Mortality and Access to Kidney Transplantation in Patients with Sickle Cell Disease–Associated Kidney Failure

Visual Abstract

Abstract

Background and objectives

Patients with sickle cell disease–associated kidney failure have high mortality, which might be lowered by kidney transplantation. However, because they show higher post-transplant mortality compared with patients with other kidney failure etiologies, kidney transplantation remains controversial in this population, potentially limiting their chance of receiving transplantation. We aimed to quantify the decrease in mortality associated with transplantation in this population and determine the chance of receiving transplantation with sickle cell disease as the cause of kidney failure as compared with other etiologies of kidney failure.

Design, setting, participants, & measurements

Using a national registry, we studied all adults with kidney failure who began maintenance dialysis or were added to the kidney transplant waiting list in 1998–2017. To quantify the decrease in mortality associated with transplantation, we measured the absolute risk difference and hazard ratio for mortality in matched pairs of transplant recipients versus waitlisted candidates in the sickle cell and control groups. To compare the chance of receiving transplantation, we estimated hazard ratios for receiving transplantation in the sickle cell and control groups, treating death as a competing risk.

Results

Compared with their matched waitlisted candidates, 189 transplant recipients with sickle cell disease and 220,251 control recipients showed significantly lower mortality. The absolute risk difference at 10 years post-transplant was 20.3 (98.75% confidence interval, 0.9 to 39.8) and 19.8 (98.75% confidence interval, 19.2 to 20.4) percentage points in the sickle cell and control groups, respectively. The hazard ratio was also similar in the sickle cell (0.57; 95% confidence interval, 0.36 to 0.91) and control (0.54; 95% confidence interval, 0.53 to 0.55) groups (interaction P=0.8). Nonetheless, the sickle cell group was less likely to receive transplantation than the controls (subdistribution hazard ratio, 0.73; 95% confidence interval, 0.61 to 0.87). Similar disparities were found among waitlisted candidates (subdistribution hazard ratio, 0.62; 95% confidence interval, 0.53 to 0.72).

Conclusions

Patients with sickle cell disease–associated kidney failure exhibited similar decreases in mortality associated with kidney transplantation as compared with those with other kidney failure etiologies. Nonetheless, the sickle cell population was less likely to receive transplantation, even after waitlist registration.

Introduction

Patients who initiate dialysis due to sickle cell disease–associated kidney failure have poor prognosis, with a 1.5- to 2.8-fold hazard of mortality compared with those with other etiologies of kidney failure (1,2). Kidney transplantation is a potential treatment option to reduce the high mortality risk in this population. Nonetheless, it is still controversial whether transplantation is desirable in patients with sickle cell disease–associated kidney failure due to their higher post-transplant mortality; historical evidence suggests two- to three-fold post-transplant mortality in recipients with sickle cell disease–associated kidney failure compared with those with other etiologies (3–5).

The high post-transplant mortality in the sickle cell population might also affect the access to transplant in this population. Clinicians could be reluctant to offer transplantation to this population because their mortality might still be higher after transplantation compared with counterparts with other kidney failure etiologies. Moreover, transplant centers are effectively disincentivized to perform transplantation in the sickle cell population because each center’s aggregate post-transplant mortality rate is closely monitored, and increases in this aggregate rate can lead to administrative consequences (6).

However, post-transplant mortality alone is an inadequate measure of candidacy for transplantation; this decision should be driven by the expected decrease in mortality associated with transplantation as compared with dialysis (7,8). Although post-transplant mortality is higher in the sickle cell population, the decrease in mortality associated with transplantation could be similar or greater in this population than in others because their dialysis mortality is also drastically higher compared with their non–sickle cell counterparts. The lower mortality associated with transplantation has not been measured in this population.

Given these knowledge gaps, we conducted a national cohort study with two objectives. First, we examined mortality rates and measured decreases in mortality associated with transplantation in patients with kidney failure due to sickle cell disease versus other etiologies. Second, we compared access to transplantation between the sickle cell and non–sickle cell populations, from dialysis initiation, to transplant waitlist registration, to receipt of a kidney transplant.

