ABSTRACT
Background
Whether severe donor acute kidney injury (AKI) requiring dialysis affects posttransplant outcomes remains debated and may influence center‐level acceptance decisions.
Methods
This retrospective cohort study of United States deceased‐donor kidney transplants (2015–2020) used OPTN data with follow‐up through September 2024. Donor AKI was staged by KDIGO 2012 criteria, and dialysis was identified using a validated rule‐based natural language processing (NLP) algorithm applied to DonorNet records. The primary outcome was all‐cause graft failure (ACGF). Secondary outcomes included delayed graft function (DGF), 1‐year estimated glomerular filtration rate (eGFR), and a composite tolerability outcome within 6 months defined as DGF, eGFR <30 mL/min/1.73 m2, graft failure, or death. Models used inverse probability weighting (IPW), clustering by transplant center, and exploratory interpretable machine learning.
Results
Of 74 939 transplants, 24.6% were from AKI donors (Stage 2 7.8%, Stage 3 without dialysis 5.8%, Stage 3 with dialysis 2.1%). DGF increased with AKI severity (Stage 3 with dialysis 54.5% vs Stage 2 38.1%), but adjusted ACGF risk was similar across groups. Donor KDPI and recipient EPTS were stronger predictors of graft failure than AKI stage. Exploratory tolerability analyses showed that early adverse outcomes were driven primarily by recipient age, BMI, dialysis duration, and organ quality, not by donor AKI severity or dialysis.
Conclusions
Among transplanted kidneys, Stage 3 AKI requiring donor dialysis was associated with higher early DGF but similar adjusted graft survival compared with Stage 2 AKI after accounting for organ quality and recipient risk. These findings reflect outcomes among AKI‐affected kidneys accepted for transplantation and should not be generalized to declined or discarded kidneys.
Keywords: acute kidney injury, deceased donor, delayed graft function, graft survival, inverse probability weighting, kidney donor profile index, kidney transplantation
Abbreviations
- ACGF
all‐cause graft failure
- AKI
acute kidney injury
- BMI
body mass index
- CIT
cold ischemia time
- DCD
donation after circulatory death
- DGF
delayed graft function
- eGFR
estimated glomerular filtration rate
- EPTS
Estimated Posttransplant Survival
- IPW
inverse probability weighting
- KDIGO
Kidney Disease: Improving Global Outcomes
- KDPI
Kidney Donor Profile Index
- NLP
natural language processing
- OPTN
Organ Procurement and Transplantation Network
- UNOS
United Network for Organ Sharing.
1. Introduction
Kidney transplantation is the most effective treatment for end‐stage kidney disease, providing superior survival and quality of life compared with dialysis [1, 2, 3, 4]. However, the continued shortage of donor organs remains a major challenge [5, 6], prompting efforts to expand the deceased donor pool through the use of higher‐risk organs, including those from expanded criteria donors and donation after circulatory death (DCD) [7]. Among these, kidneys from donors with acute kidney injury (AKI) represent a controversial but potentially valuable source. AKI affects up to 30% of deceased donors and nearly half of DCD donors, often resulting from pre‐recovery factors such as hemodynamic instability or cardiac arrest [8, 9].
While donor AKI has been linked to higher rates of delayed graft function (DGF) and other early complications, growing evidence indicates that long‐term graft survival is often comparable to transplants from non‐AKI donors [10, 11]. Despite this, kidneys from donors with severe AKI, particularly those meeting KDIGO Stage 3 criteria, are frequently discarded due to concerns about primary non‐function and early graft loss [10, 11, 12, 13]. DGF, defined as the need for dialysis within the first week after transplant, is influenced by donor, perioperative, and recipient factors. Severe donor AKI, prolonged cold or warm ischemia time, and recipient factors such as older age, obesity, and long dialysis duration increase this risk [14, 15, 16, 17]. Understanding how these factors interact is essential to identifying which recipients can best tolerate kidneys from donors with significant injury.
This study evaluates the impact of moderate‐to‐severe donor AKI on short‐ and long‐term posttransplant outcomes in a large national cohort of deceased‐donor kidney transplants. Because OPTN recipient data capture transplanted organs, these analyses reflect outcomes among AKI‐affected kidneys that proceeded to transplantation and do not include kidneys that were declined or discarded. We also examine donor and recipient factors, including the need for dialysis at AKI Stage 3, that may influence graft survival to inform evidence‐based strategies for allocation and matching of AKI‐affected donor kidneys.
2. Materials and Methods
2.1. Study Design and Data Source
This retrospective cohort study included adult recipients (≥18 years) of solitary deceased donor kidney transplants performed between January 1, 2015 and December 31, 2020. Data were obtained from the Organ Procurement and Transplantation Network (OPTN) database. The OPTN data system captures all U.S. organ donors, wait‐listed candidates, and transplant recipients, as submitted by OPTN member institutions. Oversight of the OPTN is provided by the Health Resources and Services Administration (HRSA), U.S. Department of Health and Human Services. Follow‐up data were available through September 30, 2024. The study was reviewed and determined to be exempt from institutional review board oversight as it involved secondary analysis of de‐identified data. The requirement for informed consent was waived.
2.2. Cohort and Exclusions
The analytic cohort was limited to adult recipients of solitary kidney transplants from deceased donors with moderate‐to‐severe AKI. This study was designed as a posttransplant outcomes analysis and therefore includes only kidneys that were transplanted; it does not evaluate center‐level offer acceptance behavior or outcomes among kidneys that were recovered but declined or discarded. Donors were excluded if they were younger than 16 years, weighed less than 40 kg, had only one serum creatinine measurement, or were part of a multi‐organ transplant.
2.3. Donor AKI Classification
Donor AKI was defined using the 2012 Kidney Disease: Improving Global Outcomes (KDIGO) criteria based on changes in serum creatinine. Consistent with prior registry‐based studies, donors with a terminal creatinine >1.5 mg/dL were also classified as having AKI to capture clinically significant injury when baseline trajectories were uncertain. Donors with fewer than two recorded creatinine values were excluded. Urine output was not used for primary staging due to inconsistent reporting but was evaluated descriptively and in sensitivity analyses using the RIFLE classification (Risk, Injury, Failure, Loss, End‐stage kidney disease) to assess robustness of findings. Creatinine values used for staging were restricted to within seven days of organ recovery. Final donor categories were:
Stage 2 AKI: increase in serum creatinine ≥2.0 to 2.9 times baseline (reference group).
Stage 3 AKI without dialysis: increase ≥3.0 times baseline or terminal creatinine ≥4.0 mg/dL without dialysis.
Stage 3 AKI with dialysis: Stage 3 AKI in donors who received dialysis before procurement, identified through natural language processing (NLP) as described below.
2.4. Ascertainment of Donor Dialysis Using NLP
Because donor dialysis is not a structured field in OPTN data, we developed a deterministic rule‐based NLP algorithm to extract dialysis information from free‐text DonorNet records. The algorithm required the co‐occurrence of a dialysis‐related term (e.g., “hemodialysis,” “CRRT,” “RRT,” “SLED,” “SLEDD”) with an active verb (“initiated,” “continued,” “performed”) within a 6‐word context window. Negations (e.g., “no dialysis,” “not on CRRT”) were excluded. Mentions clearly outside the donation timeline were filtered using temporal anchors referencing the admission‐to‐procurement window (±72 h). Validation was performed in a stratified random sample of 200 donors (100 NLP‐positive and 100 NLP‐negative) reviewed independently by two clinical adjudicators blinded to NLP results. Agreement between NLP classification and manual adjudication was assessed by percent agreement and Cohen's κ. Discrepancies were resolved by consensus, and overall accuracy exceeded 90%.
2.5. Urine Output Data Extraction
When reported, urine output was extracted from DonorNet in cubic centimeters per hour and converted to milliliters per kilogram per hour using donor admission weight. Three windows were analyzed: 6‐, 12‐, and 24‐h periods ending at the time of cross‐clamp or the closest documented interval within 6 h prior. When proximal data were unavailable, the most complete contiguous 6‐, 12‐, or 24‐h intervals within the preceding 48 h were used and flagged for sensitivity analysis.
2.6. Study Outcomes
The primary outcome was all‐cause graft failure (ACGF), defined as return to dialysis, re‐transplantation, or recipient death during follow‐up.
Secondary outcomes included:
DGF: need for dialysis within 7 days posttransplant.
