Elevated urinary fatty acid-binding protein 4 level predicts future renal dysfunction and poor prognosis in Japanese patients with diabetes: a longitudinal cohort study

ABSTRACT

Background

Fatty acid-binding protein 4 (FABP4) is an adipokine secreted from adipocytes and macrophages and is also expressed in injured, but not normal, glomerular endothelial cells. Elevated levels of urinary FABP4 (U-FABP4) have been reported to be associated with glomerular damage and increased proteinuria.

Methods

The associations of levels of U-FABP4 at baseline with future events including renal dysfunction defined by a 30% decline in estimated glomerular filtration rate (eGFR) and all-cause death were investigated in 660 patients with diabetes (type 1/2, 57/603).

Results

During a follow-up period (median: 62 months), 90 patients (13.6%) developed renal dysfunction, and 66 patients (10.0%) died (median follow-up period 65 months). Kaplan–Meier survival curves showed that there were significant differences in cumulative incidences for a 30% decline in eGFR and all-cause death in patients divided by the tertiles of U-FABP4 level. Furthermore, multivariable Cox proportional hazard models with a restricted cubic spline showed that hazard ratios for a 30% decline in eGFR and all-cause death increased with a higher level of logarithmically transformed (log) U-FABP4 after adjustment for age, sex, type of diabetes, body mass index, current smoking habit, duration of diabetes, comorbidities of hypertension and dyslipidemia, eGFR, and the categorical classification of urinary albumin-creatinine ratio. The addition of log U-FABP4 to traditional risk factors significantly increased the discriminatory capacities for renal dysfunction in net reclassification improvement and integrated discrimination improvement and for all-cause death in NRI.

Conclusion

U-FABP4 is a predictive biomarker for future renal dysfunction and poor prognosis in patients with diabetes.

Graphical Abstract

Open in new tabDownload slide

INTRODUCTION

The presence of diabetes has a major impact on healthy life expectancy as well as the occurrence of diabetes-related complications including cardiovascular and renal events [1, 2]. Recent advances in treatment strategies for achieving recommended glycemic management, blood pressure, and serum lipids have resulted in a reduction in the proportion of patients with diabetes [3]. However, the problems of diabetes-related cardiovascular, renal, and metabolic complications have not been fully overcome yet [4, 5]. Among the complications, diabetic kidney disease (DKD), defined as chronic kidney disease caused by diabetes with or without proteinuria/albuminuria, remains a major complication, leading to end stage renal disease (ESRD) and various cardiovascular events [6]. Therefore, early diagnosis and appropriate management of DKD are urgent issues for improving the healthy life expectancy of patients with diabetes [3].

Fatty acid-binding proteins (FABPs) are soluble proteins with a molecular weight of about 14-15 kDa that bind hydrophobic ligands such as long-chain fatty acids [7–9]. FABPs have been reported to function as chaperones of fatty acids, and some FABPs act as extracellularly secreted bioactive molecules [10, 11]. Among the FABPs, FABP1 is expressed in proximal tubular epithelial cells of the human kidneys [12, 13]. It has been reported that urinary FABP1 (U-FABP1) can be a marker of damaged proximal tubular epithelial cells in DKD [14]. However, since FABP1 is not expressed in glomeruli of the kidneys [12, 13], U-FABP1 does not directly reflect glomerular damage. On the other hand, albuminuria, a sensitive biomarker for an early stage of DKD, reflects abnormalities of both glomerular filtration and reabsorption in proximal tubular cells [15].

FABP4, classically known as adipocyte FABP, is mainly expressed in adipocytes and macrophages [7–10] and it can be a therapeutic target for insulin resistance and atherosclerosis [16, 17]. Previous studies showed that FABP4 is secreted from cells via a non-canonical pathway [18] and that circulating levels of FABP4 are increased in various cardiovascular and metabolic diseases including obesity-induced insulin resistance [19], diabetes [20], hypertension [21], dyslipidemia [22, 23], hyperuricemia [24], steatotic liver disease [25], coronary artery disease [22], carotid atherosclerosis [26], and chronic heart failure [27]. It has recently been reported that FABP4 is also expressed in injured vascular endothelial cells [28, 29].

