This monocentric retrospective study analyses all 208 patients who received a deceased-donor kidney transplant via the ESP at our transplant center over 20 years (27 June 1999 to 27 August 2019), with a mean follow-up of 5.7 years. Approved by the local IRB and registered with the German Clinical Trials Register (protocol 20-1199; registration: DRKS00037372), patients consented to record use for research.

All 208 postmortal transplants were ABO- and Human Leukocyte Antigen (HLA)-compatible, defined as negative complement-dependent cytotoxicity (CDC) crossmatch, with CDC remaining the mandatory pre-transplant compatibility test throughout the study period per local protocol. ESP allocation prioritized ABO blood-group matching and negative crossmatch over HLA matching according to Eurotransplant guidelines. As per center policy, kidneys for immunized patients were only accepted when preformed DSA could be excluded preoperatively.

Graft preparation, surgical procedure, and follow-up were as described [22]. All kidneys were preserved by static cold storage (SCS), using Histidine-Tryptophane-Ketoglutarate (HTK) solution (Custodiol®) or University of Wisconsin (UW) solution. Hypothermic machine perfusion (HMP) was not used during the study period (1999-2019), as it was not adopted in the Eurotransplant region until post-2015. Recipients stayed in intermediate care for 5–7 days post-transplant. Figure 1 shows the study profile. In 74 cases, both kidneys from a single donor were allocated to our center and, due to availability of operating theatres and surgeons, transplanted consecutively, with recipients labeled as “rank 1” and “rank 2.” In 60 cases, only one kidney was received, which is included in graft and survival analysis but not in mixed logistic regression.

The patient cohort was stratified by CIT. While a CIT >18 h is reported to impair graft survival, this was rare in our case [23, 24]. With a median CIT of 9.35 h in our cohort, we performed an ROC analysis to determine the optimal CIT cut-offs. We then categorized CIT 1 (0–8 h), CIT 2 (8–12 h), and CIT 3 (equal to or greater than 12 h) according to ROC analysis and clinical practicability.

Clinical and follow-up data were acquired from our nephrological transplant outpatient clinic and the local nephrologists. Delayed graft function was defined as the need for ≥1 dialysis within 7 days postoperatively. PNF was defined as permanent dysfunction requiring dialysis and was excluded from DGF analyses to avoid misclassification. Graft loss was defined as permanently resuming dialysis due to irreversible graft failure. Acute reversible graft failure was excluded.

Results with p-values <0.05 were considered statistically significant and assessed on a two-sided basis. The dataset was tested for normality; results are shown in Supplementary Table S1; Supplementary Figure S1 (Supplementary Material). Categorical data were assessed using Fisher’s exact test; continuous data were assessed using paired t-test, Wilcoxon matched-pairs signed-rank test, Mann-Whitney-U test and Kruskal-Wallis, presented either as mean ± SD or as median and 95% CI.

A post-hoc power analysis was performed to assess the statistical power of the study given the available sample size. For survival outcomes (patient survival and death-censored graft survival), power was calculated using the Schoenfeld formula based on the observed number of events and hazard ratios derived from Cox proportional hazards models. For the comparison of delayed graft function rates between CIT groups, Cohen’s h was used as the effect size measure. All calculations assumed a two-sided significance level of α = 0.05. Statistical power was calculated using the pwr package (version 1.3) in R (version 2024.09.1 + 394).

ROC curve analysis was performed to determine the optimal CIT cut-off for predicting DGF, graft loss, and patient survival. Optimal cut-off was determined by Youden index. Recipients with primary non-function (PNF, n = 15) were excluded (193 recipients remaining).

Delayed graft function (DGF) was the primary endpoint, evaluated through a mixed effects logistic regression with random intercept for donor identity (74 donors) to account for the non-independence of outcomes in kidney pairs from the same donor. The model specification (generalized linear mixed model (GLMM) with binominal distribution and logit link function) was logit P DGF ij = β 0 + β 1 ( CIT _ group 2 ) ij + β 2 ( CIT _ group 3 ) ij + β 3 ( DM ) j + β 4 ( CHD ) j + β 5 ( HT ) j + β 6 ( Age ) j + β 7 ( BMI ) j + u j where “i” denotes the recipient (1 or 2) and “j” denotes the donor (1–74).

“CIT_group2” and “CIT_group3” are indicator variables for CIT 8–12 h and >12 h, respectively (reference: CIT 0–8 h). We included the following donor-specific covariates: diabetes mellitus (DM), coronary heart disease (CHD), hypertension (HT), age (Age), and body mass index (BMI). The model was fitted using maximum likelihood estimation with the Laplace approximation in R (lme4 package, version 1.1–35).

Secondary endpoints included graft and patient survival, analyzed via the Kaplan-Meier method and log-rank test, considering the paired cohort (148 patients) for graft survival (functional integrity of individual graft) and all 208 patients for patient survival (limited sample size, donor-specific characteristics not as important), focusing on a five-year follow-up. PNF was included as immediate failure (time = 0). Cox proportional hazards regression assessed CIT impact on secondary endpoints. The regression model included confounder adjustments based on clinical relevance and biological plausibility. Goodness-of-fit was tested with Log-likelihood ratio, Wald, and Score tests. Multicollinearity was checked using variance inflation factors (Supplementary Table S5).

