Role of Age and Competing Risk of Death in the Racial Disparity of Kidney Failure Incidence after Onset of CKD

Visual Abstract

Abstract

Significance Statement

Black adults in the United States have 2–4 times higher incidence of kidney failure than White adults. Yet, the reasons underlying this disparity remain poorly understood. Among 547,188 US veterans with new-onset CKD, according to a new race-free GFR equation, Black veterans had a 2.5-fold higher cumulative incidence of kidney failure, compared with White veterans, in any follow-up period from CKD onset. This disparity resulted from a combination of higher hazards of progression to kidney failure and lower hazards of competing-risk death in Black veterans. Both, in turn, were largely explained by the younger age at CKD onset in Black veterans, underscoring an urgent need to prevent early onset and slow progression of CKD in younger Black adults.

Background

The Black adult population is well known to have higher incidence of kidney failure than their White counterpart in the United States, but the reasons underlying this disparity are unclear. We assessed the racial differences in kidney failure and death from onset of CKD on the basis of the race-free 2021 CKD Epidemiology Collaboration equation and examined the extent to which these differences could be explained by factors at the time of CKD onset.

Methods

We analyzed a national cohort consisting of 547,188 US veterans (103,821 non-Hispanic Black and 443,367 non-Hispanic White), aged 18–85 years, with new-onset CKD between 2005 and 2016 who were followed through 10 years or May 2018 for incident kidney failure with replacement therapy (KFRT) and pre-KFRT death.

Results

At CKD onset, Black veterans were, on average, 7.8 years younger than White veterans. In any time period from CKD onset, the cumulative incidence of KFRT was 2.5-fold higher for Black versus White veterans. Meanwhile, Black veterans had persistently >2-fold higher hazards of KFRT throughout follow-up (overall hazard ratio [95% confidence interval], 2.38 [2.31 to 2.45]) and conversely had 17%–48% decreased hazards of pre-KFRT death. These differences were reduced after accounting for the racial difference in age at CKD onset.

Conclusions

The 2.5-fold higher cumulative incidence of kidney failure in Black adults resulted from a combination of higher hazards of progression to kidney failure and lower hazards of the competing risk of death, both of which can be largely explained by the younger age at CKD onset in Black compared with White adults.

Introduction

For more than four decades, it has been consistently observed that Black adults in the United States have 2–4 times higher incidence of kidney failure than White adults.^1,2 In 2018, the adjusted incidences of kidney failure with replacement therapy (KFRT) for Black and White adults were reported to be 834 and 313 new cases per million, respectively.² Despite extensive research, the underlying drivers of this disparity remain unclear.

Previous studies that examined the racial difference in kidney failure have been limited by the use of prevalent populations and outdated GFR estimating equations, both of which can induce bias. For example, many prior studies used US general adult populations,^3–5 community-based populations,^6,7 or high-risk populations that might or might not have had CKD at baseline.^8–10 Thus, the reported rates did not measure disparities in patients with CKD. A few other studies reported disparities in the rates of kidney failure among patients with CKD,^11,12 but their use of prevalent patients with CKD is a limitation. Using prevalent patients can only provide a partial, not full, view of the disparity because of missing the early course of CKD. Furthermore, using prevalent patients may yield biased estimates of the disparity because outcome events before study entry (such as early death with CKD) are not captured and thus not included as an integral part of the overall disparity. Finally, prior studies have relied on outdated GFR estimating equations that included race terms, causing overestimation of GFR in Black adults. For instance, the seminal study by Choi et al.¹¹ included patients with a broad range of eGFR but used the Modification of Diet in Renal Disease (MDRD) equation that assigned a 21% higher eGFR for Black adults than non-Black adults at the same serum creatinine.¹³ These limitations may substantially change the inference compared with newer equations that do not include race.^14,15

Moreover, racial differences in risk factors at the time of CKD onset may contribute to the subsequent racial difference in developing kidney failure, but this has not been well studied because prior studies on prevalent patients lacked patient information at the time of CKD onset. Overall, examining outcome events from a common starting point anchored at the onset of CKD using contemporary GFR estimating equations can provide an accurate and complete view of the disparity in the entire course of the disease, help identify when disparities in events occur, and determine early potentially modifiable factors behind these disparities.

Using our recently established nationwide cohort of veterans with new-onset CKD defined using the updated race-free 2021 CKD Epidemiology Collaboration (CKD-EPI) creatinine equation,¹⁶ we, therefore, performed this study to (1) assess the racial difference in cumulative incidence of KFRT from the time of CKD onset and determine whether the difference persists or changes over the time course of CKD; (2) update prior analyses that examined racial differences drawn from previous race-based equations, including a potential role of competing risk of death; and (3) assess how much of the racial difference in KFRT could be explained by differences in demographic or other factors at CKD onset.

Methods

Data Source and Study Cohort

This was a retrospective cohort study of veterans with new onset of CKD identified in the US veterans Health Administration (VHA), the largest integrated health system in the United States providing care at 1255 health care facilities for nine million veterans nationwide each year.^17,18 This analysis used data from the VHA Corporate Data Warehouse, VHA Vital Status File, and VHA/Centers for Medicare & Medicaid Services (CMS)/United States Renal Data System (USRDS) Data for Research. This study was approved by the institutional review board at the University of Virginia and the joint institutional review board at the University of Utah and the veterans Affairs Salt Lake City Healthcare System.

