Abstract
Background:
Greater variability in estimated glomerular filtration rate (eGFR) is associated with higher mortality in patients with chronic kidney disease (CKD). Heart failure (HF) is common in CKD and may increase variability through changes in hemodynamic and volume regulation. We sought to determine if patients with versus without HF have higher kidney function variability in CKD, and to define the association with mortality.
Methods:
Patients undergoing coronary angiography from 2003–2013 with eGFR<60 mL/min/1.73m2 were evaluated from the Duke Databank for Cardiovascular Disease. eGFR variability, measured as the coefficient of variation (CV) of residuals from the regression of eGFR versus time, was calculated spanning 3 months to 2 years after catheterization. Mortality was assessed 2 to 7 years after catheterization. Patients were grouped into 3 HF phenotypes: HF with reduced ejection fraction (HFrEF), HF with preserved ejection (HFpEF), and no HF. Regression was used to evaluate associations between HF phenotypes and eGFR variability and between eGFR variability and mortality rate with stratification by HF phenotype.
Results:
Among 3767 participants, median (interquartile range) eGFR at baseline was 45 (33–53) mL/min/1.73m2, and longitudinal measures of eGFR over 21 months had within-patient residual variability (CV) of 14% (9–20%). In adjusted analyses, eGFR variability was greater in those with HFpEF (n=695, CV difference: 0.98%, 95% confidence interval [CI] 0.14% – 1.81%) or HFrEF (n=800, CV difference: 2.51%, 95% CI 1.66% – 3.37%) relative to no HF (n=2272). In 3068 participants eligible for mortality analysis, the presence of HF and greater eGFR variability were each independently associated with higher mortality, but there was no evidence of interaction between eGFR variability and any HF phenotype (all p for interaction ≥0.49).
Conclusions:
eGFR variability is greater in patients with HF and associated with mortality. Prediction algorithms and classification schemes should consider not only static but also dynamic eGFR variability in HF and CKD prognostication.
Keywords: Heart failure, chronic kidney disease, eGFR variability, outcomes
Introduction
Kidney dysfunction is common among patients with heart failure (HF). In a recent HF registry analysis, over 50% of participants had an estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m2, consistent with at least moderate chronic kidney disease (CKD).[1] Patients with co-morbid HF and CKD experience poor outcomes relative to either condition alone. For instance, patients with both HF and CKD may have up to a 50% greater risk of mortality relative to HF patients without CKD, a relationship particularly evident among patients with HF with reduced ejection fraction (HFrEF).[2] Likewise, HF patients with CKD are more likely to have increased morbidity[3, 4] and report more advanced HF symptoms compared with those without CKD.[5]
While prior studies have shown that poor baseline kidney function and worsening kidney function are associated with adverse outcomes in patients with HF, the role of kidney function variability over time is not well established. Clinical experience suggests changes in kidney function are common during longitudinal care of HF patients and potentially related to changes in volume status, hemodynamics, medications, or intrinsic renal disease progression. Previous work in patients with CKD has found greater variability in kidney function, as measured by eGFR, to be associated with greater mortality and adverse kidney outcomes, independent of the absolute value or trend.[6–8] Understanding the complex relationships between kidney function and outcomes in HF, including baseline, worsening, and variability in kidney function, can inform development of predictive algorithms aimed at classification and prediction in CKD and HF.
In this context, the goal of the current study was to determine if patients with HFrEF or HF with preserved ejection fraction (HFpEF) demonstrate greater kidney function variability relative to non-HF populations, and to determine the relationship between kidney function variability and clinical outcomes. In those with normal kidney function, a stable GFR is generally maintained despite fluctuation in volume and hemodynamics through autoregulation. eGFR variability may reflect impairment in autoregulation that portends worse kidney function than evident from static eGFR alone, particularly in those with impaired kidney function.[9] Thus, our hypothesis was that the degree of hemodynamic decompensation and impaired autoregulation present in the context of the various HF and renal function phenotypes would result in the following associations: patients with co-morbid HFrEF and CKD would demonstrate the highest levels of kidney function variability, patients with co-morbid HFpEF and CKD would have intermediate levels of variability, and patients with CKD without HF would have the lowest levels of variability. Similarly, we also hypothesized a varying association between kidney function variability and mortality, with the greatest mortality risk attributable to kidney function variability among patients with HFrEF and CKD, followed by those with HFpEF and CKD, with lesser association among those with CKD without HF.
Methods
Study Population
Patients were identified from the Duke Databank for Cardiovascular Disease (DDCD), a database of all patients who underwent diagnostic coronary angiography at the Duke University Medical Center beginning in 1969, with cessation of study follow-up in 2014. The study was approved by the Institutional Review Board at the Duke University Health System, Durham, NC. No extramural funding was used to support this work. Participants were included in this retrospective cohort analysis if they underwent coronary angiogram from 2003–2013 (to capture a more contemporary cohort with long-term follow-up), had left ventricular ejection fraction (LVEF) data available, and met study criteria for CKD, defined as a baseline eGFR less than 60 mL/min/1.73m2. If a patient had multiple qualifying catheterization procedures during the study period, the first coronary angiogram was selected as the index procedure and included in the analysis.
