Patients.

The prospective study was approved by the local institutional review board, and written informed consent was obtained from all participants. Between July 2012 and April 2014, 46 selected adult kidney transplant recipients were included in the study. Patient characteristics, medical histories, and transplantation details were recorded. The clinical follow-up period was 6 mo in all subjects recruited and 12 mo in 39/46 patients. One patient was lost to follow-up, and six patients have not yet reached the 12-mo follow-up point. DGF was defined as a failure of S-creatinine to decrease by at least 10% daily on 3 consecutive days or need for dialysis during the first week after transplantation.

MRI protocol and analysis.

MRI was performed in patients 4–11 days after kidney transplantation using a 1.5-T magnet (MAGNETOM Avanto, Siemens Healthcare). T2-weighted turbo spin echo sequences were acquired in axial and coronal planes to assess renal morphology. For renal perfusion measurement, a flow-alternating inversion recovery (FAIR) true FISP ASL technique was used as described previously (18, 21). In brief, images were acquired in an oblique sagittal orientation to avoid covering the aorta or the pelvic arteries. Images were obtained after global and slice-selective inversion pulses using the following parameters: TR/TE = 4.6/2.3 ms, TI = 1,200 ms, flip angle = 70°, averages = 30, FOV = 340 × 340 mm², matrix = 128 × 128, slice thickness = 5 mm, and acquisition time = 4.5 min. In addition, a proton density true FISP image without inversion was acquired to determine the equilibrium magnetization. Motion artifacts were compensated by registration of individual MRI images, and parameter maps of renal perfusion were calculated with Matlab (version 7.11.0.584, MathWorks, Natick, MA) according to

as described previously, where ΔM is the signal difference between FAIR images with global and slice-selective inversion, M₀ is the equilibrium magnetization per unit mass, and λ is the blood-tissue water partition coefficient, which was set to 80 ml/100 g (18, 21). Three regions of interest were placed manually on perfusion maps into the renal cortex in the top, middle, and bottom third of the kidney by one author who was blinded to clinical, laboratory, and histological outcome data, and mean perfusion was calculated. Visible perfusion defects of renal parenchyma were identified and were excluded from the analysis. To evaluate intraobserver variability, a second reader, who was blinded to the results of the first reader and clinical data, analyzed renal perfusion in 20 randomly selected patients.

Clinical data.

S-creatinine was determined daily during the first 10 days after kidney transplantation as well as 6 wk and 3, 6, and 12 mo thereafter. Glomerular filtration rate (eGFR) was calculated based on S-creatinine according to the MDRD formula (19). Urine output, blood pressure, and immunosuppressive medication including serum levels of calcineurin inhibitors (cyclosporine or tacrolimus) were determined on the day of MRI.

Doppler ultrasound.

Ultrasound of the renal transplant was performed by a nephrologist or transplant surgeon within the first and second week after transplantation as described previously (25). The renal resistance index (RI) was measured in two or three proximal segmental arteries and was calculated from peak systolic velocity (V_max) and minimal diastolic velocity (V_min) according to the formula

The RI in the early postoperative period was available in 38/46 patients.

Histology.

Renal allograft biopsies within 5 days of MRI were considered for the analysis. Indication biopsies were performed in 17/26 of DGF patients 7.8 ± 0.5 days after transplantation In 2/20 patients who had initial graft function but slower S-creatinine reduction than expected, biopsies were performed in the second week after transplantation. Biopsy specimens were stained with hematoxylin and eosin and periodic acid Schiff. Sirius red staining was used to visualize collagen fibers. For quantitation, the total stained area of each biopsy specimen after exclusion of major vessels and glomeruli was determined using image J software (National Institutes of Health). Biopsy specimens were analyzed by a nephropathologist and classified according to Banff criteria (30). AKI was diagnosed by tubular necrosis, vacuolization, lysis or loss of the brush border, and tubular epithelial cell flattening. AKI score was defined as follows: 0 = no AKI; 1 = <10%; 2 = 11–25%; 3 = 26–50%; and 4 = >51% affected tubuli.

