Urinary Biomarkers and Kidney Injury in VA NEPHRON-D: Phenotyping Acute Kidney Injury in Clinical Trials

Abstract

Rationale & Objective:

Urinary biomarkers of injury, inflammation and repair may help phenotype observed AKI in clinical trials. We evaluated differences in biomarkers between those randomized to monotherapy or to combination RAAS blockade in VA NEPHRON-D, acknowledging an increased proportion of observed AKI in the combination arm.

Study Design:

Longitudinal analysis.

Setting & Participants:

A sub-study of the VA NEPHRON-D trial.

Predictor:

Primary exposure is treatment arm (combination [RAASi] vs monotherapy). AKI is used as a stratifying variable.

Outcomes:

Urinary biomarkers including albumin, EGF, MCP-1, YKL-40, and KIM-1.

Analytical Approach:

Biomarkers measured at baseline and at 12 months in trial participants were compared between treatment groups and by AKI. AKI events occurring during hospitalization were predefined safety end points in the original trial. The results were included in a meta-analysis with other large chronic kidney disease (CKD) trials to assess global trends in biomarker changes.

Results:

In 707 participants followed for a median of 2.2 years, AKI incidence was higher in the combination (20.7%) versus the monotherapy group (12.7%; RR: 1.64; 95% CI, 1.16–2.30). Compared with the monotherapy arm, urine biomarkers at 12 months were either unchanged (MCP-1 −3% [−13% to 9%] p_adj=0.8; KIM-1 −10% [−20%, 1%] p_adj=0.2; EGF −7% [−12 to −1%] p_adj=0.08) or lower (albuminuria −24% [, −37% to −8%] p_adj=0.02; YKL-40 −44% [, −58% to - 25%] p_adj<0.001) in the combination arm. Pooled meta-analysis demonstrated reduced albuminuria in the intervention arm across 3 trials and similar trajectories in other biomarkers.

Limitations:

Biomarker measurement limited to two time points independent of AKI events.

Conclusions:

Despite increased risk of sCr–defined AKI, combination RAASi therapy was associated with unchanged or decreased urinary biomarkers at 12 months. This suggests a possible role for kidney biomarkers to further characterize kidney injury in clinical trials.

Introduction

Acute kidney injury (AKI) is among the most common in-hospital diagnoses and is associated with increased length of stay¹, hospital readmissions², chronic kidney disease (CKD)³, and increased risk of cardiovascular mortality and events^4,5. Diagnosis of AKI necessitates adjustment of important medications, frequently delays or alters clinical care^6–8, and is a common safety outcome or adverse event in therapeutic clinical trials of CKD⁹.

The most recent KDIGO guidance from 2012 defines AKI as an abrupt rise in serum creatinine (sCr) >1.5x baseline, increase of sCr by 0.3 mg/dL, or presence of oliguria/anuria and is thus highly dependent on serum creatinine (sCr) measurement¹⁰. Despite widespread clinical reliance on sCr, much evidence raises concerns regarding its reliability as a marker of true intrinsic kidney injury^10–17. Briefly, sCr is variable between and within individuals, demonstrates a 48 to 72-hour delay in rise after AKI, and does not rise until significant loss of kidney function has ocurred^15–20. Additionally, sCr fails to distinguish the site of kidney injury or clinical context, such as with transient changes in eGFR with RAAS-inhibition or SGLT-2i inhibition that portend favorable long-term outcomes. Given these limitations, alternatives or adjuncts to sCr are crucial for clinical advancement in AKI diagnosis and prognostication. Extensive research has shown promise in novel biomarkers such as KIM-1^21–23, MCP-1^24,25, YKL-40^26,27 and EGF^28,29 which identify tubular injury, tubular function or health, fibrosis or adaptive repair, and kidney inflammation.

