Significance Statement
Among individuals with CKD, decreases in eGFR, or increases that are steeper than the average loss, have been associated with increased risks of death and cardiovascular events. Sarcopenia and chronic illness might explain why apparent eGFR improvement relates to poor outcomes. The authors investigated the association between slopes of eGFR (defined with yearly eGFR) and the risks of death and cardiovascular events in a prospective cohort of adults with CKD. They found that declines in eGFR greater than the average eGFR loss were associated with increased risks of death and cardiovascular events, despite multiple adjustments. However, neither rises in eGFR nor losses smaller than the average eGFR loss were associated with increases in such risks. These findings suggest that, in CKD, improving eGFR might not be associated with concerning outcomes.
Keywords: chronic kidney disease, epidemiology and outcomes, glomerular filtration rate, risk factors, mortality risk, cardiovascular events
Visual Abstract
Abstract
Background
Slopes of eGFR have been associated with increased risks of death and cardiovascular events in a U-shaped fashion. Poor outcomes in individuals with rising eGFR are potentially attributable to sarcopenia, hemodilution, and other indicators of clinical deterioration.
Methods
To investigate the association between eGFR slopes and risks of death or cardiovascular events, accounting for multiple confounders, we studied 2738 individuals with moderate to severe CKD participating in the multicenter Chronic Renal Insufficiency Cohort (CRIC) Study. We used linear, mixed-effects models to estimate slopes with up to four annual eGFR assessments, and Cox proportional hazards models to investigate the association between slopes and the risks of death and cardiovascular events.
Results
Slopes of eGFR had a bell-shaped distribution (mean [SD], −1.5 [−2] ml/min per 1.73 m2 per year). Declines of eGFR that were steeper than the average decline associated with progressively increasing risks of death (hazard ratio [HR], 1.23; 95% confidence interval [95% CI], 1.09 to 1.39; for a slope 1 SD below the average) and cardiovascular events (HR, 1.19; 95% CI, 1.03 to 1.38). Rises of eGFR or declines lower than the average decline were not associated with the risk of death or cardiovascular events.
Conclusions
In a cohort of individuals with moderate to severe CKD, we observed steep declines of eGFR were associated with progressively increasing risks of death and cardiovascular events; however, we found no increased risks associated with eGFR improvement. These findings support the potential value of eGFR slopes in clinical assessment of adults with CKD.
Declining, stable, or recovering eGFR has been observed among individuals with CKD.1,2 Whereas declining or stable eGFR is expected,3 increasing eGFR has surprisingly been described in 10%–30% of individuals across CKD cohort studies.1,2,4,5 Increases in eGFR may indicate hyperfiltration or mild forms of recovering kidney disease, but they may also arise from muscle wasting or hemodilution, in the setting of chronic illness, that causes eGFR to overestimate true levels of kidney function. Reports of an elevated risk of death and coronary heart disease associated with rising eGFR are consistent with this hypothesis.6–10 However, most of these studies have had access to limited clinical data reflective of the severity of chronic illness.
We sought to enhance the clinical assessment of the CKD population by characterizing the independent association of slope of eGFR with the risks of death and cardiovascular events. We studied a cohort of adults with moderate to severe CKD for whom extensive clinical, physiologic, and biologic data were available. We compared the adjusted risk of these events across the full distribution of slope of eGFR, with a particular focus on the subpopulation whose eGFR rose over time.
Methods
Study Population
The Chronic Renal Insufficiency Cohort (CRIC) Study is a multicenter, prospective, cohort study that enrolled 3939 racially/ethnically diverse adults with CKD (eGFR of 20–70 ml/min per 1.73 m2 at screening) in its first phase. A detailed description of the design and methods of the CRIC Study have been published.11,12 The age of enrolled patients varied between 21 and 74 years, and 48% had a history of diabetes. Individuals with polycystic kidney disease, GN receiving immunosuppression, a history of HIV/AIDS, cirrhosis, or severe heart failure, and pregnant women, were excluded. All participants provided written informed consent. An institutional review board approved this study protocol. The study has been conducted in accordance with the principles of the Declaration of Helsinki.
A total of 2738 CRIC Study participants, 70% of those initially enrolled, had serum creatinine and cystatin C measured at the year-3 visit and were included in this study (Figure 1). The 1201 CRIC participants excluded from this study were, on average, younger, had lower eGFR, and higher proteinuria compared with those included (Supplemental Table 1).
Figure 1.
Cohort definition. (A) Total Number of participants after exclusions. (B) The primary exposure was defined within the first 3 years from baseline, while the time-to outcomes started at the end of this interval.
Primary Exposure
Slopes of eGFR were calculated using annual eGFR values collected between the baseline and year-3 visit (up to four assessments). The CRIC Study–derived equation13,14 was used to estimate GFR. This equation estimates urinary iothalamate sodium clearance using age, sex, race, standardized serum creatinine, and serum cystatin C.
Primary Outcomes
Time to death and time to cardiovascular events were the primary outcomes, with follow-up starting at the year-3 visit. For analysis of death, study participants were censored when they withdrew or were at the end of the follow-up period, whichever occurred first. For cardiovascular events, participants were censored when they died (of a noncardiovascular event), withdrew, or were at the end of the follow-up, whichever occurred first. Both outcomes were assessed at annual CRIC Study visits and by telephone midway between visits. Deaths were ascertained from next of kin, death certificates, obituaries, hospital records, the Social Security Death Master File, and the National Death Index. Potential cardiovascular events, including coronary artery disease, heart failure, stroke, and peripheral vascular disease events, were adjudicated using dual medical chart review.
Confounders
Potential confounders included age, sex, educational attainment, race/ethnicity, smoking status, history of cardiovascular events (coronary artery disease, heart failure, stroke, or peripheral vascular disease), diabetes (defined by fasting glucose ≥126 mg/dl, random blood glucose ≥200 mg/dl, or self-reported use of insulin or oral diabetes medication), and use of angiotensin-converting enzyme inhibitors/angiotensin receptor blockers or diuretics in the previous 30 days. Assessments performed during the study visits included systolic BP (SBP), fatfree mass (FFM; measured with bioimpedance15), body mass index, and waist circumference. Laboratory markers included eGFR, 24-hour urinary protein, hemoglobin, serum albumin, phosphate, fibroblast growth factor 23, intact parathyroid hormone, high-sensitivity C-reactive protein, serum sodium, HDL cholesterol, high-sensitivity Troponin-T, N-terminal probrain natriuretic peptide, urinary neutrophil gelatinase–associated lipocalin, and left ventricular ejection fraction estimated using echocardiography.16 Number of hospitalizations and number of days hospitalized between baseline and year-3 study visits were also included. A timeline for the assessment of covariates is provided in Supplemental Figure 1.
Potential Effect Modifiers
We explored if subgroups of age, sex, race/ethnicity, diabetes, history of cardiovascular events, and average eGFR had different associations between the slopes of eGFR and the risks of death and cardiovascular events. Also, we explored if the change in SBP, 24-hour urinary protein, FFM, or serum albumin also modified these associations.
