The Impact of Cardiovascular Risk Factors on Graft Outcome Disparities in African American Kidney Transplant Recipients

Abstract

Although outcome inequalities for non-Hispanic Black (NHB) kidney transplant recipients are well documented, there is paucity in data assessing the impact of cardiovascular disease (CVD) risk factors on this disparity in kidney transplantation. This was a longitudinal study of a national cohort of veteran kidney recipients transplanted between Jan 2001 and Dec 2007. Data included baseline characteristics acquired through the USRDS linked to detailed clinical follow up information acquired through the VA electronic health records. Analyses were conducted using sequential multivariable modeling (Cox regression), incorporating blocks of variables into iterative nested models; 3,139 patients were included (2,095 NHW [66.7%] and 1,044 NHBs [33.3%]). NHBs had a higher prevalence of hypertension (100% vs. 99%, p<0.01) and post-transplant diabetes (59% vs. 53%, p<0.01) with reduced control of hypertension (BP <140/90, 60% vs. 69% p<0.01), diabetes (A1c <7%, 35% vs. 47%, p<0.01) and LDL (<100 mg/dL, 55% vs. 61%, p<0.01). Adherence to medications used to manage CVD risk was significantly lower in NHBs. In the fully adjusted models, the independent risk of graft loss in NHBs was substantially reduced (unadjusted HR 2.00 vs. adjusted HR 1.49). CVD risk factors and control reduced the influence of NHB race by 9–18%. Similar trends were noted for mortality and estimates were robust across in sensitivity analyses. These results demonstrate that NHB kidney transplant recipients have significantly higher rates of CVD risk factors and reduced CVD risk control. These issues are likely partly related to medication non-adherence and meaningfully contribute to racial disparities for graft outcomes.

INTRODUCTION

Kidney transplantation is a life-prolonging procedure that substantially improves quality of life in those with end stage renal disease (ESRD).^1–3 A recent analysis of the United Network for Organ Sharing (UNOS) registry demonstrated that over the past 25 years, kidney transplantation has led to an estimated 1.37 million life-years saved, when compared to patients remaining on the waiting list.² However, because kidney graft survival is significantly shorter in African-American (AA) recipients and graft function is highly correlated with mortality, this survival advantage is likely substantially lower in AA recipients.⁴ AA kidney transplant recipients have significantly higher rates of acute rejection and graft loss. The average kidney transplant functions just over half as long in an AA patient.⁵ Despite 40 years of focused research endeavors into this disparity, little has changed in this racial inequality. In 1977, Opelz et al demonstrated a 10% absolute difference in three-year graft survival rates between AA and Caucasian recipients (25% vs. 35%, p<0.001).⁶ In the 2012 SRTR annual report, the graft survival inequity between AA and non-AA recipients was 12% at five years post-transplant.⁵

Racial disparities in kidney transplantation have primarily been attributed to biologic and immunologic differences leading to higher rejection rates,^7–10 socioeconomic status (SES) disadvantages,^11,12 reduced access to healthcare, medication non-adherence,^13,14 and comorbidities.^15–17 Research focused on reducing the higher acute rejection rates in AA transplant recipients has been successful, yet graft loss disparities persist. AA renal transplant recipients have disadvantageous immunologic characteristics placing them at higher risk for graft loss.^18–23 Therefore, most of the past research trying to eliminate outcome disparities in AA patients was focused on reducing acute rejection rates through immunosuppressant pharmacotherapy.^24,25 The impact of SES and medication adherence on racial disparities has been well-studied with conflicting results.^12,26–31

A potentially significant but under-studied factor impacting racial disparities in kidney transplant outcomes is cardiovascular disease (CVD) and CVD risk factors. In the general population, AAs have a significantly higher prevalence of diabetes and hypertension, which occur at an earlier age with a more progressive phenotype.^32–37 Yet studies assessing the impact of CVD risk factors control on racial disparities in transplantation are quite limited.^38,39 Because of this discrepancy in evidence, addressing this area of potential risk is likely to provide strong opportunities to assess modifiable factors that may influence racial disparities in transplant. Thus, the objective of this study was to assess the burden of CVD risk factors and risk control in AA renal transplants and determine the influence of CVD risk factors on racial disparities for long-term post-transplant outcomes.

CONCISE METHODS

Study Design and Patients

This was a longitudinal cohort study of national data obtained through developing a unique dataset by linking the United States Renal Data System (USRDS) and Veterans Affairs (VA) electronic health records. The study population included veteran recipients of solitary kidneys, transplanted between January 1, 2001 and December 31, 2007 (7 years) with longitudinal follow up through December 31, 2010. Pediatrics, non-renal transplant recipients, those that were not NHW or NHB, transplant events that occurred outside the study time frame and those with graft loss or follow up <1 year were excluded. After local IRB, VA HSR&D and USRDS approval, the VA system was queried for eligible kidney transplant recipients using ICD-9 codes (V42.0 or 996.81) without a history of a liver, heart, lung, pancreas or intestine transplant. This list of patients was submitted to the USRDS which linked data through scrambled social security numbers to create the study cohort which contained detailed baseline and clinical follow up data. Further details regarding the data sources to create this cohort have previously been published.⁴⁰

Outcome Measures

The primary outcome for this analysis was graft loss, which was defined as either a return to chronic dialysis or retransplant. For this outcome, death was considered a competing risk event and accounted for through modeling. The outcome was assessed as a time to event analyses, with the reference time being the date of transplant and the end time being the date of either graft loss (event), death (competing event) or the end of the study period (censored). All-cause mortality was a secondary outcome.

