Association of urinary post-translationally modified fetuin-a and outcomes in acute kidney disease

Abstract

Background

Risk stratification during the acute kidney disease period remains suboptimal with existing staging systems.

Methods

We used prospectively collected clinical data and samples from 130 patients with acute kidney disease to evaluate the association between a novel biomarker, urinary post-translationally modified fetuin-A (uPTM-FetA), and clinical outcomes. Traditional staging systems based on the serum creatinine ratio, estimated glomerular filtration rate, or albuminuria were also compared. The outcome of interest was death or dialysis within 180 days.

Results

The level of uPTM-FetA was significantly associated with the outcome in both logistic regression (odds ratio 3.41, 95% confidence interval 1.35–8.62, P = 0.01) and Cox models (hazard ratio 2.91, 95% confidence interval 1.30–6.51, P = 0.009). The crude area under the receiver operating characteristic curve was 0.75. Grouping with tertiles of uPTM-FetA levels showed discrimination between patients with and without the outcome. Furthermore, integrating uPTM-FetA into the traditional staging system of acute kidney disease was associated with improvement in model fit, risk classification, and clinical utility.

Conclusions

uPTM-FetA during the acute kidney disease period offers the potential to provide additional prognostic information. Given the small sample size, the results should be viewed as exploratory. Future studies with independent cohorts are warranted to validate this finding.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12882-026-04955-5.

Introduction

Acute kidney disease (AKD) is defined as abnormalities of kidney function for less than 3 months [1–3]. AKD may occur regardless of preceding acute kidney injury (AKI), and its clinical outcomes vary widely [4–6]. Staging systems or validated biomarkers have been proposed to help predict outcomes in patients with AKD.

A serum creatinine ratio-based AKD staging system (referred to as AKD_sCr) has been proposed by the Acute Dialysis Quality Initiative (ADQI) 16 Workgroup [1] to predict the prognosis. Later, an estimated glomerular filtration rate (eGFR)-based and an albuminuria-based AKD staging system (referred to as AKD_eGFR and AKD_UACR) were also advocated by a Kidney Disease: Improving Global Outcomes (KDIGO) consensus conference [2], utilizing the same definitions for the G and A stages as the chronic kidney disease (CKD) staging system. The purpose of this eGFR and albuminuria-based AKD staging system is to harmonize the definitions and stages of AKD and CKD. While the AKD_sCr staging system has been reported to predict outcomes in AKD patients [7, 8], the performance of the AKD_eGFR staging system is less clear. Notably, albuminuria is associated with kidney disease progression in the post-AKI period [9].

Numerous biomarkers have been shown to predict outcomes in patients with AKI [10–12]. However, most studies measured biomarkers within 7 days of AKI onset (during the AKI period) [12], and their ability to predict outcomes in patients with AKD remains unclear. Recently, a novel biomarker, urinary post-translationally modified fetuin-A (uPTM-FetA), has been identified to assist risk stratification in patients with diabetes [13], chronic kidney disease (CKD) [14], or kidney allograft [15]. In this study, we aimed to compare uPTM-FetA measured during the AKD period with traditional AKD staging systems regarding their associations with clinical outcomes.

Methods

Patients

This study is a post-hoc prognostic analysis of participants enrolled in a prospective clinical trial (201811079MINC approved at National Taiwan University Hospital on February 13, 2019; NCT05064904, retrospectively registered on October 1, 2021). The objectives of the original trial were to investigate the effects of renin-angiotensin system (RAS) inhibitors and multidisciplinary health education on major adverse kidney events at 180 days in patients with AKD. The trial included adult outpatient AKD patients with prior hospital-acquired AKI, as defined by the KDIGO criteria [1, 16]. Patients requiring maintenance dialysis treatment after discharge from the index hospitalization were excluded. At the time of enrollment, patients with baseline eGFR < 5 mL/min/1.73m², obstructive AKI, major hemorrhage, respiratory failure requiring mechanical ventilation, hypotension (systolic blood pressure < 110 mmHg), a history of organ transplantation, or pregnancy were also excluded. Blood and urine samples were obtained 7 to 14 days after discharge from the index hospitalization with the consent of the participants. The enrollment of participants was slower than expected during the COVID-19 pandemic, and the primary results of the parent randomized trial have not yet been published. This ancillary biomarker analysis was conducted independently of the primary efficacy evaluation. One hundred and thirty patients enrolled from February 2019 to May 2023 with available urinary samples during the AKD period were included in the analysis. The use of urinary and blood samples for research was approved by the Research Ethics Committee at the National Taiwan University Hospital (202308085RINC).

