (173). The incidence of POPF was significantly higher in the HR-PD group (50% versus 9%; P < 0.001), as were the rates of PPAP (18% versus 2%; P < 0.001), PPH (18% versus 11%; P = 0.005), DGE (32% versus 18%; P < 0.001), intra-abdominal collections (deep surgical site infection) (39% versus 11%; P < 0.001) and sepsis (16% versus 7%; P < 0.001), percutaneous drainage (15% versus 5%; P < 0.001), relaparotomy (15% versus 7%; P < 0.001), unplanned intensive care unit admission (21% versus 9%; P < 0.001), and 90-day mortality (5% versus 2%; P = 0.043). Overall, the HR-PD group showed a longer median LOS (15 (i.q.r. 9–25) versus 9 (i.q.r. 7–13) days; P < 0.001), with significantly higher rates of major morbidity (34% versus 15%; P < 0.001).
Median POD of drain removal was significantly delayed in the HR-PD group (10 (i.q.r. 5–23) versus 4 (i.q.r. 3–6); P < 0.001), and 24% of HR-PD patients were discharged with drains, compared with 6% of non-HR-PD (P < 0.001). Early drain removal (PODs 1–3) was performed in 40% of non-HR-PD patients, compared with 4% of HR-PD patients (P < 0.001).
Postoperative POPF predictors
The median values of each postoperative predictor across different PODs, stratified by POPF risk, are shown in Table S2, and a detailed report of missing values is shown in Table S3. The only variable with missing rates > 10% was POD 1 SL (94 (12%) missing).
Median DFA and SA/SL values were consistently higher in the HR-PD group across all PODs (all P < 0.001). However, in both the HR-PD and non-HR-PD groups, median DFA and SA/SL values decreased over time. Median C-RP was also higher in the HR-PD group (on PODs 3 and 5), though it followed a similar time-dependent trend, peaking on POD 3 in both groups.
The diagnostic performance (ROC curves with AUC) of each clinical predictor—at different PODs and according to POPF risk—is displayed in Fig. 2. In the HR-PD group, DFA had a significantly lower AUC on POD 1 (AUC 0.73 versus 0.85; P = 0.003) and POD 3 (0.74 versus 0.88; P = 0.002) compared with the non-HR-PD group, but a comparable AUC on POD 5 (0.83 versus 0.86; P = 0.535). In the non-HR-PD group, SA/SL consistently had a higher AUC than in the HR-PD group (all P < 0.001). C-RP demonstrated similar ROC patterns in both the non-HR-PD and HR-PD groups, with the highest AUC on POD 5 (0.77 versus 0.84; P = 0.182).
Fig. 2.
ROC curves and AUC of each predictor for POPF, stratified by POD and POPF risk
a Drain fluid amylase, b serum amylase, c serum lipase, d C-reactive protein. ROC, receiver operating characteristic; AUC, area under the curve; POPF, postoperative pancreatic fistula; POD, postoperative day; HR-PD, high-risk pancreatoduodenectomy.
The diagnostic performance of different predictor cut-offs for POPF across different PODs and according to POPF risk is shown in Table S4. In the non-HR-PD group, a DFA = 200 UI/l was identified as the best cut-off on POD 1 (SENS 85%, SPEC 74%, PPV 25%, NPV 98%, Prev+ 31%, Youden 0.593), whereas a DFA = 150 UI/l was identified as the best POD 3 cut-off (SENS 80%, SPEC 88%, PPV 40%, NPV 98%, Prev+ 18%, Youden 0.687). No DFA cut-offs were identified on POD 1 and 3 for the HR-PD group, whereas a DFA = 150 UI/ml was identified as the POD 5 best cut-off (SENS 86%, SPEC 72%, PPV 75%, NPV 84%, Prev+ 57%, Youden 0.584). The presence of SA and/or SL above the upper limit of normal on POD 1 was identified as the best cut-off for SA/SL in the non-HR-PD group (SENS 75%, SPEC 81%, PPV 49%, NPV 93%, Prev+ 30%, Youden 0.560). No cut-offs for SA/SL were identified in the HR-PD group. A C-RP = 150 mg/l was identified as the optimal cut-off on POD 5 in the HR-PD group (SENS 67%, SPEC 80%, PPV 80%, NPV 67%, Prev+ 46%, Youden 0.472).
