Rethinking procurement biopsy in donors with normal renal function: A barrier to kidney utilization without survival benefit

Abstract

BACKGROUND

Procurement biopsy remains widely used in deceased donor kidney transplantation, yet its value is uncertain in kidneys with preserved renal function. Concerns persist regarding sampling error, interpretive variability, and the potential for biopsy findings to contribute directly to organ discard. Whether procurement biopsy improves decision-making or graft outcomes in donors with normal renal function is not known.

AIM

To investigate whether biopsy adds clinical value or causes delay, misclassification without improving outcomes in donors with preserved renal function.

METHODS

Using national Organ Procurement and Transplantation Network data (2014-2024), we evaluated deceased donor kidneys with terminal epidermal growth factor receptor ≥ 60. Donor, recipient, and transplant variables were linked at the center level to reflect clinical decision pathways. Propensity-score matching (1:1 nearest neighbor, caliper 0.2 SD) generated balanced biopsy and non-biopsy cohorts. Center-clustered Cox models estimated the association between biopsy and death-censored graft failure. Mediation through cold ischemia time and delayed graft function was examined using sequential attenuation, with robustness assessed via E-values, restricted mean survival time, and decision curve analysis.

RESULTS

Of 158365 recovered kidneys, 46.6% underwent biopsy and were over six times more likely to be discarded (28.2% vs 4.4%), including 5.9% discarded solely due to biopsy findings, more than half with only mild abnormalities. The matched cohort comprised 52094 transplants with excellent covariate balance. Center-clustered Cox models showed biopsy was associated with higher graft-failure risk (hazard ratio = 1.12, 95% confidence interval: 1.05-1.19). Biopsy prolonged cold ischemia time by 42 minutes, mediating approximately half of its adverse survival association. Restricted mean survival time demonstrated a small but significant absolute reduction in five-year survival (-0.03 years). Decision curve analysis showed no meaningful improvement in clinical benefit from biopsy beyond standard clinical variables.

CONCLUSION

In deceased donors with normal renal function, procurement biopsy adds no prognostic value, increases discard and ischemic delay, and may hinder optimal utilization of otherwise transplantable kidneys.

Key Words: Kidney transplant; Procurement biopsy; Death-censored graft survival; Donor kidney discard; Renal function

Core Tip: Using national Organ Procurement and Transplantation Network data, we show that procurement biopsy in deceased donors with preserved renal function does not improve prognostic accuracy or graft survival. Instead, biopsy is associated with substantially higher discard rates and prolonged cold ischemia, with much of its adverse impact mediated through ischemic delay and early graft dysfunction. Advanced analytic approaches, including mediation, restricted mean survival time, and decision-curve analyses, demonstrate no added clinical benefit beyond standard donor and recipient characteristics. These findings challenge routine procurement biopsy in this donor population and support re-evaluating biopsy-driven acceptance practices to improve kidney utilization.

INTRODUCTION

Kidney transplantation remains the optimal therapy for end-stage kidney disease, yet persistent organ shortage continues to limit access, with thousands of waitlisted candidates dying annually. Despite this scarcity, approximately 20%-30% of deceased donor kidneys recovered in the United States are not transplanted, now termed “non-utilized”, and procurement biopsy is frequently cited as a contributing factor[1].

Procurement biopsy is often requested by transplant centers to inform acceptance decisions, even when donors exhibit otherwise favorable clinical characteristics. However, its value remains controversial. Substantial sampling error, interobserver variability, and limited reproducibility have raised concerns that biopsies may misclassify clinically acceptable kidneys and contribute to unnecessary discard[2-4]. Although several histologic scoring systems have been proposed to standardize interpretation, these tools remain debated, and their incremental predictive value beyond established donor risk indices is modest at best[5,6].

Prior national studies have shown that biopsy use is highly variable across the United States and often increases the likelihood of discard. Lentine et al[7] reported that more than half of recovered kidneys from 2014-2018 underwent biopsy, with Organ Procurement Organization (OPO)-level biopsy frequency ranging widely from 22.8% to 77.5%. After adjustment for donor risk, including Kidney Donor Profile Index (KDPI), biopsied kidneys had over threefold higher odds of discard, including among clinically low-risk donors. Wang et al[8] demonstrated that higher glomerulosclerosis (GS) percentages were associated with reduced utilization and inferior post-transplant outcomes, but GS percentage offered only marginal improvement over existing donor risk models.

A critical limitation of prior work, however, is that it aggregates donors across the full spectrum of kidney quality. As a result, biopsy findings in marginal or high-KDPI donors often drive conclusions, leaving unanswered the central question faced daily by transplant clinicians and OPOs: Does procurement biopsy provide meaningful clinical value in donors who already have preserved renal function? In this group, arguably the donors for whom biopsy should be least necessary, any operational delay, misclassification, or downstream shift in recipient selection may incur harm without corresponding benefit.

