To identify and quantify factors contributing to racial and ethnic differences in kidney transplant outcomes, the relative risks (RRs) of DD kidney graft failure at 5 years posttransplant were examined using multiple Cox proportional hazards regression models for adult recipients of primary solitary DD kidney transplants performed between January 1, 2000 and December 31, 2007. Data were recorded at the time of transplant. Recipients were excluded from analysis if they were younger than 18 years of age at transplantation, had previously received a kidney or extra-renal transplant, or underwent an LD or multiorgan transplant.
Graft failure was defined as the earliest date of graft failure (as determined by OPTN/SRTR or CMS data) or death (as determined by OPTN/SRTR, CMS or SSDMF data). Individuals were censored at the earliest of the date of 5 years posttransplant, last follow-up, or the end of the study (December 31, 2008). The RR of graft failure for three racial/ethnic groups (African American, Hispanic/Latino and Asian) were determined and referenced to the RR for whites. The RR for Other racial/ethnic group is not shown due to the small population size.
The main analyses focus on the RR of graft failure for the racial/ethnic groups of white, African American, Hispanic/Latino and Asian kidney transplant recipients, comparing an unadjusted model and several adjusted models. Each adjusted model includes combinations of categories of variables as described in . To better understand the differential effects of patient- and center-related characteristics on kidney transplant outcomes, variables were categorized and ranked, from those that were most associated with the characteristics of the transplant center and most removed from the individual characteristics of patients to those that were most specific to individual patients and distant from the characteristics of the transplant center. Accordingly, variables were arranged in ordered categories as: center factors, OPO factors, organ factors, transplant factors, treatment protocols, socioeconomic factors, dialysis time, disease burden and patient demographics. The variables that are included in each category are collected by the OPTN at the time of transplantation. The RR of graft failure at 5 years and the number and percentage of DD kidney transplant recipients by racial/ethnic group for selected factors related to graft failure are shown in . The individual effects of adjustment for each category of variables are shown in . To quantify the incremental effect of different categories of factors, RRs were sequentially analyzed using unadjusted models and then models that adjusted for specific categories of variables beginning with center characteristics and progressing to patient characteristics (). Finally, the models were rerun in reverse sequence, from those most patient-specific to those most center-specific to determine whether the order by which variables were added would influence the direction and magnitude of each category of variables captured in the model (). All statistical analyses were performed using SAS 9.2.
Factors used in the graft outcome models
Relative risk of deceased donor graft failure at 5 years and number and percentage of deceased donor kidney transplant recipients by racial/ethnic group for selected factors related to graft failure, 2000–2008
Relative risk of graft failure at 5 years by race/ethnicity among deceased donor kidney transplant recipients, 2000–2008 (single factors into model)
Relative risk of graft failure at 5 years by race/ethnicity among deceased donor kidney transplant recipients, 2000–2008
Relative risk of graft failure at 5 years by race/ethnicity among deceased donor kidney transplant recipients, 2000–2008 (reverse factor entry into model)
summarizes the number and percentage of 58 978 kidney transplant recipients by racial/ethnic group for selected categories of statistically important variables (age, duration of ESRD, diagnosis, hepatitis C status and insurance type at transplant), and reports the overall RR of graft failure associated with subcategories of these factors. Whites were, on average, older and all other racial/ethnic groups younger, than the average age of the entire kidney transplant recipient population. The best outcomes were seen among transplant patients between 35 and 49 years of age; lesser and greater age was associated with higher rates of graft failure. There was a dose-related effect of duration of ESRD as measured by dialysis time, with an overall twofold difference in the RR of graft loss at 5 years between those receiving a preemptive kidney transplant and those who undergo transplantation after seven or more years of dialysis. In general, whites were far more likely to receive a preemptive kidney graft and less likely to be exposed to greater than 3 years of dialysis before transplantation. Also of significant importance were primary ESRD diagnosis, hepatitis C status and insurance. Compared with the reference group of recipients with glomerulonephritis, those with ESRD from diabetes had an increased RR of graft loss of 1.11; those with hypertension had an RR of 1.08; and patients with other causes of ESRD had a slightly reduced RR of 0.94. Asians were most likely to have ESRD from glomerulonephritis, African Americans from hypertension and whites from other causes. Hepatitis C was also associated with an increased RR of 1.37, and African Americans were far more likely to be hepatitis C antibody positive at the time of transplant than recipients of the other racial/ethnic groups. Finally, in comparison with private insurance only, all other types of insurance were associated with increased RR of graft loss. Medicare-only insurance was most common in African Americans, Medicaid-only in Hispanic/Latinos, and private-only and private-primary insurance in whites.
The RRs of graft failure at 5 years are shown in – by race/ethnicity for one unadjusted and several adjusted models. In each table, the RR of graft failure for African Americans, Hispanic/Latinos and Asians are compared with the RR for whites (RR = 1.00). Without adjustments, the RR of graft failure at 5 years when compared with whites was 1.35 for African Americans, 0.83 for Hispanic/Latinos and 0.75 for Asians. All comparisons in – between the RR of graft failure for whites and other race/ethnicities are statistically significant with a p-value < 0.05.
shows the independent effect of adjustment for each category of variables. When compared with the unadjusted results, adjustments for age, disease burden, socioeconomic status, time on dialysis prior to transplantation and treatment factors modified the differences in the RR of graft failure compared with whites among each of the other racial/ethnic groups. Surprisingly, adjustments for center factors and demographic characteristics (without age) had little or no effect on the rates of graft failure. Transplant factors and center-alone adjustments affected relative survival for African Americans and Asians, and organ factors influenced outcomes for Asians.