A secondary analysis was performed to compare DFA median values and AUC between patients with and without an ETS (Table S5 and Fig. S1), with similar results between HR-PD patients and patients with an ETS.
Validation of DFA predictive ability and cut-offs
The predictive ability of DFA for HR versus non-HR-PD and the previously defined best DFA cut-offs were validated in a retrospective cohort of 398 patients, of whom 113 (28%) were classified as HR-PD (Table S6). In the validation cohort, POPF rates were 20% (56) in the non-HR-PD group versus 55% (62) in the HR-PD group. In the HR-PD group, DFA confirmed a significantly lower AUC on POD 1 (AUC 0.62 versus 0.86; P < 0.001) and POD 3 (0.65 versus 0.85; P < 0.001) compared with the non-HR-PD group, but a comparable AUC on POD 5 (0.74 versus 0.79; P = 0.384). In the non-HR-PD group, DFA = 200 UI/l was confirmed the best cut-off on POD 1 with a Youden index of 0.586 (SENS 87%, SPEC 72%, PPV 42%, NPV 96%), whereas DFA = 150 UI/l was the best cut-off on POD 3 with a Youden index of 0.558 (SENS 75%, SPEC 81%, PPV 48%, NPV 93%). In the HR-PD group, no DFA cut-off was identified on POD 1 and 3, whereas DFA = 150 UI/l was the best cut-off on POD 5 with a Youden index of 0.487 (SENS 66%, SPEC 83%, PPV 83%, NPV 66%).
Dynamic drain management algorithm: criteria for early versus late drain removal
Criteria for early (POD 3) and late (POD 5) drain removal were selected, as shown in Table 2. Criteria for early drain removal in non-HR-PD patients were DFA < 200 UI/l on POD 1, DFA <150 UI/l on POD 3, and non-elevated SA and/or SL on POD 1. These combined criteria accurately predicted the absence of POPF in the non-HR-PD group (SENS 72%, SPEC 96%, PPV 99%, NPV 25%, Prev+ 66%, POPF 0.5%, Youden 0.674). Criteria for late drain removal in HR-PD patients were DFA < 150 UI/l and C-RP < 150 mg/ml on POD 5. These combined criteria accurately predicted the absence of POPF in the HR-PD group (SENS 64%, SPEC 95%, PPV 91%, NPV 77%, Prev+ 31%, POPF 9%, Youden 0.593).
Table 2.
Diagnostic performances for POPF absence, for POD 3 and POD 5 drain removal criteria, according to POPF risk
| Criteria | SENS | SPEC | PPV | NPV | Prev+ | POPF | Youden | |
|---|---|---|---|---|---|---|---|---|
| Non-HR-PD | ||||||||
| POD 3 drain removal |
POD 1 DFA < 200 UI/l; POD 1 SA/SL not elevated; POD 3 DFA < 150 U/l | 72% | 96% | 99% | 25% | 66% | 0.5% | 0.674 |
| HR-PD | ||||||||
| POD 5 drain removal |
POD 5 DFA < 150 UI/l; POD 5 C-RP < 150 mg/l | 64% | 95% | 91% | 77% | 31% | 9% | 0.593 |
POPF, postoperative pancreatic fistula; POD; postoperative day; SENS, sensitivity; SPEC, specificity; PPV, positive predictive value; NPV, negative predictive value; Prev+, prevalence of positive cut-offs in the examined population; Youden, Youden’s index; HR-PD, high-risk pancreatoduodenectomy; DFA, drain fluid amylase; SA, serum amylase; SL, serum lipase; C-RP, c-reactive protein.
Based on these findings, a proposal for a tailored drain management protocol was developed, as shown in Fig. 3a. A Sankey diagram of drain removal criteria applied to HR-PD and non-HR-PD, with POPF outcomes, is displayed in Fig. 3b.
