Abstract
Rationale and Objective.
While low estimated glomerular filtration rate (eGFR) is associated with cardiovascular disease (CVD) events and mortality, the clinical significance of variability in eGFR over time is uncertain. We aimed to evaluate the association between variability in eGFR and the risk of CVD events and all-cause mortality.
Study Design.
Longitudinal analysis of clinical trial participants.
Settings and Participants.
7,520 Systolic Blood Pressure Intervention Trial (SPRINT) participants aged ≥ 50 year with 1 or more CVD risk factors.
Predictors.
eGFR variability, estimated by the coefficients of variation of eGFR measurements at the 6, 12, and 18-month study visits.
Outcomes.
SPRINT primary CVD composite outcome (myocardial infarction, acute coronary syndrome, stroke, heart failure, or CVD death) and all-cause mortality from month 18 to end of follow-up.
Analytical Approach.
Cox models evaluated associations between eGFR variability and CVD outcomes and all-cause mortality. Models were adjusted for demographics, randomization arm, CVD risk factors, albuminuria and month 18 eGFR.
Results.
Mean age was 68±9 years, 65% were men, and 58% were white. The mean eGFR was 73±21 ml/min/1.73m2 at 6 months. There were 370 CVD events and 154 deaths during a median follow-up of 2.4 years. Greater eGFR variability was associated with higher risk for all-cause mortality (hazard ratio (HR) per standard deviation (SD) greater variability, 1.29; 95% confidence interval (CI) 1.14 to 1.45) but not CVD events (HR 1.05; 95% CI 0.95 to 1.16) after adjusting for albuminuria at baseline, eGFR at month 18, and other CVD risk factors. Associations were similar when stratified by treatment arm and baseline CKD status, when accounting for concurrent systolic BP changes, use of angiotensin-converting enzyme (ACE) inhibitors or angiotensin receptor blockers (ARBs), and diuretic medications during follow-up.
Limitations.
Persons with diabetes and proteinuria > 1 g/day were excluded.
Conclusions.
In trial participants at high risk for CVD with hypertension, greater eGFR variability was independently associated with all-cause mortality but not CVD events.
Index Words: estimated glomerular filtration rate, variability, cardiovascular events, all-cause mortality
Plain-language Summary.
Physicians observe considerable visit-to-visit variability in eGFR in clinical practice. Little is known about the clinical significance of eGFR variability over time. We evaluated associations between eGFR variability and subsequent CVD events and all-cause mortality among SPRINT participants. We find that greater eGFR variability was associated with higher risk of all-cause mortality independent of baseline eGFR, albuminuria, and other risk factors. Greater eGFR variability was not associated with CVD events in a fully-adjusted model.
INTRODUCTION
Multiple studies in a variety of settings have demonstrated that lower estimated glomerular filtration rate (eGFR) is strongly associated with cardiovascular disease (CVD) and all-cause mortality.1–3 In clinical practice, some patients have greater visit-to-visit variability in eGFR than others, and much less is known about the clinical significance of this eGFR variability. Greater eGFR variability may indicate the presence of worse macro- or micro-vascular disease within the renal arteries or the kidney itself, or reduced renal reserve/nephron mass resulting in impaired ability to auto-regulate blood flow in response to modest changes in fluid status or renal perfusion.4–5 If so, then eGFR variability may mark subclinical vascular disease, and may therefore be a risk factor for subsequent CVD events.
Some prior studies have linked eGFR variability with risk of kidney disease progression, CVD events and all-cause mortality.6–11 These studies have typically evaluated populations with prevalent chronic kidney disease (CKD) or kidney transplant recipients. A broader evaluation in other high-risk populations, in particular hypertensive patients, is needed to assess the clinical implications of eGFR variability in these settings.
The Systolic Blood Pressure Intervention Trial (SPRINT) was a randomized clinical trial where hypertensive individuals at high risk for CVD with and without CKD were randomized to intensive (<120 mm Hg) or standard (<140 mm Hg) systolic blood pressure (BP) targets for prevention of CVD events.12 Per protocol, all participants had repeat assessment of kidney function at pre-specified follow-up visits, providing a uniform platform to assess intra-individual eGFR variability. Here, we determine the relationship of eGFR variability with subsequent CVD events and all-cause mortality in SPRINT. We also evaluated whether or not observed relationships were similar in persons with and without CKD at baseline, and in the standard and the intensive BP arms of the trial.
METHODS
Study Participants
The design and primary outcomes of the SPRINT trial have been described elsewhere.13 Briefly, participants were recruited from 102 clinical centers in the United States and Puerto Rico. Inclusion criteria required age ≥50 years, systolic BP 130–180 mmHg depending on the number of antihypertensive medications used at the time of recruitment, and high risk for CVD events (prior clinical or subclinical cardiovascular disease other than stroke, 10-year risk of cardiovascular disease of 15% or greater based on the Framingham risk score, CKD defined as eGFR 20–59 ml/min/1.73m2, or age ≥75 years). Major exclusion criteria included diabetes mellitus, proteinuria >1 g/d, polycystic kidney disease, prior stroke or transient ischemic attack, symptomatic heart failure, or a left ventricular ejection fraction <35%.