Kidney function at 1 year: estimated glomerular filtration rate (eGFR, mL/min/1.73 m2).
Composite tolerability outcome: occurrence within 6 months of any of the following: DGF, eGFR <30 mL/min/1.73 m2, graft failure, or death.
2.7. Recipient “Tolerability” Analysis
To identify recipient characteristics that modify the effect of donor AKI on posttransplant outcomes, we conducted an exploratory logistic regression evaluating the composite tolerability outcome. Prespecified interactions between donor AKI stage and recipient factors were tested, including age, sex, body mass index (BMI), preemptive listing, dialysis duration, functional status, and travel distance between the recipient residence and transplant center (ZIP‐to‐ZIP straight‐line distance). Additionally, interpretable machine learning models were tested to validate the logistic findings and assess complex interactions. These included:
Explainable Boosting Machines (EBM) to capture nonlinear effects and global importance measures.
Filter‐based variable selection using mutual information gain to retain variables with the strongest relationship to the outcome before modeling.
Global and patient‐level interpretability metrics, including variable importance and partial dependence plots, were examined to visualize protective or adverse recipient profiles.
2.8. Statistical Analysis
Continuous variables were summarized as mean ± standard deviation or median (interquartile range) and compared using Student's t‐test, one‐way ANOVA, or Kruskal–Wallis tests. Categorical variables were summarized as frequencies (percentages) and compared using chi‐square tests. Survival probabilities were estimated using Kaplan–Meier methods and compared using the log‐rank test. Multivariable Cox proportional hazards models were used to estimate hazard ratios (HRs) for ACGF across donor AKI groups. The proportional hazards assumption was verified using log‐log plots and Schoenfeld residuals.
To reduce confounding, inverse probability weighting (IPW) was applied to create a weighted pseudo‐population balanced across AKI exposure groups. Covariates in the weighting model included donor factors (age, KDPI, biopsy findings of glomerulosclerosis, arteriosclerosis, tubular atrophy, and interstitial fibrosis; DCD status; use of machine perfusion; donor hospital‐to‐transplant center distance; and cold ischemia time category) and recipient factors (age, diabetes, dialysis duration, re‐transplantation, and BMI). Weights were stabilized and truncated at the first and 99th percentiles to avoid extreme values. Covariate balance was confirmed by standardized mean differences (<0.1) and variance ratios (∼1.0). Three modeling approaches were compared:
Univariate (unadjusted) analysis,
IPW‐adjusted model, and
Fully adjusted multivariable model incorporating additional immunologic variables (cPRA, HLA mismatch, CMV serostatus).
All models were clustered by transplant center to account for center‐level variation. Interaction terms between AKI stage and key donor or recipient variables (KDPI category, cold ischemia time, DCD status) were included to test for effect modification. Missing covariate data were handled using indicator categories. Statistical significance was defined as two‐tailed p < 0.05. Analyses were performed using R version 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria) and Stata version 17 (StataCorp, College Station, TX).
3. Results
3.1. Cohort Selection
Between 2015 and 2020, 74 939 deceased donor kidney‐only transplants were performed in the United States. Of these, 18.6% involved donors with KDIGO Stage 1 AKI, 7.8% Stage 2, and 7.9% Stage 3. The analytic cohort comprised 11 756 recipients whose donors had moderate to severe AKI (Stage 2 or Stage 3) as defined in Methods (Table S1). Because OPTN recipient registry data capture transplanted organs, these analyses reflect outcomes among AKI‐affected kidneys that proceeded to transplantation and do not include declined or discarded kidneys.
3.2. Cohort Characteristics
Baseline donor and transplant features by AKI category are provided in Table 1. Donors with Stage 3 AKI on dialysis were younger than other groups (median 34 years, interquartile range [IQR] 26–42) compared with Stage 2 (40 years, IQR 30–50) and Stage 3 without dialysis (38 years, IQR 29–48; p < 0.001). Nearly 36.4% of Stage 3 on dialysis donors were younger than 30 years versus 24.7% in Stage 2 and 27.5% in Stage 3 without dialysis.
TABLE 1.
Study cohort characteristics.
| K‐DIGO AKI staging | ||||
|---|---|---|---|---|
| Variable | Stage 2 N = 5810 |
Stage 3 no dialysis N = 4361 |
Stage 3 on dialysis N = 1585 |
p‐value |
| Donor Variables | ||||
| Urine output 24 h prior to recovery surgery, n (%) | <0.001 | |||
| >800 mL | 2592 (44.6) | 1680 (38.5) | 334 (21.1) | |
| 400–800 mL | 382 (6.6) | 286 (6.6) | 84 (5.3) | |
| <400 mL | 440 (7.6) | 587 (13.5) | 407 (25.7) | |
| No UOP data | 2396 (41.2) | 1808 (41.5) | 760 (48.0) | |
| Creatinine minimum, median [IQR] | 1.2 [0.86, 1.50] | 0.94 [0.64, 1.30] | 1.10 [0.70, 1.60] | <0.001 |
| Donor peak creatinine, median [IQR] | 2.50 [2.00, 3.27] | 4.20 [2.90, 5.80] | 4.00 [1.80, 6.40] | <0.001 |
| Donor terminal creatinine, median [IQR] | 2.05 [1.70, 2.84] | 4.10 [2.78, 5.64] | 2.21 [1.02, 3.87] | <0.001 |
| Donor age, median [IQR] | 40 [30, 50] | 38 [29, 48] | 34 [26, 42] | <0.001 |
| Donor age category | <0.001 | |||
| <30 | 1433 (24.7) | 1199 (27.5) | 577 (36.4) | |
| 30–50 | 2850 (49.1) | 2276 (52.2) | 796 (50.2) | |
| >50 | 1527 (26.3) | 886 (20.3) | 212 (13.4) | |
| Donor gender (male) | 4183 (72.0) | 2733 (62.7) | 983 (62.0) | <0.001 |
| Donor race/ethnicity | <0.001 | |||
| White | 3716 (64.0) | 2680 (61.5) | 995 (62.8) | |
| African American | 1064 (18.3) | 709 (16.3) | 252 (15.9) | |
| Hispanic | 799 (13.8) | 799 (18.3) | 271 (17.1) | |
| Asian | 147 (2.5) | 125 (2.9) | 42 (2.7) | |
| Other | 84 (1.5) | 48 (1.1) | 25 (1.6) | |
| Donor height (m), median [IQR] | 1.75 [1.68, 1.80] | 1.73 [1.65, 1.80] | 1.73 [1.65, 1.80] | <0.001 |
| Donor height category | <0.001 | |||
| <5’9’’ | 4315 (74.3) | 3415 (78.3) | 1227 (77.4) | |
| ≥5’9’’ | 1495 (25.7) | 946 (21, 7) | 358 (22.6) | |
| Donor KDPI, median [IQR] | 49.0 [30.0, 69.0] | 46.0 [29.0, 66.0] | 33.0 [20.0, 50.0] | <0.001 |
| KDPI category | <0.001 | |||