In 2013, The Veterans Affairs Nephropathy in Diabetes (VA NEPHRON-D) targeted U.S Veterans with established diabetes and CKD who were deemed most likely to benefit from RAAS blockade using combination angiotensin-converting enzyme inhibition and angiotensin receptor blockade. The trial was stopped early due to increased rates of serious adverse events, hyperkalemia and AKI in the combination group³⁰. A secondary analysis of hospitalized AKI events in VA NEPHRON-D showed that amongst participants who developed AKI, those in the monotherapy group had lower rates of kidney recovery, higher mortality, and higher likelihood of reaching the primary composite endpoint compared to those in the combination therapy group^18,31. This finding suggests that elevations of sCr observed in the trial were not universally representative of true pathologic kidney injury.

In the present study, we evaluated urinary biomarkers to understand how kidney tubular cells responded to RAAS combination therapy and their association with AKI in the VA NEPHRON-D trial. Four urinary biomarkers were examined, which included KIM-1 (proximal tubular damage marker), MCP-1 (marker of intrarenal inflammation), YKL-40 (marker of tubular repair), and EGF (distal tubular damage marker), in addition to albuminuria (marker of glomerular damage). The principal aim was to characterize urinary biomarkers amongst the VA NEPHRON-D population, specifically comparing groups randomized to receive combination therapy (losartan plus lisinopril) versus monotherapy (losartan only) and between trial participants who developed AKI as an adverse event and those who did not³². We pooled results from the VA NEPHRON-D trial with results from the Action to Control Cardiovascular Risk in Diabetes (ACCORD) and the Systolic Blood Pressure Intervention Trial (SPRINT) clinical trials, which also had treatment arms associated with higher proportion of AKI. The ACCORD and SPRINT trials randomized participants with and without diabetes, respectively, to intensive versus standard systolic blood pressure lowering. Independent studies have examined AKI events and changes in urine biomarkers of kidney tubule cell injury in both trials^33–35. We hypothesize that urinary biomarkers of injury, inflammation and repair will not differ between those randomized to monotherapy or to combination RAAS blockade in VA NEPHRON-D, despite an increased proportion of observed AKI in the combination arm.

Methods

Study Population

VA NEPHRON-D was a multi-center, prospective, double-blind randomized controlled trial examining the efficacy of the combination treatment of an angiotensin converting enzyme inhibitor (ACEi; lisinopril) with an angiotensin receptor blocker (ARB; losartan) vs. the mono therapy of an ARB alone, on the progression of CKD in participants with proteinuric diabetic kidney disease. Full study design and primary trial results have been previously described and are briefly reviewed in Item S1^30,32. All participants provided informed consent and protocols were approved by the institutional review board of each study site³⁰. The present work is a secondary analysis of the VA NEPHRON-D trial.

Biomarker Measurements

Urine samples were collected at time of randomization (before titration of lisinopril or placebo) and again at the 12-month visit. Samples were processed in accordance with standard operating procedures of the CKD Biomarkers Consortium³⁶ and stored at −80°C.

We measured urinary biomarkers (KIM-1, MCP-1, YKL-40, EGF, albumin and creatinine) at baseline and 12-month visits. Urinary KIM-1, MCP-1, YKL-40 and EGF levels were measured in duplicate using a Luminex multiplex assay (Luminex Corporation, Austin, TX). To improve precision, results were averaged and the mean value was used for analysis. Biomarkers were measured en bloc with samples from baseline and follow-up timepoints on the same plate for each participant, minimizing measurement bias between timepoints. For KIM-1, MCP-1, YKL-40, and EGF, the intra-assay CVs were 4.8%, 6.6%, 7.1%, and 6.6% and inter-assay CVs were 7.8%, 8.6%, 9.7%, and 9.2% respectively. Urine creatinine was measured by the Jaffe reaction using a Randox Imola instrument with CV of <4%. Personnel performing biomarker measurements were blinded to participant data.

Outcome Ascertainment

In the original VA NEPHRON-D trial, AKI was a pre-defined safety outcome: serious adverse events requiring hospitalization or occurring during hospitalization in accordance with safety guidelines³⁰. The time from randomization to AKI event or last follow-up was recorded. In the present study (a secondary analysis of VA NEPHRON-D) outcomes of interest were biomarker levels on follow-up and change in urinary biomarkers from baseline. In a supplementary analysis, development of AKI before the time of sample collection at 12 months was included as a stratifying variable. For eGFR calculations, the Modification of Diet in Renal Disease was used as is consistent with the original trial³⁰.

