Urine Injury Biomarkers and Risk of Adverse Outcomes in Recipients of Prevalent Kidney Transplants: The Folic Acid for Vascular Outcome Reduction in Transplantation Trial

Participant Characteristics

The mean age (±SD) of the random subcohort was 51 (±9) years old, and eGFR was 46 (±18) ml/min per 1.73 m². Median (interquartile range) urine albumin-to-creatinine ratio (ACR) was 24.5 (9.5–104.7) μg/mg, urine NGAL was 20.2 (8.2–51.3) ng/ml, urine KIM-1 was 658 (319–1364) pg/ml, urine IL-18 was 29.1 (11.5–62.5) pg/ml, and urine L-FABP was 6.1 (3.0–17.6) ng/ml (Supplemental Table 1, column 1). Participants with the highest quartile of urine NGAL/creatinine were more likely to be women, have a lower eGFR, and have a higher urine ACR (Supplemental Table 1a). Participants with the highest quartile of urine KIM-1/creatinine were more likely to be women and have higher urine ACR and less likely to have received a living donor kidney (Supplemental Table 1b). Participants with the highest quartile of urine IL-18/creatinine were more likely to be women and have higher urine ACR. Graft vintage was longer among participants in the lowest quartile of urine IL-18/creatinine, and the biomarker distributions differed by country (Supplemental Table 1c). Participants with the highest quartile of urine L-FABP/creatinine were more likely to use sirolimus and have higher BP, lower eGFR, and higher urine ACR (Supplemental Table 1d). Overall, only 3% of KTRs (16 of 489 from the random subcohort) were in the highest quartile of all four urine biomarkers (as illustrated in Figure 1, with the Venn diagram showing overlap).

Participants who had a cardiovascular event or a graft failure or died were more likely to be from the United States, have older graft vintage, have higher systolic BP, have lower eGFR, and have higher urine ACR (Table 1).

Correlations among Measures of Kidney Function and Urine Injury Biomarkers

Among participants in the subcohort, eGFR level was negatively correlated with urine NGAL and urine L-FABP (Supplemental Figure 1, Table 2). Urine albumin was positively correlated with all four biomarkers. There were moderately strong positive correlations among urine NGAL, KIM-1, IL-18, and L-FABP; the strongest correlation was between urine NGAL and L-FABP (correlation coefficient =0.47; P<0.001) (Supplemental Figure 1, Table 2).

Urine Injury Biomarkers and Risk of Cardiovascular Events

Overall mean follow-up time was 3.9 (±1.6) years. Among participants in the subcohort, the crude event rate was 2.7 per 100 person-years for cardiovascular events. The cumulative incidence of cardiovascular events was greater for participants in the highest quartile of urine NGAL/creatinine, KIM-1/creatinine, and L-FABP/creatinine but not IL-18/creatinine and seemed fairly constant over the follow-up period (Figure 2).

In the full study population, in unadjusted analyses, there was a statistically significant association of urine NGAL/creatinine, urine KIM-1/creatinine, and urine L-FABP/creatinine (but not IL-18/creatinine) with risk of cardiovascular events (Table 3). These associations remained statistically significant after adjustment for participant demographics and other cardiovascular risk factors. With additional adjustment for urine ACR, only the association between urine NGAL/creatinine and cardiovascular events remained statistically significant. For each log higher urine NGAL/creatinine, the risk of cardiovascular events was 24% higher (adjusted hazard ratio [HR], 1.24; 95% confidence interval [95% CI], 1.06 to 1.45) (Table 3).

When participants with prevalent CVD at study entry were excluded, none of the urine biomarkers were associated with risk of cardiovascular events (although the 95% CIs were wider) (Supplemental Table 2).

Urine Injury Biomarkers and Risk of Graft Failure

Within the subcohort, the crude graft failure event rate was 2.7 per 100 person-years. The cumulative incidence of graft failure was greatest among participants in the highest quartile of each urine biomarker/urine creatinine (Figure 3).

Among participants in the full study population, in unadjusted models, there was a statistically significant association of higher urine NGAL/creatinine, urine KIM-1/creatinine, urine IL-18/creatinine, and urine L-FABP/creatinine with increased risk of graft failure (Table 4). These associations remained robust after adjustment for participant demographics and other risk factors. After adjustment for urine ACR, only the association between urine NGAL/creatinine and graft failure remained statistically significant. For each log increase in urine NGAL/creatinine, there was a 40% (adjusted HR, 1.40; 95% CI, 1.16 to 1.68) higher risk of graft failure (Table 4).

Urine Injury Biomarkers and Risk of All-Cause Mortality

Within the subcohort, the crude mortality rate was 3.3 per 100 person-years. The cumulative incidence of all-cause mortality was greatest among participants in the highest quartile of each urine biomarker/creatinine (Figure 4).

Among participants in the full study population, in unadjusted analyses, all four urine injury biomarkers/creatinine, considered as continuous measures or quartiles, were associated with an increased risk of all-cause mortality (Table 5). These associations remained statistically significant after adjustment for demographic characteristics and other cardiovascular risk factors. After adjustment for urine ACR, the associations of urine NGAL/creatinine (HR, 1.44; 95% CI, 1.26 to 1.65 per log increase), urine KIM-1/creatinine (HR, 1.29; 95% CI, 1.03 to 1.61 per log increase), and urine IL-18/creatinine (HR, 1.25; 95% CI, 1.04 to 1.49 per log increase) remained statistically significant (Table 5).

