Visual Abstract
Keywords: chronic heart failure, mortality risk, renal function decline, hypertension and cardiorenal disorders
Abstract
Key Points
The optimal balance between decongestion and kidney function preservation remains uncertain in outpatients with heart failure.
Improved congestion is generally associated with lower mortality risk with eGFR declines and only higher risk when eGFR exceeds 40%.
Background
Although both volume overload and reduced level of kidney function are associated with higher mortality in heart failure, decongestion can lead to kidney function decline. The optimal balance between sustaining decongestion and preserving kidney function remains uncertain among outpatients with heart failure. We compared associations of postdischarge changes in kidney function and congestion status with mortality in the Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study with Tolvaptan trial.
Methods
This post hoc analysis of a randomized controlled trial included 3404 participants discharged from a heart failure hospitalization. Compared with eGFR and clinical congestion score at discharge, eight time-varying exposure groups were defined: improved or worsened congestion, with varying degrees of eGFR decline (no decline, 1%–20%, 21%–40%, and 41% or greater). The association of these groups with all-cause mortality was assessed using marginal structural models to account for time-dependent confounding.
Results
The mean (SD) age and eGFR at discharge were 66 (12) years and 59.6 (22.3) ml/min per 1.73 m2, respectively. Over a median (interquartile range) follow-up of 44 (25–71) weeks, 740 patients died. Both higher degrees of eGFR decline and worsened congestion were associated with higher mortality risk. Compared with patients with worsened congestion and no eGFR decline, those with improved congestion had lower mortality risk (hazard ratio [HR], 0.51 [95% confidence interval (CI), 0.35 to 0.74] for no eGFR decline; HR, 0.56 [95% CI, 0.38 to 0.85] for 1%–20% eGFR decline; and HR, 0.80 [95% CI, 0.46 to 1.39] for 21%–40% eGFR decline), whereas those with improved congestion and 41% or greater eGFR decline had higher risk (HR, 2.23; 95% CI, 1.06 to 4.66).
Conclusions
Compared with worsened congestion and no eGFR decline, improved congestion is generally associated with lower mortality with eGFR declines, unless eGFR decline exceeds 40%.
Clinical Trial registry name and registration number:
Introduction
Despite various guideline-recommended therapies over the past several decades, heart failure remains a major cause of mortality, morbidity, and hospitalization.1 Among patients with heart failure, those with a history of heart failure hospitalization are considered a particularly vulnerable population, in whom the risk of mortality was three-fold higher than that in those never hospitalized.2 In addition, a major portion of the medical costs associated with heart failure is due to recurrent admissions.3 Hence, increasing attention has been paid to postdischarge care in heart failure.
Both volume overload1,4 and reduced level of kidney function5,6 are associated with higher mortality in heart failure. Importantly, decongestion can lead to a kidney function decline, which represents a challenge for the clinician. The question is which should be prioritized—further decongestion or preservation of kidney function at the expense of decongestion—in the treatment of outpatients with heart failure. In the treatment of hospitalized patients with acute decompensated heart failure (ADHF), recent studies suggest that an acute decline in kidney function is not necessarily associated with worse cardiovascular (CV) outcomes within the setting of effective decongestion,7,8 suggesting the primacy of decongestion in this population. However, the question becomes particularly challenging among discharged patients because their volume status is assumed to be better controlled than hospitalized patients with ADHF, and the clinical significance of further decongestion at the expense of kidney function remains uncertain. There are few data on the optimal balance between sustaining decongestion and preserving kidney function in the outpatient setting.
In this post hoc, time-dependent analysis of the Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study with Tolvaptan (EVEREST) trial, we examined and compared the associations of changes in kidney function and volume status over time after discharge with heart failure outcomes. We also sought to explore the relation of potential thresholds of eGFR decline, irrespective of decongestion, with outcomes.
Methods
Study Design and Population
EVEREST was an international, multicenter, randomized controlled trial that investigated the short-term and long-term efficacy and safety of tolvaptan, a vasopressin V2 receptor antagonist, among patients hospitalized for ADHF (NCT00071331).9,10 Detailed methods of the trial have been described elsewhere.11 In brief, adult patients (18 years and older) with a history of chronic heart failure who had a reduced left ventricular ejection fraction (LVEF; ≤40%) and had been hospitalized primarily for worsening congestive heart failure no more than 48 hours earlier were enrolled. Congestion based on two or more clinical signs or symptoms was required for eligibility. Key exclusion criteria included cardiac mechanical support, systolic BP (SBP) of <90 mm Hg, serum creatinine of >3.5 mg/dl, and any comorbid condition with an expected survival of less than 6 months. Participants were randomized to either 30 mg/d of tolvaptan or placebo in addition to standard heart failure therapies in a 1:1 ratio. Among a total of 4133 eligible patients, the present analysis was restricted to 3943 patients who survived hospitalization and were followed after discharge. Those who had no measurements of the exposure variables or any of covariates (n=539) were additionally excluded from the main analysis (Figure 1) but were included in a sensitivity analysis using multiple imputation. Patients were followed with scheduled visits from the date of discharge (baseline date in this study) to the date of death, loss to follow-up, or the end of the trial period (July 2006), whichever occurred first. The scheduled visits were at 1, 4, and 8 weeks after discharge and then every 8 weeks thereafter. EVEREST was conducted in accordance with the Declaration of Helsinki, and its protocol was approved by the institutional review board or ethics committee at each site. All participants gave written informed consent before the enrollment; the present analysis was deemed exempt from review by the Tufts Health Sciences Institutional Review Board.
Figure 1.
Flow diagram of the study. EVEREST, Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study with Tolvaptan.
Exposures of Interest
Time-varying kidney function and volume status exposures were defined. The kidney function exposure was defined as changes in eGFR after discharge and categorized into four groups: no eGFR decline in comparison with the value at discharge, 1%–20% decline, 21%–40% decline, and 41% or more decline. The eGFR was calculated using the 2021 CKD Epidemiology Collaboration formula (Supplemental Methods)12 and was ascertained at every scheduled visit.
The volume status exposure was defined as changes in the clinical congestion score (CCS; 0–12) after discharge. CCS, initially developed in EVEREST,13–15 was the sum on four-point scale for each of orthopnea (none, seldom, frequent, continuous), jugular venous distension (≤6, 6–9, 10–15, 15 cm or greater), rales (none, bases, up to <50%, to >50%), and pedal edema (absent/trace, slight, moderate, marked). Although a validated and widely accepted scoring system for assessing volume status has not yet been established, CCS incorporates findings considered most specific to volume overload.16 In this study, CCS was also assessed at every scheduled visit. Changes in CCS from discharge was categorized into two groups: congestion improved (CCS decreased) and congestion worsened (CCS increased). Scores remaining the same as at discharge were categorized as improved.