Statistical Analyses
Slopes Definition
We fit unadjusted, linear, mixed-effects models of longitudinal eGFR to estimate the slopes of eGFR for each participant. We used a random-intercept and random-slope model with an independent covariance matrix to minimize the correlation between slopes and intercepts.17 We also chose an independent correlation structure for the residuals on the basis of a series of likelihood ratio tests.
Descriptive Analysis
We described baseline characteristics of study participants and tested for trends across quintiles of the slope of eGFR. To characterize other factors that may have changed in tandem with eGFR over the 3-year exposure window, we described slopes of covariates assessed annually across quintiles of slopes of eGFR. We used linear, mixed-effects models to assess the slopes for covariates. We also estimated incidence rates of death and cardiovascular events as the number of incident events per 1000 person-years. Finally, because slopes were derived from the same eGFR estimates included as covariates in our models, we evaluated the collinearity between them. To enable this evaluation, all participants were required to have an eGFR at baseline and at the year-3 visit.
Cox Models
We performed Cox proportional-hazards (PH) regression, in which the primary exposure was the slope of eGFR, modeled with restricted cubic splines (knots at 10th, 50th, and 90th percentiles) or categorized into quintiles. We fit a hierarchic set of eight models to evaluate the effect of the different hypothesized pathways that may explain the association between slopes of eGFR and the time to death and cardiovascular events. The included models were as follows: model 1, unadjusted; model 2, adjusted for demographic and clinical characteristics, including age, sex, race, educational attainment, smoking status, history of cardiovascular disease (CVD), diabetes, and SBP; model 3, adjusted for variables in model 2 plus eGFR and 24-hour urinary protein; model 4, adjusted for variables in model 3 plus body composition18,19 measured as FFM, body mass index, and waist circumference; model 5, adjusted for variables in model 3 plus cardiovascular function and volume overload, including left ventricular ejection fraction, high-sensitivity Troponin-T, N-terminal probrain natriuretic peptide, serum sodium,20,21 and reported use of angiotensin-converting enzyme inhibitors/angiotensin receptor blockers and diuretics; model 6, adjusted for variables in model 3 plus markers of CKD complications and inflammation, including hemoglobin, serum albumin,22 phosphate, fibroblast growth factor 23, intact parathyroid hormone, high-sensitivity C-reactive protein, and urinary neutrophil gelatinase–associated lipocalin; model 7, adjusted for variables in model 3 plus the number of hospitalizations and the number of days spent in the hospital; and model 8, adjusted for variables included in models 3–7. All models were stratified by study site, allowing for variability in the baseline hazards across the seven CRIC Study centers. We explored potential effect modification by the prespecified subgroups described above. We assessed the PH assumption by examining interactions by time in the models.
Sensitivity Analysis
We reimplemented our primary analysis using the first 5 years of CRIC data to define the slope of eGFR and beginning follow-up at the end of the year-5 visit. We also refit models using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation23 for eGFR that incorporated both serum creatinine and cystatin C. Consistent with most previous studies of the slope of eGFR,8–10 we re-estimated slopes using ordinary least-squares regression and then refit the Cox models.
Missing Data and Software
Covariates were imputed assuming they were missing at random (Supplemental Table 2). Results from different imputed datasets were combined using the Rubin rules.24 No variables intended to be included in the analysis were excluded because of a high level of missing data. When estimating slopes, missing values of eGFR were not imputed. All analyses were performed using Stata version 14.2 for macOS.
Results
Description of the Population
The slopes of eGFR were estimated for 2738 individuals between the baseline and the end of 3 years of follow-up. Slopes ranged from −11.4 to 9.02, with a mean value of −1.5 (±2) ml/min per 1.73 m2 per year (Supplemental Figure 2). Moving from quintile 1 (steep decliners) to quintile 5 (improvers) of the slope of eGFR, there was a trend of decreasing values for various characteristics including FFM, SBP, urinary protein, and prevalence of diabetes (P<0.01) (Table 1). Because the range of the slopes of eGFR was extensive for the quintile 5, we subdivided it at its median (Supplemental Figure 3), but no differences in baseline characteristics (Supplemental Table 3) were observed across subquintiles.
Table 1.
Baseline characteristics of study participants according to quintiles of slopes of eGFR
| Characteristics | Quintile 1 | Quintile 2 | Quintile 3 | Quintile 4 | Quintile 5 | P Valuea | P Trendb |
|---|---|---|---|---|---|---|---|
| Slope range (ml/min per 1.73 m2 per year)c | −11.4 to −3.03 | −3.02 to −1.89 | −1.88 to −1.09 | −1.09 to 0.00 | 0.00–9.02 | — | — |
| Number of observations | 548 | 548 | 547 | 548 | 547 | — | — |
| Demographics | |||||||
| Age (yr), mean±SD | 60.2±10.7 | 62.1±10.3 | 61.3±11.2 | 62.1±10.1 | 60.0±9.9 | 0.05 | 0.47 |
| Female sex, n (%) | 212 (39) | 268 (49) | 267 (49) | 250 (46) | 240 (44) | 0.10 | 0.29 |
| Racial/ethnic group, n (%) | |||||||
| Non-Hispanic White | 183 (33) | 241 (44) | 283 (51) | 298 (54) | 256 (47) | 0.44 | <0.001d |
| Non-Hispanic Black | 268 (49) | 217 (39) | 201 (37) | 187 (34) | 221 (40) | ||
| Hispanic | 72 (13) | 64 (12) | 48 (9) | 41 (8) | 52 (10) | ||
| Other | 25 (5) | 26 (5) | 15 (3) | 22 (4) | 18 (3) | ||
| ApoL1 recessive genetic model, n (%) | |||||||
| Zero or one copy of APOL1 risk variants | 363 (86) | 401 (91) | 403 (93) | 424 (96) | 416 (93) | 0.81 | <0.001d |
| Two copies of APOL1 risk variants | 59 (14) | 39 (9) | 32 (7) | 19 (4) | 31 (7) | ||