When participants with CVD at study entry were excluded, in multivariable analysis including adjustment for urine ACR, the association between NGAL/creatinine and all-cause mortality remained statistically significant but not for urine KIM-1/creatinine or urine IL-18/creatinine (although the 95% CIs were wider) (Supplemental Table 2).

Sensitivity Analyses: Analyses of Urine Biomarkers Not Standardized to Urine Creatinine

In the full study population, in multivariable models that examined raw urine biomarker concentrations (i.e., not standardizing them to urine creatinine concentration) and after adjusted for urine ACR, urine NGAL was associated with increased risk of cardiovascular events, graft failure, and all-cause mortality(Supplemental Table 3), similar to the main analysis (Tables 3–5). In multivariable models, urine IL-18 was significantly associated with increased risk of all-cause mortality (Supplemental Table 3), similar to the main analysis, whereas the association between urine KIM-1 and mortality was attenuated and no longer statistically significant (Supplemental Table 3).

Sensitivity Analyses: Composite Outcomes

To assess whether our conclusions regarding cardiovascular events or graft failure could be biased by competing risk of death, we repeated our multivariable models examining two composite outcomes: cardiovascular events or all-cause mortality and graft failure or all-cause mortality. Overall, reassuring results were obtained (Supplemental Table 4).

Post Hoc Exploratory Analyses: Interactions by Sex

Given the substantial differences in biomarker levels in men and women (Supplemental Table 1), we tested for interaction by sex in exploratory post hoc analyses. Across four biomarkers and three outcomes, we found a statistically significant interaction by sex only in the association of IL-18/creatinine with all-cause death (P value =0.01). In stratified analyses, the association was stronger in men.

Study Population

This report is a post hoc analysis of the FAVORIT Trial (clinicaltrials.gov: NCT00064753), a multicenter, double–blind, randomized, controlled clinical trial conducted to determine whether lowering homocysteine levels with vitamin therapy reduced the rate of pooled arteriosclerotic cardiovascular outcomes. The FAVORIT Study protocol was approved by the Institutional Review Boards at the participating institutions, and all participants provided written informed consent.

The design of the trial and the primary results have been described in detail elsewhere.^54–56 In total, 4110 KTRs were enrolled from August of 2002 to January of 2007 and randomized to either a standard multivitamin with high doses of folic acid, vitamin B6, and vitamin B12 or a multivitamin containing low doses of vitamin B6 and vitamin B12 with no folic acid. Men and women ages 35–75 years old who were at least 6 months postkidney transplant were enrolled at 30 transplant centers in the United States, Canada, and Brazil. Entry criteria included elevated serum homocysteine level (≥11 μmol/L for women and ≥12 μmol/L for men) and stable kidney function defined by an estimated creatinine clearance ≥30 ml/min in men and ≥25 ml/min in women. Follow-up contacts occurred every 6 months through January 31, 2010 to obtain study-related outcomes through June 24, 2009. The primary outcome was pooled incident or recurrent CVD events. Graft failure and all-cause mortality were secondary outcomes. As reported previously, there was no significant difference between treatment groups for primary or secondary outcomes.⁵⁷

Urine Injury Biomarkers

Urine NGAL, urine KIM-1, urine IL-18, and urine L-FABP were measured at a single laboratory at Cincinnati Children’s Hospital in previously unthawed urine samples obtained at baseline and stored at −80°C. All assays were performed in June of 2013. The urine NGAL ELISA was performed using a commercially available assay (NGAL ELISA Kit 036; Bioporto, Grusbakken, Denmark) that specifically detects human NGAL.⁵⁸ The intra–assay coefficient of variation (CV) was 2.1%, and interassay CV was 9.1%. The urine KIM-1 ELISA was constructed using commercially available reagents (Duoset DY1750; R&D Systems, Inc., Minneapolis, MN) as described previously.⁵⁹ Intra- and interassay CVs for KIM-1 were 2% and 7.8%, respectively. Urine IL-18 and L-FABP were measured using commercially available ELISA kits (Medical & Biologic Laboratories Co., Nagoya, Japan and CMIC Co., Tokyo, Japan, respectively). CVs for IL-18 and L-FABP were 7.3% and 6.1% (interassay) and 7.5% and 10.9% (intra-assay), respectively. There were 128 participants whose samples were below the lower limit of detection (LLD; 10.3 pg/ml) for IL-18 and 19 below the LLD (3 ng/ml) for L-FABP. For these participants, values halfway between zero and the LLD were assigned. Urine creatinine was assayed using the modified Jaffe method. Intra-assay CV was 2%, and interassay CV was 4%.

To correct for the effects of urine concentration and dilution, all biomarker levels were standardized by urine creatinine concentration in our main analysis. Because of the skewed distribution, urine biomarker/urine creatinine was log transformed. To avoid assumptions of linearity, biomarkers were modeled in quartiles as well as continuously.

