**3. Results**

#### *3.1. Overall Results Regarding ICU Admission*

Overall, 17 (16.2%) patients were admitted to the ICU for a median length of 2 days (range 1–27). The main reason for ICU admission was significant hypotension requiring catecholamines in the absence of acute bleeding in five (29.4%) patients. Three (17.6%) patients were admitted for respiratory insu fficiency, three (17.6%) for sepsis with multiple organ failure, and two (11.7%) for cardiac infarction. One (5.9%) patient had hyperkalemia, another a compartment syndrome due to occlusion of iliac arteries. One (5.9%) patient had a significant bleeding requiring surgical re-exploration, another had his graft surgically removed because of arterial stenosis and consecutive graft necrosis, and required intensive care thereafter. The median time between KT and ICU admission was 0 days, as 10 (58.8%) patients were admitted to the ICU immediately after KT. In total, 4 (23.5%) patients were admitted on postoperative day (POD) 2 to 4, and 2 (11.7%) patients on POD 8 and 9. One (5.9%) patient was admitted to the ICU on POD 36; he su ffered from a late-onset sepsis. The admission rate to the ICU did not di ffer between the two transplant centers.

Patients admitted to the ICU were insignificantly older than patients without an ICU stay (71 vs. 69 years, n.s.) (see Table 1). They had a lower BMI (24.2 vs. 26.7, *p* < 0.05) and CAD twice as often (64.7% vs. 35.2%, *p* < 0.05). Regarding the underlying renal disease, hypertensive nephropathy was more common in patients admitted to the ICU (35.3% vs. 10.2%, *p* < 0.05). In both groups, the median number of HLA-mismatches was four (range 1–6, n.s.). There was a tendency towards longer CIT for patients admitted to the ICU (667.8 vs. 552.3 min), but it was not significant.


**Table 1.** Comparison of patient characteristics with or without ICU stay after kidney transplantation in the ESP program.

> \* see Appendix A Table A3 for further information.

Regarding KTs, patients admitted to the ICU had slightly longer operating times (212 vs. 180 min, *p* = 0.053), and neither WIT nor intraoperative complication rates di ffered (see Table 2). During the postoperative course, patients with an ICU stay su ffered from more frequent and higher complications based on Clavien Dindo, although this was not significant. Although there were fewer minor complications, 9 (52.9%) patients admitted to the ICU had more complications at grade 5 (17.6% vs. 0, *p* < 0.01). Patients with an ICU stay were discharged insignificantly later (21.5 vs. 18 days, n.s.).


**Table 2.** Perioperative outcome.

DGF rates were higher for patients admitted to the ICU (52.9% vs. 37.5%, n.s.) (see Table 2). Serum creatinine significantly declined after KT (*p* < 0.05) and did not di ffer between patients with or without ICU stay (see Figure 1).

**Figure 1.** Graft function during follow-up. w: week; mo: month; y: year.

#### *3.2. Donor- and Recipient-Age-Dependent Comparison*

In total, 28 (26.7%) patients received a graft from very old donors ≥75 years, compared to 77 (73.3%) old donors ('old-for-old') (see Table 3). When stratifying for donor age (very old-for-old vs. old-for-old), neither recipient nor graft characteristics di ffered. Grafts from very old donors had a tendency towards a longer CIT, which was not significant (677.1 vs. 540.6 min). Kidney recipients of very old donors had a tendency to be admitted to the ICU more frequently (21.4% vs. 14.3%, n.s.), but were discharged significantly earlier (16 vs. 20 days, *p* < 0.05). Neither DGF rates nor the kidney function di ffered during follow-up.

When stratifying for recipient age (old-for-very old vs. old-for-old), 47 (44.7%) recipients were ≥70 years old, and thereby were considered as very old (see Table 3). Regarding recipient characteristics, only the history of smoking di ffered, as fewer very old recipients had a history of smoking (8.5% vs. 24.1%, *p* < 0.05). Neither graft nor transplantation-specific factors were di fferent. Very old recipients were admitted to the ICU insignificantly more often (21.3% vs. 12.1%). Graft function one week after KT was the only parameter that di ffered when comparing very old to old recipients, as very old recipients had a lower serum creatinine than old recipients (3.35 vs. 5.36, *p* < 0.01). During follow-up, the kidney function became equivalent.


