Patient-specific evolution of renal function in chronic heart failure patients dynamically predicts clinical outcome in the Bio-SHiFT study

Renal dysfunction is an important component of chronic heart failure (CHF), but its single assessment does not sufficiently reflect clinically silent progression of CHF prior to adverse clinical outcome. Therefore, we aimed to investigate temporal evolutions of glomerular and tubular markers in 263 stable patients with CHF, and to determine if their patient-specific evolutions during this clinically silent period can dynamically predict clinical outcome. We determined the risk of clinical outcome (composite endpoint of Heart Failure hospitalization, cardiac death, Left Ventricular Assist Device placement, and heart transplantation) in relation to marker levels, slopes and areas under their trajectories. In each patient, the trajectories were estimated using repeatedly measured glomerular markers: creatinine/estimated glomerular filtration rate (eGFR), cystatin C (CysC), and tubular markers: urinary N-acetyl-beta-D-glucosaminidase (NAG) and kidney injury molecule (KIM)-1, plasma and urinary neutrophil gelatinase-associated lipocalin (NGAL). During 2.2 years of follow-up, we collected on average 8 urine and 9 plasma samples per patient. All glomerular markers predicted the endpoint (univariable hazard ratio [95% confidence interval] per 20% increase: creatinine: 1.18[1.07–1.31], CysC: 2.41[1.81–3.41], and per 20% eGFR decrease: 1.13[1.05–1.23]). Tubular markers, NAG, and KIM-1 also predicted the endpoint (NAG: 1.06[1.01–1.11] and KIM-1: 1.08[1.04–1.11]). Larger slopes were the strongest predictors (creatinine: 1.57[1.39–1.84], CysC: 1.76[1.52–2.09], eGFR: 1.59[1.37–1.90], NAG: 1.26[1.11–1.44], and KIM-1: 1.64[1.38–2.05]). Associations persisted after multivariable adjustment for clinical characteristics. Thus, during clinically silent progression of CHF, glomerular and tubular functions deteriorate, but not simultaneously. Hence, patient-specific evolutions of these renal markers dynamically predict clinical outcome in patients with CHF. Renal dysfunction is an important component of chronic heart failure (CHF), but its single assessment does not sufficiently reflect clinically silent progression of CHF prior to adverse clinical outcome. Therefore, we aimed to investigate temporal evolutions of glomerular and tubular markers in 263 stable patients with CHF, and to determine if their patient-specific evolutions during this clinically silent period can dynamically predict clinical outcome. We determined the risk of clinical outcome (composite endpoint of Heart Failure hospitalization, cardiac death, Left Ventricular Assist Device placement, and heart transplantation) in relation to marker levels, slopes and areas under their trajectories. In each patient, the trajectories were estimated using repeatedly measured glomerular markers: creatinine/estimated glomerular filtration rate (eGFR), cystatin C (CysC), and tubular markers: urinary N-acetyl-beta-D-glucosaminidase (NAG) and kidney injury molecule (KIM)-1, plasma and urinary neutrophil gelatinase-associated lipocalin (NGAL). During 2.2 years of follow-up, we collected on average 8 urine and 9 plasma samples per patient. All glomerular markers predicted the endpoint (univariable hazard ratio [95% confidence interval] per 20% increase: creatinine: 1.18[1.07–1.31], CysC: 2.41[1.81–3.41], and per 20% eGFR decrease: 1.13[1.05–1.23]). Tubular markers, NAG, and KIM-1 also predicted the endpoint (NAG: 1.06[1.01–1.11] and KIM-1: 1.08[1.04–1.11]). Larger slopes were the strongest predictors (creatinine: 1.57[1.39–1.84], CysC: 1.76[1.52–2.09], eGFR: 1.59[1.37–1.90], NAG: 1.26[1.11–1.44], and KIM-1: 1.64[1.38–2.05]). Associations persisted after multivariable adjustment for clinical characteristics. Thus, during clinically silent progression of CHF, glomerular and tubular functions deteriorate, but not simultaneously. Hence, patient-specific evolutions of these renal markers dynamically predict clinical outcome in patients with CHF. Heart failure (HF) is the leading cause of hospitalization worldwide.1Burchfield J.S. Xie M. Hill J.A. Pathological ventricular remodeling: mechanisms: part 1 of 2.Circulation. 