A total of 9,361 participants were enrolled between November 2010 and March 2013.9 All participants provided written informed consent and were randomly assigned (1:1 ratio) to the intensive or standard systolic BP arm. The antihypertensive regimens were adjusted to achieve and maintain systolic BP according to randomized targets. Participants attended visits monthly for the first 3 months and every 3 months thereafter, and venous blood was subsequently obtained every 6 months and stored at −80° Celsius at the central laboratory at the University of Minnesota for future analysis. Institutional Review Boards at all participating institutions approved the study.
Prior analyses from SPRINT demonstrated more rapid declines in eGFR from baseline to month 6 in participants randomized to the intensive arm relative to the standard arm,14 which likely reflects the hemodynamic changes induced by BP lowering.15 We therefore initiated this analysis from the month 6 visit onwards. We excluded participants who had a CVD event prior to the 18-month study visit (n=177), and those who had missing eGFR measurements at 6, 12 and 18- month study visits (n=1635). We further excluded 29 participants who dropped out of the trial prior to 18 months, leaving 7,520 participants for our analytic sample (Figure 1).
Figure 1.
Flow chart of study participants.
Assessment of eGFR Variability and Clinical Outcomes
Serum creatinine was measured by an enzymatic procedure (Roche, Indianapolis, IN) at the SPRINT central laboratory and eGFR was calculated using the 4-variable Modification of Diet in Renal Disease (MDRD) equation.14 We defined eGFR variability as the coefficient of variation (standard deviation (SD) divided by mean eGFR measured at 6, 12, and 18-month study visits) for each participant. The study outcomes included: 1) the SPRINT primary CVD composite (myocardial infarction, acute coronary syndrome, stroke, heart failure, or CVD death), and 2) all-cause mortality. The study participants were followed from the 18-month visit to the date of the first occurrence of the outcome of interest (CVD event or death) or end of SPRINT follow-up.13
Statistical Analysis
Continuous variables were described as the mean (SD) or median and interquartile ranges (IQR) and categorical variables as absolute (n) and relative (%) frequency. The eGFR variability was expressed as coefficient of variation and slope of eGFR change over time was calculated by means of linear regression of eGFR by time. We computed the distribution of participant characteristics by quartiles of eGFR variability. To explore independent correlates of eGFR variability, we set eGFR variability as the dependent variable in a linear regression model, and we identified the factors that remained independently associated with eGFR variability with multivariable adjustment. Cox proportional hazards models were employed to examine associations between eGFR variability and subsequent CVD events and all-cause mortality, and evaluated eGFR variability as a continuous variable in our primary analysis. We also evaluated associations across quartiles with the lowest quartile set as the reference group. A series of nested models were developed: Model 1 adjusted for age, sex, race, and randomization arm. Model 2, our fully adjustment model, additionally included prevalent CVD, prevalent heart failure, current smoking, body mass index, serum total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, systolic BP at month 18, albuminuria at baseline, eGFR at month 18, , systolic BP at month 18, diastolic BP at month 18, medications (angiotensin-converting enzyme inhibitors (ACEI) or angiotensin II receptor blockers (ARB), calcium channel blockers (CCBs) or diuretic use) at month 18, and fasting status. To investigate the functional form, we also constructed natural piecewise-cubic spline functions with the specified sequence of interior knots placed at the quartiles of eGFR variability. To determine whether eGFR variability had similar relationships with subsequent CVD events and all-cause mortality in each trial arm and in those with eGFR < 60 ml/min/1.73m2 vs. without CKD at baseline, we explored stratified analyses and tested multiplicative interaction terms. Finally, we performed sensitivity analysis based on multiple imputation of missing variables. Missing covariates were multiply imputed using chained equations. The multiple analyses over the imputations were combined using Rubin’s rules to account for the variability in the imputation procedure.16 All analyses were conducted using Stata (StatCorp 2013, Stata Statistical Software: Release 13, StataCorp LP, College Station, TX, USA) and SPSS (IBM Corp. Released 2016. IBM SPSS Statistics for Windows, Version 24.0. Armonk, NY: IBM Corp.). P-values <0.05 were considered statistically significant for all analyses including interaction terms.
RESULTS
Among 7,520 SPRINT participants, the mean age was 68±9 years, 65% were male, 58% were non-Hispanic whites, and 20% had prevalent CVD. The mean baseline (month 0) systolic and diastolic BP were 139±15 mm Hg and 78±12mm Hg, respectively. The mean eGFR was 73±21 ml/min/1.73m2 at the 6-month and median urine albumin to creatinine ratio (ACR) was 9 (IQR 6–20) mg/g at the baseline SPRINT visit. Table 1 shows demographic, clinical, and laboratory characteristics across quartiles of eGFR variability. Compared with participants with lower eGFR variability, those in the highest quartile were more likely to be female, African-American, and current smokers, to have prevalent CVD and heart failure, to have been randomized to the intensive BP arm, and to be users of ACEI/ARB and diuretics. They also had slightly lower mean eGFR and slightly higher ACR than those with lower eGFR variability.