Materials and Methods

Data Sources

This study used two national registries: the United States Renal Disease System (USRDS) and the Scientific Registry of Transplant Recipients (SRTR). USRDS collects data on all individuals diagnosed with kidney failure, in collaboration with the Centers for Medicare and Medicaid Services (CMS), the United Network for Organ Sharing (UNOS), and the ESKD networks. The SRTR data system includes data on all donors, waitlisted candidates, and transplant recipients in the United States, submitted by the members of the Organ Procurement and Transplantation Network (OPTN). The Health Resources and Services Administration, US Department of Health and Human Services, provides oversight to the activities of the OPTN and SRTR contractors. Details on these datasets are provided elsewhere (9). The clinical and research activities being reported are consistent with the Principles of the Declaration of Istanbul as outlined in the “Declaration of Istanbul on Organ Trafficking and Transplant Tourism.”

Both USRDS and SRTR receive their waitlist and transplant datasets from UNOS/OPTN. However, USRDS data include everyone diagnosed with kidney failure, unlike SRTR data, which only include those who were waitlisted for or received a transplant. Additionally, SRTR data include more transplantation-related variables, allowing for deeper analyses on waitlisted candidates and transplant recipients. Throughout this study, we used either of the two datasets, depending on the scope of each specific analysis, without an explicit linkage between the two.

Study Population

The first study population (“dialysis cohort”) included all adults (≥18 years) who started maintenance dialysis with incident diagnosis of kidney failure between January 1, 1998 and December 31, 2017, from USRDS data. We identified the patients who had sickle cell disease as the “primary cause of renal failure” on the ESKD Medical Evidence Report (Form CMS-2728), with International Classification of Diseases, Ninth Edition, Clinical Modification diagnosis codes 282.6 or International Classification of Diseases, Tenth Edition, Clinical Modification diagnosis codes D57 (“sickle cell group”). All others were included in the controls.

The second study population (“waitlist cohort”) included all adults added to the national kidney transplant waiting list (“waitlist registration”) between January 1, 1998 and December 31, 2017, from SRTR data. We identified the candidates whose primary diagnoses were sickle cell disease (diagnosis code, 3029) or “other” (diagnosis code, 999), accompanied by a clear specification of sickle cell disease on the UNOS/OPTN Transplant Candidate Registration Form. All others were included in the controls. Multiple waitlist registrations of a single person were consolidated into one record. We excluded waitlisted candidates with missing dialysis start date (n=18,501; 4%) (Supplemental Figure 1).

Mortality

We compared dialysis, waitlist, and post-transplant mortality in the sickle cell group versus controls. All three outcomes were defined as the time until death, with different time origins: dialysis initiation, waitlist registration, and receipt of a kidney transplant, respectively. We used the dialysis cohort to study dialysis mortality, and the waitlist cohort to study waitlist and post-transplant mortality. Post-transplant mortality analysis was restricted to those who received a transplant during the study period. Death was ascertained from multiple sources, including the ESKD Death Notification Form and Social Security Death Master File (8).

Decreases in Mortality Associated with Transplantation

To quantify the decreases in mortality associated with transplantation, we used sequential stratification (10) to match recipients of transplants and waitlisted candidates on a 1:1 ratio (Supplemental Appendix 1 and Supplemental Figure 2). We created match sets of transplant recipients from the sickle cell group with transplant recipients from the controls using methods analogous to stepwise model building. Match set 1 included all controls. Match set 2 restricted the controls to only those identified as Black or African American (AA). Match set 3 also involved variable-ratio (up to 1:5) matching of transplant recipients from the sickle cell group with transplant recipients from the controls, on the basis of age, sex, panel-reactive antibody (PRA), diabetes, previous kidney transplants, time on dialysis, and donor type (deceased versus living), as captured on the OPTN/UNOS Kidney Transplant Candidate Registration Worksheet.

Access to Kidney Transplantation

We first examined the overall access to transplantation, defined as the time from dialysis initiation to receipt of a kidney transplant. Then, we divided this process into two segments: dialysis initiation to waitlist registration, and waitlist registration to receipt of a transplant. Patients may receive a transplant from a living donor without waitlist registration, but such events were rare (<1% of the population) and, therefore, not studied. We used the dialysis cohort to study the overall access to transplantation and the segment from dialysis initiation to waitlist registration, and the waitlist cohort to study the segment from waitlist registration to receipt of a transplant.