Although FABP4 is not expressed in normal glomeruli of the kidneys, FABP4 has been reported to be ectopically expressed in glomerular endothelial cells of injured glomeruli [30]. It has also been reported that ectopic expression of FABP4 is associated with renal dysfunction and proteinuria in mice and humans [31, 32] and that urinary FABP4 (U-FABP4) is associated with albuminuria and annual decline in estimated glomerular filtration rate (eGFR) in a general population [33] and patients with diabetes [34], suggesting that U-FABP4 is a promising biomarker for glomerular damage. However, it remains unclear whether U-FABP4 is a predictive biomarker for future renal dysfunction and mortality in patients with diabetes. In the present study, we investigated the associations of U-FABP4 with future events of renal outcome and all-cause death and using a prospective cohort of Japanese patients with diabetes.

MATERIALS AND METHODS

The present study was a prospective cohort study using patients with diabetes and was approved by the ethics committees of Kure Medical Center and Chugoku Cancer Center (26–06) and Sapporo Medical University (3–1-77). Written informed consent was obtained from all the patients. The study conformed to the principles outlined in the Declaration of Helsinki.

Study patients

Japanese patients with diabetes were recruited at Kure Medical Center and Chugoku Cancer Center during the period from 1 July 2014 to 31 March 2016 (n = 738). A flow chart of the study patients is shown in Supplementary Fig. S1. At the recruitment, anthropometric and clinical characteristics were recorded, and samples of blood and urine were obtained from participants and were stored at −80°C until biochemical analyses. Prespecified exclusion criteria were the absence of data for U-FABP4 and patients with diabetes due to specific causes and gestational diabetes. Furthermore, patients with eGFR ≤15 ml/min at baseline were also excluded to investigate further progression of renal dysfunction. After exclusion, a total of 660 patients, including those with type 1 diabetes (T1D) (n = 57) and those with type 2 diabetes (T2D) (n = 603), who received checkups at least once until 31 March 2021 were enrolled in the present study. Most of the recruited patients were regularly followed up at least every 3 months. The clinical endpoints were set as a 30% decline in eGFR and all-cause death.

Measurements

Concentrations of FABP4 and FABP1 were measured by using enzyme-linked immunosorbent assay kits for FABP4 (BioVendor R&D, Modrice, Czech Republic) and FABP1 (CIMIC Co., Tokyo, Japan), respectively. Urinary albumin-to-creatinine ratio (U-ACR, mg/gCr) was used as a marker of microalbuminuria, and levels of U-ACR in patients with diabetes were categorized in accordance with the guidelines of the Japanese Society of Nephrology as follows [35]: stages A1 (U-ACR <30 mg/gCr), A2 (30 mg/gCr ≤ U-ACR <300 mg/gCr), and A3 (U-ACR ≥300 mg/gCr). eGFR was calculated using an equation for Japanese individuals [36]: eGFR (ml/min/1.73 m²) = 194 × serum creatinine ^(−1.094) × age ^(−0.287) × 0.739 (if female). A self-administered questionnaire survey was performed to obtain information on current smoking habit, alcohol drinking habit, and family history of diabetes.