Table 1:
Baseline characteristics of study participants by quartiles of eGFR variability
| Coefficient of variation of eGFR variability | |||||
|---|---|---|---|---|---|
| Variables | Q1 | Q2 | Q3 | Q4 | Total |
| Range | < 0.05 | 0.05 – 0.07 | 0.08 – 0.11 | > 0.11 | |
| N | 1771 | 1965 | 1933 | 1851 | 7520 |
| Age | 68 ± 9 | 68 ± 9 | 68 ± 9 | 68 ± 10 | 68 ± 9 |
| Female | 555 (31) | 669 (34) | 693 (36) | 697 (38) | 2614 (35) |
| Race | |||||
| white | 1072 (61) | 1115 (67) | 1140 (59) | 1066 (58) | 4393 (58) |
| black | 466 (26) | 581 (30) | 562 (29) | 596 (32) | 2205 (29) |
| hispanic | 198 (11) | 235 (12) | 193 (10) | 167 (9) | 793 (11) |
| other | 35 (2) | 34 (2) | 38 (2) | 22 (1) | 129 (2) |
| Randomization Arm | |||||
| Standard | 966 (55) | 1010 (51) | 981 (51) | 790 (43) | 3747 (50) |
| Intensive | 805 (46) | 955 (49) | 952 (49) | 1061 (57) | 3773 (50) |
| BMI | 29.7 ± 5.4 | 30.0 ± 5.6 | 29.9 ± 5.8 | 30.0 ± 5.9 | 29.9 ± 5.7 |
| Smoking | |||||
| never | 824 (47) | 864 (44) | 857 (44) | 808 (44) | 3353 (45) |
| former | 772 (44) | 885 (45) | 825 (43) | 749 (41) | 3231 (43) |
| current | 172 (10) | 216 (11) | 247 (13) | 293 (16) | 928 (12) |
| prevalent CVD | 339 (19) | 348 (18) | 385 (20) | 402 (22) | 1474 (20) |
| prevalent HF | 50 (3) | 47 (2) | 58 (3) | 88 (5) | 243 (3) |
| # anti-HTN medications | |||||
| 0 | 187 (11) | 203 (10) | 177 (9) | 163 (9) | 730 (10) |
| 1 | 568 (32) | 579 (30) | 576 (30) | 490 (27) | 2213 (29) |
| 2 | 619 (35) | 706 (36) | 652 (34) | 662 (36) | 2639 (35) |
| 3 | 329 (19) | 384 (20) | 412 (21) | 401 (22) | 1526 (20) |
| ≥ 4 | 68 (4) | 93 (5) | 116 (6) | 135 (7) | 412 (6) |
| Diuretic use (at 6m) | 970 (55) | 1162 (59) | 1160 (60) | 1181 (64) | 4473 (60) |
| ACEI/ARB use (at 6m) | 1182 (67) | 1331 (68) | 1333 (69) | 1328 (72) | 5174 (69) |
| SBP | 139 ± 15 | 139 ± 15 | 140 ± 16 | 141 ± 16 | 139 ± 15 |
| DBP | 78 ± 11 | 78 ± 12 | 78 ± 12 | 78 ± 12 | 78 ± 12 |
| eGFR (at 6m) | 74 ± 19 | 74 ± 20 | 73 ± 21 | 70 ± 21 | 73 ± 21 |
| ACR (at baseline) | 8.8 [5.5,18.9] | 8.6 [5.3, 17.5] | 9.0 [5.7, 20.3] | 11.1 [6.1,27.0] | 9.3 [5.6,20.2] |
| Total cholesterol | 188 ± 39 | 190 ± 41 | 190 ± 40 | 191 ± 43 | 190 ± 41 |
| LDL cholesterol | 112 ± 34 | 113 ± 35 | 112 ± 34 | 112 ± 36 | 112 ± 35 |
| HDL cholesterol | 52 ± 13 | 53 ± 15 | 53 ± 14 | 53 ± 15 | 53 ± 14 |
Data presented as median [interquartile range], mean ± standard deviation, or numbers (percent).
Abbreviations: BMI, body mass index; CVD, cardiovascular disease; HF, heart failure; HTN, hypertension; ACEI, angiotensin-converting-enzyme inhibitor; ARB, angiotensin II receptor blocker; eGFR, estimated glomerular filtration rate; LDL, low-density lipoprotein, HDL, high-density lipoprotein; BP, blood pressure; SBP, systolic blood pressure; DBP, diastolic blood pressure; ACR, albumin-to-creatinine ratio.
Using the beta coefficients, characteristics independently associated with greater eGFR variability were African-American race, current smoking, prevalent heart failure, higher systolic BP, lower diastolic BP, use of ACE inhibitors, ARB, or diuretics, higher ACR, and lower eGFR at the month 6 visit (Table 2).
Table 2.
Association of clinical variables with eGFR variability
| Variables | Multivariate | |
|---|---|---|
| β (95% CI) | p-value | |
| Age (per SD =9) | −0.023 (−0.054, 0.009) | 0.16 |
| Female | 0.051 (−0.002, 0.104) | 0.06 |
| Race | ||
| white | 0 (ref) | |
| black | 0.068 (0.012, 0.124) | 0.02 |
| hispanic | −0.054 (−0.134, 0.025) | 0.18 |
| other | −0.089 (−0.268, 0.091) | 0.33 |
| BMI (per SD=5.7) | 0.007 (−0.017, 0.032) | 0.56 |
| Smoking | ||
| never | 0 (ref) | |
| former | −0.005 (−0.055, 0.046) | 0.86 |
| current | 0.220 (0.142, 0.298) | <0.001 |
| Prevalent CVD | 0.045 (−0.015, 0.105) | 0.14 |
| Prevalent HF | 0.226 (0.096, 0.356) | 0.001 |
| Anti-HTN medications | ||
| Diuretics | 0.052 (0.004, 0.10) | 0.04 |
| ACEI/ARB | 0.093 (0.042, 0.143) | <0.001 |
| SBP (per SD=15) | 0.043 (0.013, 0.072) | 0.005 |
| DBP (per SD=12) | −0.040 (−0.072, −0.007) | 0.02 |
| eGFR (at 6m; per SD= 21 decrease) | 0.098 (0.072, 0.123) | <0.001 |
| ACR (at baseline; per doubling) | 0.048 (0.033, 0.063) | <0.001 |
| Total cholesterol (per SD=41) | 0.024 (−0.001, 0.050) | 0.06 |
Abbreviations: SD, standard deviation; CVD, cardiovascular disease; HF, heart failure; HTN, hypertension; ACEI, angiotensin-converting-enzyme inhibitor; ARB, angiotensin II receptor blocker; SBP, systolic blood pressure; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; ACR, albumin-to-creatinine ratio.