Based on these four-group kidney function and two-group volume status exposures, the following eight time-varying groups were defined as our main exposure of interest: Congestion improved with no eGFR decline, with 1%–20% decline, with 21%–40% decline, and with 41% or more decline group and congestion worsened with no eGFR decline, 1%–20% decline, 21%–40% decline, and 41% or more decline group.
Study Outcomes
The primary outcome was all-cause mortality, consistent with the prespecified primary outcome in the original trial.9 The secondary outcomes included CV mortality and the composite of CV mortality or first rehospitalization for heart failure. These events were adjudicated by a blinded clinical event committee.
Statistical Analyses
Data are presented as means±SD or medians and interquartile ranges (IQRs) for continuous variables and as numbers and percentages for categorical variables, as appropriate.
Association between Group Designation and Clinical Outcomes
When examining the association of eGFR or CCS over time with heart failure outcomes in an observational analysis, time-varying confounding is a major concern. For example, an eGFR (or CCS) value (the exposure) may affect the prescription of diuretic agents (a time-dependent confounder), which in turn may influence eGFR (or CCS) at the next visit (Supplemental Figure 1). Under such a causal chain relation between the time-dependent exposure and confounder, biased estimates will be produced by time-dependent Cox regression models.17–19 To address this issue, we used marginal structural models (MSMs). The MSM can minimize time-dependent confounding by applying time-varying inverse probability weights (IPWs).20,21 Specifically, time-varying IPWs are used to create a balanced pseudo-population (IPW-weighted population), in which time-dependent confounding and collider stratification bias are minimized, and the estimated effect of an exposure (e.g., a treatment or individual's condition) on an outcome theoretically reflects that in a real-world population.18–20,22–25 In other words, by incorporating IPW-weighting derived at each follow-up visit, the present MSM enables designation of each patient remaining in one of the eight exposure groups throughout follow-up (i.e., balanced pseudo-population), as a means to examine the effect of the exposure on study outcomes while attempting to limit bias from time-varying confounders to the smallest degree possible (Figure 2).
Figure 2.
The MSM of the study. CCS, clinical congestion score; IPW, inverse probability weight; MSM, marginal structural model.
The IPW, which is the product of the stabilized inverse probability of treatment weight (sIPTW) times the stabilized inverse probability of censoring weight (sIPCW), was estimated at each time point. The sIPTW (or sIPCW) was the reciprocal of the probability of having each patient's own exposure history (or being uncensored) predicted by a multinomial logistic (or logistic) regression model with baseline and time-varying covariates, as appropriate. The sIPTW (or sIPCW) was generated by multiplying the sIPTW (or sIPCW) by the probability of having the exposure history (or being uncensored) predicted by another multinomial logistic (or logistic) regression model with baseline covariates only. The IPWs were truncated at the 1st and 99th percentiles.
Hazards of study outcomes were first compared among the four eGFR decline groups (no decline; 1%–20% decline; 21%–40% decline; and 41% or more decline) and between the two congestion groups (congestion improved group versus congestion worsened group). Subsequently, hazards were compared among the eight exposure groups in the primary analysis (See Figure 2 for details). Here, the congestion worsened with no eGFR decline group served as the reference group to focus on comparing each congestion improved group to this reference. Hazard ratios (HRs) were estimated by IPW-weighted pooled logistic regression models, which emulate time-dependent Cox regression models with the ability to incorporate time-dependent weighting of the observations.20,21 Follow-up time was included in the models as a restricted cubic spline term with three knots. These outcome models were clustered by individual patients using the cluster option of the robust estimator of variance.20,21 In addition to incorporating IPWs, which were estimated using baseline and time-varying covariates as aforementioned, these outcome models included further regression adjustment for all baseline covariates used to estimate IPWs, aiming to minimize potential residual confounding from them.26
Baseline covariates included demographic and behavioral characteristics (age, sex, race, and smoking status), history (heart failure hospitalization preceding the latest one, percutaneous coronary intervention, coronary artery bypass grafting, and stroke), comorbidities (hypertension, diabetes mellitus [DM], dyslipidemia, and ischemic heart disease [IHD]), and covariates ascertained at randomization immediately after hospital admission (LVEF, the New York Heart Association classification, B-type natriuretic peptide [BNP] levels, loop diuretic dose, and randomization group [tolvaptan]). Covariates ascertained at discharge were also included as baseline covariates: physical characteristics (weight and body mass index), vital signs (SBP and heart rate), CCS, laboratory data (BNP, eGFR, hematocrit, serum albumin, sodium, and potassium), and medications (angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers, β-blockers, and diuretic agents).
Time-varying covariates included age, SBP, heart rate, CCS, laboratory data (eGFR, hematocrit, serum albumin, sodium, and potassium), medications (angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers, β blockers, and diuretic agents), and lagged variables, which had values at the most recent visit prior, of all these time-varying covariates.
Sensitivity Analyses
The association of the eight congestion-eGFR grouping with all-cause mortality was re-examined using the MSM approach in several sensitivity analyses. First, we repeated the analysis, redefining the volume status exposure using BNP, perhaps a more objective surrogate of volume status,27 which was measured at every scheduled visit in EVEREST. Worsened congestion was defined as an increase in plasma BNP in comparison with the discharge value, whereas improved congestion was defined as no increase in BNP. Second, we recategorized CCS that remained the same as at discharge as worsened congestion rather than improved congestion. Third, we used IPWs truncated at the 5th and 95th percentiles in the MSM. Fourth, we redefined eight time-varying exposure groups: improved or worsened congestion, with alternate degrees of eGFR decline (no decline, 1%–30%, 31%–40%, and 41% or greater). Last, we included 539 patients who had been excluded from the main analysis because of missing any exposure variables or covariates (Figure 1) by imputing missing values using multiple imputation by chained equations.28 Linear regression imputation was performed for continuous variables (body mass index, weight, SBP, heart rate, LVEF, hematocrit, serum albumin, sodium, potassium, chloride, plasma BNP, and eGFR), ordinal logistic regression imputation for ordinal categorical variables (New York Heart Association classification, orthopnea, jugular venous distension, rales, and pedal edema), and logistic regression imputation for dichotomous variables (current smoking, DM, and IHD). Ten imputed datasets were created, analyzed separately, and combined according to Rubin's rules. Missing data during the follow-up period were imputed by the last observation carried forward method.
Effect Modification
Effect modifications were assessed by CCS (median or above versus below median) at discharge and randomization group (tolvaptan versus placebo). P values for interaction were computed by adding each interaction term to the MSM in the primary analyses.29
Statistical tests were two-tailed with P < 0.05 considered significant. All statistical analyses were performed using SAS Enterprise Guide (version 7.12; SAS Institute Inc.) and Stata/MP (version 18.0; Stata Corp LLC) software.
