Urinary L-FABP, a Biomarker for Disease Progression, Dialysis and Overall Mortality in Chronic Kidney Disease

1. Introduction

Affecting approximately 10-13% of the population worldwide1 the epidemic of chronic kidney disease (CKD) has soared through the ranking of causes of global mortality, rising from 27th place in 1999 to 18th in 20102. Established as a major public health concern3, CKD is diagnosed, classified and monitored based on estimates of renal function and creatinine clearance. Using the Kidney Outcomes Quality Initiative (KDOQI) classification4 CKD can be defined in stages (1-5) according to disease severity, based on estimates of glomerular filtration rate (eGFR) derived using either the Modification of Diet in Renal Disease Study Group (MDRD)5 methodology, or more recently the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI)6 equations, and the presence or absence of proteinuria. The risk of disease progression increases at advanced CKD stages with 8.9% of CKD stage 4 patients and 39.1% of CKD stage 5 patients likely to show a decline in eGFR of 5ml/min/1.73m2/year or more7. However, only 0.5% of the general population progress to these advanced stages of CKD; additionally, the majority of patients with CKD have stage 3 disease (7-9% of the general population; 80-90% of all CKD) and will never progress to end-stage disease8. Nevertheless, 6% of CKD stage 3 patients show a decline in eGFR of 5ml/min/1.73m2/year9. Globally, 50-400 per million population of CKD patients will become dependent on renal replacement therapy per year10, alongside an increased risk of cardiovascular disease11,12. In this context early identification of patients at risk of CKD progression is imperative.

Although eGFR is a useful measure for stratifying disease severity, it provides little information about the potential risk of disease progression13. Proteinuria has been an important prognostic marker in glomerular pathologies14, especially diabetic nephropathy15, but is limited as a biomarker for CKD progression and response to treatment16. A number of alternative biomarkers or risk prediction equations based on creatinine and uPCR (Kidney Failure Risk Equation17) have been proposed as predictors of disease progression13. Urinary liver-type fatty acid binding protein (uL-FABP) is one such biomarker with the potential for predicting CKD progression. Expressed by the proximal renal tubule during ischemia and under oxidative stress18,19, uL-FABP is already established as a reliable biomarker of acute kidney injury20 and, in a number of small studies, has been linked with CKD progression in diabetic21,22 and non-diabetic23 patients.

Our aim was to assess whether uL-FABP levels can predict a decline in renal function, and thus disease progression, in patients with CKD attending outpatient clinics in two large tertiary centres in the United Kingdom.

2. Materials and Methods

This was a 2-yr prospective follow-up study of participants with CKD stages 1-5 (not receiving renal replacement therapy) attending outpatient nephrology clinics in 2 large tertiary university renal centres in the northwest of the UK. The study received approval from the Yorkshire and Humber-Leeds West, NHS Research Ethics Committee (15/YH/0516) and was adopted on the NIHR clinical research network portfolio (CRN ID: 20484). Adult patients, with capacity to consent, and receiving their care in the participating centres were eligible to participate in the study and were approached during their attendance to clinic. Participation in the study was voluntary. Informed written consent was obtained prior to enrolment.

Definitions

A number of definitions of CKD progression were considered, however based on the definitions used by NICE²⁴ and KDIGO²⁵ , for purposes of this study at the time, CKD progression was defined as either

A) at Yr 1 follow up i) decline in eGFR by the equivalent of 5ml/min/1.73m², ii) increase in serum creatinine by 10%, iii) initiation of renal replacement therapy for CKD 5 or

B) at yr 2 follow up i) decline in eGFR by the equivalent of 6 ml/min/1.73m², ii) increase in serum creatinine by 20% or iii) initiation of renal replacement therapy.

The combined outcome of initiating renal replacement therapy or death was also evaluated.