To address this gap, we conducted a contemporary national analysis of deceased donor kidneys with terminal epidermal growth factor receptor (eGFR) ≥ 60 mL/minute/1.73 m² from 2014-2024. By restricting our cohort to donors with preserved function and evaluating biopsy practice in this clinically favorable population, we examine an area of particular operational and policy relevance and limit the confounding introduced when lower-quality kidneys dominate biopsy patterns. Using propensity-matched analyses and a harm-augmented decision-curve framework, we evaluated biopsy frequency, discard, histologic findings, cold ischemia time (CIT), delayed graft function (DGF), and graft survival. Our objective was not to infer the causal effect of biopsy, but rather to determine whether biopsy meaningfully adds clinical value, or whether, in donors with preserved renal function, the practice introduces avoidable delays and misclassification without improving post-transplant outcomes.

MATERIALS AND METHODS

Donor and recipient populations and study source data

Two Organ Procurement and Transplantation Network (OPTN) datasets were analyzed. The deceased donor dataset included all United States deceased donors from December 4, 2014, to December 31, 2024, from whom at least one kidney was recovered for transplantation and who’s terminal eGFR was ≥ 60 mL/minute/1.73 m². The dataset was converted into a kidney-level file including kidneys offered as single or dual/en bloc grafts. Donor variables included biopsy status (yes/no), laterality (left, right, dual/en bloc), age, sex at birth, ABO blood type, height, weight, history of hypertension and diabetes mellitus, donation after circulatory death (DCD) status, cause of death (cerebrovascular accident vs other), terminal creatinine, Kidney Donor Risk Index (KDRI), and KDPI. Terminal eGFR was calculated using the chronic kidney disease epidemiology collaboration equation[9].

Kidney disposition (transplanted vs non-utilized) was recorded. Kidneys transplanted as kidney-only or as part of a multiorgan transplant were classified as transplanted. For non-utilized kidneys, OPTN-coded reasons were categorized as anatomy, biopsy, organ function, medical/social history, infection, prolonged preservation, other, or no recipient identified.

Biopsy findings included interstitial fibrosis/tubular atrophy (IFTA) and vascular changes (both categorized as absent-mild vs moderate-severe). Percent GS was recorded as continuous or categorized per OPTN into 0%-5%, 6%-10%, 11%-15%, 16%-20%, and > 20%. Hyalinosis was not analyzed due to substantial missingness. Based on preliminary Cox models showing similar hazards across higher GS strata and small numbers with GS > 10%, GS was dichotomized into 0%-5% vs ≥ 6% for analysis.

The second dataset included all United States recipients of deceased donor kidney-only transplants during the same period, with follow-up to March 31, 2025. Recipients of living donor or multiorgan transplants were excluded. The primary outcome was death-censored graft survival. Secondary outcomes included DGF, defined as dialysis in the first week, and recipient serum creatinine at 6 months and 12 months.

Recipient variables included age, sex at birth, height, weight, primary renal diagnosis, terminal calculated panel reactive antibody (cPRA), peripheral vascular disease, dialysis duration (including preemptive status), waiting time, and Estimated Post-Transplant Survival (EPTS) score[10]. Transplant-related variables included human leukocyte antigen mismatch, use of machine perfusion, cold ischemia time (CIT), distance between donor hospital and transplant center, and sharing status (local, regional, national). OPTN data were obtained from files released April 11, 2025. As these data are de-identified and publicly available, the Division of Human Subjects, University of Washington classified this study as exempt from human subjects’ review.

Statistical analysis

Continuous variables are reported as medians with interquartile ranges and categorical variables as n (%). Comparisons used the Kruskal-Wallis test or χ²/Fisher’s exact test as appropriate.

Missing donor data were minimal. Height (< 0.1%) was imputed using the median. Weight (n = 365) and CIT (n = 545) were imputed using linear regression models with clinically relevant predictors. History of donor hypertension and diabetes mellitus were missing in < 0.5% and imputed using multiple imputation. Recipient height and weight (< 15 cases total) were imputed using the median; missing cPRA values (1.5%) were imputed with the cohort median.

Propensity score matching

To reduce confounding between biopsy and non-biopsy groups, we developed a robust propensity score model incorporating donor, recipient, and transplant characteristics. This approach reflects real-world allocation patterns in which transplant centers frequently direct biopsied kidneys to older or higher-risk recipients and adjust acceptance behavior based on donor, recipient, and logistical factors.

The donor-side covariates included age, sex, eGFR, hypertension, diabetes, DCD status, cause of death, and KDRI. Recipient-side covariates included age, sex, primary diagnosis, peripheral vascular disease, cPRA, dialysis duration, and EPTS. Transplant/Logistical factors included transplant year, human leukocyte antigen mismatch, use of machine perfusion, transport distance, and sharing level. Including these covariates avoids omitting clinically important confounders that influence both biopsy use and subsequent allocation decisions.