Results
Study Population and Patient Characteristics
A total of 3404 patients were analyzed (Figure 1). The mean (SD) age, SBP, eGFR, and median (IQR) BNP at discharge were 65.7 (11.6) years, 115.9 (16.7) mm Hg, 59.6 (22.3) ml/min per 1.73 m2, and 439 (192–938) pg/ml, respectively (Supplemental Table 1). Twenty six percent were women, 38% had DM, and 66% had IHD. Characteristics were not substantially different between included and excluded patients (Supplemental Table 1). Throughout a median (IQR) follow-up of 44.0 (24.7–70.6) weeks, groups with higher degree of eGFR decline exhibited lower eGFRs, and the congestion worsened groups exhibited higher CCSs and BNPs than the congestion improved groups in the weighted population (Table 1 and Supplemental Table 2). Patient characteristics were generally well balanced throughout follow-up among the IPW-weighted eight groups. The mean IPW, truncated at 1 and 99 percentiles, was 1.0.
Table 1.
Characteristics of the weighted population at 1 week after discharge
| Characteristics of Weighted Population (n=3386a) | Congestion Improved | Congestion Worsened | ||||||
|---|---|---|---|---|---|---|---|---|
| eGFR Decline None (n=1334a) |
eGFR Decline 1%–20% (n=1060a) |
eGFR Decline 21%–40% (n=158a) |
eGFR Decline ≥41% (n=51a) |
eGFR Decline None (n=376a) |
eGFR Decline 1%–20% (n=339a) |
eGFR Decline 21%–40% (n=57a) |
eGFR Decline ≥41% (n=11a) |
|
| Demographic characteristics | ||||||||
| Age (yr) | 65.6±11.4 | 64.9±11.7 | 67.1±13.9 | 66.1±11.0 | 65.5±12.5 | 66.6±11.6 | 65.9±12.2 | 65.3±10.7 |
| Sex (female) | 362 (27) | 256 (24) | 53 (34) | 10 (19) | 81 (22) | 88 (26) | 10 (17) | 1 (6) |
| Race | ||||||||
| Black | 77 (6) | 74 (7) | 19 (12) | 8 (16) | 42 (11) | 33 (10) | 6 (10) | 1 (7) |
| Others | 98 (7) | 70 (7) | 9 (6) | 0 (1) | 36 (10) | 22 (7) | 7 (13) | 0 (4) |
| White | 1159 (87) | 916 (86) | 130 (82) | 42 (84) | 298 (79) | 284 (84) | 44 (77) | 10 (89) |
| BMI (kg/m2) | 27.5±5.3 | 27.4±5.4 | 27.6±5.9 | 27.0±4.3 | 28.5±6.1 | 28.4±5.7 | 27.5±5.3 | 29.3±6.0 |
| SBP (mm Hg) | 117.2±18.3 | 118.7±18.6 | 117.4±18.3 | 116.7±19.3 | 114.6±19.4 | 116.0±19.0 | 109.5±17.2 | 112.2±20.3 |
| Heart rate (bpm) | 74.8±12.0 | 75.2±13.0 | 76.6±13.2 | 74.6±10.0 | 75.0±13.2 | 74.5±13.2 | 78.2±15.5 | 63.7±13.7 |
| LVEF (%)b | 27.9±7.8 | 28.0±7.8 | 26.9±7.4 | 23.3±10.6 | 26.2±8.5 | 27.4±8.2 | 25.8±9.1 | 28.7±10.4 |
| CCS | 0 (0–1) | 0 (0–1) | 0 (0–1) | 0 (0–1) | 2 (1–4) | 2 (1–4) | 3 (2–5) | 2 (2–5) |
| Current smoking | 177 (13) | 137 (13) | 27 (17) | 2 (4) | 47 (12) | 37 (11) | 5 (9) | 0 (0) |
| History | ||||||||
| Heart failure hospitalizationc | 813 (61) | 622 (59) | 88 (56) | 22 (43) | 220 (59) | 210 (62) | 32 (56) | 11 (93) |
| PCI | 115 (9) | 104 (10) | 16 (10) | 9 (17) | 36 (10) | 54 (16) | 12 (21) | 1 (9) |
| CABG | 107 (8) | 97 (9) | 15 (10) | 2 (5) | 36 (10) | 43 (13) | 6 (10) | 3 (23) |
| Stroke | 148 (11) | 106 (10) | 17 (11) | 1 (3) | 58 (16) | 48 (14) | 3 (4) | 0 (1) |
| Comorbid conditions | ||||||||
| Hypertension | 980 (74) | 750 (71) | 116 (73) | 42 (84) | 261 (69) | 246 (73) | 32 (57) | 9 (77) |
| DM | 493 (37) | 358 (34) | 63 (40) | 21 (41) | 164 (44) | 154 (46) | 26 (46) | 6 (56) |
| Dyslipidemia | 633 (48) | 471 (44) | 75 (47) | 13 (26) | 182 (48) | 188 (56) | 29 (51) | 7 (58) |
| IHD | 866 (65) | 700 (66) | 87 (55) | 23 (46) | 241 (64) | 231 (68) | 35 (62) | 11 (95) |
| Medications | ||||||||
| ACEI or ARB | 1144 (86) | 917 (87) | 132 (84) | 47 (93) | 289 (77) | 285 (84) | 48 (84) | 9 (83) |
| β blocker | 1021 (77) | 792 (75) | 124 (78) | 30 (59) | 270 (72) | 271 (80) | 43 (75) | 8 (71) |
| Diuretic | 1260 (95) | 979 (92) | 154 (98) | 44 (87) | 352 (94) | 319 (94) | 54 (95) | 11 (100) |
| Laboratory data | ||||||||
| Hematocrit (%) | 42.6±5.6 | 42.6±5.7 | 40.4±6.1 | 42.9±5.4 | 41.3±6.3 | 41.3±5.7 | 40.4±5.1 | 38.4±3.2 |
| Albumin (g/dl) | 3.9±0.5 | 3.9±0.5 | 3.8±0.6 | 3.7±0.5 | 3.9±0.5 | 3.9±0.5 | 3.8±0.5 | 4.1±0.3 |
| Sodium (mEq/L) | 140.8±4.7 | 140.6±4.7 | 140.5±4.4 | 138.9±5.3 | 140.1±4.6 | 140.2±4.9 | 139.2±6.1 | 139.6±4.3 |
| Potassium (mEq/L) | 4.7±0.6 | 4.6±0.6 | 4.5±0.7 | 4.3±0.8 | 4.7±0.7 | 4.6±0.6 | 4.5±0.5 | 4.4±0.5 |
| eGFR (ml/min per 1.73 m2) | 64.3±22.6 | 62.8±22.3 | 45.5±18.3 | 33.6±12.7 | 59.7±23.2 | 57.2±21.8 | 45.7±17.6 | 29.1±12.6 |
| BNP (pg/ml) | 429 (203–815) | 470 (216–948) | 419 (241–895) | 378 (244–735) | 637 (279–1327) | 615 (244–1253) | 798 (403–1824) | 403 (148–1447) |
| Randomization group | ||||||||
| Tolvaptan | 687 (52) | 527 (50) | 85 (54) | 29 (57) | 186 (49) | 161 (48) | 22 (39) | 1 (10) |
Values are presented as either n (%), mean±SD, or median (interquartile range). ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker; BMI, body mass index; BNP, B-type natriuretic peptide; CABG, coronary artery bypass grafting; CCS, clinical congestion score; DM, diabetes mellitus; IHD, ischemic heart disease; LVEF, left ventricular ejection fraction; PCI, percutaneous coronary intervention; SBP, systolic BP.