Data Collection
Using methods described previously for the DDCD,[10, 11] baseline clinical and patient care variables such as demographics, clinical diagnoses and symptoms, and laboratory findings were systematically collected as part of routine patient care by clinicians at admission for or at the time of catheterization. Follow-up at 6 months, 12 months, and annually thereafter was obtained through self-administered questionnaires, with telephone follow-up for non-responders for a subset of DDCD participants with significant coronary artery disease. Baseline medication usage was collected for descriptive purposes and was recorded if a medication from within the designated classes was used within a window three months prior to and three months after catheterization in either the DDCD baseline dataset or the electronic medical record.
To define HF phenotypes, LVEF measurements were obtained primarily from echocardiography by linking the DDCD with the Duke Echocardiography Laboratory Database, and was augmented by visual EF from left ventriculogram and other modalities (e.g., nuclear) when echocardiographic data were not available.[12] LVEF measurements were collected within a window extending three months prior to and one month after catheterization; LVEF values were excluded if there was an intervening myocardial infarction between catheterization and LVEF evaluation. LVEF values collected via echocardiography prior to catheterization were prioritized over post-catheterization collection and LVEF measurement from other modalities to limit the influence of catheterization indication and outcome on measured LVEF. Given that LVEF may vary among different imaging modalities, we performed sensitivity analyses including only participants with LVEF collected by echocardiography. In participants from the DDCD with New York Heart Association class II-IV symptoms assessed prior to catheterization, HFrEF was defined as an LVEF ≤40% and HFpEF was defined as LVEF >40%. Patients with no history of symptoms of HF by questionnaire or NYHA class I symptoms in the two weeks prior to baseline were the control phenotype.
To calculate eGFR in this study, longitudinal measures of serum creatinine evaluated in the outpatient setting were obtained from the electronic medical record. Baseline eGFR was calculated using the CKD-EPI equation from the mean of all outpatient creatinine values collected within a window from six months prior to and three months after catheterization.[13]
Kidney Function Variability
eGFR variability was defined as the variability of outpatient eGFR measurements irrespective of absolute eGFR level or trend over time. To quantify variability, we used the coefficient of variation (CV) of the residuals from the fitted regression of eGFR versus time. This was selected as the primary variability measurement because it quantifies the magnitude of within-subject variability in eGFR while removing any systematic trend in eGFR over time; it also standardizes variation relative to the mean eGFR and is measured as a percentage of mean eGFR (Figure IA). To evaluate the robustness of results, a sensitivity analysis incorporating root-mean-square error as an alternative measurement of eGFR variability was also performed and reported for all applicable analyses. The root-mean-square error is equal to the estimated residual standard deviation from the regression line of eGFR versus time and is measured in mL/min/1.73 m2 (Figure IA).
Figure I.
Study Timeline and eGFR Variability Metrics. A. eGFR variability calculation for a hypothetical participant. Coefficient of variation was the primary variability metric utilized, and root-mean-square error was used for sensitivity analyses. B. Timeline of collection for eGFR measurements and outcomes data relative to the time of coronary angiography, noted as “index cath.” Abbreviations: Estimated glomerular filtration rate (eGFR)
eGFR variability was determined primarily for outpatient serum creatinine values obtained during a 21-month exposure period from 3–24 months after catheterization (Figure IB). This exposure window time period was pre-specified to balance obtaining a robust number of laboratory measurements through routine outpatient heart failure care while minimizing exclusion from subsequent outcomes analysis due to interval mortality. A 3-month blanking period after catheterization was pre-specified to remove any impact of iodinated contrast exposure on eGFR variability on individual serum creatinine values; the baseline eGFR, calculated as a mean value as specified above, was included in eGFR variability calculations. eGFR variability was determined for each individual participant with at least three outpatient creatinine values during the exposure period. Further, to ensure temporal spread of included lab values, participants were excluded if they did not have at least two outpatient creatinine values greater than three months apart. If there were multiple outpatient creatinine values from the same date, only one random value was selected for analysis. Serum creatinine values obtained during hospitalization were excluded from analysis to reduce the capture of episodes of acute kidney injury.
Mortality Outcome
The primary outcome was all-cause mortality. Follow-up for mortality was obtained through December 2014 and was based on deaths reported in the health system, from the 2011 Social Security Death Master File (DMF), as well as subsequent weekly DMF updates. Mortality was also evaluated through annual follow-up surveys for the subset of DDCD patients who qualified for follow-up (those with significant coronary artery disease, percutaneous coronary intervention, or coronary artery bypass grafting), with vital status of non-responders verified by querying the National Death Index.[14]
Statistical Methods
Participants were divided into three groups based on tertiles of the distribution of eGFR variability in the overall CKD cohort. Baseline characteristics evaluated at the time of catheterization, and measurements of kidney function before and during the subsequent 24 months, were summarized by eGFR variability tertile. Comparisons across groups were described using Pearson’s Chi-square test for categorical variables and Kruskal-Wallis test for continuous variables.