Exposure Measurements

There were two primary exposures for this study, race and CVD risk factors and risk control. The study cohort was restricted to two racial groups, NHWs and NHBs. Race information was gathered from the USRDS dataset and cross-validated with the VA vital records data. CVD risk factors and control was focused on the three predominant risks: hypertension, diabetes and dyslipidemia. Hypertension was defined as a blood pressure of greater than 140/90 mmHg or the use of anti-hypertensive medications. Diabetes was defined as a hemoglobin A1c of >7% or use of anti-glycemic therapy. Dyslipidemia was defined as a LDL of >100 mg/dL, triglyceride level of >150 mg/dL or use of anti-lipemic therapy. These were defined based on national standards and guidelines specific to transplant.^41–44

CVD risk factor control assessment was defined as follows: uncontrolled diabetes was classified as a mean follow up A1c of ≥8% (referent group) vs. <8% (also assessed A1c of ≥7% vs. <7% as well). Uncontrolled hypertension control was classified as a mean follow up BP ≥140/90 (referent group) vs BP <140/90 mmHg. For uncontrolled lipids, patients were classified as a mean follow up LDL ≥100 mg/dL or TG ≥150 mg/dL (referent group), vs. LDL<100 mg/dL and TG <150 mg/dL. Medications to treat CVD comorbidities were grouped according to class and patients were categorized based on the prescribing of this agent (yes vs no) and their adherence to the therapy. Medication possession ratios (MPRs) were calculated using patient and medication-specific sums for days’ supply (numerator) divided by total days in the year (denominator) and aggregated by year post-transplant, using previously validated algorithms.^45,46 For modeling, MPR was averaged for the entire follow up time and dichotomized as <0.8 vs. ≥0.8 to classify patients as adherent vs. non-adherent; 0.8 has been well established as a valid cut point to define medication adherence as it correlates with outcomes.⁴⁷ Medications were groups as follows: oral anti-hyperglycemic agents, insulin, HMG CoA reductase inhibitors (statins), other anti-lipid therapy, beta blockers, angiotensin converting enzyme inhibitors [ACE inh] or angiotensin receptor blockers [ARBs], calcium channel blockers (CCBs), diuretics and anti-platelet therapies.

Additional covariates that were included in multivariable models included recipient sociodemographics (age, gender and marital status), recipient comorbidities (history of CAD/angina, previous transplant and years on dialysis), donor characteristics (living donor, expanded criteria donor [ECD] and donor after cardiac death [DCD]), immunologic characteristics (HLA mismatches, current panel reactive antibody level [PRA]), baseline immunosuppression (induction therapy, calcineurin inhibitors, adjunctive agents and corticosteroids) and post-transplant events (delayed graft function [DGF] and acute rejection).

Statistical Analysis

Univariate analyses compared baseline variables and clinical outcomes by race (NHBs vs NHWs), assessed the baseline and follow up prevalence and control of hypertension, diabetes and dyslidemia and determined CVD medication use and adherence using the chi square test, the Student’s t-test or Mann Whitney U test, depending on data type and normality assumptions. Multivariable modeling was performed using Cox proportional hazard regression, using the Fine and Gray competing risk model for graft loss to estimate the cumulative incidence function (CIF) and standard Cox models to estimate the survival function for death.⁴⁸ Model assumptions of proportionality of hazard over time and linearity for continuous variables were first tested and confirmed, followed by residual assessments to ensure we used the most appropriate data fit and to identify any potential outliers or influential observations.

Sequential forward entry of blocks of variables were added to the model and at each step the overall impact of NHB race on the outcome was assessed by comparing the change in β to the previous model. The Akaike information criterion (AIC) was used to assess and compare models as the goodness of fit measure. Large decreases in the AIC after the addition of a block of variables suggested strong influence of that block on explaining the variability associated with the outcome. Covariance between repeated transplants in the same individual was accounted for during modelling to prevent under-estimating variance. Sensitivity analyses were conducted to ensure the robustness of the estimates, which included standard Cox regression vs. competing risk modeling, varying the entry of variable blocks during sequential modelling, using multiple imputation for missing data and using longitudinal analyses through joint modelling to first estimate individual-level random intercepts and slopes for A1c, BP and lipid levels, followed by entry of these estimates in Cox models. As a final sensitivity analysis, we also assessed the impact of BP control in means and longitudinal models with and without anti-hypertensive therapies as covariates. Effect modification was assessed by entering interaction terms in the model, with the primary variables of interest being race, delayed graft function, rejection and CVD risk factors. All data analyses were conducted using SAS (version 9.4, SAS Institute Inc, Cary, NC). Statistical significance was defined using two-sided tests with α set at 0.05.