Fig. 3.
Proposal of tailored drain management protocol according to POPF risk, and Sankey diagram of drain removal criteria applied to HR-PD and non-HR-PD and POPF outcomes
a Tailored drain management protocol according to POPF risk, b drain removal criteria applied to HR-PD, non-RH-PD, and POPF outcomes. POPF, postoperative pancreatic fistula; HR-PD, high-risk pancreatoduodenectomy; DFA, drain fluid amylase, SA/SL, serum amylase/lipase; POD, postoperative day; PD, pancreatoduodenectomy.
Discussion
The present study proposed a novel, risk-based protocol for the postoperative drain management of patients undergoing PD with a specific focus on those at high risk for POPF. To do so, the diagnostic performance of different postoperative POPF predictors were evaluated dynamically, stratifying by POPF risk. In HR-PD patients, of whom 50% eventually developed POPF, DFA was not a reliable predictor in the early postoperative period (PODs 1–3), and neither SA/SL nor C-RP showed sufficient discriminatory ability. Therefore, whereas an early drain removal policy could be recommended in non-HR-PD patients based on DFA and SA/SL, HR-PD patients would require a delayed approach, with drain removal postponed until POD 5, when DFA and C-RP reach acceptable diagnostic accuracy.
The existence of a subgroup of patients with an intrinsically higher risk for POPF after PD is universally accepted. The intraoperative classification of such risk was first provided by the fistula risk score (FRS)22. Since then, it became clearer that POPF risk is mainly driven by pancreas-specific characteristics, such as gland texture and duct diameter17. Those parameters, on which the four-tier ISGPS classification is based, eventually represent surrogates of the secretive ability of the gland, and the technical difficulty of the anastomosis. Patients in the highest risk class (soft gland, main duct ≤ 3 mm), typically one-third of the total, have POPF rates between 30% and 40%30–32, increased risk of other concomitant pancreas-specific complications, and consequently worsened postoperative outcomes. Similarly, patients considered at higher risk during surgery in the present series experienced a five-fold increase in POPF rates compared with lower-risk patients. As expected, these patients also experienced higher rates of other pancreas-specific complications (including 18% of PPAP), major morbidity, and even mortality. However, the application of mitigation strategies (such as ETS) is a well recognized key element in reducing failure-to-rescue in this subgroup, as confirmed by the relatively low 90-day mortality, despite the dramatically high complications rates. The significant differences between HR-PD and non-HR-PD outcomes confirm the ability of currently available risk stratification tools to identify properly high-risk patients who can benefit from mitigation strategies.
Despite the vital importance of a tailored approach to these patients, a dynamic, multiparametric, risk-based assessment of postoperative drain management after HR-PD was missing. Some previous publications already explored tailored protocols based on intraoperative risk stratification, primarily relying on different DFA cut-offs adjusted for a patient’s risk level, generally adopting higher thresholds for higher-risk glands8,9,11,12,33. However, previous evaluations were rarely extended beyond POD 3, were limited to DFA without considering other predictors, and were not based on the most recent ISGPS risk classification. Two studies (Zureikat et al.9 and AlMasri et al.33) evaluated DFA kinetics between POD 1 and POD 3, reporting good predictive performance of a decreasing trajectory for excluding POPF. However, trajectories were not clearly stratified according to baseline POPF risk, and the overall model performance may have been driven by the higher prevalence of non-HR-PD (in which PODs 1–3 DFA have good predictive values) compared with HR-PD (where PODs 1–3 DFA have reduced predictive ability). Notably, a significant correlation between increasing FRS and higher POD 1/POD 3 DFA values was reported9, supporting the present findings.
