Racial/ethnic disparities in access and outcomes of simultaneous liver-kidney transplant among liver transplant candidates with renal dysfunction in the United States

Abstract

Background:

Since the Model for End-Stage Liver Disease (MELD) allocation system was implemented, the proportion of simultaneous liver-kidney transplantation (SLKT) has increased significantly. However, whether racial/ethnic disparities exist in access to SLKT and post-SLKT survival remains understudied.

Methods:

A retrospective cohort of patients age ≥18 with renal dysfunction on the liver transplant (LT) waiting list was obtained from Organ Procurement and Transplantation Network. Renal dysfunction was defined as estimated glomerular filtration rate <60 mL/min/1.73m² at listing for LT. Multilevel time-to-competing-events regression adjusting for center effect was used to examine the likelihood of receiving SLKT. Inverse probability of treatment weighted survival analyses were used to analyze post-transplant mortality outcomes.

Results:

For patients with renal dysfunction at listing for LT, not listed for simultaneous kidney transplant, non-Hispanic black (NHB) and Hispanic patients were more likely to receive SLKT than non-Hispanic white (NHW) patients (NHB: multivariable-adjusted hazard ratio, aHR 2.57; 95% confidence interval, CI 1.42–4.65; Hispanic: aHR 2.03; 95% CI 1.14–3.60). For post-SLKT outcomes, compared to NHW patients, NHB patients had a lower mortality risk prior to 24 months (aHR 0.80; 95% CI 0.65–0.97), but had a higher mortality risk (aHR 2.00; 95% CI 1.59–2.55) afterwards; in contrast, Hispanic patients had a lower overall mortality risk than NHW patients (aHR 0.61; 95% CI 0.51–0.74).

Conclusions:

In the MELD era, racial/ethnic differences exist in access and survival of SLKT for patients with renal dysfunction at listing for LT. Future studies are warranted to examine whether these differences remain in the post-SLK allocation policy era.

1. INTRODUCTION

Racial disparities exist in the prevalence, diagnosis, and treatments of liver disease, access to liver transplantation (LT), and post-LT survival.^1–4 Such disparities in waitlist mortality and LT rate have also been observed, especially before the era of the Model for End-Stage Liver Disease (MELD) allocation system.^5,6 The MELD allocation system was adopted by the United Network for Organ Sharing to follow the Institute of Medicine’s recommendations for a system based on objective criteria, as opposed to the subjective assessments of disease and emphasis on waiting time.⁷

Since the MELD allocation system was implemented on February 27, 2002, the proportion of simultaneous liver-kidney transplantation (SLKT) has increased significantly^8,9 due to the following factors. First and foremost, because serum creatinine is used as a key determinant in the prioritization for LT, the prevalence of renal dysfunction in waitlisted candidates has steadily risen.¹⁰ Second, the increasing trend of indications for LT that are highly correlated with renal disease, e.g., non-alcoholic steatohepatitis, hepatitis B and C, and hepatocellular carcinoma (HCC),^11,12 results in an increasing need for SLKT. Last, for these patients SLKT is often a preferred treatment to LT alone, because studies have shown that waitlisted patients with renal dysfunction who received LT alone and continued to have renal dysfunction after transplant had a higher risk of bacterial and fungal infection¹³ and poor graft and overall survival outcomes.^14–18 The rising demand and number of SLKT performed in the recent years have prompted the recent implementation of the simultaneous liver-kidney (SLK) allocation policy in August 2017 to establish eligibility guidelines for SLKT for more equitable allocation of kidney allografts.¹⁹

While equity in transplant access is an over-arching goal for both the MELD allocation system and the SLK allocation policy through providing objective and standardized criteria for organ allocation, racial/ethnic parity was never an explicit objective of either of these policies. Previous studies have shown that patients of black race used to be less likely than whites to receive LT in the pre-MELD era,^5,6 but that racial disparities no longer exist in the post-MELD era.⁶ However, to our knowledge, no studies have examined racial and ethnic disparities in SLKT access and outcomes.

There is an increasing demand for SLKT and number of SLKT in the post-MELD era. Due to the recent implementation of the SLK allocation policy and lack of studies on racial/ethnic differences in access to SLKT and outcomes, we studied whether racial/ethnic disparities exist in access and outcomes of SLKT in the post-MELD and pre-SLK policy era, using a large, contemporary cohort of U.S. LT candidates with renal dysfunction at listing for LT. This subset of patients were selected as our study population because they were potential candidates of SLKT. The findings from this study could shed light on identifying disparities in the pre-SLK allocation policy era for future comparisons when data in the post-SLK allocation policy era are available.

2. MATERIALS AND METHODS

2.1. Data

A retrospective cohort of patients who registered on the LT waiting list between February 28, 2002 and December 31, 2013 was obtained from the Organ Procurement and Transplantation Network (OPTN) database. Institutional Review Boards at the Washington University School of Medicine approved the study.