Propensity scores were estimated using logistic regression. Matching was performed 1:1 using nearest-neighbor matching with a caliper width of 0.2 SD of the logit of the propensity score, without replacement. Covariate balance was assessed using standardized mean differences (SMDs), with SMD < 0.2 indicating adequate balance. Because of the large sample size, SMDs were favored over P-values as size-independent measures of balance.

Cox proportional hazards model on the matched cohort

The matched cohort was analyzed using Cox proportional hazards models clustered by transplant center, not OPO. Transplant centers mainly direct biopsy practice, request biopsy type, interpret pathology findings, and ultimately determine acceptance or discard. Therefore, center-level clustering correctly accounts for the decision-making unit that governs both biopsy use and organ acceptance.

Models adjusted for donor, recipient, and transplant covariates. Multicollinearity was assessed using variance inflation factors (> 10), and collinear variables were removed. Schoenfeld residuals were examined to verify proportional hazards, with minor violations retained to preserve adjustment. Cluster-robust standard errors accounted for within-center correlation.

Linear mixed-effects regression for CIT

To evaluate whether biopsy contributed to prolonged CIT, we constructed a weighted linear mixed-effects model with CIT as the dependent variable. Biopsy status was the primary fixed effect. Donor variables (age, sex, hypertension, diabetes, eGFR, DCD status, cause of death, height, weight), logistical factors (distance, sharing), and transplant center (random intercept) were included. Center-level modeling reflects heterogeneity in operational processes surrounding biopsy and organ acceptance.

Mixed-effects logistic regression for DGF

We analyzed DGF using a weighted mixed-effects logistic regression model. CIT was the primary predictor, with adjustment for donor, recipient, and transplant variables and a random intercept for transplant center to account for clustering.

Sequential Cox attenuation analysis

To explore whether the association between biopsy and graft failure was mediated through ischemic delay and early graft dysfunction, we constructed three nested, center-clustered Cox models: Model A: Biopsy + all covariates except CIT; model B: Model A + CIT; model C: Model B + DGF.

DGF occurs in a fixed, early postoperative window and was modeled as an early event influencing later graft survival, a widely accepted approach in transplant research. The percent change in the biopsy coefficient quantified the proportion of the association potentially explained by CIT and DGF.

E-value analysis

To evaluate the robustness of the association between biopsy and graft failure to unmeasured confounding, we calculated the E-value[11]. This represents the minimum strength of association an unmeasured confounder would need to have with both biopsy and graft failure to fully explain the observed hazard ratio (HR).

Restricted mean survival time

Restricted mean survival time RMST to 5 years (τ = 5) was estimated to complement HRs and avoid reliance on proportional hazards assumptions[12]. Differences in RMST were calculated using cluster bootstrap resampling by transplant center.

Decision curve analysis

We performed decision curve analysis (DCA)[13-15] to assess whether biopsy or pathology findings improved clinical decision-making beyond standard donor and recipient characteristics. Net benefit was computed for three models: (1) Clinical variables alone; (2) Clinical + biopsy status; and (3) Clinical + biopsy-pathology findings. Inverse-probability-of-censoring weighting was applied to account for variable follow-up. Overlapping curves indicated limited added clinical value. DCA was applied to assess whether biopsy provides clinically meaningful benefit by changing organ acceptance decisions, not merely by improving statistical model performance. All analyses were conducted using JMP Pro 17.0 and R 4.4.0. Propensity matching was performed using the MatchIt package[16].

RESULTS

Patterns of biopsy use and organ discard

Among 158365 deceased donor kidneys with terminal eGFR ≥ 60 mL/minute/1.73 m² recovered for transplantation, 84589 (53.4%) were not biopsied and 73776 (46.6%) underwent procurement biopsy (Table 1). As expected, based on contemporary practice, kidneys selected for biopsy came from donors with older age, markedly higher prevalence of hypertension and diabetes, greater DCD proportion, and more frequent cerebrovascular death. Consequently, KDRI and KDPI were substantially higher in the biopsied group. Transplant rates differed sharply: 95.6% of non-biopsied kidneys were transplanted vs 71.8% of biopsied kidneys.

Table 1 Demographic data for deceased donor kidneys recovered for transplantation with normal renal function, n (%)/median (interquartile rage).