Trans-anastomotic stents (internal or external) are considered optimal mitigation strategies for patients in the high-risk categories, to reduce not only POPF incidence but also the general and POPF-related morbidity burden, especially in the high-risk setting23,34–40. However, such mitigation strategies may also influence the accuracy and interpretation of traditional postoperative POPF predictors. Intuitively, the presence of a stent may significantly impact the predictive ability of DFA, potentially reducing its diagnostic performance. When an ETS is placed—diverting nearly all pancreatic juice outside the abdominal cavity—DFA activity may not reliably exclude the occurrence of POPF13. Interestingly, as showed by a dedicated subanalysis in the present study, median DFA activity was nonetheless higher in HR-PD patients and patients with a ETS, with poor prognostic performance, irrespective of subsequent development of POPF. Despite the external diversion of pancreatic juice, what seems to be the main driver of higher DFA values is the presence of a high-risk gland. In HR-PD, the DFA diagnostic performance increases only at POD 5, when DFA activity starts to decline in patients without POPF. Unfortunately, no other examined predictors were able to exclude POPF early enough in HR-PD patients to safely apply early drain removal policies. Even PODs 1–3 SA/SL activity were not useful early POPF predictors in HR-PD patients, as the prevalence of elevated SA/SL was very high in this group independently of POPF development. Finally, for both HR and non-HR patients, C-RP was confirmed to be a valuable indicator of POPF clinical relevance, but mostly from POD 3 onward, as a compendium of other more specific predictors14.
These results reinforce the importance of avoiding ‘blind’ early drain removal policies in high-risk patients, deferring drain removal to POD 5 according to DFA and C-RP levels. Conversely, in non-HR patients, PODs 1–3 DFA activity remains highly predictive and can safely drive early drain removal in most patients. When implemented with POD 1 SA/SL28, the predictive accuracy of the model might be further improved.
This study has some limitations. Most importantly, it represents an observational series in which postoperative decisions for HR-PD patients were not based on a predefined dedicated protocol, reflecting the absence of previous evidence and validated removal criteria in this group. Notably, despite an established institutional practice favouring DFA-based early drain removal, such policy was applied to only nine patients in the HR-PD group based on a very low drain volume output, reflecting a cautious surgical management, whereas it was maintained in almost half of the non-HR-PD patients. However, not performing early removal in the HR-PD group allowed comprehensive collection of dynamic postoperative data across PODs 1 to 5. A randomized clinical trial comparing early versus late removal policies in HR-PD patients would provide higher-level evidence, but the present analysis suggests that early drain removal is not supported by reliable indicators in this high-risk population, and might even be harmful given the high rate of POPF. On the other hand, in non-HR-PD patients with a low POPF risk and negative POD 1 DFA, a ‘super-early’ drain removal on POD 1 might be non-inferior—or even superior—relative to the conventional POD 3 removal. Despite this new policy not being represented in the present series, the present institutions are running a bicentric randomized clinical trial (FIDEL Trial: NCT06468917), which will hopefully bring new insights into early removal policies. When applied to the few selected patients with a negligible or low risk of fistula, even ‘drainless’ policies are described as safe by the literature24. However, drain omission was not part of the usual clinical practice in the present series. Finally, trans-anastomotic stents also include internal stents, widely used during minimally invasive surgery41 but excluded in the present analysis due to the relatively small sample size. It is not clear whether the results of the present analysis should also be applied to internal stents, given the macroscopic differences in their design and functioning. Future developments of the existing policies might come after the integration of novel tools into risk stratification, such as pathological data and artificial intelligence. For example, the acinar content calculated during the histologic assessment of pancreatic section margins was recently used to dichotomize retrospectively the risk of pancreas-specific complications into either high or low risk, surpassing the accuracy of existing intraoperative scores32. The diffusion of real-time, available pathological data during surgery may change traditional risk stratification.
In conclusion, postoperative drain management after PD remains challenging and warrants continuous refinement. Emerging approaches, such as the application of data-driven algorithms16, based on daily assessment of clinical and serological parameters and a proactive approach towards complications, have shown promising results in preventing the most severe sequelae of POPF, including mortality. However, tailored postoperative care should consider both the pancreas-specific risk of complications, and the use of available mitigations strategies like ETSs. In patients undergoing PD with a high intraoperative risk for POPF, early DFA and SA/SL activity are not reliable POPF predictors, and avoiding early drain removal—within the first 72 hours after surgery—appears more appropriate. In non-HR patients, early drain removal remains safe when guided by dynamic DFA and SA/SL assessment.
