Change in Multiple Filtration Markers and Subsequent Risk of Cardiovascular Disease and Mortality

Abstract

Background and objectives

Kidney disease progression, assessed by change in eGFR on the basis of creatinine, is an independent risk factor for cardiovascular disease and death. This study aimed to evaluate whether changes in multiple filtration markers, individually and combined, were associated with cardiovascular disease and death.

Design, setting, participants, & measurements

Creatinine, cystatin C, and β₂-microglobulin were measured among 9716 Atherosclerosis Risk in Communities Study participants in 1990–1992 and 1996–1998. Percentage change in three filtration markers (eGFR on the basis of creatinine, eGFR on the basis of cystatin C, and 1/β₂-microglobulin) individually and the average of percentage change across all three filtration markers were calculated. Cardiovascular events and deaths were ascertained from 1996 to 2011. Cox regression models were adjusted for established risk factors for cardiovascular disease and mortality and first measurement of eGFR on the basis of creatinine.

Results

During a median follow-up of 14 years, there were 1922 cardiovascular events and 2285 deaths from any cause. Decline of >30% in each filtration marker was significantly associated with higher risk of mortality compared with stable kidney function (−9.9% to +9.9% change in the filtration marker) with hazard ratios (95% confidence intervals) of 1.91 (1.67 to 2.18) for eGFR on the basis of creatinine, 2.29 (1.99 to 2.63) for eGFR on the basis of cystatin C, and 2.48 (2.15 to 2.86) for 1/β₂-microglobulin, with similar associations for cardiovascular disease. An average decline of >30% across the three markers was strongly associated with higher risk of all-cause mortality (hazard ratio, 2.82; 95% confidence interval, 2.42 to 3.29).

Conclusions

Kidney disease progression was assessed using >30% decline in eGFR on the basis of creatinine, eGFR on the basis of cystatin C, and 1/β₂-microglobulin and average decline of >30% across the three filtration markers is strongly associated with risk of cardiovascular disease and death.

Introduction

Percentage change in eGFR on the basis of creatinine (eGFR_Cr) and in particular, the threshold of ≥30% eGFR_Cr decline are strongly associated with risk of incident ESRD (1,2). Cardiovascular disease and death may occur sooner and more often than ESRD and thus, are important outcomes to quantify within the context of CKD progression (3–6). In this context, a few studies have documented a strong relationship between change in eGFR_Cr and risk of cardiovascular disease and mortality (1,7–9).

Assessment of kidney function solely using serum creatinine to estimate GFR has limitations. Serum creatinine can be influenced by non-GFR determinants, namely muscle mass and dietary protein intake (10). Although not yet widely used in clinical practice, cystatin C and β₂-microglobulin are low molecular weight proteins considered to be filtration markers that are strongly associated with clinical outcomes (11–13). Use of these alternative filtration markers instead of or in addition to creatinine could improve estimation of GFR and prognosis in CKD (14–18). The use of multiple filtration markers may ameliorate bias introduced by non-GFR determinants and reduce imprecision caused by measurement error. We previously showed that change across three filtration markers (creatinine, cystatin C, and β₂-microglobulin) was strongly associated with risk of incident ESRD (2). However, to the best of our knowledge, no study has examined the relationship of cardiovascular disease and mortality to change in β₂-microglobulin or a combination of multiple markers.

We assessed percentage change in three filtration markers (creatinine, cystatin C, and β₂-microglobulin) over a 6-year period, individually and combined, and the subsequent risk of cardiovascular disease and mortality.

Materials and Methods

Study Design

The Atherosclerosis Risk in Communities (ARIC) Study is a community-based cohort study of 15,792 middle-aged, predominantly black and white men and women from four United States communities who enrolled in 1987–1989. The ARIC Study has been described in detail previously (19).

Study Population

This study included ARIC Study participants with measurements of all three filtration markers at two time points approximately 6 years apart (1990–1992, ARIC Study Visit 2 and 1996–1998, ARIC Study Visit 4). Of those who attended ARIC Study Visit 2 (n=11,449), we excluded those with missing serum measurements of creatinine, cystatin C, or β₂-microglobulin at Study Visit 2 (n=766) or 4 (n=965) and those with missing follow-up information (n=2). For analysis of incident events, those with preexisting coronary heart disease, ischemic stroke, and heart failure at or before Study Visit 4 were excluded from the respective analyses. Informed consent was obtained at each study visit, and the study protocol was approved by all participating centers.