| 0–20 | 729 (12.6) | 467 (10.7) | 384 (24.2) | |
| 20–35 | 1136 (19.6) | 961 (22.0) | 447 (28.2) | |
| 35–85 | 3425 (59.0) | 2606 (59.8) | 710 (44.8) | |
| 85–100 | 520 (9.0) | 327 (7.5) | 44 (2.8) | |
| Donation after circulatory death (DCD) n (%) | 782 (13.5) | 504 (11.6) | 232 (14.6) | 0.002 |
| Donor cause of death | <0.001 | |||
| CVA | 1250 (21.5) | 785 (18.0) | 166 (10.5) | |
| Others | 4560 (78.5) | 3576 (82.0) | 1419 (89.5) | |
| Recipient variables | ||||
| Median (IQR) posttransplant follow up, yr | 4.1 (3.0, 6.0) | 4.1 (3.0, 5.8) | 4.1 (3.0, 5.7) | 0.01 |
| Recipient age, median [IQR] | 57.0 [47.0, 65.0] | 56.0 [47.0, 64.0] | 55.0 [44.0, 64.0] | <0.001 |
| Recipient age Category | <0.001 | |||
| <40 | 744 (12.8) | 562 (12.9) | 275 (17.4) | |
| 40–60 | 2621 (45.1) | 2003 (45.9) | 719 (45.4) | |
| >60 | 2445 (42.1) | 1796 (41.2) | 591 (37.3) | |
| Recipient gender (male) | 3578 (61.6) | 2744 (62.9) | 980 (61.8) | 0.38 |
| Recipient race/ethnicity | <0.001 | |||
| White | 2020 (34.8) | 1413 (32.4) | 504 (31.8) | |
| African American | 2068 (35.6) | 1387 (31.8) | 498 (31.4) | |
| Hispanic | 1144 (19.7) | 1110 (25.5) | 404 (25.5) | |
| Asian | 459 (7.9) | 359 (8.2) | 129 (8.1) | |
| Other | 119 (2.1) | 92 (2.1) | 50 (3.2) | |
| EPTS, median [IQR] | 56.0 [29.0, 81.0] | 55.0 [28.0, 79.0] | 51.0 [21.0, 77.0] | <0.001 |
| EPTS Category | <0.001 | |||
| 0–20 | 932 (16.0) | 722 (16.6) | 379 (23.9) | |
| 20–35 | 835 (14.4) | 622 (14.3) | 198 (12.5) | |
| 35–85 | 2820 (48.5) | 2204 (50.6) | 742 (46.8) | |
| 85–100 | 1223 (21.1) | 812 (18.6) | 266 (16.8) | |
| On dialysis >3 years | 4159 (71.6) | 3087 (70.8) | 1101 (69.5) | 0.24 |
| Diabetes type | 0.14 | |||
| None | 3459 (59.5) | 2536 (58.2) | 970 (61.2) | |
| Type I | 137 (2.4) | 106 (2.4) | 45 (2.8) | |
| Type II | 2214 (38.1) | 1719 (39.4) | 570 (36.0) | |
| CPRA, median [IQR] | 0.0 [0.0, 41.7] | 0.0 [0.0, 26.6] | 0.0 [0.0, 26.5] | <0.001 |
| CPRA category | <0.001 | |||
| <20 | 4008 (69.0) | 3153 (72.3) | 1148 (72.4) | |
| 20–80 | 904 (15.6) | 676 (15.5) | 245 (15.5) | |
| 80–95 | 353 (6.1) | 257 (5.9) | 85 (5.4) | |
| ≥95 | 545 (9.4) | 275 (6.3) | 107 (6.8) | |
| Retransplant | 642 (11.1) | 363 (8.3) | 133 (8.4) | <0.001 |
| BMI, median [IQR] | 28.0 [25.0, 32.0] | 28.0 [25.0, 32.0] | 28.0 [24.0,32.0] | 0.14 |
| BMI category | 0.51 | |||
| <25 | 1406 (24.2) | 1057 (24.2) | 423 (26.7) | |
| 25–30 | 1985 (34.2) | 1512 (34.7) | 536 (33.8) | |
| 30–35 | 1536 (26.4) | 1140 (26.1) | 393 (24.8) | |
| >35 | 883 (15.2) | 652 (15.0) | 233 (14.7) | |
| Cold ischemia time (h), median [IQR] | 18.0 [13.0, 24.0] | 20.0 [15.0, 26.0] | 20.0 [15.0, 26.0] | <0.001 |
| Cold ischemia time category | <0.001 | |||
| <12 | 1020 (17.6) | 621 (14.2) | 204 (12.9) | |
| 12–24 | 3177 (54.7) | 2228 (51.1) | 814 (51.4) | |
| 24–36 | 1255 (21.6) | 1171 (26.9) | 468 (29.5) | |
| >36 | 358 (6.2) | 341 (7.8) | 99 (6.3) | |
| Pump | 3385 (58.3) | 2880 (66.0) | 997 (62.9) | <0.001 |
| HLA mismatch, median [IQR] | 4.0 [4.0, 5.0] | 4.0 [4.0, 5.0] | 5.0 [4.0, 5.0] | 0.03 |
| HLA mismatch category | 0.01 | |||
| 0–2 | 568 (9.8) | 404 (9.3) | 109 (6.9) | |
| 3–4 | 2430 (41.8) | 1798 (41.2) | 678 (42.8) | |
| 5–6 | 2812 (48.4) | 2159 (49.5) | 798 (50.4) | |
| CMV high risk mismatch | 1014 (17.5) | 735 (16.9) | 291 (18.4) | 0.38 |
| Biopsy | 4392 (75.6) | 3785 (86.8) | 1345 (84.9) | <0.001 |
| Glomerulosclerosis category | <0.001 | |||
| <5 | 3282 (56.5) | 2975 (68.2) | 1123 (70.9) | |
| 5–10 | 667 (11.5) | 480 (11.0) | 143 (9.0) | |
| 10–15 | 225 (3.9) | 187 (4.3) | 37 (2.3) | |
| >15 | 214 (3.7) | 132 (3.0) | 38 (2.4) | |
| Biopsy not done or missing | 1422 (24.5) | 587 (13.5) | 244 (15.4) | |
| IFTA | <0.001 | |||
| Absent | 3349 (57.6) | 2887 (66.2) | 1021 (64.4) | |
| Mild | 869 (15.0) | 746 (17.1) | 284 (17.9) | |
| Moderate | 98 (1.7)) | 64 (1.5) | 19 (1.2) | |
| Biopsy not done or missing | 1494 (25.7) | 664 (15.2) | 261 (16.5) | |
| Arteriosclerosis (AS) | <0.001 | |||
| Absent | 3282 (56.5) | 2860 (65.6) | 1037 (65.4) | |
| Mild | 748 (12.9) | 585 (13.4) | 192 (12.1) | |
| Moderate | 173 (3.0) | 138 (3.2) | 43 (2.7) | |
| Biopsy not done or missing | 1607 (27.7) | 778 (17.8) | 313 (19.8) | |
| Induction agents | 0.01 | |||
| Thymoglobulin | 3501 (60.3) | 2663 (61.1) | 921 (58.1) | |
| Basiliximab | 1111 (19.1) | 845 (19.4) | 292 (18.4) | |
| Alemtuzumab | 750 (12.9) | 582 (13.4) | 249 (15.7) | |
| No induction | 448 (7.7) | 271 (6.2) | 123 (7.8) | |
| Transplant laterality | 0.83 | |||
| Left kidney | 2820 (48.5) | 2146 (49.2) | 795 (50.2) | |
| Right kidney | 2919 (50.2) | 2165 (49.6) | 776 (49.0) | |
| Enbloc | 7 (0.1) | 4 (0.1) | 2 (0.1) | |
| Dual kidney | 64 (1.1) | 46 (1.1) | 12 (0.8) | |
Kidney Donor Profile Index differed across groups. Stage 3 on dialysis had the lowest median KDPI at 33% (IQR 20–50) compared with 49% (IQR 30–69) for Stage 2 and 46% (IQR 29–66) for Stage 3 without dialysis (p < 0.001). High‐quality organs were more common in Stage 3 on dialysis (0%–0% KDPI in 24.2%) than Stage 2 (12.6%) and Stage 3 without dialysis (10.7%). Oliguria less than 400 mL per day was more frequent in Stage 3 on dialysis (25.7%) than Stage 2 (7.6%) and Stage 3 without dialysis (13.5%). Distribution of DCD and machine perfusion use was similar across AKI categories.
Donors with Stage 3 AKI requiring dialysis were younger (median 34 years [IQR 26–42]) and had lower KDPI (median 33% [IQR 20–50]) compared with Stage 2 and Stage 3 without dialysis (p < 0.001). Oliguria was more frequent among Stage 3 dialysis donors (25.7%), but donation type and machine perfusion use were similar across groups. Within this subgroup (Table 2), donors with functioning grafts were younger (33 vs. 36.5 years, p < 0.001), more often male (64.7% vs. 54.0%, p < 0.001), and had lower KDPI (32 vs. 40, p < 0.001). Recipients with functioning grafts were also younger (53 vs. 60 years, p < 0.001), had lower EPTS (44 vs. 66, p < 0.001), and were less likely to have diabetes (35.5% vs. 48.4%, p < 0.001). Other factors, including cold ischemia time, HLA mismatch, and biopsy findings, did not differ significantly between groups. These results suggest that younger age, lower KDPI, and favorable metabolic profiles in donors and recipients were associated with better graft outcomes among Stage 3 dialysis kidneys.
TABLE 2.