Characteristics	Not biopsied (n = 84589)	Biopsied (n = 73776)	P value1
Kidney			< 0.001
Dual/en bloc	2266 (2.7)	1014 (1.4)
Left	41450 (49.0)	36067 (48.9)
Right	40873 (48.3)	36695 (49.7)
Age, years	30 (21-40)	53 (43-60)	< 0.001
Male	55255 (65.3)	43675 (59.2)	< 0.001
Blood type			< 0.001
A	30963 (36.6)	28175 (38.2)
AB	2901 (3.4)	2680 (3.6)
B	9576 (11.2)	8326 (11.3)
O	41144 (48.6)	34591 (46.9)
Height, cm	173 (163-180)	170 (163-178)	< 0.001
Weight, kg	75.7 (63.5-89.9)	82.0 (69.0-98.6)	< 0.001
History of hypertension	10295 (12.2)	38077 (51.6)	< 0.001
Diabetes mellitus	1709 (2.0)	13900 (18.8)	< 0.001
DCD	18677 (22.1)	33835 (45.9)	< 0.001
COD: CVA/stroke	11818 (14.0)	26300 (35.7)	< 0.001
Creatinine terminal	0.8 (0.6-1.0)	0.8 (0.6-1.0)	0.16
eGFR terminal, mL/minute/1.73 m2	115.7 (93.8-130.6)	100 (80.0-111.9)	< 0.001
KDRI	1.00 (0.88-1.17)	1.44 (1.21-1.71)	< 0.001
KDPI, %	20 (9-38)	62 (42-79)	< 0.001
Disposition			< 0.001
Transplanted	80876 (95.6)	52952 (71.8)
Discarded	3713 (4.4)	20824 (28.2)

¹Categorical variables were compared using χ² or Fisher’s exact test, as appropriate. Continuous variables were compared using the Wilcoxon rank-sum test.

DCD: Donor after circulatory death; COD: Cause of death; CVA: Cerebrovascular accident; eGFR: Epidermal growth factor receptor; KDRI: Kidney Donor Risk Index; KDPI: Kidney Donor Profile Index.

Discard reasons are shown in Table 2. Discard was significantly more common among biopsied kidneys (28.2%) compared with non-biopsied kidneys (4.4%). “No recipient found” remained the leading reason in both groups; however, in the biopsied group, biopsy findings themselves accounted for 5.9% of all kidneys (n = 4371), directly contributing to discard despite preserved donor renal function.

Table 2 Nonuse reasons for deceased donor kidneys with normal kidney function recovered for transplant, n (%).

	Not biopsied (n = 84589)	Biopsied (n = 73776)	P value1
Nonuse	3713 (4.4)	20824 (28.2)	< 0.001
Nonuse reason
Anatomy	870 (1.0)	1086 (1.5)	< 0.001
Biopsy	0 (0)	4371 (5.9)	< 0.001
Function	187 (0.2)	603 (0.8)	< 0.001
Medical social history	64 (0.1)	344 (0.5)	< 0.001
Infection	240 (0.3)	384 (0.5)	< 0.001
Too long preservation	74 (0.1)	263 (0.4)	< 0.001
Other	984 (1.2)	2340 (3.2)	< 0.001
No recipient found	1294 (1.5)	11433 (15.5)	< 0.001

¹Compared using χ² or Fisher’s exact test, as appropriate.

Pathology results are summarized in Table 3. Among kidneys not used due to biopsy findings, 53.7% had minimal abnormalities, defined as 0%-5% GS, absent-to-mild vascular changes, and absent-to-mild IFTA, highlighting substantial misclassification of acceptable kidneys as unsuitable based on histology alone.

Table 3 Pathology of kidney biopsied and pathology reason for discard, n (%).

Pathology results of biopsy		Non used for biopsy reason
Total biopsies	n = 73776	4371 (5.9)
Combination of lesions
0%-5% GS/absent-mild vascular/absent-mild IFTA	60251 (81.7)	2346 (53.7)
0%-5% GS/moderate-severe IFTA	1979 (2.7)	357 (8.2)
0%-5% GS/moderate-severe vascular	3049 (4.1)	436 (10.0)
0%-5% GS /moderate-severe vascular/moderate-severe IFTA	1320 (1.8)	451 (10.3)
≥ 6% GS/absent-mild vascular/absent-mild IFTA	5771 (7.8)	507 (11.6)
≥ 6% GS /moderate-severe IFTA	360 (0.5)	77 (1.8)
≥ 6% GS /moderate-severe vascular	728 (1.0)	114 (2.6)
≥ 6% GS/moderate-severe vascular/moderate-severe IFTA	318 (0.4)	83 (1.9)

GS: Glomerulosclerosis; IFTA: Interstitial fibrosis/tubular atrophy.

Clinical impact of biopsy on transplant outcomes: Baseline characteristics and propensity matching

Before matching, biopsied and non-biopsied kidneys exhibited major differences across donor, recipient, and transplant variables (Table 4). Donors of biopsied kidneys were older, more likely to have hypertension, diabetes, and DCD procurement, and had higher KDRI. Recipients of biopsied kidneys were older, had more comorbidities, longer wait times, and higher EPTS scores. These differences reflect real-world allocation patterns where centers frequently direct biopsied kidneys to older or higher-risk recipients. Robust propensity score matching produced two balanced groups (n = 26047 each) with SMDs < 0.02 across all key covariates (Supplementary Table 1).

Table 4 Comparison of donor and recipient characteristics between biopsied and non-biopsied kidneys, n (%)/median (interquartile rage).