Statistical analysis

Variables are presented as means ± standard deviations for normal distributions or medians (interquartile ranges) for skewed variables. Normality of each variable was tested by the Shapiro–Wilk W test. Comparisons between two groups for parametric and nonparametric parameters were performed by Student's t-test and the Mann–Whitney U-test, respectively. The chi-square test was performed for intergroup differences in percentages of parameters. Clinical parameters were divided into three subgroups according to tertiles of U-FABP4 at baseline (T1–T3). One-way analysis of variance was used for detecting significant differences between data in multiple groups. To investigate whether U-FABP4 level affects outcomes step by step, the cumulative incidences of a 30% decline in eGFR and all-cause death were analyzed by the log-rank test of Kaplan–Meier survival curves in the three subgroups according to tertiles of U-FABP4 at baseline. Multivariable Cox proportional hazard models for a 30% decline in eGFR and all-cause death with a restricted cubic spline (spline knot 4) were analyzed after adjustment for confounders, which were determined by univariable analyses as well as possible candidates [37], including age, sex, type of diabetes, body mass index (BMI), current smoking habit, duration of diabetes, comorbidities of hypertension and dyslipidemia, eGFR, and stage of U-ACR. Non-normally distributed variables were logarithmically transformed for regression analyses. Interactions of sex or type of diabetes with U-FABP4 for the endpoint of a 30% decline in eGFR or all-cause death were also investigated, and hazard ratios (HRs), 95% confidence intervals (CIs), and Akaike's information criterion (AIC) were calculated. Parameters with a lower AIC score constitute a better-fit model. To compare the discrimination, C statistics, analogous to the area under the curve (AUC) in receiver operating characteristic curve analysis, were investigated using the method from DeLong et al. [38]. Moreover, the different discriminatory value was investigated by using continuous net reclassification improvement (NRI) [39] and integrated discrimination improvement (IDI) [40]. A P value of <0.05 was considered statistically significant. All data were analyzed by using EZR [41] and R version 3.6.1.

RESULTS

Characteristics of the studied patients

Baseline characteristics of the enrolled patients (n = 660) and excluded patients (n = 78) with diabetes are shown in Supplementary Table S1. The enrolled patients were significantly older than the excluded patients and they had significantly higher BMI and systolic blood pressure and included a higher percentage of patients with hypertension than did the excluded patients. Levels of U-ACR and eGFR were comparable in the enrolled patients and excluded patients.

Baseline characteristics of the enrolled patients with T1D and those with T2D are shown in Table 1. The median duration of diabetes at baseline was 14 (interquartile range 7–22) years. Patients with T2D were significantly older than those with T1D. Patients with T2D had higher BMI, lower eGFR, and higher levels of total protein, triglycerides, and U-ACR, than those with T1D. The percentage of patients with dyslipidemia was higher in patients with T2D than in patients with T1D. Levels of serum FABP4 (S-FABP4), U-FABP1, and U-FABP4 were significantly higher in patients with T2D than in those with T1D.

Baseline characteristics of patients divided by tertiles (T1–T3) of U-FABP4 at baseline are shown in Table 2. Higher tertiles of U-FABP4 were accompanied by older patients, lower percentage of men, longer duration of diabetes, lower eGFR, and higher levels of S-FABP4, U-ACR, and U-FABP1. There were no significant differences in systolic blood pressure and BMI among the three subgroups.

Incidence rates and cumulative incidences for a 30% decline in eGFR and all-cause death during a follow-up period

Incidence rates for clinical endpoints of a 30% decline in eGFR and all-cause death are shown in Table 3. For a 30% decline in eGFR, the median follow-up period was 62 months, and follow-up summation was 34 637 patient-months. The number of patients who had only one checkup during the follow-up period was 13 (2.0%). Among the 660 enrolled patients, 90 patients (13.6%) progressed to a 30% decline in eGFR during the follow-up period. The incidence rate for a 30% decline in eGFR was 2.6 per 10 000 person-months. Baseline characteristics of patients divided by the presence and absence of a 30% decline in eGFR are shown in Supplementary Table S2. U-FABP4 level was significantly higher in patients with a 30% decline in eGFR during the follow-up period than in those without a 30% decline in eGFR. In Kaplan–Meier survival curves, there was a significant difference in cumulative incidences for a 30% decline in eGFR among the T1 (U-FABP4 ≤0.27 µg/gCr), T2 (U-FABP4 0.28–0.91 µg/gCr), and (T3) (U-FABP4 ≥0.92 µg/gCr) groups (log-rank test P < .001) (Fig. 1A).