Multivariable linear regression modeling was employed to evaluate differences in eGFR variability among different HF phenotypes. Models were adjusted for the following pre-specified covariates measured at baseline: baseline eGFR, age, race, sex, history of relevant comorbidities such as diabetes, hypertension, hyperlipidemia, tobacco use, peripheral vascular disease, or cerebrovascular disease, body mass index, coronary artery disease, baseline systolic blood pressure, and year of coronary angiography. Model assumptions were tested by analysis of residuals.
The association between eGFR variability during the exposure window and mortality risk during the subsequent 5 years of follow-up was evaluated using a landmark approach incorporating Kaplan-Meier plots, log rank tests, and negative binomial models, starting at 24 months after catheterization to allow exposure ascertainment. Cumulative incidence curves and 5-year estimates for mortality were calculated for each eGFR variability tertile using the Kaplan-Meier method in the overall cohort. Curves and 5-year estimates were also generated separately for each HF phenotype group for descriptive purposes.
To assess associations between eGFR variability and mortality rate, negative binomial regression models were generated. This approach was selected over Cox regression modeling, because for this landmark analysis, comparison of average mortality rates was felt to be more clinically interpretable than comparison of hazard rates with a time scale starting 24 months after catheterization. Further, when the hazard is constant over time the rate ratios estimated by negative binomial regression yield similar results to hazard ratios from Cox proportional hazards models. The constant hazard rate assumption was first tested and satisfied; analyses were also repeated using Cox proportional hazards models beginning at the origin of 24 months after catheterization, which provided nearly identical effect estimates; thus, only our primary approach is reported. The main variable of interest, eGFR variability, was modeled using the tertile groupings, with mortality rate ratios estimated for medium and high tertiles relative to the lowest tertile. Adjusted models included HF phenotype group, baseline eGFR as well as other pre-selected baseline covariates. The linearity assumption was tested for all continuous variables, and two-piece linear spline transformations were applied to measures of systolic blood pressure and baseline eGFR. We then tested for interaction between HF phenotype and eGFR variability on the outcome of mortality in the adjusted model. A post-hoc sensitivity analysis incorporating a binary covariate indicating the receipt of all four baseline medication classes (aspirin, ace-inhibitor or angiotensin-receptor blocker, beta-blocker, and diuretic) was performed for the mortality analysis.
Data were largely complete for key variables except for baseline systolic blood pressure which was missing in 8.4% (318) of the cohort. We conducted complete case analyses using all available data for univariate descriptive summaries. For adjusted models, inverse probability weighting was used to account for missing values for systolic blood pressure. Statistical significance was based on a p value of ≤0.05, with no adjustment for multiple hypothesis tests. All statistical analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC).
Results
Patient Characteristics
Overall, 3767 participants met study inclusion criteria and were included in the eGFR variability analysis; 699 participants from this group died or were lost to follow up during the 24-month window after catheterization, resulting in 3068 participants eligible for the mortality outcome analysis (Figure II).
Figure II.
Inclusion and Exclusion Criteria for eGFR Variability Analysis and Mortality Analysis. Composition of studied cohort based on inclusion and exclusion criteria. Abbreviations: Estimated glomerular filtration rate (eGFR), left ventricular ejection fraction (LVEF), New York Heart Association (NYHA), chronic kidney disease (CKD)
Baseline characteristics for the overall cohort are summarized by eGFR variability tertile in Table I. Participants with higher eGFR variability were more likely to be female and have diabetes, lower systolic blood pressures, reduced LVEF, higher blood urea nitrogen levels, and have higher use of diuretics. Supplemental Table I presents baseline characteristics by HF phenotype.
Table I.
Demographics and Baseline Characteristics Stratified by eGFR Variability Tertile.