| Educational attainment, n (%) | |||||||
| Less than high school | 119 (22) | 110 (20) | 95 (17) | 66 (12) | 82 (15) | 0.70 | <0.001d |
| High school graduate | 122 (22) | 90 (16) | 84 (16) | 110 (20) | 90 (16) | ||
| Some college | 158 (29) | 173 (32) | 166 (30) | 142 (26) | 175 (32) | ||
| College graduate or higher | 149 (27) | 175 (32) | 202 (37) | 230 (42) | 199 (36) | ||
| Body composition | |||||||
| BMI (kg/m2), mean±SD | 32.9±7.9 | 32.1±7.5 | 31.5±7.3 | 32.2±7.8 | 31.7±6.8 | 0.72 | 0.03d |
| FFM (kg), mean±SD | 63.0±15.8 | 58.6±14.2 | 58.0±13.4 | 59.5±14.3 | 59.3±14.3 | 0.11 | 0.006d |
| Waist circumference (cm), mean±SD | 108.4±17.4 | 106.1±16.0 | 104.7±16.4 | 106.6±17.0 | 105.4±15.5 | 0.49 | 0.02d |
| BP (mm Hg), mean±SD | |||||||
| SBP | 133±17 | 128±17 | 123±15 | 122±15 | 123±15 | 0.6111 | <0.001d |
| Diastolic BP | 71±11 | 70±10 | 69±10 | 69±10 | 71±10 | <0.001d | <0.001d |
| Medical history, n (%) | |||||||
| Diabetes | 342 (62) | 282 (51) | 232 (42) | 224 (41) | 247 (45) | 0.36 | <0.001d |
| Hypertension | 529 (97) | 514 (94) | 469 (86) | 479 (87) | 480 (88) | 0.29 | <0.001d |
| History of CVD | 262 (48) | 219 (40) | 191 (35) | 192 (35) | 179 (33) | 0.44 | <0.001d |
| History of atrial fibrillation | 123 (22) | 119 (22) | 105 (19) | 110 (20) | 97 (18) | 0.53 | 0.04d |
| History of COPD | 37 (7) | 39 (7) | 35 (6) | 32 (6) | 48 (9) | 0.16 | 0.46 |
| Current smoker | 68 (12) | 72 (13) | 61 (11) | 42 (8) | 59 (11) | 0.85 | 0.04d |
| History of cancer | 65 (12) | 71 (13) | 76 (14) | 80 (15) | 78 (14) | 0.86 | 0.17 |
| eGFR (CRIC equation) (ml/min per 1.73 m2), mean±SD | |||||||
| Index date (year-3 visit) | 28.0±13.0 | 35.7±14.7 | 42.4±15.7 | 48.2±14.7 | 60.6±17.8 | <0.001d | <0.001d |
| Baseline | 46.9±14.5 | 44.9±16.1 | 47.0±17.1 | 48.3±16.1 | 52.2±17.4 | <0.001d | <0.001d |
| Mean eGFR | 37.0±13.2 | 40.0±15.0 | 44.6±16.0 | 48.0±14.8 | 56.1±16.4 | <0.001d | <0.001d |
| Other laboratory markers | |||||||
| 24-H urinary protein (g/d), median (IQR) | 1.0 (0.2–2.7) | 0.3 (0.1–1.0) | 0.2 (0.1–0.5) | 0.1 (0.1–0.3) | 0.1 (0.1–0.2) | <0.001d | <0.001d |
| Urinary NGAL (ng/ml), median (IQR) | 16 (7–37) | 14 (6–30) | 14 (6–28) | 10.8 (5–23) | 10.5 (5–22) | 0.01d | <0.001d |
| Hemoglobin (g/dl), mean±SD | 12.2±1.5 | 12.6±1.6 | 13.0±1.6 | 13.1±1.5 | 13.3±1.5 | <0.001d | <0.001d |
| Phosphate (mg/dl), mean±SD | 3.7±0.6 | 3.6±0.6 | 3.6±0.6 | 3.6±0.6 | 3.7±0.6 | 0.03d | 0.981 |
| Calciume (mg/dl), mean±SD | 9.3±0.4 | 9.3±0.4 | 9.3±0.3 | 9.3±0.3 | 9.3±0.3 | 0.71 | 0.601 |
| iPTH (pg/ml), median (IQR) | 56 (36–87) | 58 (38–91) | 48 (33–80) | 43 (31–69) | 41 (29–61) | 0.001d | <0.001d |
| FGF-23 (pg/ml), median (IQR) | 146 (102–213) | 135 (94–211) | 130 (82–203) | 120 (86–190) | 120 (86–180) | 0.06 | <0.001d |
| Serum bicarbonate (mmol/L), mean±SD | 23.6±2.8 | 23.9±2.7 | 24.3±2.4 | 24.6±2.3 | 24.9±2.3 | <0.001d | <0.001d |
| Glucose (mg/dl), mean±SD | 121±40 | 114±39 | 110±36 | 108±32 | 112±36 | 0.31 | <0.001d |
| HbA1C (%), mean±SD | 7.0±1.7 | 6.6±1.5 | 6.4±1.4 | 6.3±1.2 | 6.3±1.3 | 0.46 | <0.001d |
| Total cholesterol (mg/dl), mean±SD | 181±37 | 182±38 | 179±34 | 180±35 | 181±36 | 0.54 | 0.77 |
| HDL (mg/dl), mean±SD | 46±13 | 48±15 | 4915 | 4815 | 4915 | 0.92 | 0.004d |
| LDL (mg/dl), mean±SD | 99±29 | 101±29 | 100±29 | 100±28 | 101±28 | 0.41 | 0.12 |
| Triglycerides (mg/dl), median (IQR) | 139 (102–189) | 133 (93–178) | 122 (89–172) | 121 (90–172) | 124 (93–178) | 0.63 | <0.001d |
| Serum albumin (g/dl), mean±SD | 3.9±0.4 | 4.0±0.3 | 4.1±0.3 | 4.1±0.3 | 4.1±0.3 | 0.003d | <0.001d |
| Urea nitrogen (mg/dl), median (IQR) | 34 (29–44) | 29 (23–40) | 26 (20–33) | 25 (19–32) | 21 (17–27) | <0.001d | <0.001d |
| Uric acid (mg/dl), mean±SD | 7.4±1.8 | 7.4±1.8 | 7.2±1.8 | 7.2±1.9 | 7.2±2.0 | 0.83 | 0.002d |
| Serum sodium (mmol/L), mean±SD | 139.6±2.1 | 139.8±2.1 | 139.5±2.0 | 139.6±2.1 | 139.5±1.9 | 0.80 | 0.10 |
| hs-CRP (mg/L), median (IQR) | 2.3 (1.0–5.5) | 2.4 (1.1–6.0) | 2.5 (1.0–6.2) | 2.5 (1.1–6.0) | 2.4 (1.0–5.9) | 0.45 | 0.82 |
| hs-Troponin T (pg/ml), median (IQR) | 14 (7–27) | 12 (6–20) | 10 (5–18) | 9 (4–17) | 9 (4–16) | 0.04d | <0.001d |
| NT-proBNP (pg/ml), median (IQR) | 175 (76–429) | 170 (76–382) | 111 (51–244) | 102 (43–251) | 84 (38–187) | 0.001d | <0.001d |
| Ankle-brachial index, mean±SD | 1.1 (1.0–1.2) | 1.1 (1.0–1.1) | 1.1 (1.0–1.1) | 1.1 (1.0–1.2) | 1.1 (1.0–1.2) | 0.35 | 0.006d |
| LV ejection fraction (%), median (IQR) | 54 (50–58) | 56 (51–60) | 56 (52–60) | 56 (52–60) | 56 (52–59) | 0.70 | 0.003d |
| Medications, n (%) | |||||||
| ACEIs and ARBs | 400 (73) | 387 (71) | 360 (66) | 366 (68) | 371 (68) | 0.38 | 0.03d |
| Diuretics | 375 (69) | 339 (62) | 291 (53) | 292 (54) | 253 (47) | 0.03d | <0.001d |
Variables susceptible to variation, planned to be repeated yearly between baseline and year-3 visit, are represented by arithmetic averages (sum of up to four values divided by up to four assessments). These include age, FFM, waist circumference, BMI, SBP, diastolic BP, 24-h urinary protein, hemoglobin, calcium, serum bicarbonate, glucose, total cholesterol, HDL, LDL, triglycerides, serum albumin, urea nitrogen, and serum sodium. The variables “number of hospitalizations” and “number of days spent in the hospital” represent the sum of values captured between the baseline and year-3 visit. The remaining variables are represented by one-time assessments: at baseline (urinary NGAL, serum phosphate, iPTH, FGF-23, HbA1C, uric acid, hs-CRP, hs-Troponin T, and NT-proBNP), at the year-1 visit (LV ejection fraction), or at the year-3 visit (use of ACEIs/ARBs and use of diuretics in the previous 30 d). BMI, body mass index; COPD, chronic obstructive pulmonary disease; IQR, interquartile range; NGAL, neutrophil gelatinase–associated lipocalin; iPTH, intact parathyroid hormone; FGF-23, fibroblast growth factor 23; HbA1C, hemoglobin A1C; hs-CRP, high-sensitivity C-reactive protein; hs-Troponin T; high-sensitivity Troponin T; NT-proBNP, N-terminal probrain natriuretic peptide; LV, left ventricular; ACEIs, angiotensin-converting enzyme inhibitors; ARBs, angiotensin receptor blockers.