Data Collection

At each study visit, trained staff administered questionnaires to collect demographics (age, sex, and race) and kidney disease risk factors (medical history, medication use, and smoking status). Body mass index was calculated as weight (kilograms)/height (meters)². Mean BP was calculated from two measurements obtained with a random zero sphygmomanometer. Blood samples were centrifuged within 30 minutes of venipuncture at 3000×g for 10 minutes and stored at −70°C for future laboratory analysis. Total cholesterol was measured colorimetrically after oxidation by cholesterol oxidase, and HDL was measured enzymatically after precipitation with dextran sulfate magnesium (20,21). Glucose was measured by the modified hexokinase/glucose-6-phosphate dehydrogenase method. Diabetes status was defined by fasting glucose ≥126 mg/dl, nonfasting glucose ≥200 mg/dl, self-reported history of diagnosed diabetes, or current diabetes medication use.

Measurement of Filtration Markers

Creatinine was measured by the modified kinetic Jaffe method according to the original study protocol in Visit 2 (1990–1992) serum specimens and Visit 4 (1996–1998) plasma specimens. In 2012 and 2013, stored Visit 2 serum specimens were used to measure cystatin C with the Gentian immunoassay (Gentian, Moss, Norway) and β₂-microglobulin with the Roche β₂-microglobulin reagent on the Roche Modular P800 Chemistry Analyzer. In 2010, stored Visit 4 plasma specimens were used to measure cystatin C and β₂-microglobulin using a particle-enhanced immune-nephelometric assay with a BNII Nephelometer (Siemens Healthcare Diagnostics).

Values of filtration markers were standardized and calibrated as appropriate (22). eGFR_Cr and eGFR on the basis of cystatin C (eGFR_Cys) were calculated using CKD Epidemiology Collaboration (CKD-EPI) equations (14,23). β₂-microglobulin was expressed as its inverse for ease of comparison with eGFR. Percentage change in filtration markers (eGFR_Cr, eGFR_Cys, and 1/β₂-microglobulin) over the 6-year period was calculated as the difference between the two measurements as a proportion of the first measurement. The average change in three markers was calculated as follows: (percentage change in eGFR_Cr + percentage change in eGFR_Cys + percentage change in 1/β₂-microglobulin)/3.

Outcome Ascertainment

Incident coronary heart disease, ischemic stroke, heart failure, and deaths from 1996 through December 31, 2011, were ascertained by active surveillance of local hospital discharge records, state death records, and linkage to the National Death Index combined with information from annual phone interviews with participants or proxies. Incident coronary heart disease was defined by adjudicated definite or probable hospitalized myocardial infarction or fatal coronary heart disease (24). Incident ischemic stroke was defined by adjudicated definite or probable ischemic stroke (25). Incident heart failure was defined by a hospitalization or death with an International Classification of Disease-9/10 code for heart failure (428, I50) (26). Total cardiovascular disease was defined as any of the aforementioned cardiovascular events. Cardiovascular disease mortality was defined as death with an International Classification of Disease-9/10 code of 390–459 or I00–I99 for underlying cause of death.

Statistical Analyses

Analyses were conducted with percentage change over 6 years expressed categorically (substantial decline: <−30%; moderate decline: −30% to −10%; stable kidney function [reference group]: >−10% to <+10% [denoted as −9.9% to +9.9% for convenience]; and increase: ≥+10%). To estimate the association between change in filtration markers and risk of cardiovascular disease and death, we used Cox regression models with adjustment for demographics (age, sex, and race), traditional risk factors (body mass index, systolic BP, antihypertensive medication use, diabetes status, total cholesterol, HDL cholesterol, and current smoking status), and eGFR_Cr, all of which were assessed at the first time point (1990–1992, ARIC Study Visit 2). eGFR_Cr was included in multivariable models as linear spline terms with two knots at 60 and 90 ml/min per 1.73 m². In sensitivity analyses, models were (1) adjusted for the first measurement of the respective filtration marker, (2) not adjusted for any filtration marker, and (3) adjusted for eGFR_Cr and covariates assessed at the last time point (1996–1998, ARIC Study Visit 4). Analyses were conducted using Stata (version 13) statistical software (StataCorp LP, College Station, TX).