We obtained patient demographic data on age, gender, race, height, weight, and body mass index (BMI), socioeconomic data on education, insurance, and zip code of residence. Zip codes were used to categorize into urban and rural based on Rural Urban Commuting Area Data version 2.0²⁰ and to proxy income level using the median household income within the zip code based on the American Community Survey.²¹

We also collected data on registered transplant center and patient clinical data on etiology of liver disease, year and region listed, comorbidities, whether a patient received dialysis, international normalized ratio (INR), MELD score, level of serum albumin, creatinine, and bilirubin. Estimated glomerular filtration rate (eGFR) was computed based on the Chronic Kidney Disease Epidemiology Collaboration equation,²² based on the following formula:

$$eGFR=141\times \text{min}{\left(\frac{Scr}{\kappa },1\right)}^{\alpha }\times \text{max}{\left(\frac{Scr}{\kappa },1\right)}^{-1.209}\times {0.993}^{Age}\times 1.018\left[if female\right]\times 1.159\left[if black\right],$$

where scr is serum creatinine in mg/dL, κ is 0.7 for females and 0.9 for males, α is −0.329 for females and −0.411 for males. To obtain more data on baseline kidney failure treatment, we linked patients on the LT waiting list to the kidney transplant (KT) waiting list by patient identifier and collected data on their dialysis status and duration of dialysis closest to transplant, if there was one. Data for the time-varying variables were obtained at all available time points.

For patients with transplant, donor data on age, height, race, cause of death, indicator of donation after cardiac death, type of graft donation (whole, partial), organ share type (local, regional, national), and cold ischemic time were also collected to compute Donor Risk Index (DRI)²³. Additionally, donor hepatitis C status was obtained. Transplant data on level of human leukocyte antigen (HLA) mismatch, whether the recipient was on ventilator and on life support and medical condition (home, hospital, intensive care unit) at transplant, and center where the transplant was performed were collected. Last, time of graft failure, if the patient experienced one, and post-transplant mortality were obtained.

2.2. Outcome measures

The primary outcome was time from listing on the LT waiting list to receiving SLKT versus LT alone. Patients who had no transplant, did not die, and were not removed from the list were assumed to be on LT list at the time of last death recorded within the cohort, March 14, 2014.

The secondary outcome was patient all-cause mortality after transplant for patients who underwent either SLKT or LT. Patients without date of death information were assumed to be alive on March 14, 2014.

2.3. Race/ethnicity

We used race and ethnicity definitions provided by transplant centers with the OPTN data collection infrastructure.²⁴ Race/ethnicity was categorized as non-Hispanic white (NHW), non-Hispanic black (NHB), Hispanic, and other.

2.4. Analytic cohort

We restricted our cohort to waitlisted patients with renal dysfunction, defined as eGFR <60 mL/min/1.73 m², at listing for LT. The rationale for this restriction was because all included patients to some extent would need SLKT (or be considered SLKT); otherwise, the estimates of disparity would be biased, if more patients of some race/ethnicity than others never developed renal dysfunction and thus never needed SLKT. We used a less stringent criterion for renal dysfunction based on published studies^19,25 to include more candidates being considered SLKT (see sensitivity analyses using a more stringent criterion for renal dysfunction).

To assess the likelihood of receiving SLKT, we applied the following exclusion criteria: (a) patients age <18 at listing; (b) patients who had any previous transplant; (c) patients who underwent living donor transplant; (d) patients’ date of removal from the list were earlier than their date at listing; (e) patients with extreme BMI at listing (BMI <12 or BMI >100); (f) patients on dialysis for >10 years; (g) patients with missing data on continuous variables or registered transplant center; (i) patients with HCC or unknown HCC status at listing; and (j) patients registering at centers, in which no SLKT was performed.

2.5. Statistical analyses

To compare racial/ethnic groups, chi-square tests were performed to examine differences in proportions for categorical variables. Analysis of variance was conducted to test differences in means for continuous variables. When categorical variables were used, an unknown category was created for individuals with missing data.

For the primary outcome, the analytic cohort was stratified by whether they were listed for simultaneous KT at the time of listing for LT. Time-to-competing-events regression was used to assess the likelihood of receiving SLKT with three possible outcomes: SLKT, LT alone, or no transplant (including patients who died on the LT waiting list). Patients with no transplant were treated as censored, and LT was regarded as the competing event of receiving SLKT. Given some transplant centers are more likely to perform SLKT than others, multilevel model was used to account for center volume effects by categorizing transplant centers into high- and low-volume SLKT clusters based on the medium number of SLKT.^26,27 Forward selection was carried out to select the covariates to be included in the multivariable-adjusted model among all variables (see candidate variables in Table 1 or Appendix Table S1). To ensure the robustness of our results, we also performed multilevel time-to-competing-events regression with four possible outcomes: SLKT, LT alone, waitlist mortality, or censored. Both LT alone and waitlist mortality were regarded as the competing events of receiving SLKT.

Table 1.