P values for comparison between quintile 5 and quintile 3.
P trend represents P value for t test of a linear trend across all quintiles.
Cutoffs for quintiles of slopes of eGFR were rounded. Original values are the following: quintile 1, −11.38492 to −3.031002; quintile 2, −3.028516 to −1.889321; quintile 3, −1.889317 to −1.091369; quintile 4, −1.087086 to −0.0016062; quintile 5, −0.0004773 to 9.020505.
P<0.05.
Total serum calcium corrected for albumin.
Changes in Markers of CKD Comorbidities
We observed a longitudinal change in some markers of CKD comorbidities during the 3-year measurement window for the eGFR slope. Overall, there was a trend toward more severe changes in markers among individuals in quintile 1 of the eGFR slope, which lessened when moving toward quintile 5 (Table 2).
Table 2.
Slopes of covariates that were annually updated according to quintiles of slopes of eGFR
| Covariate | Quintile 1 | Quintile 2 | Quintile 3 | Quintile 4 | Quintile 5 | P Valuea | P Trendb |
|---|---|---|---|---|---|---|---|
| Number of observations | 548 | 548 | 547 | 548 | 547 | — | — |
| Slope range (ml/min per 1.73 m2 per year) | −11.4 to −3.03 | −3.02 to −1.89 | −1.88 to −1.09 | −1.09 to 0.00 | 0.00–9.02 | — | — |
| Log 24-h urinary protein (g/d per yr) | 0.06±0.14 | 0.05±0.12 | 0.05±0.11 | 0.05±0.11 | 0.05± 0.12 | 0.38 | 0.25 |
| BMI (kg/m2 per yr) | 0.07±0.71 | 0.09±0.50 | 0.06±0.61 | 0.06±0.55 | 0.01±0.67 | 0.20 | 0.24 |
| FFM (kg/yr) | −0.41±0.66 | −0.39±0.46 | −0.39±0.44 | −0.39±0.41 | −0.37±0.45 | 0.49 | 0.19 |
| Waist circumference (cm/yr) | 0.29±1.00 | 0.24±0.92 | 0.27±0.93 | 0.27±0.91 | 0.18±1.00 | 0.14 | 0.40 |
| SBP (mm Hg/yr) | 0.15±1.26 | 0.20±1.14 | 0.17±1.01 | 0.16±0.88 | 0.41±1.02 | <0.001c | <0.001c |
| Diastolic BP (mm Hg/yr) | −0.74±0.51 | −0.64±0.49 | −0.62±0.46 | −0.60±0.43 | −0.45±0.46 | <0.001c | <0.001c |
| Ankle-brachial index (U/yr) | 0.008±0.018 | 0.008±0.016 | 0.009±0.014 | 0.009±0.013 | 0.010±0.013 | 0.20 | 0.003c |
| Hemoglobin (g/dl per yr) | −0.10±0.14 | −0.04±0.12 | 0.001±0.11 | 0.01±0.11 | 0.02±0.11 | 0.02c | <0.001c |
| Phosphate (mg/dl/ per yr)d | 0.18±0.08 | 0.16±0.08 | 0.14±0.07 | 0.14±0.07 | 0.13±0.07 | 0.17 | <0.001c |
| Calcium (mg/dl per yr) | 0.048±0.015 | 0.050±0.012 | 0.050±0.011 | 0.050±0.011 | 0.049±0.012 | 0.24 | 0.27 |
| Log total PTH (pg/ml per yr)d | 0.10±0.10 | 0.07±0.08 | 0.06±0.06 | 0.05±0.06 | 0.05±0.05 | 0.44 | <0.001c |
| Log FGF-23 (pg/ml per yr)d | 0.11±0.11 | 0.06±0.08 | 0.03±0.08 | 0.13±0.08 | −0.01±0.08 | <0.001c | <0.001c |
| Bicarbonate (mmol/L per yr) | −0.32±0.16 | −0.28±0.14 | −0.26±0.13 | −0.24±0.12 | −0.23±0.12 | <0.001c | <0.001c |
| Glucose (mg/dl per yr) | 0.66±5.18 | 0.68±4.29 | 0.64±4.07 | 0.65±3.89 | 1.04±±4.56 | 0.13 | 0.08 |
| HbA1C (%/yr)e | −0.18±0.16 | −0.14±0.16 | −0.15±0.14 | −0.14±0.13 | −0.13±0.16 | 0.12 | <0.001c |
| Total cholesterol (mg/dl per yr) | −1.27±4.32 | −0.72±3.74 | −0.74±3.59 | −0.75±3.50 | −0.98±3.39 | 0.26 | 0.36 |
| HDL (mg/dl per yr) | −0.38±0.82 | −0.30±0.75 | −0.21±0.73 | −0.21±0.78 | −0.22±0.74 | 0.90 | <0.001c |
| LDL (mg/dl per yr) | −1.29±2.42 | −0.89±2.19 | −0.91±2.06 | −0.93±2.03 | 1.05±1.94 | 0.10 | 0.09 |
| Log triglycerides (mg/dl per yr) | −0.002±0.030 | −0.002±0.030 | −0.004±0.025 | −0.004±0.024 | −0.005±0.027 | 0.40 | 0.02c |
| Serum albumin (g/dl per yr) | −0.002±0.025 | −0.001±0.021 | 0.003±0.019 | 0.002±0.019 | −0.002±0.020 | <0.001c | 0.64 |
| Urea nitrogen (mg/dl per yr) | 3.62±2.63 | 1.69±1.97 | 0.83±1.27 | 0.18±1.10 | −0.52±1.21 | <0.001c | <0.001c |
| Serum sodium (mmol/L per yr)f | 0.15±0.00 | 0.15±0.00 | 0.15±0.00 | 0.15±0.00 | 0.15±0.00 | — | >0.99 |
BMI, body mass index; PTH, parathyroid hormone; FGF-23, fibroblast growth factor 23; HbA1C, hemoglobin A1C.