| Low eGFR Variability CV: 0–11% (n=1255) | Medium eGFR Variability CV: 11–18% (n=1256) | High eGFR Variability CV: 18–122% (n=1256) | All Participants (n=3767) | p-value | |
|---|---|---|---|---|---|
|
| |||||
| Demographics | |||||
| Age at baseline (mean, SD) | 65.9 (12.44) | 66.6 (11.19) | 63.9 (12.59) | 65.5 (12.14) | <.001 |
| Race | 0.063 | ||||
| White | 844 (67.7%) | 898 (71.7%) | 842 (67.2%) | 2584 (68.9%) | |
| Black | 377 (30.3%) | 328 (26.2%) | 386 (30.8%) | 1091 (29.1%) | |
| Native American | 6 (0.5%) | 3 (0.2%) | 9 (0.7%) | 18 (0.5%) | |
| Other | 19 (1.5%) | 23 (1.8%) | 16 (1.3%) | 58 (1.5%) | |
| Male | 774 (61.7%) | 721 (57.4%) | 634 (50.5%) | 2129 (56.5%) | <.001 |
| BMI, kg/m2 (median, IQR) | 28 (25, 33) | 29 (26, 33) | 29 (25, 33) | 29 (25, 33) | 0.020 |
| Medical History | |||||
| History of Hypertension | 973 (77.5%) | 903 (71.9%) | 903 (71.9%) | 2779 (73.8%) | 0.001 |
| History of Diabetes | 433 (34.5%) | 468 (37.3%) | 519 (41.3%) | 1420 (37.7%) | 0.002 |
| History of Hyperlipidemia | 724 (57.7%) | 716 (57.0%) | 695 (55.3%) | 2135 (56.7%) | 0.472 |
| History of Smoking | 454 (36.2%) | 499 (39.7%) | 470 (37.4%) | 1423 (37.8%) | 0.176 |
| History of Peripheral Vascular Disease | 121 (9.6%) | 113 (9.0%) | 120 (9.6%) | 354 (9.4%) | 0.835 |
| History of Cerebrovascular Disease | 161 (12.8%) | 150 (11.9%) | 165 (13.1%) | 476 (12.6%) | 0.646 |
| Systolic Blood Pressure, mm Hg (median, IQR) | 146 (130, 163) | 141 (125, 161) | 137 (120, 157) | 141 (124, 161) | <.001 |
| Significant CAD | 670 (62.6%) | 584 (61.9%) | 572 (62.9%) | 1826 (62.5%) | 0.912 |
| Multivessel CAD | 474 (44.3%) | 436 (46.2%) | 413 (45.4%) | 1323 (45.3%) | 0.682 |
| LVEF (median, IQR) | 55 (45,55) | 55 (35,55) | 50 (30,55) | 55 (36,55) | <.001 |
| No HF* | 877 (38.6%) | 741 (32.6%) | 654 (28.8%) | 2272 | --- |
| HFpEF* | 208 (29.9%) | 239 (34.4%) | 248 (35.7%) | 695 | |
| HFrEF* | 170 (21.3%) | 276 (34.5%) | 354 (44.3%) | 800 | |
| Baseline Labs/Medications | |||||
| Serum Blood Urea Nitrogen, mg/dL (median, IQR) | 24 (19, 33) | 24 (19, 31) | 26 (20, 36) | 25 (19, 33) | <.001 |
| Serum Sodium, mEq/L (median, IQR) |
140 (138, 141) | 139 (138, 141) | 139 (137, 141) | 139 (138, 141) | <.001 |
| Renin-Angiotensin System Inhibitors | 939 (74.8%) | 977 (77.8%) | 1007 (80.2%) | 2923 (77.6%) | 0.006 |
| Beta-blockers | 973 (77.5%) | 988 (78.7%) | 1017 (81.0%) | 2978 (79.1%) | 0.097 |
| Aspirin | 990 (78.9%) | 1014 (80.7%) | 1035 (82.4%) | 3039 (80.7%) | 0.082 |
| Diuretics (loop, thiazide, etc.) | 752 (59.9%) | 958 (76.3%) | 979 (77.9%) | 2689 (71.4%) | <.001 |
Unless otherwise specified, cells display counts (%).
Parentheses show row percentage for HF phenotype.
Abbreviations: Estimated glomerular filtration rate (eGFR), coefficient of variation (CV), standard deviation (SD), interquartile range (IQR), coronary artery disease (CAD), left ventricular ejection fraction (LVEF), heart failure (HF), heart failure with preserved ejection fraction (HFpEF), heart failure with reduced ejection fraction (HFrEF)
Kidney Function Variability and HF Status
Summary characteristics for measurements of kidney function during the exposure window stratified by either eGFR variability tertile or HF phenotype are reported in Supplemental Table II. In the overall cohort, the median (interquartile range) eGFR was 45 (33–53) mL/min/1.73m2 at baseline, with a similar distribution for the within-patient mean eGFR during the subsequent 24 months. Median (interquartile range) within-patient eGFR variability measured by CV was 14% (9–20%), and by root-mean-square error was 5.8 (3.6–8.6) mL/min/1.73 m2. The median number of lab measurements for eGFR calculation for each tertile was 6 for the low variability tertile, 8 for the medium variability tertile, and 10 for the high variability tertile. Participants with greater eGFR variability had lower minimum and higher maximum eGFR values during the exposure period than other groups; those with either low or high eGFR variability had a reduced baseline and mean eGFR compared to those with medium eGFR variability. Participants with HFrEF had a higher number of lab measurements.
Table IIA presents unadjusted and adjusted associations between HF phenotype and eGFR variability. In unadjusted models, significant associations between the presence of either HFpEF or HFrEF and higher eGFR variability were observed. For HFpEF, the CV was 1.54% higher (95% confidence interval [CI] 0.75% – 2.32%) compared with those without HF; for HFrEF, the CV was 3.28% higher (95% CI 2.53% – 4.02%) than in those without HF. When adjusted for pre-selected covariates, significant associations between the presence of HF and higher eGFR variability remained, but the magnitude of the effect was diminished. After adjustment, among patients with HFpEF, the CV was 0.98% higher (95% CI 0.14% – 1.81%) compared to those without HF. Likewise, the CV was 2.51% higher (95% CI 1.66% – 3.37%) in those with HFrEF relative to those without HF. In sensitivity analysis using root-mean-square error as the eGFR variability metric shown in Table IIB, the significant association between eGFR variability and HFrEF versus no HF remained, while the association between HFpEF versus no HF and eGFR variability was no longer significant. Furthermore, in a sensitivity analyses investigating associations between HF phenotype and eGFR variability in those with LVEF measured by echocardiography, we found similar results (Supplemental Table IV).