There were 7.9 million eligible veterans in the VHA, either non-Hispanic White or non-Hispanic Black, who had at least one serum creatinine measurement between January 1, 2000, and May 30, 2018. eGFR values were calculated using the 2021 CKD-EPI creatinine equation that does not include race¹⁶ for all serum creatinine measurements; those measured in inpatient acute care settings were excluded.¹⁹ From this source population, we identified veterans with new onset of CKD stage G3 or higher between January 1, 2005, and May 30, 2016. Onset of CKD was determined by the earliest pair of two eGFR values <60 ml/min per 1.73 m² that met two conditions: (1) They were measured ≥91 days apart²⁰ but within 18 months and (2) the second (confirmatory) one, which served as the index date of the CKD onset, had to be the initial eGFR value assessed between 91 days and 18 months from the first one. To maximize validity of the index date in representing the veteran's first lifetime occurrence of decreased eGFR lasting >3 months (i.e., CKD G3 or higher), eligible veterans had to be in the VHA system for at least 2 years before their first documented eGFR value <60 ml/min per 1.73 m² in the VHA database. With this criterion, during this prior period of >2 years, 92.4% of veterans in the cohort had all prior eGFR values ≥60 ml/min per 1.73 m² and the remaining 7.6% had no eGFR measurements. veterans who had reached KFRT, as determined by the USRDS records,² before the index date were also excluded.

The main analyses excluded veterans (1.8%) with missing baseline data on body mass index (BMI) or smoking status as well as veterans older than 85 years (8.7%) assessed at the index date because very few people advanced to KFRT during follow-up when CKD first occurred at this advanced age. After exclusions, 547,188 veterans aged 18–85 years at CKD onset were included and followed through May 31, 2018, by which time KFRT records were available through linkage to the USRDS² (see the flowchart for the cohort construction in Supplemental Figure 1).

Follow-Up, Outcomes, Exposure, and Covariates

The follow-up period started at the index date and ended at the first occurrence of the following events: incident KFRT, death, 10 years, or the end of the study period (May 31, 2018). Incident KFRT was defined as initiation of dialysis or receipt of kidney transplantation on the basis of USRDS records. Death status was based on the VHA Vital Status File and VHA Corporate Data Warehouse Patient Domain that integrate death data from multiple official sources inside and outside the VHA. The two primary outcomes for this report were incident KFRT and death before KFRT (i.e., pre-KFRT death).

The exposure was self-reported race and ethnicity, classified as non-Hispanic White or non-Hispanic Black, ascertained using all patient encounters in records of both VHA and the CMS (through linkage to CMS). Baseline covariates were grouped into (1) age at CKD onset (assessed at index date), sex, and calendar year of CKD onset; (2) eGFR (at index date) and BMI; (3) 13 comorbidities (hypertension, diabetes, heart failure, coronary artery disease, cardiac dysrhythmia, other cardiac diseases, cerebrovascular accident or transient ischemic attack, peripheral vascular disease, chronic obstructive pulmonary disorder, anemia, cancer, gastrointestinal bleeding disorders, and liver disease) and individual uses of three medication classes (angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, and statins); (4) smoking status (never/former/current), alcohol abuse, and drug abuse (herein called behavioral factors); and (5) Area Deprivation Index (ADI). ADI is a composite measure of socioeconomic disadvantage at the Census block group neighborhood level, which was developed using the 2015 American Community Survey 5-year data.^21,22 The ADI ranges from 1 to 100, with higher values indicating higher levels of neighborhood deprivation. The nine-digit ZIP code crosswalk was built to correspond directly to Census block groups.²² Because the VHA database only includes residential five-digit ZIP codes, we calculated ADI at five-digit ZIP codes using the median of all ADI scores at nine-digit ZIP codes within each five-digit ZIP code.²³ The comorbidities, alcohol abuse, and drug abuse were ascertained on the basis of the presence of at least one International Classification of Disease code, Ninth and Tenth editions, in either VHA records or CMS claims.²⁴ Data on urine albumin-to-creatinine ratio (UACR) are presented, but were not included in regression analyses because they were available in only 40% of patients. All these covariates were defined using records during the 2-year baseline period, with the exception of smoking and UACR that were based on the last record before the index date. Finally, we also assessed eGFR levels at the time of KFRT initiation using the last eGFR value before the day of KFRT initiation for patients who progressed to KFRT.