Comparison of the cohort characteristics of patients with K‐DIGO AKI Stage 3 with dialysis by primary outcome during the study period.
| K‐DIGO AKI Stage 3 on dialysis cohort during study period | |||
|---|---|---|---|
| Variable |
Alive/graft functioning N = 1189 (75.0%) |
All‐cause graft failure N = 396 (25.0%) |
p‐value |
| Donor variables | |||
| Urine output 24 h prior to recovery surgery, n (%) | 0.75 | ||
| >800 mL | 247 (20.8) | 87 (22.0) | |
| 400–800 mL | 64 (5.4) | 20 (5.1) | |
| <400 mL | 313 (26.3) | 94 (23.7) | |
| No UOP data | 565 (47.5) | 195 (49.2) | |
| Creatinine minimum, median [IQR] | 1.2 [0.7, 1.6] | 1.0 [0.7, 1.6] | 0.05 |
| Donor peak creatinine, median [IQR] | 4.0 [1.8, 6.5] | 3.7 [1.7, 6.2] | 0.11 |
| Donor terminal creatinine, median [IQR] | 2.3 [1.1, 4.0] | 2.1 [1.0, 3.7] | 0.24 |
| Donor age, median [IQR] | 33.0 [26.0, 41.0] | 36.5 [27.0, 47.0] | <0.001 |
| Donor age category | 0.01 | ||
| <30 | 449 (37.8) | 128 (32.3) | |
| 30–50 | 598 (50.3) | 198 (50.0) | |
| >50 | 142 (11.9) | 70 (17.7) | |
| Donor gender (male) | 749 (64.7) | 214 (54.0) | <0.001 |
| Donor race/ethnicity | 0.50 | ||
| White | 756 (63.6) | 239 (60.4) | |
| African American | 178 (15.0) | 74 (18.7) | |
| Hispanic | 203 (17.1) | 68 (17.2) | |
| Asian | 33 (2.8) | 9 (2.3) | |
| Other | 19 (1.6) | 6 (1.5) | |
| Donor height (m), median [IQR] | 1.7 [1.7, 1.8] | 1.7 [1.6, 1.8] | 0.001 |
| Donor height category | 0.11 | ||
| <5’9’’ | 909 (76.5) | 318 (80.3) | |
| ≥5’9’’ | 280 (23.6) | 78 (19.7) | |
| Donor KDPI, median [IQR] | 32.0 [19.0, 48.0] | 40.0 [23.0, 59.0] | <0.001 |
| KDPI category | 0.001 | ||
| 0–20 | 306 (25.7) | 78 (19.7) | |
| 20–35 | 349 (29.4) | 98 (24.8) | |
| 35–85 | 507 (42.6) | 203 (51.3) | |
| 85–100 | 27 (2.3) | 17 (4.3) | |
| Donation after circulatory death, n (%) | 172 (14.5) | 60 (15.2) | 0.74 |
| Donor cause of death | 0.07 | ||
| CVA | 115 (9.7) | 51 (12.9) | |
| Others | 1074 (90.3) | 345 (87.1) | |
| Recipient variables | |||
| Median (IQR) posttransplant follow up, y | 4.4 (3.5, 6.0) | 2.9 (1.3, 4.4) | <0.001 |
| Recipient age, median [IQR] | 53.0 [43.0, 62.0] | 60.0 [48.0, 67.0] | <0.001 |
| Recipient age category | <0.001 | ||
| <40 | 219 (18.4) | 56 (14.1) | |
| 40–60 | 582 (49.0) | 137 (34.6) | |
| >60 | 388 (32.6) | 203 (51.3) | |
| Recipient gender (male) | 728 (61.2) | 252 (63.6) | 0.39 |
| Recipient race/ethnicity | 0.31 | ||
| White | 368 (31.0) | 136 (34.3) | |
| African American | 369 (31.0) | 129 (32.6) | |
| Hispanic | 312 (26.2) | 92 (23.2) | |
| Asian | 104 (8.8) | 25 (6.3) | |
| Other | 36 (3.0) | 14 (3.5) | |
| EPTS, median [IQR] | 44.0 [18.0, 72.0] | 66.0 [34.5, 85.5] | <0.001 |
| EPTS Category | <0.001 | ||
| 0–20 | 316 (26.6) | 63 (15.9) | |
| 20–35 | 162 (13.6) | 36 (9.1) | |
| 35–85 | 549 (46.2) | 193 (48.7) | |
| 85–100 | 162 (13.6) | 104 (26.3) | |
| Dialysis duration category | 0.20 | ||
| Preemptive | 123 (10.3) | 29 (7.3) | |
| <3 years | 361 (30.4) | 121 (30.6) | |
| ≥3 years | 705 (59.3) | 246 (62.1) | |
| Diabetes type | <0.001 | ||
| None | 766 (64.4) | 204 (51.5) | |
| Type I | 29 (2.4) | 16 (4.0) | |
| Type II | 394 (33.1) | 176 (44.4) | |
| CPRA, median [IQR] | 0.0 [0.0, 26.2] | 0.0 [0.0, 36.9] | 0.05 |
| CPRA category | 0.10 | ||
| <20 | 872 (73.3) | 276 (69.7) | |
| 20–80 | 181 (15.2) | 64 (16.2) | |
| 80–95 | 66 (5.6) | 19 (4.8) | |
| ≥95 | 70 (5.9) | 37 (9.3) | |
| Retransplant | 99 (8.3) | 34 (8.6) | 0.87 |
| BMI, median [IQR] | 28.0 [24.0, 32.0] | 29.0 [25.0, 33.0] | <0.001 |
| BMI category | 0.17 | ||
| <25 | 219 (18.4) | 60 (15.2) | |
| 25–30 | 259 (21.8) | 81 (20.5) | |
| 30–35 | 536 (45.1) | 181 (45.7) | |
| >35 | 175 (14.7) | 74 (18.7) | |
| Transplant related factors | |||
| Cold ischemia time (h), median [IQR] | 20.0 [15.0, 26.0] | 20.0 [14.5, 26.0] | 0.83 |
| Cold ischemia time category | 0.31 | ||
| <12 | 144 (12.1) | 60 (15.2) | |
| 12–24 | 615 (51.7) | 199 (50.3) | |
| 24–36 | 359 (30.2) | 109 (27.5) | |
| >36 | 71 (6.0) | 28 (7.1) | |
| Pump | 752 (63.3) | 245 (61.9) | 0.62 |
| HLA mismatch, median [IQR] | 5.0 [4.0, 5.0] | 4.0 [4.0, 5.0] | 0.35 |
| HLA mismatch category | 0.92 | ||
| 0–2 | 81 (6.8) | 28 (7.1) | |
| 3–4 | 506 (42.6) | 172 (43.4) | |
| 5–6 | 602 (50.6) | 196 (49.5) | |
| CMV high risk mismatch | 206 (17.3) | 85 (21.5) | 0.07 |
| Biopsy (Yes) | 1019 (85.7) | 326 (82.3) | 0.10 |
| Glomerulosclerosis category | 0.52 | ||
| <5 | 853 (71.7) | 270 (68.2) | |
| 5–10 | 105 (8.8) | 38 (9.6) | |
| 10–15 | 28 (2.4) | 9 (2.3) | |
| >15 | 30 (2.5) | 8 (2.0) | |
| Biopsy not done or missing | 173 (14.6) | 71 (17.9) | |
| IFTA | 0.27 | ||
| Absent | 781 (65.7) | 240 (60.6) | |
| Mild | 209 (17.6) | 75 (18.9) | |
| Moderate | 14 (1.2) | 5 (1.3) | |
| Biopsy not done or missing | 185 (15.6) | 76 (19.2) | |
| Arteriosclerosis (AS) | 0.32 | ||
| Absent | 789 (66.4) | 248 (62.6) | |
| Mild | 146 (12.3) | 46 (11.6) | |
| Moderate | 32 (2.7) | 11 (2.8) | |
| Biopsy not done or missing | 222 (18.7) | 91 (23.0) | |
| Induction agents | 0.08 | ||
| Thymoglobulin | 703 (59.1) | 218 (55.1) | |
| Basiliximab | 202 (17.0) | 90 (22.7) | |
| Alemtuzumab | 188 (15.8) | 61 (15.4) | |
| No induction | 96 (8.1) | 27 (6.8) | |
| Transplant laterality | 0.64 | ||
| Left kidney | 602 (50.6) | 193 (48.7) | |
| Right kidney | 575 (48.4) | 201 (50.8) | |
| Pediatric enbloc kidney | 2 (0.2) | 0 (0.0) | |
| Dual kidney | 10 (0.8) | 2 (0.5) | |
3.3. Unadjusted Outcomes
Early and intermediate outcomes by AKI category are summarized in Table 3. DGF increased with AKI severity: 38.1% for Stage 2, 53.3% for Stage 3 without dialysis, and 54.5% for Stage 3 on dialysis (p < 0.001). Primary nonfunction within 90 days did not differ (2.0% Stage 2, 2.0% Stage 3 without dialysis, 2.3% Stage 3 on dialysis; p = 0.73).