**Table 3.** Age-dependent comparison stratifying for donor age (very old donors ≥75 years vs. old donors) or recipient age (very old recipients ≥70 years vs. old recipients).

#### *3.3. Risk Model for ICU Stay*

Among recipient and graft characteristics as well as transplantation-specific outcomes, the BMI of the recipient, an underlying hypertensive nephropathy and CAD were the only significant predictors for ICU admission in univariate and multivariate analysis (see Table 4). A higher BMI lowered the OR for ICU admission (OR 0.8, *p* < 0.01), but a hypertensive nephropathy (OR 4.0, *p* < 0.05) and CAD (OR 4.46, *p* < 0.05) significantly increased the OR for ICU admission during the hospital stay. Donor or recipient age did not impact the risk for ICU admission.

**Table 4.** Multivariate regression analysis to predict an ICU admission during the hospital stay.


When combining these three factors in a risk model to estimate the probability for ICU admission, the c-index reached 0.789 (*p* < 0.001) (see Figure A1). When setting the cut-o ff for the predicted probability of ICU admission to 0.08, which had highest Youden-index, the risk model reached a sensitivity of 94.1%, specificity of 51.1%, false positive rate (FPR) of 48.9%, false negative rate (FNR) of 5.9%, positive predictive value (PPV) of 27.1% and negative predictive value (NPV) of 97.8% (see Table A1).

## *3.4. Survival Analysis*

For all 105 patients, the median length of follow-up was 49.5 months. The overall graft survival at 1, 5 and 9 years was 84%, 73% and 42%, respectively, with a median death-censored graft survival of 113.9 months. Median patient survival was 108.2 months, with a 1-, 5- and 9-year survival of 85%, 62% and 38%, respectively.

When stratifying for ICU admission, patients admitted to the ICU had a significantly shorter graft survival (59.1 vs. 115.7 months, *p* = 0.049) (see Figure 2a). Their 1- and 5-year graft survivals were 75% and 49%, and thereby worse compared to patients without an ICU stay (86% and 77%). Over the whole study period, the death-rate for patients with an ICU stay was almost three times higher compared to patients without an ICU stay (70.6% vs. 26.1%, *p* < 0.001). Consequently, the median patient survival for patients admitted to the ICU was significantly shorter (ICU 36.9 vs. 114.9 months, *p* < 0.001) (see Figure 2b). 1- and 5-year patient survival for patients admitted to an ICU was 57% and 0% and for patients without an ICU stay 90% (1 year), 72% (5 years) and 44% after 9 years, respectively. In total, 17 (48.6%) patients died with a functioning graft, and the DWFG rate did not differ between groups (ICU 50% vs. 47.8%, n.s.). Neither the age of the donor nor the recipient affected graft or patient survival (see Table A2).

**Figure 2.** Death-censored graft survival (**a**) and patient survival (**b**) comparing patients admitted to the ICU (**dashed line**) vs. patients not admitted to the ICU (**solid line**) after kidney transplantation in the ESP program.

In a multivariate cox regression, higher numbers of prior KTs and HLA-mismatches significantly shortened graft survival (hazard ratio (HR) for graft loss 9.66, *p* = 0.001; HR 1.53, *p* < 0.05) (see Table 5). Additionally, higher serum creatinine 1 month after KT was associated with worse graft survival (HR 1.37, *p* < 0.05). ICU admission during the hospital stay after KT did not affect graft survival. Regarding patient survival, a pre-transplant diabetes mellitus and an ICU admission during the hospital stay were significant predictors for worse outcomes in the multivariate analysis (HR for patient death 2.22, *p* < 0.05, HR 4.7, *p* < 0.001). Major complications during the hospital stay and the serum creatinine 1 month after KT were only associated with patient survival in univariate analysis.