P values for comparison between quintile 5 and quintile 3.
P trend represents the P value for t test of a linear trend across all quintiles.
P<0.05.
Within mineral metabolism subcohort, including 907 subjects.
Among patients with diabetes.
Exceptionally, the variable sodium, assessed yearly as an integer, presented extremely low variability in the random coefficients used to derive the individual slopes (−1.73e−16 to 1.17e−16). As a result, the average slopes of sodium were virtually the same across quintiles, and no variability was observed, allowing up to six decimal digits (estimated average across quintiles was 0.152469, with undetected variability for all quintiles).
Rates of CKD Complications
eGFR improvers (quintile 5) had the lowest crude rates of death and cardiovascular events, whereas steep decliners (quintile 1) had the highest (Supplemental Table 4).
Correlations between slopes (main exposure) and eGFR (covariate) were 0.04 for baseline (year-0 visit) eGFR, 0.41 for mean eGFR (average between all available eGFR assessment between year-0 and year-3 visits), and 0.66 for the index eGFR (year-3 visit) (Figure 1, Supplemental Figure 4). To minimize the correlation between eGFR level and the slope of eGFR, adjustments in the main analysis included eGFR assessed at baseline. Given the high level of correlation between the eGFR assessed at the year-3 visit (index eGFR) and the slopes of eGFR, we did not pursue survival models adjusted for index eGFR.
Association between Slopes and Outcomes
Slopes of eGFR (continuous scale) were associated with the risks of death (P=0.004) and cardiovascular events (P=0.02), despite adjustments, in a nonlinear fashion (Figure 2, Supplemental Figure 5). In unadjusted models, steeper declines of eGFR were associated with increasingly higher risks, whereas lower declines/rises of eGFR were associated with progressively lower risks. For example, a slope of −3.5 ml/min per 1.73 m2 per year (1 SD below the average slope) was associated with a risk of death 36% higher than the average slope (hazard ratio [HR], 1.36; 95% CI, 1.24 to 1.48), whereas a slope of 0.5 ml/min per 1.73 m2 per year (1 SD above the average slope) was associated with a risk of death 24% lower than the average slope (HR, 0.76; 95% CI, 0.65 to 0.88).
Figure 2.
Association between slopes of eGFR and the relative risks of death and CVD events. (A) Unadjusted association between slopes of eGFR and the risk of death (model 1). (B) Multivariable-adjusted association between slopes of eGFR and the risk of death (model 8). (C) Unadjusted association between slopes of eGFR and the risk of cardiovascular events (model 1). (D) Multivariable-adjusted association between slopes of eGFR and the risk of cardiovascular events (model 8). Model 8 includes adjustments for age, sex, race, educational attainment, smoking status, history of CVD, diabetes, SBP, urinary protein, eGFR, FFM, body mass index, waist circumference, ejection fraction, serum sodium, HDL cholesterol, high-sensitivity Troponin-T, N-terminal probrain natriuretic peptide, reported use of angiotensin-converting enzyme inhibitors/angiotensin receptor blockers and diuretics, hemoglobin, serum bicarbonate, albumin, phosphate, fibroblast growth factor 23, intact parathyroid hormone, high-sensitivity C-reactive protein, urinary neutrophil gelatinase–associated lipocalin, number of hospitalizations, and the number of days spent in the hospital in the 3 years between baseline and the index date. The solid black lines represent HRs. Boundaries of the 95% CIs for the HRs are shown by the solid gray lines. Reference risk for a slope of eGFR of −1.5 ml/min per 1.73 m2 per year (mean slope of eGFR for the study population) is depicted by dashed lines. Reported P values are the results of chi-squared tests, checking for the overall association between the slope of eGFR and the risk of all-cause mortality. Slopes of eGFR (CRIC Study equation) were defined using annual assessments of eGFR over a 3-year period, using linear, mixed-effects models. Analyses were performed in multiple imputed datasets. The estimated HRs for the slopes of −5.5 (2 SD below the average), −3.5 (1 SD below the average), 0.5 (1 SD above the average), and 2.0 (2 SD above the average) are presented for each model in the tables below the graphs. Ref, reference.
In multiple adjusted models, steeper declines of eGFR were consistently associated with increased risk of death and CVD events. For example, in model 8, including all proposed adjustments, a slope of −3.5 ml/min per 1.73 m2 per year was associated with a risk of death 23% higher than the average slope (HR, 1.23; 95% CI, 1.09 to 1.39). However, for lower declines/rises of eGFR, the protective relationship initially observed was attenuated, becoming flat and null across all adjusted models. For example, in model 8, a slope of 0.5 ml/min per 1.73 m2 per year was associated with a risk of death not different from the average slope (HR, 0.95; 95% CI, 0.81 to 1.11).
The trend in adjusted associations of slopes of eGFR, categorized in quintiles, with the risks of death and CVD events were similar. Compared with quintile 3, we observed higher risks in quintile 1 and lower risks in quintile 5 (Supplemental Table 5).
Subgroups
We observed significant effect modification for the association between slopes of eGFR and risks of death and cardiovascular events (Supplemental Table 6). Subgroups formed on the basis of the rate of change in urinary protein experienced different associations between the slope of eGFR and death. Among those with relatively stable urinary protein (change rate <0.05 g/24 h per year, log scale), declines of eGFR steeper than the mean decline were associated with progressively higher risks of death, whereas lower declines/rises were not. Among those with increasing urinary protein (change rate ≥0.05 g/24 h per year, log scale), no association between slopes of eGFR and the risk of death was observed (Figure 3). Subgroups of race/ethnicity experienced different associations between the slope of eGFR and the risk of cardiovascular events. Among those of racial minorities, including non-Hispanic Black and Hispanic patients, declines/rises of eGFR lower than the mean decline were associated with progressively lower risks of cardiovascular events, whereas steeper declines were not associated with increased risk. Among non-Hispanic White patients, slopes of eGFR were not associated with the risk of cardiovascular events (Figure 3).
Figure 3.