Study Outcomes

Three outcomes were considered in our analysis: adjudicated CVD events, graft failure, and all-cause mortality. The CVD outcome was a composite of cardiovascular death, myocardial infarction, resuscitated sudden death, and stroke. Each of these outcomes was centrally reviewed and adjudicated by the FAVORIT Clinical Endpoints Committee.^54–57 The same committee also reviewed medical records for occurrence of unstable angina and urgent coronary revascularization procedures to identify additional patients with myocardial infarction. Graft failure was defined as initiation of dialysis ascertained by local study staff. Study staff also identified deaths from review of medical records, regular participant contact, and contact with family. For each outcome, participants were censored at either the time of last follow-up visit or the end of study period. For the outcomes of CVD and graft failure, participants were also censored at time of death.

Covariates

At time of study enrollment (baseline), demographic characteristics (age, sex, race, and country of origin [United States, Canada, or Brazil]); smoking status (current, former, or never); past medical history by self-report (baseline CVD and diabetes mellitus); transplant characteristics (living donor kidney and time since transplant); physical examination findings (body mass index [BMI] and systolic and diastolic BPs); and laboratory measurements (total cholesterol, HDL cholesterol, and triglycerides) were obtained. Race was recorded as white, black, or other, including mixed race. LDL cholesterol was estimated using the Friedewald equation at triglyceride levels <400 and measured directly in 82 participants with triglyceride levels >400 mg/dl. Baseline BP was the average of two measurements. Diabetes was defined by the use of insulin or oral hypoglycemic medications or participant self-report. Prior history of CVD was determined by self-report at baseline and included prior myocardial infarction, coronary artery revascularization, stroke, carotid arterial revascularization, abdominal or thoracic aortic aneurysm repair, and/or lower extremity arterial revascularization or amputation above the ankle. BMI was calculated using the formula weight (kg)/height (m)². GFR was estimated from serum creatinine using the Chronic Kidney Disease Epidemiology Collaboration 2009 equation.⁶⁰ Serum creatinine was measured using an alkaline picrate kinetic method on an Olympus AU 400e (Olympus America Inc., Center Valley, PA) instrument that was calibrated to an isotope dilution mass spectrometry traceable standard. Urine albumin and creatinine were measured in spot urine samples to calculate urine ACR. Urine albumin was measured using an immunoturbidimetric assay. Intra-assay CV was 2%, and interassay CV was 4%.

Study Design

To evaluate the association between urine kidney injury biomarkers and outcomes, we performed a case-cohort study. Both patients and nonpatients are drawn from the entire cohort, with patients identified at study completion. Participants can contribute to more than one type of case; for example, one participant can be included because of both CVD disease and death. Members of the subcohort are randomly selected from the entire cohort and can include individuals who are also selected as patients. After excluding 104 participants with missing urine samples at baseline (n=53) or missing key covariates (n=51), all patients with cardiovascular events (n=291), graft failure (n=257), and all-cause mortality (n=359) were included. We also sampled a random subcohort (n=489) of FAVORIT Trial participants, with a final study population of 1027 for this analysis.

Statistical Methods

Chi-squared test, ANOVA, and Kruskal–Wallis test were used as appropriate to compare baseline data across quartiles of each urine injury biomarker/urine creatinine in the random subcohort of 489 participants. We also reported characteristics of study participants who did and did not have an outcome of interest. We generated a Venn diagram to evaluate overlap among participants in the highest quartiles of each urine biomarker. Among the random subcohort participants, we then calculated the Spearman correlation coefficients and generated scatterplots between eGFR, urine albumin, urine NGAL, urine KIM-1, urine IL-18, and urine L-FABP.

Kaplan–Meier survival analysis was used to estimate the cumulative incidence of each outcome among subcohort study participants by quartile of each urine injury biomarker/urine creatinine. To take into account the case-cohort study design, weighted Cox proportional hazards regression^61,62 was used to examine the association between baseline urine injury biomarkers and time to study outcomes. We examined the Schoenfeld residual plots from model 2 for each biomarker and outcome to check the proportional hazard assumption (which did not seem to be violated). Models were adjusted for baseline age, sex, race, randomly assigned treatment, country from which participants were recruited (United States, Canada, or Brazil), diabetes, smoking, graft vintage, donor type, diastolic BP, systolic BP, BMI, and eGFR. Models for the outcomes of CVD and death were additionally adjusted for history of CVD and lipids. In a final model, we further adjusted for urine ACR to test whether the association of each urine injury biomarker with our outcome of interest was independent of this established urine biomarker.

We performed three sensitivity analyses. First, we examined the association of each urine injury biomarker with risk of cardiovascular events and all-cause mortality, excluding participants with prevalent CVD at study entry. Second, we repeated our analyses without standardizing each urine biomarker to urine creatinine concentration. Third, to evaluate for possible competing risk, we repeated our multivariable models, examining the association of each urine biomarker with risk of (1) cardiovascular events or all-cause mortality and (2) graft failure or all-cause mortality.

All analyses were performed using SAS 9.3, SAS 9.4, and JMP.