Table II.
Difference in eGFR Variability Across HF Phenotypes.
| Unadjusted | Adjusted* | |||
|---|---|---|---|---|
| Comparison† | Estimate‡ (95% CI) | p-value | Estimate (95% CI) | p-value |
|
| ||||
| a. eGFR Coefficient of variation in % | ||||
| HFpEF | 1.54 (0.75, 2.32) | <.001 | 0.98 (0.14, 1.81) | 0.021 |
| HFrEF | 3.28 (2.53, 4.02) | <.001 | 2.51 (1.66, 3.37) | <.001 |
| b. eGFR Root-mean-square error in ml/min/1.73m2 | ||||
| HFpEF | 0.15 (−0.21, 0.51) | 0.416 | 0.26 (−0.09, 0.62) | 0.142 |
| HFrEF | 1.11 (0.76, 1.45) | <.001 | 0.69 (0.33, 1.05) | <.001 |
Adjustments were made for age, race, sex, history of diabetes, history of hypertension, history of hyperlipidemia, history of smoking, history of peripheral vascular disease, history of cerebrovascular disease, coronary artery disease, body mass index, baseline eGFR, systolic blood pressure, and year of coronary angiography.
Reference group is those without HF for each comparison.
Estimate is the beta-coefficient from the linear regression model and quantifies the mean difference in residual CV or root-mean-square error between the compared groups.
Abbreviations: Estimated glomerular filtration rate (eGFR), heart failure (HF), confidence interval (CI), heart failure with preserved ejection fraction (HFpEF), heart failure with reduced ejection fraction (HFrEF)
All-Cause Mortality
Supplementary Table III shows that the unadjusted cumulative incidence of death was higher in participants with either HF phenotype and in participants with greater eGFR variability in all HF groups. In unadjusted analysis of the overall cohort in Figure III, there was a higher rate of mortality in participants with medium eGFR variability (rate ratio [RR] 1.30, 95% CI 1.11 – 1.53) or high eGFR variability (RR 1.80, 95% CI 1.54 – 2.10) compared to those with low eGFR variability. The mortality associations persisted after adjustment for covariates in the overall cohort in those with medium eGFR variability (RR 1.33, 95% CI 1.12 – 1.58) or high eGFR variability (RR 1.72, 95% CI 1.45 – 2.04) compared to those with low eGFR variability. In the adjusted sensitivity analysis employing root-mean-square error, there was a higher mortality rate in those with medium eGFR variability (RR 1.26 95% CI 1.06 – 1.50) or high eGFR variability (RR 1.52 RR 1.27 – 1.81) relative to those with low eGFR variability.
Figure III.
Unadjusted and Adjusted Cumulative Mortality Rates Stratified by eGFR Variability in the Overall Cohort. A. Cumulative probability of all-cause mortality in participants during the mortality outcome window, stratified by tertile of eGFR variability. B. Unadjusted and adjusted associations between different levels of eGFR variability, measured as coefficient of variation or root-mean-square error, and all-cause mortality are presented. Comparison is to low eGFR variability tertile for each subgroup. Adjustments were made covariates as in Table 2, with the addition of HF phenotype. Abbreviations as in Table 2.
Cumulative mortality curves for participants with either no HF, HFpEF, or HFrEF are displayed in Figure IV; in all groups, higher eGFR variability showed a numerically higher rate of mortality relative to those with low eGFR variability. However, when participants within eGFR variability tertiles were stratified by HF phenotype, there was no evidence of interaction between HF phenotype and eGFR variability on the mortality rate in either the primary or sensitivity analysis (Table III). Furthermore, we saw no evidence of interaction in sensitivity analyses incorporating the baseline medication composite as a covariate (Supplemental Table V) or participants with LVEF measured by echocardiography (Supplemental Table VI).
Figure IV.
Cumulative Mortality Based on eGFR Variability Stratified by HF Phenotype. Cumulative probability of all-cause mortality in participants during the mortality outcome window, defined as years 2–7 after coronary angiography. Adjusted 5-year mortality rate ratios for each subgroup are presented, assessed by including in the regression model terms for HF phenotype, eGFR variability tertile, and the interaction between these factors. Comparison is low eGFR variability tertile for each subgroup. Adjustments were made for covariates as in Table 2. Abbreviations as in Table 2.
Table III.