Study population size

Katoh et al ²⁶ suggest that a cut-off of 19 μg/g.cr for urinary L-FABP carries 100% sensitivity and 81.8% specificity for the detection of contrast-induced AKI. Their mean study population uL-FABP level was 59.8± 45.6 μg/g.cr. Shingai et al²⁷ used a detection cut-off 8.4 μg/g.cr base on the study of 420 healthy volunteers by Kamijo-Ikemori et al ²². Although published data from Kamijo et al ²³ suggest that a uL-FABP level of 17.4 μg/g.cr can be used as a cut-off for monitoring progression in CKD the use of L-FABP as a predictor of progression of CKD is unclear and remains under investigation. Although the literature failed to inform any power calculation, we believe that based on previous study population sizes a population of 500 CKD participants would provide us with an adequate sample size to evaluate the potential of uLFABP as a CKD biomarker.

Data Collection

The past medical history and medication lists were collected directly from participants and through their electronic medical records. These were used to calculate the Davies’ Comorbidity Score for all participants²⁸. Prospective follow-up data on CKD was collected from patient electronic records at 1-yr and 2-yr from enrolment.

Blood Sampling, eGFR calculation and CKD classification

Serum samples were collected as part of participants’ routine attendance to clinic and were analysed for urea and creatinine at the individual participating centres’ NHS biochemistry laboratories. The eGFR was calculated using the MDRD⁵ formula. Patients were categorised in CKD stages according to KDOQI⁴. The difference in creatinine and eGFR over the 1^st and 2^nd years of the study was taken to be the difference in the values of these parameters at the data collection time points and at baseline.

Urine sampling and analysis

A 20ml urine sample was provided as part of the participants’ routine clinical care. 10ml was analysed by the participating centres’’ local NHS biochemistry laboratories to measure urinary protein, creatinine and protein: creatinine ratio (uPCR). The remaining 10ml sample was stored at -80^o pending measurement of uL-FABP levels using an Enzyme-Linked Immunosorbent Assay (ELISA).

ELISA analysis for the measurement of uL-FABP levels

Urine samples were processed in batches. Samples were thawed overnight and centrifuged at 2400 rpm at room temperature for 10 minutes to separate and remove impurities. UL-FABP levels were quantified using a commercially available ELISA kit (CMIC Co Ltd, Tokyo, Japan) as per manufacturers' instructions. All samples were assessed in duplicate within the same ELISA plate. Intra-assay coefficient of variation was 5.5%. Ten samples with detectable levels of uL-FABP were analysed separately to assess inter-plate assay variability. The inter-assay coefficient of variance was 7.8%. Calculations of uL-FABP levels were adjusted for urinary creatinine concentration obtained from the uPCR data that was measured within the same sample.

Statistical analysis

The demographic and biochemical characteristics of participants were analysed using descriptive epidemiology. Categorical variables were presented as frequencies and percentages. Continuous variables with normal distribution were presented as means and standard deviation (SD) while variables with skewed distribution were presented as medians with minimum and maximum values. Normality of distribution was assessed using the Shapiro-Wilk method. The relationship between baseline urine uL-FABP levels, serum Creatinine, eGFR and urine PCR was explored using stepwise linear regression modelling adjusted for age and sex. Skewed outcome variables underwent logarithmic transformation. Longitudinal analysis was performed from the change in eGFR, increase in serum creatinine, CKD progression and combined renal mortality. Participants that received kidney transplantation between baseline and 1-yr were removed from year one analysis unless they had initiated dialysis prior to transplantation. The same was true for participants receiving renal transplantation between years 1 and 2 for year two analysis. The association between uL-FABP and uPCR and the change in serum creatinine and eGFR was assessed using stepwise linear regression modelling adjusted for age and sex. The demographic characteristics and biochemical profiles of participants that exhibited CKD progression were compared to those without disease progression using group comparison. Categorical variables were compared using Pearson’s Chi² with observation of the standardised residuals as post hoc analysis. Continuous variables were skewed in distribution, so differences between the assessed groups were analysed using Kruskall-Wallis test. The predictive potential of uL-FABP was evaluated using logistic regression modelling adjusted for age, sex and proteinuria. The predictive sensitivity and specificity of uL-FABP were evaluated through ROC curve analysis and was compared to that of uPCR and uPCR in combination with uL-FABP. The interaction between uL-FABP and uPCR and changes in creatinine was also explored using heat map assessment. Subgroup analysis stratified by individuals' CKD stage at baseline and absence of proteinuria was performed using the same logistic regression model utilised for the entire study population. Combined outcome analysis was performed using Cox regression modelling using 8mcg/gCr of uL-FABP as a cut-off for a significant level. Models were adjusted for age, sex and uPCR. Statistical analyses were performed by IBM SPSS Statistics, version 23 (IBM Corp., USA).