Statistical Analysis

Means of continuous and frequencies of categorical variables were calculated for the overall sample and separately for combination and monotherapy treatment groups. Continuous variables were compared with t-tests or Mann-Whitney U tests and categorical variables were compared with chi-squared or Fisher’s exact tests, as appropriate. AKI events in treatment groups were compared using a chi-squared test and the effect size further quantified with the sample relative risk and its associated Wald 95% confidence interval. To correct for skewness, biomarker measurements were log₂-transformed in all analyses. In univariate comparisons of biomarkers between treatment groups, urinary biomarkers were indexed to urinary creatinine. In a supplementary analysis, the geometric means of urinary biomarkers not indexed to urinary creatinine were compared. There were minimal missing data so a complete case analysis was performed.

In accordance with prior studies^33,34, a linear mixed-effects model was used to estimate percent change in ACR and the urinary biomarkers KIM-1, MCP-1, YKL-40, and EGF, with assigned treatment arm in VA NEPHRON-D as exposure. Log₂-adjusted biomarker values at baseline and 12 months were used as the outcome, while assigned treatment arm, time of measurement in years, and their interaction were used as covariates. Additionally, linear and quadratic terms for urine creatinine (UCr) were included as time-varying covariates when modeling the urinary biomarkers. Each patient was modeled with an unstructured covariance structure for two repeated measures of each biomarker, one at each time point. The formula is shown below:

log₂(biomarker) = Treatment + Time + Treatment*Time + UCr + UCr² + Intercept; where treatment was coded as 1=combination therapy, 0=monotherapy and time as 1=12 months and 0=baseline.

Means and 95% confidence intervals of biomarker measurements at each time point and treatment group were calculated by taking the antilog of the least squares mean. Geometric mean ratios of biomarker measurements between the combination and monotherapy groups were calculated by taking the antilog of the difference in least squares means between groups. To account for multiple comparisons, Benjamini-Hochberg adjusted p-values are presented³⁷. Five urinary biomarkers at each of two time points were considered in the adjusted p-values comparing the two treatment arms, for a total of ten. The same model was used in a sensitivity analysis where the urine creatinine terms were removed and the biomarkers were indexed for urine creatinine³⁷. Supplementary analyses report geometric mean ratios of biomarker measurements between the 12-month and baseline time points, as well as the ratio of ratios between the treatment arms and time points, for both methods of modeling urine creatinine. In another supplementary analysis, we examined participants who did and did not develop AKI before the time of sample collection at 12 months. Percent changes in urine creatinine indexed biomarker from baseline to 12 months were calculated as [12-month value/baseline value - 1] and compared with Mann-Whitney U tests across treatment groups, stratifying by those who did or did not develop AKI before 12 months. A final supplementary analysis examines the geometric mean ratios between therapy arms in an unadjusted analysis, without modeling UCr.

To understand the observed biomarker results in the primary analysis, we identified and abstracted data from two large clinical trials and evaluated intensive versus standard blood pressure regimens (ACCORD and SPRINT)^{33–35,38,39}. These studies represent large randomized hemodynamic interventions with available urinary biomarker data collected in a similar manner to VA NEPHRON-D, and are described in Item S1. The urinary biomarkers for which data were available from all three trials were albumin, KIM-1, MCP-1 and YKL-40 The ratio of the intervention to standard of care arm (from VA Nephron D, SPRINT, and ACCORD) were reported as least-squares means from linear mixed models adjusted for UCr (both linear and quadratic terms). Results were reported at 1 year for Nephron D and SPRINT and at 2 years for ACCORD. For the meta-analysis, results were back transformed to the coefficient parameter estimate and its standard error. The R function metamean from the meta package was used to pool results and calculate heterogeneity statistics. Inverse-weighting random effects meta-analysis was performed to pool estimates across studies while accounting for within-study and between-study heterogeneity. Between-study variance was estimated using the Hartung-Knapp method. Pooled estimates were back transformed to report as the ratio between intervention arm and standard of care arm and were only reported in the presence of low heterogeneity (I-squared<50%). All analyses were performed using SAS 9.4 and R 4.2.2. Figure 1 was produced with the ggplot2 R package⁴⁰.