The number in each category was based on an inverse probability weight-weighted population.
Left ventricular ejection fraction values were those at randomization.
Heart failure hospitalization does not include the latest one that was just before the baseline in the present study.
Association between Group Designation and Clinical Outcomes
Over a median follow-up of 44.0 weeks, there were 740 deaths, of which 559 were due to CV causes (Figure 1). Of a total of 3404 eligible patients, 1072 experienced heart failure rehospitalization, and 1408 experienced the composite outcome of CV mortality or heart failure rehospitalization. In the fully adjusted MSM, a higher degree of eGFR decline was associated with a higher risk of all-cause mortality (P for trend < 0.001), with the highest risk in the 41% or greater eGFR decline group (versus the no eGFR decline group; HR, 3.27; 95% confidence interval [CI], 2.05 to 5.20; Table 2). The congestion worsened group (versus the congestion improved group) was associated with a 116% higher risk of all-cause mortality (HR, 2.16; 95% CI, 1.79 to 2.60; Table 2). There was an association between the eight congestion-eGFR groups and mortality, with the lowest risk in the congestion improved with no eGFR decline group and the highest risk in the congestion worsened with 41% or greater eGFR decline group (Figure 3 and Table 3). In both congestion improved and worsened groups, a higher degree of eGFR decline was associated with a higher risk of mortality (P for trend < 0.001). As compared with the congestion worsened with no eGFR decline group, the risks of all-cause mortality were significantly lower in the congestion improved with no eGFR decline group (HR, 0.51; 95% CI, 0.35 to 0.74) and the congestion improved with 1%–20% decline group (HR, 0.56; 95% CI, 0.38 to 0.85), and lower (but not significantly different) in the congestion improved with 21%–40% decline group (HR, 0.80; 95% CI, 0.46 to 1.39). In the congestion improved with 41% or greater decline group, the risk was higher than congestion worsened with no GFR decline (HR, 2.23; 95% CI, 1.06 to 4.66; Figure 3 and Table 3).
Table 2.
Association of kidney function and volume status exposures with study outcomes
| Outcome: All-Cause Mortality | No. of Events | Model 1 | Model 2 | Model 3 | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| HR (95% CI) | P Value | P for Trend | HR (95% CI) | P Value | P for Trend | HR (95% CI) | P Value | P for Trend | ||
| Kidney function exposure | <0.001 | <0.001 | <0.001 | |||||||
| No eGFR decline | 289 | Ref. | — | Ref. | — | Ref. | — | |||
| 1%–20% eGFR decline | 251 | 1.04 (0.89 to 1.23) | 0.62 | 1.06 (0.88 to 1.27) | 0.56 | 1.14 (0.89 to 1.46) | 0.31 | |||
| 21%–40% eGFR decline | 120 | 1.50 (1.21 to 1.85) | <0.001 | 1.38 (1.09 to 1.76) | 0.01 | 1.50 (1.04 to 2.17) | 0.03 | |||
| ≥41% eGFR decline | 80 | 4.18 (3.26 to 5.34) | <0.001 | 3.76 (2.86 to 4.96) | <0.001 | 3.27 (2.05 to 5.20) | <0.001 | |||
| Volume status exposure | — | — | — | |||||||
| Congestion improved | 345 | Ref. | — | Ref. | — | Ref. | — | |||
| Congestion worsened | 395 | 2.79 (2.42 to 3.21) | <0.001 | 2.86 (2.44 to 3.35) | <0.001 | 2.16 (1.79 to 2.60) | <0.001 | |||
| Outcome: CV Mortality | No. of Events | Model 1 | Model 2 | Model 3 | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| HR (95% CI) | P Value | P for Trend | HR (95% CI) | P Value | P for Trend | HR (95% CI) | P Value | P for Trend | ||
| Kidney function exposure | <0.001 | <0.001 | <0.001 | |||||||
| No eGFR decline | 214 | Ref. | — | Ref. | — | Ref. | — | |||
| 1%–20% eGFR decline | 193 | 1.09 (0.90 to 1.31) | 0.40 | 1.13 (0.91 to 1.40) | 0.26 | 1.20 (0.90 to 1.61) | 0.22 | |||
| 21%–40% eGFR decline | 93 | 1.64 (1.29 to 2.08) | <0.001 | 1.58 (1.20 to 2.08) | 0.001 | 1.69 (1.09 to 2.60) | 0.02 | |||
| ≥41% eGFR decline | 59 | 4.21 (3.15 to 5.62) | <0.001 | 3.98 (2.88 to 5.51) | <0.001 | 3.04 (1.65 to 5.60) | <0.001 | |||
| Volume status exposure | — | — | — | |||||||
| Congestion improved | 246 | Ref. | — | Ref. | — | Ref. | — | |||
| Congestion worsened | 313 | 3.02 (2.57 to 3.55) | <0.001 | 3.12 (2.60 to 3.75) | <0.001 | 2.37 (1.92 to 2.94) | <0.001 | |||
| Outcome: Composite of CV Mortality or Heart Failure Rehospitalization | No. of Events | Model 1 | Model 2 | Model 3 | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| HR (95% CI) | P Value | P for Trend | HR (95% CI) | P Value | P for Trend | HR (95% CI) | P Value | P for Trend | ||
| Kidney function exposure | <0.001 | <0.001 | 0.001 | |||||||
| No eGFR decline | 520 | Ref. | — | Ref. | — | Ref. | — | |||
| 1%–20% eGFR decline | 515 | 1.15 (1.01 to 1.30) | 0.03 | 1.24 (1.08 to 1.42) | 0.002 | 1.13 (0.95 to 1.35) | 0.17 | |||
| 21%–40% eGFR decline | 203 | 1.87 (1.58 to 2.20) | <0.001 | 1.76 (1.46 to 2.12) | <0.001 | 1.54 (1.18 to 2.02) | 0.002 | |||
| ≥41% eGFR decline | 65 | 2.76 (2.12 to 3.60) | <0.001 | 2.45 (1.84 to 3.26) | <0.001 | 1.74 (1.05 to 2.87) | 0.03 | |||
| Volume status exposure | — | — | — | |||||||
| Congestion improved | 636 | Ref. | — | Ref. | — | Ref. | — | |||
| Congestion worsened | 667 | 3.27 (2.92 to 3.66) | <0.001 | 3.23 (2.85 to 3.66) | <0.001 | 2.87 (2.50 to 3.29) | <0.001 | |||
Model 1: unweighted, unadjusted time-dependent model. CI, confidence interval; CV, cardiovascular; HR, hazard ratio; Ref., reference.