RESULTS

There were 5,757 kidney transplants performed between Jan 1, 2001 and Dec 31, 2007 that met inclusion criteria and were capable of being linked between the VA and USRDS databases. Of these, 382 were excluded for being Hispanic, 112 were excluded for being Asian or other races, 345 were excluded for receiving a transplant outside the time period and 275 for excluded for having graft loss or follow up <1 year post-transplant, leaving 4,643 eligible for inclusion; 1,504 were missing data for at least one variable, leaving 3,139 cases with complete data, which included NHWs (66.5%) and NHBs (33.5%, see Figure 1 for the study Consort flow diagram). Baseline sociodemographics, donor criteria and transplant characteristics are summarized and compared between NHWs vs. NHBs in Table 1. At the time of transplant, NHBs were significantly younger (59.5±10.0 vs. 57.5±10.4 years), less likely to be married (57.3% vs. 68.4%), had differences in the level of education (finished high school and attended college), differences in the primary cause of ESRD (hypertension, diabetes, and FSGS) and pre-transplant comorbidities (coronary artery disease, hypertension and previous transplant). There were also significant differences between NHBs and NHWs for donor criteria (age, gender, race, less living donors and more donors after cardiac death), immunologic risks (more HLA mismatches, longer cold ischemic times), baseline immunosuppression and post-transplant outcomes; particularly death censored graft loss (NHBs 14.7% vs. NHWs 8.0%), acute rejection (NHBs 14.8% vs. NHWs 11.4%) and delayed graft function (NHBs 24.0% vs. NHWs 11.7%).

Figure 1.

Flowchart summarizing the creation of the study cohort of NHW and NHB adult solitary kidney transplants from Jan 1, 2001 to Dec 31, 2007 with data available within the VA Health System

Table 1.

Baseline and outcomes variables for the entire cohort and stratified by recipient race

Characteristics	Total Cases with Complete Data (n=3,139)	Non-Hispanic Whites (n=2,095)	Non-Hispanic Blacks (n=1,044)
Age (mean±SD)	57.5±10.4	59.5±10.0	53.6±10.1*
Male Gender	98.0%	98.2%	97.7%
Married	64.6%	68.4%	57.3%*
Education, n=2,660
Less than high school	1.6%	2.1%	0.6%*
High school	47.0%	46.7%	47.6%
Attended college/technical school	29.8%	28.3%	32.9%*
Received associate/bachelor degree	16.5%	17.5%	14.5%
Attended graduate school	5.1%	5.4%	4.5%
Primary Cause of ESRD
Hypertension	26.8%	18.5%	43.4%*
Diabetes	29.0%	30.2%	26.4%*
Focal Segmental Glomerulosclerosis	6.7%	5.3%	9.6%*
Any type of Glomerulonephritis	13.2%	12.6%	14.6%
Comorbidities
Angina or Coronary Artery Disease	13.0%	15.3%	8.5%*
Hypertension	93.7%	92.3%	96.4%*
Diabetes	38.5%	38.4%	38.8%
Previous Kidney Transplant	3.4%	4.3%	1.6%*
Donor Information
Age (mean±SD)	40.1±14.9	40.5±14.9	39.1±14.8*
Living	34.5%	39.7%	23.9%*
Female	47.4%	49.3%	43.6%*
Non-Hispanic Black	15.1%	4.3%	36.9%*
Expanded Criteria Donor	13.4%	13.3%	13.6%
Donor after Cardiac Death	4.6%	3.9%	5.9%*
Immunologic Risks
Cold Ischemic Time (mean hrs±SD), n=2,575	14.2±10.4	13.3±10.5	15.8±10.0*
Warm Ischemic Time (mean min±SD), n=1,972	40.1±24.6	39.5±24.7	41.3±24.2
HLA mismatches (mean±SD)	3.4±1.8	3.2±1.8	4.0±1.5*
Panel Reactive Antibody (mean±SD)	2.7±11.1	2.6±10.9	3.0±11.5
Years on Dialysis (mean±SD)	2.3±2.5	1.9±2.1	3.3±2.9*
Pre-emptive Transplant	21.0%	24.5%	14.1%*
Baseline Immunosuppression
IL-2 Receptor Antagonist Antibody Induction	31.5%	34.5%	25.6%*
Cytolytic Induction	34.1%	31.6%	39.2%*
Tacrolimus	71.9%	70.2%	75.5%*
Cyclosporine	22.1%	23.7%	18.8%*
Mycophenolate	86.6%	86.7%	86.2%
Azathioprine	1.5%	1.8%	0.9%*
mTOR	10.3%	10.2%	10.4%
Prednisone	94.5%	94.5%	94.5%
Clinical Outcomes
Death	19.6%	21.7%	15.4%*
Overall Graft Loss	25.1%	24.9%	25.6%
Death-Censored Graft Loss	10.2%	8.0%	14.7%*
Acute Rejection	12.5%	11.4%	14.8%*
Delayed Graft Function	15.8%	11.7%	24.0*