Measurements of uPTM-FetA, albuminuria, and creatinine

The measurements of uPTM-FetA were assisted by Bio Preventive Medicine Corp., Hsinchu, Taiwan, using a DNlite uPTM-FetA ELISA Kit (CE IVD Marked, Trade name: DNlite-IVD103). The kit has a detection range of 5.428-250 ng/mL, with intra-assay coefficients of variation below 10% and inter-assay coefficients of variation below 15%. Bio Preventive Medicine was blinded to the clinical information of the participants providing urinary samples when performing the tests. The urine albumin and urine creatinine were measured by an ADVIA Chemistry XPT.

Covariates and the outcome

The clinical parameters in Table 1 were retrieved from the record of the clinical trial (NCT05064904). The timeframe of the study was illustrated in Supplementary Figure S1. The baseline eGFR was defined as the highest level of eGFR within three months before the index hospitalization estimated by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) method [17]. AKI stage was defined by the KDIGO criteria [18]. The AKD_sCr stage was defined according to the ADQI 16 Workgroup [1]. The AKD ratio was defined as the ratio of serum creatinine level between the AKD period and the baseline period. The AKD_eGFR stage and the AKD_UACR stage were defined according to the 2021 KDIGO Consensus Conference [2]. The AKD_eGFR was defined as the eGFR level at D7-14 after discharge (Supplementary Figure S1). Because the levels of uPTM-FetA and urinary albumin were right-skewed, they were adjusted for urine creatinine and natural log-transformed to fit the logistic regression or Cox models. uPTM-FetA, albuminuria, and serum creatinine were measured using samples obtained 7 to 14 days after discharge from the index hospitalization (i.e., during the AKD period). The levels of serum creatinine at baseline and at D7-14 were used to estimate the AKD_sCr stage, AKD ratio, and AKD_eGFR stage. The Sequential Organ Failure Assessment (SOFA) Score and Charlson’s Comorbidity Index were calculated based on prior reports [19, 20]. In Table 1, comorbid cardiovascular disease included coronary artery disease, heart failure, stroke, and peripheral artery disease. The outcome of interest was death or dialysis within 180 days after discharge from the index hospitalization. Since patients requiring maintenance dialysis after discharge were excluded, the dialysis endpoint specifically captured those who initiated new treatment post-enrollment. Inclusion of patients who received only temporary dialysis during their index hospitalization was allowed.

Table 1.