Model 2: unweighted time-dependent model with regression adjustment for baseline covariates: age, sex race, smoking status, hypertension, diabetes mellitus, dyslipidemia, ischemic heart disease, history of heart failure hospitalization, percutaneous coronary intervention, coronary artery bypass grafting, and stroke, variables at hospital discharge (clinical congestion score, weight, body mass index, systolic BP, heart rate, B-type natriuretic peptide, eGFR, hematocrit, serum albumin, sodium, potassium, angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers, β blockers, and diuretic agents), and variables at randomization (left ventricular ejection fraction, New York Heart Association classification, B-type natriuretic peptide, loop diuretic dose, and randomization group [tolvaptan]).
Model 3 (marginal structural model): inverse probability weight-weighted time-dependent model with further regression adjustment for the same covariates as model 2. Inverse probability weight was estimated at each patient visit using baseline and time-varying covariates.
Figure 3.
Association of the eight congestion-eGFR grouping with all-cause mortality in the fully adjusted MSM. HRs were estimated by IPW-weighted pooled logistic regression model. IPW was estimated at each patient visit using baseline and time-varying covariates. The IPW-weighted model included further regression adjustment for baseline covariates: age, sex race, smoking status, hypertension, DM, dyslipidemia, IHD, history of heart failure hospitalization, PCI, CABG, and stroke, variables at hospital discharge (CCS, weight, BMI, SBP, heart rate, BNP, eGFR, hematocrit, serum albumin, sodium, potassium, ACEIs/ARBs, β blockers, and diuretic agents), and variables at randomization (LVEF, the NYHA classification, BNP, loop diuretic dose, and randomization group [tolvaptan]). ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker; BMI, body mass index; BNP, B-type natriuretic peptide; CABG, coronary artery bypass grafting; CI, confidence interval; DM, diabetes mellitus; HR, hazard ratio; IHD, ischemic heart disease; LVEF, left ventricular ejection fraction; NYHA, New York Heart Association; PCI, percutaneous coronary intervention; SBP, systolic BP.
Table 3.
Association of the eight congestion-eGFR grouping with all-cause mortality
| Outcome: All-Cause Mortality | Event, n | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|---|
| HR (95% CI) | P Value | HR (95% CI) | P Value | HR (95% CI) | P Value | ||
| Congestion improved | |||||||
| No eGFR decline | 147 | 0.38 (0.30 to 0.47) | <0.001 | 0.42 (0.33 to 0.54) | <0.001 | 0.51 (0.35 to 0.74) | <0.001 |
| 1%–20% eGFR decline | 118 | 0.38 (0.30 to 0.47) | <0.001 | 0.39 (0.30 to 0.51) | <0.001 | 0.56 (0.38 to 0.85) | 0.01 |
| 21%–40% eGFR decline | 54 | 0.55 (0.40 to 0.75) | <0.001 | 0.52 (0.37 to 0.74) | <0.001 | 0.80 (0.46–1.39) | 0.42 |
| ≥41% eGFR decline | 26 | 1.36 (0.93 to 2.00) | 0.12 | 1.04 (0.67 to 1.60) | 0.87 | 2.23 (1.06 to 4.66) | 0.03 |
| Congestion worsened | |||||||
| No eGFR decline | 142 | Ref. | — | Ref. | — | Ref. | — |
| 1–20% eGFR decline | 133 | 1.04 (0.82 to 1.31) | 0.75 | 1.19 (0.92 to 1.53) | 0.18 | 1.24 (0.83 to 1.84) | 0.29 |
| 21–40% eGFR decline | 66 | 1.32 (0.98 to 1.76) | 0.07 | 1.36 (0.98 to 1.89) | 0.06 | 1.92 (1.18 to 3.15) | 0.01 |
| ≥41% eGFR decline | 54 | 3.81 (2.74 to 5.29) | <0.001 | 4.16 (2.92 to 5.93) | <0.001 | 3.25 (1.81 to 5.83) | <0.001 |
Model 1: unweighted, unadjusted time-dependent model. CI, confidence interval; HR, hazard ratio; Ref., reference.
Model 2: unweighted time-dependent model with regression adjustment for baseline covariates: age, sex race, smoking status, hypertension, diabetes mellitus, dyslipidemia, ischemic heart disease, history of heart failure hospitalization, percutaneous coronary intervention, coronary artery bypass grafting, and stroke, variables at hospital discharge (clinical congestion score, weight, body mass index, systolic BP, heart rate, B-type natriuretic peptide, eGFR, hematocrit, serum albumin, sodium, potassium, angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers, β blockers, and diuretic agents), and variables at randomization (left ventricular ejection fraction, New York Heart Association classification, B-type natriuretic peptide, loop diuretic dose, and randomization group [tolvaptan]).
Model 3 (marginal structural model): inverse probability weight-weighted time-dependent model with further regression adjustment for the same covariates as model 2. Inverse probability weight was estimated using baseline and time-varying covariates.
There were no significant effect modifications by CCS at discharge (P for interaction = 0.22) or randomization group (P for interaction = 0.26; Supplemental Figures 2 and 3). The association was generally consistent when the outcome was CV mortality or the composite of CV mortality or heart failure rehospitalization (Table 2 and Supplemental Figures 4 and 5).
The results of sensitivity analyses showed consistent relations with the primary analyses (Supplemental Table 3). When redefining the volume status changes using BNP, patients in the congestion improved groups were at significantly lower risk of all-cause mortality up to an eGFR decline of 40% in comparison with those in the congestion worsened with no eGFR decline group. In addition, redefining eight exposure groups using alternate degrees of eGFR decline did not change the results substantially.
Discussion
This study examined and compared the associations of postdischarge changes in kidney function and changes in volume status with heart failure outcomes among patients who survived ADHF hospitalization in EVEREST. After adjustment for baseline and time-dependent confounders in the IPW-weighted model (MSM), both higher degree of eGFR decline and worsened congestion after discharge were associated with higher risk of all-cause mortality. The principal finding of this study was that patients in the congestion improved groups were at lower risk of all-cause mortality up to an eGFR decline of 20% but were at higher risk when eGFR decline exceeds 40% compared with those in the congestion worsened with no eGFR decline group.