Statistical Analyses

To compare mortality, we conducted Cox regression with stepwise model building to identify factors that may explain any observed differences in survival between the sickle cell group and the controls. Model 1 was unadjusted. Model 2 only included those identified as Black or AA in the controls, because sickle cell disease is strongly associated with ancestry, and race is associated with mortality in patients with kidney failure (11–14). Model 3 included multiple covariables on model 2. For dialysis mortality, we adjusted for age, sex, comorbidities (hypertension, diabetes, heart disease, stroke, and current tobacco use), body mass index (BMI), hemoglobin, serum albumin, dialysis type, for-profit status of the dialysis center, and primary payer, collected from Form CMS-2728. For waitlist mortality, we adjusted for age, sex, diabetes, BMI, history of previous kidney transplant, and time from dialysis initiation to waitlist registration, collected from the OPTN/UNOS Kidney Transplant Candidate Registration Worksheet. For post-transplant mortality, we adjusted for recipient factors (age at transplant, sex, diabetes, hypertension, BMI, PRA, time on dialysis at transplant, serum albumin, peripheral vascular disease, malignancy, hepatitis C virus and hepatitis B virus seropositivity, previous kidney transplant, education level, primary payer, and number of HLA-A/B/DR mismatches), donor factors (age, sex, race, BMI, hepatitis C virus seropositivity, diabetes, and hypertension), and additional factors for deceased donors (donation after cardiac death, stroke as the cause of donor death, terminal serum creatinine, and cold ischemic time), collected from the OPTN/UNOS Kidney Transplant Recipient Registration Worksheet.

To estimate the decreases in mortality associated with transplantation, we used the absolute risk difference for mortality because the proportional-hazards assumption is invalid in comparing mortality between transplant recipients and waitlisted candidates (8,15,16). Specifically, we first estimated the differences in Kaplan–Meier mortality estimates at 1, 3, 5, and 10 years post-transplant between the transplant recipients and waitlisted candidates in the sickle cell and control groups. For this analysis, we estimated 98.75% confidence intervals (CIs) after Bonferroni correction. Additionally, we conducted Cox regression to estimate the hazard ratio (HR) for mortality in the transplant recipients versus the matched waitlisted candidates, which can be interpreted as the average association over the course of follow-up, even when the proportional-hazards assumption is invalid (10,17). We used interaction terms to test whether the amount of decrease in mortality associated with transplantation differed between the sickle cell group and the controls.

To evaluate access to transplantation, we estimated subdistribution HRs (sHRs; E. S. Kawaguchi et al., unpublished observations) (18) using the Fine and Gray method (treating death as a competing risk) and cause-specific HRs using Cox models (censoring at death). The former measures the difference in access in the presence of any mortality differences, whereas the latter does so while adjusting any mortality differences away (19). Comparing the two measures may reveal whether an apparent decrease in access to transplantation is a consequence of higher mortality in that population rather than an actual disparity. This analysis was also conducted in stepwise models. Model 1 was unadjusted. Model 2 restricted the non–sickle cell disease group to only those identified as Black or AA. Model 3 included multiple covariables on model 2. For dialysis initiation to kidney transplant and to waitlist registration, we adjusted for age, sex, comorbidities (hypertension, diabetes, heart disease, stroke, and current tobacco use), BMI, hemoglobin, serum albumin, dialysis type, for-profit status of the dialysis center, and primary payer, as captured on Form CMS-2728. For waitlist registration to kidney transplant, we adjusted for candidate PRA, age, ABO blood type, maximum acceptable cold ischemic time and number of HLA mismatches, and willingness to accept expanded criteria donors and donors positive for hepatitis B/C virus stratified by organ procurement organization, as captured on the OPTN/UNOS Kidney Transplant Candidate Registration Worksheet.

Lastly, because we identified the sickle cell group from the primary diagnosis reported on Form CMS-2728, we conducted a sensitivity analysis where we used Medicare claims data to identify patients with diagnoses of sickle cell disease (Supplemental Appendix 2).