Adjusted Associations* between eGFR Variability and Cumulative Mortality Rates Stratified by HF Phenotype with interaction testing by HF Phenotype.
| Comparison† | Subgroup | Adjusted 5-year Mortality Rate Ratio (95% CI) | Interaction Test |
|---|---|---|---|
|
| |||
| a. Coefficient of variation | |||
| Medium eGFR variability | No HF | 1.35 (1.08 – 1.68) | 0.67 |
| HFpEF | 1.15 (0.80 – 1.66) | ||
| HFrEF | 1.46 (0.99 – 2.14) | ||
| High eGFR variability | No HF | 1.84 (1.48 – 2.29) | 0.49 |
| HFpEF | 1.44 (1.02 – 2.03) | ||
| HFrEF | 1.78 (1.22 – 2.59) | ||
|
| |||
| b. Root-mean-square error | |||
| Medium eGFR variability | No HF | 1.27 (1.01 – 1.59) | 0.12 |
| HFpEF | 0.95 (0.66 – 1.38) | ||
| HFrEF | 1.62 (1.12 – 2.33) | ||
| High eGFR variability | No HF | 1.48 (1.17 – 1.86) | 0.60 |
| HFpEF | 1.44 (1.04 – 1.99) | ||
| HFrEF | 1.81 (1.24 – 2.65) | ||
Discussion
In a cohort of participants with CKD who underwent coronary angiography, the presence of HF, particularly HFrEF, was associated with greater eGFR variability in the outpatient setting. High eGFR variability was associated with a substantial and independent greater rate of mortality over long-term follow-up. This higher relative risk of death with high kidney function variability was consistent irrespective of HF status, with similar degrees of excess risk seen among patients with HFrEF, HFpEF, and no HF. The relationship between mortality and eGFR variability in our study is similar to those in prior studies in a general CKD population,[6, 7] but to our knowledge this is the first study to describe a relationship between HF and eGFR variability, and to evaluate the interplay of both conditions on clinical outcomes.
A relationship between HF and kidney function is well-established, with kidney dysfunction having the potential to worsen HF, and vice versa, via multiple mechanisms.[15] Hemodynamic and neurohormonal aberrations are interwoven and result in abnormal physiology that damage both organ systems.[16, 17] Patients with HF may be more susceptible to fluctuations in kidney function due to a reduced capacity to augment cardiac output in response to kidney hypoperfusion and concurrent renin-angiotensin-aldosterone system overactivation, resulting in insufficient capacity for autoregulation. In particular, in a subset of patients with HFrEF, a global low flow state may increase susceptibility to transient hypotension, leading to reduced kidney perfusion and decreased glomerular filtration. Likewise, patients with HFpEF have been shown to have reduced stroke volume reserve to augment cardiac output in response to increased circulatory demands,[18] and increased arterial stiffness in HFpEF may also result in impaired vasodilatory mechanisms to compensate for reduced kidney blood flow.[19]
When controlling for factors likely to affect neurohormonal or hemodynamic homeostasis, as well as baseline eGFR function, mean eGFR, and the overall eGFR trend, we observed a significant increase in eGFR variability in a CKD population with HF compared to those without HF. A potential mechanism is that patients with HF may be more likely to experience small deviations in eGFR for which those without HF may rapidly compensate for by increasing cardiac output. Further, in the entire population with CKD, greater eGFR variability was associated with higher mortality during follow-up. These findings suggest monitoring of eGFR variability, which could be easily calculated and reported by the electronic health record for each individual patient, may aid clinicians identifying patients at higher risk for poor outcomes and better inform decisions regarding outpatient monitoring. eGFR variability may also be an important clinical feature to include in predictive algorithms and classification schemes in CKD and HF, both high priority areas for ongoing research.[20–22]
While greater eGFR variability was associated with higher mortality in the overall cohort, and greater eGFR variability was observed in participants with either HF phenotype, we did not observe a differential effect of eGFR variability on mortality stratified by HF phenotype. Nonetheless, given the higher baseline risk of the HF cohorts, relative risk translated into higher increases in absolute risk among HF patients, particularly HFrEF. Indeed, rates of mortality among those with high kidney function variability were >50% among patients with HFrEF or HFpEF. Likewise, comparing patients with HFrEF with low versus high eGFR variability showed a ~20% absolute increase in risk of death. Thus, although relative risk was consistent irrespective of HF status, the combination of higher prevalence of eGFR variability and higher absolute risk supports patients with HF as a population where longitudinal monitoring of eGFR variability may offer particularly high clinical value.
While we attempted to describe medication usage for participants at the time of catheterization and effects of baseline medication use on outcomes, it is possible fluctuations in eGFR in the mortality window were reflective of a differential response to changes in medical therapy over time. Many guideline-directed and decongestive therapies for HF may be poorly tolerated in patients with CKD due to physiologic alterations that directly impact kidney hemodynamics, reduce capacity for autoregulation and subsequent measured kidney function. In contrast, previous studies have shown that worsening kidney function in the hospital setting with diuresis may be associated with improved mortality in the context of adequate decongestion.[23] Overall, eGFR variability in this context may act a summary indicator of various physiologic and medication-related changes, particularly in patients with HF known to have hemodynamic compromise. This study relied on data prior to the introduction of sodium glucose cotransporter-2 inhibitors and sacubitril/valsartan in management of HF and further studies to examine the applicability of eGFR variability monitoring during use of these medications will be warranted.