TABLE 3.
Unadjusted outcomes of the study cohort based on AKI stage and dialysis need.
| AKI group | ||||
|---|---|---|---|---|
| Outcome variable |
Stage 2 N = 5810 |
Stage 3 no dialysis N = 4361 | Stage 3 on dialysis N = 1585 | p‐value |
| DGF | 2213 (38.1) | 2324 (53.3) | 864 (54.5) | <0.001 |
| PNF (death censored graft failure) within 90 days posttransplant | 114 (2.0) | 87 (2.0) | 36 (2.3) | 0.73 |
| Death censored graft failure at one‐year | 177 (3.0) | 137 (3.1) | 58 (3.7) | 0.46 |
| Mortality at one‐year | 238 (4.1) | 165 (3.8) | 33 (2.1) | 0.001 |
| All cause graft failure at one‐year (ACGF‐ graft failure and death) | 388 (6.7) | 279 (6.4) | 85 (5.4) | 0.17 |
| eGFR at one‐year, median [IQR] | 58 [45,73] | 60 [47, 75] | 62 [48, 77] | <0.001 |
| Composite outcome at one year (all cause graft failure and eGFR<30 at one year) | 757 (13.0) | 479 (11.0) | 161 (10.2) | 0.001 |
| Death censored graft failure at 3 years | 366 (6.3) | 249 (5.7) | 104 (6.6) | 0.34 |
| Mortality at 3 years | 609 (10.5) | 432 (9.9) | 116 (7.3) | 0.001 |
| All cause graft failure at 3 years | 895 (15.4) | 619 (14.2) | 203 (12.8) | 0.02 |
At one year, death‐censored graft failure was similar (3.0%, 3.1%, and 3.7%; p = 0.46). Mortality at one year was lowest in the Stage 3 on dialysis group (2.1% vs. 4.1% Stage 2 and 3.8% Stage 3 without dialysis; p = 0.001). The observed 1‐year mortality rate in the Stage 2 group is consistent with contemporary national posttransplant mortality rates for recipients with similar KDPI and EPTS distributions during this era, suggesting that the difference likely reflects underlying recipient risk rather than AKI severity alone. The combined endpoint of ACGF and eGFR less than 30 mL per min per 1.73 m2 at one year was lowest in Stage 3 on dialysis (10.2%) compared with Stage 3 without dialysis (11.0%) and Stage 2 (13.0%; p = 0.001). Median one‐year eGFR was modestly higher with increasing AKI category: 58 [45–73], 60 [47–75], and 62 [48–77] mL per min per 1.73 m2 (p < 0.001).
At three years, death‐censored graft failure remained similar across groups (6.3%, 5.7%, and 6.6%; p = 0.34). Mortality at three years was lowest for Stage 3 on dialysis (7.3% vs. 10.5% Stage 2 and 9.9% Stage 3 without dialysis; p = 0.001). ACGF at three years was 15.4% for Stage 2, 14.2% for Stage 3 without dialysis, and 12.8% for Stage 3 on dialysis (p = 0.02). Kaplan–Meier curves showed a small separation favoring Stage 3 on dialysis for ACGF and patient survival, with no separation for death‐censored graft failure (Figures S1, S2 and S3).
3.4. Multivariable and Weighted Analyses
Cox models using Stage 2 as reference are presented in Table 4. In unadjusted analysis, Stage 3 on dialysis had a lower hazard of ACGF (HR 0.85, 95% confidence interval [CI] 0.75–0.96, p = 0.01), while Stage 3 without dialysis did not differ (HR 0.95, 95% CI 0.88–1.02, p = 0.15). After IPW and in fully adjusted models, neither Stage 3 category was associated with ACGF compared with Stage 2 (IPW adjusted HR 0.92, 95% CI 0.81–1.05 for Stage 3 on dialysis; HR 0.97, 95% CI 0.90–1.04 for Stage 3 without dialysis). Results were unchanged after inclusion of KDPI and EPTS categories (Model 4). Balance diagnostics showed standardized differences near 0 and variance ratios near 1 after weighting (Table S2).
TABLE 4.
Unadjusted outcomes of the AKI Stage 3 on dialysis cohort by KDPI and EPTS category.
| K‐DIGO AKI stage 3 on dialysis cohort during study Period | ||||
|---|---|---|---|---|
| Outcome variable |
KDPI & EPTS ≤35% N = 408 |
KDPI & EPTS >35% N = 548 |
KDPI≤35% & EPTS>35 or KDPI>35 & EPTS≤35% N = 629 |
p‐value |
| DGF | 186 (45.6) | 313 (57.1) | 365 (58.0) | <0.001 |
| PNF (death censored graft failure) within 90 days posttransplant | 3 (0.7) | 15 (2.7) | 18 (2.9) | 0.05 |
| Death censored graft failure at one‐year | 7 (1.7) | 26 (4.7) | 25 (4.0) | 0.04 |
| Mortality at one‐year | 2 (0.5) | 19 (3.5) | 12 (1.9) | 0.01 |
| All cause graft failure at one‐year (ACGF‐ graft failure and death) | 9 (2.2) | 42 (7.7) | 34 (5.4) | 0.001 |
| eGFR at one‐year, median [IQR] | 69 [56, 84] | 54 [43, 70] | 63 [49, 77] | <0.001 |
| Composite outcome at one year (all cause graft failure and eGFR<30 at one year) | 22 (5.4) | 81 (14.8) | 58 (9.2) | <0.001 |
| Death censored graft failure at 3 years | 21 (5.2) | 43 (7.9) | 40 (6.4) | 0.24 |
| Mortality at 3 years | 7 (1.7) | 60 (11.0) | 49 (7.8) | <0.001 |
| All cause graft failure at 3 years | 27 (6.6) | 94 (17.2) | 82 (13.0) | <0.001 |
3.5. Stage 3 on Dialysis Donor Subgroup
Within the Stage 3 on dialysis cohort, donor recipient pairings with higher KDPI and higher EPTS were associated with worse outcomes (Table 5 and Table 4). Relative to the low risk reference (KDPI 35% or less and EPTS 35% or less), hazard of ACGF was higher when either KDPI or EPTS exceeded 35% (adjusted HR 1.79, 95% CI 1.27–2.51) and highest when both exceeded 35% (adjusted HR 2.79, 95% CI 1.95–3.99). Unadjusted one year eGFR was highest in the low risk pairing (median 69 [56–84] mL per min per 1.73 m2) versus intermediate risk 63 [49–77] and high risk 54 [43–70] mL per min per 1.73 m2 (p < 0.001). One year mortality and three year ACGF were also lowest in the low risk pairing (both p < 0.001). Detailed characteristics by graft outcome in this subgroup are in Table 2.
TABLE 5.