Supplementary Material
Contributor Information
Giampaolo Perri, Hepato-pancreato-biliary and Liver Transplant Surgery Unit, Department of Surgical, Oncological and Gastroenterological Sciences (DiSCOG), University of Padua, Padua, Italy.
Elisa Bannone, Hepato-pancreato-biliary Surgery Department, Pederzoli Hospital, Peschiera Del Garda, Italy.
Isabella Frigerio, Hepato-pancreato-biliary Surgery Department, Pederzoli Hospital, Peschiera Del Garda, Italy; Collegium Medicum, SAN University, Lodz, Poland.
Riccardo Pellegrini, Hepato-pancreato-biliary and Liver Transplant Surgery Unit, Department of Surgical, Oncological and Gastroenterological Sciences (DiSCOG), University of Padua, Padua, Italy.
Martina Guerra, Hepato-pancreato-biliary Surgery Department, Pederzoli Hospital, Peschiera Del Garda, Italy.
Roberta Vella, Hepato-pancreato-biliary Surgery Department, Pederzoli Hospital, Peschiera Del Garda, Italy.
Alice Cattelani, Hepato-pancreato-biliary Surgery Department, Pederzoli Hospital, Peschiera Del Garda, Italy.
Elvira Adinolfi, Hepato-pancreato-biliary Surgery Department, Pederzoli Hospital, Peschiera Del Garda, Italy.
Umberto Cillo, Hepato-pancreato-biliary and Liver Transplant Surgery Unit, Department of Surgical, Oncological and Gastroenterological Sciences (DiSCOG), University of Padua, Padua, Italy.
Giovanni Butturini, Hepato-pancreato-biliary Surgery Department, Pederzoli Hospital, Peschiera Del Garda, Italy.
Giovanni Marchegiani, Hepato-pancreato-biliary and Liver Transplant Surgery Unit, Department of Surgical, Oncological and Gastroenterological Sciences (DiSCOG), University of Padua, Padua, Italy.
Funding
The BJSO open access fee was funded by the European Union – NextGenerationEU and STARS@UNIPD program (MARC_STARS_MUR24_01, ‘Pancreas PREDICT’).
Acknowledgements
The authors acknowledge the Department of Surgery, Oncology and Gastroenterology (DiSCOG) of the University of Padua.
Author contributions
Giampaolo Perri (Conceptualization, Data curation, Formal analysis, Writing—original draft), Elisa Bannone (Data curation, Writing—review & editing), Isabella Frigerio (Supervision, Writing—review & editing), Riccardo Pellegrini (Data curation), Martina Guerra (Data curation), Roberta Vella (Data curation), Alice Cattelani (Data curation), Elvira Adinolfi (Data curation), Umberto Cillo (Project administration, Supervision, Writing—review & editing), Giovanni Butturini (Supervision, Writing—review & editing), and Giovanni Marchegiani (Conceptualization, Methodology, Supervision, Writing—review & editing)
Disclosure
The authors declare no conflict of interest.
Supplementary material
Supplementary material is available at BJS Open online.