Association between slopes of eGFR and the relative risks of death and CVD events according to significant subgroups. (A) Association between slopes of eGFR and the risk of death among subjects who presented a rate of change of (log) urinary protein over time <0.05 g/24 h per year during the period in which slopes were defined. (B) Association between slopes of eGFR and the risk of death among subjects who presented a rate of change of (log) urinary protein over time ≥0.05 g/24 h per year during the period in which slopes were defined. (C) Association between slopes of eGFR and the risk of CVD events among non-Hispanic White subjects. (D) Association between slopes of eGFR and the risk of CVD events among non-Hispanic Black and Hispanic subjects. Models were adjusted for age, sex, race, educational attainment, smoking status, history of CVD, diabetes, SBP, urinary protein, eGFR, FFM, body mass index, waist circumference, ejection fraction, serum sodium, HDL cholesterol, high-sensitivity Troponin-T, N-terminal probrain natriuretic peptide, reported use of angiotensin-converting enzyme inhibitors/angiotensin receptor blockers and diuretics, hemoglobin, serum bicarbonate, albumin, phosphate, fibroblast growth factor 23, intact parathyroid hormone, high-sensitivity C-reactive protein, urinary neutrophil gelatinase–associated lipocalin, number of hospitalizations, and the number of days spent in the hospital in the 3 years between baseline and the index date. HRs are depicted by the solid black lines. Boundaries of the 95% CIs for the HRs are shown by the solid gray lines. Reference risk for a slope of eGFR of −1.5 ml/min per 1.73 m2 per year (mean slope of eGFR for the study population) is depicted by dashed lines. Reported P values are the results of chi-squared tests, checking for the overall association between the slope of eGFR and the risk of all-cause mortality. Slopes of eGFR (CRIC Study equation) were defined using annual assessments of eGFR over a 3-year period, using linear, mixed-effects models. Analyses were performed in multiple imputed datasets. The estimated HRs for the slopes of −5.5 (2 SD below the average), −3.5 (1 SD below the average), 0.5 (1 SD above the average), and 2.0 (2 SD above the average) are presented for each model in the tables below the graphs. Subgroup analyses were performed on the basis of significant interactions on the association between slopes of eGFR and the risks of death and CVD events. Ref, reference.
Sensitivity Analyses
Cox models examining the association between slopes defined over 5 years and the risks of death and cardiovascular events showed similar results to those observed in the main analysis, with increased risks for steep decliners and lower risks for lower decliners/improvers. However, the adjusted associations between the 5-year slopes of eGFR and longitudinal outcomes were significant for death, but not for cardiovascular events (Supplemental Figure 6, Supplemental Table 7). Models using eGFR slopes defined using the CKD-EPI equation yielded results consistent with those obtained with the CRIC equation estimates (Supplemental Figures 7 and 8, Supplemental Tables 8 and 9). Slopes estimated through ordinary least squares also yielded similar results (Supplemental Figure 9, Supplemental Table 10). No violation of the PH assumption was observed. Lastly, models restricted to complete case analysis were fit and demonstrated associations consistent with those that incorporated imputed data.
Discussion
Among participants in the CRIC Study, declines in eGFR greater than the mean decline were associated with progressively increased risks of death and cardiovascular events. These associations remained despite adjustment for clinical, physiologic, and biologic factors associated with these outcomes. These findings support the potential value of eGFR slopes in the clinical assessment of adults with CKD. Contrary to previously published work,6–10 a loss of eGFR less than the average loss and a rise in eGFR was associated with better clinical outcomes. The slope of eGFR was particularly associated with the risk of death among individuals with stable levels of urinary protein over time. Further, among non-Hispanic Black and Hispanic patients, reduced risk of CVD was observed after lower declines/rising levels of eGFR.
In a series of meta-analyses, the CKD Prognosis Consortium9,10,25 examined the associations between slopes of eGFR and longitudinal risks of various outcomes. In a study focusing on the risk of death, 250,000 individuals with moderate CKD from 12 cohorts were included.10 They observed an increased adjusted risk of death, both for declining and improving eGFR. However, >85% of the individuals studied were from the Veterans Affairs health system, comprising an older (average age 76±10 years), less diverse population (9% women, 9% Black), with less severe CKD than we examined here. Despite the robust number of participants and the consistency of conclusions across sensitivity analyses, results were heterogeneous from cohort to cohort. Further, much of the data in that study was obtained from medical records, where patients who were sicker tend to be retained and reassessed more often. Also, clinical events, like AKI episodes, may have contributed to variations in eGFR. Lastly, subgroups associated with more rapid loss of kidney function—such as Hispanic and non-Hispanic Black individuals, those with a history of diabetes, those with a history of CVD, and those with higher levels of proteinuria—were less represented in that work.
Many studies have reported no differences in levels of association between slopes of eGFR and the risk of longitudinal outcomes across strata of urinary protein measured at baseline.6,8,10 We assessed how different levels of longitudinal changes in urinary protein modified the association between slopes of eGFR and the risk of death. Consistent with our hypothesis, in the setting of stable proteinuria over time, there was an association between steeper declining eGFR and the risk of death, although this association was attenuated when proteinuria worsened over time. These findings underscore the importance of proteinuria as a risk factor for CVD and death, independent of level and changes in eGFR.
No differences in the association between slopes of eGFR and longitudinal outcomes across race subgroups have been previously identified.7,26 The small proportions of individuals from racial minority groups included in these studies, varying from 10% to 13%, may have accounted for why their findings differed from those presented here. Although these minorities have been strongly associated with steeper declines of eGFR,27,28 in our population, they were well represented across all quintiles of slopes of eGFR. Also, the discordance between a protective association with cardiovascular events and no protective association with death among racial minorities was unexpected.29 We speculate that other causes of death, like infections and cancer, may be increased among improvers from racial minorities, counteracting the benefit arising from their lower relative risk of cardiovascular events. Studying a larger population of individuals of the non-White races is needed to better understand this.
Overall, the various explanatory models we proposed did not change the relationship between the slopes of eGFR and outcomes. These analyses were implemented mainly to explore a possible association between positive slopes and increased risks observed by other investigators. Although rising slopes of eGFR may be associated with increased risks of death and CVD events in different populations of individuals with moderate to severe CKD, we did not observe this association in a selected group of the CRIC Study population. Also, the associations between slopes of eGFR and outcomes were consistent using different filtration markers to assess the main exposure. Defining slopes using the combination of cystatin C and creatinine or cystatin C alone yielded similar associations with death and CVD events, suggesting that confounding by muscle mass did not play a large role in the observed associations.
Strengths of our study include the fact that eGFR was repeated several times, using both creatinine and cystatin C. In addition, the CRIC Study had a high rate of participant retention, and its participants were evaluated yearly in planned study visits, minimizing the effect of clinically driven eGFR assessments that may be performed in response to clinical status related to eGFR fluctuations. Additionally, the use of linear, mixed-effects models, as opposed to ordinary least squares to estimate slope, provided for precise estimation of eGFR slope. Residual confounding was minimized compared with previous studies because we were able to adjust for numerous factors, including FFM, direct assessments of cardiovascular function, inflammatory markers, markers of metabolic complications, and hospitalizations in our analyses. Also, changes in a variety of laboratory markers were assessed to characterize different levels of slopes of eGFR. However, our findings should be interpreted in light of our study’s limitations. First, the population included in this study is a nonrandom subset of the entire CRIC Study population. Second, the estimation of slopes was made on the basis of eGFR assessments performed yearly, limiting our ability to assess eGFR variation over shorter intervals. Third, we were not able to assess if increases in eGFR were accompanied by changes in kidney histopathology, reducing our ability to infer the reasons for improving eGFR, such as the resolution of fibrosis or recovery from AKI. Finally, we were unable to evaluate the frequency of AKI episodes across our population.
In a population of racially/ethnically diverse adult men and women with moderate to severe CKD followed yearly, the slope of eGFR was independently associated with the risks of death and cardiovascular events after extensive adjustment for clinical, physiologic, and biologic factors. Declines of eGFR steeper than the average decline were associated with an increased risk of death and CVD events. However, rises of eGFR or declines lower than the average decline were not associated with levels of risk higher than those presented by the average slope. Declining eGFR was particularly associated with an increased risk of death among individuals with stable urinary protein, whereas improving eGFR was associated with a reduced risk of cardiovascular events among racial minorities. Further validation of these findings requires studies that include larger proportions of racial minorities.