^*p<0.05 comparing NHB vs. NHW

The prevalence and control of CVD risk factors at baseline and at 1, 3 and 5 years post-transplant significantly differed by race (see Table 2). At baseline, NHBs had a higher prevalence of hypertension (96.4% vs. 92.3%) and lower prevalence of dyslipidemia (40.7% vs. 46.4%). During post-transplant follow-up, the prevalence of hypertension remained significantly higher in NHBs at 1, 3 and 5 years, while the prevalence of diabetes significantly increased in NHBs, as compared to NHWs. Dyslipidemia prevalence substantially increased in both groups, but always remained higher in NHWs until 5 years post-transplant. The baseline and post-transplant control of hypertension, diabetes and LDL was significantly better in NHWs vs. NHBs at most time points. The use of medication classes to treat CVD risk factors and adherence to these medications also differed by race. NHBs were more likely to be prescribed anti-hypertensive therapy but, in general, had lower rates of adherence to these therapies. This was also demonstrated for post-transplant insulin therapy. Post-transplant statin prescribing was similar, but adherence to this therapy was lower in NHBs. Figure S1 and figure S2 display the data from Table 2 in graphical format.

Table 2.

CVD risk factor prevalence and control, compared between NHB and NHW adult kidney transplant recipients

CVD Variable	Baseline		One Year		Three Years		Five Years
	NHW	NHB	NHW	NHB	NHW	NHB	NHW	NHB
Prevalence of Hypertension†	92.3%	96.4%‡	97.2%	99.0%‡	98.7%	99.9%‡	99.2%	100.0%‡
Prevalence of Diabetes§	38.4%	38.8%	46.9%	50.3%	49.4%	52.9%	53.3%	58.9%‡
Prevalence of Dyslipidemia‖	46.4%	40.7%‡	69.2%	63.3%‡	81.4%	77.2%‡	86.3%	83.9%
Mean Systolic BP (±SD)	140±18	141±19	137±14	139±15‡	136±12	138±12*	135±11	137±12‡
Mean Diastolic BP (±SD)	75±10	79±11‡	73±9	76±9‡	73±8	76±8‡	73±8	77±8‡
Blood Pressure <140/90 mmHg	51%	46%*	60%	52%‡	65%	58%‡	69%	60%‡
Mean Hemoglobin A1c (±SD)	6.6±1.3	6.5±1.5	7.1±1.	7.8±1.7‡	7.2±1.3	7.8±1.6‡	7.2±1.2	7.7±1.6‡
Hemoglobin A1c <7%¶	57%	61%	53%	35%‡	46%	31%‡	47%	35%‡
Hemoglobin A1c <8%¶	92%	93%	87%	78%‡	83%	74%‡	86%	83%
Mean LDL (±SD)	86±32	86±31	100±34	103±34	97±30	102±32‡	95±28	99±30‡
LDL <100 mg/dL	72%	70%	52%	51%	56%	52%*	61%	55%‡
Mean TG (±SD)	200±149	164±105‡	198±150	161±104‡	194±126	160±98‡	187±120	154±86‡
TG < 150 mg/dL	46%	57%‡	44%	59%‡	43%	57%‡	45%	60%‡
Prescribed ACE inhibitor or ARB#	48%	50%	35%	38%*	50%	54%*	56%	61%‡
MPR ≥80%	23%	18%*	56%	49%*	53%	51%	52%	40%‡
Prescribed Beta Blocker#	49%	51%	59%	64%*	70%	73%*	73%	78%*
MPR ≥80%	34%	23%‡	61%	56%*	56%	47%‡	51%	46%
Prescribed Oral Anti-hyperglycemic¶	34%	35%	30%	33%	36%	38%	39%	42%
MPR ≥80%	17%	19%	60%	49%*	44%	47%	49%	43%
Prescribed Insulin¶	48%	44%	62%	66%	70%	73%	70%	74%
MPR ≥80%	69%	45%‡	86%	78%‡	83%	82%	84%	76%*
Prescribed Statin**	84%	80%*	74%	70%	77%	76%	78%	78%
MPR ≥80%	37%	22%‡	62%	52%‡	57%	50%*	58%	43%‡
Prescribed Other Anti-Lipemic**	19%	17%	13%	7%‡	19%	12%‡	20%	12%‡
MPR ≥80%	18%	7%‡	58%	60%	46%	43%	51%	40%
Prescribed Anti-platelet	10%	7%‡	7%	5%	10%	7%‡	13%	9%‡
MPR ≥80%	27%	21%	51%	42%	51%	35%*	44%	24%

^*p<0.05

^‡p<0.01

^†Defined as documentation, blood pressure >140/90 mmHg or treatment with anti-hypertensive therapy

