Extracellular fluid volume is associated with incident end-stage kidney disease and mortality in patients with chronic kidney disease

Volume overload has been shown to be an independent risk factor for mortality in patients receiving chronic dialysis, but data in non-dialysis patients are scarce. Therefore we evaluated the prognostic value of extracellular fluid (ECF) volume for chronic kidney disease (CKD) progression and mortality in a prospective hospital-based cohort with CKD stage 1-4 (NephroTest Study). ECF (scaled to body surface area) and the measured glomerular filtration rate (mGFR) were determined using the distribution volume and clearance of 51Cr-EDTA, respectively. Cause-specific Cox and linear mixed-effect regression models were used to analyze the association of ECF with end-stage kidney disease (ESKD) and mortality, and with mGFR decline, respectively. The 1593 patients were mean age 58.8 years, 67% were men, mean mGFR of 43.6 mL/min/1.73m2 and mean ECF 15.1 L/1.73m2. After a median follow-up of 5.3 years, ESKD occurred in 324 patients and 185 patients died before ESKD. In multivariable analysis, ECF was significantly associated with the risk of ESKD (hazard ratio per 1L/1.73m2 increase: 1.14; 95% confidence interval [1.07; 1.21]) and with a faster GFR decline (adjusted mean difference in mGFR slope per 1L/1.73m2 increase -0.14 [-0.23; -0.05] mL/min/year). The relationship of ECF with mortality was non-linear and not significant (per 1L/1.73m2 increase 0.92, [0.73; 1.16]), below 15L/1.73m2, but significant (1.28; [1.14-1.45]) above 15L/1.73m2. Thus, in this large cohort of carefully phenotyped patients with CKD, ECF was an independent risk factor of CKD progression and mortality. Hence, close monitoring and treatment of fluid overload are important for the clinical management of patients with non-dialysis CKD. Volume overload has been shown to be an independent risk factor for mortality in patients receiving chronic dialysis, but data in non-dialysis patients are scarce. Therefore we evaluated the prognostic value of extracellular fluid (ECF) volume for chronic kidney disease (CKD) progression and mortality in a prospective hospital-based cohort with CKD stage 1-4 (NephroTest Study). ECF (scaled to body surface area) and the measured glomerular filtration rate (mGFR) were determined using the distribution volume and clearance of 51Cr-EDTA, respectively. Cause-specific Cox and linear mixed-effect regression models were used to analyze the association of ECF with end-stage kidney disease (ESKD) and mortality, and with mGFR decline, respectively. The 1593 patients were mean age 58.8 years, 67% were men, mean mGFR of 43.6 mL/min/1.73m2 and mean ECF 15.1 L/1.73m2. After a median follow-up of 5.3 years, ESKD occurred in 324 patients and 185 patients died before ESKD. In multivariable analysis, ECF was significantly associated with the risk of ESKD (hazard ratio per 1L/1.73m2 increase: 1.14; 95% confidence interval [1.07; 1.21]) and with a faster GFR decline (adjusted mean difference in mGFR slope per 1L/1.73m2 increase -0.14 [-0.23; -0.05] mL/min/year). The relationship of ECF with mortality was non-linear and not significant (per 1L/1.73m2 increase 0.92, [0.73; 1.16]), below 15L/1.73m2, but significant (1.28; [1.14-1.45]) above 15L/1.73m2. Thus, in this large cohort of carefully phenotyped patients with CKD, ECF was an independent risk factor of CKD progression and mortality. Hence, close monitoring and treatment of fluid overload are important for the clinical management of patients with non-dialysis CKD. Impaired renal salt and water excretion, in combination with other factors such as hypo-albuminemia, often result in chronic ECF overload with CKD.1Ellison D.H. Treatment of disorders of sodium balance in chronic kidney disease.Adv Chronic Kidney Dis. 2017; 24: 332-341Abstract Full Text Full Text PDF PubMed Scopus (21) Google Scholar In hemodialysis patients, several large-scale studies have shown that fluid overload is a strong and independent risk factor for mortality.2Zoccali C. Moissl U. Chazot C. et al.Chronic fluid overload and mortality in ESRD.J Am Soc Nephrol. 