Cox proportional hazard model for predicting ACGF with and without inverse probability weightage by AKI staging and need for dialysis.
| The study cohort (N = 11 756) models (Reference group AKI Stage 2 category) | Hazard ratio | 95% confidence interval | p‐value |
|---|---|---|---|
| Model 1: Univariate cox analysis without IPW adjustment | |||
| AKI Stage 3 no dialysis | 0.95 | 0.88–1.02 | 0.15 |
| AKI Stage 3 on dialysis | 0.85 | 0.75–0.96 | 0.01 |
| Model 2: Univariate cox analysis adjusted to IPW* | |||
| AKI Stage 3 no dialysis | 0.97 | 0.90–1.04 | 0.40 |
| AKI Stage 3 on dialysis | 0.92 | 0.81–1.05 | 0.21 |
| Model 3: Multivariable cox ACGF model adjusted for IPW and other variables** not included in the IPW calculation | |||
| AKI Stage 3 no dialysis | 0.98 | 0.92–1.06 | 0.65 |
| AKI Stage 3 on dialysis | 0.92 | 0.81–1.05 | 0.21 |
| Model 4***: Model 3 and the variables including KDPI and EPTS categories | |||
| AKI Stage 3 no dialysis | 0.98 | 0.91–1.05 | 0.56 |
| AKI Stage 3 on dialysis | 0.96 | 0.84–1.09 | 0.50 |
| AKI Stage 3 on dialysis cohort (N = 1,585) (Reference group KDPI≤ 35% and EPTS≤ 35% category) | Hazard ratio | 95% confidence interval | p‐value |
| Model 1: Univariate Cox analysis without IPW* adjustment | |||
| Either KDPI or EPTS>35% | 1.88 | 1.38–2.58 | <0.001 |
| Both KDPI and EPTS >35% | 2.65 | 1.92–3.65 | <0.001 |
| Model 2: Univariate cox analysis with IPW* adjustment | |||
| Either KDPI or EPTS>35% | 1.84 | 1.32–2.57 | <0.001 |
| Both KDPI and EPTS >35% | 2.82 | 1.98–4.02 | <0.001 |
| Model 3: Multivariable cox ACGF model adjusted for IPW and other variables** not included in the IPW calculation | |||
| Either KDPI or EPTS>35% | 1.79 | 1.27–2.51 | <0.001 |
| Both KDPI and EPTS >35% | 2.79 | 1.95–3.99 | <0.001 |
| Model 4***: Model 3 and the variables including KDPI and EPTS categories | |||
| Either KDPI or EPTS>35% | 1.79 | 1.27–2.51 | <0.001 |
| Both KDPI and EPTS >35% | 2.79 | 1.95–3.99 | <0.001 |
*Model 2: The variables used for inverse probability weighting (IPW) calculation: donor age category, procurement biopsy glomerulosclerosis/ arteriosclerosis/ tubular atrophy interstitial fibrosis categories, machine perfusions status, donation after circulatory death status, donor distance to transplant center category, cold ischemia category, recipient age category, diabetes status, dialysis exposure duration, re‐transplant status, and recipient body mass index category, adjusted for transplant center cluster effect.
**Model 3: The variables including HLA mismatch category, CMV mismatch category, calculated PRA category, induction immunosuppression category, recipient race and gender category, adjusted for IPW and transplant center cluster effect.
***Model 4: The variables including Model 3 and the ones including the KDPI and EPTS categories.
3.6. Subgroup and Interaction Analyses
Adjusted subgroup analyses for ACGF are displayed in Figure 1. Across strata of KDPI, EPTS, donor urine output, DCD, and cold ischemia time, HRs for Stage 3 versus Stage 2 typically crossed 1.0 and showed no excess risk. In prespecified interaction models, AKI stage did not significantly modify the effect of KDPI or EPTS on ACGF (Table S4a). For cold ischemia time, an interaction consistent with attenuation of risk for Stage 3 on dialysis was observed at 24–36 h (interaction term p = 0.04), whereas estimates for greater than 36 h were imprecise and did not reach significance (Table S4b).
FIGURE 1.
Posttransplant HRs for ACGF in the subgroups of the study cohort by AKI stages and dialysis status.
3.7. Recipient Tolerability Analysis
Exploratory analyses were performed to evaluate whether recipient characteristics modified the relationship between donor AKI severity and early posttransplant outcomes. Among 11 708 recipients with complete data, the multivariable logistic regression model for the composite tolerability outcome (DGF, eGFR <30 mL/min/1.73 m2 at 6 months, graft failure, or death within 6 months) demonstrated that recipient factors, rather than donor AKI stage, were the primary drivers of early adverse events.
Older recipient age, higher BMI, and longer pre‐transplant dialysis duration were each associated with higher odds of the composite outcome (all p < 0.01). In contrast, donor AKI stage, including Stage 3 AKI with dialysis, was not independently associated with the composite outcome after adjustment for donor, recipient, and transplant factors. Donor urine output and dialysis status similarly demonstrated no significant association. Among continuous variables, KDPI, recipient BMI, cold ischemia time, and dialysis duration were the four strongest contributors to model fit in the ANOVA decomposition. The model discriminative performance was modest (c‐statistic 0.66).
Pre‐specified interaction terms (AKI × age, AKI × BMI, AKI × dialysis duration, AKI × preemptive listing, AKI × travel distance, and AKI × functional status) did not reveal significant effect modification. These findings indicate that recipient risk factors exert larger and more consistent influence on tolerability than donor AKI stage itself.
To validate these results and evaluate nonlinear patterns, multiple interpretable machine learning models were tested. After hyperparameter tuning, an XGBoost model performed best (cross‐validated c‐statistic 0.66). SHAP feature importance rankings closely mirrored logistic regression results, with KDPI, donor age, and recipient age emerging as the most influential variables. Donor AKI stage, including dialysis exposure, ranked low in importance across all models. No clinically meaningful donor AKI × recipient interactions emerged from SHAP dependence analyses. Taken together, these results suggest that early vulnerability to AKI‐affected kidneys is driven primarily by recipient frailty indicators (age, BMI, dialysis vintage) and overall organ quality (KDPI) rather than by AKI severity or donor dialysis itself.
3.8. Kaplan–Meier Survival
Unadjusted survival curves are shown in Figures S1–S3. For ACGF, the log rank p value was 0.01 with the best trajectory in Stage 3 on dialysis. Death censored graft failure curves overlapped (log rank p = 0.07). Patient survival favored Stage 3 on dialysis (log rank p = 0.003).
4. Discussion
This study shows that kidneys from donors with severe AKI, including KDIGO Stage 3 with dialysis, can achieve outcomes comparable to kidneys from donors with moderate AKI when organ quality and recipient risk are accounted for. After IPW and multivariable adjustment, donor AKI stage was not an independent predictor of ACGF, whereas Kidney Donor Profile Index and recipient estimated posttransplant survival carried greater prognostic weight. These findings support careful use of selected severely injured donor kidneys within evidence‐based matching strategies rather than categorical decline based solely on AKI stage.
Interpretation of these findings must be framed within the context of organ selection. Because this cohort includes only kidneys that proceeded to transplantation, our results reflect outcomes among AKI‐affected kidneys deemed acceptable for use by transplant centers rather than the broader population of donors with severe AKI. Donor age, KDPI, biopsy interpretation, hemodynamic trajectory, and center‐specific risk tolerance likely influenced which Stage 3 AKI kidneys were ultimately used. Accordingly, these findings should not be generalized to all AKI‐affected donor kidneys, including those declined or discarded. Rather, they demonstrate that among the selected subset of AKI kidneys deemed acceptable for transplantation, AKI severity and donor dialysis were not independent predictors of graft failure after accounting for organ quality and recipient risk. This reflects center‐level clinical judgment and survivorship of organs that successfully progressed to transplant.
In addition to donor‐level risk, our study incorporated a comprehensive assessment of recipient tolerability, informed by both logistic regression and interpretable machine learning methods. Across approaches, donor AKI stage, including Stage 3 with dialysis, did not significantly interact with key recipient factors such as age, BMI, dialysis duration, or preemptive listing status. Instead, early vulnerability was most strongly associated with recipient age, metabolic profile, and dialysis vintage, along with kidney quality as captured by KDPI. Machine learning models confirmed these findings, identifying KDPI, donor age, and recipient age as the primary drivers of risk, with donor AKI stage contributing minimally to outcome prediction. These results emphasize that the decision to transplant kidneys affected by severe AKI should be guided more by recipient fitness and organ quality than by AKI severity alone.
Our results align with contemporary reports that dialysis in the donor increases DGF without worsening adjusted long‐term outcomes [19, 20, 21, 22]. Although 1‐year mortality was numerically higher in the Stage 2 group, this rate aligns with expected national benchmarks for similar recipient risk profiles and appears driven by underlying recipient characteristics rather than donor AKI severity. Donors in the Stage 3 dialysis group were younger and had lower KDPI than those in other AKI categories, indicating that these organs were selectively accepted on the basis of favorable overall quality despite severe AKI. The favorable outcomes observed in this subgroup likely reflect this careful donor selection rather than an absence of biological risk associated with severe injury. These findings reinforce that organ quality metrics, particularly KDPI and donor age, remain central to posttransplant prognosis. We extend this literature by using a large, recent national cohort and by explicitly testing interactions that identify recipient profiles more likely to tolerate challenging kidneys. Younger recipients, preemptive or with shorter dialysis vintage, and those with lower BMI demonstrated attenuated early risk in exploratory analyses. This provides a practical allocation signal that can be operationalized alongside KDPI and EPTS.