3. Results

A total of 641 patients were recruited into the study. Of those recruited the data of 583 participants were analysed at baseline, while 484 were available in year one and 335 in year two. Participants’ exclusions and dropout reasons are shown in Figure 1 and Figure 2. The demographic characteristics and biochemical profiles of these participants are shown in Table 1 and Figure 3.

At baseline uL-FABP levels correlated to those of uPCR (beta= 0.383, CI 0.305; 0.457, p= <0.001, R² =0.142). uL-FABP also correlated to serum creatinine (beta= 0.379, CI 0.303; 0.455, p= <0.001, R² =0.145) and eGFR (beta= -0.380, CI -0.454; -0.302, p= <0.001, R² =0.143). This association was independent of proteinuria, age and sex (linear regression models shown in Table 2).

Year 1&2 analysis

The median percentage increase in serum Creatinine between baseline and year 1 was 3.5% (-6.9; 18.8) with a median decrease in MDRD eGFR of 1 ml/min (-2.0; 5.0).

Both uL-FABP (Beta= 0.140, CI 0.045; 0.203, p=0.002, R² =0.020) and uPCR (Beta= 0.202, CI 0.102; 0.258, p <0.001, R² = 0.041) correlated with increases in serum creatinine. UPCR (beta= 0.101, CI 0.010; 0.159, p= 0.026, R² = 0.010) predicted a decrease in eGFR over 1 year but this relationship was not observed with uL-FABP (beta= -0.045, CI -0.112; 0.037, p= 0.324, R² = 0.002).

By the end of year 1, 208 participants (43%) showed progression of CKD defined as a reduction in eGFR by 5ml/min, an increase in serum creatinine by 10% or initiation of renal replacement therapy. Their characteristics are described in Table 3. Both uL-FABP and uPCR predicted CKD progression (logistic regression model adjusted for sex and age) (Table 4).

By the end of year 2, 153 participants (42%) showed CKD progression defined as a reduction in eGFR by 6ml/min, an increase in creatinine by 20% and start of renal replacement therapy. uL-FABP did not correlate with the increase in creatinine over 2 years (beta= 0.041, p=0.464) or the decrease in eGFR (beta= -0.030, p=0.602) over the same period. uPCR did correlate with both of these outcomes in a linear model adjusted for age, sex and uL-FABP (ΔCreatinine: beta=0.310, p<0.001, ΔeGFR; beta= 0.253, p<0.001). Both L-FABP and uPCR predicted CKD progression (logistic regression model adjusted for sex and age). The results are shown in Supplementary Table 1.

Sensitivity & Specificity

The area under the curve for sensitivity and specificity of uL-FABP in predicting CKD progression was calculated using ROC curve analysis as 0.623 (CI 0.572; 0.675, p<0.01) while that for uPCR was 0.706 (CI 0.658; 0.753, p<0.01) (Figure 4a and 4b). Heat map analysis (Figure 5) also showed that uL-FABP in the absence of significant proteinuria can predict modest increases in serum creatinine (10% at 1 year and 20% at 2 years). The combined elevation of uL-FABP and uPCR had the highest predictive power for increases in creatinine at both 1 year (≤20% increase; Figure 5a) and 2 years (≤40% increase; Figure 5b).