Statistical Analyses

The main analyses included the following three sets. First, we examined racial differences in cumulative incidence of KFRT after CKD onset using the Aalen–Johansen estimator with pre-KFRT death treated as a competing risk.²⁵ Similarly, cumulative incidences of pre-KFRT death were also presented with occurrence of KFRT treated as a competing risk. Second, we conducted cause-specific hazard analysis to examine the racial differences in cause-specific hazards of progression to KFRT and pre-KFRT death. To capture changes over time, we used time-varying hazard analysis, in which the cause-specific hazard of outcome was modeled as a function of time. We focused on unadjusted cause-specific hazards (herein referred to as hazards) and cause-specific hazard ratios (HRs) (referred to as HRs, Black versus White veterans) to reflect actual risks experienced in the real-world setting. Finally, we examined how much of the racial differences in hazards of KFRT and of death could be explained by differences in factors at the time of CKD onset by performing the following eight cause-specific hazard regression models: (1) base model, adjusted for sex and calendar year of CKD onset; (2) base model+age at CKD onset; (3) base model+BMI+eGFR; (4) base model+comorbidities+medications; (5) base model+behavioral factors; (6) base model+ADI; (7) base model+all above factors but ADI; and (8) base model+all above factors, including ADI (full model). The ADI variable was included in the model as a four-categorical variable defined by its quartiles. To quantify the contribution of each factor set, we calculated the percentage of excess relative risk (hazard) of incident KFRT (and decreased relative risk of pre-KFRT death) in Black versus White veterans that was explained by that factor set. Excess relative risk was defined as HR minus 1.0, and decreased relative risk was defined as 1.0 minus HR, both denoted by Δ. We then calculated the explainable percentage using the formula (Δ_base−Δ_base+factor)/Δ_base×100%,^5,26,27 where Δ_base and Δ_base+factor are the values of Δ from the base model and each of the above models 2–8. In addition, we calculated the proportional reduction in the Wald chi-square statistic associated with the HR in each of the models 2–8 from the base model.

Because age at CKD onset differed greatly by race and was associated with both outcomes, two additional sets of analyses were performed. We constructed an age-matched cohort by retaining all White veterans and sampling a subset of Black veterans from the actual group of Black veterans such that the resulting Black group and the White group were similar in the distributions of age at CKD onset, sex, and year of CKD onset (see Supplemental Item 1). We then assessed the racial difference in KFRT for this age-matched cohort.

The second additional analysis addressed whether the age difference between races in our incident CKD cohort reflected similar age differences between races among overall US veterans (source population) or instead was related to different rates of incident CKD experienced by different age and racial groups in the VHA source population. In the former case, the racial difference in age at CKD onset would be very difficult to modify because it reflects Veteran demographic characteristics. On the other hand, if the latter scenario was indeed the case, a racial difference in age at CKD onset can be potentially modified by reducing group-specific rates of incident CKD. Thus, we examined two subpopulations in 2011 (a midpoint in the study period): (1) the subset of veterans from our incident CKD cohort who had CKD onset in 2011; those older than 85 years excluded from the main analyses were now included; and (2) all eligible veterans in the 2011 VHA system (Supplemental Item 2). We compared age differences between races for these two subpopulations and calculated race- and age-specific rates of incident CKD cases per 1000 VHA veterans.

Results

Baseline Characteristics

The main analyses included 103,821 Black and 443,367 White veterans with 96.8% being male (Table 1). At the time of CKD onset, more than half (57.9%) of the Black veterans were aged 18–65 years, in contrast to only 23.8% of White veterans aged 18–65 years (Figure 1), resulting in a mean age of 7.8 years younger at CKD onset (Table 1). The two groups had similar mean eGFR levels at the index date (51 ml/min per 1.73 m²) as a result of our incident cohort criteria. They also had similar prevalence of UACR >300 mg/g among those with available urine protein data. At CKD onset, the two groups had similar prevalence of hypertension (93%) and diabetes (53%), the two most common risk factors of CKD. Compared with White peers, Black veterans had a much lower prevalence of many other comorbidities, especially cardiovascular conditions, and were more likely to smoke, suffer from both alcohol and drug abuse, and live in areas with higher ADI. Finally, at the time of KFRT initiation, eGFR levels were on average lower in Black than White veterans (median: 12.2 and 15.6 ml/min per 1.73 m², respectively; Table 1).

Table 1.