2013; 128: 388-400Crossref PubMed Scopus (506) Google Scholar Despite declines in HF-related mortality as a result of current therapies, re-hospitalization rates for decompensation of chronic heart failure (CHF) remain high.1Burchfield J.S. Xie M. Hill J.A. Pathological ventricular remodeling: mechanisms: part 1 of 2.Circulation. 2013; 128: 388-400Crossref PubMed Scopus (506) Google Scholar, 2Eapen Z.J. Liang L. Fonarow G.C. et al.Validated, electronic health record deployable prediction models for assessing patient risk of 30-day rehospitalization and mortality in older heart failure patients.JACC Heart Fail. 2013; 1: 245-251Crossref PubMed Scopus (88) Google Scholar Several blood biomarkers that predict re-hospitalization and mortality have been identified in patients with CHF.3Ponikowski P. Voors A.A. Anker S.D. et al.2016 ESC guidelines for the diagnosis and treatment of acute and chronic heart failure: the task force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC) Developed with the special contribution of the Heart Failure Association (HFA) of the ESC.Eur Heart J. 2016; 37: 2129-2200Crossref PubMed Scopus (8759) Google Scholar Still their predictive capabilities in practice are limited, and adequate risk assessment remains a challenge.3Ponikowski P. Voors A.A. Anker S.D. et al.2016 ESC guidelines for the diagnosis and treatment of acute and chronic heart failure: the task force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC) Developed with the special contribution of the Heart Failure Association (HFA) of the ESC.Eur Heart J. 2016; 37: 2129-2200Crossref PubMed Scopus (8759) Google Scholar Estimation of renal dysfunction, which coexists and interacts with HF,3Ponikowski P. Voors A.A. Anker S.D. et al.2016 ESC guidelines for the diagnosis and treatment of acute and chronic heart failure: the task force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC) Developed with the special contribution of the Heart Failure Association (HFA) of the ESC.Eur Heart J. 2016; 37: 2129-2200Crossref PubMed Scopus (8759) Google Scholar may improve risk stratification. Baseline glomerular dysfunction, as assessed by estimated glomerular filtration rate (eGFR), entails an unfavorable prognosis in CHF.4Damman K. Valente M.A. Voors A.A. et al.Renal impairment, worsening renal function, and outcome in patients with heart failure: an updated meta-analysis.Eur Heart J. 2014; 35: 455-469Crossref PubMed Scopus (574) Google Scholar, 5Dries D.L. Exner D.V. Domanski M.J. et al.The prognostic implications of renal insufficiency in asymptomatic and symptomatic patients with left ventricular systolic dysfunction.J Am Coll Cardiol. 2000; 35: 681-689Crossref PubMed Scopus (727) Google Scholar, 6Hillege H.L. Girbes A.R. de Kam P.J. et al.Renal function, neurohormonal activation, and survival in patients with chronic heart failure.Circulation. 2000; 102: 203-210Crossref PubMed Scopus (856) Google Scholar Besides glomerular impairment, such patients often have tubular damage due to tubulo-interstitial injury by renal tissue hypoperfusion or due to damaged glomerular barrier.7Goldfarb M. Abassi Z. Rosen S. et al.Compensated heart failure predisposes to outer medullary tubular injury: studies in rats.Kidney Int. 2001; 60: 607-613Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar, 8Tsuruya K. Eriguchi M. Cardiorenal syndrome in chronic kidney disease.Curr Opin Nephrol Hypertens. 2015; 24: 154-162Crossref PubMed Scopus (22) Google Scholar Notably, a single assessment of damaged tubules predicts adverse outcome in CHF independently of eGFR.9Damman K. Voors A.A. Navis G. et al.Current and novel renal biomarkers in heart failure.Heart Fail Rev. 2012; 17: 241-250Crossref PubMed Scopus (35) Google Scholar, 10Damman K. Masson S. Hillege H.L. et al.Tubular damage and worsening renal function in chronic heart failure.JACC Heart Fail. 2013; 1: 417-424Crossref PubMed Scopus (69) Google Scholar, 11Damman K. Masson S. Hillege H.L. et al.Clinical outcome of renal tubular damage in chronic heart failure.Eur Heart J. 2011; 32: 2705-2712Crossref PubMed Scopus (150) Google Scholar It is clear that both glomerular and tubular function are important in patients with CHF, but their single assessment does not sufficiently reflect deterioration along the cardio-renal axis that occurs over time preceding adverse events. Yet the temporal evolution of renal function preceding the event may dynamically ascertain the clinically silent progression of the disease. Specifically, it would enable accurate investigation of whether, and to which degree, increasing (or decreasing) levels of renal biomarkers contribute to the patient’s risk, regardless of whether these levels exceed established cut points at study baseline (i.e., a random point in time prior to event). In the context of cardio-renal interplay, patients with CHF also display large biological heterogeneity. Renal function not only changes dynamically within a patient over time, but also differs from patient to patient. Hence, the true potential of renal markers in ascertaining individual disease progression and their accurate relation with clinical outcome can only be revealed if their patient-specific evolutions are considered. However, detailed individual temporal evolutions of renal function in CHF have never been described. To overcome these issues, our aim was 2-fold: (i) to investigate the average (population) temporal evolutions of glomerular function (measured with plasma creatinine [Cr], eGFR, and cystatin C [CysC]) and tubular status (measured with urinary kidney injury molecule [KIM]-1, N-acetyl-beta-d-glucosaminidase [NAG], and urinary and plasma neutrophil gelatinase-associated lipocalin [NGAL]) in stable patients with CHF; and (ii) to determine whether patient-specific (individual) evolutions of these renal biomarkers during a clinically silent period can dynamically predict clinical outcome. For this purpose we examined several aspects of the temporal evolution of each renal biomarker that may be relevant for clinical prediction. Table 1 displays the baseline characteristics. At baseline, patients who later experienced the endpoint were older; more frequently had diabetes and atrial fibrillation; had lower systolic blood pressure, higher New York Heart Association (NYHA) class, higher levels of N-terminal prohormone of brain natriuretic peptide (NT-proBNP), cardiac troponin T, CysC, urinary N-acetyl-beta-D-glucosaminidase (NAG), and plasma urinary neutrophil-gelatinase-associated-lipocalin (NGAL); and were more frequently on diuretics than the patients who remained endpoint-free.Table 1Patient characteristics in relation to the occurrence of the composite endpointVariableTotalComposite endpoint reachedP valueYesNoN (%)263 (100)70 (27)193 (73)Demographics Age, yr (mean ± SD)67 ± 1369 ± 1366 ± 120.05 Men, n (%)189 (72)53 (76)136 (70)0.41Clinical characteristics BMI, kg/m2 (mean ± SD)27.5 ± 4.727.6 ± 4.827.4 ± 4.70.80 Heart rate, bpm (mean ± SD)67 ± 1269 ± 1367 ± 110.31 SBP, mm Hg (mean ± SD)122 ± 20117 ± 17124 ± 210.02 DBP, mm Hg (mean ± SD)72 ± 1170 ± 1073 ± 110.06Features of heart failure NYHA class III or IV, n (%)69 (26)31 (44)38 (20)<0.001 HF-rEF n (%)250 (95)66 (94)184 (95)0.75 HF-pEF n (%)13 (5)4 (6)9 (5) LVEF, % (mean ± SD)32 ± 1130 ± 1133 ± 100.18 NT pro-BNP (pmol/l)∗All biomarkers levels were presented as median (interquartile range).137.3 (51.7–272.6)282.4 (176.4–517.4)95.3 (31.72–207.7)<0.001 Hs-TnT (ng/l)∗All biomarkers levels were presented as median (interquartile range).18.0 (9.5–33.2)31.9 (20.6–49.7)13.9 (8.4–26.7)<0.001Etiology of heart failure, n (%) Ischemic117 (44)36 (51)81 (42)0.17 Hypertension34 (13)10 (14)24 (12)0.70 Secondary to valvular disease12 (5)5 (7)7 (4)0.23 Cardiomyopathy68 (26)15 (21)53 (28)0.32 Unknown or Others32 (12)4 (6)28 (15)Medical history, n (%) Prior MI96 (36)32 (46)64 (33)0.06 Prior PCI82 (31)27 (39)55 (28)0.12 Prior CABG43 (16)13 (19)30 (15)0.57 Atrial fibrillation106 (40)36 (51)70 (36)0.03 Diabetes81 (31)32 (46)49 (25)0.002 Hypercholesterolemia96 (36)30 (43)66 (34)0.20 Hypertension120 (46)38 (54)82 (42)0.09 COPD31 (12)12 (17)19 (10)0.10Medication use, n (%) Beta-blocker236 (90)61 (87)175 (91)0.40 ACE-I or ARB245 (93)63 (90)182 (94)0.22 Diuretics237 (90)68 (97)169 (88)0.02Loop diuretics236 (90)68 (97)168 (87)0.02Thiazides7 (3)3 (4)4 (2)0.28 Aldosterone antagonist179 (68)53 (76)126 (65)0.11Glomerular function markers∗All biomarkers levels were presented as median (interquartile range). Creatinine, mg/dl1.18 (0.99–1.49)1.30 (1.02–1.52)1.17 (0.98–1.45)0.18 eGFR, ml/min per 1.73m258 (43–76)53 (40–73)59 (44–77)0.16 Cystatin C, mg/l0.73 (0.57–0.97)0.87 (0.71–1.03)0.70 (0.53–0.90)<0.001KDOQI classification, n (%) eGFR ≥ 90 ml/min per 1.73m228 (11)7 (10)21 (11)0.18 eGFR 60–89 ml/min per 1.73m295 (36)20 (28)75 (39) eGFR 30–59 ml/min per 1.73m2119 (45)37 (53)82 (42) eGFR < 30 ml/min per 1.73m221 (8)6 (9)15 (8)Tubular markers∗All biomarkers levels were presented as median (interquartile range). NAG, U/gCr (urine)5.9 (3.8–9.3)8.0 (6.0–11.0)5.1 (3.3–8.0)<0.001 KIM-1, ng/gCr (urine)477.2 (247.0–938.6)589.0 (255.0–957.2)465.1 (237.6–911.5)0.10 NGAL, μg/gCr (urine)17.4 (9.2–32.6)18.2 (10.0–50.5)17.4 (9.0–31.4)0.20 NGAL, ng/ml (plasma)190.1 (133.5–280.0)260.8 (169.5–355.4)179.2 (127.9–244.5)<0.001ACE-I, angiotensin-converting enzyme inhibitors; ARB, angiotensin II receptor blockers; BMI, body mass index; bpm, beats per minute; CABG, coronary artery bypass grafting; COPD, chronic obstructive pulmonary disease; CVA, cerebrovascular accident; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; HF-pEF, heart failure with preserved ejection fraction; HF-rEF, heart failure with reduced ejection fraction; LVEF, left ventricular ejection fraction; MI, myocardial infarction; NYHA, New York Heart Association; PCI, percutaneous coronary intervention; SBP, systolic blood pressure; TIA, transitory ischemic attack.Normally distributed continuous variables are presented as mean ± SD, and non-normally distributed variables as median and interquartile range. Categorical variables are presented as numbers and percentages.∗ All biomarkers levels were presented as median (interquartile range). Open table in a new tab ACE-I, angiotensin-converting enzyme inhibitors; ARB, angiotensin II receptor blockers; BMI, body mass index; bpm, beats per minute; CABG, coronary artery bypass grafting; COPD, chronic obstructive pulmonary disease; CVA, cerebrovascular accident; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; HF-pEF, heart failure with preserved ejection fraction; HF-rEF, heart failure with reduced ejection fraction; LVEF, left ventricular ejection fraction; MI, myocardial infarction; NYHA, New York Heart Association; PCI, percutaneous coronary intervention; SBP, systolic blood pressure; TIA, transitory ischemic attack. Normally distributed continuous variables are presented as mean ± SD, and non-normally distributed variables as median and interquartile range. Categorical variables are presented as numbers and percentages. From 263 patients with CHF, a total of 1912 urine and 1984 blood samples were collected with a median (interquartile range, IQR) of 8 (5–10) urine and 9 (5–10) plasma samples per patient. During a median (IQR) follow-up of 2.2 (1.4–2.5) years, 70 (27%) patients reached the primary endpoint: 56 patients were re-hospitalized for acute or worsened HF, 3 patients underwent heart transplantation, 2 patients underwent left ventricle assist device (LVAD) placement, and 9 patients died of cardiovascular causes. In patients who reached the composite endpoint, Cr levels on average showed an increasing pattern over time preceding the endpoint. In endpoint-free patients Cr levels were lower and remained stable during follow-up (Figure 1a). eGFR displayed similar dynamics (Figure 1b). Independently of baseline levels, repeatedly measured Cr and eGFR predicted the endpoint (per 20% increase of Cr levels: hazard ratio [95% confidence interval] 1.18 [1.07–1.31], P = 0.004, and per 20% eGFR decrease: 1.13 [1.05–1.1.23], P = 0.002) (Table 2). Similarly, their larger slopes and larger area under the curve of the marker’s trajectory (AUCm) predicted the endpoint (per 20% increase of Cr slope: 1.57 [1.39–1.84], P < 0.001, per 20% decrease of eGFR slope: 1.59 [1.37–1.90], P < 0.001) (per 20% increase of Cr’s AUCm: 1.10 [1.03–1.18], P = 0.010, and eGFR’s AUCm: 1.07 [1.02–1.11], P < 0.001). These risk estimates remained significant even after adjustment for clinical characteristics and dose changes of HF medications during follow-up. After adjustment for cardiac markers, Cr’s levels and AUCm lost precision, whereas eGFR remained significant (Table 2). Table S1 shows similar results for HF hospitalizations (secondary endpoint).Table 2Associations between glomerular function markers and the composite endpointCreatinineeGFRCystatin CHR (95% CI)P valueHR (95% CI)P valueHR (95% CI)P valueBaseline levelaHRs and 95% CIs are given per 20% increase of creatinine and cystatin C, and 20% eGFR decrease. Model A is unadjusted. Model B is adjusted for age, sex, diabetes, atrial fibrillation, baseline New York Heart Association class, diuretics, and systolic blood pressure. Model C is adjusted for baseline NT-proBNP and hs-cTnT.Model A1.04 (0.99–1.09)0.141.03 (0.99–1.07)0.131.09 (1.05–1.14)<0.001Model B1.02 (0.97–1.07)0.491.02 (0.97–1.06)0.481.07 (1.02–1.12)0.007Model C0.98 (0.93–1.03)0.460.98 (0.94–1.02)0.281.00 (0.95–1.06)0.89Temporal evolutionbHRs and 95% CIs are given per 20% increase of the level, slope, and AUCm of creatinine and cystatin C, and 20% decrease of the level, slope, and AUCm of eGFR. Model 1 is Cox model–adjusted for marker’s baseline levels and LME model–adjusted for sampling time. Model 2 is Cox and LME model–adjusted for the clinical variables age, sex, diabetes, atrial fibrillation, baseline New York Heart Association class, diuretics, systolic blood pressure, and sampling time (LME). Model 3 is Cox and LME model–adjusted for baseline NT-proBNP and hs-cTnT and sampling time (LME). Model 4 is time-dependent Cox–adjusted for total daily equivalent doses of carvedilol, enalapril, furosemide, and spironolactone during follow-up.Repeatedly measured levelsModel 11.18 (1.07–1.31)0.0041.13 (1.05–1.23)0.0022.41 (1.81–3.41)<0.001Model 21.12 (1.02–1.23)0.0221.12 (1.06–1.20)<0.0012.16 (1.44–3.72)<0.001Model 31.05 (0.96–1.15)0.281.09 (1.04–1.14)<0.0011.63 (1.35–2.30)<0.001Model 41.15 (1.08–1.24)<0.0011.10 (1.04–1.16)<0.0012.27 (1.99–2.59)<0.001Annual slopeModel 11.57 (1.39–1.84)<0.0011.59 (1.37–1.90)<0.0011.76 (1.52–2.09)<0.001Model 21.65 (1.40–1.98)<0.0011.64 (1.38–2.02)<0.0012.00 (1.66–2.51)<0.001Model 31.37 (1.22–1.57)<0.0011.30 (1.16–1.46)0.0021.47 (1.32–1.66)<0.001Model 41.28 (1.16–1.43)<0.0011.18 (1.07–1.31)0.0011.63 (1.50–1.77)<0.001AUCmModel 11.10 (1.03–1.18)0.0101.07 (1.02–1.11)<0.0011.32 (1.17–1.54)<0.001Model 21.08 (1.01–1.15)0.0201.07 (1.02–1.12)<0.0011.23 (1.13–1.36)<0.001Model 31.04 (0.98–1.10)0.171.06 (1.02–1.10)<0.0011.17 (1.08–1.28)<0.001AUCm, area under the curve of marker’s trajectory; CI, confidence interval; eGFR, estimated glomerular filtration rate; HR, hazard ratio; LME, linear mixed effects.a HRs and 95% CIs are given per 20% increase of creatinine and cystatin C, and 20% eGFR decrease. Model A is unadjusted. Model B is adjusted for age, sex, diabetes, atrial fibrillation, baseline New York Heart Association class, diuretics, and systolic blood pressure. Model C is adjusted for baseline NT-proBNP and hs-cTnT.b HRs and 95% CIs are given per 20% increase of the level, slope, and AUCm of creatinine and cystatin C, and 20% decrease of the level, slope, and AUCm of eGFR. Model 1 is Cox model–adjusted for marker’s baseline levels and LME model–adjusted for sampling time. Model 2 is Cox and LME model–adjusted for the clinical variables age, sex, diabetes, atrial fibrillation, baseline New York Heart Association class, diuretics, systolic blood pressure, and sampling time (LME). Model 3 is Cox and LME model–adjusted for baseline NT-proBNP and hs-cTnT and sampling time (LME). Model 4 is time-dependent Cox–adjusted for total daily equivalent doses of carvedilol, enalapril, furosemide, and spironolactone during follow-up. Open table in a new tab AUCm, area under the curve of marker’s trajectory; CI, confidence interval; eGFR, estimated glomerular filtration rate; HR, hazard ratio; LME, linear mixed effects. In patients who reached the composite endpoint, CysC showed on average higher baseline levels that increased further as the endpoint approached. In endpoint-free patients, CysC levels were lower and slightly decreased during follow-up (Figure 1c). Independently of baseline levels, CysC levels at any time during follow-up were associated with the endpoint (per 20% increase of CysC levels: 2.41 [1.81–3.41], P < 0.001) (Table 2). Similarly, larger slope and larger AUCm predicted the endpoint (1.76 [1.52–2.09], P < 0.001 and 1.32 [1.17–1.54], P < 0.001). These risk estimates remained significant after multivariable adjustments (Table 2). Supplementary Table S1 shows similar results for HF hospitalizations. Overall, we found substantial associations between NAG, KIM-1, and NGAL, but only mild associations between these tubular markers and glomerular function markers (namely CysC), when assessed during follow-up (Table S2). In patients who reached the composite endpoint, NAG showed on average higher baseline levels that increased further as the endpoint approached. In endpoint-free patients, NAG levels were lower and decreased during follow-up (Figure 2a). Independently of baseline levels, higher NAG levels at any time during follow-up were associated with the endpoint (per 20% increase of NAG levels: 1.06 [1.01–1.11], P = 0.018). Similarly, larger NAG slope predicted the endpoint (1.26 [1.11–1.44], P = 0.004).These risk estimates remained significant after multivariable adjustments, except for NAG slope that became insignificant after controlling for cardiac markers (Table 3). Table S3 shows similar results for HF hospitalizations, except for NAG levels that lost significance after adjusting for cardiac markers.Table 3Associations between tubular markers, urinary NAG and KIM-1, and the composite endpointUrinary NAGUrinary KIM-1HR (95% CI)P valueHR (95% CI)P valueBaseline levelsaHRs and 95% CIs are given per 20% increase of urinary NAG and KIM-1. Model A is unadjusted. Model B is adjusted for age, sex, diabetes, atrial fibrillation, baseline New York Heart Association class, diuretics, systolic blood pressure, and estimated glomerular filtration rate. Model C is adjusted for baseline NT-proBNP and hs-cTnT.Model A1.07 (1.05–1.09)<0.0011.02 (1.00–1.04)0.06Model B1.06 (1.03–1.09)<0.0011.01 (0.99–1.03)0.26Model C1.03 (1.00–1.06)0.0500.99 (0.97–1.01)0.44Temporal evolutionbHRs and 95% CIs are given per 20% increase of the level, slope, and AUCm of urinary NAG and KIM-1. Model 1 is Cox model–adjusted for marker’s baseline levels and LME model–adjusted for sampling time. Model 2 is Cox and LME model–adjusted for age, sex, diabetes, atrial fibrillation, baseline New York Heart Association class, diuretics, systolic blood pressure, estimated glomerular filtration rate, and sampling time (LME). Model 3 is Cox and LME model–adjusted for baseline NT-proBNP and hs-cTnT and sampling time (LME). Model 4 is time-dependent Cox–adjusted for total daily equivalent doses of carvedilol, enalapril, furosemide, and spironolactone during follow-up.Repeatedly measured levelsModel 11.06 (1.01–1.11)0.0181.08 (1.04–1.11)<0.001Model 21.07 (1.03–1.12)<0.0011.06 (1.03–1.10)<0.001Model 31.05 (1.00–1.10)0.0481.04 (1.01–1.07)0.016Model 41.13 (1.09–1.17)<0.0011.06 (1.03–1.09)<0.001Annual slopeModel 11.26 (1.11–1.44)0.0041.64 (1.38–2.05)<0.001Model 21.50 (1.18–2.00)0.0021.78 (1.41–2.39)<0.001Model 30.81 (0.65–1.41)0.161.52 (1.25–1.98)<0.001Model 41.10 (1.02–1.20)0.0091.12 (1.04–1.20)0.002AUCmModel 11.02 (0.99–1.05)0.111.01 (0.99–1.02)0.23Model 21.04 (1.01–1.07)0.011.01 (0.99–1.03)0.10Model 31.01 (0.98–1.05)0.331.01 (0.99–1.02)0.38AUCm, area under the curve of marker’s trajectory; CI, confidence interval; HR, hazard ratio; KIM-1, kidney injury molecule-1; LME, linear mixed effects; NAG, N-acetyl-beta-D-glucosaminidase.a HRs and 95% CIs are given per 20% increase of urinary NAG and KIM-1. Model A is unadjusted. Model B is adjusted for age, sex, diabetes, atrial fibrillation, baseline New York Heart Association class, diuretics, systolic blood pressure, and estimated glomerular filtration rate. Model C is adjusted for baseline NT-proBNP and hs-cTnT.b HRs and 95% CIs are given per 20% increase of the level, slope, and AUCm of urinary NAG and KIM-1. Model 1 is Cox model–adjusted for marker’s baseline levels and LME model–adjusted for sampling time. Model 2 is Cox and LME model–adjusted for age, sex, diabetes, atrial fibrillation, baseline New York Heart Association class, diuretics, systolic blood pressure, estimated glomerular filtration rate, and sampling time (LME). Model 3 is Cox and LME model–adjusted for baseline NT-proBNP and hs-cTnT and sampling time (LME). Model 4 is time-dependent Cox–adjusted for total daily equivalent doses of carvedilol, enalapril, furosemide, and spironolactone during follow-up. Open table in a new tab AUCm, area under the curve of marker’s trajectory; CI, confidence interval; HR, hazard ratio; KIM-1, kidney injury molecule-1; LME, linear mixed effects; NAG, N-acetyl-beta-D-glucosaminidase. In patients who reached the composite endpoint, KIM-1 levels showed an average increasing pattern over time preceding the endpoint. In endpoint-free patients, KIM-1 levels were lower and slightly decreased during follow-up (Figure 2b). Independently of baseline levels, higher KIM-1 levels at any time during follow-up were associated with the endpoint (per 20% increase of KIM-1 levels: 1.08 [1.04–1.11], P < 0.001). Similarly, larger KIM-1 slope predicted the endpoint (1.64 [1.38–2.05], P < 0.001). These risk estimates remained significant after multivariable adjustments (Table 3). Table S3 shows similar results for HF hospitalizations, except for KIM-1 levels that lost significance after adjusting for cardiac markers. Although baseline plasma NGAL levels were higher in patients who reached the endpoint, this difference declined during follow-up (Supplementary Figure S1A). The evolution of urinary NGAL levels of patients who reached the endpoint and those who did not substantially overlapped during follow-up (Supplementary Figure S1B). No clear associations were found between NGAL and primary and secondary endpoints during follow-up (Supplementary Tables S4 and S5). Supplementary Table S6 shows the time-dependent area under the receiver operating curve (AUC) for the different renal markers for the composite endpoint. After the 1-year collection time period, markers showed reasonably good discriminatory power both for the 6- and 12-month risk window, with slightly better accuracy for the 6-month window. The highest accuracy was found for clinical models using levels of CysC, NAG, and KIM-1 (6-month AUCs: 0.80, 0.81, and 0.80, respectively). Figure S2 shows the temporal patterns of eGFR and NAG in several individual patients from our cohort, together with their corresponding individual survival probabilities as estimated by the joint model. The figure shows that each time an additional measurement is performed in the patient, the individual survival probability is updated. Specifically, rising marker levels and worsening prognosis can be seen in the example patients who ultimately reached the composite endpoint versus stable or decreasing marker levels and more favorable prognosis in the example patients who stayed event-free. We have shown that in patients with CHF both glomerular function (as assessed by repeatedly measured creatinine, eGFR, and CysC), and tubular function (as assessed by repeatedly measured urinary NAG and KIM-1) deteriorate over time preceding clinical outcome. Importantly, patient-specific trajectories of all glomerular markers dynamically predicted the event, and CysC was the strongest predictor. Similarly, patient-specific trajectories of urinary NAG and KIM-1 indicated progression of tubular damage in patients who later suffered adverse events. No clear associations were found between repeatedly measured plasma or urinary NGAL and the event. Therefore, the current study does not justify its use for clinical prediction in patients with CHF. Our findings confirm that renal function is an indivisible component of HF, and that it is clinically relevant for the monitoring of stable patients with CHF. Importantly, our results show that temporal changes in renal function remain predictive for clinical outcome despite controlling for NYHA class, cardiac markers and other clinical features, which suggests that renal dysfunction may drive adverse clinical outcomes independently of cardiac dysfunction. In addition, the results demonstrate the predictive value not only of GFR levels (single value or cumulative effects), but also of GFR slope. These findings are supported by other studies.4Damman K. Valente M.A. Voors A.A. et al.Renal impairment, worsening renal function, and outcome in patients with heart failure: an updated meta-analysis.Eur Heart J. 2014; 35: 455-469Crossref PubMed Scopus (574) Google Scholar, 10Damman K. Masson S. Hillege H.L. et al.Tubular damage and wors