Disclosures
A. Anderson reports receiving personal fees from Kyowa Hakko Kirin, outside the submitted work. H. Feldman reports having consultancy agreements with DLA Piper LLP, InMed Inc., Kyowa Hakko Kirin Co., Ltd. (ongoing), and National Kidney Foundation; receiving research funding from Regeneron; receiving honoraria from Rogosin Institute (invited speaker); being a scientific advisor for and a member of the CRIC Study (steering committee); and being a scientific advisor and editor-in-chief for the National Kidney Foundation. D. Raj reports having consultancy agreements with Kaleido, Medimmune, and Relypsa; receiving research funding from Carvedia, Kaleido Pharmaceuticals, Medimmune, National Institutes of Health (NIH), Relypsa; receiving honoraria from AstraZeneca, Carvedia, and Relypsa; and being a scientific advisor for or a member of the National Institute of Diabetes and Digestive and Kidney Diseases, National Heart, Lung, and Blood Institute, Relypsa, and Washington Academy of Medicine. All remaining authors have nothing to disclose.
Funding
Funding for the CRIC Study was obtained under a National Institute of Diabetes and Digestive and Kidney Diseases cooperative agreement, grants U01DK060990, U01DK060984, U01DK061022, U01DK061021, U01DK061028, U01DK060980, U01DK060963, and U01DK060902. In addition, this work was supported in part by the National Center for Advancing Translational Sciences (NCATS) grants UL1TR000003 (Perelman School of Medicine at the University of Pennsylvania Clinical and Translational Science Award) and UL1TR000439 (from the NCATS component of the NIH and NIH Roadmap for Medical Research); Johns Hopkins University grant UL1TR000424; University of Maryland General Clinical Research Center grant M01RR-16500; Michigan Institute for Clinical and Health Research grant UL1TR000433; Center for Clinical and Translational Science, University of Illinois at Chicago grant UL1RR029879; Tulane University grant P20 GM109036 (COBRE for Clinical and Translational Research in Cardiometabolic Diseases); and Kaiser Permanente grant UL1 RR-024131 (NIH/National Center for Research Resources University of California, San Francisco–Clinical and Translational Science Institute).
Supplementary Material
Acknowledgments
Dr. Paula F. Orlandi designed the study; Dr. Harold I. Feldman and Dr. Amanda H. Anderson supervised the elaboration of the study design; Dr. Paula F. Orlandi analyzed the data; Dr. Dawei Xei and Dr. Wei Yang supervised the data analysis; Dr. Paula F. Orlandi made the figures, drafted and revised the paper; and all authors revised and approved the final version of the manuscript. This manuscript was developed under the scope of a Master of Clinical Epidemiology protocol and considered for the obtention of the MsCE degree by the corresponding author, Dr. Paula F. Orlandi, under the mentorship of Dr. Harold I. Feldman. Many thanks for the CRIC study team, participants, and their families for making this work possible. This work is dedicated to Chloe and Joseph.
Footnotes
Published online ahead of print. Publication date available at www.jasn.org.
Contributor Information
Collaborators: Chronic Renal Insufficiency Cohort (CRIC) Study Investigators, Lawrence J. Appel, Alan S. Go, James P. Lash, Panduranga S. Rao, and Mahboob Rahman
Supplemental Material
This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2020040476/-/DCSupplemental.
Supplemental Figure 1. Timeline for assessment of covariates applied in the Cox models.
Supplemental Figure 2. Distributions of slopes according to different methods.
Supplemental Figure 3. Mean eGFR over time across subgroups of the study population.
Supplemental Figure 4. Correlation between slopes and eGFR assessments.
Supplemental Figure 5. Relative risk for death and CVD events (main exposure).
Supplemental Figure 6. Relative risk for death and CVD events (sensitivity analysis: 5-year slopes).
Supplemental Figure 7. Relative risk for death and CVD events (sensitivity analysis: CKD-EPI creatinine).
Supplemental Figure 8. Relative risk for death and CVD events (sensitivity analysis: CKD-EPI cystatin C).
Supplemental Figure 9. Relative risk for death and CVD events (sensitivity analysis: OLS).
Supplemental Table 1. Comparison between included and excluded CRIC participants.
Supplemental Table 2. Proportion of missing data imputed for the Cox models.
Supplemental Table 3. Characteristics of study participants within the fifth quintile.
Supplemental Table 4. Incidence rates of death and CVD events according to quintiles of slope.
Supplemental Table 5. Association between slopes and outcomes by quintiles (main exposure).
Supplemental Table 6. Effect modification on the association between slopes and outcomes.
Supplemental Table 7. Association between slopes and outcomes by quintiles (5-year slopes).
Supplemental Table 8. Association between slopes and outcomes by quintiles (CKD-EPI creatinine).
Supplemental Table 9. Association between slopes and outcomes by quintiles (CKD-EPI cystatin C).
Supplemental Table 10. Association between slopes and outcomes by quintiles (OLS).