Baseline characteristics of veterans with incident CKD by race

Characteristic	Overall (N=547,188)	Black (n=103,821)	White (n=443,367)
Age at CKD onset, yr, mean (SD)	71.0 (9.3)	64.7 (9.7)	72.5 (8.5)
Male, No. (%)	529,507 (96.8)	98,327 (94.7)	431,180 (97.3)
eGFR at CKD onset, ml/min per 1.73 m2, mean (SD)	51.3 (8.0)	51.0 (8.6)	51.4 (7.8)
eGFR at KFRT onset, ml/min per 1.73 m2, mean (SD)/median (IQR)a	19.0 (13.8)/14.3 (9.6–23.9)	16.8 (13.8)/12.2 (8.4–19.1)	20.2 (13.7)/15.6 (10.5–26.4)
BMI, kg/m2, mean (SD)	30.1 (6.0)	29.9 (6.3)	30.1 (6.0)
Comorbidities, No. (%)
Hypertension	508,738 (93.0)	96,917 (93.4)	411,821 (92.9)
Diabetes	287,570 (52.6)	56,125 (54.1)	231,445 (52.2)
Heart failure	148,290 (27.1)	22,167 (21.4)	126,123 (28.4)
Coronary artery disease	272,711 (49.8)	34,057 (32.8)	238,654 (53.8)
Cardiac dysrhythmia	214,522 (39.2)	27,181 (26.2)	187,341 (42.3)
Other cardiac diseases	203,197 (37.1)	29,085 (28.0)	174,112 (39.3)
CVA/TIA	145,318 (26.6)	19,855 (19.1)	125,463 (28.3)
PVD	183,733 (33.6)	24,548 (23.6)	159,185 (35.9)
COPD	189,490 (34.6)	26,579 (25.6)	162,911 (36.7)
Anemia	209,164 (38.2)	38,984 (37.5)	170,180 (38.4)
Cancer	142,551 (26.1)	23,949 (23.1)	118,602 (26.8)
GI bleeding disorders	79,952 (14.6)	14,733 (14.2)	65,219 (14.7)
Liver disease	31,028 (5.7)	6545 (6.3)	24,483 (5.5)
ACEi use, No. (%)	356,242 (65.1)	67,604 (65.1)	288,638 (65.1)
ARB use, No. (%)	90,217 (16.5)	17,615 (17.0)	72,602 (16.4)
Statins use, No. (%)	411,487 (75.2)	69,306 (66.8)	342,181 (77.2)
Smoking status, No. (%)
Never	117,776 (21.5)	22,529 (21.7)	95,247 (21.5)
Former	303,563 (55.5)	48,498 (46.7)	255,065 (57.5)
Current	125,849 (23.0)	32,794 (31.6)	93,055 (21.0)
Alcohol abuse, No. (%)	61,941 (11.3)	21,719 (20.9)	40,222 (9.1)
Drug abuse, No. (%)	41,420 (7.6)	20,342 (19.6)	21,078 (4.8)
UACR, No. (%)
<30 mg/g	133,942 (24.5)	24,504 (23.6)	109,438 (24.7)
30–300 mg/g	67,714 (12.4)	12,477 (12.0)	55,237 (12.5)
>300 mg/g	20,082 (3.7)	4760 (4.6)	15,322 (3.5)
Missing	325,450 (59.5)	62,080 (59.8)	263,370 (59.4)
ADI, median (IQR)	62 (42–78)	67 (45–84)	61 (42–77)

ACEi, angiotensin-converting enzyme inhibitors; ADI, Area Deprivation Index; ARB, angiotensin II receptor blockers; BMI, body mass index; COPD, chronic obstructive pulmonary disorder; CVA/TIA, cerebrovascular accident/transient ischemic attack; GI, gastrointestinal; IQR, interquartile range; KFRT, kidney failure with replacement therapy; PVD, peripheral vascular disease; UACR, urinary albumin-to-creatinine ratio.

^aOn the basis of the last eGFR measurement before the day of KFRT initiation. The median intervals between these two dates were 13 and 38 days, respectively, among 6443 Black and 11,608 White veterans who developed KFRT during the follow-up period.

Figure 1.

Age distribution of veterans at CKD onset by race. Texts for the percentages of veterans in the age group of 18–45 years at CKD onset, which were 2.6% in Black veterans and 0.36% in White veterans, were not shown because of insufficient space. Black veterans had a disproportionately larger number of veterans aged 65 years and younger at CKD onset than White veterans (57.9% and 23.8%, respectively).

Cumulative Incidences of KFRT and Pre-KFRT Death after CKD Onset

The cumulative incidence of KFRT in any time period from CKD onset was approximately 2.5-fold higher or 150% excess for Black compared with White veterans (Figure 2A). Specifically, the 10-year cumulative incidence of KFRT was 9.1% in Black and 3.6% in White veterans, representing 153% excess in the Black group. By contrast, the cumulative incidence of pre-KFRT death was much higher in both groups and was conversely approximately 30% lower in the Black group throughout the follow-up period (Figure 2B). Both incidences were age-dependent; as age decreased, the incidence of KFRT increased (Figure 3A) while the incidence of pre-KFRT death decreased (Figure 3B). However, the excess incidence of KFRT and the lower incidence of pre-KFRT death in Black veterans persisted in all age groups.

Figure 2.

Racial differences in the cumulative incidences of KFRT and pre-KFRT death over the course of 10 years from CKD onset. Relative excess (decrease) is the incidence ratio of Black to White minus 1.0. Compared with the White group, the Black group persistently had a 2.5-fold higher cumulative incidence of KFRT in any time period from CKD onset (A) while conversely had a lower cumulative incidence of pre-KFRT death (B). KFRT, kidney failure with replacement therapy.

Figure 3.

Racial differences in the cumulative incidences of KFRT and pre-KFRT death by age group. Y axis scales differ in A and B. Both the higher cumulative incidence of KFRT (A) and the lower cumulative incidence of pre-KFRT death (B) in Black compared with White veterans remained persistent in all age groups.