There were 370 composite CVD events and 154 deaths during a median follow-up of 2.4 years after the 18-month visit. In a model adjusted for age, sex, race and randomization arm, greater eGFR variability (per SD greater = 0.06) was associated with higher risk of CVD events (hazard ratio (HR), 1.12; 95% confidence interval (CI), 1.03 to 1.23). However, in subsequent models that additionally included additional CVD risk factors, and month-18 eGFR and baseline ACR, this association was attenuated and rendered no longer statistically significant (HR, 1.05; 95% CI, 0.95 to 1.16) (Table 3). When categorized into quartiles, there was a relatively monotonic associations of increasing eGFR variability quartiles with CVD events. Relative to the lowest quartile (coefficient of variation < 0.05), the highest quartile (coefficient of variation > 0.11) had a non-significant association with CVD events after adjustment (HR 1.34; 95% CI 0.98 to 1.84) (Table 3).
Table 3.
Association of eGFR variability with CVD events and all-cause mortality
| Cardiovascular Events | |||||
|---|---|---|---|---|---|
| CONTINUOUS | DISTRIBUTION OF EGFR VARIABILITY | ||||
| per SD greater =0.06 | < 0.05 | 0.05 – 0.07 | > 0.07 – 0.11 | > 0.11 | |
| HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | |
| N | 7520 | 1771 | 1965 | 1933 | 1851 |
| CVD EVENTS | 370 | 71 | 88 | 100 | 111 |
| RATE (%/YR) | 2.06 | 1.69 | 1.88 | 2.15 | 2.52 |
| UNADJUSTED | 1.12 (1.03, 1.23) | 1.00 (ref) | 1.15 (0.84, 1.58) | 1.29 (0.95, 1.76) | 1.53 (1.13, 2.07) |
| MODEL 1 | 1.15 (1.05, 1.26) | 1.00 (ref) | 1.18 (0.85, 1.62) | 1.33 (0.97, 1.81) | 1.62 (1.20, 2.21) |
| MODEL 2 | 1.05 (0.95, 1.16) | 1.00 (ref) | 1.17 (0.85, 1.62) | 1.19 (0.87, 1.63) | 1.34 (0.98, 1.84) |
| All-Cause Mortality | |||||
| CONTINUOUS | DISTRIBUTION OF EGFR VARIABILITY | ||||
| per SD greater =0.06 | < 0.05 | 0.05 – 0.07 | > 0.07 – 0.11 | > 0.11 | |
| HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | |
| N | 7520 | 1771 | 1965 | 1933 | 1851 |
| MORTALITY | 154 | 30 | 32 | 34 | 58 |
| RATE (%/YR) | 1.09 | 0.92 | 0.88 | 0.93 | 1.66 |
| UNADJUSTED | 1.30 (1.18, 1.44) | 1.00 (ref) | 0.89 (0.53, 1.49) | 0.91 (0.55, 1.52) | 1.77 (1.13, 2.77) |
| MODEL 1 | 1.35 (1.21, 1.50) | 1.00 (ref) | 0.92 (0.55, 1.54) | 0.95 (0.57, 1.58 | 1.90 (1.21, 2.99) |
| MODEL 2 | 1.29 (1.14, 1.45) | 1.00 (ref) | 0.89 (0.53, 1.49) | 0.87 (0.52, 1.45) | 1.57 (0.99, 2.49) |
Model 1: adjusted for age, sex, race, randomized arm.
Model 2: Model 1 + baseline history of CVD, history of heart failure, current smoking status, BMI, LDL, total cholesterol, ACR (baseline), eGFR (18m), systolic blood pressure (18m), diastolic blood pressure (18m), ACE/ARB (18m), diuretics (18m), calcium channel blocker (18m) and fasting status.
Abbreviations: SD, standard deviation; CVD, cardiovascular disease; CI, confidence interval; BMI, body mass index; LDL, low-density lipoprotein; ACEI, angiotensin-converting-enzyme inhibitor; ARB, angiotensin II receptor blocker; eGFR, estimated glomerular filtration rate; ACR, albumin-to-creatinine ratio.