Missing covariables (Supplemental Table 1) were handled via the conventional missing-indicator method by using product terms of the covariables and their missing indicators (20). All analyses were censored at the end of study on December 31, 2017. We used a familywise significance level of 0.05. All analyses were performed using Stata 15.1/MP for Linux (21) and R version 3.6.1 (22).

Results

Study Populations

The dialysis cohort included 1970 patients in the sickle cell group, and 2,047,790 patients in the control group. Compared with the controls, the sickle cell group was younger (median [interquartile range; IQR] age, 44 [35–53] versus 65 [54–75] years), more likely to be Black/AA (92% versus 27%), and had fewer comorbidities (Table 1). The waitlist cohort included 507 patients in the sickle cell group and 463,298 in the control group. The sickle cell group was younger (median [IQR] age, 39 [31–49] versus 53 [42–61] years), more likely to be Black/AA (94% versus 28%) or have PRA ≥80% (19% versus 10%), and less likely to have diabetes (3% versus 42%) than the control group (Table 1).

Table 1.

Population characteristics

Characteristics	Dialysis Cohort		Waitlist Cohort
	Sickle Cell (n=1970)	Control (n=2,047,790)	Sickle Cell (n=507)	Control (n=463,298)
Age at dialysis initiation, yr (IQR)	44 (35–53)	65 (54–75)	39 (31–49)	53 (42–61)
Female sex, %	48	44	47	39
Race, %
White	4	53	3	47
Black or African American	92	27	94	28
Hispanic/Latino	3	14	2	17
Other/multiracial	1	5	1	8
Comorbidities, %
Hypertension	70	84	78	86
Diabetes	6	55	3	42
Heart disease	36	47
Stroke	8	9
Current tobacco user	7	6
Body mass index, kg/m2 (IQR)	21.4 (19.2–24.9)	27.2 (23.3–32.6)	22.3 (20.3–25.5)	27.8 (24.2–32.0)
Serum albumin, g/dl (IQR)	3.0 (2.3–3.5)	3.1 (2.4–3.6)	3.8 (3.4–4.1)	3.9 (3.5–4.2)
Hemoglobin, g/dl (IQR)	7.9 (7.0–9.1)	9.7 (8.7–10.8)
Primary insurance, %
Private	29	45	34	46
Public	65	48	66	53
None/other	6	7	0.4	1
Panel-reactive antibody, %
0%			48	58
1%–9%			10	12
10%–79%			24	20
80%–100%			19	10
Previous kidney transplantation, %			8	8

Continuous variables are shown in median (IQR). Blank cells indicate no available data. See Supplemental Table 1 for information on missing values. IQR, interquartile range.

Mortality

The sickle cell group showed higher mortality in all analyses (Table 2). The 10-year Kaplan–Meier estimates for dialysis, waitlist, and post-transplant mortality were, respectively, 85%, 61%, and 50% in the sickle cell group, and 81%, 41%, and 32% in the control group (Supplemental Figure 3). Before any adjustments (model 1), the sickle cell group showed 1.21-fold (95% CI, 1.15 to 1.27) dialysis mortality, 1.98-fold (95% CI, 1.75 to 2.23) waitlist mortality, and 1.81-fold (95% CI, 1.45 to 2.27) post-transplant mortality. The fully adjusted model (model 3) indicated greater differences in mortality, with adjusted HRs of 2.14 (95% CI, 2.03 to 2.25) for dialysis mortality, 3.21 (95% CI, 2.84 to 3.62) for waitlist mortality, and 3.03 (95% CI, 2.42 to 3.80) for post-transplant mortality. Our sensitivity analysis suggested these findings are robust to how we identified the sickle cell group (Supplemental Appendix 2).

Table 2.