Cause-Specific Hazard Analysis for KFRT and Pre-KFRT Death

Over a median follow-up of 4.9 years, the overall hazard rate of KFRT (cases per 1000 patient-years) was 11.6 in Black and 4.9 in White veterans. Despite the differences in time-varying hazards (Figure 4A), the Black-to-White HR for KFRT exceeded 2.0 from the beginning and persisted throughout follow-up (overall unadjusted HR [95% confidence interval]: 2.38 [2.31 to 2.45]). In other words, Black individuals always had >2-fold increased risk of progressing to KFRT than White individuals at any given time point in the course of CKD progression among those who were alive and not yet developed KFRT. By contrast, the overall hazard rate of pre-KFRT death after CKD onset was lower for Black compared with White veterans (47.8 and 74.3 cases per 1000 patient-years, respectively). The time-varying hazards of pre-KFRT death ranged from 17% to 48% lower in the Black group throughout follow-up (overall unadjusted HR [95% confidence interval]: 0.64 [0.64 to 0.65]) (Figure 4B).

Figure 4.

Racial differences in the cause-specific hazards and cumulative incidences of KFRT and pre-KFRT death over time from CKD onset. Y axis scales differ in all panels; shaded areas represent the 95% CI bands of the hazard lines; HR, cause-specific HR for Black (or matched Black) veterans versus White veterans; relative excess (decrease) is the incidence ratio of Black to White minus 1.0. The hazard of progressing to KFRT was >2-fold greater (A) and the hazard of pre-KFRT death was conversely lower (B) at any given time point in disease progression for Black veterans compared with White veterans among those who remained alive without KFRT at that time point. The combination of the racial differences in these two hazards resulted in the excess cumulative incidence of KFRT among Black veterans throughout the follow-up period (C). The racial gap in cumulative incidence of KFRT decreased approximately 70% when the two racial groups were comparable in age at CKD onset (C). CI, confidence interval; HR, hazard ratio.

Baseline Factors Explaining the Racial Differences in Hazards of KFRT and Pre-KFRT Death

As presented in Table 2, for KFRT, adjusting for age at CKD onset reduced the Black-to-White HR from 2.42 to 1.41, due to the Black group being younger and younger age was associated with higher risk of KFRT. Statistically, age explained 71% of the excess relative risk of KFRT for Black versus White veterans in the model while other factor sets, including ADI, explained 1.4%–20% of the excess KFRT risk (models 3–6). All these factors but ADI collectively (model 7) explained 75% of excess KFRT risk, leaving 25% unexplained. For pre-KFRT death, age at CKD onset also explained 71% of the decreased relative risk of death among Black veterans. In the full model (model 8), higher ADI was significantly associated with high risk of death, but not with KFRT. Thus, while ADI explained a small fraction of the racial difference in death hazards, it did not explain further the excess KFRT risk in Black veterans after all the patient factors were taken into account. The results of the proportional reduction in the chi-square statistic similarly suggest that the racial difference in age at CKD onset was the strongest factor that explained the racial differences in both hazards of KFRT and pre-KFRT death.

Table 2.

Contributions of factors to racial differences in the hazards of KFRT and pre-KFRT death, controlling for sex and year of CKD onset

KFRT
Model	HR (95% CI)	Excess Relative Riska	Excess Relative Risk Explained (%)b	Chi-Square Statistic Associated with the HR	Proportional Reduction (%)
1. Base model, adjusted for sex and year of CKD onset	2.42 (2.35 to 2.50)	1.42	—	3222	—
2. Base model+age at CKD onset	1.41 (1.36 to 1.46)	0.41	71.1	415	87.1
3. Base model+BMI+eGFR	2.24 (2.17 to 2.31)	1.24	12.7	2672	17.1
4. Base model+comorbidities+medicationsc	2.23 (2.16 to 2.30)	1.23	13.4	2471	23.3
5. Base model+behavioral factorsd	2.14 (2.08 to 2.21)	1.14	19.7	2169	32.7
6. Base model+ADI	2.40 (2.32 to 2.47)	1.40	1.4	3097	3.9
7. Base model+all above factors but ADI	1.35 (1.31 to 1.40)	0.35	75.4	311	90.3
8. Base model+all above factors including ADI (full model)	1.35 (1.31 to 1.40)	0.35	75.4	304	90.6

Pre-KFRT Death
Model	HR (95% CI)	Decreased Relative Riske	Decreased Relative Risk Explained (%)b,f	Chi-Square Statistic Associated with the HR	Proportional Reduction (%)f
1. Base model, adjusted for sex and year of CKD onset	0.65 (0.64 to 0.66)	0.35	—	4197	—
2. Base model+age at CKD onset	0.90 (0.88 to 0.91)	0.10	71.4	253	94.0
3. Base model+BMI+eGFR	0.64 (0.63 to 0.65)	0.36	0	4531	0
4. Base model+comorbidities+medicationsc	0.77 (0.76 to 0.78)	0.23	34.3	1553	63.0
5. Base model+behavioral factorsd	0.62 (0.61 to 0.63)	0.38	0	4884	0
6. Base model+ADI	0.65 (0.64 to 0.65)	0.35	0	4379	0
7. Base model+all above factors but ADI	0.88 (0.87 to 0.89)	0.12	65.7	322	92.3
8. Base model+all above factors including ADI (full model)	0.87 (0.86 to 0.89)	0.13	62.9	363	91.4

ADI, Area Deprivation Index; BMI, body mass index; CI, confidence interval; HR, hazard ratio; KFRT, kidney failure with replacement therapy.

^aExcess relative risk was calculated by subtracting 1.0 from hazard ratio, denoted by Δ.