2017; 28: 2491-2497Crossref PubMed Scopus (212) Google Scholar, 3Kalantar-Zadeh K. Regidor D.L. Kovesdy C.P. et al.Fluid retention is associated with cardiovascular mortality in patients undergoing long-term hemodialysis.Circulation. 2009; 119: 671-679Crossref PubMed Scopus (399) Google Scholar In contrast, studies evaluating the role of fluid overload on renal function and mortality in patients with non-dialysis CKD yielded conflicting results.4Bansal N. Zelnick L.R. Himmelfarb J. et al.Bioelectrical impedance analysis measures and clinical outcomes in CKD.Am J Kidney Dis. 2018; 72: 662-672Abstract Full Text Full Text PDF PubMed Scopus (25) Google Scholar, 5Tsai Y.C. Chiu Y.W. Tsai J.C. et al.Association of fluid overload with cardiovascular morbidity and all-cause mortality in stages 4 and 5 CKD.Clin J Am Soc Nephrol. 2015; 10: 39-46Crossref PubMed Scopus (94) Google Scholar, 6Tsai Y.C. Tsai J.C. Chen S.C. et al.Association of fluid overload with kidney disease progression in advanced CKD: a prospective cohort study.Am J Kidney Dis. 2014; 63: 68-75Abstract Full Text Full Text PDF PubMed Scopus (67) Google Scholar In addition, ECF was estimated using single- or multi-frequency bioelectrical impedance analysis, and not isotope dilution, which is the most direct and accurate method, although not routinely available.7Kyle U.G. Bosaeus I. De Lorenzo A.D. et al.Bioelectrical impedance analysis—part I: review of principles and methods.Clin Nutr. 2004; 23: 1226-1243Abstract Full Text Full Text PDF PubMed Scopus (1829) Google Scholar, 8Kyle U.G. Bosaeus I. De Lorenzo A.D. et al.Bioelectrical impedance analysis—part II: utilization in clinical practice.Clin Nutr. 2004; 23: 1430-1453Abstract Full Text Full Text PDF PubMed Scopus (1380) Google Scholar, 9Bird N.J. Peters C. Michell A.R. et al.Extracellular distribution volumes of hydrophilic solutes used to measure the glomerular filtration rate: comparison between chromium-51-EDTA and iohexol.Physiol Meas. 2007; 28: 223-234Crossref PubMed Scopus (5) Google Scholar, 10Ellis K.J. Human body composition: in vivo methods.Physiol Rev. 2000; 80: 649-680Crossref PubMed Scopus (594) Google Scholar Similarly, GFR was estimated and not measured by a reference method. Therefore, the relationship between ECF and renal outcome and mortality during CKD remains uncertain. The aim of this prospective observational study was to evaluate the association of ECF with progression of CKD, with ESKD, and with mortality occurring before ESKD, in non-dialysis CKD patients, using the “gold-standard” measurements of ECF and GFR. A total of 1593 patients with available ECF and CKD stage 1 to 4 from the prospective observational Nephrotest cohort were included in the study. Characteristics of the patients are reported in Table 1. Mean age was 58.8 ± 15.1 years; 66.7% were men; 87.8% had a history of hypertension; and 27.0% had diabetes. Mean mGFR was 43.6 ± 18.6 ml/min per 1.73 m2. Mean ECF was 15.1 ± 2.2 L/1.73 m2. Distribution of ECF, crude or scaled to body surface area (BSA), according to sex, is shown in Supplementary Figure S1. ECF, anthropometric parameters, and diuretic prescription, for GFR subgroups, are reported in Supplementary Figure S2. BSA and crude ECF decreased as GFR decreased, so the relationship