Characteristics	Not biopsied (n = 66057)	Biopsied (n = 51450)	P value1
Donor
Kidney			< 0.001
Dual/en bloc	1846 (2.8)	632 (1.2)
Left	29520 (44.7)	25037 (48.7)
Right	34691 (52.5)	25781 (50.1)
Age, years	31 (21-41)	51 (41-58)	< 0.001
Male	42927 (65.0)	31246 (60.7)	< 0.001
Blood type			< 0.001
A	24178 (36.6)	19457 (37.8)
AB	2437 (3.7)	1701 (3.3)
B	7437 (11.3)	5899 (11.5)
O	32005 (48.5)	24393 (47.4)
Height, cm	172 (163-179)	170 (164-178)	< 0.001
Weight, kg	76.5 (63.5-90.6)	83.0 (70-99.2)	< 0.001
History of hypertension	8080 (12.2)	23896 (46.5)	< 0.001
Diabetes mellitus	1277 (1.9)	7922 (15.4)	< 0.001
DCD	16099 (24.4)	22600 (43.9)	< 0.001
COD: CVA/stroke	9303 (14.1)	16890 (32.8)	< 0.001
eGFR terminal	115.6 (93.9-130.6)	102.3 (82.2-114.0)	< 0.001
KDRI	1.01 (0.89-1.17)	1.37 (1.16-1.59)	< 0.001
Recipient
Age, years	47 (35-59)	59 (50-66)	< 0.001
Male	37986 (57.5)	31525 (61.3)	< 0.001
Height, cm	168 (160-178)	170 (163-178)	< 0.001
Weight, kg	78.0 (64.0-93.3)	81.6 (69.0-95.3)	< 0.001
Diagnosis			< 0.001
Autoimmune/systemic	3122 (4.7)	1503 (2.9)
Cystic/genetic	5311 (8.0)	3903 (7.6)
Diabetes mellitus	13250 (20.1%	18007 (35.0)
Glomerular	10890 (16.5)	5992 (11.7)
Other	9943 (15.1)	5227 (10.2)
Retransplant	9595 (14.5)	5185 (10.1)
Vascular/hypertensive	13946 (21.1)	11633 (22.6)
cPRA	2.3 (0-72.3)	0 (0-48)	< 0.001
Peripheral vascular disease	6003 (9.1)	7108 (13.8)	< 0.001
Dialysis groups			< 0.001
Preemptive	6881 (10.4)	5059 (9.8)
≤ 1 year	5579 (8.5)	3690 (7.2)
> 1 year ≤ 3 years	15586 (23.6)	11935 (23.2)
> 3 years ≤ 5 years	14268 (21.6)	12275 (23.9)
> 5 years	23743 (35.9)	18492 (35.9)
Wait list, days	479 (121-1240)	535 (128-1305)	< 0.001
EPTS, %	33 (13-68)	62 (35-83)	< 0.001
Transplant
Year of transplant			< 0.001
2014-2019	29559 (44.7)	21139 (41.1)
2020-2024	36498 (55.3)	30311 (58.9)
HLA mismatches			< 0.001
0	3394 (5.1)	2542 (4.9)
1-2	4734 (7.2)	2937 (5.7)
3-4	28145 (42.6)	21267 (41.3)
5-6	29784 (45.1)	24704 (48.0)
Machine perfusion	19375 (29.3)	32399 (63.0)	< 0.001
Cold ischemia time, hours	16.5 (11.5-21.6)	19.5 (15.0-24.0)	< 0.001
Distance donor to transplant hospital, miles	88 (13-203)	97 (16-212)	< 0.001
Shared organ			< 0.001
Local	39177 (59.3)	28868 (56.1)
Regional	11703 (17.7)	10630 (20.7)
National	15177 (23.0)	11952 (23.2)

¹Categorical variables were compared using χ² or Fisher’s exact test, as appropriate. Continuous variables were compared using the Wilcoxon rank-sum test.

DCD: Donor after circulatory death; COD: Cause of death; CVA: Cerebrovascular accident; eGFR: Epidermal growth factor receptor; KDRI: Kidney Donor Risk Index; cPRA: Calculated panel reactive antibody; EPTS: Estimated post-transplant survival; HLA: Human leukocyte antigen.

Graft survival

In a center-clustered Cox model on the propensity-matched cohort, procurement biopsy remained significantly associated with decreased graft survival [HR = 1.12, 95% confidence interval (CI): 1.05-1.19, P < 0.001; Table 5]. Adjustment for donor, recipient, and transplant characteristics, including CIT, did not explain the association. Multicollinearity diagnostics and proportional hazards testing supported the validity of the fully adjusted model. Sensitivity analyses using stratified and time-varying Cox approaches produced nearly identical estimates.

Table 5 Multivariable Cox proportional hazard model clustered by transplant center for death-censored graft survival in the propensity-matched cohort.