References
- 1.O’Hare AM, Batten A, Burrows NR, Pavkov ME, Taylor L, Gupta I, et al.: Trajectories of kidney function decline in the 2 years before initiation of long-term dialysis. Am J Kidney Dis 59: 513–522, 2012. 10.1053/j.ajkd.2011.11.044 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Hu B, Gadegbeku C, Lipkowitz MS, Rostand S, Lewis J, Wright JT, et al.; African-American Study of Kidney Disease and Hypertension Group: Kidney function can improve in patients with hypertensive CKD. J Am Soc Nephrol 23: 706–713, 2012. 10.1681/ASN.2011050456 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Levey AS, Coresh J, Greene T, Marsh J, Stevens LA, Kusek JW, et al.; Chronic Kidney Disease Epidemiology Collaboration: Expressing the Modification of Diet in Renal Disease Study equation for estimating glomerular filtration rate with standardized serum creatinine values. Clin Chem 53: 766–772, 2007. 10.1373/clinchem.2006.077180 [DOI] [PubMed] [Google Scholar]
- 4.Hunsicker LG, Adler S, Caggiula A, England BK, Greene T, Kusek JW, et al.: Predictors of the progression of renal disease in the modification of diet in renal disease study. Kidney Int 51: 1908–1919, 1997. [DOI] [PubMed] [Google Scholar]
- 5.Khan NA, Ma I, Thompson CR, Humphries K, Salem DN, Sarnak MJ, et al.: Kidney function and mortality among patients with left ventricular systolic dysfunction. J Am Soc Nephrol 17: 244–253, 2006. 10.1681/ASN.2005030270 [DOI] [PubMed] [Google Scholar]
- 6.Perkins RM, Bucaloiu ID, Kirchner HL, Ashouian N, Hartle JE, Yahya T: GFR decline and mortality risk among patients with chronic kidney disease. Clin J Am Soc Nephrol 6: 1879–1886, 2011. 10.2215/CJN.00470111 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Al-Aly Z, Zeringue A, Fu J, Rauchman MI, McDonald JR, El-Achkar TM, et al.: Rate of kidney function decline associates with mortality. J Am Soc Nephrol 21: 1961–1969, 2010. 10.1681/ASN.2009121210 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Matsushita K, Selvin E, Bash LD, Franceschini N, Astor BC, Coresh J: Change in estimated GFR associates with coronary heart disease and mortality. J Am Soc Nephrol 20: 2617–2624, 2009. 10.1681/ASN.2009010025 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Coresh J, Turin TC, Matsushita K, Sang Y, Ballew SH, Appel LJ, et al.: Decline in estimated glomerular filtration rate and subsequent risk of end-stage renal disease and mortality. JAMA 311: 2518–2531, 2014. 10.1001/jama.2014.6634 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Naimark DMJ, Grams ME, Matsushita K, Black C, Drion I, Fox CS, et al.; CKD Prognosis Consortium: Past decline versus current eGFR and subsequent mortality risk. J Am Soc Nephrol 27: 2456–2466, 2016. 10.1681/ASN.2015060688 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Lash JP, Go AS, Appel LJ, He J, Ojo A, Rahman M, et al.; Chronic Renal Insufficiency Cohort (CRIC) Study Group: Chronic Renal Insufficiency Cohort (CRIC) Study: Baseline characteristics and associations with kidney function [published correction appears in Clin J Am Soc Nephrol 6: 2548–2553, 2011]. Clin J Am Soc Nephrol 4: 1302–1311, 2009. 10.2215/CJN.00070109 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Feldman HI, Appel LJ, Chertow GM, Cifelli D, Cizman B, Daugirdas J, et al.: Chronic Renal Insufficiency Cohort (CRIC) Study Investigators : The chronic renal insufficiency cohort (CRIC) study: Design and methods. J Am Soc Nephrol 14[Suppl 2]: S148–S153, 2003. [DOI] [PubMed] [Google Scholar]
- 13.Joffe M, Hsu CY, Feldman HI, Weir M, Landis JR, Hamm LL; Chronic Renal Insufficiency Cohort (CRIC) Study Group: Variability of creatinine measurements in clinical laboratories: Results from the CRIC Study. Am J Nephrol 31: 426–434, 2010. 10.1159/000296250 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Anderson AH, Yang W, Hsu CY, Joffe MM, Leonard MB, Xie D, et al.; CRIC Study Investigators: Estimating GFR among participants in the chronic renal insufficiency cohort (CRIC) Study. Am J Kidney Dis 60: 250–261, 2012. 10.1053/j.ajkd.2012.04.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Wilson FP, Xie D, Anderson AH, Leonard MB, Reese PP, Delafontaine P, et al.; CRIC Study Investigators: Urinary creatinine excretion, bioelectrical impedance analysis, and clinical outcomes in patients with CKD: The CRIC Study. Clin J Am Soc Nephrol 9: 2095–2103, 2014. 10.2215/CJN.03790414 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Dubin RF, Deo R, Bansal N, Anderson AH, Yang P, Go AS, et al.; CRIC Study Investigators: Associations of conventional echocardiographic measures with incident heart failure and mortality: The chronic renal insufficiency cohort. Clin J Am Soc Nephrol 12: 60–68, 2017. 10.2215/CJN.02700316 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Shou H, Hsu JY, Xie D, Yang W, Roy J, Anderson AH, et al.; Chronic Renal Insufficiency Cohort (CRIC) Study Investigators: Analytic considerations for repeated measures of eGFR in cohort studies of CKD. Clin J Am Soc Nephrol 12: 1357–1365, 2017. 10.2215/CJN.11311116 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Madero M, Katz R, Murphy R, Newman A, Patel K, Ix J, et al.: Comparison between different measures of body fat with kidney function decline and incident CKD. Clin J Am Soc Nephrol 12: 893–903, 2017. 10.2215/CJN.07010716 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.D’Agati VD, Chagnac A, de Vries APJ, Levi M, Porrini E, Herman-Edelstein M, et al. : Obesity-related glomerulopathy: Clinical and pathologic characteristics and pathogenesis. Nat Rev Nephrol 12: 453–471, 2016. 10.1038/nrneph.2016.75 [DOI] [PubMed] [Google Scholar]
- 20.Kovesdy CP, Lott EH, Lu JL, Malakauskas SM, Ma JZ, Molnar MZ, et al.: Hyponatremia, hypernatremia, and mortality in patients with chronic kidney disease with and without congestive heart failure. Circulation 125: 677–684, 2012. 10.1161/CIRCULATIONAHA.111.065391 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Huang H, Jolly SE, Airy M, Arrigain S, Schold JD, Nally JV, et al.: Associations of dysnatremias with mortality in chronic kidney disease. Nephrol Dial Transplant 32: 1204–1210, 2017. 10.1093/ndt/gfw209 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Kovesdy CP, George SM, Anderson JE, Kalantar-Zadeh K: Outcome predictability of biomarkers of protein-energy wasting and inflammation in moderate and advanced chronic kidney disease. Am J Clin Nutr 90: 407–414, 2009. 10.3945/ajcn.2008.27390 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF 3rd, Feldman HI, et al.; CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration): A new equation to estimate glomerular filtration rate [published correction appears in Ann Intern Med 155: 408, 2011]. Ann Intern Med 150: 604–612, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Rubin DB: Multiple Imputation for Nonresponse in Surveys, 1st Ed., Hoboken, NJ, John Wiley & Sons, 1987 [Google Scholar]
- 25.Kovesdy CP, Coresh J, Ballew SH, Woodward M, Levin A, Naimark DM, et al.; CKD Prognosis Consortium: Past decline versus current eGFR and subsequent ESRD risk. J Am Soc Nephrol 27: 2447–2455, 2016. 10.1681/ASN.2015060687 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Rifkin DE, Shlipak MG, Katz R, Fried LF, Siscovick D, Chonchol M, et al. : Rapid kidney function decline and mortality risk in older adults. Arch Intern Med 168: 2212–2218, 2008. 10.1001/archinte.168.20.2212 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Parsa A, Kao WHL, Xie D, Astor BC, Li M, Hsu CY, et al.; AASK Study Investigators; CRIC Study Investigators: APOL1 risk variants, race, and progression of chronic kidney disease. N Engl J Med 369: 2183–2196, 2013. 10.1056/NEJMoa1310345 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Michael J, Hsu JY, Lora CM, Ricardo AC, Anderson AH, Bazzano L, et al. ; Chronic Renal Insufficiency Cohort (CRIC) Study Investigators: CKD progression and mortality among Hispanics and non-Hispanics. J Am Soc Nephrol 27: 3488–3497, 2016. 10.1681/ASN.2015050570 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Thompson S, James M, Wiebe N, Hemmelgarn B, Manns B, Klarenbach S, et al.; Alberta Kidney Disease Network: Cause of death in patients with reduced kidney function. J Am Soc Nephrol 26: 2504–2511, 2015. 10.1681/ASN.2014070714 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.