Results

In the study population of 9716 participants at Study Visit 2, mean age was 57 years old, 57.4% were women, and 21.4% were black (Table 1). Median eGFR_Cr was 97 ml/min per 1.73 m², and 1.4% of participants had eGFR_Cr<60 ml/min per 1.73 m².

Table 1.

Demographics, clinical characteristics, and concentrations of filtration markers assessed at the Atherosclerosis Risk in Communities Study visit 2 (1990–1992) in the overall study population (n=9716)

Variable	Percent (n) or Mean (SD)a
Age, yr	56.8 (5.7)
Women	57.4% (5580)
Black	21.4% (2078)
Body mass index, kg/m2	27.9 (5.3)
Systolic BP, mmHg	120.3 (17.9)
Antihypertensive medication use	30.3% (2940)
Diabetes mellitus	12.8% (1241)
Total cholesterol, mg/dl	209.0 (38.4)
HDL cholesterol, mg/dl	50.2 (16.7)
Current cigarette smoker	19.0% (1843)
eGFRCr<60 ml/min per 1.73 m2	1.4% (137)
eGFRCr, ml/min per 1.73 m2	97.3 (89.2, 105.3)
Creatinine, mg/dl	0.73 (0.63, 0.83)
eGFRCys, ml/min per 1.73 m2	94.5 (80.2, 105.1)
Cystatin C, mg/L	0.85 (0.75, 0.95)
β2-Microglobulin, mg/L	1.82 (1.61, 2.08)

eGFR_Cr, eGFR on the basis of creatinine; eGFR_Cys, eGFR on the basis of cystatin C.

^aMedian (25th percentile, 75th percentile) for concentrations of filtration markers (eGFR_Cr, creatinine, eGFR_Cys, cystatin C, and β₂-microglobulin).

Over the 6-year period, mean percentage change was −10.4% for eGFR_Cr, −8.9% for eGFR_Cys, −6.6% for 1/β₂-microglobulin, and −8.7% for the three-marker average (Figure 1). The peak of the distribution was higher and the tails of the distribution were flatter for the three-marker average, representing less variability compared with the individual filtration markers. For the individual filtration markers, more participants had a >30% decline in eGFR_Cr (6.9%) than >30% decline in any other marker (Table 2). More participants had an increase in 1/β₂-microglobulin (10.3%) than any other marker; 36.9% of participants had a 10%–30% decline in eGFR_Cr.

Figure 1.

Distribution of the percentage change in filtration markers over a 6-year period. eGFR_Cr, eGFR on the basis of creatinine; eGFR_Cys, eGFR on the basis of cystatin C.

Table 2.

Frequencies and hazard ratios (95% confidence intervals) for cardiovascular disease and mortality by category of percentage change in filtration markers over a 6-year period adjusted for first measurement of covariates and eGFR