^b(Δ_base−Δ_base+factor)/Δ_base×100%.

^cComorbidities included hypertension, diabetes, heart failure, coronary artery disease, cardiac dysrhythmia, other cardiac diseases, cerebrovascular accident/transient ischemic attack, peripheral vascular disease, chronic obstructive pulmonary disorder, anemia, cancer, gastrointestinal bleeding disorders, and liver disease. Medications included use of angiotensin-converting enzyme inhibitors, use of angiotensin II receptor blockers, and statins use.

^dBehavioral factors included smoking status, alcohol abuse, and drug abuse.

^eDecreased relative risk was calculated by subtracting hazard ratio from 1.0, denoted by Δ.

^fThe value of zero indicates that the factor set that was examined did not explain the decreased relative risk of death in Black veterans after adjusting for sex and year of CKD onset. A smaller value of the percentage explained or proportional reduction in the expanded model (model 7 or 8) as compared with model 2 is possible because of the combined effect of protective factors and risk factors.

Effect of Age at CKD Onset on the Racial Difference in Cumulative Incidence of KFRT

To directly assess this, we used an age-matched cohort in which the matched Black group had the same mean age as the White group at 72 years and was older than the actual Black group by 7.6 years (Supplemental Table 1). As shown in Figure 4, A and B, racial differences in both hazards (KFRT and pre-KFRT death) were smaller throughout the follow-up period in the age-matched cohort compared with those in the actual cohort. Collectively, age matching reduced the racial gap in both the cumulative incidences of KFRT and pre-KFRT death by 70% (Figure 4, C and D). For example, the racial gap in the 10-year cumulative incidence of KFRT decreased to 47% excess in the matched Black group from the 153% excess in the actual Black group as compared with the White group (Figure 4C), representing 69% reduction.

Factors Accounting for Younger Age in the Black Group in Our Incident CKD Cohort

The age difference between races, with the Black group being younger, was much more pronounced in the incident CKD cohort than in the overall VHA Veteran population in 2011 (Supplemental Figure 2). In veterans with incident CKD in 2011, the proportion of veterans who were 65 years and younger at CKD onset was 61% in the Black group versus only 25% in the White group, whereas in the VHA population in 2011, the proportion of veterans 65 years and younger was 82% in the Black group versus 59% in the White group. Thus, the much greater age difference between races in veterans with incident CKD was attributed to the higher rates of incident CKD among younger Black versus younger White VHA veterans. As shown in Figure 5, the rates of incident CKD were >2-fold higher among VHA Black compared with VHA White veterans in all age groups 65 years and younger.

Figure 5.

Age- and race-specific rates of incident CKD cases per 1000 VHA veterans in 2011. In all age groups 65 years and younger, Black veterans in the VHA system experienced two-fold increased rates of incident CKD compared with their VHA White age counterparts. By contrast, in age groups older than 80 years, Black veterans experienced decreased rates of incident CKD. Of note, the two racial groups had similar overall rates of incident CKD (the right most 2 bars). VHA, veterans Health Administration.

Discussion

In this national study of veterans with incident CKD, we report several important findings. First, over the course of 10 years after CKD onset, Black veterans persistently had 2.5-fold higher cumulative incidence of KFRT than their White counterparts. Second, for all times over the follow-up period, the higher hazards of KFRT in Black relative to White patients were accompanied by lower hazards of pre-KFRT death in Black relative to White patients. Third, the racial difference in age at CKD onset, with Black veterans being 7.8 years younger on average, accounted for 70% of the excess cumulative incidence of KFRT because of the robust associations of age with both hazards of KFRT and pre-KFRT death. This substantial age gap at CKD onset stemmed from the two-fold increased rates of incident CKD among younger Black compared with White VHA veterans. These updated findings on racial disparities using an incident cohort and a contemporary GFR equation suggest the need to focus on early prevention to reduce the stark racial difference in age at CKD onset.

Our results build on several prior studies conducted in the US general or community-based populations^3–10 or prevalent patients with CKD.^11,12 During 7.1 years of median follow-up for prevalent patients with CKD in the Chronic Renal Insufficiency Cohort, 30% of Black patients and 15% of White patients developed KFRT.¹² These KFRT incidence rates were higher than the rates in our incident patients and reflect the progression to KFRT from moderate or advanced CKD because Chronic Renal Insufficiency Cohort baseline eGFRs were lower (mean 43.7 and 46.2 ml/min per 1.73 m² for Black and White participants, respectively). Choi et al.¹¹ assessed the disparities in KFRT and death in veterans with a broad range of eGFR on the basis of the MDRD GFR equation and a single creatinine measurement. They reported that the hazards of KFRT among Black veterans exceeded those among White veterans in all CKD stages. However, veterans who entered the study with eGFR <60 ml/min per 1.73 m² might have had a prior eGFR <60, reflecting a prevalent study design and inherent limitations as previously noted. By using time-varying hazard analysis and implementing a uniform disease entry at CKD onset for all patients, our study fills an important gap in the available literature by demonstrating that a large racial gap in the incidence of KFRT was present from CKD onset and forward throughout the entire duration of CKD. This new finding highlights the importance of recognition and intervention early in the disease continuum, particularly in younger Black adults.