between ECF and mGFR was attenuated when ECF was scaled to BSA: ECF decreased slightly from 15.7 to 14.5 L/1.73 m2 when GFR decreased from >60 ml/min to 15–30 ml/min. Loop diuretics increased from 16% in patients with GFR >60 ml/min to 43% in patients with GFR of 15–30 ml/min.Table 1Baseline clinical and biological characteristics in total population and according to tertiles of extracellular fluid volumeTotal n = 1593ECF (L/1.73 m2)P1st tertile n = 531 [8.88–14.03]2nd tertile n = 531 [14.04–15.82]3rd tertile n = 531 [15.83–23.29]Demographics and clinical characteristicsAge (yr)58.8 ± 15.153.2 ± 15.258.8 ± 14.364.2 ± 13.6<0.001Sex (men)1063 (66.7)249 (46.9)365 (68.7)449 (84.6)<0.001Ethnicity (Sub-Saharan African origin)217 (14.3)80 (15.8)80 (15.8)57 (11.2)0.053Height (cm)167.1 (9.4)164.6 (9.6)167.7 (9.4)169.1 (8.6)<0.001Weight (kg)74.6 (16.3)67.1 (13.7)74.4 (13.9)82.2 (17.3)<0.001Body surface area (m2)1.83 (0.22)1.73 (0.20)1.83 (0.20)1.92 (0.21)<0.001Body mass index (kg/m2)26.6 ± 5.124.7 ± 4.326.4 ± 4.428.7 ± 5.8<0.001Tobacco consumption<0.001 Nonsmoker873 (54.8)345 (65.0)284 (53.5)244 (46.0) Former smoker506 (31.8)113 (21.3)168 (31.6)225 (42.4) Current smoker214 (13.4)73 (13.7)79 (14.9)62 (11.7)Systolic blood pressure (mm Hg)135 ± 20.2130 ± 18.6135 ± 19.7141 ± 20.8<0.001Diastolic blood pressure (mm Hg)75 ± 11.574 ± 11.875 ± 11.976 ± 10.80.024Elevated blood pressure (≥140 and/or 90 mm Hg)573 (37.3)144 (28.3)187 (36.7)242 (46.8)<0.001Medical historyHypertension1399 (87.8)441 (83.1)459 (86.4)499 (94.0)<0.001Diabetes mellitus430 (27.0)77 (14.5)120 (22.6)233 (43.9)<0.001Dyslipidemia277 (18.0)110 (21.4)84 (16.3)83 (16.3)0.047Previous cardiovascular event288 (18.4)61 (11.7)81 (15.5)146 (28.1)<0.001Underlying renal disease<0.001 Diabetic nephropathy154 (9.7)18 (3.4)34 (6.4)102 (19.2) Glomerular224 (14.1)102 (19.2)69 (13.0)53 (10.0) Vascular410 (25.7)110 (20.7)146 (27.5)154 (29.0) Polycystic kidney disease100 (6.3)35 (6.6)39 (7.3)26 (4.9) Interstitial150 (9.4)69 (13.0)51 (9.6)30 (5.6) Other or unknown conditions555 (34.8)197 (37.1)192 (36.2)166 (31.3)TreatmentNumber of antihypertensive drugs2.3 ± 1.62.0 ± 1.42.3 ± 1.62.7 ± 1.6<0.001ACEi and/or ARB (%)1186 (74.5)378 (71.2)386 (72.7)422 (79.5)0.004Diuretics759 (47.7)209 (39.4)253 (47.6)297 (56.0)<0.001 Loop diuretic454 (28.5)114 (21.5)137 (25.8)203 (38.3)<0.001 Thiazide diuretic326 (20.5)96 (18.1)121 (22.8)109 (20.6)0.169 Amiloride17 (1.1)5 (0.9)6 (1.1)6 (1.1)0.943 Aldosterone antagonist43 (2.7)13 (2.5)19 (3.6)11 (2.1)0.291Statin693 (43.6)196 (37.0)217 (40.9)280 (52.8)<0.001Biological parameterseGFR CKD-EPI (ml/min per 1.73 m2)aCalculated using the CKD-EPI formula.45.9 ± 21.845.2 ± 23.246.1 ± 21.746.3 ± 20.40.668mGFR (ml/min per 1.73 m2)43.6 ± 18.640.6 ± 17.944.5 ± 18.545.6 ± 19.0<0.001mGFR (ml/min per 1.73 m2)<0.001 mGFR ≥60299 (18.8)77 (14.5)105 (19.8)117 (22.0) 45 ≤ mGFR < 60366 (23.0)108 (20.3)129 (24.3)129 (24.3) 30 ≤ mGFR < 45503 (31.6)168 (31.6)171 (32.2)164 (30.9) 15 ≤ mGFR < 30425 (26.7)178 (33.5)126 (23.7)121 (22.8)Measured ECF (L)16.1 ± 3.612.9 ± 2.015.8 ± 1.819.6 ± 3.1<0.001Measured ECF (L/1.73 m2)15.1 ± 2.212.8 ± 1.014.9 ± 0.517.6 ± 1.6<0.001Hemoglobin (g/dl)12.74 (1.60)12.68 (1.63)12.83 (1.59)12.70 (1.60)0.267Protein (g/L)70.2 ± 6.0)70.5 ± 6.370.5± 5.769.6 ± 6.00.026Albumin (g/L)39.5 ± 4.439.9 ± 4.140.0 ± 4.138.8 ± 4.8<0.00124-h urinary sodium excretion (mmol/24 h)155 ± 73.0145 ± 71.7153 ± 68.4168 ± 77.3<0.00124-h urinary potassium excretion (mmol/24 h)65 ± 2659 ± 22.965 ± 24.772 ± 29.1<0.00124-h urinary sodium/potassium ratio2.6 ± 1.52.7 ± 1.72.6 ± 1.52.6 ± 1.30.210Protein-to-creatinine ratio (mg/mmol)80.8 ± 144.172.3 ± 114.767.0 ± 113.9103.2 ± 188.8<0.001ACEi, angiotensin-converting enzyme inhibitors; ARB: angiotensin II receptor blocker; CKD-EPI, Chronic Kidney Disease–Epidemiology Collaboration; ECF, extracellular fluid volume; eGFR: estimated glomerular filtration rate; mGFR: measured glomerular filtration rate;Continuous data are expressed as mean ± SD; categorical data are expressed as n (%). Diabetes was either self-reported or defined as fasting glycemia ≥7 mmol/L or antidiabetic drug treatment. Previous cardiovascular event was defined as a history of stroke, ischemic heart disease (angioplasty, surgical coronary bypass, or myocardial infarction), or heart failure. Dyslipidemia was defined as total cholesterol >6 mmol/L or >5 mmol/L in case of a previous cardiovascular event.a Calculated using the CKD-EPI formula. Open table in a new tab ACEi, angiotensin-converting enzyme inhibitors; ARB: angiotensin II receptor blocker; CKD-EPI, Chronic Kidney Disease–Epidemiology Collaboration; ECF, extracellular fluid volume; eGFR: estimated glomerular filtration rate; mGFR: measured glomerular filtration rate; Continuous data are expressed as mean ± SD; categorical data are expressed as n (%). Diabetes was either self-reported or defined as fasting glycemia ≥7 mmol/L or antidiabetic drug treatment. Previous cardiovascular event was defined as a history of stroke, ischemic heart disease (angioplasty, surgical coronary bypass, or myocardial infarction), or heart failure. Dyslipidemia was defined as total cholesterol >6 mmol/L or >5 mmol/L in case of a previous cardiovascular event. Compared with patients in the first tertile of ECF, patients in the third tertile were older, more often men, more likely to have a history of hypertension and diabetes, had higher systolic and diastolic blood pressure (BP), body mass index, mGFR, and urinary protein-to-creatinine ratio (uPCR), a higher 24-hour urine sodium excretion, and a lower plasma albumin concentration (Table 1). Associations were similar in multivariable analysis (Supplementary Table S1). After a median follow-up of 5.3 (interquartile range: 3.0, 7.4) years, 324 (20.3%) patients reached ESKD, and 185 (11.6%) of those who did not reach ESKD died (67 [4.2%] from cardiovascular causes). Cumulative incidence of death at 5 years was significantly higher in the third tertile of ECF (3.8%, 95% confidence interval [CI] [2.3, 5.9]; 5.4%, 95% CI [3.5, 7.8], and 13.2%, 95% CI [10.2, 16.6], for the 1st, 2nd, and 3rd tertile, respectively, P < 0.001), with a similar pattern for cardiovascular mortality (Supplementary Figure S3), whereas no difference was observed across tertiles of ECF for ESKD occurrence (Figure 1). Penalized splines representing the relationship between ECF and adjusted hazard ratios (HRs) of death and ESKD are shown in Figure 2. The relationship between ECF and ESKD was linear (Figure 2, left panel). In contrast, the relationship between ECF and mortality before ESKD showed an inflection point, with an increasing risk observed for ECF values above 15 L/1.73 m2 (Figure 2, right panel). Cause-specific Cox regression models showed that ECF was significantly associated with ESKD after adjustment for confounders (adjusted HR per 1 L/1.73 m2 increase of ECF: 1.14, 95% CI [1.07, 1.21], P < 0.001), and with mortality before ESKD above a threshold of 15 L/1.73 m2 (adjusted HR per 1 L/1.73 m2 increase of ECF below 15 L/1.73 m2, 0.92, 95% CI [0.73, 1.16], P = 0.49, and above 15 L/1.73 m2, 1.28, 95% CI [1.14, 1.45], P < 0.001) (Table 2). When the model was adjusted for systolic BP as a time-varying covariate, the association between ECF and ESKD was slightly weaker (HR per 1 L/1.73 m2 increase of ECF: 1.12, 95% CI [1.05, 1.19], P < 0.001 (Table 2). E-values were 1.42 for the association between ECF and ESKD, and 1.88 for the association between ECF and mortality before ESKD. Associations between ECF and the risks of ESKD or mortality did not depend on uPCR, BP, diabetes, age, sex, or mGFR (P values for interaction tests nonsignificant). Similar results were observed using ECF categorized in tertiles (Supplementary Figure S4), stratification for baseline estimated glomerular filtration rate instead of baseline mGFR (Supplementary Table S2), and multiple imputations for missing data (Supplementary Table S2). Similar associations were also observed when ECF was expressed as a percentage of body weight (Supplementary Table S3).Figure 2Estimated adjusted hazard ratios and 95% confidence intervals for the association of ECF with end-stage kidney disease (ESKD) and with mortality, using penalized-splines estimators. Cause-specific Cox regression models were adjusted for the following covariates: age, sex, site of inclusion, ethnicity, body mass index, diabetic status (no diabetes, diabetes without diabetic nephropathy, diabetes with diabetic nephropathy), elevated blood pressure, urinary protein-to-creatinine ratio (log-transformed), 24-hour urinary sodium excretion, diuretics, and renin–angiotensin system inhibitors. For mortality, models were also adjusted for previous cardiovascular events (myocardial infarction or angioplasty or stroke or heart failure) and plasma albumin concentration. Models were stratified for baseline measured glomerular filtration rate. Single imputations were used for missing data. ECF, extracellular fluid volume, scaled to body surface area (L/1.73 m2).View Large Image Figure ViewerDownload Hi-res image Download (PPT)Table 2Cause-specific Cox regression models: effects of 1 L/1.73 m2 increase in extracellular fluid volume on end-stage kidney disease and mortalityOutcomeModelsHR [95% CI]P valueESKDn = 1593Events (n)324Crude model1.03 [0.98, 1.08]0.29Adjusted model (1)aWhen the model was fully adjusted, but neither stratified nor adjusted for measured glomerular filtration rate, HR for ESKD was 0.98 (95% CI [0.92, 1.04], P = 0.44, showing that glomerular filtration rate is the covariate explaining that the crude analysis reveals no significant association between extracellular fluid volume and ESKD.1.14 [1.07, 1.21]<0.001Adjusted model (2)aWhen the model was fully adjusted, but neither stratified nor adjusted for measured glomerular filtration rate, HR for ESKD was 0.98 (95% CI [0.92, 1.04], P = 0.44, showing that glomerular filtration rate is the covariate explaining that the crude analysis reveals no significant association between extracellular fluid volume and ESKD.1.10 [1.04, 1.17]0.001Adjusted model (3)1.12 [1.05, 1.19]<0.001Mortality≤15 L/1.73 m2Events (n)63(n = 827)Crude1.00 [0.81, 1.24]0.99Adjusted model0.92 [0.73, 1.16]0.49>15 L/1.73 m2Events (n)122(n = 766)Crude model1.27 [1.16, 1.39]<0.001Adjusted model (1)1.28 [1.14, 1.45]<0.001Adjusted model (2)1.28 [1.14, 1.44]<0.001Adjusted model (3)1.29 [1.15, 1.46]<0.001CI, confidence interval; ESKD, end-stage kidney disease; HR, hazard ratio.Crude and adjusted HRs are indicated for 1 L/1.73 m2 increase in extracellular fluid volume. Analyses were adjusted for the following covariates: age, sex, site of inclusion, ethnicity, body mass index, diabetic status (no diabetes, diabetes without diabetic nephropathy, diabetes with diabetic nephropathy), elevated blood pressure, urinary protein-to-creatinine ratio (log-transformed), 24-hour urinary sodium excretion, diuretics, and renin–angiotensin system inhibitors. For mortality, models were also adjusted for previous cardiovascular events (myocardial infarction or angioplasty or stroke or heart failure) and plasma albumin concentration. Single imputations were used for missing data. Model (1) was stratified for baseline measured glomerular filtration rate. Model (2) was adjusted for measured glomerular filtration rate (expressed as a continuous variable, as a time-dependent coefficient). Model (3) was adjusted for systolic blood pressure as a continuous time-varying covariate.a When the model was fully adjusted, but neither stratified nor adjusted for measured glomerular filtration rate, HR for ESKD was 0.98 (95% CI [0.92, 1.04], P = 0.44, showing that glomerular filtration rate is the covariate explaining that the crude analysis reveals no significant association between extracellular fluid volume and ESKD. Open table in a new tab CI, confidence interval; ESKD, end-stage kidney disease; HR, hazard ratio. Crude and adjusted HRs are indicated for 1 L/1.73 m2 increase in extracellular fluid volume. Analyses were adjusted for the following covariates: age, sex, site of inclusion, ethnicity, body mass index, diabetic status (no diabetes, diabetes without diabetic nephropathy, diabetes with diabetic nephropathy), elevated blood pressure, urinary protein-to-creatinine ratio (log-transformed), 24-hour urinary sodium excretion, diuretics, and renin–angiotensin system inhibitors. For mortality, models were also adjusted for previous cardiovascular events (myocardial infarction or angioplasty or stroke or heart failure) and plasma albumin concentration. Single imputations were used for missing data. Model (1) was stratified for baseline measured glomerular filtration rate. Model (2) was adjusted for measured glomerular filtration rate (expressed as a continuous variable, as a time-dependent coefficient). Model (3) was adjusted for systolic blood pressure as a continuous time-varying covariate. Analyses of longitudinal data (median number of visits: 2 [1–4] per patient, median duration between 2 consecutive visits 1.1 [IQR: 1.0–1.5 years]) showed that the mean mGFR slope was –1.64, 95% CI (–1.82, –1.45) ml/min per year in the total population, and –1.31, 95% CI (–1.60, –1.01) ml/min per year; –1.49, 95% CI (–1.81, –1.17) ml/min per year, and –2.28, 95% CI (–2.63, –1.92) ml/min per year for the first, second, and third tertiles of ECF, respectively. In the fully adjusted linear mixed-effect model, ECF was significantly associated with a faster mGFR decline (mean difference in mGFR slope per 1 L/1.73 m2 increase in ECF: –0.14, 95% CI [–0.23, –0.05] ml/min per year, P = 0.002) (Table 3; Figure 3). Similar results were observed when ECF was analyzed in tertiles (Table 3) and in the subgroup of patients with at least 2 visits (Supplementary Table S4). Analyses yielded similar trends when GFR was estimated using the deindexed Chronic Kidney Disease–Epidemiology Collaboration (CKD-EPI) formula (Supplementary Table S5).Table 3Decline of measured glomerular filtration rate according to ECF (n = 1593)Mean difference in mGFR slopes (ml/min per yr)Model 0Model 1Model 2Model 3ECF analyzed as a continuous variableECF (per 1 L/1.73 m2)–0.19 [–0.28, –0.10]–0.20 [–0.29, –0.1]–0.15 [–0.25, –0.06]–0.14 [–0.23, –0.05]Elevated blood pressure–0.78 [–1.19, –0.37]–0.61 [–1.01, –0.21]Protein-to-creatinine ratioaThe indicated coefficient corresponds to a 2.72-fold increase in protein-to-creatinine ratio.–0.43 [–0.55, –0.32]ECF analyzed in tertilesECF1st tertile [8.88–14.03]RefRefRefRef2nd tertile [14.04–15.82]–0.09 [–0.54, 0.35]–0.10 [–0.55, 0.35]0.02 [–0.43, 0.47]–0.02 [–0.46, 0.41]3rd tertile [15.83–23.29]–0.86 [–1.34, –0.39]–0.90 [–1.37, –0.42]–0.72 [–1.20, –0.24]–0.55 [–1.02, –0.08]Elevated blood pressure–0.85 [–1.25, –0.44]–0.68 [–1.08, –0.29]Protein-to-creatinine ratioaThe indicated coefficient corresponds to a 2.72-fold increase in protein-to-creatinine ratio.–0.42 [–0.54, –0.31]ECF, extracellular fluid volume; mGFR, measured glomerular filtration rate; Ref, referent.ECF was analyzed as a continuous variable (results are indicated per 1 L/1.73 m2 increase in ECF), or in tertiles. Model 0: time, baseline values of ECF, and mGFR levels (>60, 45–60, 30–44, 15–29 ml/min per 1.73 m2), and interaction terms between time and GFR and time and ECF. Model 1: Model 0 + all baseline covariates (age, sex, ethnicity, recruitment site, body mass index, diabetic status (no diabetes, diabetes without diabetic nephropathy, diabetes with diabetic nephropathy), elevated blood pressure (< or ≥140/90 mm Hg), urinary protein-to-creatinine ratio (log-transformed, per 1-log unit increase), 24-h urinary sodium excretion, diuretics, and renin–angiotensin system inhibitors). Model 2: Model 1 + interaction term between time and elevated blood pressure. Model 3: Model 2 + interaction terms between time and urinary protein-to-creatinine ratio and between time and site of inclusion. mGFR decline was modeled using a linear mixed-effect regression model. Mean differences in mGFR slopes are expressed in ml/min per year.a The indicated coefficient corresponds to a 2.72-fold increase in protein-to-creatinine ratio. Open table in a new tab ECF, extracellular fluid volume; mGFR, measured glomerular filtration rate; Ref, referent. ECF was analyzed as a continuous variable (results are indicated per 1 L/1.73 m2 increase in ECF), or in tertiles. Model 0: time, baseline values of ECF, and mGFR levels (>60, 45–60, 30–44, 15–29 ml/min per 1.73 m2), and interaction terms between time and GFR and time and ECF. Model 1: Model 0 + all baseline covariates (age, sex, ethnicity, recruitment site, body mass index, diabetic status (no diabetes, diabetes without diabetic nephropathy, diabetes with diabetic nephropathy), elevated blood pressure (< or ≥140/90 mm Hg), urinary protein-to-creatinine ratio (log-transformed, per 1-log unit increase), 24-h urinary sodium excretion, diuretics, and renin–angiotensin system inhibitors). Model 2: Model 1 + interaction term between time and elevated blood pressure. Model 3: Model 2 + interaction terms between time and urinary protein-to-creatinine ratio and between time and site of inclusion. mGFR decline was modeled using a linear mixed-effect regression model. Mean differences in mGFR slopes are expressed in ml/min per year. In this multicenter study conducted in 1593 patients with CKD stage 1 to 4, with a median follow-up of 5.3 years, a higher ECF was independently associated with ESKD (linear association) and death (nonlinear association, with an increasing risk as ECF increased above a threshold of 15 L/1.73 m2). These findings are robust because they rely on gold-standard methods for ECF and GFR measurement9Bird N.J. Peters C. Michell A.R. et al.Extracellular distribution volumes of hydrophilic solutes used to measure the glomerular filtration rate: comparison between chromium-51-EDTA and iohexol.Physiol Meas. 2007; 28: 223-234Crossref PubMed Scopus (5) Google Scholar, 10Ellis K.J. Human body composition: in vivo methods.Physiol Rev. 2000; 80: 649-680Crossref PubMed Scopus (594) Google Scholar, 11Zdolsek J.H. Lisander B. Hahn R.G. Measuring the size of the extracellular fluid space using bromide, iohexol, and sodium dilution.Anesth Analg. 2005; 101: 1770-1777Crossref PubMed Scopus (28) Google Scholar, 12Peters A.M. Allison H. Ussov W. Simultaneous measurement of extracellular fluid distribution and renal function with a single injection of 99mTc DTPA.Nephrol Dial Transplant. 1995; 10: 1829-1833PubMed Google Scholar, 13Visser F.W. Muntinga J.H. Dierckx R.A. et al.Feasibility and impact of the measurement of extracellular fluid volume simultaneous with GFR by 125I-iothalamate.Clin J Am Soc Nephrol. 