Mortality and RRT initiation risk Analysis

Cox regression for the combined outcomes of death and initiation of RRT over the 2-year study period showed that the cumulative risk for patients with high uL-FABP levels (using uL-FABP of 8 as the differentiator) was significantly higher than those with low levels (p<0.001) in a model adjusted for age, sex and uPCR (Figure 6)

4. Discussion

We describe the findings of a prospective 2-year follow-up study exploring the potential of uL-FABP as a biomarker for disease progression in CKD. This is the largest study of a CKD population using uL-FABP. Our baseline analysis demonstrated an association between uL-FABP levels and disease severity, as demonstrated both by raised creatinine and reduced eGFR but also proteinuria. The presence of elevated levels of uL-FABP at baseline appears to predict CKD progression both at 1 and 2 years of follow-up. Although proteinuria appears to have a higher sensitivity and specificity than uL-FABP alone in predicting CKD progression, our ROC curve analysis would suggest that the best prediction for sensitivity and specificity is offered by the combination of the 2 biomarkers. As apparent from the heat map analysis, patients with elevated uL-FABP levels were more likely to experience a rise in serum creatinine both at 1 and 2 years of follow-up. In addition to disease progression, elevated uL-FABP carried a 24% higher risk of RRT and overall mortality independent of the presence of proteinuria.

Serum creatinine levels, and by extension creatinine-based estimations of GFR, might portray a reliable insight into CKD severity but their overall value in predicting disease progression is limited¹³. uPCR and its risk prediction equations (KFRE¹⁷) remain the gold standard in identifying individuals more likely to progress^16,29–31 but our findings suggest that the absence of proteinuria does not exclude progressive pathology. In these individuals, the presence of elevated uL-FABP might be a better predictor of disease progression.

Urinary PCR and L-FABP biomarkers are likely to represent different pathological processes involving the kidney. While uPCR tends to be predominantly glomerular in its origin at early stages of CKD¹⁴, uL-FABP is produced by the renal tubules in response to ischemia and oxidative stress. A number of studies have linked urinary levels of L-FABP with tubular injury both in the acute and chronic phases of kidney disease. In health, L-FABP is produced in the proximal renal tubule and binds fatty acids reabsorbed by the renal tubular cells enabling their transposition into the tubular mitochondria¹⁹. Under ischaemic conditions, this process is compromised by lipid peroxidation products binding to L-FABP instead leading to increase secretion of L-FABP into the urine¹⁹. Whether this process is pathological or protective is unclear. All studies to date have shown that high levels of uL-FABP are generally linked with adverse outcomes. This has been predominantly in the setting of acute kidney injury^32–36, perioperative ischaemic^37,38, contrast-induced acute tubular injury³⁹ and peri-haematopoetic stem cell transplantation associated acute kidney injury²⁷. In the setting of CKD, studies have been few, with limited study populations^21,23,40,41. Some studies have investigated specific CKD groups such as those with diabetes or cardiovascular disease^21,23,41. They do all, however, conclude that uL-FABP is a useful biomarker in CKD¹⁹ and its potential should be explored further. Our results are suggestive that in addition to an additive predictive value to that of uPCR, uL-FABP might play a role in risk stratification of patients, especially with negligible or absent proteinuria. Our attempts to characterise this group was limited by a smaller subgroup size (Supplementary Table 2).

Although the association between uL-FABP and ESKD/RRT has been noted before in a study of diabetic Pima Indians⁴¹, this is the first time that such an association has been shown in a large population of unselected patients with CKD.

Another potential benefit when considering uL-FABP as a potential biomarker of CKD progression in the clinical setting is its potential availability as a semi-quantitative point-of-care test. The technology can be adapted for screening the CKD population in clinics.

The main limitation of our study was the lack of repeated measurements of uL-FABP during the follow-up period. The limited number of studies in the field made the design of this study and any population size estimation difficult. Efforts were made to minimise selection biases during the recruitment of participants. Participants were randomly approached for recruitment during their attendance at routine CKD clinical appointments. Laboratory analysis was also blinded to the personal information of participants.

In summary, uL-FABP appears to be a highly sensitive and specific biomarker of renal dysfunction and an effective predictor of CKD progression. A rapid assay of elevated uL-FABP in routine CKD care could improve rapid diagnosis and feasibility in clinical practice, especially in CKD patients with minimal or no proteinuria.

Further studies involving large registries are needed to evaluate the role of uL-FABP in risk stratification of CKD patients and disease progression. Future research should also investigate interventions that can modify the expression of the biomarker and help improve the prognosis of chronic kidney disease.