In contrast to the CVD endpoint, we found that eGFR variability was associated with all-cause mortality across the series of models, such that each SD greater variability was associated with an approximately 29% higher risk of mortality in the final model (HR 1.29; 95% CI 1.14 to 1.45) (Table 3). In Figure 2, adjusted spline functions depict the associations of eGFR variability with both CVD and mortality outcomes in the fully adjusted model. The association with mortality appeared limited to those in approximately the highest quartile of eGFR variability, whereas the nature of the relationship with CVD appeared more linear. The association of eGFR variability was similar across quartiles 1 to 3, and the higher risk for mortality appeared limited to individuals in quartile 4 (coefficient of variation > 0.11), who had an approximately 57% higher risk of mortality relative to the lowest quartile (HR 1.57; 95% CI 0.99 to 2.49) (Table 3). Findings for the CVD endpoint and all-cause mortality were similar if month 6 variables were used in the adjusted models (Table S1). Results were mostly similar when we re-analyzed the models following multiple imputation for missing covariates (Table S2), although the association between eGFR variability per 1 SD with all-cause mortality had a 95% confidence interval that included 1.0. Finally, findings for both endpoints were similar in sensitivity analyses where we explored whether systolic BP changes or use of ACEI/ARB, CCB and diuretics between months 6 to 18 were mediators of the observed associations (Table S3).
Figure 2.
Adjusted spline plots for risk of CVD events and all-cause mortality over the range of eGFR variability. Curves represent adjusted hazard ratio (solid line) and the 95% CI (dashed lines) based on restricted cubic splines with knots at quartiles of eGFR variability. Model adjusted for age, sex, randomization arm, baseline history of CVD, history of heart failure, current smoking status, BMI, LDL, total cholesterol, systolic blood pressure (6m), ACR (6m), eGFR (6m), ACE/ARB (6m), diuretics (6m), and fasting status. CVD, cardiovascular disease; CI, confidence interval; BMI, body mass index; LDL, low-density lipoprotein; ACEI, angiotensin-converting-enzyme inhibitor; ARB, angiotensin II receptor blocker; eGFR, estimated glomerular filtration rate; ACR, albumin-to-creatinine ratio.
When evaluating the individual components of the composite CVD outcome, associations of eGFR variability appeared to be strongest for heart failure and stroke, but the number of events within each subcategory was small and the associations were not statistically significantly associated with any individual CVD endpoint (Table 4). As eGFR variability was associated with all-cause mortality, but not significantly with the composite CVD endpoint or CVD death, we evaluated other causes of death that were available in SPRINT. We found that eGFR variability was strongly associated with cancer related (HR 1.35; 95% CI 1.13 to 1.60) and for “non-cancer, non-vascular” deaths (HR 1.35; 95% CI 1.13 to 1.63), and that these associations appeared stronger than the association with mortality related to CVD (HR 1.10; 95% CI 0.83 to 1.47) (Table 4).
Table 4.
Association of eGFR variability with individual components of the composite CVD outcome
| Continuous Per SD increase = 0.06 HR (95% CI) |
||||||||
|---|---|---|---|---|---|---|---|---|
| Heart Failure | CVD death | MI | MI (silent) | Stroke | ACS | Cancer related deaths | Other deaths (non-cardiac, non-stroke, accident, unclassifiable) | |
| # EVENTS | 89 | 43 | 98 | 26 | 84 | 30 | 57 | 54 |
| Model* | 1.03 (0.85, 1.25) | 1.10 (0.83, 1.47) | 0.96 (0.77, 1.19) | 1.29 (0.97, 1.70) | 1.10 (0.90, 1.35) | 0.86 (0.56, 1.32) | 1.35 (1.13, 1.60) | 1.35 (1.13, 1.63) |
adjusted for age, sex, race, randomized arm, baseline history of CVD, history of heart failure, current smoking status, BMI, LDL, total cholesterol, ACR (baseline), eGFR (18 m), systolic blood pressure (18m), diastolic blood pressure (18m), ACEI/ARB (18m), diuretics (18m), calcium channel blocker (18m) and fasting status.
Abbreviations: CVD, cardiovascular disease; BMI, body mass index; LDL, low density lipoprotein; ACR, albumin creatinine ratio; ACEI, angiotensin converting enzyme inhibitor; ARB, angiotensin II receptor blocker; MI, myocardial infarction; ACS, acute coronary syndrome.
No statistically significant interactions were observed for the systolic BP intervention or CKD status on the relationship of eGFR variability with either the composite CVD endpoint or all-cause mortality (Table S4 and Table S5). Finally, we considered whether eGFR variability may be driven by those who are losing eGFR quickly. In sensitivity analyses, we found that the associations of eGFR variability with each endpoint remained similar, irrespective of whether the slope of eGFR change was one of decreasing or increasing eGFR (Table S6).
DISCUSSION
Among a large sample of hypertensive trial participants who were at high risk for CVD, eGFR variability was independently associated with risk of mortality. This finding was independent of eGFR, albuminuria, and other risk factors, and was not explained by concurrent changes in BP or BP medications. We found that greater eGFR variability was not associated with CVD events when adjusted for eGFR, albuminuria, and CVD risk factors. Interestingly, we observed that eGFR variability was associated with cancer related and “non-cancer, non-vascular” mortality, and yet not associated with death due to CVD causes. Collectively, these findings demonstrate that variability in eGFR measurements observed within individuals over time holds important physiological and prognostic information, above and beyond the severity of CKD and presence of related risk factors.