Demographic, socioeconomic, and clinical characteristics of 29,182 patients with renal dysfunction on the liver waiting list from February 28, 2002 to December 31, 2013

Variable	Listed for simultaneous KT: n=5,823 (19.95%)				Not listed for simultaneous KT n=23,359 (80.05%)
	NHW	NHB	Hispanic		NHW	NHB	Hispanic
N (%)	3,616 (62.10)	834 (14.32)	1,047 (17.98)		17,488 (74.87)	1,862 (7.97)	3,009 (12.88)
Age at listing
Mean (Std)	55.54 (9.57)	54.84 (8.40)	54.71 (9.19)		55.62 (9.79)	53.26 (10.39)	54.84 (9.68)
Gender (%)
Female	36.75	38.49	36.01		43.81	49.79	45.36
Male	63.25	61.51	63.99		56.19	50.21	54.64
BMI at listing
Mean (Std)	28.70 (6.06)	27.04 (5.62)	28.45 (5.59)		29.30 (6.20)	29.62 (6.65)	29.41 (5.92)
Ever had dialysis (%)
No	35.84	28.06	28.08		88.43	87.27	84.88
Yes	64.16	71.94	71.92		11.57	12.73	15.12
Duration of the dialysis prior to transplant or censoring (%)
None	41.37	33.93	32.57		90.71	89.96	87.60
≤60	19.41	14.15	17.57		0.43	0.81	0.56
>60	35.20	46.64	45.56		0.29	0.64	0.53
Unknown	4.01	5.28	4.30		8.58	8.59	11.30
Diabetes (%)
No	62.33	51.80	44.70		71.64	72.29	64.34
Yes	36.42	46.64	53.49		25.89	24.81	33.27
Unknown	1.24	1.56	1.81		2.46	2.90	2.39
Hypertension (%)
No	90.43	86.21	88.44		90.02	85.98	89.00
Yes	8.57	12.59	9.93		7.36	11.12	7.58
Unknown	1.00	1.20	1.62		2.62	2.90	3.42
Albumin
Mean (Std)	3.04 (0.71)	2.99 (0.77)	3.03 (0.76)		2.99 (0.70)	2.74  (0.75)	2.95 (0.74)
Serum creatinine
Mean (Std)	3.69 (2.23)	5.30 (3.11)	4.04 (2.50)		2.05 (1.23)	2.50 (1.55)	2.10 (1.26)
Bilirubin
Mean (Std)	5.19 (8.88)	4.75 (8.56)	5.15 (9.36)		7.87 (10.86)	11.24 (12.43)	10.02 (12.80)
International normalized ratio (%)
0-<1.7	69.52	74.34	71.82		54.93	38.78	48.12
1.7-<3.4	28.18	21.82	25.79		37.36	45.49	43.84
3.4-<6	1.77	2.64	1.91		5.65	11.98	6.55
Unknown	0.53	1.20	0.48		2.05	3.76	1.50
Ascites (%)
Absent	14.02	22.78	15.00		12.95	16.81	10.34
Slight	43.92	44.84	46.80		50.06	46.78	50.22
Moderate	42.04	32.25	38.11		36.96	36.41	39.42
Unknown	0.03	0.12	0.10		0.03	0.00	0.03
MELD score
Mean (Std)	24.86 (8.36)	25.41 (8.06)	24.70 (8.11)		23.74 (10.35)	28.70 (11.26)	25.23 (10.86)
On ventilator (%)
No	96.79	97.60	96.75		92.73	89.04	92.79
Yes	3.21	2.40	3.25		7.27	10.96	7.21
On life support
No	95.74	96.52	95.70		92.01	88.08	91.79
Yes	4.26	3.48	4.30		7.99	11.92	8.21
eGFR
Mean (Std)	22.96 (13.72)	18.17 (12.97)	21.41 (13.79)		39.43 (14.56)	37.40 (15.38)	38.85 (14.91)
Donor risk index
Mean (Std)	1.41 (0.37)	1.42 (0.38)	1.43 (0.34)		1.54 (0.46)	1.53 (0.42)	1.54 (0.42)
Days on the liver waiting list
Mean (Std)	263.63 (432.76)	270.19 (404.17)	335.55 (481.13)		261.84 (513.62)	161.40 (383.24)	269.96 (522.05)
Etiology of liver disease (%)
Biliary atresia	3.37	0.36	1.05		2.60	0.48	0.73
Cholestatic cirrhosis	3.01	2.40	2.20		4.76	3.76	3.72
Fulminant	2.74	3.36	1.43		7.07	13.37	6.25
Malignant neoplasm	10.34	10.79	6.78		7.37	9.83	4.75
Metabolic disease	0.11	0.36	0.00		0.20	0.11	0.10
Noncholestatic cirrhosis	80.42	82.73	88.54		78.01	72.45	84.45
Medical Condition
ICU	8.60	6.35	8.50		11.09	15.68	13.53
Inpatient	13.38	11.15	14.61		15.28	18.31	18.01
Home	33.13	39.09	24.36		30.54	27.07	22.17
Unknown	44.88	43.41	52.53		43.08	38.94	46.29
Year at listing (%)
2002	3.40	2.16	4.01		5.39	4.99	5.38
2003	5.09	3.96	4.30		6.99	6.39	6.91
2004	5.97	4.92	5.44		8.22	7.25	8.24
2005	7.05	6.47	7.16		8.69	7.47	8.64
2006	7.74	9.47	7.45		8.34	8.38	7.84
2007	9.04	9.35	9.65		8.50	6.98	7.98
2008	8.02	8.39	8.21		8.50	9.24	8.04
2009	8.27	7.67	9.26		8.50	9.40	8.64
2010	9.29	9.95	10.51		9.38	10.26	9.34
2011	11.23	11.51	11.27		9.18	11.44	9.90
2012	12.53	12.23	11.94		8.77	9.18	9.70
2013	12.36	13.91	10.79		9.54	9.02	9.37
Insurance (%)
Medicaid	10.65	17.03	26.07		12.43	19.76	27.78
Medicare	32.36	40.89	38.40		23.24	22.13	23.43
Other	3.60	2.16	2.29		5.52	7.68	5.48
Private Insurance	53.40	39.93	33.24		58.82	50.43	43.30
Education (%)
None	0.19	0.00	1.81		0.18	0.27	1.10
Some school (< 5y)	0.00	0.00	0.00		0.05	0.05	0.00
Grade school (0–8y)	2.27	2.76	21.59		2.09	1.77	19.28
High school (9–12y)/GED	40.98	46.40	42.02		37.29	41.46	38.98
Bachelor degree	15.74	10.07	6.02		13.93	10.20	6.75
College/technical school	19.99	20.98	12.70		19.71	18.42	13.76
Graduate degree	6.42	3.36	1.05		5.67	3.97	2.33
Unknown	14.41	16.43	14.80		21.09	23.85	17.81
Region (%)
1	5.75	2.64	2.96		4.66	2.58	2.86
2	11.17	19.18	4.58		14.40	19.92	5.65
3	13.50	17.63	8.02		13.56	15.52	11.07
4	9.21	11.03	21.68		8.48	7.84	17.41
5	13.30	8.39	42.50		13.27	8.27	37.35
6	1.60	0.48	0.76		2.86	0.81	1.53
7	17.04	14.27	9.36		9.44	9.51	6.11
8	7.05	3.12	1.81		6.88	3.92	4.52
9	4.09	6.83	5.83		6.78	9.94	10.54
10	9.60	9.35	1.43		8.89	8.32	1.53
11	7.69	7.07	1.05		10.77	13.37	1.43
Income (%)
1st quartile	18.50	46.76	35.05		20.67	44.31	34.26
2nd quartile	25.47	20.62	24.83		26.61	19.98	24.96
3rd quartile	25.69	16.79	23.88		25.14	17.88	22.00
4th quartile	27.46	14.51	11.46		24.94	15.57	13.92
Unknown	2.88	1.32	4.78		2.64	2.26	4.85
Urban (%)
No	7.72	1.92	2.67		8.79	2.15	2.29
Yes	90.68	97.60	94.17		89.94	96.94	94.12
Unknown	1.60	0.48	3.15		1.27	0.91	3.59
Center volume (%)
Low volume (1–18)	15.87	17.51	15.76		30.06	29.27	25.36
High volume (19–146)	84.13	82.49	84.24		69.94	70.73	74.64
Transplant status (%)
SLKT	49.94	52.52	44.22		0.43	0.91	0.63
LT	5.17	4.20	3.25		56.48	60.15	53.07
NO transplantation	44.88	43.29	52.53		43.08	38.94	46.29

