Maximum cold ischemia duration for a kidney allograft: a prediction model for allograft failure at the time of organ allocation

Abstract

Background: Many determinants of kidney allograft failure are established at the time of allocation by organ distribution agencies. At this point, the main modifiable factor is the duration of cold ischemia (CIT). Currently, no practical tool exists to determine the maximum permissible cold ischemia time for a specific recipient at allocation.

Methods: We analyzed two prospective cohorts of kidney transplant recipients from European centers: a derivation cohort of 7040 patients from 10 centers (Barcelona; Leuven; Oslo; Paris Necker, Lyon, Nantes, Nancy, Montpellier, Nice, Paris Saint Louis) with data collected between 2005 and 2020, and a validation cohort of 6131 patients from 6 French centers (Paris Necker, Lyon, Nantes, Nancy, Montpellier, Nice) with data collected between 2008 and 2019. The main outcome was allograft failure (return to dialysis or pre-emptive retransplantation). We assessed 26 determinants of allograft failure available at the time of allograft allocation including cold ischemia time as a modifiable factor. Prediction models were developed using a classical survival analysis and a competing risk framework.

Findings: Allograft failure occurred in 16% (1113) of the derivation cohort and 14% (832) of the validation cohort. Independent determinants of allograft failure were donor age (HR 2.2 [1.9-2.6] for donors above 65 years old), previous allografts (HR 1.5 [1.3-1.6]), dialysis history (HR 1.7 [1.3-2] for Hemodialysis), diabetes (HR 1.4 [1.2-1.6]), vascular disease (HR 1.3 [1.1-1.5]), HLA-DR incompatibility (HR 1.2 [1.1-1.3]), donor serum creatinine (HR 1 [1-1]), and cold ischemia time (HR 1 [1-1]). Donor age was the strongest contributor, while cold ischemia was the only modifiable factor. These factors were combined into two predictive models of kidney allograft failure (Cox regression and Fine Gray) showing accurate calibration, and discrimination with a C-Index of 0.66 (95% CI: 0.63-0.70 at year one) on the validation cohort for the Fine Gray model. The Fine-Gray model, which accounts for the competing risks between allograft failure and patient death, was used to develop a practical tool for predicting allograft failure based on cold ischemia.

Interpretation: Prediction model at the time of allocation provides a simple and practical tool which may guide organ distribution agencies and medico-surgical teams by customizing cold ischemia time for a kidney allograft transplantation.

Funding: Centaure Foundation (SIREN 499,947,398-http://www.fondation-centaure.org) none of the funding sources had any role in study.

Keywords: Allograft allocation; Cold ischemia; Kidney transplantation; Predictive model.

Fig. 1

Cumulative incidence curves for allograft failure in the derivation cohort EKITE. A. Kaplan–Meier estimates of time to kidney allograft failure with censoring for death. B. Aalen-Johansen estimates of time to kidney allograft failure or death with functioning allograft as a competing event. C. Aalen-Johansen estimates of time to kidney allograft failure or death with functioning allograft as a competing event according to the donor group of age (<50 years old; 50–64 years old; ≥65 years old).

Fig. 2

Cold ischemia in the derivation and the validation cohorts. A. Distribution of cold ischemia time in the derivation cohort EKITE and on the validation cohort DIVAT. B. Kaplan–Meier estimates of time to kidney allograft failure in the derivation cohort EKITE stratified by cold ischemia time (hours). C. Kaplan–Meier estimates of time to kidney allograft failure in the validation cohort DIVAT stratified by cold ischemia time (hours).

Fig. 3

Out of bag (OOB) variable importance (VIMP) results from the prediction of kidney allograft failure using survival random forests on the derivation cohort EKITE.

Fig. 4

Proportion of individuals with a risk of failure under 0.2 predicted using the Fine Gray model. Predictions were computed for cold ischemia time ranging from 0 to 40 h at 1, 3, 5 and 10 years post-transplant in each panel. Presenting individual risk of failure for the different CIT and time points would result in saturated plots, hence, we decided to present the proportion of individuals in the dataset with a predicted risk of failure smaller than 0.2.

Fig. 5

Practical application of Fine Gray model: Individual prediction of kidney allograft risk of failure at time points post-transplant: one, three, five, and ten years based on cold ischemia time. Patient 1: 36-year-old donor, creatinin 77 μmol/L, Transplant #1, Diabetes: No, Hemodialysis: Yes. Patient 2: 54-year-old donor, creatinin 104 μmol/L, Transplant #2, Diabetes: Yes, Hemodialysis: Yes. Patient 3: 77-year-old donor, creatinin 64 μmol/L, Transplant #1, Diabetes: Yes, Hemodialysis: Yes. Patient 4: 59-year-old donor, creatinin 68 μmol/L, Transplant #1, Diabetes: No, Hemodialysis: No. Patient 5: 68-year-old donor, creatinin 57 μmol/L, Transplant #1, Diabetes: No, Peritoneal Dialysis: Yes. Patient 6: 48-year-old donor, creatinin 144 μmol/L, Transplant #2, Diabetes: No, Hemodialysis: Yes.

Supplementary Fig. S2

Supplementary Fig. S2: Calibration plots at one, three, five and ten years of prediction of risk of allograft failure for the validation cohort DIVAT predictions using Cox model A) and Fine Gray model (B).

Supplementary Fig. S3

Supplementary Fig. S3: Proportion of individuals with a predicted risk of allograft failure under 0·2 using the Cox model. Predictions were computed for cold ischemia times ranging from 0 to 40 hours at 4 time points post-transplant: one, three, five, and ten years.

Supplementary Fig. S4

Supplementary Fig. S4: Individual prediction of risk of kidney allograft failure at time points post-transplant: one, three, five, and ten years based on cold ischemia time using Cox model. Patient 1: 36-year-old donor, Transplant #1, Diabetes: No, Vascular disease: No, Hemodialysis: Yes, DR incompatibility: 1. Patient 2: 54-year-old donor, Transplant #2, Diabetes: Yes, Vascular disease: No, Hemodialysis: Yes, DR incompatibility: 1. Patient 3: 77-year-old donor, Transplant #1, Diabetes: Yes, Vascular disease: Yes, Hemodialysis: Yes, DR incompatibility: 0. Patient 4: 59-year-old donor, Transplant #1, Diabetes: No, Vascular disease: No, Hemodialysis: No, DR incompatibility: 0. Patient 5: 68-year-old donor, Transplant #1, Diabetes: No, Vascular disease: Yes, Peritoneal Dialysis: Yes, DR incompatibility: 1. Patient 6: 48-year-old donor, Transplant #2, Diabetes: No, Vascular disease: Yes, Hemodialysis: Yes, DR incompatibility: 2.