Baseline characteristics of the included AKD patients

		All	uPTM-FetA Low		uPTM-FetA Medium	uPTM-FetA High	P
No.		130	43		43	44
Age (years old)		64.1 ± 14.6	62.1 ± 14.9		65.9 ± 14.2	64.2 ± 14.8	0.50
Male Sex (%)		95 (73%)	32 (74%)		30 (70%)	33 (75%)	0.84
Baseline eGFR (mL/min/1.73m2)		53.8 ± 30.6	62.7 ± 31.1		51.6 ± 30.1	47.4 ± 29.1	0.06
AKI stage (%)							0.16
Stage 1		34 (26%)	7 (16%)		12 (28%)	15 (34%)
Stage 2		24 (18%)	7 (16%)		9 (21%)	8 (18%)
Stage 3		27 (21%)	14 (33%)		9 (21%)	4 (9%)
Dialysis		45 (35%)	15 (35%)		13 (30%)	17 (39%)
1 uPTM-FetA (ng/mg)		3.41 ± 1.88	1.38 ± 1.59		3.67 ± 0.41	5.21 ± 0.61	< 0.001
1 Albuminuria (mg/g)		4.67 ± 2.12	2.98 ± 1.62		4.77 ± 1.61	6.23 ± 1.75	< 0.001
2AKDsCr stage (%)							0.008
Stage 0		96 (74%)	40 (93%)		30 (70%)	26 (59%)
Stage 1		23 (18%)	3 (7.0%)		9 (21%)	11 (25%)
Stage 2		7 (5.4%)	0 (0%)		3 (7.0%)	4 (9.1%)
Stage 3		4 (3.1%)	0 (0%)		1 (2.3%)	3 (6.8%)
3 AKD ratio		1.38 ± 0.86	1.10 ± 0.29		1.34 ± 0.56	1.70 ± 1.28	0.004
4AKDeGFR (mL/min/1.73m2)		43.7 ± 29.0	59.8 ± 29.4		40.1 ± 26.8	31.7 ± 23.6	< 0.001
5AKDeGFR stage (%)							< 0.001
Stage 0		31 (24%)	19 (43%)		8 (19%)	4 (9.3%)
Stage 3		51 (39%)	16 (36%)		20 (47%)	15 (35%)
Stage 4		27 (21%)	8 (18%)		9 (21%)	10 (23%)
Stage 5		21 (16%)	1 (2.3%)		6 (14%)	14 (33%)
SOFA score		6.3 ± 2.8	6.1 ± 2.5		6.4 ± 2.9	6.5 ± 3.0	0.79
ICU admission (%)		28 (22%)	10 (23%)		7 (16%)	11 (26%)	0.56
Charlson’s Comorbidity Index		5.1 ± 3.0	4.0 ± 2.3		5.4 ± 2.7	6.1 ± 3.5	0.004
Comorbidity (%)
Diabetes mellitus		61 (47%)	14 (33%)		18 (42%)	29 (66%)	0.006
Hypertension		80 (62%)	22 (51%)		28 (65%)	30 (68%)	0.22
Cardiovascular disease		41 (32%)	18 (42%)		15 (35%)	8 (18%)	0.050
Malignancy		18 (14%)	7 (16%)		6 (14%)	5 (11%)	0.80
Intervention (%)
RAS inhibitor		66 (51%)	23 (53%)		19 (44%)	24 (55%)	0.57
Multidisciplinary health education		69 (53%)	18 (42%)		29 (67%)	22 (50%)	0.052

¹uPTM-FetA and albuminuria were adjusted for urine creatinine and then natural log-transformed. Patients were grouped according to tertiles of uPTM-FetA levels. ²The AKDsCr stages were defined according to the ADQI 16 Workgroup. ³The AKD ratio was defined as the ratio of serum creatinine between the AKD period and the baseline period. ⁴The AKD_eGFR represented the eGFR level at D7-14 in Supplementary Figure S1. ⁵The AKD_eGFR stages were defined according to the 2021 KDIGO Consensus Conference. Patients with eGFR ≥ 60 mL/min/1.73m² were classified as AKD_eGFR stage 0. Continuous variables were expressed as mean (standard deviation, SD). Categorical variables were expressed as n (%)

Abbreviations: AKD, acute kidney disease; AKI, acute kidney injury; eGFR, estimated glomerular filtration rate; ICU, intensive care unit; RAS, renin-angiotensin system; SOFA, Sequential Organ Failure Assessment; uPTM-FetA, urinary post-translationally modified fetuin-A

Statistical analysis

Continuous variables were presented as mean ± standard deviation. Categorical variables were presented as percentages. Statistical significance was defined as P < 0.05. The distribution of baseline characteristics across uPTM-FetA tertiles was assessed using one-way ANOVA tests for continuous variables and Chi-square tests for categorical variables. Logistic regression models and Cox models were used to test the association between uPTM-FetA and the outcome of death or dialysis within 180 days after discharge from the index hospitalization. In addition to uPTM-FetA, we also tested the associations between the outcome and other AKD staging systems, including the AKD_sCr staging system, the AKD_eGFR staging system, and albuminuria. Because no patients with AKD_sCr stage 3 (defined as AKD ratio ≥ 3) had death or dialysis within 180 days, the level of AKD ratio was used to represent AKD_sCr stages to fit the logistic regression or the Cox models. Similarly, the natural log-transformed urinary creatinine-adjusted level of albuminuria (mg/g) was used to represent the albuminuria stage because no patients with albuminuria stage A1 (defined as albuminuria < 30 mg/g of creatinine) experienced death or dialysis within 180 days. The unadjusted area under the curve (AUC) of the receiver operating characteristic curve (ROC) was used to assess the prognostic performance of uPTM-FetA and other AKD staging systems. The Youden Index was used to derive an empirical uPTM-FetA threshold for within-cohort discrimination between patients with and without the outcome [21]. The Kaplan-Meier plot was used to visualize event-free survival across uPTM-FetA tertiles or an empirically derived uPTM-FetA threshold level. To account for the low event rate and high risk of overfitting, we used the Least Absolute Shrinkage and Selection Operator (LASSO) method to select variables among nine potential predictors, including age, sex, baseline eGFR, AKI stage, SOFA score, intensive care unit (ICU) admission, Charlson’s Comorbidity Index, RAS inhibitor use, and multidisciplinary health education, while forcing uPTM-FetA into the model. Continuous variables were standardized by their respective standard deviations. AKI stage was defined as a binary variable with AKI stage 1 or 2 as the reference group. At λ_1se, only uPTM-FetA was retained in the model. At λ_min, only RAS inhibitor use was removed from the model. Consequently, we selected uPTM-FetA, AKI stage, age, and sex to construct a parsimonious model. Akaike information criterion (AIC) and the likelihood ratio test were used to determine the fitness of multivariable logistic regression models, while the C-statistic was used to evaluate the fitness of multivariable Cox models. A lower AIC or a higher C-statistic indicates better model fit. The continuous net reclassification improvement (NRI) [22] was used to assess the risk reclassification after adding uPTM-FetA. Decision curve analysis was used to assess clinical utility [23]. As internal validation, bootstrap resamples were performed to obtain optimism-corrected performance estimates. All data were analyzed using R version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Association of uPTM-FetA with outcomes in patients with acute kidney disease