Adjustment for age alone resulted in an apparent increase in the RR of graft failure for minorities in comparison with whites (). Because white recipients were older, on average, than recipients from other racial/ethnic groups (), adjustment for age likely increased the expected number of graft failures for whites relative to minorities and, therefore, resulted in higher RRs for African Americans, Hispanic/Latinos and Asians. In contrast, whites had less pretransplant exposure to dialysis than did minority recipients. Because longer time on dialysis is associated with poorer graft survival, adjustment for time on dialysis increased the number of expected graft failures for African Americans, Asians and Hispanic/Latinos and decreased the RR of graft failure for these minority groups relative to whites. Therefore, adjusting for age tends to increase and, for duration of dialysis exposure, to decrease the RR of graft failure for each minority in comparison with whites.
It is also notable that adjustment for the center factors included in these models closely parallels the results of an adjustment for center alone. In contrast, adjustment for age alone does not mimic the effect of adjustment for additional demographic factors among African Americans and Hispanic/Latinos. Therefore, center-alone is not maintained as a separate category in the stepwise analyses displayed in and , whereas age in these analyses continues to be analyzed separately from other demographic factors.
The RR of graft failure by race/ethnicity at 5 years is shown for 11 models (1 unadjusted model and 10 models adjusted for an increasing number of factors) in and . These tables demonstrate a different method of comparison of graft failure rates than that used in . As in , unadjusted, or crude rates for African Americans, Hispanic/Latinos and Asians were compared with the rates for whites (1.00). However, unlike , the unadjusted models were followed, not by separate, stand-alone adjusted models, but by models with stepwise adjustments for the sets of variables under consideration. In these stepwise models, each subsequent model incorporates all of the adjustments preceding it, as summarized in the row immediately above. For example, in , the model for OPO factors also includes the adjustments for center factors, and the model for organ factors includes the adjustment factors for center and OPO, etc.
In , adjustment for center factors alone reduced the difference in the RR of graft failure observed in the unadjusted models between African Americans and whites. Stepwise adjustments for OPO factors and organ characteristics had little incremental effect. Further stepwise adjustment for transplant factors demonstrated additional effects only for African Americans, but no additive effect was demonstrated through adjustment for treatment protocol. With progressive adjustments for socioeconomic factors, dialysis time and disease burden, differences in the RR of graft failure between the unadjusted and adjusted results for Hispanic/Latinos and Asians when compared with whites increased; differences between whites and African Americans decreased. Taking all of the preceding adjustments into account, further adjustments for age modestly increased the differences observed between whites and African Americans and modestly decreased the differences between whites and other groups. Surprisingly, once all of the other stepwise adjustments in these models are performed, the addition of demographic factors, per se, exerted no incremental effects on graft survival. In the final complete model that accounts for all of the variables listed in , the overall adjusted RR of graft failure was 1.12 for African Americans, 0.77 for Hispanic/Latinos and 0.74 for Asians compared with whites.
As discussed earlier, the sequentially adjusted models were tested in reverse order. These results are shown in , and were in general similar to those seen in . Adjustment for demographic factors had little impact on the relationships seen in the unadjusted models. Adjustment for age increased the differences observed between whites and African Americans and decreased the differences between whites and other groups. Sequential adjustments for disease burden, dialysis time, socioeconomic factors, treatment protocols and transplant factors reduced differences between whites and African Americans, but increased differences between whites and the other race/ethnicity groups. Incremental adjustment for OPO slightly decreased differences in RR of graft failure between whites and African Americans. When all other adjustments were taken into account, there was no discernible incremental effect for center factors.
These analyses show that African Americans have graft failure rates at 5 years that are higher than those of all other racial/ethnic groups, even after adjusting for the variables listed in . The adjusted RR of graft failure for African Americans compared with whites varied from 1.07 to 1.45, depending upon the manner in which variables were incorporated into the adjusted models (–). The RR for Hispanic/Latinos and Asians compared with whites varied much less across the models considered (Hispanic/Latino: unadjusted RR = 0.83, adjusted RR range 0.74–0.88; Asian: unadjusted RR = 0.75, adjusted RR range 0.69–0.81). Adjustments for demographic characteristics, other than age, and for center factors explain few of these differences. In general, those categories (age, disease burden, duration of pretransplant dialysis exposure, socioeconomic factors and treatment protocols) that are more closely associated with individual patients have a greater effect on the RR of graft failure than do organ or transplant-related factors or variables that are more closely associated with the transplant center or OPO.
Overall, graft survival at 5 years was best for Asians and Hispanic/Latinos, intermediate for whites and poorest for African Americans. These differences were not explained in their entirety by the data from the time of transplant currently available in the OPTN/SRTR database. The range of RRs when compared with whites in the unadjusted analyses ranged from an RR of 0.75 for Asians to an RR of 1.35 in African Americans (first line of and ). In the fully adjusted models, this range is compressed to an RR of 0.74 for Asians and 1.12 in African Americans (last line of and ). Thus, the covariates explain a greater portion of the differences observed between African Americans and whites than between whites and Hispanic/Latinos or between whites and Asians ( and ). Age and duration of pretransplant dialysis exposure seem to exert the greatest effect on differential outcomes ().