Greater eGFR variability is observed commonly in clinical practice. In some individuals, eGFR variability may be explained by changes in renal plasma flow due to initiation or titration of ACEIs, ARBs, or diuretics, intensification of antihypertensive therapy, heart failure, or volume depletion.17,18 Whereas in others, no clear culprit is identifiable. In such cases, eGFR variability may reflect a reduced renal reserve/nephron mass or athero/arteriosclerotic disease of renal arteries.4 We hypothesize, therefore, that eGFR variability would associate with CVD events. By contrast, we found that eGFR variability was strongly associated with cancer related, and “non-cancer, non-vascular deaths”, but not CVD. Reasons for this discrepancy warrant further investigation. It is possible that the relationship of eGFR variability with both cancer, and “non-cancer, non-vascular” mortality may also be related to changes in hemodynamic status. It could potentially also be related to changes in muscle mass, however rapid changes in muscle mass (particularly gain of muscle) seems unlikely over 6 months among individuals at high risk for cancer or other causes of death. The potential explanation may be that eGFR variability reflects impairment of homeostatic/homeodynamic control in renal blood flow, thereby reflecting a “vascular aging” phenotype, in and above chronological age, and therefore with multiple disease processes and causes of death. Whatever the mechanism, we believe it is more likely that eGFR is a marker reflecting this process, and less likely due to creatinine production. We also believe eGFR variability is more likely serving as a marker of this physiology, rather than directly causal with these endpoints. Nonetheless, as eGFR variability is easily measured and a frequent observation in clinical practice, its recognition to identify individuals with unique phenotypes linked with higher risk of death warrants clinical recognition and additional investigation.
To our knowledge, only a few prior studies have evaluated the relationship of eGFR variability with clinical endpoints.5–7 One evaluated a cohort of 3361 participants with stage 3 CKD or greater, almost all of whom were Caucasian. The investigators found that the highest quartile of eGFR variability was associated with mortality (HR 1.40; 95% CI 1.05 to 1.87) similar to our findings.7 We extend these findings by confirming that eGFR variability has similar associations among persons without prevalent CKD. Moreover, the aforementioned prior study lacked medication data after baseline. Thus, the present study demonstrates that concurrent changes in systolic BP changes, and use of ACEI/ARBs and diuretic medications during follow-up did not meaningfully attenuate or explain the association of eGFR variability with mortality.
We found no significant association of eGFR variability with CVD events in fully adjusted models. To our knowledge, only two studies have evaluated the association of eGFR variability and CVD events previously.9,19 Of the two studies, only Suzuki et al. reported that the highest tertile of eGFR variability was significantly associated with CVD events (HR 1.90; 95% CI 1.03 to 3.71) among 2869 participants.9 Our discrepant findings may be explained by differences in the study populations, as all participants in the study by Suzuki et al. had CKD, 33% had diabetes, and all were Asian. In addition, selection of adjustment covariates differed. We adjusted for ACR and prevalent heart failure, which were not included in the Suzuki study. It remains possible that an association exists in truth, but was missed by chance in this study, despite being well powered for CVD events.
There are some inconsistencies in the literature with regards to the associations of eGFR variability with kidney disease progression and dialysis initiation. One prior study reported that greater eGFR variability was independently associated with CKD progression,6 but this was not confirmed in another study.8 We did not evaluate this outcome, in part because we only had 37 CKD progression events after month 18 in our study, and also because eGFR would have been used as both the predictor variable and the outcome in an analysis of CKD progression.
Our study has several strengths including the well characterized, large and diverse study population, and the novelty of assessing eGFR variability for clinical endpoints in persons without CKD. We evaluated clinical trial participants, which provided the unique advantage that the protocol dictated measurements of eGFR at pre-specified time points during the trial, allowing consistent assessment of eGFR variability in all participants. The CVD endpoint was centrally adjudicated according to pre-specified rules. The clinical trial setting and available data allowed us to adjust for anti-hypertensive medications, especially use of ACEI/ARB and diuretics as well as for concurrent changes in BP, during follow-up. All of these factors were demonstrated to be independent determinants of eGFR variability and were plausible confounders, although in the end, associations remained similar even after accounting for these factors.
This study also has important limitations. SPRINT excluded persons with diabetes mellitus, stroke, or proteinuria >1 g/d, and participants were volunteers in a clinical trial, which may limit the generalizability. We calculated the coefficient of variation of eGFR using only three eGFR data points that were each 6 months-apart, so the prognostic value of shorter or longer-term eGFR variability remains uncertain. We had limited power to detect moderate associations between the exposure and some of our outcomes (as demonstrated by the wide confidence intervals). Sensitivity analyses accounting for missing data attenuated the association with all-cause mortality, suggesting this may have led to a bias.
In summary, in a large cohort of hypertensive trial participants at higher risk of CVD, greater eGFR variability was associated with mortality independent of baseline eGFR, albuminuria, and other risk factors, an association that does not appear to be explained by concurrent changes in BP or use of anti-hypertensive medications. Associations appeared stronger for cancer mortality, and “non-cancer, non-vascular” mortality, than mortality related to CVD events. Future studies should test these hypotheses and evaluate mechanisms underlying these associations. Clinicians should be aware that greater eGFR variability may identify people at higher mortality risk, irrespective of the severity of their kidney function at baseline.
Supplementary Material
Table S1. Association of eGFR variability with CVD events and all-cause mortality (adjustment with month 6 variables).
Table S2. Association of eGFR variability with CVD events and all-cause mortality (imputation of missing covariates).