Limitations
This study had several limitations. First, the database utilized was not HF-specific. While this allowed for comparison to those without HF, there may have been limited power to detect differential mortality effects attributable to greater eGFR variability among patients with HF. Second, our results must be interpreted in the context of the pre-specified definition of phenotypes utilized, which relied on LVEF and NYHA symptomatology. Requiring symptoms by NYHA class resulted in a population of stage C and stage D HF within the defined HF groups, whereas patients with asymptomatic LV dysfunction or stage A or B HF were included in the group without HF. For example, 258 (11.4%) of patients in the group without HF had an LVEF <40%; thus, the results of this study may be more applicable to a HF population with an increased symptom burden. Furthermore, eGFR in this study was based on available serum creatinine values from routine practice, which are influenced by multiple factors such as age, body mass, and medication usage, and may temporally lag behind kidney injury. Future studies may attempt to incorporate novel, specific biomarkers for kidney injury such as NGAL or cystatin C, and may investigate the role that either the level of or changes in albuminuria has in those with HFrEF and CKD. Also, the clinical indications for creatinine value measurements is not known, which may predispose to measurement bias as those with abnormal or concerning values are more likely to have a higher number of measurements.
Third, this cohort was restricted to those undergoing coronary angiography at a university hospital. While this is a common procedure and likely selects for a generalizable population with suspected cardiovascular disease, participants in this study may have had a greater incidence of ischemic cardiomyopathy compared to a general HF population. Also, the indication for coronary angiography was not recorded for this study, and it is possible participants experienced sequelae of their condition requiring cardiac angiography, such as acute coronary syndrome, that would affect cardiac function during the study windows; thus, we were unable to capture temporal changes in cardiac function that may affect patient characterization. Further, given DDCD inclusion criteria, the cohort studied received iodinated contrast dye, which carries a risk of nephrotoxicity. To minimize the potential impact of contrast-induced nephropathy on the results, the pre-specified window for measuring eGFR variability began 3 months after coronary angiography, and the baseline eGFR used in calculations utilized a mean of outpatient creatinine values from a window surrounding coronary angiography to reduce the effect of peri-procedural acute kidney episodes.
Finally, though we saw no differences in mortality when adjusted for a composite of baseline medications received, we were unable to accurately capture medication usage and changes throughout the course of the study; while there was increased baseline usage of renin-angiotensin system inhibitors and diuretics in those with HF or increased eGFR variability, the changes in prescription and dose of these medications or their effects on variability an mortality in our study are not known and may contribute to eGFR variability and any mortality interaction testing. Similarly, this retrospective study cannot determine cause-effect relationship, and residual confounding may remain despite rigorous multivariable adjustment.
Conclusions
In a cohort of patients with CKD, the presence of HF, and particularly HFrEF, was an independent risk factor for greater variability in kidney function. Greater kidney function variability was independently associated with risk of death over long-term follow-up and was consistent among those with CKD regardless of HF status. Given the heightened degree of kidney function variability among HF patients and the prognostic significance, future studies are required to understand mechanisms, triggers, and mitigation strategies for kidney function variability among patients with HF. Furthermore, eGFR variability may be an important feature of kidney function to incorporate in predictive algorithms focused on classification or prognostication in CKD and HF.
Supplementary Material
Acknowledgements:
No acknowledgements. The authors are solely responsible for the design and conduct of this study, all study analyses and drafting and editing of the paper.
Abbreviations
- (HF)
Heart failure
- (CKD)
chronic kidney disease
- (eGFR)
estimated glomerular filtration rate
- (HFrEF)
heart failure with reduced ejection fraction
- (HFpEF)
heart failure with preserved ejection fraction
- (DDCD)
Duke Databank for Cardiovascular Diseases
- (LVEF)
left ventricular ejection fraction
- (CV)
coefficient of variation
- (CI)
confidence interval
- (RR)
rate ratio
Footnotes
Declarations of interest: AMH, JLS, LKS, and KC have no disclosures to report. JJS has received consulting fees from Tricida and modest research support for clinical event activities related to trials sponsored by Eli Lilly, Sanofi and GlaxoSmithKline. SJG has received a Heart Failure Society of America/ Emergency Medicine Foundation Acute Heart Failure Young Investigator Award funded by Novartis; has received research support from Amgen, AstraZeneca, Bristol-Myers Squibb, Merck and Novartis; has served on advisory boards for Amgen and Cytokinetics; and serves as a consultant for Amgen and Merck. PHP receives research support from the National Institutes of Health (R34HL140477, R34HL140477) and has served on advisory boards for AstraZeneca, Relypsa. RJM receives research support from the National Institutes of Health (U01HL125511–01A1 and R01AG045551–01A1), Amgen, AstraZeneca, Bayer, GlaxoSmithKline, Gilead, InnoLife, Luitpold/American Regent, Medtronic, Merck, Novartis and Sanofi; honoraria from Abbott, Amgen, AstraZeneca, Bayer, Boston Scientific, Janssen, Luitpold Pharmaceuticals, Medtronic, Merck, Novartis, Roche, Sanofi and Vifor; and has served on an advisory board for Amgen, AstraZeneca, Luitpold, Merck, Novartis and Boehringer Ingelheim.