NHW: non-Hispanic white; NHB: non-Hispanic black; std: standard deviation; WLKIY: listed for simultaneous kidney; BMI: body mass index; SLKT: simultaneous liver-kidney transplant; LT: liver transplant; KT: kidney transplant; MELD: Model of End-Stage Liver Disease; eGFR: estimated glomerular filtration rate; DRI: Donor risk index.

Data at listing were used unless otherwise specified.

For the secondary outcome, we restricted to patients receiving either SLKT or LT alone. To account for the potential bias arising from confounding variables that affect both the likelihood of patients being selected for SLKT and survival, we plotted inverse probability of treatment weighted (IPTW) survival curves and performed IPTW survival analyses using propensity scores to generate the weights.^28,29 To predict a patient’s propensity score of receiving SLKT, we used a logistic regression model with a binary outcome indicating whether a patient received SLKT and potential predictors including blood type and log-transformed DRI in addition to the potential variables considered in the time-to-competing-events model (see Appendix Table S2). We then kept the covariates with p <0.1 in the final model.³⁰ The propensity scores were predicted through the estimated logistic regression model. Patients with extremely high or low propensity scores in both the SLKT and the LT groups were excluded to ensure overlap in the range of propensity scores across the two groups, a required condition for propensity score, i.e., common support.³¹