Filtration Marker	Category of Percentage Change in Filtration Marker
	<−30%	−30% to −10%	−9.9% to +9.9%	≥+10%
eGFRCr
Frequency, % (n)	6.9 (666)	36.9 (3589)	51.5 (5005)	4.7 (456)
eGFRCys
Frequency, % (n)	5.8 (564)	36.7 (3563)	50.9 (4948)	6.6 (641)
1/β2-Microglobulin
Frequency, % (n)	5.3 (511)	34.4 (3342)	50.1 (4863)	10.3 (1000)
Three-marker averagea
Frequency, % (n)	3.7 (356)	37.4 (3638)	55.5 (5396)	3.4 (326)
Total cardiovascular disease (1922/9712)b
eGFRCr
Events, % (n)	38.6 (257)	19.4 (696)	17.3 (867)	22.4 (102)
HR (95% CI)	2.02 (1.75 to 2.33)c	1.06 (0.96 to 1.17)	1 (Reference)	1.09 (0.87 to 1.36)
eGFRCys
Events, % (n)	31.9 (180)	20.0 (714)	17.4 (862)	25.9 (166)
HR (95% CI)	1.73 (1.46 to 2.03)c	1.17 (1.06 to 1.29)c	1 (Reference)	1.17 (0.99 to 1.39)
1/β2-Microglobulin
Events, % (n)	35.0 (179)	21.8 (728)	16.8 (818)	19.7 (197)
HR (95% CI)	2.09 (1.77 to 2.47)c	1.31 (1.19 to 1.45)c	1 (Reference)	1.13 (0.97 to 1.32)
Three-marker averagea
Events, % (n)	41.6 (148)	20.8 (758)	17.3 (931)	26.1 (85)
HR (95% CI)	2.30 (1.92 to 2.75)c	1.24 (1.12 to 1.36)c	1 (Reference)	1.20 (0.96 to 1.51)
Coronary heart disease (835/9452)b
eGFRCr
Events, % (n)	23.4 (156)	11.2 (401)	9.7 (483)	12.9 (59)
HR (95% CI)	2.19 (1.77 to 2.70)c	1.06 (0.90 to 1.23)	1 (Reference)	1.17 (0.84 to 1.63)
eGFRCys
Events, % (n)	20.6 (116)	11.1 (396)	9.9 (491)	15.0 (96)
HR (95% CI)	1.66 (1.29 to 2.13)c	1.12 (0.96 to 1.30)	1 (Reference)	1.08 (0.83 to 1.40)
1/β2-Microglobulin
Events, % (n)	20.2 (103)	12.3 (412)	9.5 (464)	12.0 (120)
HR (95% CI)	1.79 (1.36 to 2.34)c	1.32 (1.13 to 1.54)c	1 (Reference)	1.27 (1.01 to 1.59)c
Three-marker averagea
Events, % (n)	27.3 (97)	12.2 (444)	9.2 (498)	18.4 (60)
HR (95% CI)	2.43 (1.85 to 3.19)c	1.35 (1.17 to 1.57)c	1 (Reference)	1.62 (1.18 to 2.22)c
Ischemic stroke (442/9609)b
eGFRCr
Events, % (n)	11.4 (76)	5.7 (205)	4.7 (237)	6.8 (31)
HR (95% CI)	1.70 (1.26 to 2.29)c	0.95 (0.77 to 1.17)	1 (Reference)	0.92 (0.56 to 1.51)
eGFRCys
Events, % (n)	9.2 (52)	5.7 (203)	4.8 (238)	8.7 (56)
HR (95% CI)	1.56 (1.09 to 2.22)c	1.18 (0.95 to 1.45)	1 (Reference)	1.30 (0.92 to 1.84)
1/β2-Microglobulin
Events, % (n)	11.6 (59)	5.8 (192)	4.7 (228)	7.0 (70)
HR (95% CI)	1.81 (1.27 to 2.60)c	1.20 (0.97 to 1.49)	1 (Reference)	1.36 (1.00 to 1.85)c
Three-marker averagea
Events, % (n)	11.8 (42)	6.1 (221)	4.7 (255)	9.5 (31)
HR (95% CI)	1.72 (1.16 to 2.57)c	1.20 (0.98 to 1.47)	1 (Reference)	1.12 (0.68 to 1.85)
Heart failure (1139/9531)b
eGFRCr
Events, % (n)	34.7 (215)	14.3 (485)	11.6 (554)	16.2 (70)
HR (95% CI)	2.33 (1.94 to 2.79)c	1.20 (1.05 to 1.37)c	1 (Reference)	1.09 (0.82 to 1.45)
