Risk of ESRD in Prior Living Kidney Donors

Jennifer L. Wainright; Amanda M. Robinson; Amber R. Wilk; David K. Klassen; Wida S. Cherikh; Darren E. Stewart


American Journal of Transplantation. 2018;18(5):1129-1139. 

In This Article

Materials and Methods


This was a retrospective observational study examining the association between characteristics at donation and development of ESRD for LKDs who donated between April 1, 1994 and September 30, 2016 in the United States.

This study used data from the OPTN, whose data system includes data on all donors, waitlisted candidates, and transplant recipients in the United States, submitted by the members of the OPTN, and has been described elsewhere (https://optn.transplant.hrsa.gov/data/about-data/). The Health Resources and Services Administration (HRSA), US Department of Health and Human Services, oversees the activities of the OPTN contractor. Dialysis dates and diagnosis data were supplemented with data from Centers for Medicare and Medicaid Services (CMS) Medical Evidence Form 2728 (certification of ESRD). Institutional review board exemption was obtained from HRSA.

We identified LKDs who developed ESRD between April 1, 1994 and September 30, 2016 by linking OPTN LKD data with OPTN kidney waiting list and transplant data, as well as CMS Form 2728 ESRD data. We linked datasets by name, Social Security Number (SSN), date of birth, and sex, manually reviewing imperfect matches. We excluded LKDs who required multiorgan transplants. Donors were considered to have developed ESRD if we found a dialysis, waiting list, or transplant record for them in the OPTN database or CMS data.

The OPTN has data on all LKDs who donated on or after October 1, 1987 and began collecting SSNs for donors in April 1994. Our experience linking datasets[6,35] indicated that including pre–1994 LKDs results in 2 types of bias in ascertainment of ESRD: (1) missed cases of ESRD, including (mostly young) female donors who married and changed their names between donation and onset of ESRD (eg, donor Jane Doe marries and changes her name to Jane Jones, making it impossible to link her donor record to her waiting list record without an SSN), and (2) false positives for patients with common names (eg, donor John Doe is incorrectly linked with a kidney candidate who shares his name and date of birth because we cannot verify the linkage without SSN). In turn, linkage to the Social Security Death Master File (SSDMF) to censor donors at death would also be biased in these ways. Knowing that these errors would bias ascertainment of both ESRD and death by factors known to be associated with ESRD in LKDs (ie, sex, age, and race), we limited our cohort to LKDs who donated on or after April 1, 1994. See Supplementary Materials for a more in–depth explanation.


Transplant programs submit donor characteristics and clinical data on the OPTN Living Donor Registration form (LDR) at donation: sex, age, body mass index (BMI), preoperative blood pressure, race, and relationship to recipient. eGFR was calculated from preoperative serum creatinine using the CKD–EPI equation.[36] Asian, American Indian/Alaskan Native, Native Hawaiian/Pacific Islander, and multiracial races were grouped as "other."

The OPTN does not collect data on donor income, and other proxies for SES, such as level of education, were not collected for donations before October 1999. As a result, we approximated donor SES as median neighborhood income at donation by linking home ZIP code at time of donation with ZIP code level median family income data from the 2010 Census.[37]

Time from donation to ESRD was calculated using donation date and the earliest of the following: date the patient began maintenance dialysis, listing date, or transplant date. The data were death censored at date of death via the SSDMF. Otherwise, observation time was censored at September 30, 2016.

Statistical Analyses

Cumulative incidence was calculated from the Kaplan–Meier estimate of the survival function, both overall and for subsets of the cohort. We fit a Cox proportional hazards model to predict risk of ESRD in LKDs, adjusting for preselected characteristics at donation including age, sex, race, neighborhood income, relationship to recipient, BMI, eGFR, preoperative systolic blood pressure, and an interaction between age and race. This model was used to estimate adjusted hazard ratios (aHR), as well as absolute risk for select hypothetical donors.

Several covariates were not collected for the entire study period, and thus are missing for some donors. ZIP code at donation—and subsequently neighborhood income at donation—was missing for 5% of the cohort. Other factors with missing data were not collected on the LDR until 1999, and included BMI (22% missing), serum creatinine (18% missing), and systolic blood pressure (25% missing). Multiple imputation was done using Moons' methodology[38] with 50 iterations.[39] Because missingness for these values was dependent upon when the donor donated, donation date was included in imputation in addition to the covariates used in the Cox model. Donor insurance type was also included in the imputation process to assist the imputation of neighborhood income.

While our cohort included 123 526 LKDs, only 218 ESRD cases were identified, which limited the number and functional form of the covariates we could include in the model. We examined martingale residuals[40] to determine appropriate functional forms for continuous variables. BMI, eGFR, and preoperative systolic blood pressure were included as linear terms based on these results, while neighborhood income was included as a linear spline with 1 knot at the median. The main effect for age at donation was included as a linear term, along with its interaction with donor race.

Analyses were performed using R version 3.3.2[41] and associated software packages.[42–52]