To examine the association between uPTM-FetA levels and the clinical outcome of death or dialysis within 180 days in patients with AKD, we analyzed 130 AKD patients enrolled between February 2019 and May 2023 from a clinical cohort (NCT05064904). Overall, the included patients were relatively elderly (age of 64.1 ± 14.6 years old), were mostly male (73%), had baseline eGFR 53.8 ± 30.6 mL/min/1.73m², had high KDIGO AKI stages (56% of stage 3 or dialysis), had high AKD_eGFR stages (37% of stage 4 or 5), had high SOFA scores (6.3 ± 2.8), had a high percentage of ICU admission (22%), had a high Charlson’s Comorbidity Index (5.1 ± 3.0), and had a high percentage of comorbid cardiovascular diseases (32%), diabetes mellitus (47%), hypertension (62%), and malignancy (14%) (Table 1). In line with prior reports that uPTM-FetA is a biomarker for diabetic kidney disease [13], comorbid diabetes mellitus and a higher Charlson’s Comorbidity Index were associated with higher uPTM-FetA levels. Interestingly, a higher uPTM-FetA level was also associated with more severe albuminuria, a higher AKD ratio, a higher AKD_sCr stage, and a higher AKD_eGFR stage. The baseline characteristics were otherwise balanced across uPTM-FetA tertiles.

Thirteen (10%) of the patients developed the outcome. Among the thirteen patients, six initiated dialysis without death, six died without dialysis, and one initiated dialysis before death within the 180-day time frame. In the univariable logistic regression model, uPTM-FetA was significantly associated with the development of death or dialysis within 180 days [odds ratio (OR) 3.41, 95% confidence interval (CI) 1.35–8.62, p = 0.01] (Supplementary Table S1). The association between uPTM-FetA and death or dialysis within 180 days remained significant (OR 4.53, 95% CI 1.49–13.76, p = 0.008) in a multivariable logistic regression model adjusted for age, sex, and AKI stage (Supplementary Table S1). The optimism-corrected OR derived from bootstrap resampling was 3.30 (Supplementary Table S1).

The Kaplan-Meier plot showed significant differences in survival free from death or dialysis within 180 days across tertiles (p = 0.005 by log-rank test) (Fig. 1). In addition, uPTM-FetA was significantly associated with the outcome in both an univariable Cox model [hazard ratio (HR) 2.91, 95% CI 1.30–6.51, p = 0.009] and a multivariable Cox model (HR 3.27, 95% CI 1.39–7.66, p = 0.006, Supplementary Table S2) adjusted for the same covariates as in the Supplementary Table S1. The optimism-corrected HR derived from bootstrap resampling was 2.63 (Supplementary Table S2).

Fig. 1.

