Renal tubular damage and worsening renal function in chronic heart failure: Clinical determinants and relation to prognosis (Bio‐SHiFT study)

Abstract Background It is uncertain that chronic heart failure (CHF) patients are susceptible to renal tubular damage with that of worsening renal function (WRF) preceding clinical outcomes. Hypothesis Changes in tubular damage biomarkers are stronger predictors of subsequent clinical events than changes in creatinine (Cr), and both have different clinical determinants. Methods During 2.2 years, we repeatedly simultaneously collected a median of 9 blood and 8 urine samples per patient in 263 CHF patients. We determined the slopes (rates of change) of the biomarker trajectories for plasma (Cr) and urinary tubular damage biomarkers N‐acetyl‐β‐d‐glucosaminidase (NAG), and kidney‐injury‐molecule (KIM)‐1. The degree of tubular injury was ranked according to NAG and KIM‐1 slopes: increase in neither, increase in either, or increase in both; WRF was defined as increasing Cr slope. The composite endpoint comprised HF‐hospitalization, cardiac death, left ventricular assist device placement, and heart transplantation. Results Higher baseline NT‐proBNP and lower eGFR predicted more severe tubular damage (adjusted odds ratio, adj. OR [95%CI, 95% confidence interval] per doubling NT‐proBNP: 1.26 [1.07‐1.49]; per 10 mL/min/1.73 m2 eGFR decrease 1.16 [1.03‐1.31]). Higher loop diuretic doses, lower aldosterone antagonist doses, and higher eGFR predicted WRF (furosemide per 40 mg increase: 1.32 [1.08‐1.62]; spironolactone per 25 mg decrease: 1.76 [1.07‐2.89]; per 10 mL/min/1.73 m2 eGFR increase: 1.40 [1.20‐1.63]). WRF and higher rank of tubular injury individually entailed higher risk of the composite endpoint (adjusted hazard ratios, adj. HR [95%CI]: WRF 1.9 [1.1‐3.4], tubular 8.4 [2.6‐27.9]; when combined risk was highest 15.0 [2.0‐111.0]). Conclusion Slopes of tubular damage and WRF biomarkers had different clinical determinants. Both predicted clinical outcome, but this association was stronger for tubular injury. Prognostic effects of both appeared independent and additive.

). 19 Patients were ambulatory and stable, that is, they had not been hospitalized for HF in the past 3 months. The study was approved by the medical ethics committees, Since 95% of the study population had heart failure with reduced ejection fraction (HFrEF), in this study, we focused on the HFrEF patients (n = 250).
All patients were evaluated by research physicians, who collected information on HF-related symptoms, New York Heart Association class (NYHA) class, and performed a physical examination and collected samples. Information on HF etiology, ejection fraction, cardiovascular risk factors, comorbidities, and treatment were retrieved from hospital records. Study follow-up visits were predefined and scheduled tri-monthly (±1 month), with a maximum of 10 study follow-up visits. All patients were also routinely followed at the outpatient clinic by treating physicians who were blinded for biomarker data. Occurrence of rehospitalizations for HF, myocardial infarction (MI), percutaneous coronary intervention (PCI), coronary artery bypass grafting (CABG), arrhythmias, cerebrovascular accident (CVA), cardiac transplantation, left ventricular assist device (LVAD)-placement, and mortality was recorded in electronic case-report forms, and associated hospital records and discharge letters were collected. A clinical event committee, blinded for biomarker data, reviewed hospital records and discharge letters, and adjudicated the study endpoints.
The composite endpoint comprised cardiac death, cardiac transplantation, LVAD implantation, and hospitalization for the management of acute or worsened HF, whichever occurred first. Cardiac death was defined as death from MI or other ischemic heart disease (implantable cardioverter defibrillator (ICD)-10: I20-I25), death from other heart disease including HF (I30-I45 and I47-I52), sudden cardiac To assess patient-specific slopes (rates of change over time) of biomarker trajectories, we performed joint modeling (JM) of linear mixed-effects (LME) and Cox regression models. 22 The LME models estimate the individual biomarker trajectory based on repeated measurements and also correct for biomarkers' sampling variability (for details, see Figure S2). 23 The JM then combines LME with the Cox regression model to adjust the biomarker trajectory for different follow-up durations between patients. 23  Covariates that were found to be different across categories of tubular damage with P < .10 (see Table 1) were entered into a multivariable ordinal regression model and those were: age, NYHA class, diabetes, use of cardiac resynchronization therapy (CRT), diastolic blood pressure, NT-proBNP, cTnT, and eGFR. b *Represents only covariates with P-value <.05 were presented in the Covariates that were found to be different between WRF patient and non-WRF patients with P < .10 (see Table 2) were entered into a multivariable binary regression model and those were: diastolic blood pressure, NT-proBNP, hs-cTnT, eGFR, urinary NAG, prior myocardial infarction, loop diuretics and MRAs doses.
F I G U R E 1 Distributions of slopes of renal biomarkers prior to study endpoints. Notes: X-axis displays number of patients who experienced the event (red) and those who did not (blue), Y-axis displays the estimated slopes on the continuous scale, where positive numbers correspond to increasing slopes and negative numbers correspond to decreasing slopes. t test was used test the average difference between patient with and without event were entered into a multivariable logistic regression model applying proportional odds ordinal regression or binary logistic regression. Data on all variables were complete, except for systolic blood pressure, which was missing in <5% of patients and for which imputations were applied using patients' clinical and outcome data.