Volume overload is an accepted risk factor associated with poor outcomes in patients with chronic heart failure.1,4 Even among discharged patients with heart failure, in whom volume status is assumed better-controlled than in in-hospital patients with ADHF, more than 75% were reported to have residual congestion, which was associated with increased risks of mortality and re-hospitalization for heart failure.4,30 These data suggest a rationale for achieving and sustaining decongestion in this population. However, adequate decongestion may not be achieved because of clinicians' concern that aggressive decongestion leads to eGFR decline,31 another well-known heart failure risk factor,1,5,6 by reducing intravascular volume and activating the renin-angiotensin-aldosterone and sympathetic nervous systems.32 Consequently, discharged patients, who are at particularly high risk of adverse heart failure events,2 are often left undertreated in terms of volume control so as to preserve kidney function. Under these circumstances, providing clinical evidence on whether preserving kidney function at the expense of further decongestion or sustaining aggressive decongestion should be prioritized remains a major unmet need.
There is little evidence on a desirable balance between changes in kidney function and changes in volume status in terms of outcome risks in the management of outpatients with heart failure. A previous study of 151 discharged patients showed the lowest CV event rates in patients with baseline N-terminal proBNP values of <1000 pg/ml and worsened kidney function, suggesting a benefit of decongestion regardless of kidney function.33 However, neither longitudinal natriuretic peptide level changes nor eGFR changes were considered, and the findings were limited because of the small sample size. When analyzing the association of changes in eGFR and volume status with outcomes, several issues need to be addressed. First, in conventional survival analyses using a time-fixed exposure, defining changes in eGFR and volume status as time-fixed exposure variables requires an exposure-defining period before time at risk, resulting in the loss of patients who had an outcome event during this period and thus the creation of immortal bias. Given the high rates of 30-day mortality among patients with heart failure just after discharge,34,35 excluding this large number of high-risk patients from the analysis can be an important limitation. Second, although time-dependent analyses can overcome this limitation, time-dependent Cox regression models produce biased estimates in the presence of time-dependent confounding.17,18 By making the most of the large repeated measures of eGFR and CCS (or BNP) in EVEREST and employing MSMs in time-dependent analyses, this study attempted to address these concerns.
We confirmed that a higher degree of eGFR decline and worsened congestion after discharge was associated with a higher risk of mortality. Patients grouped into the congestion improved with no eGFR decline showed the lowest risk, whereas those grouped as the worsened congestion with 41% or greater eGFR decline showed the highest risk. Importantly, patients grouped into the congestion improved had a lower mortality rate up to an eGFR decline of 20%, compared with those grouped as the congestion worsened with no eGFR decline. This lower risk extended to an eGFR decline of 40% when volume status was defined using BNP. These findings suggest that prioritizing sustaining decongestion over preserving kidney function is preferable up to an eGFR decline of 20% and likely extends to approximately 40% among recently discharged patients with heart failure. Although these findings cannot be directly applied to clinical practice because of the observational nature of the study, they may encourage clinicians to consider more aggressive decongestion strategies in appropriate patients, potentially reducing the number of discharged patients with heart failure who remain undertreated with respect to volume control. Importantly, our data also underscore the importance of close monitoring of eGFR during decongestion therapy because substantial declines may be associated with harm regardless of volume status.
To the best of our knowledge, this study is the first to account for the associations of changes in kidney function and changes in volume status over time with outcomes in outpatients with heart failure. The strengths of this study included that the analysis was performed by making use of the large repeated measures of eGFR and CCS in EVEREST at prespecified monthly intervals over the course of follow-up. To correctly estimate these associations, we performed extensive adjustment for various potential baseline and time-dependent confounders, employing the MSM. We confirmed the robustness of the findings through sensitivity analyses. The study analyzed a large sample size (N=3404) from EVEREST, where outcome events of interest were adjudicated. The present findings provide the clinically important information about volume control in recently discharged patients with heart failure, who are particularly vulnerable2 and whose care represents a major economic burden.3
This study has limitations. First, as with any observed association, causality cannot be assumed, despite the pathophysiologic rationale. Despite extensive covariate adjustment, residual confounding by measured and unmeasured covariates may remain. For example, our analysis did not account for the confounding effects of medical adherence, frailty, or changes in loop diuretic dosing, all of which may have influenced the results. Second, although CCS has been used in several other studies,13–15 it has not been validated, and change in CCS is a subjective measure of volume status evaluated by clinicians. However, a sensitivity analysis using BNP as an objective alternate measure of volume status showed consistent and perhaps even stronger results with regard to the primacy of decongestion. Third, although the MSM theoretically estimates the same effect size of an exposure on an outcome in a pseudo-population as in a real-world population,22 our observed associations can be exaggerated because the designation of patients into each of the eight exposure categories is fixed on the basis of IPW from each study visit. For example, patients designated by IPW into the congestion improved with no eGFR decline group are being analyzed as having the congestion improved with no eGFR decline at each follow-up visit (Figure 2). Fourth, the present threshold analysis was exploratory and the statistical power may have been limited. Larger studies are warranted to explore a threshold of eGFR decrease beyond which is associated with adverse heart failure outcomes, irrespective of decongestion. Fifth, EVEREST was conducted nearly two decades ago, and heart failure management has evolved to include newer therapies such as Sodium–glucose cotransporter 2 inhibitors and angiotensin receptor-neprilysin inhibitors. The principal approach to decongestion—the use of diuretic agents—has not however changed substantially in outpatients with heart failure, and clinicians still face the question about the optimal balance between sustaining decongestion and kidney function preservation, and we have no a priori reason to suspect that these relations would differ in the current era. Sixth, data on the outcome of ESKD were not available in EVEREST. Finally, because EVEREST enrolled hospitalized patients with heart failure with reduced LVEF, the generalizability of the present findings to other heart failure populations is uncertain.
In summary, a higher degree of eGFR decline and worsened congestion were both associated with a higher risk of all-cause mortality among patients who survived heart failure hospitalization. Patients in the congestion improved groups generally had a lower risk of mortality but a higher risk when eGFR decline exceeds 40%, compared with the congestion worsened with no eGFR decline group. This suggests the importance of sustaining congestion even with moderate declines in eGFR among discharged patients with heart failure. Although our analysis suggests that prioritizing sustaining decongestion likely extends up to an eGFR decline of approximately 40%, further studies in larger external cohorts are needed to validate the relation of potential thresholds of eGFR decline in the context of decongestion with clinically relevant outcomes.
Supplementary Material
Acknowledgments
We are grateful to all participants involved in the present study.