The Kaplan-Meier plot for survival free from death or dialysis within 180 days across uPTM-FetA tertiles. The Kaplan-Meier curves depict survival free from death or dialysis in AKD patients stratified by uPTM-FetA tertiles. The survival analysis was censored at day 180. p = 0.005 by log-rank test

An empirically derived uPTM-FetA threshold (40.7 ng/mg, corresponding to 3.7 ng/mg after natural logarithmic transformation in Table 1) was identified using the Youden index [21]. Kaplan–Meier analysis (Supplementary Figure S2) demonstrated that patients above this threshold had a higher risk for the composite outcome. These findings suggest the discriminatory capacity of uPTM-FetA within this cohort.

Incremental prognostic information provided by uPTM-FetA beyond the AKD_eGFR staging system

We next asked whether integrating uPTM-FetA into existing AKD staging systems, such as the creatinine ratio-based AKD_sCr staging system [1] and the eGFR-based AKD_eGFR staging system [2], can further enhance their predictive ability for the outcome. First, we examined the distribution of patients who developed the outcome of death or dialysis within 180 days across uPTM-FetA tertiles or AKD stages in different AKD staging systems. Except for the AKD_sCr staging system, the percentages of patients with the outcome generally increased with a higher uPTM-FetA category or a higher AKD stage (Supplementary Table S3). Then, we investigated the association between AKD staging systems and the outcome in logistic regression and Cox models. In this study, the AKD_sCr staging system was represented by the AKD ratio (the ratio of serum creatinine between the AKD period and the baseline period), the AKD_eGFR staging system was represented by the AKD_eGFR (eGFR obtained during the AKD period), and the AKD_UACR staging system was represented by albuminuria during the AKD period, respectively, because traditional AKD staging systems could not fit the models.

In the univariable logistic regression models, the AKD ratio was not significantly associated with the outcome, while AKD_eGFR and albuminuria were (Table 2). The AUC was 0.721, 0.764, 0.722, and 0.751 for AKD ratio, AKD_eGFR, albuminuria, and uPTM-FetA, respectively (Table 2 and Supplementary Figure S3). Six multivariable logistic regression models incorporating different predictors and three common covariates were then generated to compare the performance of traditional AKD staging systems without uPTM-FetA (models 1–3) or with uPTM-FetA (models 4–6) (Table 3). Comparison between models with the likelihood ratio test suggested that models integrating uPTM-FetA had better model fit (model 1 vs. model 4, p = 0.002; model 2 vs. model 5, p = 0.02; model 3 vs. model 6, p = 0.04). Adding uPTM-FetA in the model also improved risk classification assessed by continuous NRI [point estimate and 95% CI; model 1 vs. model 4, 0.82 (0.28–1.31); model 2 vs. model 5, 0.77 (0.18–1.28); model 3 vs. model 6, 0.36 (-0.25-1.01)]. Furthermore, decision curve analysis supported the clinical benefit of models that added uPTM-FetA across threshold probabilities between 15% and 40% (Fig. 2). The results in the univariable Cox models (Supplementary Table S4) and multivariable Cox models were similar (Supplementary Table S5).

Table 2.

The crude area under the curve and the crude odds ratios for uPTM-FetA and other AKD staging systems

1Predictor	2AUC (95% CI)	3OR (95% CI)	3 P
AKD ratio	0.721 (0.592–0.849)	1.26 (0.83–1.90)	0.28
4AKDeGFR	0.764 (0.623–0.905)	3.85 (1.32–11.1)	0.01
Albuminuria	0.722 (0.601–0.843)	2.53 (1.24–5.17)	0.01
uPTM-FetA	0.751 (0.648–0.854)	3.41 (1.35–8.62)	0.01

¹The predictors were defined the same as in Table 1. Continuous variables were standardized by their standard deviations. ²Bootstrap resampling with 2,000 replicates was performed to estimate 95% CIs for AUCs. ³OR, CI, and P were estimated in the univariable logistic regression model. ⁴The OR and 95% CI were estimated per 1-standard-deviation decrease in AKD_eGFR

Abbreviations: CI, confidence interval; OR, odds ratio

Table 3.