Figure 1.

Comparison of 12-Months Biomarkers Between ACE-i/ARB Combination Therapy and Mono Therapy, Urine Creatinine Indexed

Biomarker distributions are represented with boxplots with the inter-quartile range (IQR) plotted as a rectangle with median as the middle line, 25^th percentile as the lower horizontal line and 75^th percentile as the upper horizontal line. Whiskers extend to the farthest value within (25^th percentile −1.5*IQR) and (75^th percentile+1.5*IQR) with points outside the range plotted individually. Values for medians and 25^th and 75^th percentiles can be found in Supplemental Table 2.

Urinary biomarkers are indexed to urine creatinine.

Results

Baseline Characteristics of Cohort

We identified 707 participants of the VA NEPHRON-D trial who had urine samples collected at both randomization and the 12-month visit (Figure S1). Excluded participants were more likely to be black with other characteristics being similar to the parent trial (Table S1). Baseline eGFR was calculated with the four-variable Modification of Diet in Renal Disease formula, to be consistent with the eGFR formula used in the original trial. The average baseline eGFR was 57 (SD 19) mL/min/1.73 m² and median baseline albuminuria was 805 (25^th percentile 420, 75^th percentile 1650) mg/g. A large proportion of the study population had one or more significant comorbidities: 17.5% of participants had coronary artery bypass surgery and 20.1% of participants were active smokers (Table 1). Participants included in the analysis were nearly universally male, characteristic of trials conducted in the VA system. There were no differences in baseline characteristics between groups randomized to monotherapy or combination therapy.

Table 1.

Baseline Characteristics of Cohort

	Overall (n=707)	Combination Therapy (n=352)	Monotherapy (n=355)
Age (years)	65 (7)	65 (8)	65 (7)
Male	703 (99%)	349 (99%)	354 (100%)
White	137 (19%)	66 (19%)	71 (20%)
Diabetic Retinopathy	310 (44%)	156 (44%)	154 (43%)
Macrovascular Event	414 (59%)	216 (61%)	198 (56%)
Smoker	142 (20%)	79 (22%)	63 (18%)
Coronary Artery Bypass Surgery	124 (18%)	71 (20%)	53 (15%)
Total Cholesterol (mg/dl)	156 (41)	157 (41)	156 (40)
HDL (mg/dl)	38 (11)	37 (11)	39 (12)
LDL (mg/dl)	81 (31)	80 (29)	83 (32)
Systolic BP (mm Hg)	136 (16)	135 (16)	136 (15)
Diastolic BP (mm Hg)	72 (10)	72 (10)	72 (10)
Triglycerides (mg/dl)	202 (157)	213 (162)	190 (151)
eGFR (mL/min/1.73 m2)	57 (19)	56 (19)	58 (19)
ACR (mg/g)	805 (420, 1650)	836 (420, 1550)	780 (425, 1749)

Data presented as mean (SD), n (%), or median (25^th percentile, 75^th percentile) in the case of ACR

^*All groups include 707 participants with exception of Total Cholesterol (n=700), HDL (n=699), LDL (n=682), Triglycerides (n=698), eGFR (n=705)

^†t-test for continuous variables, chi-squared for categorical, Mann-Whitney U for ACR, Fisher’s exact test for Male sex

Comparison of Biomarkers by Treatment Group Throughout VA NEPHRON-D Trial

The VA NEPHRON-D trial investigators evaluated serum creatinine and potassium every 3 months throughout the trial period of 2.2 years and urinary biomarkers were measured at 12 months. The mean eGFR at 12 months was lower in the combination therapy arm (48.9 mL/min/1.73 m², SD 19.6) relative to the monotherapy arm (52.3 mL/min/1.73 m², SD 20.0) (Figure 1). Urine ACR and all UCr-indexed urinary biomarkers at 12-months were either reduced or not different in the combination arm compared to the monotherapy arm (Figure 1, Table S2). Urine ACR was lower at 12 months in the combination arm (609 (247, 1270) mg/g) as compared to the monotherapy arm (675 (360, 1560) mg/g). YKL-40 at 12 months was lower in the combination arm (0.39 (0.14, 1.44) ng/mg) relative to the monotherapy arm (0.64 (0.22, 2.74) ng/mg). EGF was also lower in the combination therapy arm (1010 (755, 1372) pg/mg) as compared to the monotherapy arm (1068 (746, 1640) pg/mg). KIM-1 and MCP-1 were similar between treatment arms (Figure 1, Table S2).