Although this analysis focused on comparisons within moderate‐to‐severe AKI categories, prior national registry studies have demonstrated that transplanted kidneys from donors with AKI, including severe AKI, have long‐term graft survival comparable to kidneys without AKI after adjustment for KDPI and recipient factors [10, 11, 12, 13]. Our findings are consistent with this broader literature and further refine the evidence by suggesting that even within severe AKI, dialysis exposure does not independently worsen adjusted graft survival when organ quality and recipient risk are considered.
Methodologically, this study improves ascertainment of donor dialysis exposure by applying a transparent, rule‐based NLP approach with blinded manual validation. While this enhances exposure classification in registry data where dialysis is not a structured field, it does not eliminate uncertainty regarding the clinical heterogeneity of severe AKI. Differences in donor physiology, timing of injury, and response to dialysis remain incompletely captured. The higher proportion of DCD among dialysis donors in our cohort likely reflects increased DCD utilization during 2015–2020 and differences in ascertainment relative to earlier studies such as Wen et al.20 Together, these observations suggest that center experience and era effects influence which severely injured kidneys progress to transplant.
Several implications follow. First, allocation and acceptance should prioritize organ quality and recipient fitness over AKI stage alone. In practice, low‐KDPI Stage 3 kidneys, even with donor dialysis, appear suitable for lower‐risk recipients, particularly those with low EPTS. Second, programs should consider integrating structured dialysis and urine output fields into donor records to reduce information bias and facilitate consistent staging. Third, policy efforts aimed at improving consistency in acceptance practices could incorporate decision support that highlights favorable donor‐recipient pairings when kidneys present with severe AKI. Whether such strategies reduce discard requires analyses that include declined and discarded kidneys.
Limitations include the observational design, potential selection of transplanted organs, and incomplete granularity on donor clinical course. Despite targeted validation, dialysis exposure may be misclassified in some cases. Duration, modality (e.g., CRRT vs. intermittent hemodialysis), and cumulative intensity of donor dialysis were inconsistently documented in free‐text records and could not be reliably quantified. We therefore could not evaluate whether duration or modality of dialysis influenced posttransplant outcomes. Urine output was heterogeneous in timing and completeness, which is why it was used descriptively and in sensitivity analyses rather than for primary KDIGO staging. These factors introduce residual confounding and information bias. This study was intentionally designed as a posttransplant outcomes analysis and therefore includes only transplanted kidneys, introducing survivorship and selection bias. We could not evaluate outcomes among AKI‐affected kidneys that were declined or discarded. As such, the findings apply to the subset of AKI kidneys accepted for transplantation rather than to all severe AKI donors.
Future work should prospectively capture standardized donor dialysis and urine output data, adjudicate AKI etiology and trajectory, and test allocation tools that use KDPI, EPTS, and simple recipient features to recommend match quality. A testable hypothesis is that a strategy directing low‐KDPI Stage 3 kidneys to recipients with low EPTS, younger age, lower BMI, and shorter dialysis vintage will reduce discard without increasing early complications. Pragmatic trials or policy pilots could evaluate this approach. Finally, external validation in systems with older donor populations would clarify generalizability and help define whether similar tolerability profiles hold across different donor ages and case mixes [18, 19, 20, 21, 22, 23].
5. Conclusion
This study demonstrates that among transplanted kidneys, those from donors with severe AKI, including those who required dialysis, can achieve favorable long‐term outcomes when organ quality is preserved and recipients are carefully selected. These findings inform ongoing debate regarding the use of kidneys affected by severe AKI and highlight their potential to expand the donor pool without evidence of increased adjusted graft failure among transplanted kidneys.
By refining donor and recipient matching, such as allocating low‐KDPI kidneys with severe AKI to younger or lower‐EPTS recipients, transplant programs may support more consistent acceptance of appropriately selected AKI kidneys. Whether such approaches reduce discard will require analyses that include declined and discarded kidneys. This supports a more individualized and evidence‐based approach to organ acceptance rather than a categorical exclusion of AKI donors.
Unanswered questions remain about the biological mechanisms that enable recovery of severely injured kidneys and whether targeted perioperative strategies, such as ischemia mitigation or tailored immunosuppression, could further improve results. Addressing these through future prospective studies, along with standardized reporting of donor AKI severity and dialysis exposure, will be essential. In the meantime, these findings support a thoughtful shift toward greater acceptance of kidneys from donors with severe AKI, guided by structured risk assessment and deliberate recipient pairing. Such an approach can expand the donor pool, reduce discard of viable kidneys, and ultimately improve access and equity in kidney transplantation.
Author Contributions
Concept and design: Amber B. Paulus, Gaurav Gupta, Bekir Tanriover, Data acquisition: Alice Toll, Alexandra Lewis, Carlos Martinez, Cassandra McCharen, Wida Cherikh, Data analysis and interpretation: Amber B. Paulus, Ahmet Gungor, Alice Toll, Alexandra Lewis, Carlos Martinez, Cassandra McCharen, Wida Cherikh, Gaurav Gupta, Bekir Tanriover, Drafting of manuscript: Amber B. Paulus, Ahmet Gungor, Gaurav Gupta, Bekir Tanriover, Critical revision of manuscript for important intellectual content: all authors, Statistical analysis: Alice Toll, Alexandra Lewis, Carlos Martinez, Cassandra McCharen, Wida Cherikh, Amber B. Paulus, Funding secured by: Gaurav Gupta, Administrative, technical, or material support: Gaurav Gupta, Bekir Tanriover, Final approval of article: all authors
Funding
This work was supported intramurally by the VCU division of nephrology. UNOS contributed analytic support but had no role in the decision to submit the manuscript for publication.
Conflicts of Interest
The authors declare no conflicts of interest.
Disclosures
Gaurav Gupta has received honoraria from Alexion, has served on the Scientific Advisory Board for Apellis, CareDx, Natera, Novartis, TGI‐Eurofins, Thermo Fisher and has received research funding from Merck. The other authors declare no conflicts of interest.
Supporting information
Supplemental Table S1: Study cohort selection, deceased donor kidneys between 2015 and 2020 in the United States by Kidney‐Disease Improving Global Outcomes (K‐DIGO) AKI classification.
Supplemental Table S2. Standardized differences and variance ratio before and after IPW adjustments.
Supplemental Table S3. Cox proportional hazard model for predicting all‐cause graft failure with inverse probability weightage by AKI staging and need for dialysis, Full Model (Model 4).
Supplemental Table S4a. Cox proportional hazard model for predicting all‐cause graft failure with inverse probability weightage by AKI staging, the interaction models between AKI stage and KDPI categories.
Supplemental Table S4b. Cox proportional hazard model for predicting all‐cause graft failure with inverse probability weightage by AKI staging, the interaction models between AKI stage and cold‐ischemia time (CIT) categories.
Supplemental Figure S1. Kaplan Meier all‐cause graft failure estimates for deceased donor kidney transplants with acute kidney injury (AKI) between 2015‐2020 in the United States.
Abbreviations: ACGF = All cause graft failure; AKI = acute kidney injury.
Supplemental Figure S2. Kaplan Meier death censored graft failure estimates for deceased donor kidney transplants with acute kidney injury (AKI) between 2015‐2020 in the United States.
Abbreviations: DCGF = Death censored graft failure; AKI = acute kidney injury.
Supplemental Figure S3. Kaplan Meier patient survival estimates for deceased donor kidney transplants with acute kidney injury (AKI) between 2015‐2020 in the United States.
Acknowledgments
The data reported here have been supplied by UNOS as the contractor for the Organ Procurement and Transplantation Network (OPTN). The interpretation and reporting of these data are the responsibility of the author(s) and in no way should be seen as an official policy of or interpretation by the OPTN or the U.S. Government.
The authors acknowledge the collaborators within the KiTE (Kidney disease and Transplant Epidemiology) Research Consortium for their invaluable support and contributions throughout the completion of this research.