Data availability
The data are not publicly available due to privacy restrictions, but are available from the corresponding author upon reasonable request
References
- 1. Melloul E, Lassen K, Roulin D, Grass F, Perinel J, Adham M et al. Guidelines for perioperative care for pancreatoduodenectomy: enhanced recovery after surgery (ERAS) recommendations 2019. World J Surg 2020;44:2056–2084 [DOI] [PubMed] [Google Scholar]
- 2. Kawai M, Tani M, Terasawa H, Ina S, Hirono S, Nishioka R et al. Early removal of prophylactic drains reduces the risk of intra-abdominal infections in patients with pancreatic head resection: prospective study for 104 consecutive patients. Ann Surg 2006;244:1–7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Bassi C, Molinari E, Malleo G, Crippa S, Butturini G, Salvia R et al. Early versus late drain removal after standard pancreatic resections: results of a prospective randomized trial. Ann Surg 2010;252:207–214 [DOI] [PubMed] [Google Scholar]
- 4. McMillan MT, Malleo G, Bassi C, Butturini G, Salvia R, Roses RE et al. Drain management after pancreatoduodenectomy: reappraisal of a prospective randomized trial using risk stratification. J Am Coll Surg 2015;221:798–809 [DOI] [PubMed] [Google Scholar]
- 5. Bassi C, Marchegiani G, Giuliani T, Di Gioia A, Andrianello S, Zingaretti CC et al. Pancreatoduodenectomy at the Verona Pancreas Institute: the evolution of indications, surgical techniques, and outcomes: a retrospective analysis of 3000 consecutive cases. Ann Surg 2022;276:1029–1038 [DOI] [PubMed] [Google Scholar]
- 6. Molinari E, Bassi C, Salvia R, Butturini G, Crippa S, Talamini G et al. Amylase value in drains after pancreatic resection as predictive factor of postoperative pancreatic fistula: results of a prospective study in 137 patients. Ann Surg 2007;246:281–287 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Fong V, Correa-Gallego Z, Ferrone CR, Veillette GR, Warshaw AL, Lillemoe KD et al. Early drain removal—the middle ground between the drain versus no drain debate in patients undergoing pancreaticoduodenectomy: a prospective validation study. Ann Surg 2015;262:378–383 [DOI] [PubMed] [Google Scholar]
- 8. Newhook TE, Vega EA, Vreeland TJ, Prakash L, Dewhurst WL, Bruno ML et al. Early postoperative drain fluid amylase in risk-stratified patients promotes tailored post-pancreatectomy drain management and potential for accelerated discharge. Surgery 2020;167:442–447 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Zureikat AH, Casciani F, Ahmad S, Bassi C, Vollmer CM. Kinetics of postoperative drain fluid amylase values after pancreatoduodenectomy: new insights to dynamic, data-driven drain management. Surgery 2021;170:639–641 [DOI] [PubMed] [Google Scholar]
- 10. Giglio MC, Spalding DRC, Giakoustidis A, Zarzavadjian Le Bian A, Jiao LR, Habib NA et al. Meta-analysis of drain amylase content on postoperative day 1 as a predictor of pancreatic fistula following pancreatic resection. Br J Surg 2016;103:328–336 [DOI] [PubMed] [Google Scholar]
- 11. Seykora TF, Maggino L, Malleo G, Lee MK, Roses R, Salvia R et al. Evolving the paradigm of early drain removal following pancreatoduodenectomy. J Gastrointest Surg 2019;23:135–144 [DOI] [PubMed] [Google Scholar]
- 12. Trudeau MT, Maggino L, Chen B, McMillan MT, Lee MK., Roses R et al. Extended experience with a dynamic, data-driven selective drain management protocol in pancreaticoduodenectomy: progressive risk stratification for better practice. J Am Coll Surg 2020;230:809–818.e1 [DOI] [PubMed] [Google Scholar]
- 13. Salvia R, Marchegiani G, Andrianello S, Balduzzi A, Masini G, Casetti L et al. Redefining the role of drain amylase value for a risk-based drain management after pancreaticoduodenectomy: early drain removal still is beneficial. J Gastrointest Surg 2021;25:1461–1470 [DOI] [PubMed] [Google Scholar]