Hazard of mortality in the sickle cell disease group compared with the controls

Treatment Modality	Model	Hazard Ratio (95% CI)
Dialysis	Model 1	1.21 (1.15 to 1.27)
	Model 2	1.48 (1.41 to 1.56)
	Model 3	2.14 (2.03 to 2.25)
Waiting list	Model 1	1.98 (1.75 to 2.23)
	Model 2	2.00 (1.78 to 2.26)
	Model 3	3.21 (2.84 to 3.62)
Post-transplant	Model 1	1.81 (1.45 to 2.27)
	Model 2	1.72 (1.38 to 2.16)
	Model 3	3.03 (2.42 to 3.80)

Model 1 was unadjusted. Model 2 restricted the non–sickle cell disease group to only those who were identified as Black or African American. Model 3 was adjusted for age, sex, comorbidities (hypertension, diabetes, heart disease, stroke, and current tobacco use), body mass index, hemoglobin, serum albumin, dialysis type, for-profit status of the dialysis center, and primary payer (dialysis mortality); age, sex, diabetes, body mass index, history of previous kidney transplant, and time from dialysis initiation to waitlist registration (waitlist mortality); and recipient factors (age at transplant, sex, diabetes, hypertension, body mass index, panel-reactive antibody, time on dialysis at transplant, serum albumin, peripheral vascular disease, malignancy, hepatitis C virus and hepatitis B virus seropositivity, previous kidney transplant, education level, primary payer, and number of HLA-A/B/DR mismatches), donor factors (age, sex, race, body mass index, hepatitis C virus seropositivity, diabetes, and hypertension), and additional factors for deceased donors (donation after cardiac death, stroke as the cause of donor death, terminal serum creatinine, and cold ischemic time) (post-transplant mortality). 95%, 95% confidence interval.

Decreases in Mortality Associated with Transplantation

In the waitlist cohort, 507 candidates had sickle cell disease as their primary diagnosis. Of these, 192 patients subsequently received transplants. Three transplant recipients from the sickle cell group could not be matched to waitlisted candidates from the sickle cell group and were excluded, yielding 189 matched pairs in the sickle cell group. Of the 463,298 candidates from the controls, 243,045 subsequently received transplants. Among these patients, two transplant recipients from the controls could not be matched to waitlisted candidates from the controls and were excluded (match set 1). Additionally, eight transplant recipients from the sickle cell group in match set 3 could not be matched to recipients from the controls and were excluded, yielding 181 matched pairs in the sickle cell group (Supplemental Table 2). In match set 1, we ascertained 122,949 deaths from 2,604,568 person-years of follow-up, with a crude mortality rate of 4.72 (95% CI, 4.69 to 4.75) per 100 person-year over median (IQR) follow-up of 4.1 (1.7–8.0) years.

Kidney transplantation was associated with similar decreases in mortality in the sickle cell and control groups (Supplemental Figure 4 and Table 3). In match set 1, transplant recipients from the sickle cell group had lower mortality, with absolute risk differences of 6.1 (98.75% CI, −0.8 to 13.0) percentage points (pp) at 1-year post-transplant, 15.3 (98.75% CI, 3.9 to 26.7) pp at 3 years, 23.8 (98.75% CI, 9.6 to 38.0) pp at 5 years, and 20.3 (98.75% CI, 0.9 to 39.8) pp at 10 years, as compared with waitlisted candidates from the sickle cell group. Similarly, in the controls, transplantation was associated with absolute risk differences of 0.7 (98.75% CI, 0.5 to 0.8) pp at 1 year, 6.6 (98.75% CI, 6.4 to 6.9) pp at 3 years, 12.7 (98.75% CI, 12.4 to 13.1) pp at 5 years, and 19.8 (98.75% CI, 19.2 to 20.4) pp at 10 years (Figure 1). After controlling for race and clinical characteristics (match sets 2 and 3), the estimates remained overall similar.

Table 3.

Decreases in mortality risk associated with kidney transplantation in the sickle cell disease and control groups