Variables	HR	95%CI	P value
Biopsy performed	1.12	1.05-1.19	< 0.001
Donor
Kidney
Left	Reference
Right	0.99	0.94-1.06	0.94
Age, years	1.014	1.01-1.02	< 0.001
Male	1.02	0.94-1.09	0.66
Blood type
A	0.94	0.88-0.99	0.046
B	0.98	0.89-1.07	0.62
O	Reference
Height, cm	0.989	0.986-0.992	< 0.001
Weight, kg	0.998	0.986-1.001	0.07
eGFR terminal	0.997	0.996-0.998	< 0.001
History of hypertension	1.14	1.07-1.22	< 0.001
Diabetes mellitus	1.60	1.43-1.79	< 0.001
DCD	1.26	1.17-1.35	< 0.001
COD: CVA/stroke	1.12	1.04-1.20	< 0.001
Recipient
Age, years	0.975	0.972-0.977	< 0.001
Male	0.92	0.86-0.99	0.049
Height, cm	1.003	0.99-1.01	0.17
Weight, kg	1.005	1.003-1.007	< 0.001
Diagnosis
Autoimmune/systemic	Reference
Cystic/genetic	0.66	0.54-0.80	< 0.001
Diabetes mellitus	0.99	0.85-1.17	0.97
Glomerular	0.96	0.81-1.13	0.63
Other	0.97	0.81-1.15	0.69
Retransplant	0.90	0.76-1.07	0.22
Vascular/hypertensive	1.02	0.87-1.20	0.81
cPRA	1.002	1.001-1.003	< 0.001
Peripheral vascular disease	1.10	0.99-1.22	0.06
Dialysis groups
Preemptive	0.77	0.66-0.88	< 0.001
≤ 1 year	0.92	0.80-1.05	0.22
> 1 year ≤ 3 years	Reference
>3 years ≤ 5 years	1.06	0.97-1.17	0.19
> 5 years	1.10	1.01-1.20	0.03
Wait list, days	1.002	1.001-1.04	0.005
Transplant
Year groups
2014-2019	Reference
2020-2024	1.03	0.96-1.11	0.40
HLA mismatches
0	Reference
1-2	1.48	1.23-1.79	< 0.001
3-4	1.51	1.28-1.77	< 0.001
5-6	1.76	1.49-2.07	< 0.001
Cold ischemia time, hours	1.009	1.005-1.014	< 0.001
Machine perfusion	0.97	0.91-1.05	0.48
Distance between donor and recipients’ hospital by 100 miles	0.99	0.98-1.01	0.12
Shared organ
Local	Reference
Regional	1.06	0.97-1.17	0.20
National	1.20	1.07-1.35	< 0.001

Hazard ratios are from a Cox proportional hazard model clustered by transplant center adjusted for all donor, recipients and transplant covariables listed. The primary variable of interest was biopsy performed (yes vs no). eGFR: Epidermal growth factor receptor; DCD: Donor after circulatory death; COD: Cause of death; CVA: Cerebrovascular accident; cPRA: Calculated panel reactive antibody; HLA: Human leukocyte antigen; HR: Hazard ratio; CI: Confidence interval.

CIT

In weighted linear mixed-effects regression, procurement biopsy was associated with a 0.70-hour (approximately 42 minutes) increase in CIT (95%CI: 0.58-0.81, P < 0.001; Table 6). Although center-level random effects captured modest variation in baseline CIT (SD 2.8 hours), the biopsy-associated delay was consistent across centers. Longer transport distance and broader sharing were also associated with prolonged CIT, but biopsy remained independently associated with ischemic delay after full adjustment.

Table 6 Weighted mixed-effect on transplant center linear regression of cold ischemia time in the propensity-matched cohort.

Variables	Effect-hours (95%CI)	Effect-minutes (95%CI)	P value
Biopsy performed	+0.70 (0.58-0.81)	+42 (35-49)	< 0.001
Distance between donor and recipients’ hospital miles	+0.0035 (0.0034-0.0037)	Approximately 21 minutes per 100 miles	< 0.001
Sharing
Local	Reference
Regional	+2.9 (1.8-3.1)	+174 (108-186)	< 0.001
National	+4.2 (4.0-4.4)	+252 (240-264)	< 0.001

Estimates represent the effect of biopsy and other covariables on cold ischemia time, expressed in both hours and minutes. All donor covariables from Table 5 were included in the model for controls; only biopsy performed, distance, and organ sharing are shown here. CI: Confidence interval.

DGF

Observed outcomes showed higher DGF rates in recipients of biopsied kidneys (30.6% vs 27.6%, P < 0.001; Table 7). Serum creatinine at 6 months and 12 months was similarly higher in the biopsy group. In weighted mixed-effect logistic regression (Table 8): Biopsy had an odds ratio = 1.16 (95%CI: 1.11-1.21, P < 0.001) and CIT (per hour) had an odds ratio = 1.025 (95%CI: 1.02-1.03, P < 0.001). Distance and national sharing also contributed to increased DGF odds, but biopsy status remained independently associated with early graft dysfunction.