Footnotes
See related editorial, “Probing the Limits of Permissive Hypercreatininemia during Decongestion in Patients with Heart Failure,” on pages 1174–1176.
Disclosures
Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/CJN/C336.
Author Contributions
Conceptualization: Tatsufumi Oka, Mark J. Sarnak.
Data curation: Hocine Tighiouart.
Formal analysis: Tatsufumi Oka, Hocine Tighiouart.
Investigation: Tatsufumi Oka, Mark J. Sarnak.
Methodology: Tatsufumi Oka, Mark J. Sarnak, Hocine Tighiouart.
Project administration: Marvin A. Konstam, James E. Udelson.
Resources: Marvin A. Konstam, Wendy McCallum, James E. Udelson.
Software: Tatsufumi Oka, Hocine Tighiouart.
Supervision: Mark J. Sarnak.
Validation: Yoshitaka Isaka, Marvin A. Konstam, Wendy McCallum, Tatsufumi Oka, Mark J. Sarnak, Marcelle Tuttle.
Visualization: Tatsufumi Oka.
Writing – original draft: Tatsufumi Oka.
Writing – review & editing: Yoshitaka Isaka, Marvin A. Konstam, Wendy McCallum, Mark J. Sarnak, Hocine Tighiouart, Marcelle Tuttle, James E. Udelson.
Funding
None.
Data Availability Statements
Partial restrictions to the data and/or materials apply. The data used in this study were from an industry funded trial and are not publicly available. Proposals to collaborate or share data will be reviewed by the principal investigator of the trial.
Supplemental Material
This article contains the following supplemental material online at http://links.lww.com/CJN/C337.
Supplemental Methods. GFR estimation using the 2021 CKD Epidemiology Collaboration formula.
Supplemental Table 1. Comparison of characteristics at discharge between included and excluded participants among those who were followed up after discharge.
Supplemental Table 2. Characteristics of the weighted population at 24 weeks after discharge.
Supplemental Table 3. Associations of the eight congestion-eGFR grouping with all-cause mortality in sensitivity analyses.
Supplemental Figure 1. A causal chain relation between time-dependent exposure and confounder.
Supplemental Figure 2. Subgroup analysis by CCS at discharge in association of the eight congestion-eGFR grouping with all-cause mortality.
Supplemental Figure 3. Subgroup analysis by randomization group in association of the eight congestion-eGFR grouping with all-cause mortality.
Supplemental Figure 4. Association of the eight congestion-eGFR grouping with CV mortality in the fully adjusted MSM.
Supplemental Figure 5. Association of the eight congestion-eGFR grouping with the composite of CV mortality or re-hospitalization for HF in the fully adjusted MSM.
References
- 1.McDonagh TA Metra M Adamo M, et al.; ESC Scientific Document Group. 2021 ESC guidelines for the diagnosis and treatment of acute and chronic heart failure. Eur Heart J. 2021;42(36):3599–3726. doi: 10.1093/eurheartj/ehab368 [DOI] [PubMed] [Google Scholar]
- 2.Solomon SD Dobson J Pocock S, et al.; Candesartan in Heart failure: Assessment of Reduction in Mortality and morbidity (CHARM) Investigators. Influence of nonfatal hospitalization for heart failure on subsequent mortality in patients with chronic heart failure. Circulation. 2007;116(13):1482–1487. doi: 10.1161/CIRCULATIONAHA.107.696906 [DOI] [PubMed] [Google Scholar]
- 3.McIlvennan CK, Eapen ZJ, Allen LA. Hospital readmissions reduction program. Circulation. 2015;131(20):1796–1803. doi: 10.1161/CIRCULATIONAHA.114.010270 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Rubio-Gracia J Demissei BG ter Maaten JM, et al. Prevalence, predictors and clinical outcome of residual congestion in acute decompensated heart failure. Int J Cardiol. 2018;258:185–191. doi: 10.1016/J.IJCARD.2018.01.067 [DOI] [PubMed] [Google Scholar]
- 5.Damman K, Valente MAE, Voors AA, O’Connor CM, van Veldhuisen DJ, Hillege HL. Renal impairment, worsening renal function, and outcome in patients with heart failure: an updated meta-analysis. Eur Heart J. 2014;35(7):455–469. doi: 10.1093/eurheartj/eht386 [DOI] [PubMed] [Google Scholar]
- 6.Dries DL, Exner DV, Domanski MJ, Greenberg B, Stevenson LW. The prognostic implications of renal insufficiency in asymptomatic and symptomatic patients with left ventricular systolic dysfunction. J Am Coll Cardiol. 2000;35(3):681–689. doi: 10.1016/s0735-1097(99)00608-7 [DOI] [PubMed] [Google Scholar]
- 7.Metra M Cotter G Senger S, et al. Prognostic significance of creatinine increases during an acute heart failure admission in patients with and without residual congestion: a post hoc analysis of the PROTECT data. Circ Heart Fail. 2018;11(5):e004644. doi: 10.1161/CIRCHEARTFAILURE.117.004644 [DOI] [PubMed] [Google Scholar]
- 8.McCallum W Tighiouart H Testani JM, et al. Acute kidney function declines in the context of decongestion in acute decompensated heart failure. JACC Heart Fail. 2020;8(7):537–547. doi: 10.1016/J.JCHF.2020.03.009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Konstam MA Gheorghiade M Burnett JC, et al.; Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study With Tolvaptan (EVEREST) Investigators. Effects of oral tolvaptan in patients hospitalized for worsening heart failure: the EVEREST outcome trial. JAMA. 2007;297(12):1319–1331. doi: 10.1001/jama.297.12.1319 [DOI] [PubMed] [Google Scholar]
- 10.Konstam MA Kiernan M Chandler A, et al.; SECRET of CHF Investigators, Coordinators, and Committee Members. Short-term effects of tolvaptan in patients with acute heart failure and volume overload. J Am Coll Cardiol. 2017;69(11):1409–1419. doi: 10.1016/j.jacc.2016.12.035 [DOI] [PubMed] [Google Scholar]
- 11.Gheorghiade M Orlandi C Burnett JC, et al. Rationale and design of the multicenter, randomized, double-blind, placebo-controlled study to evaluate the efficacy of vasopressin antagonism in heart failure: outcome study with tolvaptan (EVEREST). J Card Fail. 2005;11(4):260–269. doi: 10.1016/J.CARDFAIL.2005.03.009 [DOI] [PubMed] [Google Scholar]