eGFRCys
Events, % (n)	32.5 (170)	14.8 (501)	11.5 (542)	18.9 (111)
HR (95% CI)	2.16 (1.77 to 2.65)c	1.30 (1.14 to 1.48)c	1 (Reference)	1.17 (0.94 to 1.47)
1/β2-Microglobulin
Events, % (n)	36.8 (175)	16.3 (517)	11.0 (510)	13.1 (122)
HR (95% CI)	2.86 (2.34 to 3.50)c	1.47 (1.29 to 1.68)c	1 (Reference)	1.08 (0.88 to 1.34)
Three-marker averagea
Events, % (n)	45.3 (149)	14.8 (511)	11.9 (607)	19.1 (57)
HR (95% CI)	3.07 (2.48 to 3.79)c	1.21 (1.07 to 1.38)c	1 (Reference)	1.04 (0.76 to 1.42)
All-cause mortality (2285/9716)b
eGFRCr
Events, % (n)	45.7 (304)	23.1 (830)	20.9 (1045)	23.3 (106)
HR (95% CI)	1.91 (1.67 to 2.18)c	1.04 (0.95 to 1.15)	1 (Reference)	0.98 (0.79 to 1.21)
eGFRCys
Events, % (n)	48.8 (275)	23.3 (831)	19.8 (979)	31.2 (200)
HR (95% CI)	2.29 (1.99 to 2.63)c	1.18 (1.07 to 1.29)c	1 (Reference)	1.28 (1.09 to 1.50)c
1/β2-Microglobulin
Events, % (n)	50.1 (256)	25.6 (857)	19.9 (966)	20.6 (206)
HR (95% CI)	2.48 (2.15 to 2.86)c	1.28 (1.16 to 1.40)c	1 (Reference)	1.01 (0.87 to 1.18)
Three-marker averagea
Events, % (n)	59.3 (211)	24.7 (899)	19.9 (1076)	30.4 (99)
HR (95% CI)	2.82 (2.42 to 3.29)c	1.25 (1.14 to 1.37)c	1 (Reference)	1.23 (1.00 to 1.53)c
Cardiovascular disease mortality (696/9716)b
eGFRCr
Events, % (n)	19.1 (127)	7.1 (256)	5.7 (284)	6.4 (29)
HR (95% CI)	2.44 (1.96 to 3.03)c	1.12 (0.94 to 1.33)	1 (Reference)	0.84 (0.56 to 1.26)
eGFRCys
Events, % (n)	16.7 (94)	7.1 (253)	5.8 (285)	10.0 (64)
HR (95% CI)	2.32 (1.82 to 2.95)c	1.21 (1.02 to 1.44)c	1 (Reference)	1.23 (0.93 to 1.63)
1/β2-Microglobulin
Events, % (n)	17.4 (89)	8.2 (275)	5.4 (263)	6.9 (69)
HR (95% CI)	2.58 (2.00 to 3.32)c	1.47 (1.24 to 1.75)c	1 (Reference)	1.19 (0.91 to 1.55)
Three-marker averagea
Events, % (n)	23.9 (85)	7.8 (283)	5.3 (288)	12.3 (40)
HR (95% CI)	3.55 (2.75 to 4.58)c	1.43 (1.21 to 1.69)c	1 (Reference)	1.67 (1.18 to 2.35)c
Noncardiovascular disease mortality (1540/9716)b
eGFRCr
Events, % (n)	25.8 (172)	15.6 (560)	14.6 (732)	16.7 (76)
HR (95% CI)	1.68 (1.42 to 2.00)c	1.03 (0.92 to 1.15)	1 (Reference)	1.09 (0.84 to 1.41)
eGFRCys
Events, % (n)	30.9 (174)	15.7 (559)	13.7 (676)	20.4 (131)
HR (95% CI)	2.24 (1.89 to 2.66)c	1.15 (1.03 to 1.29)c	1 (Reference)	1.28 (1.06 to 1.55)c
1/β2-Microglobulin
Events, % (n)	31.5 (161)	17.0 (568)	14.0 (679)	13.2 (132)
HR (95% CI)	2.42 (2.03 to 2.89)c	1.22 (1.09 to 1.36)c	1 (Reference)	0.94 (0.78 to 1.13)
Three-marker averagea
Events, % (n)	34.8 (124)	16.3 (594)	14.2 (765)	17.5 (57)
HR (95% CI)	2.53 (2.07 to 3.08)c	1.18 (1.06 to 1.31)c	1 (Reference)	1.05 (0.80 to 1.39)