Treatment Modality	Group	Absolute Risk Difference, Percentage Points (98.75% CI)
		1 Year Post-KT	3 Years Post-KT	5 Years Post-KT	10 Years Post-KT
Match set 1	Sickle cell	6.1 (−0.8 to 13.0)	15.3 (3.9 to 26.7)	23.8 (9.6 to 38.0)	20.3 (0.9 to 39.8)
	Control	0.7 (0.5 to 0.8)	6.6 (6.4 to 6.9)	12.7 (12.4 to 13.1)	19.8 (19.2 to 20.4)
Match set 2	Sickle cell	6.1 (−0.8 to 13.0)	15.3 (3.9 to 26.7)	23.8 (9.6 to 38.0)	20.3 (0.9 to 39.8)
	Control	0.6 (0.3 to 0.8)	5.6 (5.1 to 6.1)	11.0 (10.3 to 11.7)	18.2 (17.1 to 19.3)
Match set 3	Sickle cell	6.4 (−0.9 to 13.6)	16.0 (4.3 to 27.8)	25.3 (10.6 to 40.0)	18.8 (−1.3 to 38.9)
	Control	−1.2 (−3.0 to 0.6)	1.0 (−2.3 to 4.3)	6.1 (1.3 to 10.9)	15.5 (6.9 to 24.1)

The 98.75% CIs were determined after Bonferroni correction for multiple comparisons. Match set 1 included all transplant recipients and waitlisted candidates in the controls. Match set 2 restricted the controls to only those who were identified as Black or African American. Match set 3 also involved variable-ratio (up to 1:5) matching of transplant recipients from the sickle cell group with transplant recipients from the controls, on the basis of age, sex, panel-reactive antibody, diabetes, previous KT, time on dialysis, and donor type (deceased versus living). CI, confidence interval; KT, kidney transplant.

Figure 1.

Survival difference associated with kidney transplantation is similar in the sickle cell disease and control groups. (A) Match set 1, (B) match set 2, and (C) match set 3. Bands indicate 95% pointwise confidence intervals, without adjusting for multiple comparisons. The horizontal line at y=0 indicates no survival difference.

Our Cox model also suggested similar decreases in mortality associated with transplantation in both groups. Compared with the waitlisted candidates, the transplant recipients had a 0.57-fold (95% CI, 0.36 to 0.91) hazard of mortality in the sickle cell group and 0.54-fold (95% CI, 0.53 to 0.55) in the controls (match set 1; P=0.8). The HRs remained similar in all other match sets (Supplemental Table 3).

Access to Kidney Transplantation

Overall, the sickle cell group had lower access to transplantation (Table 4). Starting from dialysis initiation, the sickle cell group was less likely to receive a transplant (model 1, sHR, 0.73; 95% CI, 0.61 to 0.87). This association remained when death was treated as a censoring event (cause-specific HR, 0.81; 95% CI, 0.68 to 0.97), suggesting the higher mortality in the sickle cell group does not fully explain this disparity. This association was subtly weaker when the controls were restricted to Black/AA patients (model 2, sHR, 0.79; 95% CI, 0.66 to 0.93), but substantially stronger when adjusted for clinical characteristics (model 3, sHR, 0.48; 95% CI, 0.40 to 0.56).

Table 4.

Access to kidney transplantation in the sickle cell disease group compared with the control group

Entry	End Point	Model	Subdistribution Hazard Ratio (95% CI)	Cause-Specific Hazard Ratio (95% CI)
Dialysis initiation	Kidney transplant	Model 1	0.73 (0.61 to 0.87)	0.81 (0.68 to 0.97)
		Model 2	0.79 (0.66 to 0.93)	1.09 (0.91 to 1.30)
		Model 3	0.48 (0.40 to 0.56)	0.69 (0.58 to 0.83)
Dialysis initiation	Waitlist registration	Model 1	1.20 (1.08 to 1.32)	1.28 (1.15 to 1.41)
		Model 2	1.12 (1.01 to 1.24)	1.33 (1.20 to 1.48)
		Model 3	0.74 (0.67 to 0.83)	0.88 (0.80 to 0.98)
Waitlist registration	Kidney transplant	Model 1	0.62 (0.53 to 0.72)	0.72 (0.63 to 0.83)
		Model 2	0.75 (0.65 to 0.87)	0.94 (0.81 to 1.08)
		Model 3	0.64 (0.54 to 0.74)	0.82 (0.71 to 0.94)

Subdistribution hazard ratios were estimated using the Fine and Gray method, treating death as a competing risk. Cause-specific hazard ratios were estimated using Cox regression, censoring at death. Model 1 is unadjusted. Model 2 restricts the controls to only those who were identified as Black or African American. Model 3 is adjusted for age, sex, comorbidities (hypertension, diabetes, heart disease, stroke, and current tobacco use), body mass index, hemoglobin, serum albumin, dialysis type, for-profit status of the dialysis center, and primary payer (dialysis initiation to kidney transplant and dialysis initiation to waitlist registration); and candidate panel-reactive antibody, age, ABO blood type, maximum acceptable cold ischemic time and number of HLA mismatches, and willingness to accept expanded criteria donors and donors positive for hepatitis B/C virus stratified by organ procurement organization (waitlist registration to kidney transplant). CI, confidence interval.