Our finding of the higher hazard of progression to KFRT among Black veterans after CKD onset is, in fact, consistent with previous reports.^{8–12,28–31} Given the similar initial eGFRs of the two racial groups in our cohort, we can attribute this higher hazard of KFRT to the higher rate of rapidly declining kidney function among Black veterans, whereas those prior studies were difficult to conclude because of the differences in baseline kidney function between racial groups. However, the current finding of the lower pre-KFRT mortality in Black versus White veterans even after adjustment is contrary to the higher or similar pre-KFRT mortality in Black versus White participants in many prior studies,^9,11,30 although not all.³¹ All earlier studies, including that by Choi et al.,¹¹ calculated eGFR using previous GFR equations that included race terms (e.g., MDRD¹³ or the 2009 CKD-EPI equation³²), which resulted in higher eGFR for Black participants at the same serum creatinine. The MDRD equation, the only widely used equation at the time of the seminal study by Choi et al., assigned a 21% higher eGFR for Black adults and also has poor accuracy at normal or near-normal eGFR.³³ The 2009 CKD-EPI equation assigned a 16% higher eGFR for Black adults. We have recently reported that the race-free 2021 CKD-EPI creatinine equation led to inclusion of more Black patients with CKD who are younger and generally healthier and removal of some healthier White patients who were previously diagnosed to have CKD with the 2009 CKD-EPI equation.¹⁴ These differences affected assessment of disparities.^14,15 Thus, we believe the use of race-based GFR equations in prior studies account for the different findings compared with what we report here.

The coexistence of the higher hazards of KFRT (Figure 4A) and lower hazards of pre-KFRT death (Figure 4B) in Black relative to White patients suggest that the excess cumulative incidence of KFRT in Black patients after CKD onset (Figure 2A) resulted from a combination of these two hazards. The higher hazard of progressing to KFRT is consistent with more rapid decline in kidney function among Black patients with CKD, as previously reported.^29,34 However, the implication of the lower hazard of pre-KFRT death among Black patients for the racial difference in the cumulative incidence of KFRT over the course of CKD is less clear because they depend in part on unobservable quantities. The larger size of the patient population that remains at risk of KFRT for the Black compared with the White group because of lower mortality in Black patients before KFRT seems likely to have contributed to additional KFRT cases in Black patients with CKD. In addition, if death and KFRT share risk factors, fewer deaths before KFRT may lead to more people at high risk of KFRT being retained in the larger Black risk set than in the White risk set, contributing additional KFRT cases in Black patients with CKD.

Our finding reveals the large racial difference in age at CKD onset as a main contributor to the KFRT disparity. This finding is intriguing because it suggests that if racial gap in age at CKD onset can be reduced, there could be a major reduction in the racial disparity in KFRT incidence after CKD onset. Further analysis of the VHA population revealed that this wide age gap at CKD onset stemmed from a two-fold increased rate of incident CKD experienced by the VHA younger Black population (Figure 5), instead of demographic characteristics of the overall VHA population. In addition to recognizing the role of age in KFRT disparity, we note that the likelihood of incident KFRT was higher than that of death in younger patients and vice versa in older patients, suggesting that age modifies differential risks of KFRT and death, which was consistent with the prior findings by O'Hare et al.³⁵ Moreover, the higher likelihood of KFRT relative to that of death was more prominent in younger Black compared with younger White patients (Figure 3), further emphasizing the need to prevent the onset of CKD and to slow the rapid decline in kidney function among younger Black adults.

As a latent social construct, racial group differences are predominantly mediated through racism and its differential effect on exposure to upstream factors (social, economic, environmental, dietary, and other individual-level or contextual stressors) that may heighten the risk of development of CKD earlier in life and its progression.^36–38 These can then activate a number of maladaptive epigenetic, neurohormonal, immunologic, and physiologic changes^39–42 that contribute to subclinical abnormalities not captured on administrative data but predispose to CKD development and progression.^36–38,41 Additional factors that may predispose younger Black adults to early onset and/or rapid progression of CKD include the presence of APOL1 polymorphisms that are more prevalent in persons with West African ancestry.⁴³

Strengths of this study include the analysis of nationwide patients with incident CKD, the first study to assess time-varying disparities over the course of CKD, a detailed examination of the role of racial differences in the hazards of KFRT and competing-risk death in the disparity of cumulative KFRT incidence, and the identification of age at CKD onset as a major contributory factor to this disparity. Moreover, our inclusion and exclusion approach (Methods Section and Supplemental Figure 1) and similar frequencies of eGFR measurements in the VHA by race should yield great accuracy in similarly identifying the first clinical presentation of CKD onset and capturing incident patients across racial groups. However, our findings should be interpreted in light of several considerations and limitations. First, we defined the onset of CKD as eGFR <60 ml/min per 1.73 m² on the basis of the 2021 CKD-EPI equation. While this is a reasonable and practical approach for large population studies, Inker et al.¹⁶ reported that the 2021 equation, on average, underestimated measured GFR for individuals of Black race and overestimated measured GFR for individuals of non-Black race in their study populations. Provided that these biases similarly existed in estimation of GFR at around 60 ml/min per 1.73 m², the racial disparity in the incidence of KFRT would likely be even larger if the onset of CKD was based on measured GFR. Second, we cannot rule out ascertainment bias because of the use of electronic health records. Third, the unavailability of consistent UACR data in a substantial subset of the cohort precludes the inclusion of patients with CKD stage G1 or G2 and examination of the role of albuminuria in the racial disparity in KFRT. Fourth, ascertainment of comorbidities relied on diagnosis codes, although our linkage to CMS claims should improve accuracy. Finally, the generalizability of our findings is unclear given that the Veteran population consists of primarily older male individuals compared with the non-Veteran population.⁴⁴