The performance of AKD staging systems with or without integration of uPTM-FetA in the multivariable logistic regression models

Model	Predictor	Optimism-corrected AUC	Optimism-corrected Brier score	Optimism-corrected calibration slope	AIC	OR (95% CI)	P
1	AKD ratio	0.791	0.092	0.729	77.8	1.46 (0.85–2.50)	0.17
2	AKDeGFR	0.839	0.081	0.723	68.5	5.54 (1.49–20.61)	0.01
3	Albuminuria	0.855	0.086	0.667	68.9	1.82 (0.72–4.61)	0.21
	AKDeGFR					3.85 (1.05–14.10)	0.04
4	uPTM-FetA	0.843	0.083	0.725	70.6	4.38 (1.40–13.69)	0.01
	AKD ratio					1.09 (0.57–2.09)	0.79
5	uPTM-FetA	0.871	0.081	0.673	64.7	4.09 (1.16–14.42)	0.03
	AKDeGFR					4.26 (1.07–16.96)	0.04
6	uPTM-FetA	0.863	0.082	0.610	66.6	4.74 (0.96–23.49)	0.06
	Albuminuria					0.83 (0.25–2.74)	0.76
	AKDeGFR					4.57 (1.03–20.34)	0.046

All the models were adjusted for three common covariates, including age, sex, and AKI stage as listed in Supplementary Table 1. Optimism-corrected AUC, Brier scores, and calibration slopes were estimated by 1,000 bootstrap resampling iterations. The OR and 95% CI in AKD_eGFR were estimated per 1-standard-deviation decrease

Abbreviation: AIC, Akaike information criterion

Fig. 2.

The decision curve analysis comparing models with and without uPTM-FetA. The decision curve analysis depicts the net clinical benefit across threshold probabilities for the baseline models and models adding uPTM-FetA. Figure 2a and b, and 2c compare figures comparing model 1 vs. model 4, model 2 vs. model 5, and model 3 vs. model 6 from Table 3, respectively. Red lines indicate the “treat all” strategy. Green lines indicate the “treat none” strategy. Teal lines indicate the baseline models, and purpule lines indicate models adding uPTM-FetA

These results suggest that incorporation of uPTM-FetA into AKD staging frameworks may provide incremental prognostic information in this cohort.

Discussion

The main finding of this study is that uPTM-FetA levels measured during the AKD period were associated with adverse 180-day outcomes in this cohort of 130 patients with AKD. Furthermore, incorporation of uPTM-FetA might provide incremental prognostic information beyond traditional AKD staging systems. These findings should be interpreted as hypothesis-generating and should be validated in independent cohorts.

We observed several findings supporting that uPTM-FetA measurement in AKD patients may have potential prognostic relevance. First, tertiles of uPTM-FetA levels showed significant discrimination between patients with and without the composite outcome in this cohort (Fig. 1). Second, the AUC of 0.751 for uPTM-FetA in this cohort was acceptable and comparable to those of traditional staging systems (Table 2). Third, integrating uPTM-FetA achieves better model fit, risk classification, and clinical benefits than using the traditional AKD staging systems alone (Table 3; Fig. 2).

In line with prior studies showing the prognostic role of urinary fetuin-A in CKD patients especially with diabetes mellitus, we observed a higher uPTM-FetA level in AKD patients with higher albuminuria, with comorbid diabetes mellitus, and with a trend for lower eGFR in our cohort (Table 1). Importantly, the level of uPTM-FetA was independently associated with death or dialysis of AKD patients even after adjustment for age, sex, and the index AKI stage. Interestingly, higher uPTM-FetA levels appeared to be inversely associated with prevalent cardiovascular disease in this cohort, despite their positive associations with diabetes mellitus and CKD. However, this observation should be interpreted cautiously, given the limited sample size. Future studies are required to validate the association between uPTM-FetA and cardiovascular diseases.

Plasma fetuin-A (alpha-2-HS glycoprotein), a multifunctional protein participating in calcium homeostasis and immune responses, is mainly produced in the liver [24]. A clinical study showed limited fetuin-A transcripts and proteins in the renal tubular cells in patients with chronic kidney disease [25]. Because fetuin-A is a relatively small protein with a molecular weight of 51 to 67 kDa, filtration of plasma fetuin-A through the damaged glomerular filtration barrier as with albumin (molecular weight of 65 kDa) may occur. However, another pre-clinical study reported that during hypoxic kidney injury renal proximal tubule produces fetuin-A, which may be an important source of urinary fetuin-A [26]. In addition, the locally produced fetuin-A was shown to mediate renoprotective effects through modulation of calcification and inflammation. As such, the source of urinary fetuin-A during kidney insult is controversial. Notably, increased urinary fetuin-A protein levels have been shown to be a biomarker associated with histological kidney fibrosis [25, 27], diabetes kidney disease [13, 28, 29], and acute kidney disease [30, 31]. In our study, fetuin-A was measured by the DNlite-IVD103 ELISA kit, which detected only the post-translationally modified connecting peptide of the precursor fetuin-A protein, uPTM-FetA. Whether uPTM-FetA in AKD patients is produced locally in the acutely injured renal tubules or is filtered through damaged renal glomeruli warrants further investigation. Nonetheless, uPTM-FetA may reflect persistent tubular injury or glomerular injury, which is distinct from filtration function (as measured by GFR). This supports the complementary role of uPTM-FetA, in addition to eGFR, in predicting outcomes.