Adjusted for age, sex, race, body mass index, systolic BP, antihypertensive medication use, diabetes status, total cholesterol, HDL cholesterol, current cigarette smoking status, and first measurement of eGFR_Cr as linear spline terms with two knots at 60 and 90 ml/min per 1.73 m² (all covariates were assessed at Study Visit 2 from 1990 to 1992). eGFR_Cr, eGFR on the basis of creatinine; eGFR_Cys, eGFR on the basis of cystatin C; HR, hazard ratio; 95% CI, 95% confidence interval.

^aThree-marker average: mean of percentage change in eGFR_Cr, eGFR_Cys, and 1/β₂-microglobulin.

^bAnalyses exclude prevalent cases of the respective outcome.

^cStatistically significant.

During a median follow-up of 14 years, there were 1922 (19.8%) total cardiovascular events and 2285 (23.5%) deaths from any cause. For individual cardiovascular outcomes, there were 835 (8.8%) incident coronary heart disease events, 442 (4.6%) incident ischemic stroke events, and 1139 (12.0%) incident heart failure events. Nearly one third (30.5%) of all deaths were attributed to an underlying cardiovascular cause (n=696; 7.2% of the entire study population).

Greater than 30% declines in individual filtration markers and the three-marker average were each statistically significantly associated with all cardiovascular and mortality outcomes compared with stable kidney function (Table 2). For example, >30% declines in the three-marker average and eGFR_Cr were independently associated with 3.55-fold (95% confidence interval, 2.75 to 4.58) and 2.44-fold (95% confidence interval, 1.96 to 3.03), respectively, higher risk of cardiovascular mortality compared with stable kidney function. Estimates of association between filtration marker decline and outcomes were weakest for ischemic stroke and strongest for cardiovascular mortality. Declines of 10%–30% in eGFR_Cys, 1/β₂-microglobulin, and the three-marker average but not eGFR_Cr were significantly associated with most cardiovascular disease and mortality outcomes relative to stable kidney function. An increase of ≥10% in filtration markers trended toward higher risk of cardiovascular and mortality outcomes but was not statistically significant in most analyses.

Three sensitivity analyses showed similar results. After adjusting for first measurement of the respective filtration marker and covariates, risk estimates for all outcomes were similar in magnitude to the primary results (Supplemental Table 1). In models unadjusted for filtration marker level, risk estimates were stronger but showed similar patterns (Supplemental Table 2). In models adjusted for last eGFR_Cr and covariate measurement (ARIC Study Visit 4, 1996–1998), estimates were similar in magnitude or slightly weaker than the primary results with the exception of a ≥10% increase in eGFR_Cr, which had significant associations with total cardiovascular disease, coronary heart disease, and heart failure (Supplemental Table 3).

Discussion

In this prospective, community-based study of middle-aged adults, percentage changes in established and novel filtration markers over a 6-year period were associated with cardiovascular morbidity and mortality; >30% declines in eGFR_Cr, eGFR_Cys, and 1/β₂-microglobulin individually and combined were associated with 2- to 3-fold higher risks of any incident cardiovascular disease event, incident coronary heart disease, incident ischemic stroke, incident heart failure, all-cause mortality, cardiovascular mortality, and noncardiovascular mortality compared with stable kidney function. These relationships were statistically significant and independent of demographic characteristics, multiple traditional risk factors, and first measurement of eGFR_Cr. Declines of 10%–30% in filtration markers were more prevalent but less strongly and less consistently associated with risk of cardiovascular disease and mortality than >30% decline.

To the best of our knowledge, this study is the first to assess percentage change in 1/β₂-microglobulin and one of few studies to assess percentage change in cystatin C in relation to cardiovascular morbidity and mortality (8,9). Our findings generally corroborate other related studies (1,7–9,27–29). In the Cardiovascular Health Study of older adults, rapid decline in eGFR_Cys (>3 ml/min per 1.73 m² per year) was associated with higher risk of heart failure, myocardial infarction, and peripheral arterial disease but not stroke (9). Unlike the Cardiovascular Health Study, our study documented significant associations between >30% decline in eGFR_Cys and risk of ischemic stroke. However, risk estimates were weaker for ischemic stroke than any other outcome, and 10%–30% decline in filtration markers was not significantly associated with ischemic stroke.