Statistical analysis.

Statistical analysis was performed with SPSS 22.0 (IBM, Armonk, NY). As no deviations from normal distribution of renal perfusion values were detected by the Kolmogorov-Smirnov test (P = 0.2), parametric statistical tests were used for analysis. Differences in renal perfusion between patients with and without DGF were assessed by unpaired t-tests. In addition, subgroup analyses were performed to evaluate renal perfusion differences between 1) patients with living and with deceased donor transplantation, 2) DGF patients with and without biopsy-proven acute allograft rejection, and 3) DGF patients with persistent severe impairment of renal function (eGFR <30 ml/min) or graft loss and DGF patients with eGFR ≥30 ml/min at 12-mo follow-up by using one-way ANOVA. Adjustment for multiple comparisons was performed with the Sidak method. The intraobserver variability of renal perfusion quantification was evaluated using the coefficient of variation. For comparison of clinical and laboratory parameters between patients with normal graft function and with DGF, unpaired t-tests and Fisher's exact test were used. The correlations of renal perfusion with various clinical parameters were determined by Pearson's coefficient of correlation.

Furthermore, receiver operating characteristic curve analyses were performed to determine the diagnostic performance of renal perfusion measurement by MRI for 1) detection of DGF, 2) prediction of the need for >2 hemodialysis sessions during the first and second week after transplantation, and 3) prediction of graft loss and severe impairment of renal allograft function (eGFR <30 ml/min) in DGF patients at 12 mo follow-up. In addition, Youden-selected thresholds were determined and sensitivities and specificities at threshold are given.

Values are given as means and SE. This is an exploratory study, thus two-sided P values <0.05 are considered statistically significant.

Transplantation details and clinical data.

In total, 20 patients with functional grafts and 26 patients with DGF were examined. As selected patients were included into the study, the proportion of patients with DGF was higher than it is on average at our center (on average 26% DGF patients). Patient characteristics are summarized in Table 1. All patients received triple immunosuppressive therapy with cyclosporine or tacrolimus in addition to mycophenolate mofetil (MMF) and prednisolone. Basiliximab was applied as induction therapy. DGF patients were significantly older (58.9 ± 2.1 vs. 48.8 ± 3.8 yr, P < 0.05) and had more coronary and peripheral artery occlusive disease (P < 0.05). Donor S-creatinine was significantly lower in patients with initial graft function (67.7 ± 4.9 vs. 116.0 ± 19.8 μmol/l, P < 0.05, Table 1). Living donor transplantation was performed in 6/26 DGF patients and in 9/20 patients with initial graft function. No differences between recipients with initial graft function and DGF in the number of human leukocyte antigen mismatches, cold ischemia time, the immunosuppressive regime, and levels of immunosuppressive drugs were observed.

Renal function measured by eGFR 7 days after transplantation was reduced in all patients with DGF compared with patients with initial graft function (14.5 ± 1.7 vs. 42.5 ± 16 ml/min, P < 0.001). Consequently, many patients in the DGF group required hemodialysis after transplantation whereas in the group with initial graft function no dialysis was necessary (16/26 vs. 0/20, P < 0.001).

Allograft histology after clinically indicated biopsies.

Indicated biopsies were performed in 17 patients in the DGF group within 5 days of the MRI examination. Histological analysis in DGF patients revealed five patients with borderline rejection according to the Banff classification, three with acute T cell-mediated rejections (Banff Ib, IIa, and III), and one with an antibody-mediated rejection. Eight patients in the DGF group had other changes such as tubular damage or glomerulosclerosis. One patient had severe thrombotic microangiopathy 6 days after transplantation in combination with signs of T cell-mediated rejection, and removal of the kidney was necessary. Two patients with initial graft function had biopsies 8 and 10 days after kidney transplantation because their S-creatinine decreased more slowly than expected. One of these patients had borderline changes and was treated with steroid pulse therapy, and one had mild acute tubular injury. Mean AKI scores were 0 in the group with initial graft function and 0.9 ± 0.3 in the group with DGF.

Renal perfusion measured by functional MRI.

Intraobserver variability of renal perfusion quantification was low, with a coefficient of variation of 2.4%. Renal perfusion in allografts from DGF patients was significantly lower than in allografts with initial renal function (231 ± 15 vs. 331 ± 15 ml·min⁻¹·100 g⁻¹, P < 0.001). Representative perfusion maps can be seen in Fig. 1. Living donor grafts exhibited higher perfusion values compared with deceased donor grafts (339 ± 19 vs. 243 ± 14 ml·min⁻¹·100 g⁻¹, P < 0.001), and renal perfusion significantly correlated with cold ischemia time (r = −0.48, P < 0.01). Furthermore, slightly lower perfusion values were observed in patients with acute renal allograft rejection at histology vs. patients with DGF but without signs of an acute allograft rejection (202 ± 35 vs. 246 ± 14 ml·min⁻¹·100 g⁻¹, not significant, P = 0.14, Fig. 1D).

Renal perfusion measured by MRI was positively correlated with eGFR on the day of MRI (r = 0.64, P < 0.001, Fig. 2A) and was negatively correlated with RI (r = −0.57 P < 0.001, Fig. 2B). In addition, renal perfusion significantly correlated with the number of dialysis sessions required in the first and second week after transplantation (r = −0.63, P < 0.001), with cold ischemia time (r = −0.48, P < 0.01, Fig. 2C), and recipient age (r = −0.42, P < 0.01). Serum levels of cyclosporine or tacrolimus, number of human leukocyte antigen mismatches, blood pressure, and recipient height and weight did not correlate significantly with renal perfusion.

The best diagnostic performance to detect patients with DGF was achieved by renal perfusion below a Youden-selected threshold of 278 ml·min⁻¹·100 g⁻¹ with a sensitivity and specificity of 81% (95% CI [61%;93%]) and 75% (95% CI [51%;91%]), respectively (Youden index at threshold 56%, Fig. 2D; CI, confidence interval). The sensitivity of renal perfusion measurement to identify patients who required more than two hemodialysis sessions was 100% (95% CI [54%;100%]), and the specificity was 80% (95% CI [64%;91%]) at a threshold of 236 ml·min⁻¹·100 g⁻¹ (Fig. 2E).

Renal perfusion and renal outcome after 1 yr.

After 12 mo, DGF patients were reevaluated and divided into two groups based on their renal function measured by eGFR: one group improved over time and reached eGFR levels ≥30 ml/min; the other group continued to have impaired renal function 12 mo after transplantaton with eGFR <30 ml/min (Fig. 3A). In addition, three patients did not reach the 12-mo time point due to allograft nephrectomy because of severe thrombotic microangiopathy (TMA) and mycotic allograft nephritis 9 days and 6 wk after transplantation, respectively. One patient with transplant glomerulosclerosis, severe tubular atrophy, and graft fibrosis died before the scheduled nephrectomy due to cerebral infarction followed by sepsis 5 mo after transplantation.

Note that DGF patients with better renal function after 12 mo (eGFR ≥30 ml/min) had significantly higher renal perfusion early after transplantation (259 ± 16 ml·min⁻¹·100 g⁻¹) than DGF patients with persistent severe impairment of renal function eGFR <30 ml/min or graft loss (163 ± 22 ml·min⁻¹·100 g⁻¹, P < 0.01, Fig. 3B). Renal allograft perfusion impairment below 186 ml·min⁻¹·100 g⁻¹ in DGF patients predicted graft loss or severe impairment of renal function (eGFR <30 ml/min) 1 yr after transplantation, with a sensitivity of 75% (95% CI [35%;97%]) and a specificity of 92% (95% CI [62%;100%]) (Fig. 3C).

For comparison, renal fibrosis quantified by the sirius red-positive tubulointerstitial area in indicated biopsy specimens early after transplantation was highest in patients with graft loss or persistent severe reduction of eGFR <30 ml/min after 1 yr compared with patients without DGF or better renal function (28.1 ± 3.4 vs. 10.3 ± 6 vs. 20.3 ± 2.9%, not significant, Fig. 4).