^§Defined as documentation, hemoglobin A1c >7% or treatment with anti-glycemic therapy

^‖Defined as a LDL >130 mg/dL, TG >150 mg/dL or treatment with anti-lipemic therapy

^¶Only in those with a diagnosis of diabetes

^#Only in those with a diagnosis of hypertension

^**Only in those with a diagnosis of dyslipidemia

The fully adjusted model for the primary outcome of graft loss is displayed in Table 3, demonstrating that after adjusting for all baseline and follow up variables, including CVD risk and control, NHB race continued to be a significant risk factor (HR 1.49, 1.11–1.99, p=0.0072). The fully adjusted model reduced the risk associated with NHB race by 26%. Beyond race, there were seven variables that were significant independent predictors of graft loss. This included receiving an expanded criteria donor kidney (HR 1.39, 1.01–1.91, p=0.0455), receiving and being adherent to ACE inh/ARB therapy (HR 0.45, 0.29–0.68, p=0.0002), being non-adherent to diuretic therapy (HR 1.96, 1.49–2.57, p<0.0001), being non-adherent to insulin therapy (HR 1.59, 1.02–2.47, p=0.0414), receiving and being adherent to statin therapy (HR 0.53, 0.36–0.80, p=0.0022), uncontrolled hypertension (HR 1.64, 1.29–2.07, p<0.0001) and developing acute allograft rejection (HR 1.91, 1.46–2.51, p<0.0001).

Table 3.

Fully adjusted competing risk model for the primary outcome of graft loss

Variable	Reference	Hazard Ratio	95% CI	p-Value
NHB Race	NHW Race	1.490	1.11–1.99	0.0072
Age (per year)	18 years old	0.997	0.99–1.01	0.6683
Female Gender	Male Gender	1.132	0.52–2.48	0.7570
Not Married	Married	0.968	0.76–1.23	0.7933
Angina or CAD History	No Cardiac History	0.947	0.68–1.33	0.7550
Previous Transplant	No Previous Transplant	1.255	0.68–2.33	0.4705
Living Donor	Deceased Donor	0.897	0.67–1.20	0.4581
Expanded Criteria Donor	Standard Deceased Donor	1.387	1.01–1.91	0.0455
Cardiac Death Donor	Brain Dead Donor	1.150	0.62–2.12	0.6536
NHB Donor	Any Other Donor Race	1.087	0.79–1.50	0.6088
HLA Mismatches – 0	1–5 HLA Mismatches	0.964	0.64–1.46	0.8626
MLA Mismatches – 6		1.204	0.86–1.68	0.2784
PRA 0%	PRA 1–20%	1.056	0.73–1.53	0.7715
PRA >20%		0.977	0.50–1.91	0.9463
Years on Dialysis	0 Years	1.044	1.00–1.09	0.0739
Preemptive Transplant	On Dialysis	0.896	0.63–1.28	0.5474
IL-2 Receptor Antagonist Induction	No Induction Therapy	0.986	0.74–1.31	0.9223
Cytolytic Induction		0.916	0.69–1.22	0.5525
Tacrolimus Maintenance Therapy	No Calcineurin Inhibitor Therapy	0.720	0.44–1.17	0.1877
Cyclosporine Maintenance Therapy		0.845	0.51–1.41	0.5186
Mycophenolate Maintenance Therapy	No Adjunct Agent	0.935	0.65–1.35	0.7213
Azathioprine Maintenance Therapy		0.762	0.29–2.02	0.5849
mTOR Maintenance Therapy	No mTOR	0.755	0.49–1.17	0.2046
Corticosteroids Maintenance Therapy	No Corticosteroids	1.422	0.75–2.68	0.2764
ACE inh/ARB Therapy MPR 0–79%	No ACE inhibitor or ARB Therapy	0.959	0.74–1.24	0.7455
ACE inh/ARB Therapy MPR ≥80%		0.446	0.29–0.68	0.0002
Beta Blocker Therapy MPR 0–79%	No Beta Blocker Therapy	1.363	0.99–1.88	0.0599
Beta Blocker Therapy MPR ≥80%		0.934	0.64–1.37	0.7276
Calcium Channel Blocker MPR 0–79%	No Calcium Channel Blocker Therapy	1.238	0.93–1.65	0.1488
Calcium Channel Blocker MPR ≥80%		0.946	0.64–1.39	0.1488
Diuretic Therapy MPR 0–79%	No Diuretic Therapy	1.956	1.49–2.57	<0.0001
Diuretic Therapy MPR ≥80%		1.025	0.63–1.67	0.9207
Anti-Platelet Therapy MPR 0–79%	No Anti-Platelet	1.198	0.83–1.73	0.3376
Anti-Platelet Therapy MPR ≥80%		1.218	0.64–2.32	0.5486
Insulin Therapy MPR 0–79%	No Insulin Therapy	1.585	1.02–2.47	0.0414
Insulin Therapy MPR ≥80%		1.220	0.80–1.86	0.3522
Oral Anti-Hyperglycemic Therapy MPR 0–79%	No Oral Anti-Hyperglycemic Therapy	1.191	0.85–1.66	0.3021
Oral Anti-Hyperglycemic Therapy MPR ≥80%		0.766	0.40–1.48	0.4258
Statin Therapy MPR 0–79%	No Statin Therapy	0.836	0.62–1.14	0.2529
Statin Therapy MPR ≥80%		0.534	0.36–0.80	0.0022
Other Dyslipidemia Therapy MPR 0–79%	No Other Dyslipidemia Therapy	1.240	0.89–1.73	0.2088
Other Dyslipidemia Therapy MPR ≥80%		1.464	0.80–2.67	0.2149
Diabetes	No Diabetes	0.657	0.43–1.01	0.0541
Dyslipidemia	No Dyslipidemia	0.851	0.56–1.29	0.4454
Uncontrolled Dyslipidemia	Dyslipidemia Controlled	1.083	0.83–1.42	0.5607
Uncontrolled Diabetes	Diabetes Controlled	0.831	0.58–1.19	0.3167
Uncontrolled Hypertension	Hypertension Controlled	1.636	1.29–2.07	<0.001
Delayed Graft Function	No Delayed Graft Function	0.925	0.68–1.27	0.6248
Acute Rejection	No Acute Rejection	1.914	1.46–2.51	<0.001