No known previous studies have combined percentage change in multiple filtration markers to estimate risk of cardiovascular morbidity and mortality. We report that decline in the average change in all three filtration markers was strongly associated with any cardiovascular disease event, heart failure, and mortality. Averaging the decline in multiple filtration markers may reduce imprecision caused by measurement error and reduce confounding caused by non-GFR determinants compared with using a single filtration marker. The association between average decline in multiple filtration markers and outcomes is robust to multiple outcomes and multivariable adjustment for many risk factors. These findings suggest the benefit of measuring alternative filtration markers in addition to creatinine for the assessment of kidney function.

In principle, improvement in kidney function should be associated with better health outcomes. However, previous studies have reported that an observed increase in eGFR_Cr is associated with higher risk of all-cause mortality (1,7,27–29). One possible explanation may be confounding from changes in non-GFR determinants of serum creatinine, such as loss of muscle mass caused by poor health (30). Other potential explanations include confounding from conditions, such as AKI or hyperfiltration (31). Interestingly, in this study, an observed increase in alternative filtration markers was generally more consistently associated with adverse outcomes than an increase in eGFR_Cr, suggesting that the latter two mechanisms may be more important. Prevention efforts targeting AKI and glomerular hyperfiltration may, in turn, prevent cardiovascular morbidity and mortality.

Approximately one half of the study population had stable levels of kidney filtration markers (−9.9% to +9.9%) over a 6-year period, a category that was associated with the lowest risk of cardiovascular events and mortality. Given that ARIC Study participants were selected without regard to disease status, it is expected that the majority would have stable kidney function. Other studies have shown that, even among individuals with advanced kidney disease, many will maintain the same level of kidney function as measured by serum creatinine over many years of follow-up (32–34). Estimating change in filtration markers can help to differentiate between those who will and will not develop clinical outcomes for both patients with kidney disease and the general population. Use of percentage change in multiple filtration markers and >30% decline in particular may be more informative for prognosis than using the standard clinical definition for classifying CKD stage on the basis of a single measurement of eGFR_Cr (35).

This study has some limitations. First, measurement error in the quantification of filtration marker levels could introduce random error or bias. However, rigorous quality control procedures were followed, including blinding of laboratory technicians to the case status of specimens, comparison of values from blind replicates, and use of reference standards (National Institute of Standards and Technology for creatinine and International Federation for Clinical Chemists for cystatin C) (36). Furthermore, a calibration study was conducted among 200 ARIC Study participants across all study visits to quantify this source of error and calibrate values accordingly (22). Second, there may be nonrandom loss to follow-up between study visits, whereby healthier participants are more likely to return for study visits. If present, this bias would lead to conservative risk estimates. Third, measured GFR was not available in this study. However, CKD-EPI eGFR equations are widely accepted as approximations of measured GFR in individuals with lower and higher levels of kidney function (14,23). Fourth, with only two estimates, we do not account for different patterns of change in levels of filtration markers (37,38). Fifth, proteinuria is a strong risk factor for adverse clinical outcomes, and measurement of albuminuria was not available at ARIC Study Visit 2 (39). As in any observational study, there is the potential for residual confounding, despite adjustment for a multitude of established cardiovascular and mortality risk factors.

This study has several characteristics that strengthen the validity and importance of the findings. First, the ARIC Study is a large, well characterized, community-based, prospective cohort. As such, these findings are likely generalizable to a broad segment of the United States population, including middle-aged black and white men and women. Second, long-term follow-up (median of 14 years) was available on the study population, with active surveillance for the ascertainment of multiple cardiovascular morbidity and mortality outcomes. Third, many important covariates were rigorously measured and included in adjusted multivariable regression models, including baseline eGFR_Cr. Fourth, this is a comprehensive and novel study of multiple filtration markers and relevant clinical end points that expands on the existing literature.

In conclusion, kidney disease progression assessed using change in filtration markers, including creatinine, cystatin C, and β₂-microglobulin, and the average change in all three markers is consistently and independently associated with risk of cardiovascular disease and mortality independent of demographics, traditional risk factors, and baseline eGFR_Cr. Furthermore, stable levels of filtration markers represents a low-risk status, but an observed increase in filtration markers should be considered a distinct risk category worthy of additional investigation. Taken together, these study findings lend additional support for the prognostic value of change in filtration markers. These results suggest that the effort to prevent and delay kidney disease progression may also result in reducing risk of heart disease morbidity and mortality.

Disclosures

None.