Examining each segment separately, we observed that the chance of waitlist registration was actually higher in the sickle cell group before adjustments (model 1, sHR, 1.20; 95% CI, 1.08 to 1.32) or after restricting the controls to Black/AA patients (model 2, sHR, 1.12; 95% CI, 1.01 to 1.24). However, after adjusting for clinical characteristics, the sickle cell group was less likely to be waitlisted (model 3, sHR, 0.74; 95% CI, 0.67 to 0.83). In other words, they were more likely to be waitlisted than the entire control group, who were older and had more comorbidities, but less likely than the comparable patients in the control group.

After waitlist registration, the sickle cell group was less likely to receive a transplant than patients in the control group (model 1, sHR, 0.62; 95% CI, 0.53 to 0.72). This association persisted after restricting the controls to Black/AA patients (model 2, sHR, 0.75; 95% CI, 0.65 to 0.87), adjusting for factors constituting the kidney allocation criteria (model 3, sHR, 0.64; 95% CI, 0.54 to 0.74), and treating death as a censoring event (model 3, cause-specific HR, 0.82; 95% CI, 0.71 to 0.94).

Discussion

In this national study including all adults who started maintenance dialysis or were added to the kidney transplant waiting list between 1998 and 2017, kidney transplantation was associated with a substantial decrease in mortality in both the sickle cell and control groups, with a decrease in 10-year mortality of 20.3 and 19.8 pp, respectively. Although transplant recipients from the sickle cell group experienced higher post-transplant mortality compared with recipients from the controls (10 year, 50% versus 32%), the amount of the decrease in mortality associated with transplantation was similar in both groups. However, the sickle cell group had worse access to transplantation than the control group. The waitlisted candidates from the sickle cell group were 38% less likely to receive a transplant than the controls, and this association persisted after adjusting for factors constituting the kidney allocation criteria.

Previous studies have found higher waitlist (1,2) and post-transplant (3–5) mortality in patients with sickle cell disease–associated kidney failure. On the other hand, there has been only suggestive evidence on whether kidney transplantation is associated with meaningful decreases in mortality in the sickle cell population, or on how the amount of the decrease in mortality in the sickle cell population compares with that in the general kidney transplant recipient population (1,3,23). Our results substantiate the prior findings by demonstrating statistically significant decreases in mortality associated with transplantation in the sickle cell group. More importantly, we observed a similar degree of decreases in mortality associated with transplantation in the sickle cell and control groups.

The worse access to kidney transplantation in the sickle cell group was the most pronounced in the candidates who were waitlisted. This was unexpected because having sickle cell disease as the primary cause of kidney failure has never been explicitly considered in the kidney allocation criteria and, therefore, should not affect the chances of waitlisted candidates receiving a transplant (24,25). We conducted stepwise model building to understand why this disparity occurred. However, neither the sickle cell group’s higher mortality nor the factors constituting the kidney allocation criteria—such as their PRA, blood type, geographic location, or age—fully explained the disparity.

Of note, the higher post-transplant mortality in the sickle cell population can be a significant burden to the provider, considering the scrutiny that occurs over post-transplant mortality. Our findings imply that performing kidney transplants on candidates with kidney failure secondary to sickle cell disease effectively incurs the risk of administrative consequences, which may include a shutdown of the kidney transplant program (6). Pursuant to the Code of Federal Regulations, Title 42, Section 482.80(c)(2)(ii)(C), CMS requires transplant centers to maintain aggregate 1-year post-transplant mortality <185% of the expected rate, which is derived from a case-mix adjustment model (26). This model includes numerous clinical variables but not the recipient’s primary diagnosis other than diabetes; i.e., it expects kidney transplant recipients whose primary diagnosis was sickle cell disease to survive at the same rate as recipients who do not have sickle cell disease and are nondiabetic. Given our observation of three-fold post-transplant mortality in the sickle cell group (Table 2; model 3), performing kidney transplants on candidates with sickle cell disease will negatively affect the center’s aggregate mortality, well above the 185% threshold. This feature is clearly a disincentive for transplant centers to perform kidney transplants on candidates with sickle cell disease.