2008; 3: 1308-1315Crossref PubMed Scopus (46) Google Scholar, 14Brochner-Mortensen J. A simple single injection method for determination of the extracellular fluid volume.Scand J Clin Lab Invest. 1980; 40: 567-573Crossref PubMed Scopus (67) Google Scholar, 15Henriksen U.L. Henriksen J.H. The clearance concept with special reference to determination of glomerular filtration rate in patients with fluid retention.Clin Physiol Funct Imaging. 2015; 35: 7-16Crossref PubMed Scopus (14) Google Scholar in a large cohort of nearly 1600 carefully phenotyped patients, and analyses were adjusted for multiple potential confounders. In addition, we found an association of a higher ECF with mGFR decline, strengthening our results. Studies evaluating the role of fluid overload on renal function and mortality in patients with non-dialysis CKD have yielded conflicting results. In the largest study to date, recently published, Bansal et al.4Bansal N. Zelnick L.R. Himmelfarb J. et al.Bioelectrical impedance analysis measures and clinical outcomes in CKD.Am J Kidney Dis. 2018; 72: 662-672Abstract Full Text Full Text PDF PubMed Scopus (25) Google Scholar showed in 3751 patients from the chronic renal insufficiency cohort (CRIC) that a shorter vector length—a bioelectrical impedance analysis–derived marker of overhydration—was significantly associated with the risk for heart failure, with adjusted HR for first (highest hydration state) versus third and fourth quartiles of 1.28, 95% CI (1.01, 1.61), but not with all-cause mortality or CKD progression. In contrast, all previous pioneer works evaluating the prognostic role of ECF in patients with non-dialysis CKD showed that fluid overload was associated with mortality and/or CKD progression. However, they suffered from a number of limitations: they were conducted in smaller cohorts; they had composite endpoints; and bioelectrical impedance analysis (and not a gold standard such as isotope dilution) was used to assess fluid volume.7Kyle U.G. Bosaeus I. De Lorenzo A.D. et al.Bioelectrical impedance analysis—part I: review of principles and methods.Clin Nutr. 2004; 23: 1226-1243Abstract Full Text Full Text PDF PubMed Scopus (1829) Google Scholar, 8Kyle U.G. Bosaeus I. De Lorenzo A.D. et al.Bioelectrical impedance analysis—part II: utilization in clinical practice.Clin Nutr. 2004; 23: 1430-1453Abstract Full Text Full Text PDF PubMed Scopus (1380) Google Scholar, 10Ellis K.J. Human body composition: in vivo methods.Physiol Rev. 2000; 80: 649-680Crossref PubMed Scopus (594) Google Scholar Thus, in a single-center cohort of 472 non-dialysis patients with CKD stage 4–5, Tsai et al.6Tsai Y.C. Tsai J.C. Chen S.C. et al.Association of fluid overload with kidney disease progression in advanced CKD: a prospective cohort study.Am J Kidney Dis. 2014; 63: 68-75Abstract Full Text Full Text PDF PubMed Scopus (67) Google Scholar reported that fluid overload was associated with an increased risk of renal replacement therapy (adjusted HR for third versus first tertile = 3.16, 95% CI [1.33, 7.50]) and with a faster GFR decline (–1.10 95% CI [–2.06, –0.13] ml/min per 1.73 m2 per year in the third vs. first tertile), after a median follow-up of 17.3 months. In the same cohort of patients, fluid overload was later reported to be associated with the combined endpoint of all-cause mortality and cardiovascular morbidity.5Tsai Y.C. Chiu Y.W. Tsai J.C. et al.Association of fluid overload with cardiovascular morbidity and all-cause mortality in stages 4 and 5 CKD.Clin J Am Soc Nephrol. 2015; 10: 39-46Crossref PubMed Scopus (94) Google Scholar In a single-