Table 4 summarizes the results for the sequential modelling for the primary outcome of graft loss with varied entry of the blocks of variables through iterative processes. The results demonstrate that donor characteristics (−13.4 to −4.2%), immunologic risks (−9.6 to −4.7%), CVD risks and control (−17.5 to −8.7%), and post-transplant events (−5.5 to −0.7%) had the largest impact on reducing the independent risks associated of NHB race on graft loss. CVD risk factors and risk control produced the largest reduction in the model AIC, suggesting that including these variables in the model substantially improved the model performance in capturing the variability associated with graft loss by enhancing model goodness of fit. Figure 2 displays the cumulative incidence estimates for graft loss, comparing these curves between NHBs vs. NHWs for the unadjusted model (Figure 2A) and the fully adjusted model (Figure 2B). After full adjustment, the estimated incidence curves significantly converge.

Table 4.

Comparison of hazard ratios for NHBs vs. NHWs and assessment of model fit across all sequential competing risk modeling analyses for the primary outcome of graft loss

Model	Domain	Fixed Entry				Varied Entry*
		HR for NHB vs. NHW	95% CI	p-Value	Model AIC	Relative change in NHB HRs vs. NHW as compared to previous model	Change in AICs from previous model
Model 1	Race Only	1.999	1.61–2.49	<0.0001	4868	NA	NA
Model 2	+Sociodemographics	2.054	1.64–2.57	<0.0001	4875	−2.8 to 2.1%	5 to 7
Model 3	+Comorbidities	2.057	1.65–2.57	<0.0001	4877	0.1 to 1.3%	2 to 4
Model 4	+Donor Characteristics	1.781	1.36–2.32	<0.0001	4870	−13.4 to −4.2%	−7 to 3
Model 5	+Immunologic Risks	1.697	1.30–2.22	0.0001	4878	−9.6 to −4.7%	6 to 8
Model 6	+Immunosuppression	1.691	1.29–2.22	0.0001	4885	−0.4 to 0.7%	7 to 9
Model 7	+CVD Risk Factors & Control	1.501	1.12–2.01	0.0059	4773	−17.5 to −8.7%	−118 to −106
Model 8	+Post-Transplant Events	1.490	1.11–1.99	0.0072	4757	−5.5 to −0.7%	−28 to −16

^*Varied entry was conducted using iterative modeling and changing the introduction of the blocks of variables into the model in order to determine their impact on racial disparities. For all scenarios, race was initially entered into the model (unadjusted risk), followed by entry of blocks of variables which varied across different iterations.

Figure 2.

Cumulative incidence function curve estimates for the primary outcome of graft loss in NHB vs. NHW kidney transplants for the unadjusted model (Figure 2A) and the fully adjusted model (Figure 2B) demonstrating a 24% relative reduction in risk for NHBs when controlling for all measured variables

The results for the secondary outcomes of mortality are displayed in table S1 (full model), table S2 (sequential modeling) and figure S3 (unadjusted and fully adjusted survival estimates). The full model for mortality demonstrated that NHB race was a significant protective factor for risk of post-transplant death (HR 0.75, 0.60–0.94, p=0.0132). Additional independent risk factors for death included not being married, deceased donor, mycophenolate therapy, mTOR therapy, non-adherence to CVD medications, the presence of diabetes and uncontrolled hypertension (table S1). Sequential modeling for death demonstrated that sociodemographics (predominantly age) were a significant suppressor, while donor characteristics, immunologic risks and CVD risk factors and control were all significant explanatory variables. Similar to modeling for other outcomes, CVD risk factors and control produced the largest reduction in AIC, suggesting it explains a substantial portion of the variability associated with the risk of death after transplant.