To further assess differences in urinary biomarkers between the combination therapy and monotherapy groups, we applied a mixed-effects model to estimate the relative differences of each urinary biomarker at baseline and after 12 months of treatment. In this initial model, we adjusted for linear and quadratic UCr. There were no differences in baseline urine ACR or urinary biomarkers (KIM-1, MCP-1, YKL-40, EGF) (p>0.05; Table 2). However, after 12 months of therapy, albuminuria was −24% lower (95% CI −37%, −8%) and YKL-40 −44% lower (95% CI −58%, −25%) in the combination therapy arm relative to the monotherapy arm. There were no differences between arms in 12 months changes in urine KIM-1, EGF and MCP-1 (Table 2). Results were similar in the sensitivity analysis indexing for UCr rather than adjusting for UCr (Table S3). To explore the interaction between treatment and time, we examined the difference of differences for time and therapy arm from the same models (Table S4). Significant differences in biomarker change from baseline to Year 1 were found in albuminuria, YKL-40, and EGF.

Table 2.

Relative Difference of Traditional and Novel Biomarkers of Kidney Injury in Combination vs Monotherapy Arms, Urine Creatinine Adjusted

Biomarker†	Time point	Combination Therapy Arm (n=352) (LS mean* (CI))	Monotherapy Arm (n=355) (LS mean* (CI))	Ratio Combination Therapy Arm vs. Monotherapy Arm	P-value‡
ACR*, mg/g	Baseline	800 (717, 892)	858 (770, 957)	−7% (−20%, 9%)	0.6
	Year 1	536 (469, 613)	705 (618, 805)	−24% (−37%, −8%)	0.02
Urine KIM-1, ng/mL	Baseline	2.63 (2.43, 2.85)	2.68 (2.48, 2.9)	−2% (−12%, 10%)	0.8
	Year 1	2.64 (2.43, 2.88)	2.94 (2.7, 3.19)	−10% (−20%, 1%)	0.2
Urine MCP-1, pg/mL	Baseline	233 (217, 251)	237 (220, 255)	−2% (−11%, 9%)	0.8
	Year 1	257 (237, 278)	264 (243, 286)	−3% (−13%, 9%)	0.8
Urine YKL-40, ng/mL	Baseline	0.29 (0.24, 0.35)	0.35 (0.29, 0.42)	−18% (−38%, 7%)	0.3
	Year 1	0.39 (0.32, 0.47)	0.69 (0.56, 0.84)	−44% (−58%, −25%)	<0.001
Urine EGF, pg/mL	Baseline	878 (849, 909)	883 (853, 914)	−1% (−5%, 4%)	0.8
	Year 1	810 (777, 845)	868 (832, 905)	−7% (−12%, −1%)	0.08

^*Data presented are antilogs of least-squares means (95% CI) from linear mixed models. Negative ratios indicate levels in the combination therapy were lower than the monotherapy arm. In this analysis, biomarkers are outcome and therapy arm is exposure.

^†All biomarkers were log₂-tranformed. Novel biomarkers were adjusted for linear and quadratic urine creatinine concentrations.

^*12 month Urine ACR was missing for 44 participants in the combination therapy arm and 35 participants in the monotherapy arm.

^‡Benjamini-Hochberg adjusted p-values to account for multiple testing (10 comparisons).

Clinical Descriptors of Participants Who Developed AKI Prior to 12 Month Sample Collection and Comparison of Biomarkers Among Participants with and without AKI

Consistent with the original VA NEPHRON-D trial³⁰, combination therapy participants in our study had a higher proportion of AKI events (20.7%) than the monotherapy group (12.7%) over 2.2 years of trial follow up. Throughout the trial period, the relative risk of developing AKI in the combination therapy group was 1.64 (CI 1.16, 2.30) compared to the monotherapy group.