- 14. Raza SS, Nutu A, Powell-Brett S, Marchetti A, Perri G, Carvalheiro Boteon A et al. Early postoperative risk stratification in patients with pancreatic fistula after pancreaticoduodenectomy. Surgery 2023;173:492–500 [DOI] [PubMed] [Google Scholar]
- 15. Smits FJ, Molenaar IQ, Besselink MG, Borel Rinkes IHM, van Eijck CHJ, Busch OR et al. Early recognition of clinically relevant postoperative pancreatic fistula: a systematic review. HPB (Oxford) 2020;22:1–11 [DOI] [PubMed] [Google Scholar]
- 16. Smits FJ, Henry AC, Besselink MG, Busch OR, van Eijck CH, Arntz M et al. Algorithm-based care versus usual care for the early recognition and management of complications after pancreatic resection in The Netherlands: an open-label, nationwide, stepped-wedge cluster-randomised trial. Lancet 2022;399:1867–1875 [DOI] [PubMed] [Google Scholar]
- 17. Schuh F, Mihaljevic AL, Probst P, Trudeau MT, Müller PC, Marchegiani G et al. A simple classification of pancreatic duct size and texture predicts postoperative pancreatic fistula: a classification of the International Study Group of Pancreatic Surgery (ISGPS). Ann Surg 2021;12:e597–e608 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Perri G, Romandini E, Marchegiani G, Ghorbani P, Sahakyan M, Holmberg M et al. Postoperative hyperamylasemia (POH) is an early predictor of pancreatic fistula occurrence and severity after distal pancreatectomy: results from a European multicentric study. Ann Surg 2025;282:291–298 [DOI] [PubMed] [Google Scholar]
- 19. Bannone E, Marchegiani G, Vollmer C, Perri G, Procida G, Corvino G et al. Postoperative Serum hyperamylasemia adds sequential value to the fistula risk score in predicting pancreatic fistula after pancreatoduodenectomy. Ann Surg 2023;278:e293–e301 [DOI] [PubMed] [Google Scholar]
- 20. Bannone E, Marchegiani G, Perri G, Procida G, Vacca PG, Cattelani A et al. Postoperative serum hyperamylasemia (POH) predicts additional morbidity after pancreatoduodenectomy: it is not all about pancreatic fistula. Surgery 2022;172:715–722 [DOI] [PubMed] [Google Scholar]
- 21. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP. Strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ 2007;335:806–808 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Callery MP, Pratt WB, Kent TS, Chaikof EL, Vollmer CM. A prospectively validated clinical risk score accurately predicts pancreatic fistula after pancreatoduodenectomy. J Am Coll Surg 2013;216:1–14 [DOI] [PubMed] [Google Scholar]
- 23. Andrianello S, Marchegiani G, Balduzzi A, Bastin A, Masini G, Esposito A et al. Pros and pitfalls of externalized trans-anastomotic stent as a mitigation strategy of POPF: a prospective risk-stratified observational series. HPB (Oxford) 2021;23:1046–1053 [DOI] [PubMed] [Google Scholar]
- 24. McMillan MT, Malleo G, Bassi C, Allegrini V, Casetti L, Drebin JA et al. Multicenter, prospective trial of selective drain management for pancreatoduodenectomy using risk stratification. Ann Surg 2017;265:1209–1218 [DOI] [PubMed] [Google Scholar]
- 25. Bassi C, Marchegiani G, Dervenis C, Sarr M, Abu Hilal M, Adham M et al. The 2016 update of the International Study Group (ISGPS) definition and grading of postoperative pancreatic fistula: 11 years after. Surgery 2017;161:584–591 [DOI] [PubMed] [Google Scholar]
- 26. Wente MN, Veit JA, Bassi C, Dervenis C, Fingerhut A, Gouma DJ et al. Postpancreatectomy hemorrhage (PPH): an International Study Group of Pancreatic Surgery (ISGPS) definition. Surgery 2007;142:20–25 [DOI] [PubMed] [Google Scholar]
- 27. Wente MN, Bassi C, Dervenis C, Fingerhut A, Gouma DJ, Izbicki JR et al. Delayed gastric emptying (DGE) after pancreatic surgery: a suggested definition by the International Study Group of Pancreatic Surgery (ISGPS). Surgery 2007;142:761–768 [DOI] [PubMed] [Google Scholar]
- 28. Marchegiani G, Barreto SG, Bannone E, Sarr M, Vollmer CM, Connor S et al. Postpancreatectomy acute pancreatitis (PPAP): definition and grading from the International Study Group for Pancreatic Surgery (ISGPS). Ann Surg 2022;275:663–672 [DOI] [PubMed] [Google Scholar]
- 29. Dindo D, Demartines N, Clavien P-A. Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg 2004;240:205–213 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Schuh F, Yildirim B, Klotz R, Pianka F, Boskovic A, Werba A et al. Prospective validation of the pancreatic fistula risk classification by the International Study Group for Pancreatic Surgery (PARIS trial). Ann Surg 2024;8:495–504 [DOI] [PubMed] [Google Scholar]
- 31. Suurmeijer JA, Emmen AM, Bonsing BA, Busch OR, Daams F, van Eijck CH et al. Nationwide validation of the ISGPS risk classification for postoperative pancreatic fistula after pancreatoduodenectomy: “less is more”. Surgery 2023;173:1248–1253 [DOI] [PubMed] [Google Scholar]
- 32. Perri G, Marchegiani G, Partelli S, Andreasi V, Luchini C, Bariani E et al. Either high or low risk: the acinar score at the resection margin dichotomizes the risk spectrum of pancreas-specific complications after pancreatoduodenectomy. Ann Surg 2023;278:e1242–e1249 [DOI] [PubMed] [Google Scholar]
- 33. AlMasri S, Kim VM, Hodges JC, Casciani F, Lee KK, Paniccia A et al. Dynamic assessment of drain fluid amylase estimates the risk of CR-POPF following pancreatoduodenectomy. Ann Surg 2024;26:841–847 [DOI] [PubMed] [Google Scholar]
- 34. Judish MM, Vollmer CM. Elucidating the optimal use of mitigation strategies for improving pancreatic fistula rates after pancreatoduodenectomy. Ann Surg; DOI: 10.1097/SLA.0000000000006995 [Epub ahead of print] [DOI] [PubMed] [Google Scholar]
- 35. Ecker BL, McMillan MT, Asbun HJ, Ball CG, Bassi C, Beane JD et al. Characterization and optimal management of high-risk pancreatic anastomoses during pancreatoduodenectomy. Ann Surg 2018;267:608–616 [DOI] [PubMed] [Google Scholar]
- 36. Andrianello S, Marchegiani G, Malleo G, Masini G, Balduzzi A, Paiella S et al. Pancreaticojejunostomy with externalized stent vs pancreaticogastrostomy with externalized stent for patients with high-risk pancreatic anastomosis: a single-center, phase 3, randomized clinical trial. JAMA Surg 2020;155:313–321 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. McMillan MT, Ecker BL, Behrman SW, Callery MP, Christein JD, Drebin JA et al. Externalized stents for pancreatoduodenectomy provide value only in high-risk scenarios. J Gastrointest Surg 2016;20:2052–2062 [DOI] [PubMed] [Google Scholar]
- 38. Huscher C, Perri G, Lazzarin G, Marchegiani G, Malleo G, Salvia R et al. Coronary artery stent for securing high-risk pancreatico-jejunal anastomosis after pancreatoduodenectomy: a pilot series. Ann Surg 2022;275:e665–e668 [DOI] [PubMed] [Google Scholar]
- 39. Pessaux P, Sauvanet A, Mariette C, Paye F, Muscari F, Cunha ASA et al. External pancreatic duct stent decreases pancreatic fistula rate after pancreaticoduodenectomy: prospective multicenter randomized trial. Ann Surg 2011;253:879–885 [DOI] [PubMed] [Google Scholar]
- 40. Motoi F, Egawa S, Rikiyama T, Katayose Y, Unno M. Randomized clinical trial of external stent drainage of the pancreatic duct to reduce postoperative pancreatic fistula after pancreaticojejunostomy. Br J Surg 2012;99:524–531 [DOI] [PubMed] [Google Scholar]
- 41. De Pastena M, Bannone E, Andreotti E, Filippini C, Ramera M, Esposito A et al. Pancreatic anastomosis in robotic-assisted pancreaticoduodenectomy: different surgical techniques. Dig Surg 2023;40:1–8 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data are not publicly available due to privacy restrictions, but are available from the corresponding author upon reasonable request