Single-center observational study of 1,119 deceased donor kidney transplant recipients suggesting the removal of 5/13 individual components of the kidney donor risk index
Giselle Guerra1,2, Luke Preczewski1, Jeffrey Gaynor1,3, Marina Tabbara1,3, Mahmoud Morsi1,3, Rodrigo Vianna1,3, Gaetano Ciancio1,3,4.
1Miami Transplant Institute, University of Miami Miller School of Medicine, Jackson Memorial Hospital , Miami, FL, United States; 2Department of Nephrology, University of Miami Miller School of Medicine, Jackson Memorial Hospital , Miami, FL, United States; 3DeWitt Daughtry Family Department of Surgery, Jackson Memorial Hospital, University of Miami Miller School of Medicine, Miami, FL, United States; 4Department of Urology, Jackson Memorial Hospital, University of Miami Miller School of Medicine, Miami, FL, United States
Background: Given our center’s desire in recent years to reduce kidney transplant waiting times by utilizing “higher risk” deceased donor(DD) kidneys, we wanted to better understand the multivariable predictors of post-transplant renal function among 1,119 adult DD recipients consecutively transplanted at our center during 2016-2019.
Methods: Three stepwise linear regression analyses of estimated glomerular filtration rate(eGFR)(using the CKD-EPI formula) at 3mo, 6mo, and 12mo post-transplant, respectively, were performed to determine the significant multivariable baseline predictors. Patients with failed allografts were included into each analysis by imputing an arbitrarily low value(5ml/min/1.73m2) for eGFR among patients experiencing death-censored graft failure prior to the time point being analyzed. A type I error<.01 was used in the attempt to avoid model selection of spuriously(or weakly) associated baseline variables. It was thought that outcomes which summarized renal function during the first year post-transplant would reasonably provide a surrogate marker for longer-term renal function/graft outcome. All DD kidneys were initially stored on ice at retrieval and were scheduled to receive machine perfusion(MP) upon arrival at our center. Sample sizes of available eGFRs at 3mo, 6mo, and 12mo post-transplant were 1,103, 1,086, and 912(including 21, 29, and 49 patients who experienced death-censored graft failure prior to each time point), respectively.
Results: Selected linear regression models for eGFR at 3mo, 6mo, and 12mo included 9, 9, and 6 variables, respectively. Unfavorable baseline characteristics selected into at least one of these models included: Older Donor Age(yr)-40(P<.000001), Longer Static Cold Storage Time(hr)(P<.000001), and Recipient BMI(P<.00003) in each model; Shorter (Donor Height(cm)-170)/10(P<.00003), Higher Natural Logarithm{Initial Donor Creatinine}(P<.001), Longer MP Pump Time(P<.003), and Greater DR Mismatches(P=.01) at 3mo and 6mo; Donor Hypertension(P<.004) and Recipient HIV+(P<.006) at 6mo and 12mo; Donation-after-Cardiac Death(DCD) Kidney(P=.002) and Cerebrovascular Donor Death(P=.01) only at 3mo; and Donor Diabetes(DM)(P=.01) only at 12mo. Interestingly, 7 components of the Kidney Donor Risk Index(KDRI) were not selected into any model: 2 splines for donor age, (Donor Age-18)*I{Donor Age <18yr}(P>.52) and (Donor Age-50)*I{Donor Age >50yr}(P>.28), 2 representations for Terminal Donor Creatinine, Terminal Donor Creatinine(mg/dl)-1(P>.57) and (TerminalDonor Creatinine-1.5)*I{TermDonor Cr >1.5}(P>.24), Black Donor(P>.08), Donor HepatitisC Virus(HCV)+(P>.06), and (Donor Weight(kg)-80)*I{Donor Wt <80}(P>.03).
Conclusion: These results suggest that 5 weaker components of KDRI, 2 splines for Donor Age, Donor Race, Donor HCV, and Donor Weight, could be eliminated without dramatically affecting its predictive power. Additionally, biochemical determinations with skewed distributions such as Donor Creatinine are best represented by natural logarithmic transformed values rather than their original scales(similar to the MELD components).