In summary, our findings demonstrate that the 2.5-fold higher cumulative incidence of KFRT in Black veterans persisted throughout the course of CKD from its onset and was not limited to a time segment in advanced CKD. This disparity resulted from a combination of higher hazards of progression to KFRT and lower hazards of the competing risk of death among Black veterans. The younger age of onset of stage 3 CKD in Black veterans was a major contributor to the KFRT disparity. These findings underscore the need to prevent early onset and slow progression of CKD in younger Black adults to reduce the persistent racial disparity in KFRT in the United States.

Supplementary Material

jasn-35-299-s001.pdf ^{(255.8KB, pdf)}

Disclosures

A.K. Cheung reports Consultancy: Alucent, Boehringer-Ingelheim, CSL Behring, and 3D Communications; Ownership Interest: Merck; Patents or Royalties: UpToDate; and Advisory or Leadership Role: KDIGO. T. Greene reports Consultancy: AstraZeneca, Invokana, Janssen Pharmaceuticals, Novartis, and Pfizer Inc.; and Research Funding: AstraZeneca, Boehringer-Ingleheim, CSL, and Vertex. R. Nee reports Ownership Interest: Mutual funds, ETFs and individual stocks to include AirBnB, Amazon, Apple, cruise and airline companies, Microsoft, PayPal, Shopify, and Teladoc; Advisory or Leadership Role: Centers for Disease Control (CDC)-Kidney Disease Surveillance System Advisory Group (voluntary position), Medical Advisory Board for the National Kidney Foundation, and National Capital Area (voluntary position); and other interests or relationships: Kidney Interagency Coordinating Committee, NIDDK, and National Institutes of Health K.C. Norris reports Consultancy: Atlantis Healthcare—Compliance, research and quality care for dialysis and CKD care in Puerto Rico; Research Funding: NIH, State of California; Advisory or Leadership Role: AAMC, AAKP, Atlantis Healthcare, CJASN, ESRD Network Forum, Ethnicity & Disease, ISN, JASN, NIDDK Council, and NKF-KEEP; and Other Interests or Relationships: AAKP, ASN, ESRD Network Forum, NKF, and SGIM. J.J. Scialla reports Advisory or Leadership Role: Deputy Editor, American Journal of Kidney Diseases; Scientific Advisory Board, National Kidney Foundation. G. Yan reports Advisory or Leadership Role: JASN Editorial Board. All remaining authors have nothing to disclose.

Funding

Research reported in this publication was supported by NIH/NIDDK under Award Number R01DK112008. K.C. Norris was supported by NIH grants P50MD017366, P30AG021684, and UL1TR000124. T. Greene received support from the University of Utah Translational Research: Implementation, Analysis, and Design (TRIAD) team, with funding in part from the National Center for Research Resources and the National Center for Advancing Translational Sciences, National Institutes of Health, through NIH grant UM1TR004409.

Author Contributions

Conceptualization: Alfred K. Cheung, Tom Greene, Robert Nee, Keith C. Norris, Julia J. Scialla, Guofen Yan.

Data curation: Wei Yu.

Formal analysis: Fei Heng, Guofen Yan.

Funding acquisition: Alfred K. Cheung, Tom Greene, Keith C. Norris, Guofen Yan, Wei Yu.

Investigation: Alfred K. Cheung, Tom Greene, Robert Nee, Keith C. Norris, Julia J. Scialla, Guofen Yan, Wei Yu.

Methodology: Tom Greene, Fei Heng, Guofen Yan.

Project administration: Guofen Yan.

Resources: Alfred K. Cheung.

Visualization: Guofen Yan, Wei Yu.

Writing – original draft: Guofen Yan.

Writing – review & editing: Alfred K. Cheung, Tom Greene, Fei Heng, Robert Nee, Keith C. Norris, Julia J. Scialla, Guofen Yan, Wei Yu.

Data Sharing Statement

Data cannot be shared. Authors are not allowed to share the data with a third party. The third party will need to request data directly to the VA Informatics and Computing Infrastructure Resource Center.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/JSN/E573.

Supplemental Item 1. Method to construct an age-matched cohort.

Supplemental Item 2. Method to extract the two subpopulations.

Supplemental Table 1. Distributions of age at CKD onset, sex, and calendar year of CKD onset in the actual cohort of White and Black veterans and the matched subset of Black veterans.

Supplemental Figure 1. Flowchart for constructing the cohort of veterans with incident CKD.

Supplemental Figure 2. Age distributions of veterans by race in the two extracted subpopulations.