These estimated propensity scores were used to compute the weights for the IPTW survival curves and the IPTW survival model.^28,29 The proportional hazard assumption was tested using the Schoenfeld residuals and time-dependent covariates.³² We stratified the IPTW survival curves by race/ethnicity based on SLKT status and used log-rank or Renyi test statistics to detect the difference between racial/ethnic groups.³² The IPTW survival model was used to analyze overall mortality (time from transplant to patient death/censoring) for patients undergoing either SLKT or LT with the following covariates: duration of the most recent dialysis prior to transplant or censoring (if any), donor Hepatitis C status, HLA mismatch, and time-varying graft failure, in addition to the covariates used in the logistic regression model (full model, see Appendix Table S2). For the survival analysis that violates the proportionality assumption, we used a step function for the time-dependent covariate and determined the change point that yielded the largest log partial likelihood.³²

All tests are two-sided. Statistical significance was evaluated at α=0.05 level. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC) and R version 3.4.3 (R Core Team, 2017).

2.6. Sensitivity analyses

To explore the impact of more stringent definitions of renal dysfunction on our conclusion, we performed sensitivity analyses (SA) using two alternative definitions of renal dysfunction for LT candidates: SA(1) either receipt of dialysis or having creatinine ≥2.0 mg/dL^30,33,34 at listing for LT; and SA(2) either receipt of dialysis or having eGFR <35 mL/min/1.73 m² at listing for LT.¹⁹ We repeated the same analyses on these patients.

3. RESULTS

3.1. Trend of SLKT

In general, throughout the study period (2002–2013), the total number of SLKT increased, declined in 2008 and 2009, and increased again (Figure 1a). The proportion of SLKT over the total number of transplants (LT+SLKT) for NHB patients was the highest in all years except for 2002–2003 and 2011 (Figure 1b).

Figure 1. Trend of SLKT and distribution of SLKT among racial/ethnic groups.

(a) Number of SLKT for each racial/ethnic group, 2002–2013

(b) Percentage of SLKT over the total number of transplants (LT+SLKT) for each racial/ethnic group, 2002–2013

3.2. Primary outcome: likelihood of receiving SLKT

We identified 116,649 patients who registered on the LT waiting list (Figure 2) between February 28, 2002 and December 31, 2013 in the OPTN database. Among them, 32,383 patients had renal dysfunction at the time of listing for LT. We excluded 366 patients age <18 at listing, 320 patients with a previous transplant, 340 patients who underwent living donor transplantation, 4 patients whose date of removal were earlier than their date at listing for LT, 173 patients with extreme BMI, 149 patients on dialysis for >10 years, 439 and 52 patients with missing data, and 1,164 patients with HCC or unknown HCC status at listing. Last, we excluded 194 patients listed in transplant centers, which did not perform SLKT. The analytic cohort included 29,182 patients, among whom, 20.0% were listed for simultaneous KT at the time of listing for LT (Table 1).

Figure 2. Data attrition diagram.

† Data at listing for these variables were used; ‡ Data at transplant for these variables were used.

Among patients listed for simultaneous KT (n=5,823) at listing for LT, 62.1% were NHW, 14.3% were NHB, and 18.0% were Hispanic; among patients not listed for simultaneous KT (n=23,359) at listing for LT, 74.9% were NHW, 8.0% were NHB, and 12.9% were Hispanic. Patients of different racial/ethnic groups were statistically significantly different in all variables, except for DRI and year at listing for patients not listed for simultaneous KT (see Appendix Table S1) as well as gender, whether the recipient was on life support and was listed in low- or high-volume transplant center, MELD scores, DRI, and year at listing for patients listed for simultaneous KT at time of listing for LT.

For patients listed for simultaneous KT at the time of listing for LT (Figure 3), compared to NHW patients, NHB and Hispanic patients did not appear to have different likelihood in receiving SLKT (NHB: multivariable-adjusted hazard ratio, aHR 1.00, 95% confidence interval, CI 0.89–1.12; Hispanic: aHR 0.99, 95% CI 0.88–1.12); but for patients only listed on LT waiting list, NHB (aHR 2.57, 95% CI 1.42–4.65) and Hispanic (aHR 2.03, 95% CI 1.14–3.60) patients had a higher likelihood of receiving SLKT. The results of the multilevel time-to-competing-events regression with four possible outcomes (SLKT, LT alone, waitlist mortality, or censored) are very similar and thus are not reported.

Figure 3. Multivariable adjusted hazard ratios for receiving SLKT, 2002–2013.

Reference group: Non-Hispanic white; Hazard ratios for other race group are not presented.