| RESULTS
Seventy-four percentage of HFrEF patients experienced incline in either tubular damage biomarker during follow-up. Of those, 44% of patients had both tubular biomarkers rising prior to the endpoint or last sampling moment. Figure 2A shows that endpoint-free rates were lowest when both tubular damage biomarkers were increased, followed by the rates when either marker was increasing (P for trend <.001). HR were almost four times higher in patients in whom Fifty-eight percent of HFrEF patients experienced incline in Cr levels during follow-up. Figure 2B shows that patients with increasing plasma Cr slope had lower endpoint-free rates than their counterparts (P = .018). The HR in these patients were also significantly higher than in those in whom Cr remained stable or decreased (adj. HR 1.9 [1.1-3.4], P = .027).
Eighteen percent of HFrEF patients experienced deteriorating patterns of both urinary tubular biomarkers as well as Cr, while 31% of patients had at least one tubular biomarker rising without a change in Cr, and only 10% of patients had neither biomarker worsening during follow-up (for details, see Figure 2C). Figure 2C displays that when tubular damage markers were stable or improving, Cr incline did not affect endpoint-free rates. However, if either NAG or KIM-1 slope increased, endpoint-free rates decreased. Finally, the lowest endpoint-free rates were in patients who had increasing slopes of all three renal biomarkers (P for trend <.001).
F I G U R E 2 Kaplan-Meier survival curves stratified by slopes of renal biomarkers. Notes: Shown are Kaplan-Meier (KM) curves for the cumulative event-free survival of the composite of HF-rehospitalization, cardiac death, LVAD placement, and heart transplantation. A, KM curves are stratified by whether both NAG and KIM-1 slopes were decreasing/stable (blue); either NAG or KIM-1 slope was increasing (red); or both NAG and KIM-1 slopes were increasing (green); B, KM curves are stratified by whether Cr slope was decreasing/stable (blue) or increasing (red). C, KM curves are stratified by whether slopes of all three renal biomarkers were decreasing/stable (blue); NAG and KIM-1 slopes were decreasing/stable, but creatinine slope was increasing (red); either NAG or KIM-1 slope was increasing but creatinine slope was decreasing/stable (green); either NAG or KIM-1 slope was increasing, and creatinine slope was increasing (orange); and slopes of all three biomarkers were increasing (purple). Hazard ratios (HR) were adjusted for age, sex, diabetes, atrial fibrillation, NYHA class, diuretics, systolic blood pressure, eGFR (only for tubular damage biomarkers), NT-proBNP, and hs-cTnT To our best knowledge, this study is the first to identify clinical determinants of progressive renal tubular damage in CHF. Of note, these determinants differ from those found for WRF, which strengthens the recommendation that glomerular and tubular compartment should be jointly assessed. This study also displays that patients in whom both renal compartments deteriorate during outpatient follow-up have the lowest endpoint-free survival.
Renal function may act as a "barometer" of the severity of CHF. 27,28 However, because of the multifactorial nature of cardiorenal interactions, merely assessing the glomerular function may be suboptimal for decision-making. Our study confirms this and provides additional evidence for the notion that each aspect of the kidney (glomerular and tubular) provides incremental prognostic information, and together they may further identify higher-risk CHF individuals. These kidney-specific signals may therefore help physicians to better and timely target medical therapy before the future event occurs.
Based on our findings, we could speculate that "renoprotective" treatment targeted at the tubules may be even more effective than treatment simply aiming at improving only eGFR or Cr values. To this end, we have previously found that higher ACE-inhibitor/ARBs doses during follow-up were associated with less renal tubular damage together with less cardiac impairment (as assessed by NT-proBNP and troponin levels). 29 However, interventional studies on these tubular damage markers are needed to provide definite answers in this matter.
Our findings suggest that patients who have reduced eGFR already at baseline are more susceptible to tubular injury during follow-up than those with higher baseline eGFR (ie, greater renal functional capacity).
This phenomenon may potentially be attributed to the "work-overload" in residual nephrons to compensate renal function in patients who had fewer functioning nephrons available. 30 Compensatory hyperfiltration in the rest of nephrons may eventually exceed tubular adaption to hypoperfusion leading to tubulointerstitial hypoxic damage. 31 Higher doses of loop diuretics and lower MRA doses were identified in patients with WRF and are supported by previous studies. 1,34 WRF was found to be associated with higher baseline eGFR which is also supported by several previous studies. [35][36][37] However, this finding is inconsistent with the general opinion that WRF (defined as delta Cr >0.03 mg/dL) occurs more frequently in CHF patients that have impaired GFR already at baseline. 38 One explanation for this discrepancy may be that closer monitoring of patients who already had impaired GFR could have also increased the likelihood of finding WRF in these patients, 34 and particularly if sampling was not fixed but left at the discretion of the treating physician. 39 As for our study, the observations were made using more than twice as many repeated measurements as in each of the previous studies, samples were collected prospectively at fixed time intervals defined by the study protocol. This further strengthens our suggestion that WRF should not be disregarded in CHF patients with relatively intact renal function.
Several limitations merit consideration. First, this study lacked direct GFR measurement. Second, we cannot comment on the effects of glomerular permeability on clinical outcomes since proteinuria was not measured. Third, causal inference is limited by the observational nature of our study. Although trials on this subject are still lacking, the repeated-measures design of this study allows for stronger claims of true associations than previous studies do.

| CONCLUSIONS
Slopes of tubular damage and WRF biomarkers had different clinical determinants. Both predicted clinical outcome, but this association was stronger for tubular injury. Prognostic effects of both appeared independent and additive. These findings are of particular interest since in current clinical practice the degree of tubular injury usually remains undetermined.

ACKNOWLEDGMENTS
This work was supported by the Jaap Schouten Foundation, the Foreest Medical School, and the Coolsingel Foundation ('Stichting Coolsingel'). Plasma and urine assays were acquired from HaemoScan BV, Groningen, The Netherlands.