To assess whether the long inclusion period (1999–2019) influenced study findings, a sensitivity analysis was performed. Patients were divided into three transplant eras: Era 1 (1999–2006, ESP introduction), Era 2 (2007–2013, tacrolimus, refined surgical techniques), and Era 3 (2014–2019, modern protocols). Eras were chosen according to major milestones and improvements in transplant protocols. First, the transplant era was included as a covariate in all multivariable Cox regression models. Second, a formal interaction test between the CIT group and transplant era was performed using a likelihood ratio test comparing models with and without the interaction term.

Missing data were minimal (<4% for any variable). Specifically, primary immunosuppression (n = 2, 0.96%), creatinine at discharge (n = 7, 3.37%), warm ischemia time (n = 1, 0.48%), time on dialysis before KTx (n = 1, 0.48%), and time on wait list (n = 1, 0.48%) were missing due to incomplete historical records (see STROBE Statement in Supplementary Material). Complete case analysis was used for multivariable models. All primary outcomes (DGF, graft loss, and death) had complete data with no loss to follow-up for the five-year observation period, as all patients were followed at our center and at the local nephrologists.

All analyses and visualizations were conducted using GraphPad Prism 11 and R version 2024.09.1 + 394.

Among the 208 patients who received an ESP allograft at our transplant center, in 74 cases both kidneys from a single donor were transplanted consecutively in two patients, designated as rank 1 and rank 2 recipients. Comparing these two groups, no differences were observed with respect to sex, body mass index (BMI), ASA classification, pre-existing medical conditions, time on dialysis prior to KTx, time on the wait list, or the number of HLA mismatches (Table 1A; for a more detailed comparison of groups, refer to Supplementary Table S2). As a result of consecutively performed transplantations, the cold ischemic period was significantly prolonged in rank 2 recipients (10.7 (9.8, 11.5) vs. 6.3 (5.6, 7.1) h, p < 0.0001, Table 1A). Similar characteristics were noted when comparing patients according to CIT groups (CIT groups): no significant differences in demographics, comorbidities, or HLA mismatches, except CIT, were found (Table 1B, for post-hoc pairwise comparisons for baseline characteristics of CIT groups see Table 1C).

Among 193 transplants with initial function, DGF occurred in 45 cases (23.3%). Cases with primary non-function (PNF, n = 15) were excluded. Among 74 kidney pairs, DGF was 20% for rank 1% and 30% for rank 2 recipients and 18.9% in single kidney allocations (kidneys with PNF excluded: rank 1: 4 kidneys; rank 2: 4 kidneys; single kidneys: 7 kidneys). No significant difference was observed concerning extended dialysis (>7 days post-transplant) (Table 2).

CIT was the only significant difference between rank 1 and rank 2 (Table 1A). Subsequently, the entire cohort was stratified into CIT groups (Table 2; Figures 2A–C), with ROC curve analysis aiding in identifying the optimal CIT cut-off for predicting DGF (Table 2A, for ROC curves and their comparison see Supplementary Material; Supplementary Figure S2). CIT demonstrated poor discriminative ability for DGF with an AUC of 0.592 (95% CI 0.497–0.687, p = 0.031). The optimal cut-off was determined using the Youden index, yielding a CIT threshold of 8.53 h, with a sensitivity of 71.1% and specificity of 48%. To take the most important secondary endpoints into account as well, we added ROC curve analyses for graft and patient survival. We found that CIT demonstrated poor discriminative ability for graft loss with an AUC of 0.597 (95% CI 0.443–0.750, p = 0.115). The optimal cut-off was determined using the Youden index, yielding a CIT threshold of 11.3 h, with a sensitivity of 50% and specificity of 73.7%. Concerning patient survival, CIT demonstrated acceptable discriminative ability with an AUC of 0.63 (95% CI 0.504–0.756, p = 0.014). The optimal cut-off was determined using the Youden index, yielding a CIT threshold of 11.6 h, with a sensitivity of 57.1% and specificity of 79.4%. With the median CIT being 9.35 h in our cohort, we set the groups as CIT 1 (0–8 h), CIT 2 (8–12 h), and CIT 3 (≥12 h), taking both clinical practicability and ROC analysis results into account.

Although KTx were performed consecutively, nearly two-thirds (63.5%) of rank 2 recipients received the allograft within 12 h (Table 2B). DGF proportions were similar in CIT groups 1 and 2 regardless of rank (rank 1 (CIT <12 h): 18.8%, rank 2 (CIT <12 h): 21.7%; Table 2B). When stratifying the cohort according to CIT, by every 4-h increment, DGF rates rose steadily, from 16.7% (CIT <8 h), to 21.6% (CIT 8–12 h), up to 36.2% (CIT >12 h; Figure 2D).