Multilevel time-to-competing-events regression model adjusting for age, gender (female, male), race (NHW, NHB, Hispanic, other), BMI, year indicators (2002–2013), ever had dialysis (yes, no), eGFR, serum albumin, creatinine, and bilirubin, MELD score, INR (<1.7, 1.7-<3.4, 3.4-<6, ≥6, missing), insurance (Medicare, Medicaid, private, other, unknown), education (some school, grade school, high school or GED, Bachelor degree, college/technical school, post-college graduate degree, none, unknown), income (1^st-4^th quartile, unknown), region (1–11, based on OPTN geographic region category), ascites (absent, slight, moderate, unknown), etiology of liver disease (biliary atresia, cholestatic cirrhosis, fulminant, malignant neoplasm, metabolic disease, noncholestatic cirrhosis, other), diabetes (yes, no, unknown). Data at listing were used unless specified otherwise.

* Statistically significant at α=0.05.

3.3. Secondary outcome: all-cause mortality after transplant

We present the estimated logistic regression model used to predict the propensity scores for receiving SLKT or LT in Appendix Table S2. After excluding patients with extreme propensity scores (n=17) and patients with missing values (n=270), we included 16,092 patients in the second analytic cohort for survival analyses (Figure 2). The IPTW survival curves (using the propensity scores as weights) (Figure 4) demonstrate statistically significant differences between the racial/ethnic groups for LT patients (log-rank test p <0.0001) and for SLKT patients (Renyi test p <0.0001). Renyi test was performed in SLKT patients, because the proportionality assumption did not hold in these patients, particularly among NHB patients. Consequently, we determined that the change point for NHB patients that yielded the largest log partial likelihood and satisfied the proportionality assumption was post-SLKT 24 months.

Figure 4. Inverse probability of treatment weighted survival curves among racial/ethnic groups with SLKT or LT alone.

LT Log-rank test p<0.0001

SLKT Renyi test p<0.0001

The multivariable-adjusted IPTW survival analyses (Table 2, upper panel) show that for SLKT patients, compared to NHW patients, NHB patients had a lower mortality risk prior to 24 months following SLKT (aHR 0.80, 95% CI 0.65–0.97), but had a higher mortality risk after 24 months (aHR 2.00, 95% CI 1.59–2.55). Among LT patients, compared to NHW patients, NHB patients had higher overall mortality (aHR 1.34, 95% CI 1.21–1.49). On the other hand, Hispanic patients had a lower mortality risk than NHW patients for both SLKT (aHR 0.61, 95% CI 0.51–0.74) and LT alone (aHR 0.78, 95% CI 0.69–0.87).

Table 2.

Inverse probability of treatment weighted survival analyses: overall mortality for patients undergoing SLKT or LT alone

		Main analyses (n=16,092)
		NHB	Hispanic
LT alone		1.34 (1.21, 1.49)	0.78 (0.69, 0.87)
SLKT	≤24 months	0.80 (0.65, 0.97)	0.61 (0.51, 0.74)
	>24 months	2.00 (1.59, 2.55)

		Sensitivity analyses (1) (n=8,513)		Sensitivity analyses (2) (n=8,739)
		NHB	Hispanic	NHB	Hispanic
LT alone		1.36 (1.18, 1.56)	0.75 (0.64, 0.88)	1.40 (1.22, 1.61)	0.75 (0.64, 0.87)
SLKT	≤15 months	0.72 (0.57, 0.91)	0.64 (0.53, 0.79)	0.74 (0.59, 0.94)	0.77 (0.64, 0.94)
	>15 months	1.95 (1.54, 2.47)		1.31 (1.01, 1.71)

Main analyses: renal dysfunction was defined as eGFR <60 at listing for LT; Sensitivity analyses (1): renal dysfunction was defined as either receipt of dialysis or having creatinine ≥2.0 mg/dL at listing for LT; Sensitivity analyses (2): renal dysfunction was defined as either receipt of dialysis or having eGFR <35 at listing for LT.

Reference group: Non-Hispanic white; Hazard ratios for Other race group are not presented.

NHB: non-Hispanic black; SLKT: simultaneous liver-kidney transplant; LT: liver transplant; aHR Multivariable-adjusted hazard ratio; CI: confidence interval.

Inverse probability of treatment weighted survival analyses adjusting for recipient age, gender (female, male), race (NHW, NHB, Hispanic, other), BMI, year indicators (2002–2013), etiology of liver disease (biliary atresia, cholestatic cirrhosis, fulminant, malignant neoplasm, metabolic disease, noncholestatic cirrhosis, other), ever had dialysis (yes, no), duration of the most recent dialysis prior to transplant or censoring (≤60 days, >60 days, none, unknown), eGFR, diabetes (yes, no), hypertension (yes, no), serum albumin, creatinine, and bilirubin, MELD score, INR (<1.7, 1.7-<3.4, 3.4-<6, ≥6, missing), ascites (absent, slight, moderate, unknown), ventilator (yes, no) and life support (yes, no), medical condition (home, hospital, ICU, unknown), insurance (Medicare, Medicaid, private, other, unknown), blood type (A, A1, A1B, A2, A2B, AB, B, O), education (some school, grade school, high school or GED, Bachelor degree, college/technical school, post-college graduate degree, none, unknown), income (1st-4th quartile, unknown), indicator of urban dwelling (yes, no, unknown), region (1–11, based on OPTN geographic region category), HLA mismatch (zero mismatch, mismatch, missing), natural log transformed DRI, donor hepatitis C (yes, no), time-varying graft failure status, days on the liver waiting list, and center volume (low volume: ≤18 SLKTs, high volume: >18 SLKTs). Data at transplant were used.

3.4. Sensitivity analyses: likelihood of receiving SLKT

The results for the SA using alternate definitions of renal dysfunction at listing for LT (Figure 3) are similar to the main analyses. Compared to NHW patients, for patients listed for simultaneous KT at the time of listing for LT (n=5,166 for SA(1) and n=5,227 for SA(2)), NHB (SA(1): aHR 0.99, 95% CI 0.88–1.12; SA(2): aHR 0.98, 95% CI 0.87–1.11) and Hispanic (SA(1): aHR 0.98, 95% CI 0.87–1.11; SA(2): aHR 0.98, 95% CI 0.87–1.11) patients did not appear to have difference in receiving SLKT; but for patients only listed on LT waiting list (n=8,801 for SA(1) and 9,364 for SA(2)), both NHB (SA(1): aHR 2.46, 95% CI 1.32–4.60; SA(2): aHR 2.55, 95% CI 1.36–4.78) and Hispanic (SA(1): aHR 2.26, 95% CI 1.24–4.11; SA(2): aHR 2.25, 95% CI 1.24–4.10) patients had a higher likelihood of receiving SLKT.

3.5. Sensitivity analyses: all-cause mortality after transplant

Consistent with the main analysis, the proportionality assumption did not hold in the survival analysis of the SLKT patients. The change point for NHB patients that yielded the largest log partial likelihood was post-SLKT 15 months in both SA.

For SLKT, compared to NHW patients (Table 2, lower panel), NHB patients had a lower mortality risk prior to 15 months in both SA – SA(1): aHR 0.72, 95% CI 0.57–0.91; SA(2): aHR 0.74, 95% CI 0.59–0.94 – and had a higher mortality risk after 15 months – SA(1): aHR 1.95, 95% CI 1.54–2.47; SA(2): aHR 1.31, 95% CI 1.01–1.71. For LT alone, consistent with the main analyses, compared to NHW patients, NHB patients had a higher overall mortality risk in both SA – SA(1): aHR 1.36, 95% CI 1.18–1.56; SA(2): aHR 1.40, 95% CI 1.22–1.61. Again, in both SA, Hispanic patients showed a lower mortality risk than NHW patients either with LT (SA(1): aHR 0.75, 95% CI 0.64–0.88; SA(2): aHR 0.75, 95% CI 0.64–0.87) or SLKT (SA(1): aHR 0.64, 95% CI 0.53–0.79; SA(2): aHR 0.77, 95% CI 0.64–0.94).

4. DISCUSSION

We investigated racial/ethnic disparities in the likelihood of receiving SLKT and survival following SLKT or LT among patients with renal dysfunction at listing for LT in the post-MELD era. We found that compared to NHW patients, both NHB and Hispanic patients were more likely to receive SLKT for patients who did not simultaneously register on both LT and KT waiting lists. We also found no evidence for racial/ethnic disparities in the likelihood of receiving SLKT for patients who simultaneously registered on both LT and KT waiting lists. Furthermore, compared to NHW patients, for SLKT recipients, NHB patients had better survival in the short-run, but had worse survival in the longer-term. Among LT recipients, NHB patients had worse overall survival than NHW patients. In contrast, among either SLKT or LT recipients, Hispanic patients had better overall post-transplant survival than NHW patients.

Inconsistent with the findings from previous studies that NHB patients had disadvantages in access to LT in the pre-MELD era^5,6 and that the disadvantages no longer appeared in the MELD era,⁶ our findings in the context of SLKT revealed that NHB patients were significantly more likely to receive SLKT than NHW patients if they were not simultaneously listed at the time of listing for LT in the MELD era, even after controlling for baseline patient characteristics, clinical conditions, and center effects. Several rationales may explain why transplant centers were more likely to provide SLKT to NHB patients than their white counterparts. First, as observed by previous studies that NHB patients get listed at a more advanced stage of disease due to barriers in referral for LT,^6,35–37 our data also demonstrate that (i) NHB patients had a higher MELD score and lower eGFR at listing than NHW patients; and (ii) a higher percentage of NHB patients (834/(834+1,862)=30.9%) than NHW patients (3,616/(3,616+17,488)=17.1%) were listed for simultaneous KT at time of listing for LT. Second, patients with renal dysfunction at listing for LT are given a higher MELD score and thus are prioritized, rather than subjective assessments, thus increasing the chance of being considered SLKT. Third, studies have shown that patients with renal dysfunction had a higher risk of developing end-stage renal disease after LT and NHB race was a risk factor for post-LT end-stage renal disease independent of pre-transplant renal dysfunction.³⁸ Without guidelines for SLKT, patients with renal dysfunction may have been more likely to receive SLKT due to the concern of renal failure following LT. Notably, the OPTN recently enacted SLK allocation policy with a goal of equitable allocation of kidney organs.^19,39 While it is too early to assess the impact of this policy, we anticipate that racial/ethnic differences in the era of post-SLK allocation policy will move closer to the results of the patients listed for simultaneous KT at the time of listing for LT, in which no evidence suggests disparities in the likelihood of receiving SLKT. Post-implementation evaluation should include monitoring of equity in transplant access by race/ethnicity.