Table 7 Observed outcomes by biopsy status in the propensity-matched cohort, n (%)/median (interquartile rage).

Outcome	Not biopsied (n = 26047)	Biopsied (n = 26047)	P value1
Delayed graft function	7177 (27.6)	7957 (30.6)	< 0.001
Serum creatinine mg/dL 6 months	1.3 (1.1-1.7)	1.4 (1.1-1.7)	< 0.001
Serum creatinine mg/dL 12 months	1.3 (1.1-1.6)	1.4 (1.1-1.7)	< 0.001

¹χ² tests for categorical variables and Wilcoxon rank-sum tests for continuous variables.

Table 8 Weighted mixed- effect logistic regression for development of delayed graft function in the propensity-matched cohort.

Characteristics	Odds ratio	95%CI	P value
Biopsy performed	1.16	1.11-1.21	< 0.001
Cold ischemia time, hour	1.025	1.02-1.03	< 0.001
Distance (per 1000 miles)	1.08	1.01-1.22	0.03
Share
Local	Ref
Regional	1.03	0.97-1.09	0.42
National	1.14	1.06-1.22	< 0.001

Odds ratios are from a weighted mixed logistic regression adjusted for all donor, recipient, and transplant covariates listed in the Cox model (Table 5). Only biopsy performed, cold ischemia time, distance, and sharing status are shown here.

Sequential attenuation analysis

Sequential Cox models examined whether the biopsy-survival association operated partly through CIT and DGF. For each model the results were - model A (base model): HR = 1.25, model B (+ CIT): HR = 1.12 and model C (+ DGF): HR = 1.10. Addition of CIT attenuated the biopsy effect by 50.4%, and adding DGF produced a total attenuation of 54.8%, indicating that much of the observed association is mediated through ischemic delay and early dysfunction. The fully adjusted model C (HR = 1.10, 95%CI: 1.02-1.19) corresponded to an E-value of 1.43 (lower CI limit 1.17). Given that over half of the association was mechanistically explained by CIT and DGF, the remaining modest effect is likely multifactorial and consistent with minor residual confounding and clinical features not fully captured in registry data.

RMST

RMST at 5 years was 4.73 years for non-biopsied kidneys and 4.70 years for biopsied kidneys. The difference (-0.034 years is approximately equivalent to 12 days; 95%CI: -0.054 to -0.012; P < 0.001) was small in magnitude and consistent with attenuation and E-value results.

DCA

Across 3-year and 5-year horizons, DCA demonstrated no improvement in net benefit when adding biopsy status or biopsy pathology to models using clinical characteristics alone (Figure 1). Net benefit curves for clinical-only, clinical + biopsy, and clinical + pathology were nearly superimposable in the full cohort and among biopsied kidneys. These findings indicate that biopsy information did not enhance decision utility for predicting graft failure.

Figure 1 Decision curve analysis for procurement biopsy in deceased donors with preserved renal function, comparing models with clinical variables alone, clinical variables plus biopsy status, and clinical variables plus biopsy pathology. Overlapping curves indicate no incremental net benefit. A: Decision curve analysis (DCA) for 3-year outcomes in the whole cohort; B: DCA for 5-year outcomes in the whole cohort; C: DCA for 3-year outcomes in the biopsied-only cohort; D: DCA analysis for 5-year outcomes in the biopsied-only cohort. DCA: Decision curve analysis.

Sub-analysis of pathology among biopsied kidneys

Within the biopsied, propensity-matched cohort, kidneys with moderate-severe vascular or IFTA lesions (0%-5% GS) had increased graft failure risk (HR = 1.48, 95%CI: 1.23-1.78). Kidneys with ≥ 6% GS, with or without additional lesions, demonstrated similar risk elevation (HR = 1.51, 95%CI: 1.15-1.98; Supplementary Table 2). These findings confirm that meaningful histologic injury is detectable but that such findings did not improve recipient-level outcome prediction in the primary DCA analysis.

DISCUSSION

In this national study of deceased donor kidneys with preserved renal function (terminal eGFR ≥ 60 mL/minute/1.73 m²), procurement biopsy was associated with increased graft failure, higher rates of DGF, and modest but consistent increases in CIT, yet it did not provide any measurable prognostic or decision-making advantage beyond information already available from clinical data. By focusing specifically on donors with good renal function, where clinical practice varies widely and where guidelines about biopsy remain under debate, this study directly tested the value of biopsy in a population where the rationale for its use is weakest. Across multiple analytic frameworks, including propensity matching, center-level clustering, sequential attenuation modeling, RMST, E-values, and DCA, the findings consistently demonstrated that procurement biopsy adds logistical and interpretive disadvantages without improving allocation decisions or recipient outcomes. The attenuation of the biopsy-graft failure association after accounting for CIT and DGF suggests that biopsy is unlikely to function solely as a surrogate for unmeasured donor risk and is more consistent with a process-related, rather than biologic, mechanism.