Clinical data reflect the characteristics of AKI events that occurred prior to 12 months (Table S5). Data was available for 37 of 38 participants who developed AKI prior to sample collection. The majority of AKIs were Stage 1 in severity (76%, n=28), with the remainder being Stage 2 (11%, n=4) or Stage 3 (14%, n=5). The most common cause of AKI was listed as prerenal azotemia (73%, n=27). Most common predisposing factors to AKI were diuretics (68%, n=25), medications (62%, n=23) and volume depletion (57%, n=21). Most participants experienced return of creatinine to 25% of outpatient baseline (86%, n=32) with zero participants that remained on dialysis and only one related fatality.

In a supplementary analysis, we examined participants who developed AKI before the time of sample collection at 12 months (n=38) and those who did not (n=669; Table S6). The number of participants who experienced AKI was small (n=38), but nonetheless revealed differences in urinary biomarkers between the combination and monotherapy arms. The percent changes of UCr-indexed biomarkers from baseline to 12 months were compared between treatment arms in the AKI and non-AKI subgroup. Amongst the 667 participants who did not experience AKI prior to the time of biomarker measurement at 12 months, combination therapy was associated with a greater reduction in albuminuria (−30% vs −9%, p=0.002), a lesser increase in YKL-40 (32% vs 81%, p=0.004), and a greater decrease in EGF (−9% vs 3%, p=0.002) as compared to baseline. In the participants who developed AKI, there was a greater reduction in albuminuria (−56 vs −3%, p=0.02) and a lesser increase in YKL-40 (21% vs 234%, p=0.02).

When comparing the geometric mean ratios of unadjusted urinary biomarkers across therapy arms (Table S7) to the ratios from adjusted mixed models (Table 2) there is a consistency in trends. In both analyses, there is a decrease in albuminuria and YKL-40 excretion in the combination arm relative to the monotherapy arm, while the ratios of KIM-1, MCP-1, and EGF do not indicate differences between treatment arms.

Meta-analysis of Urinary Biomarkers in CKD Clinical Trials Evaluating Therapies with Known Hemodynamic Effects on eGFR

In meta-analysis of three trials studying intensive vs. standard blood pressure regimens, urinary biomarkers including albuminuria, YKL-40, KIM-1, MCP-1, YKL-40 did not reflect a pattern of injury associated with hemodynamic change (Figure 2). In all 3 trials, eGFR was significantly lower between treatment and standard arms by pooled estimate of −8%. The ratio of ACR was significantly lower by −24 to −32% with intensive therapy in all trials. Consistent with our findings, biomarkers suggestive of kidney injury were not increased in intensive treatment arms relative to standard treatment arms. The decrease in urinary YKL-40 was directionally consistent in all studies.

Figure 2.

Effect of Trial Therapies on Kidney Function and Urinary Biomarkers in Large Clinical Trials

*Urinary biomarkers from NEPHRON-D at 12-months. ACCORD trial participants with urinary biomarkers at 24-months. SPRINT trial participants with urinary biomarkers at 12-months. The Ratio presented is the transformed least-square means (95% CI) from linear mixed models. Negative ratios indicate levels in the intervention arm were lower than the standard of care arm. The squares represent the mean value and error bars represent the 95% confidence interval. The error bars not crossing zero demonstrates statistical significance. Pooled estimates and heterogeneity estimates are shown.

Discussion

In this study of a large randomized clinical trial evaluating dual vs. single RAAS blockade, we found that participants randomized to combination therapy experienced more AKI events and greater mean reductions in eGFR, yet did not exhibit a pattern of urinary biomarker changes indicative of tubule cell damage and injury. Indeed, participants randomized to the combination therapy arm demonstrated no changes or reductions in urine biomarkers at 12 months. Furthermore, differences in biomarkers suggest that participants in the monotherapy arm were more likely to display elevations in biomarkers of tubule cell injury including albuminuria and YKL-40. A meta-analysis of urinary biomarkers in VA NEPHRON-D, ACCORD, and SPRINT showed that treatment arms associated with greater risk for AKI were not associated with elevations of biomarkers of tubule cell injury, consistent with our findings. In fact, intensive treatment arms were associated with reduction of albuminuria across trials.