Table S3. Association of eGFR variability with CVD events and all-cause mortality, evaluating the mediation effects of SBP, and use of ACEI/ARB and diuretics at the month 6, 12, and 18 visits.
Table S4. Association of eGFR variability with CVD events stratified by randomization treatment arm and baseline CKD status.
Table S5. Association of eGFR variability with mortality stratified by randomization treatment arm and eGFR category.
Table S6. Association of eGFR slope# with CVD events and all-cause mortality.
Acknowledgements:
For a full list of contributors to SPRINT, please see https://www.sprinttrial.org/public/dspScience.cfm. All components of the SPRINT study protocol were designed and implemented by the investigators. The investigative team collected, analyzed, and interpreted the data.
Financial Disclosure:
Dr. Shlipak is a Scientific Advisor and stock in TAI Diagnostics and has received compensation from Cricket Health, Inc. Dr. Ix is principal investigator of an investigator initiated research study sponsored by Baxter International. Dr. Cushman received and institutional grant from Eli Lilly. The remaining authors declare that they have no relevant financial interests.
Support:
This work was supported by the National Institutes of Diabetes and Digestive and Kidney Diseases (NIDDK) (R01DK098234-04 and K24DK110427 for JHI), the American Heart Association (14EIA18560026 for JHI), the Satellite Coplon award for RM and the Academic Community UCSD grant for RM. The Systolic Blood Pressure Intervention Trial (SPRINT) is funded with federal funds from the NIH, including the National Heart, Lung, and Blood Institute (NHLBI), the NIDDK, the National Institute on Aging (NIA), and the National Institute of Neurological Disorders and Stroke (NINDS), under Contract Numbers HHSN268200900040C, HHSN268200900046C, HHSN268200900047C, HHSN268200900048C, HHSN268200900049C, and Inter-Agency Agreement Number A-HL-13-002-001. It was also supported in part with resources and use of facilities through the Department of Veterans Affairs. The SPRINT investigators acknowledge the contribution of study medications (azilsartan and azilsartan combined with chlorthalidone) from Takeda Pharmaceuticals International, Inc. We also acknowledge the support from the following CTSAs funded by NCATS: CWRU: UL1TR000439, OSU: UL1RR025755, U Penn: UL1RR024134& UL1TR000003, Boston: UL1RR025771, Stanford: UL1TR000093, Tufts: UL1RR025752, UL1TR000073 & UL1TR001064, University of Illinois: UL1TR000050, University of Pittsburgh: UL1TR000005, UT Southwestern: 9U54TR000017-06, University of Utah: UL1TR000105-05, Vanderbilt University: UL1 TR000445, George Washington University: UL1TR000075, University of CA, Davis: UL1 TR000002, University of Florida: UL1 TR000064, University of Michigan: UL1TR000433, Tulane University: P30GM103337 COBRE Award NIGMS, Wake Forest University: UL1TR001420. The funders had no role in study design, data collection, analysis, reporting, or the decision to submit for publication.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Publisher's Disclaimer: Disclaimer: All aspects of manuscript writing and revision were carried out by the coauthors. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH, the U.S. Department of Veterans Affairs, or the United States Government.
Peer Review: Received April 27, 2020. Evaluated by 2 external peer reviewers and a statistician, with direct editorial input from an International Editor, who served as Acting Editor-in-Chief. Accepted in revised form October 16, 2020. The involvement of an Acting Editor-in-Chief was to comply with AJKD’s procedures for potential conflicts of interest for editors, described in the Information for Authors & Journal Policies.
REFERENCES:
- 1.Weiner DE, Tighiouart H, Amin MG, Stark PC, MacLeod B, Salem DN, Levey AS, Sarnak MJ. Chronic kidney disease as a risk factor for cardiovascular disease and all-cause mortality: a pooled analysis of community-based studies. J Am Soc Nephrol. 2004; 15(5): 1307–1315. [DOI] [PubMed] [Google Scholar]
- 2.Go AS, Chertow GM, Fan D, McCulloch CE, Hsu CY. Chronic kidney disease and the risks of death, cardiovascular events, and hospitalization. N Engl J Med. 2004; 351(13): 1296–1305. [DOI] [PubMed] [Google Scholar]
- 3.Ford I, Bezlyak V, Stott DJ, Sattar N, Packard CJ, Buckley BM, Jukema JW, de Craen AJ, Westendorp RG, Shepherd J. Reduced glomerular filtration rate and its association with clinical outcome in older patients at risk of vascular events: secondary analysis. PLoS Med. 2009; 6(1): 76–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Denic A, Glassock RJ, Rule AD. Structural and functional changes with the aging kidney. Adv Chronic Kidney Dis. 2016; 23(1): 19–28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Ljungman S, Laragh JH, Cody RJ. Role of the kidney in congestive heart failure. Relationship of cardiac index to kidney function. Drugs. 1990; 39(S4):10–21. [DOI] [PubMed] [Google Scholar]