For all-cause death, the median follow-up period was 65 months, and follow-up summation was 37 128 patient-months. Among the 660 enrolled patients, 66 patients (10.0%) died during the follow-up period. The incidence rate for all-cause death was 1.8 per 10 000 person-months. Baseline characteristics of patients divided by the presence and absence of all-cause death are shown in Supplementary Table S3. U-FABP4 level was significantly higher in patients with all-cause death during the follow-up period than in those without all-cause death. In Kaplan–Meier survival curves, there was a significant difference in cumulative incidences for all-cause death among the T1, T2, and T3 groups (log-rank test P < .001) (Fig. 1B). Cumulative incidences of the T2 and T3 groups were similar but markedly higher than that of the T1 group.

HRs for a 30% decline in eGFR and all-cause death in patients with diabetes

In univariable Cox regression analyses, current smoking habit, hypertension, eGFR, and U-ACR stage were significantly associated with a 30% decline in eGFR. On the other hand, age, sex, BMI, duration of diabetes, hypertension, eGFR, and U-ACR stage were significantly associated with all-cause death (Supplementary Table S4).

In multivariable Cox regression analysis after adjustment for age, sex, and type of diabetes (Model 1), a high level of logarithmically transformed (log) U-FABP4 was a significant risk factor for a 30% decline in eGFR [HR (95% CI): 1.53 (1.29–1.81)] and for all-cause death [HR (95% CI): 1.31 (1.07–1.61)] (Table 4). After additional incorporation of BMI, current smoking habit, duration of diabetes, comorbidities of hypertension and dyslipidemia, eGFR and stage of U-ACR into Model 1 (Model 2), a high level of log U-FABP4 was also a significant risk factor for a 30% decline in eGFR [HR (95% CI): 1.22 (1.02–1.46)] and for all-cause death [HR (95% CI): 1.25 (1.01–1.56)] (Table 4). The cubic spline of HR for a 30% decline in eGFR in Model 2 gradually increased with a higher level of log U-FABP4 (Fig. 1C). On the other hand, the cubic spline of HR for all-cause death in Model 2 rapidly and then gently increased with a higher level of log U-FABP4 (Fig. 1D).

After further additional incorporation of S-FABP4 into Model 2 (Model 3), a high level of log U-FABP4, but not that of log S-FABP4, was still a significant risk factor for a 30% decline in eGFR [HR (95% CI): 1.21 (1.01–1.45)] and for all-cause death [HR (95% CI): 1.25 (1.01–1.56)] (Supplementary Table S5). Further additional incorporation of U-FABP1 into Model 2 (Model 4) showed that the level of log U-FABP1, but not that of log U-FABP4, was significantly associated with a 30% decline in eGFR or all-cause death (Supplementary Table S5). As a related information, an elevated U-FABP1 level was independently associated with a 30% decline in eGFR and all-cause death after adjustment for age, sex, type of diabetes, BMI, current smoking habit, duration of diabetes, comorbidities of hypertension and dyslipidemia, eGFR, and stage of U-ACR in patients with diabetes (Supplementary Table S6).

In all Models 1–4, there were no significant interactions of sex or type of diabetes with log U-FABP4 for the adjusted HR of a 30% decline in eGFR and that of all-cause death (Table 4, Supplementary Table S5).

Discriminatory capacity of the addition of U-FABP4 for predicting a 30% decline in eGFR or all-cause death

The addition of log U-FABP4 to traditional risk factors, including age, sex, type of diabetes, BMI, current smoking habit, duration of diabetes, comorbidities of hypertension and dyslipidemia, eGFR, and stage of U-ACR, significantly improved the discriminatory capacity for a 30% decline in eGFR in NRI (value 0.242, P = .046) and IDI (value 0.012, P = .031) (Table 5). On the other hand, the discriminatory capacity for all-cause death was significantly improved in NRI (value 0.300, P = .024) but not in IDI. There were no significant differences in the discriminatory capacity for a 30% decline in eGFR or all-cause death between values of AUC for traditional risk factors and those in addition to U-FABP4.

DISCUSSION

The present study showed that elevated U-FABP4 level predicted future renal dysfunction and mortality in patients with diabetes independently of conventional risk factors including age, sex, type of diabetes, BMI, current smoking habit, duration of diabetes, comorbidities of hypertension and dyslipidemia, eGFR, and U-ACR. Kaplan–Meier survival curves showed that there were significant differences in cumulative incidences for a 30% decline in eGFR and all-cause death in patients divided by the tertiles of U-FABP4 level. Furthermore, restricted cubic splines clearly showed that HR for a 30% decline in eGFR gradually increased with a higher level of log U-FABP4 and that HR for all-cause death increased rapidly and then gently increased with a higher level of log U-FABP4. Since progression of renal dysfunction is one of the important prognostic factors in patients with diabetes [42], the results of the present study support the notion that measurement of U-FABP4 is useful for estimation of the risk for both diabetes-related complications and mortality in patients with diabetes.

To the best of our knowledge, this study is the first study showing an association between U-FABP4 level and future prognosis. It was previously shown that elevated circulating FABP4 concentration is a predictive biomarker for all-cause death and cardiovascular death during a 12-year follow-up period in a general population [43]. However, U-FABP4 level was not investigated in the previous study [43]. The precise reason why U-FABP4 was a stronger risk marker than S-FABP4 for all-cause death in the present study remains unknown. However, patients with diabetes are more susceptible than the general population to the development of renal dysfunction [1]. In addition, the pathophysiology of DKD can be greatly influenced by glomerular endothelial cell damage [44]. Ectopic expression and subsequent secretion of FABP4 into urine due to glomerular damage may be greater than those of FABP4 into blood. It has been shown that the presence of chronic kidney disease or DKD not only precedes ESRD but also increases the risk for cardiovascular events, leading to death even in early stages of DKD [35, 45, 46]. Therefore, level of U-FABP4 may directly reflect the progression of DKD, resulting in the association with high mortality in patients with diabetes.

Another reason for the discrepancy for all-cause death between levels of U-FABP4 and S-FABP4 may be the influence of adiposity and nutritional condition. S-FABP4 level has been reported to be independently associated with the amount of adipose tissue in the body [7–9]. In the present study, patients with diabetes who died during the follow-up had significantly lower BMI at baseline than did those who did not die. However, S-FABP4 levels were comparable between patients with diabetes who died and those who did not die (Supplementary Table S3), suggesting that S-FABP4 levels are underestimated in elder patients with diabetes who may have sarcopenia. On the other hand, it has been reported that U-FABP4 level is not significantly correlated with BMI [33, 34]. Therefore, U-FABP4 level would be a better predictor of prognosis than S-FABP4 level independent of adiposity especially in elder patients with diabetes.

In a cross-sectional study [47], S-FABP4 concentration was shown to be independently associated with eGFR as well as levels of soluble tumor necrosis factor receptors (sTNFRs) including sTNFR1 and sTNFR2, which are potent markers for renal dysfunction [48, 49], in patients with diabetes. In the present study, S-FABP4 was not significantly associated with future renal dysfunction and all-cause death (Model 3, Supplementary Table S5). The reason for this discrepancy in results is unclear, but it might be explained by the possibility that specific expression of FABP4 in the kidneys directly reflects U-FABP4 rather than S-FABP4. It has been reported that stimulation of vascular endothelial cells with vascular endothelial growth factor or hydrogen peroxide induced the expression of FABP4 [29] and that FABP4 is secreted from cultured glomerular endothelial cells [32]. We also previously showed that U-FABP4 level and expression of FABP4 in glomeruli were independently associated with each other after adjustment of various related factors in patients with chronic glomerulonephritis [32]. Since the molecular weight of FABP4 is ∼14–15 kDa [7], a part of circulating FABP4 may cross the glomerular filtration barrier, leading to a source of U-FABP4. Indeed, U-FABP4 level was significantly, but not strongly, correlated with S-FABP4 level in the present study (r = 0.186, P < .001, data not shown). Therefore, U-FABP4 rather than S-FABP4 may reflect glomerular damage, presumably resulting in a promising predictive marker for a decline in eGFR in patients with diabetes.