The pattern of biomarker changes in the combination RAAS blockade arm suggested a lack of tubular damage and possible signal of kidney protection. Principal among these effects is the consistent decrease in albuminuria observed in the combination group compared to the monotherapy group. Regarding biomarkers of inflammation and repair, a significant reduction in the absolute value of YKL-40 was seen in the combined treatment arm. YKL-40 is a protein which is up-regulated in response to ischemia/reperfusion injury and functions to limit apoptosis in tubular epithelium. Biopsy and clinical cohorts have demonstrated that YLK-40 upregulation follows significant kidney injury of any cause, and elevated urinary or blood levels of YK-40 have been negatively associated with kidney prognosis^41,42. The reduction of YKL-40 in the combination therapy arm therefore suggests absence of kidney injury and adaptive repair. There were no significant differences in the absolute value or relative change of MCP-1, which suggests no increased ongoing intrarenal inflammation in the combination arm at the time of sample collection^24,43. KIM-1 is not expressed in healthy kidney tubular cells and may be a sensitive and persistent marker of proximal tubular damage associated with tubulointerstitial injury and proteinuric nephropathy^16,44,45. KIM-1 was elevated in both arms without significant relative or absolute difference between treatment groups. EGF, a marker of distal tubule structural integrity and known protective cytokine, reportedly repairs epithelial barrier function in damaged tubules^46,47. Higher EGF, therefore, is associated with preserved eGFR. In the main analysis, EGF was modestly lower in the combination therapy arm relative to monotherapy arm, suggesting less repair in the combination therapy arm. However, there was no difference in EGF across treatment arms in the sub-group analysis of participants who developed AKI prior to the 12 months trial time point for biomarker collection. These inconsistent findings may be related to AKI events that do not usually affect the distal tubule or a complex physiology given a population with baseline diabetic kidney disease and merits further study.

A meta-analysis of urinary biomarker data from this analysis and the SPRINT and ACCORD trials showed that our findings were consistent in other large cohorts which had elevations in sCr after participants are randomized to therapies with known kidney hemodynamic reductions. In the SPRINT and ACCORD trials, blood pressure therapy was used to target a goal SBP <120mmHg compared with SBP <140mmHg^{34,35,48–50}. Like VA NEPHRON-D, a higher rate of AKI was observed in the intensive arm. However, KIM-1, MCP-1 and YKL-40 were not different in the intensive arms as compared to standard arm. Additionally, a protective signal of reduced albuminuria persisted across trials.

Large cohort studies investigating the kinetics of urinary biomarkers after AKI are relevant to this work. To our knowledge, the Assessment, Serial Evaluation and Subsequent Sequelae in Acute Kidney Injury (ASSESS-AKI) study represents the strongest prospective data supporting the utility of biomarkers in the recovery period after the AKI event. ASSESS-AKI showed that patients experiencing AKI develop sustained increase in proteinuria and biomarkers⁵¹, and that elevations in biomarkers 3 months and 12 months after the AKI episode are associated with increased events including incident CKD, progression of CKD, heart failure and all cause death^25,52. Important retrospective work in the Chronic Renal Insufficiency Cohort (CRIC) study has shown mixed results. In parallel papers, McCoy et. al showed greater elevations of plasma biomarkers (TNFR1, TNFR2, KIM-1) among those experiencing AKI compared to those without AKI⁵³, but urinary biomarkers in the same group were unchanged⁵⁴. Finally, recent data using present VA NEPHRON-D cohort support the association between higher absolute levels of biomarkers at 1 year and kidney function decline and mortality⁵⁵. Collectively, an evolving body of literature supports urinary and plasma biomarkers as an important signal of kidney damage during and after AKI event, though more work is needed to define the kinetics of individual biomarkers.

Strengths of our analyses include the randomized trial design, evaluation of biomarkers both at baseline and after 1 year of therapy, the ability to compare changes separately in those who experienced AKI, and our ability to compare biomarker changes across cohorts using meta-analysis. In addition, biomarkers were measured in duplicate at each time point and averaged, improving precision.