- 6.Al-Aly Z, Balasubramanian S, McDonald JR, Scherrer JF, O’Hare AM. Greater variability in kidney function is associated with an increased risk of death. Kidney Int. 2012; 82(11): 1208–1214. [DOI] [PubMed] [Google Scholar]
- 7.Perkins RM, Tang X, Bengier AC, Kirchner HL, Bucaloiu ID. Variability in estimated glomerular filtration rate is an independent risk factor for death among patients with stage 3 chronic kidney disease. Kidney Int. 2012; 82(12): 1332–1338. [DOI] [PubMed] [Google Scholar]
- 8.Tseng CL, Lafrance JP, Lu SE, Soroka O, Miller DR, Maney M, Pogach LM. Variability in estimated glomerular filtration rate values is a risk factor in chronic kidney disease progression among patients with diabetes. BMC Nephrol. 2015; 16: 34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Suzuki A, Obi Y, Hayashi T, Kotani N, Uemura Y, Imai E, Makino H, Hishida A. Visit-to-visit variability in estimated glomerular filtration rate predicts hospitalization and death due to cardiovascular events. Clin Exp Nephrol. 2019; 23(5): 661–668. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Turin TC, Coresh J, Tonelli M, Stevens PE, de Jong PE, T Farmer CK, Matsushita K, Hemmelgarn BR. Change in the estimated glomerular filtration rate over time and risk of all-cause mortality. Kidney Int. 2013; 83(4): 684–691. [DOI] [PubMed] [Google Scholar]
- 11.Picth NA, Rohan V, Rao V, Mauldin PD, Su Z, Dubay DA, Moinelli TA, Taber DJ. Renal Function Variability: An Independent Risk Factor for Graft Loss and Death following Kidney Transplantation. Am J Nephrol. 2018; 47(3): 191–199. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Ambrosius WT, Sink KM, Foy CG, Berlowitz DR, Cheung AK, Cushman WC, Fine LJ, Goff DC Jr., Johnson KC, Killeen AA, Lewis CE, Oparil S, Reboussin DM, Rocco MV, Snyder JK, Williamson JD, Wright JT Jr., Whelton PK; SPRINT Study Research Group: The design and rationale of a multicenter clinical trial comparing two strategies for control of systolic blood pressure: The Systolic Blood Pressure Intervention Trial (SPRINT). Clin Trials. 2014; 11(5): 532–546. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Wright JT Jr., Williamson JD, Whelton PK, Snyder JK, Sink KM, Rocco MV, Reboussin DM, Rahman M, Oparil S, Lewis CE, Kimmel PL, Johnson KC, Goff DC Jr., Fine LJ, Cutler JA, Cushman WC, Cheung AK, Ambrosius WT; SPRINT Research Group: A randomized trial of intensive versus standard blood-pressure control. N Engl J Med. 2015; 373(22): 2103–2116. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Cheung AK, Rahman M, Reboussin DM, Craven TE, Greene T, Kimmel PL, Cushman WC, Hawfield AT, Johnson KC, Lewis CE, Oparil S, Rocco MV, Sink KM, Whelton PK, Wright JT Jr., Basile J, Beddhu S, Bhatt U, Chang TI, Chertow GM, Chonchol M, Freedman BI, Haley W, Ix JH, Katz LA, Killeen AA, Papademetriou V, Ricardo AC, Servilla K, Wall B, Wolfgram D, Yee J; SPRINT Research Group: Effects of intensive BP control in CKD. J Am Soc Nephrol. 2017; 28(9): 2812–2823. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Malhotra R, Craven T, Ambrosius WT, Killeen AA, Haley WE, Cheung AK, Chonchol M, Sarnak M, Parikh CR, Shlipak MG, Ix JH; SPRINT Research Group. Effects of Intensive Blood Pressure Lowering on Kidney Tubule Injury in CKD: A Longitudinal Subgroup Analysis in SPRINT. Am J Kidney Dis. 2019; 73(1): 21–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Rubin DB. Multiple imputation for nonresponse in surveys. Wiley, New York: 1987. [Google Scholar]
- 17.Ichikawa I, Pfeffer JM, Pfeffer MA, Hostetter TH, Brenner BM. Role of angiotensin II in the altered renal function of congestive heart failure. Circulation research. 1984; 55(5): 669–675. [DOI] [PubMed] [Google Scholar]
- 18.Ruggenenti P, Remuzzi G. Worsening kidney function in decompensated heart failure: treat the heart, don’t mind the kidney. European heart journal. 2011; 32(20): 2476–2478. [DOI] [PubMed] [Google Scholar]
- 19.Dimitriadis K, Tsioufis C, Kasiakogias A, Konstantinidis D, Michas C, Iliakis P, Fragoulis C, Kouremeti M, Syrmali K, Kintis K, Voutsinas M, Chatzis ST, Kallikazaros I, Tousoulis D. The predictive role of visit-to-visit glomerular filtration rate variability for cardiovascular and renal outcomes in essential hypertension: data from a Greek 8-year-follow-up study. European Heart Journal. 2018; 39(1): P1278. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1. Association of eGFR variability with CVD events and all-cause mortality (adjustment with month 6 variables).
Table S2. Association of eGFR variability with CVD events and all-cause mortality (imputation of missing covariates).
Table S3. Association of eGFR variability with CVD events and all-cause mortality, evaluating the mediation effects of SBP, and use of ACEI/ARB and diuretics at the month 6, 12, and 18 visits.
Table S4. Association of eGFR variability with CVD events stratified by randomization treatment arm and baseline CKD status.
Table S5. Association of eGFR variability with mortality stratified by randomization treatment arm and eGFR category.
Table S6. Association of eGFR slope# with CVD events and all-cause mortality.