It has been reported that U-FABP1 is associated with tubulointerstitial damage [13] and that U-FABP4 can reflect glomerular damage [30]. In the present study, an elevated U-FABP1 level was independently associated with a 30% decline in eGFR and all-cause death (Supplementary Table S6). Furthermore, only U-FABP1 was selected as a significant factor for future renal dysfunction and mortality in a model using both U-FABP4 and U-FABP1 as covariates (Model 4, Supplementary Table S5). One interpretation of the result is that not only glomerular damage but also tubulointerstitial damage contributes to major pathophysiological factors in the progression of DKD [50]. Another possible reason is that the effects of U-FABP4 on the outcomes were underestimated in the model since there might be a multicollinearity between levels of U-FABP1 and U-FABP4 [33, 34]. Indeed, there was a significant positive correlation between U-FABP4 and U-FABP1 in the present study (r = 0.631, P < .001, data not shown). Since FABPs including FABP4 and FABP1 are molecules that coordinate lipid responses in cells [7], the expression of FABPs might be influenced by the state of metabolic dysfunction including diabetes, partially affecting urinary excretion of FABP4 and FABP1. Nevertheless, simultaneous evaluation of U-FABP4 and U-FABP1 would be useful for prediction of renal outcomes, and the development of novel formulas using levels of U-FABP4 and U-FABP1 may improve the accuracy for estimation of renal outcomes in DKD.

The present study has several limitations. First, since only Japanese people were enrolled in the present study, the generalizability of the findings to other ethnic groups would be limited by any genetic and environmental factors. Second, the possibility of sample selection bias cannot be ruled out since all patients were recruited form a single hospital. Third, it has been reported that several agents affect circulating levels of FABP4 [22, 51–54], although there has been no study focused on the regulation of U-FABP4 by treatment with drugs. Potential effects of unknown therapeutic agents affecting levels of U-FABP4 cannot be ruled out in the present study. Fourth, since the present study was started before showing beneficial effects of sodium-glucose co-transporter 2 inhibitors and glucagon-like peptide-1 analogs on DKD [55], the number of patients treated with those drugs was small in the present study. Further exploration of how existing treatments might influence U-FABP4 levels could provide valuable context for future research directions. Finally, events for the development of ESRD and for cardiovascular death as clinical endpoints were not evaluated in the present study.

In conclusion, U-FABP4 is a promising predictive biomarker for both renal dysfunction and poor prognosis in patients with diabetes. Incorporating U-FABP4 measurement into clinical protocols as an indicator of early therapeutic intervention is expected to improve prognosis and healthy life expectancy in not only patients with DKD but also patients with diabetes who have not yet developed DKD. Further studies are needed to investigate whether U-FABP4 can be a useful biomarker in a large number of patients with diabetes including those who have ESRD.

FUNDING

M.T. (22K08313), T.G. (23K07681), and M.F. (23K07993) were supported by grants from Japan Society for the Promotion of Science.

AUTHORS’ CONTRIBUTIONS

M.T. and M.F. designed the study, performed data collection and statistical analyses, and wrote the paper. T.S., T.G., N.K., M.M., K.N., Y.A., K.E., W.K., H.A., T.S., and Y.S. performed data collection and discussed the data. E.I., M.K., and M.S. performed data collection. All authors approved the final version of manuscript.

DATA AVAILABILITY STATEMENT

The data underlying this article will be shared on reasonable request to the corresponding author.

CONFLICT OF INTEREST STATEMENT

None declared.

REFERENCES

Author notes