Our study has important limitations. First, urinary biomarker data were collected only at baseline and the 12-month visit, and therefore measurement of biomarkers is not directly proximate to AKI events. As noted in the supplementary analysis, only 38 patients had an AKI event before collection of the 1-year urine sample. Thus, changes in biomarkers should be viewed as reflective of after an AKI event only amongst the AKI subset. Similarly, the single study-pre-determined sample collection time point may conceal dynamic changes in urinary biomarkers which may have occurred prior to or normalized before the 12-month collection. To compare differences, our supplemental analysis of biomarker changes is stratified by presence of AKI by 12 months, which subjects the analysis to post-randomization effects and limits the strength of findings. Second, while evaluation of 4 distinct biomarkers is a key strength, similar analyses of large clinical trials have examined different sets of biomarkers, limiting our direct comparison of results for other biomarkers. Third, our meta-analysis is limited with only 3 included studies. Furthermore, while we are assured by the robust design of clinical trials included in this analysis, we consider the possibility that post-randomization confounding may have caused regression to the mean and led to under-represented differences between groups. We suspect this would be most likely to occur in our analysis of data from the SPRINT trial, where arms were determined by BP target values rather than by randomization to therapy. Finally, we did not have longer follow-up from our cohort of participants as the trial was stopped early after 2.2 years.

Overall, this study suggests that combination RAAS-i therapy in VA NEPHRON-D was not universally associated with tubule cell injury, despite being associated with AKI episodes and greater decreases in eGFR. Indeed, the top line results of the overall VA NEPHRON-D trial for the kidney composite endpoint (secondary endpoint) trended favorably for dual RAAS blockade, but marginally missed statistical significance (HR 0.78, 95% CI 0.58–1.05). These findings in VA NEPHRON-D pertain to those with chronic kidney disease exposed to RAAS blockade. However, the findings are similar to studies of urinary biomarkers in other clinical trials where therapies alter hemodynamics of the kidney. These data demonstrate the heterogeneity of kidney response to hemodynamic therapy, and limitations of defining kidney injury by sCr alone. Further research is warranted to investigate the kinetics, predictive and associative nature of urinary biomarkers and acute and subacute episodes of GFR decline. Our work supports a role for urine albumin and suggests that novel biomarkers such as YKL-40 and KIM-1 may offer additional phenotyping to assess tubular health and differentiate benign cases of hemodynamic change from true intrinsic kidney injury. Careful consideration of the nature of acute eGFR decline will likely alter the landscape of trials and has broad implications for standard of clinical care of hypertension, CKD and diabetes.

Supplementary Material

1

Figure S1. Participant flow chart

Item S1. Supplemental Methods

Table S1. Comparison of participants included or excluded from analysis due to the availability of urine samples

Table S2. Urinary Biomarkers at Baseline and 12 Months, Urine Creatinine Indexed

Table S3. Relative Difference of Traditional and Novel Biomarkers of Kidney Injury in Combination vs Monotherapy Arms, Urine Creatinine Indexed

Table S4. Relative Difference of Traditional and Novel Biomarkers of Kidney Injury in 12M vs Baseline Time Points, Urine Creatinine Adjusted and Indexed

Table S5. AKI Details for participants that developed AKI prior to sample collection at 12 months

Table S6. Percent Change‡ in Urine Creatinine Indexed Biomarkers from Baseline AKI Status at 12 Months

Table S7. Unadjusted Urinary Biomarkers at Baseline and 12 Months

Plain Language Summary.

The VA NEPHRON-D trial investigated inhibition of the RAAS hormonal axis on kidney outcomes in a large population of diabetic CKD patients. The trial was stopped early due to increased events of serum creatinine-defined acute kidney injury in the combination therapy arm. Urine biomarkers can serve as an adjunct to serum creatinine in identifying kidney injury. We found that urinary biomarkers in the combination therapy group were not associated with a pattern of harm and damage to the kidney, despite the increased number of kidney injury events in that group. This suggests that serum creatinine alone may be insufficient for defining kidney injury and supports further exploration of how other biomarkers might improve identification of kidney injury in clinical trials.