Paulus A. B., Gupta G., Gungor A., et al. “Outcomes of Kidney Transplants From Donors With Stage 3 Acute Kidney Injury Requiring Dialysis: A National Cohort Study.” Clinical Transplantation 40, no. 4 (2026): e70535. 10.1111/ctr.70535
Social Media Post (≤200 characters): Among transplanted kidneys, severe donor AKI (even with dialysis) was not independently associated with worse graft survival after accounting for KDPI and EPTS. #KidneyTransplant #AKI
References
- 1. Tonelli M., Wiebe N., Knoll G., et al., “Systematic Review: Kidney Transplantation Compared With Dialysis in Clinically Relevant Outcomes,” American Journal of Transplantation 11, no. 10 (2011): 2093–2109, 10.1111/j.1600-6143.2011.03686.x. [DOI] [PubMed] [Google Scholar]
- 2. Kaballo M. A., Canney M., O'Kelly P., Williams Y., O'Seaghdha C. M., and Conlon P. J., “A Comparative Analysis of Survival of Patients on Dialysis and After Kidney Transplantation,” Clinical Kidney Journal 11, no. 3 (2017): 389–393, 10.1093/ckj/sfx117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Shi B., Ying T., and Chadban S. J., “Survival After Kidney Transplantation Compared With Ongoing Dialysis for People Over 70 Years of Age: A Matched‐Pair Analysis,” American Journal of Transplantation 23, no. 10 (2023): 1551–1560, 10.1016/j.ajt.2023.07.006. [DOI] [PubMed] [Google Scholar]
- 4. Strohmaier S., Wallisch C., Kammer M., et al., “Survival Benefit of First Single‐Organ Deceased Donor Kidney Transplantation Compared With Long‐Term Dialysis Across Ages in Transplant‐Eligible Patients With Kidney Failure,” JAMA Network Open 5, no. 10 (2022): e2234971, 10.1001/jamanetworkopen.2022.34971. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. McCormick F., Held P. J., Chertow G. M., Peters T. G., and Roberts J. P., “Reducing the Shortage of Transplant Kidneys: A Lost Opportunity for the US Health Resources and Services Administration (HRSA),” American Journal of Kidney Diseases 77, no. 6 (2021): 963–966, 10.1053/j.ajkd.2020.11.030. [DOI] [PubMed] [Google Scholar]
- 6. Wang J. H. and Hart A., “Global Perspective on Kidney Transplantation: United States,” Kidney 2, no. 11 (2021): 1836–1839, 10.34067/KID.0002472021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Singh S. K. and Kim S. J., “Epidemiology of Kidney Discard From Expanded Criteria Donors Undergoing Donation After Circulatory Death,” Clincal Journal of the American Society of Nephrology 11, no. 2 (2015): 317–323, 10.2215/CJN.07190715. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Wang C. J., Wetmore J. B., and Israni A. K., “Old Versus New: Progress in Reaching the Goals of the New Kidney Allocation System,” Human Immunology 78 (2017): 9–15, 10.1016/j.humimm.2016.08.007. [DOI] [PubMed] [Google Scholar]
- 9. Zens T. J., Danobeitia J. S., Leverson G., et al., “The Impact of Kidney Donor Profile Index on Delayed Graft Function and Transplant Outcomes: A Single‐Center Analysis,” Clinical Transplantation 32 (2018): e13190, 10.1111/ctr.13190. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Boffa C., Van De Leemkolk F., Curnow E., et al., “Transplantation of Kidneys From Donors With Acute Kidney Injury: Friend or Foe?,” American Journal of Transplantation 17 (2016): 411–419, 10.1111/ajt.14001. [DOI] [PubMed] [Google Scholar]
- 11. Hall I. E., Schröppel B., Doshi M. D., et al., “Associations of Deceased Donor Kidney Injury With Kidney Discard and Function After Transplantation,” American Journal of Transplantation 15 (2015): 1623–1631, 10.1111/ajt.13150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Hall I. E., Akalin E., Bromberg J. S., et al., “Deceased‐Donor Acute Kidney Injury is Not Associated With Kidney Allograft Failure,” Kidney International 95 (2019): 199–209, 10.1016/j.kint.2018.08.032. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Heilman R. L., Smith M. L., Smith B. H., et al., “Long‐Term Outcomes Following Kidney Transplantation From Donors With Acute Kidney Injury,” Transplantation 103 (2019): e263–e272, 10.1097/TP.0000000000002634. [DOI] [PubMed] [Google Scholar]
- 14. Mannon R. B., “Delayed Graft Function: The AKI of Kidney Transplantation,” Nephron 140 (2018): 94–98, 10.1159/000491556. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Bahl D., Haddad Z., Datoo A., and Qazi Y. A., “Delayed Graft Function in Kidney Transplantation,” Current Opinion in Organ Transplantation 24 (2019): 82–86, 10.1097/MOT.0000000000000592. [DOI] [PubMed] [Google Scholar]
- 16. Palmisano A., Gandolfini I., Delsante M., et al., “Acute Kidney Injury (AKI) Before and After Kidney Transplantation: Causes, Medical Approach, and Implications for the Long‐Term Outcomes,” Journal of Clinical Medicine 10, no. 7 (2021): 1484, 10.3390/jcm10071484, Published April 2, 2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Post Hospers G., Visser W. J., Verhoeven J., et al., “Delayed Graft Function After Kidney Transplantation: The Role of Residual Diuresis and Waste Products, as Oxalic Acid and Its Precursors,” Transplant International 37 (2024): 13218, 10.3389/ti.2024.13218, Published 2024 Jul 19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. United States Renal Data System 2024 USRDS Annual Data Report: Epidemiology of Kidney Disease in the United States (National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, 2024). [Google Scholar]
- 19. Kaufman D. M., Perkins J. D., Leca N., Bakthavatsalam R., and Sibulesky L., “Tailoring Kidney Transplant Strategies: Unraveling the Impact of Acute Kidney Injury and Preservation Methods Across Different Strata of Kidney Risk,” Clinical Transplantation (2024), Published online December 24. [DOI] [PubMed] [Google Scholar]
- 20. Wen Y., Mansour S. G., Srialluri N., et al., “Kidney Transplant Outcomes From Deceased Donors Who Received Dialysis,” JAMA 332, no. 3 (2024): 215–225, 10.1001/jama.2024.8469. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Liu C., Hall I. E., Mansour S., Thiessen Philbrook H. R., Jia Y., and Parikh C. R., “Association of Deceased Donor Acute Kidney Injury With Recipient Graft Survival,” JAMA Network Open 3, no. 1 (2020): e1918634, 10.1001/jamanetworkopen.2019.18634. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Hall I. E., Akalin E., Bromberg J. S., et al., “Deceased‐Donor Acute Kidney Injury is Not Associated With Kidney Allograft Failure,” Kidney International 95, no. 1 (2019): 199–209, 10.1016/j.kint.2018.08.047. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Sonnenberg E. M., Hsu J. Y., Cohen J. B., et al., “Acute Kidney Injury in Deceased Organ Donors and Kidney Transplant Outcomes: A National Cohort Study Using a Novel Data Source,” Annals of Surgery 276, no. 6 (2022): e982–e990, 10.1097/SLA.0000000000004597. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental Table S1: Study cohort selection, deceased donor kidneys between 2015 and 2020 in the United States by Kidney‐Disease Improving Global Outcomes (K‐DIGO) AKI classification.
Supplemental Table S2. Standardized differences and variance ratio before and after IPW adjustments.
Supplemental Table S3. Cox proportional hazard model for predicting all‐cause graft failure with inverse probability weightage by AKI staging and need for dialysis, Full Model (Model 4).
Supplemental Table S4a. Cox proportional hazard model for predicting all‐cause graft failure with inverse probability weightage by AKI staging, the interaction models between AKI stage and KDPI categories.
Supplemental Table S4b. Cox proportional hazard model for predicting all‐cause graft failure with inverse probability weightage by AKI staging, the interaction models between AKI stage and cold‐ischemia time (CIT) categories.
Supplemental Figure S1. Kaplan Meier all‐cause graft failure estimates for deceased donor kidney transplants with acute kidney injury (AKI) between 2015‐2020 in the United States.
Abbreviations: ACGF = All cause graft failure; AKI = acute kidney injury.
Supplemental Figure S2. Kaplan Meier death censored graft failure estimates for deceased donor kidney transplants with acute kidney injury (AKI) between 2015‐2020 in the United States.
Abbreviations: DCGF = Death censored graft failure; AKI = acute kidney injury.
Supplemental Figure S3. Kaplan Meier patient survival estimates for deceased donor kidney transplants with acute kidney injury (AKI) between 2015‐2020 in the United States.