Our study has limitations. First, confounding may still exist due to the observational nature of our study. Our analyses on the decreases in mortality associated with transplantation were restricted to waitlisted candidates because all of them were clinically evaluated and cleared for kidney transplantation. Within this population, receipt of a kidney transplant is determined by factors that are less relevant to the patient’s survival, such as regional organ availability, blood type, and HLA, reducing the risk of confounding in the context of our study. However, as a downside of restricting our study to waitlisted candidates, our findings may not be generalizable to patients who have never been added to the kidney transplant waiting list; rather, our study suggests that those deemed suitable for transplantation would exhibit the decreases in mortality associated with transplantation. Lastly, our registry dataset did not have detailed information on certain clinical factors. For example, the only information available to identify patients with sickle cell disease was the primary cause of kidney failure; therefore, patients who have sickle cell disease, but whose kidney failure was secondary to other causes, were assigned to the controls. Additionally, some of the key covariables, including hemoglobin, were not available on our dialysis cohort and waitlist cohort datasets, prohibiting a uniform selection of regression covariables.

In this national study, kidney transplantation was associated with similar and substantial decreases in mortality in the sickle cell and control groups. Nonetheless, the sickle cell group had worse access to transplantation compared with the control group, even after being placed on the national kidney transplant waiting list. Our findings suggest that access to transplantation in the sickle cell population should be improved.

Disclosures

M.E. Grams reports receiving travel support from Dialysis Clinic, Inc. in May 2019 to speak at the annual director’s meeting and grants from the National Institutes of Health and National Kidney Foundation, both during the conduct of the study and outside the submitted work. M. Johnson reports being employed by Johns Hopkins School of Medicine. S. Lanzkron reports receiving research funding from Global Blood Therapeutics, Ironwood, Novartis, Pfizer, and Shire, and being employed by Johns Hopkins School of Medicine. X. Luo reports being employed by University Hospitals Cleveland Medical Center. T.S. Purnell reports being employed by Johns Hopkins University; serving as a scientific advisor for, or member of, the American Society of Transplant Surgeons; and receiving honoraria from LiveOnNY. All remaining authors have nothing to disclose.

Funding

The authors were supported by an ASN Foundation for Kidney Research Pre-Doctoral Fellowship Award (principal investigator [PI]: S. Bae); Agency for Healthcare Research and Quality grant K01HS024600 (PI: T.S. Purnell); National Institute of Diabetes and Digestive and Kidney Diseases grants K01DK101677 (PI: A. Massie), K08DK092287 (PI: M.E. Grams); National Institute of Allergy and Infectious Diseases grant K24AI44954 (PI: D. Segev); National Heart, Lung, and Blood Institute grant K01HL108832 (PI: C. Haywood); and a Patient-Centered Outcomes Research Institute research project grant (PI: S. Lanzkron).

Supplemental Material

This article contains the following supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.02720320/-/DCSupplemental.

Supplemental Appendix 1. Sequential stratification matching between KT recipients and waitlisted candidates.

Supplemental Appendix 2. Sensitivity analysis on identifying the sickle cell group.

Supplemental Figure 1. Patient inclusion flow chart.

Supplemental Figure 2. Matching transplant recipients with waitlisted candidates via sequential stratification.

Supplemental Figure 3. Kaplan–Meier curves for dialysis, waitlist, and post-transplant mortality.

Supplemental Figure 4. Kaplan–Meier curves for mortality in matched cohorts.

Supplemental Table 1. Observations with missing covariables.

Supplemental Table 2. Population characteristics of matched cohorts.

Supplemental Table 3. Relative hazard of mortality associated with kidney transplantation in matched cohorts.