In terms of post-transplant survival, despite NHB patients having a higher likelihood of receiving SLKT and better survival ≤2 years after SLKT, they had worse long-term survival than their white counterparts; moreover, NHB patients with LT alone had worse survival. Our SA using more a stringent criterion for renal dysfunction confirmed these results, except that the change point of survival advantage in NHB patients with SLKT moved earlier to 15 months post-SLKT. This could possibly be due to worse renal dysfunction at listing for LT. While racial/ethnic disparity in post-SLKT survival has not been studied, several studies have demonstrated consistent racial disparities in post-LT survival outcomes.^3,40 Except for a study by Lee et al.,⁴¹ whose patient population was from four academic centers, the patient population of the studies that found post-LT disadvantage in NHB patients, including our study, used the OPTN database. Despite controlling for all patient demographic, socioeconomic, and clinical factors at transplant, as well as center effects, the higher post-transplant mortality in NHB patients may result from unobserved heterogeneity between racial/ethnic groups both at the time of transplant and after transplant, e.g., loss of insurance after transplant, post-transplant care and medical adherence.

Interestingly, although Hispanic patients belong to a minority population with a higher prevalence of liver disease than NHW patients, Hispanic patients tend to present with more advanced disease at listing⁴ and have a higher likelihood of receiving SLKT, possibly for the same reasons as previously noted for NHB patients. However, Hispanic patients who received SLKT or LT alone had better survival (both short- and long-run) than NHW patients. It remains unclear why Hispanic patients had better post-transplant survival than NHB patients and even NHW patients. Nonetheless, this finding for post-LT survival outcomes for Hispanic patients is not unique and is consistent with several studies that separated out Hispanic ethnicity from other racial group.^3,27,40 Although Hispanic patients may suffer from the aforementioned reasons thought to explain worse post-transplant survival in NHB patients, biologic and/or disease factors have been proposed to explain this benefit,³ warranting further clinical investigations.

To our knowledge, our study is the first to explore racial/ethnic disparities in access and outcomes of SLKT among patients with renal dysfunction awaiting LT. Further, we used national data with a large sample size to draw conclusion on minority patients. To account for the center effect in the estimation of the likelihood of receiving SLKT, we carefully excluded patients listed in transplant centers, which did not perform SLKT, and used multilevel time-to-competing-events regression. To account for factors selecting patients into SLKT as opposed to LT in the evaluation of post-transplant survival between racial/ethnic groups, we used propensity scores and present both IPTW survival curves and multivariable-adjusted IPTW survival analyses.

Despite these strengths, our study has several limitations. First, like other retrospective studies using the OPTN data, our study is limited by the available variables and the existing data quality. Second, our main analysis used a generous definition of pretransplant renal dysfunction as eGFR <60 mL/min/1.73 m² at listing for LT, which allowed us to include more patients being considered for SLKT. Nonetheless, our sensitivity analyses using alternate definitions yielded the same conclusion. Last, although center effect was controlled for in the multilevel time-to-competing-events regression, racial/ethnic distribution within transplant centers may play a role in the observed racial/ethnic differences for patients not listed for simultaneous KT. In this group, we observed a slightly higher proportion of Hispanic patients listed in high-volume SLKT clusters, despite similar proportions of NHW and NHB patients (NHW: 69.9%; NHB: 70.7%; Hispanic: 74.6% as shown in Table 1).

5. CONCLUSION

In the MELD era, for patients with renal dysfunction at listing for LT, compared to NHW patients, NHB and Hispanic patients are more likely to receive SLKT. Compared to their NHW counterparts, these NHB patients have better short-term survival, although the survival benefit disappears in the long-term; in contrast, these Hispanic patients have better survival overall. Future studies are warranted to examine whether these differences remain in the post-SLK allocation policy era.

Abbreviations:

aHR

multivariable-adjusted hazard ratio

BMI

body mass index

CI

confidence interval

eGFR

estimated glomerular filtration rate

HLA

human leukocyte antigen

INR

international normalized ratio

IPTW

inverse probability of treatment weighted

KT

kidney transplant

LT

liver transplantation

MELD

Model of End Stage Liver Disease

NHB

non-Hispanic black

NHW

non-Hispanic white

OPTN

Organ Procurement and Transplantation Network

SA

sensitivity analyses

SLKT

simultaneous liver-kidney transplantation

HCC

hepatocellular carcinoma

DRI

Donor Risk Index