- 12.Inker LA Eneanya ND Coresh J, et al.; Chronic Kidney Disease Epidemiology Collaboration. New Creatinine- and cystatin C-Based equations to estimate GFR without race. N Engl J Med. 2021;385(19):1737–1749. doi: 10.1056/NEJMOA2102953 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Ambrosy AP Pang PS Khan S, et al.; EVEREST Trial Investigators. Clinical course and predictive value of congestion during hospitalization in patients admitted for worsening signs and symptoms of heart failure with reduced ejection fraction: findings from the EVEREST trial. Eur Heart J. 2013;34(11):835–843. doi: 10.1093/EURHEARTJ/EHS444 [DOI] [PubMed] [Google Scholar]
- 14.McCallum W Tighiouart H Testani JM, et al. Association of volume overload with kidney function outcomes among patients with heart failure with reduced ejection fraction. Kidney Int Rep. 2020;5(10):1661–1669. doi: 10.1016/j.ekir.2020.07.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.McCallum W Tighiouart H Testani JM, et al. Rates of reversal of volume overload in hospitalized acute heart failure: association with long-term kidney function. Am J Kidney Dis. 2022;80(1):65–78. doi: 10.1053/j.ajkd.2021.09.026 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Thibodeau JT, Drazner MH. The role of the clinical examination in patients with heart failure. JACC Heart Fail. 2018;6(7):543–551. doi: 10.1016/j.jchf.2018.04.005 [DOI] [PubMed] [Google Scholar]
- 17.Mansournia MA, Etminan M, Danaei G, Kaufman JS, Collins G. Handling time varying confounding in observational research. BMJ. 2017;359:j4587. doi: 10.1136/bmj.j4587 [DOI] [PubMed] [Google Scholar]
- 18.Oka T Sakaguchi Y Hattori K, et al. Mineralocorticoid receptor antagonist use and hard renal outcomes in real-world patients with chronic kidney disease. Hypertension. 2022;79(3):679–689. doi: 10.1161/HYPERTENSIONAHA.121.18360 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Oka T Sakaguchi Y Hattori K, et al. Association of longitudinal B-type natriuretic peptide monitoring with kidney failure in patients with CKD: a cohort study. Am J Kidney Dis. 2023;82(5):559–568. doi: 10.1053/J.AJKD.2023.05.003 [DOI] [PubMed] [Google Scholar]
- 20.Fewell Z, Hernán MA, Wolfe F, Tilling K, Choi H, Sterne JAC. Controlling for time-dependent confounding using marginal structural models. Stata J. 2004;4:402–420. doi: 10.1177/1536867X0400400403 [DOI] [Google Scholar]
- 21.Hernán MA, Brumback B, Robins JM. Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-Positive men. Epidemiology. 2000;11(5):561–570. doi: 10.1097/00001648-200009000-00012 [DOI] [PubMed] [Google Scholar]
- 22.Toh S, Hernán MA. Causal inference from longitudinal studies with baseline randomization. Int J Biostat. 2008;4(1):Article 22. doi: 10.2202/1557-4679.1117 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Xie D Yang W Jepson C, et al.; Chronic Renal Insufficiency Cohort (CRIC) Study Investigators. Statistical methods for modeling time-updated exposures in cohort studies of chronic kidney disease. Clin J Am Soc Nephrol. 2017;12(11):1892–1899. doi: 10.2215/CJN.00650117 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Asahina Y Sakaguchi Y Kajimoto S, et al. Association of time-updated anion gap with risk of kidney failure in advanced CKD: a cohort study. Am J Kidney Dis. 2022;79(3):374–382. doi: 10.1053/j.ajkd.2021.05.022 [DOI] [PubMed] [Google Scholar]
- 25.Asahina Y Sakaguchi Y Kajimoto S, et al. Time-updated anion gap and cardiovascular events in advanced chronic kidney disease: a cohort study. Clin Kidney J. 2022;15(5):929–936. doi: 10.1093/ckj/sfab277 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Lukowsky LR, Mehrotra R, Kheifets L, Arah OA, Nissenson AR, Kalantar-Zadeh K. Comparing mortality of peritoneal and hemodialysis patients in the first 2 years of dialysis therapy: a marginal structural model analysis. Clin J Am Soc Nephrol. 2013;8(4):619–628. doi: 10.2215/CJN.04810512 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Daniels LB, Maisel AS. Natriuretic peptides. J Am Coll Cardiol. 2007;50(25):2357–2368. doi: 10.1016/j.jacc.2007.09.021 [DOI] [PubMed] [Google Scholar]
- 28.Royston P. Multiple imputation of missing values. Stata J Promoting Commun Stat Stata. 2004;4(3):227–241. doi: 10.1177/1536867x0400400301 [DOI] [Google Scholar]
- 29.Bansal N Xie D Tao K, et al.; CRIC Study. Atrial fibrillation and risk of ESRD in adults with CKD. Clin J Am Soc Nephrol. 2016;11(7):1189–1196. doi: 10.2215/CJN.10921015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Boorsma EM ter Maaten JM Damman K, et al. Congestion in heart failure: a contemporary look at physiology, diagnosis and treatment. Nat Rev Cardiol. 2020;17(10):641–655. doi: 10.1038/S41569-020-0379-7 [DOI] [PubMed] [Google Scholar]
- 31.de Silva R Nikitin NP Witte KKA, et al. Incidence of renal dysfunction over 6 months in patients with chronic heart failure due to left ventricular systolic dysfunction: contributing factors and relationship to prognosis. Eur Heart J. 2006;27(5):569–581. doi: 10.1093/eurheartj/ehi696 [DOI] [PubMed] [Google Scholar]
- 32.Wang DJ, Gottlieb SS. Diuretics: still the mainstay of treatment. Crit Care Med. 2008;36(1 suppl):S89–S94. doi: 10.1097/01.CCM.0000296272.68078.6B [DOI] [PubMed] [Google Scholar]
- 33.Ibrahim NE, Gaggin HK, Rabideau DJ, Gandhi PU, Mallick A, Januzzi JL. Worsening renal function during management for chronic heart failure with reduced ejection fraction: results from the Pro-BNP outpatient tailored chronic heart failure therapy (PROTECT) study. J Card Fail. 2017;23(2):121–130. doi: 10.1016/J.CARDFAIL.2016.07.440 [DOI] [PubMed] [Google Scholar]
- 34.Goldgrab D, Balakumaran K, Kim MJ, Tabtabai SR. Updates in heart failure 30-day readmission prevention. Heart Fail Rev. 2019;24(2):177–187. doi: 10.1007/S10741-018-9754-4 [DOI] [PubMed] [Google Scholar]
- 35.Dharmarajan K Wang Y Lin Z, et al. Association of changing hospital readmission rates with mortality rates after hospital discharge. JAMA. 2017;318(3):270–278. doi: 10.1001/JAMA.2017.8444 [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.
Data Availability Statement
Partial restrictions to the data and/or materials apply. The data used in this study were from an industry funded trial and are not publicly available. Proposals to collaborate or share data will be reviewed by the principal investigator of the trial.