A mixed logistic regression analysis examined the relationship between CIT and DGF. To mitigate potential confounding factors pertaining to allograft characteristics, analysis was restricted to the 74 kidney pairs and further adjustment for donors’ pre-existing medical conditions was made (Table 3A). Kidneys with PNF were excluded from analysis. CIT group 3 had a 6.3-times higher risk of DGF than CIT group 1 (adjusted OR 6.30; 95% CI 1.52–26.06; p = 0.011, Table 3B). Additionally, the risk of DGF increased with donor age (OR: 1.23; CI: 1.05–1.43; p = 0.01, Table 3B). To underscore the need for minimizing CIT, the analysis was replicated with CIT measured in minutes: DGF still correlated with increasing CIT (OR: 1.0002; CI: 1.0–1.01; p = 0.012; Supplementary Table S3A), although not as pronounced as with categorical groups. Rank of transplantation did not affect DGF (Supplementary Table S3B).

After identifying CIT as a risk factor for DGF, rather than the chronological sequence of graft transplantation, we analyzed its impact on graft function and patient survival. Creatinine levels were similar at discharge and last follow-up (Supplementary Table S4). Graft loss incidence was similar across CIT groups but occurred earlier in CIT group 3 (57 (1, 639) days vs. 674 (129, 1664) days, p = 0.042, Supplementary Table S4).

In matched patients without DGF, graft survival rates were comparable among CIT groups (Figure 3A). For those with DGF, CIT group 3 exhibited a numerically (though not statistically significant, p = 0.345) lower five-year graft survival rate (67.1% vs. CIT 1: 78.4%/CIT 2: 83.3%), which may reflect insufficient power (Figure 3B).

Cox multivariable regression analysis identified DGF as the only independent risk factor for graft loss (adjusted HR: 3.71, CI: 1.14–12.53, p = 0.029, Table 4), meaning CIT, transplantation rank (Supplementary Figure S3) and recipients’ comorbidities were not risk factors.

Patient survival was assessed in the whole cohort. Kaplan-Meier analysis indicated higher mortality in CIT group 3 (log-rank P = 0.0003, Figure 4A), whereas survival rates between rank 1 and rank 2 were comparable (Supplementary Figure S3). DGF worsened outcomes in CIT group 3 (log-rank P < 0.0001, Figure 4B), with a 23.5% five-year survival rate that was markedly lower than the other groups (CIT 1: 69.5%, with DGF: 67.5%; CIT 2: 69.4%, with DGF: 87.1%; CIT 3: 56.7%).

To understand why survival within CIT group 3, even without DGF, was significantly worse, we compared postoperative complications. Significantly more cardiac complications were seen in CIT group 3 (9.1% vs. 22.2%, p = 0.017, Table 5). Nonetheless, using a Cox proportional hazards regression model, the sole independent risk factor for impaired survival remained a CIT >12 h, increasing mortality by 2.98 times (adjusted HR 2.98, 95% CI: 1.34–6.97, p = 0.009, Table 6).

To address the potential confounding effect of the long inclusion period, a sensitivity analysis was performed. After adjustment for transplant era (Era 1: 1999-2006, Era 2: 2007-2013, Era 3: 2014-2019), the association between CIT group 3 and patient survival remained significant (HR = 4.81, 95% CI: 1.21–19.99, p = 0.025, Supplementary Table S6). No significant interaction between CIT group and transplant era was observed (Likelihood ratio test (Era × CIT interaction): patient survival: p = 0.583; death-censored graft survival: p = 0.254, Supplementary Table S6), indicating that the prognostic effect of prolonged CIT was consistent across all three transplant eras.

HLA-DR matching for ESP patients has been advocated for, with de Fijter et al. positing it enhances five-year graft and patient survival [25]. Accordingly, we assessed outcomes of our Freiburg ESP cohort considering HLA-mismatches. Among 28 patients with zero HLA-DR mismatches, 21% experienced DGF, matching the whole cohort’s proportion (21.6%). Kaplan-Meier analysis showed no survival differences between patients with 0–3 and 4-6 mismatches (Figures 5A,C). However, differentiating mismatch groups by CIT, CIT group 3 patients, irrespective of mismatches, had lower survival compared to CIT groups 1 and 2 (log-rank P = 0.0002, Figure 5D). This trend was also present for death-censored graft survival, though the analysis did not achieve statistical significance (log-rank P = 0.617, Figure 5B). A comparable analysis of the 74 kidney pairs yielded consistent findings (Supplementary Figure S4).

To supplement this retrospective analysis, a Cox proportional hazards regression model identified risk factors for mortality, adjusting for confounders like ASA category and DGF. Regardless of mismatch count, CIT >12 h tripled mortality risk (Table 7A: CIT 3: adjusted HR 2.89, 95% CI: 1.33–6.65, p = 0.009; Table 7B: CIT 3: adjusted HR 3.07, 95% CI: 1.43–6.97, p = 0.005).

Post-hoc power analysis revealed that the study was underpowered for several comparisons, with power ranging from 11.9% to 68.3% across outcomes, primarily reflecting the limited number of events inherent to this low-volume transplant program (Supplementary Table S7).