References
- [1].Lofman I, Szummer K, Hagerman I, Dahlström U, Lund LH, Jernberg T. Prevalence and prognostic impact of kidney disease on heart failure patients. Open Heart. 2016;3:e000324. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [2].Lofman I, Szummer K, Dahlstrom U, Jernberg T, Lund LH. Associations with and prognostic impact of chronic kidney disease in heart failure with preserved, mid-range, and reduced ejection fraction. European journal of heart failure. 2017;19:1606–14. [DOI] [PubMed] [Google Scholar]
- [3].Galil AG, Pinheiro HS, Chaoubah A, Costa DM, Bastos MG. Chronic kidney disease increases cardiovascular unfavourable outcomes in outpatients with heart failure. BMC nephrology. 2009;10:31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Smith DH, Thorp ML, Gurwitz JH, McManus DD, Goldberg RJ, Allen LA, et al. Chronic kidney disease and outcomes in heart failure with preserved versus reduced ejection fraction: the Cardiovascular Research Network PRESERVE Study. Circulation Cardiovascular quality and outcomes. 2013;6:333–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].McCullough PA, Franklin BA, Leifer E, Fonarow GC. Impact of reduced kidney function on cardiopulmonary fitness in patients with systolic heart failure. Am J Nephrol. 2010;32:226–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].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;82:1332–8. [DOI] [PubMed] [Google Scholar]
- [7].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;82:1208–14. [DOI] [PubMed] [Google Scholar]
- [8].Uehara K, Yasuda T, Shibagaki Y, Kimura K. Estimated Glomerular Filtration Rate Variability Independently Predicts Renal Prognosis in Advanced Chronic Kidney Disease Patients. Nephron. 2015;130:256–62. [DOI] [PubMed] [Google Scholar]
- [9].Schefold JC, Filippatos G, Hasenfuss G, Anker SD, von Haehling S. Heart failure and kidney dysfunction: epidemiology, mechanisms and management. Nature reviews Nephrology. 2016;12:610–23. [DOI] [PubMed] [Google Scholar]
- [10].Rosati RA, McNeer JF, Starmer CF, Mittler BS, Morris JJ Jr., Wallace AG. A new information system for medical practice. Arch Intern Med. 1975;135:1017–24. [PubMed] [Google Scholar]
- [11].Harris PJ, Lee K, Harrell F Jr, Behar V, Rosati R. Outcome in medically treated coronary artery disease. Ischemic events: nonfatal infarction and death. Circulation. 1980;62:718–26. [DOI] [PubMed] [Google Scholar]
- [12].Banks A, Broderick S, Chiswell K, Shaw L, Devore A, Fiuzat M, et al. Comparison of Clinical Characteristics and Outcomes of Patients With Versus Without Diabetes Mellitus and With Versus Without Angina Pectoris (from the Duke Databank for Cardiovascular Disease). Am J Cardiol. 2017;119:1703–9. [DOI] [PubMed] [Google Scholar]
- [13].Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF, Feldman HI, et al. A new equation to estimate glomerular filtration rate. Annals of internal medicine. 2009;150:604–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Boyle CA, Decouflé P. National sources of vital status information: extent of coverage and possible selectivity in reporting. American Journal of Epidemiology. 1990;131:160–8. [DOI] [PubMed] [Google Scholar]
- [15].Ronco C, Bellasi A, Di Lullo L. Cardiorenal Syndrome: An Overview. Adv Chronic Kidney Dis. 2018;25:382–90. [DOI] [PubMed] [Google Scholar]
- [16].Zannad F, Rossignol P. Cardiorenal Syndrome Revisited. Circulation. 2018;138:929–44. [DOI] [PubMed] [Google Scholar]
- [17].Rangaswami J, Mathew RO. Pathophysiological Mechanisms in Cardiorenal Syndrome. Adv Chronic Kidney Dis. 2018;25:400–7. [DOI] [PubMed] [Google Scholar]
- [18].Abudiab MM, Redfield MM, Melenovsky V, Olson TP, Kass DA, Johnson BD, et al. Cardiac output response to exercise in relation to metabolic demand in heart failure with preserved ejection fraction. European journal of heart failure. 2013;15:776–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Reddy YNV, Andersen MJ, Obokata M, Koepp KE, Kane GC, Melenovsky V, et al. Arterial Stiffening With Exercise in Patients With Heart Failure and Preserved Ejection Fraction. Journal of the American College of Cardiology. 2017;70:136–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Triposkiadis F, Butler J, Abboud FM, Armstrong PW, Adamopoulos S, Atherton JJ, et al. The continuous heart failure spectrum: moving beyond an ejection fraction classification. Eur Heart J. 2019;40:2155–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].House AA, Wanner C, Sarnak MJ, Piña IL, McIntyre CW, Komenda P, et al. Heart failure in chronic kidney disease: conclusions from a Kidney Disease: Improving Global Outcomes (KDIGO) Controversies Conference. Kidney Int. 2019;95:1304–17. [DOI] [PubMed] [Google Scholar]
- [22].Shah SJ, Katz DH, Selvaraj S, Burke MA, Yancy CW, Gheorghiade M, et al. Phenomapping for novel classification of heart failure with preserved ejection fraction. Circulation. 2015;131:269–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [23].Testani JM, McCauley BD, Chen J, Coca SG, Cappola TP, Kimmel SE. Clinical characteristics and outcomes of patients with improvement in renal function during the treatment of decompensated heart failure. J Card Fail. 2011;17:993–1000. [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.