Five sensitivity analyses were performed to ensure robustness of the estimates from the modelling. This included varied entry of blocks of variables into the sequential modelling, using standard Cox regression for the outcome of graft loss, modeling random and fixed effects for CVD risk factor control variables through joint modeling, assessing BP control with and without anti-hypertensives included in the modelling and using multiple imputation for missing data. The results of these analyses demonstrated consistent estimates, trends and results. The full model for the outcome of graft loss with the imputed missing data (n=4,643) is displayed in table S4, with the rate of missingness in table S3. Race continued to be an independent risk factor for graft loss and after adjustment, the reduction in risk for NHBs was similar to the analysis using cases with complete data (25.5% vs. 27.8%); in the imputed model, similar to the model using cases with complete data, CVD risk factors and CVD risk control had the largest impact on model performance and significantly reduced the independent influence of NHB race on graft loss. In the final sensitivity analysis, the estimated impact of uncontrolled hypertension on graft loss was similar in both longitudinal and means models, with and without the inclusion on anti-hypertensive medications (means model with anti-hypertensive medications HR 1.64 (1.29–2.07), without anti-hypertensive medications HR 1.79 (1.42–2.25); longitudinal model with anti-hypertensive medications intercept HR 3.04 (1.73–5.33), slope HR 3.20 (1.07–9.54, without anti-hypertensive medications intercept HR 3.12 (1.80–5.40) slope HR 3.61 (1.24–10.54).

DISCUSSION

The results of this study demonstrate that NHB transplant recipients have a considerably larger burden of CVD risk factors, which meaningfully contribute to racial disparities in graft outcomes. Specifically, NHBs have a higher prevalence of hypertension and post-transplant diabetes with reduced control of hypertension, diabetes and LDL, when compared to NHWs. In addition, adherence to medications used to manage CVD risk factors appears to be significantly lower in NHBs, as compared to NHWs. These data provide novel information to demonstrate that focusing on improving the management of CVD risk factors within kidney transplant recipients offers a promising mechanism to both enhance graft outcomes for the entire population while also potentially impacting racial disparities. To our knowledge, this is the first study to comprehensively assess CVD risk factors and risk control in a large national cohort of contemporary U.S. kidney transplant recipients and specifically assess the impact of CVD risk factors on racial disparities.

Racial disparities for outcomes in kidney transplant recipients have been long-standing and well-documented, with the literature exposing this issue dating back to the late 1970s, when kidney transplantation was still considered an experimental procedure. Recent evidence suggests that over the past 20 years, the absolute disparity in graft survival has slightly improved. Yet, in relative terms, NHBs are still at substantially higher risk of late graft loss (5 year relative risk of graft loss is 2.2 in living donors and 1.8 in deceased donors, comparing NHBs to NHWs).⁴⁹ Some aspects that contribute to this disparity have been well-studied. It is clear that NHBs have unique immunologic characteristics that increase their risk of acute rejection and have socioeconomic disadvantages which create barriers to access to transplant, access to living donors and access to optimal post-transplant care.^12,50 Thus, to date, most of the interventions designed to reduce this disparity have focused on using potent immunosuppression to mitigate immunologic risk and improve access to deceased and living donor transplants.²⁵ There have been some documented successes in this capacity. Recent evidence suggests that acute rejection rate disparities in NHBs have been mostly resolved and access to deceased donor organ transplant is improving in NHBs. However, improving living donation rates within the NHB community has been an area of marginal regional success, with national data demonstrating reduced rates of living donation within NHBs, when compared to NHWs.^49,51 Our results corroborate these previous findings, in that immunologic risks and donor characteristics continue to be significant explanatory variables for racial disparities.

There is paucity in the research assessing CVD risk factors and risk factor control as a potential cluster of explanatory factors for racial disparities in kidney transplant outcomes. This is despite the fact that these risks are decidedly mutable and that there is strong evidence to suggest these issues drive disparate outcomes in the general non-transplant population with diabetes and hypertension.^52,53 Previous single-center studies from our research group has demonstrated similar results, and the data from this analysis further support these findings.^38,39 Hypertension, diabetes and LDL have significantly lower rates of control post-transplant in NHBs, as compared to NHWs. This is despite the increased use of anti-hypertensives in NHBs. The higher rates of medication non-adherence in these patients likely contribute to this issue, and represents an opportunity for future interventions designed to improve CVD risk factor control in NHBs. It is interesting to note that our study only included veterans obtaining medications through the VA system. Thus, the issues surrounding access and cost of medications should not be a significant factor leading to medication non-adherence in this population. Other factors, included regimen complexity, health literacy and self-efficacy may be producing the higher rates of non-adherence that were seen within NHBs in this study.⁵³ We focused on graft loss after the first year post-transplant, because previous studies have demonstrated that racial disparities in transplantation occur later post-transplant and CVD risk factors are most likely to influence these later events.

The prescribing of two medication classes, ACE inh/ARBs and statins, demonstrated a substantial protective benefit for graft loss and death. In our study, only about 50–60% of the population was prescribed ACE inh/ARBs at five years post-transplant; and of these, only 40–50% were adherent (MPR ≥80%). For statins, about three-quarters were prescribed this therapy by five years and 40–60% were adherent. One could argue that due to the prevailing CVD and CVD risks that are highly correlated with CKD, these therapies should be prescribed to nearly all kidney transplant recipients that do not have absolute contraindications.⁵⁴ Yet, data from this study and other non-VA national and single-center studies of kidney transplant recipients demonstrate the utilization of these therapies is on the rise, but remains substantially below optimal levels.^55,56 Given the body of evidence in the general non-transplant population and the inherent CVD risk within the kidney transplant population, these results provide evidence to support future interventional studies designed to optimize utilization and adherence to these therapies in kidney transplant recipients as a mechanism to improve long-term graft and patient outcomes.^54–56

It is also noteworthy that non-adherence to a number of medication classes was a significant and independent risk-factor for both graft loss and death. This included diuretics, insulin, beta blockers, calcium channel blockers and anti-platelet therapy. This risk was independent of CVD disease and CVD risk factor control. These results may be a true cause and effect relationship, in that medication non-adherence to these therapies may actually lead to increased rates of graft loss and death, or could also be a proxy for other risks that are highly correlated to medication non-adherence and which were not measured as a part of this study.⁵⁷ In particular, these include social determinants of health (finances, living conditions, social support), health literacy and self-efficacy.⁵⁸ Thus, further research is needed to ascertain whether non-adherence to these medication classes truly causes increased rates of graft loss and death in kidney transplant recipients.

There are a number of limitations to this analysis that are worthy of discussion. First, this analysis was confined to a veteran transplant population and the number of female recipients is quite low. Thus, generalizability of these results to female transplant recipients is not warranted. Using a veteran population also restricts the ability of this analysis to assess the impact of insurance coverage and access to care on outcomes. However, previous studies have demonstrated that racial disparities in kidney transplant recipients are similar in magnitude between VA and non-VA populations, and using VA data will allow us to include numerous clinical variables that are missing or not available from previous studies that solely utilized the USRDS or CMS datasets.²⁸ Therefore, for the purposes of this analysis, which focuses on racial disparities as it relates to CVD risk factor control, the benefits provided by the use of VA data greatly outweigh the limitations of solely focusing on veteran patients.

Another limitation of this analysis is its retrospective design, which may increase the risk for confounding and misclassification, potentially biasing the results. However, because we were able to link two databases with unique and overlapping data in a longitudinal format, we dramatically minimized these risks. We validated data elements by comparing them between the two databases for accuracy and were able to include the largest number of covariates ever reported in a transplant racial disparities analysis.⁴⁰ In addition, retrospectively accurately assessing control of CVD risk factors is a difficult endeavor, and using means during the entire follow up may not truly capture the time dependency of this exposure. However, using the means does allow for a more straightforward analysis and presentation of the data. To assess if longitudinal analyses improve the primary exposure classification, we also conducted a separate analysis, using a time-varying analysis technique to assess for the fixed and random effects (intercept and trajectory) of the BP, A1c, LDL and TG on the outcomes.⁵⁹ The results of this assessment were similar with respect to direction and magnitude; thus, for ease of display and interpretation, we utilized the mean values, which appear to be an accurate reflection of CVD risk control. Another limitation with this analysis is that we fail to measure and account for biologic and socioeconomic differences between NHB and NHW transplant recipients that may explain some of the disparity. This includes genetic variants that are more prevalent in NHB recipients, including cytochrome P450 3A5 and apolipoprotein L1 and the aforementioned social determinants of health.^60–63 Although we recognize there are a number of limitations with using a veteran dataset, the overall objectives of this analysis require the use of longitudinal clinical data that is currently unavailable within any other national dataset accessible at this level of comprehension and detail.

Perspectives

In summary, these results demonstrate that NHB kidney transplant recipients have significantly higher rates of CVD risk factors and reduced CVD risk control, as compared to NHWs. These issues may be partly related to medication non-adherence, meaningfully contribute to disparities for graft outcomes within NHBs and represent a promising area for future interventional studies designed to improve CVD risk factor control as a mechanism to optimize graft outcomes and reduce racial disparities in kidney transplantation.

Novelty and Significance.

1. What is new: the results provide new insights demonstrating that disparities in outcomes for African American kidney transplant recipients are significantly influenced by cardiovascular disease risk factors and risk factor control; in particular, the treatment and control of hypertension and adherence to medications used to manage this condition impact these disparities.

2. What is relevant: improving the management and control of cardiovascular risk factors, particularly hypertension, is likely to significantly improve outcomes in African American kidney transplant recipients and reduce disparities in graft survival rates.

3. Summary: these results demonstrate that African American kidney transplant recipients have significantly higher rates of cardiovascular disease risk factors and reduced cardiovascular disease risk control, as compared to Caucasians. These issues may be partly related to medication non-adherence and meaningfully contribute to disparities for graft outcomes.