Biopsy use was common; nearly half of all kidneys in this favorable subset underwent biopsy, and kidneys that were biopsied were more than six times as likely to be non-utilized. Importantly, nearly 6% of all biopsied kidneys were not used solely on the basis of biopsy findings, and more than half of these biopsy-driven discard occurred in kidneys showing only minimal abnormalities that were not associated with inferior graft performance. This pattern aligns with prior national studies demonstrating that biopsy findings disproportionately influence discard even when donor quality overlaps substantially with transplanted kidneys. Earlier work by Mohan et al[1] showed that biopsy findings were one of the leading drivers of kidney discard in the United States, while Lentine et al[7] documented wide center-level variation in biopsy use and a substantial association between biopsy and discard even after adjustment for donor risk. The present analysis extends these findings to a clinically favorable donor population, reinforcing that for donors with preserved renal function, biopsy frequently serves as a procedural filter rather than as a clinically informative tool.

The sequential attenuation analysis provides additional insight into why biopsy is associated with inferior outcomes despite adjustment for donor and recipient characteristics. Biopsy prolonged CIT by approximately 0.7 hours, even after accounting for distance, sharing, and center-level practice patterns. Although the approximately 42-minute increase in CIT associated with procurement biopsy may appear modest at the individual donor level, its impact becomes substantial when applied across the transplant system. When multiplied across thousands of deceased donor kidneys, such incremental delays represent a meaningful system-level inefficiency that adversely affects overall graft performance. In the mixed-effects logistic model, each additional hour of CIT increased the odds of DGF by two to three percent, and biopsy itself independently increased the odds of DGF by sixteen percent. When CIT and DGF were incorporated into the sequential Cox models, over half of the association between biopsy and graft survival was explained. The remaining modest HR of approximately 1.10 likely reflects a mix of minor residual confounding and subtle donor or procurement factors not captured within national registry data. Together, these results support a process-related mechanism rather than a biologic one; the biopsy procedure is not inherently harmful, but the delay and interpretive challenges surrounding it contribute to downstream ischemic injury, DGF, and graft failure.

DCA further strengthens the conclusion that biopsy does not improve clinical decision-making in this population. At both three and five years, models incorporating biopsy status or pathology findings did not provide greater net benefit than models using only clinical data. The decision-curve plots showed nearly complete overlap of the net benefit curves, indicating that biopsy results, whether the presence or severity of pathology, did not meaningfully change decisions or improve the accuracy of identifying kidneys at risk for graft failure. This finding mirrors prior evaluations of histologic scoring systems, which have consistently shown limited reproducibility, substantial interobserver variability, and minimal incremental prognostic value beyond standard donor risk indices[17]. International experience is consistent with this perspective; European programs that limit the use of procurement biopsy report higher utilization rates without compromising graft outcomes, suggesting that routine biopsy may reflect entrenched United States practice patterns rather than evidence-based necessity[17].

Several aspects of the study design strengthen confidence in these conclusions. Restricting the cohort to donors with terminal eGFR ≥ 60 mitigates much of the traditional indication bias that complicates analyses of biopsy use in more heterogeneous donor populations. Propensity matching incorporated donor, recipient, and transplant factors, reflecting the real-world dynamics in which transplant centers, not OPOs, directly influence biopsy decisions and recipient selection. The analytic framework accounted for center-level clustering, acknowledging that centers differ substantially in their biopsy thresholds and allocation pathways. The convergence of results across multiple methodologies, including sequential attenuation modeling, RMST, E-values, and DCA, provides consistent evidence pointing toward the same interpretation: Biopsy offers minimal added value while introducing quantifiable process-related disadvantages.

Limitations are inherent to any observational study using registry data. The dataset lacks granular information on biopsy technique, slide preparation, pathologist training, and procurement conditions, all of which may influence interpretation. Reasons for discard may be subject to misclassification. Unmeasured donor attributes such as intraoperative hemodynamic instability or procurement timing could contribute to the decision to biopsy. However, given the restricted donor phenotype, excellent covariate balance, and the findings from the attenuation and decision-analytic approaches, these limitations are unlikely to overturn the conclusion that biopsy adds minimal prognostic or decision-making benefit in this donor population.

CONCLUSION

In summary, procurement biopsy among deceased donors with preserved renal function is associated with higher discard rates, modest delays in CIT, increased DGF, and slightly reduced graft survival, yet offers no incremental value in prognostication or decision-making beyond sound clinical judgement from clinical factors already known to the transplant program. The adverse association between biopsy and graft failure is primarily mediated through process-related delays rather than intrinsic pathology. These findings support a shift away from routine procurement biopsy in clinically favorable donors and toward allocation practices centered on functional and clinical data. Reducing unnecessary biopsies has the potential to expand kidney utilization, decrease ischemic injury, and improve overall system efficiency without compromising recipient outcomes.