We observed zero events in patients with AKD_sCr stage 3, while nearly 40% of events occurred in patients with AKD_sCr stage 2 (Supplementary Table S3). This might be seen as paradoxical to the common belief that a higher creatinine increment is associated with more severe kidney injury and poorer outcomes. However, because patients who died or underwent maintenance dialysis after discharge were excluded, the results in patients with AKD_sCr stage 3 might reflect a selection of survivors without long-term dialysis requirements. Notably, only 4 patients (3.1%) fell within this category.

This study has several limitations. First, the sample size of 130 is small. The follow-up duration is relatively short, and the absolute number of outcome events is limited. The multivariable models are subject to overfitting, given the low ratio of events per variable (EPV) of 3.3 when AKI stage, age, and sex are adjusted for (Supplementary Tables S1 and S2). Furthermore, the insignificant associations between the outcome and the predictors in the models (Tables 2 and 3) may arise from insufficient statistical power. The current study should be viewed as exploratory, and future studies with larger sample sizes are warranted to validate the present findings. Nonetheless, this study used biological samples from AKD patients enrolled in a prospective clinical trial with extensive and accurate data collection. The consistent results obtained with parsimonious models and internal validation via bootstrap resamples support the robustness of the associations between uPTM-FetA and the outcome. Second, the comparison of performance between uPTM-FetA and other biomarkers such as proenkephalin A 119–159 (PENK) [32] and C-C motif chemokine ligand 14 (CCL14) [33] is lacking. However, we reported the comparative performance of the AKD_eGFR staging system, albuminuria, and the AKD_sCr staging system, which are essential standard-of-care for AKD patients. Third, we were unable to fit the traditional AKD_eGFR staging system, AKD_sCr staging system, or AKD_UACR staging system in our models due to the small sample size and sparse events in specific stages. We used the continuous variables of the AKD_eGFR (representing the eGFR level), the AKD ratio (representing the ratio of serum creatinine), and the albuminuria to estimate the associations between these variables and the outcome. Interestingly, the AKD_eGFR seemed to demonstrate more consistent risk stratification than the AKD ratio in our cohort. The AUC and the model fit were generally lower for models incorporating the AKD ratio than for models incorporating the AKD_eGFR. In our prior analysis of AKD patients with hospital-acquired dialysis-requiring AKI, we also observed more consistent risk stratification using the AKD_eGFR staging system than the AKD_sCr staging system [34]. More studies are needed to validate the prognostic performance of different AKD staging systems.

Conclusion

In this prospective exploratory study, higher uPTM-FetA levels measured during the AKD period were associated with an increased risk of death or dialysis within 180 days. Incorporation of uPTM-FetA into traditional AKD staging systems provided incremental prognostic information in this cohort. Given the small sample size and low number of events in this study, external validation in larger and independent cohorts is required before clinical application can be considered.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Szu-Yu Pan, Vin-Cent Wu, and Thomas Tao-Min Huang conceived the article. Szu-Yu Pan drafted the article. Thomas Tao-Min Huang, Yi-De Huang, Yu-Ting Hsieh, and Tsung-Lin Wu revised the article. Shi-Heng Wang, Heng-Chih Pan, Chih-Hsiang Chang, and Shuei-Liong Lin provided intellectual content of critical importance to the work. All authors approved the final version to be published.

Funding

We thank Mrs. Hsiu-Chin Lee Kidney Research Fund for supporting this study. Dr. Szu-Yu Pan is supported by the National Taiwan University Hospital (114-SS0022, 114-UN0058) and the Taiwan National Science and Technology Council (NSTC 113-2314-B-002-277-MY3).

Data availability

All data generated or analyzed during this study are included in this article and its supplementary material files. