This multicenter retrospective observational study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [9]. Consecutive patients undergoing PT, including simultaneous pancreas-kidney (SPK), pancreas-after-kidney (PAK), and pancreas transplant alone (PTA), at three European pancreas transplant centers between 2010 and 2024 were considered for inclusion. Patients with missing data required to apply the PTAP definition, including postoperative serum amylase measurements, were excluded. Patients undergoing graft explantation on postoperative day (POD) 0 or 1 were also excluded because sustained postoperative hyperamylasemia could not be assessed and the adapted ISGPS-based definition could therefore not be applied. The patient selection process is summarized in Figure 1.

The study was conducted in accordance with the Declaration of Helsinki. Given the retrospective observational design and the use of anonymized routinely collected clinical data, approval by the local ethics committees was waived at all participating centers.

PTAP was defined using an adapted ISGPS framework for PPAP [7, 8]. Specifically, PTAP was defined by the presence of both of the following objective criteria:

Sustained postoperative hyperamylasemia (POH): persistent elevation of serum pancreatic amylase above the institutional upper limit of normal (Supplementary Table 1) for at least 48 h postoperatively, defined by elevated values on both POD1 and POD2;

Postoperative CE-CT scans were interpreted locally by radiologists at each participating center as part of routine clinical care. No central radiological review was performed. Because imaging was obtained and interpreted in the clinical setting, radiologists were not systematically blinded to serum amylase values or clinical information. For study purposes, PTAP status was assigned retrospectively according to prespecified biochemical and radiological criteria.

Serum pancreatic amylase was routinely measured on POD1 and POD2 in all patients. CE-CT was not performed systematically in all patients with sustained POH but was obtained according to clinical indication or center-specific protocols. CE-CT was the standard postoperative cross-sectional imaging modality across the three participating centers and was performed according to center-specific imaging strategies. In one center, CE-CT was performed according to clinical indication, whereas the other two centers adopted protocol-driven CE-CT for all PT recipients, implemented from 2013 onward in one center and routinely performed throughout the study period in the other. Patients with sustained POH who did not undergo postoperative cross-sectional imaging were classified as having POH without radiological confirmation, because the radiological criterion for PTAP could not be assessed. These patients were therefore considered not to fulfill the PTAP definition, but were not regarded as radiologically negative. This pragmatic imaging strategy, which reflects real-world clinical practice, may have led to underdiagnosis of mild or subclinical forms of PTAP but was expected to enhance specificity for clinically meaningful PTAP, which was the primary focus of the present study.

The clinical criterion included in the original ISGPS definition, namely clinically relevant changes in postoperative management, was intentionally not applied. This choice was made to avoid incorporation bias, because postoperative management decisions are intrinsically linked to short-term clinical outcomes, which constituted primary study endpoints. By relying exclusively on objective biochemical and radiological criteria, the analysis aimed to preserve independence between exposure (PTAP) and outcome assessment.

The clinical impact of PTAP was evaluated using two primary outcomes: (i) severe postoperative complications and (ii) length of index hospital stay (LOS). Postoperative complications occurring within 30 days after transplantation or during the index hospitalization were graded according to the Clavien-Dindo classification [10] and were considered severe if grade IIIA or higher. This threshold was selected to capture complications of clear clinical relevance requiring invasive management. Both systemic and pancreas-related complications were included, whereas complications exclusively related to the kidney graft were excluded. LOS was calculated from the day of transplantation (POD0) to hospital discharge and dichotomized using the third quartile of the cohort distribution; prolonged LOS was defined as more than 23 days.

Secondary outcomes included death-censored pancreas graft survival (GS) and patient survival. Death-censored pancreas GS was defined as the time from transplantation to pancreas graft loss from any cause, excluding death with a functioning graft. Pancreas graft loss was defined as graft explantation or persistent return to insulin dependence during follow-up. Temporary insulin therapy during the early postoperative period was not considered graft loss if graft endocrine function subsequently recovered. Patient survival was defined as the time from transplantation to death from any cause. Patients without events were censored at last available follow-up.

Clinical data were retrospectively collected from prospectively maintained institutional databases using standardized data collection procedures. Collected variables included recipient- and donor-related intraoperative and postoperative characteristics.

Recipient-related variables included sex, age at transplantation, body mass index (BMI), duration and type of diabetes, comorbidities, preoperative dialysis status and modality, and history of previous transplants. Donor-related variables included donor type (deceased donor after brain death [DBD] or after circulatory death [DCD]), sex, age and BMI.

Intraoperative variables collected included transplant type (SPK, PAK, PTA), cold ischemia time (CIT), anastomotic time, duration of surgery, and intraoperative blood transfusions. CIT was defined as the interval between initiation of cold perfusion in the donor and removal of the graft from ice for implantation. Anastomotic time was defined as the interval between the end of cold preservation and restoration of graft blood flow in the recipient.

Postoperative variables included early and late complications, relaparotomy, hemorrhage, thrombosis (occurring from POD2 onwards), graft explantation, duodenal leakage, and graft rejection before discharge. Clinically relevant graft exocrine leakage was captured as duodenal leakage and recorded as a distinct graft-related complication. Duodenal leakage was defined on the basis of clinical findings, drain characteristics including suspicious drain output and increased drain amylase levels, operative findings, and radiological findings when CT imaging was available. Complications occurring within 30 days after surgery were classified as early, whereas those occurring between 30 days and 3 months were classified as late. Long-term outcomes included rejection episodes, graft loss (with or without graft pancreatectomy), and patient survival.

Induction immunosuppression consisted of rabbit anti-thymocyte globulin (rATG) and corticosteroids in all patients. Maintenance immunosuppression consisted of a triple-drug regimen including tacrolimus, mycophenolate mofetil (MMF), and corticosteroids.

Perioperative anticoagulation protocols differed across centers and over time, and included prophylactic-dose low molecular weight heparin (LMWH), therapeutic-dose LMWH, or continuous intravenous unfractionated heparin (UFH), with escalation to therapeutic-dose anticoagulation in selected patients with thrombotic defects detected on postoperative CE-CT, as summarized in Supplementary Table 2. Antimicrobial prophylaxis did not differ across centers and followed standard institutional transplant protocols.

Categorical variables are reported as absolute numbers and percentages and were compared using the χ² test or Fisher’s exact test, as appropriate. Continuous variables are reported as median with interquartile range (IQR); normality was assessed using the Kolmogorov-Smirnov test. Group comparisons were performed using Student’s t test or Mann-Whitney U test, as appropriate.

Univariable and multivariable logistic regression analyses were performed to identify determinants of severe postoperative complications, prolonged LOS, and PTAP. LOS was primarily analyzed as a binary outcome using the third quartile as the threshold for prolonged hospitalization. As a sensitivity analysis, LOS was also analyzed as a continuous outcome using multivariable linear regression after logarithmic transformation to account for its right-skewed distribution. Logistic regression was selected for early postoperative outcomes occurring within a fixed 30-day window. Results are reported as odds ratios (OR) with 95% confidence intervals (CIs). Multivariable models were constructed by entering variables significantly associated with the outcome on univariable analysis, followed by backward elimination for covariate selection. Center-specific effects were included in multivariable models to account for between-center heterogeneity. The number of covariates included in each model was limited according to the number of outcome events, aiming for up to 10 outcome events per covariate, to reduce the risk of overfitting. Missing data were not imputed; multivariable analyses were performed as complete-case analyses for the variables included in each model. Variables with low event counts were interpreted cautiously because of the potential for unstable estimates. To address the potential impact of non-systematic imaging, a sensitivity analysis restricted to patients who underwent postoperative CE-CT was performed for severe postoperative complications and prolonged LOS.

Death-censored pancreas GS and patient survival were estimated using the Kaplan-Meier method and compared using the log-rank test. Predictors of death-censored graft loss were assessed using a Fine-Gray competing-risk regression [11], with death with a functioning graft treated as a competing event. Results are reported as subdistribution hazard ratios (sHRs) with 95% CIs.

Receiver operating characteristic (ROC) curve analysis was used to assess the discriminatory performance of serum amylase levels for predicting radiological GP. Optimal cut-off values were determined using the Youden index.

All tests were 2-sided, and P < 0.05 was considered statistically significant. Analyses were performed using SPSS software version 29.0.2 for Mac (IBM Corp., Armonk, NY, USA), GraphPad Prism version 10.2.2 for Mac (GraphPad Software, Boston, MA, USA) and R version 4.4.2 (R Foundation for Statistical Computing, Vienna, Austria) through RStudio version 2024.12.1 + 563 (RStudio, Boston, MA, USA), using the “survival” and “cmprsk” packages.

A total of 453 consecutive patients undergoing PT were screened across the three participating centers. Twenty-one patients were excluded because of missing critical data required to apply the PTAP definition, namely unavailable serum pancreatic amylase measurements on POD1/POD2. The study cohort therefore included 432 PT recipients. Among these, 16 patients underwent graft pancreatectomy on POD0-1 because of early graft thrombosis and were excluded from the PTAP analysis because sustained POH over 48 h could not be assessed. The final analyzable cohort included 416 patients. All analyses assessing PTAP and its association with postoperative and long-term outcomes were performed in this cohort. Sustained postoperative hyperamylasemia (POH) occurred in 196 patients (47%), and PTAP was diagnosed in 86 (21%). Overall, postoperative CE-CT was performed in 371/416 patients (89%) in the analyzable cohort. Center-level CE-CT rates were 97/128 (76%) in the center performing CE-CT according to clinical indication, 138/152 (91%) in the center where protocol-driven CE-CT was implemented from 2013 onward, and 136/136 (100%) in the center performing routine protocol-driven CE-CT throughout the study period. Center-level PTAP rates were 38/128 (30%), 32/152 (21%), and 16/136 (12%), respectively (P = 0.002). PTAP rates were stable across transplant eras: 15/84 patients (18%) in 2010–2014, 29/141 patients (21%) in 2015–2019, and 42/191 patients (22%) in 2020–2024 (P = 0.738).

Among the remaining patients with sustained POH (n = 110), 85 had no radiological evidence of GP, whereas 25 did not undergo CE-CT. The latter patients were classified as having POH without radiological confirmation. Among these patients, 9 experienced a major postoperative complication, including 7 hemorrhagic and 2 thrombotic events. Among patients without sustained POH (n = 220), CE-CT was performed postoperatively in 200 (91%), and radiological findings consistent with GP were identified in 58 (29%). The study flow diagram is shown in Figure 1.

Recipient and donor baseline characteristics were largely comparable between patients with and without PTAP (Table 1). In contrast, operative and preservation-related variables differed between groups. Median CIT was significantly longer in patients with PTAP than in those without PTAP [538 min (IQR, 458–630) vs. 489 min (IQR, 410–560); P < 0.001]. Median anastomotic time was also longer in the PTAP group [35 min (IQR, 30–40) vs. 30 min (IQR, 27–36), P = 0.036], as was overall duration of surgery (P = 0.022).

With respect to early postoperative outcomes, graft explantation between POD2 and POD30 was more frequent in patients with PTAP than in those without PTAP [13/86 (15%) vs. 13/330 (4%), P < 0.001]. Duodenal leakage was also more common in the PTAP group [7/86 (8%) vs. 6/330 (2%), P = 0.007], whereas rates of early hemorrhage and thrombosis did not differ significantly between groups. Among the 13 patients with duodenal leakage, 11 had sustained POH and 7 fulfilled the adapted ISGPS-based PTAP definition.

Severe postoperative complications (Clavien-Dindo grade ≥ IIIA) occurred in 112 patients (27%) and were significantly more frequent in patients with PTAP than in those without PTAP [33/86 (38%) vs. 79/330 (24%); P = 0.008]. Univariable and multivariable logistic regression analyses assessing determinants of severe postoperative complications are reported in Table 2. After adjustment for center-specific effects and clinically relevant covariates, PTAP was the only independent predictor of severe postoperative complications (OR, 2.482; 95% CI, 1.432–4.305; P = 0.001).

Median LOS was significantly longer in patients with PTAP than in the remainder of the cohort (20 vs. 15 days, P < 0.001). On multivariable analysis (Table 3), PTAP (OR, 3.817; 95% CI, 2.146–6.792, P < 0.001) and graft rejection before discharge (OR, 16.662; 95% CI, 3.063–90.621; P = 0.001) were independently associated with prolonged LOS (>23 days–n = 98). PTAP remained independently associated with longer hospital stay when LOS was analyzed as a continuous log-transformed outcome (adjusted exp[β], 1.301; 95% CI, 1.166–1.451; P < 0.001 – Supplementary Table 3).

In a sensitivity analysis restricted to the 371 patients who underwent postoperative CE-CT, severe postoperative complications occurred in 33/86 patients with PTAP (38%) and in 67/285 patients without PTAP (23%; P = 0.008). Median hospital stay was longer in patients with PTAP than in those without PTAP (20 days [IQR, 14–33] vs. 15 days [IQR, 13–22]; P < 0.001). On multivariable analysis, PTAP remained independently associated with severe postoperative complications (OR, 2.362; 95% CI, 1.369–4.077; P = 0.002) and prolonged hospital stay (OR, 3.767; 95% CI, 2.140–6.632; P < 0.001).

To further characterize the relative contribution of sustained POH and radiological GP to postoperative outcomes, a four-group comparison was performed among patients who underwent postoperative CE-CT. Overall, 142 (38%) patients had neither sustained POH nor radiological GP, 85 (23%) had sustained POH without radiological GP, 58 (16%) had radiological GP without sustained POH, and 86 (23%) fulfilled the PTAP definition. CE-CT timing was comparable across groups (P = 0.442). Later serum pancreatic amylase trajectories (POD3, POD5, and POD10), normalized to center-specific ULN, differed significantly across groups (P < 0.001), with the highest median values observed in patients with sustained POH, either with or without radiological GP. Patients with radiological GP only did not show a delayed median enzyme rise above the ULN.

Patients with radiological GP without sustained POH showed intermediate outcomes, with rates of severe postoperative complications (26%) and hospital stay (median 18 days; IQR, 13–28) closer to those observed in the no POH/no radiological GP and POH-only groups than to those observed in patients fulfilling the combined PTAP definition (severe complications, 38%; median LOS, 20 days). The full four-group comparison is reported in Supplementary Table 4.

After a median follow-up of 51 months (IQR, 24–87), graft loss occurred in 85 patients (20%), and 30 patients (7%) died. Death-censored pancreas GS was significantly worse in patients with PTAP than in those without PTAP (P < 0.001). Three-year death-censored pancreas graft survival was 86% in patients without PTAP and 65% in patients with PTAP (Figure 2A). Overall pancreas graft rejection during follow-up occurred in 58/416 patients (14%).

Results of univariable and multivariable competing-risk regression analyses for predictors of death-censored pancreas graft loss are reported in Table 4. On multivariable analysis, independent predictors of pancreas graft loss included donor type (DCD; sHR, 6.586; 95% CI, 1.409–30.774; P = 0.017), CIT (sHR, 1.004 per minute; 95% CI, 1.001–1.006; P = 0.003), transplant type (PAK/PTA; sHR, 2.722; 95% CI, 1.314–5.641; P = 0.007), PTAP (sHR, 3.412; 95% CI, 2.005–5.808; P < 0.001), early thrombosis (sHR, 2.821; 95% CI, 1.626–4.895; P < 0.001), and graft rejection (sHR, 3.975; 95% CI, 2.338–6.616; P < 0.001).

To further characterize death-censored pancreas graft loss, timing, associated graft-related events, and main causes of graft loss were assessed according to PTAP status (Supplementary Table 5). Among patients experiencing graft loss (n = 85), median time to graft loss was shorter in patients with PTAP than in those without PTAP (1.0 months [IQR, 0–7.5] vs. 9.3 months [IQR, 1–41]; P = 0.015). Early graft explantation between POD2 and POD30 was more frequent among patients with PTAP (13/31, 42%) than among those without PTAP (13/54, 24%), although this difference did not reach statistical significance (P = 0.085). Pancreas graft rejection before discharge occurred in 10/416 patients (2%), while overall pancreas graft rejection during follow-up occurred in 58/416 patients (14%). The distribution of graft-loss causes differed according to PTAP status (P = 0.036), with early postoperative complication-related graft loss accounting for a larger proportion of events among patients with PTAP, whereas rejection-related graft loss was more common among patients without PTAP.

To assess whether the association between PTAP and graft loss was driven by very early technical failures, a sensitivity analysis was performed after excluding patients who underwent early graft explantation between POD2 and POD30. In Kaplan-Meier analysis, death-censored pancreas graft survival remained significantly worse in patients with PTAP (P = 0.003). In a center-adjusted competing-risk sensitivity analysis, PTAP remained associated with a higher cumulative incidence of pancreas graft loss (sHR, 1.964; 95% CI, 1.087–3.550; P = 0.025).

Patient survival was also significantly worse among patients with PTAP (P < 0.001, Figure 2B).

Overall, 371 patients (89%) underwent at least one CE-CT scan during the postoperative period, at a median of POD8 (IQR, 4–10). Among these patients, 171 (45%) had sustained POH. Radiological signs consistent with GP were identified in 144 patients (35%), of whom 86 met the biochemical criterion for sustained POH and were therefore classified as having PTAP.

Serum amylase levels on POD1 and POD2, expressed as multiples of the upper limit of normal (ULN), were significantly higher in patients with radiological GP than in the remainder of the cohort (P < 0.001, Figure 3). The discriminatory performance of POD1 and POD2 amylase levels for predicting radiological GP was modest, with an AUC of 0.630 (95% CI, 0.570–0.690; P < 0.001) and 0.648 (95% CI, 0.589–0.708; P < 0.001), respectively (Supplementary Figure 1). Optimal cut-off values were 3.25×ULN on POD1 and 2.25×ULN on POD2. Radiological GP was present in 74% of patients (37/50) who exceeded both these thresholds, compared with 33% of those below at least one threshold (P < 0.001).

Determinants of PTAP were further explored using logistic regression analyses (Supplementary Table 6). On multivariable analysis, CIT was the only independent determinant of PTAP (OR, 1.004 per minute; 95% CI 1.002–1.006; P < 0.001).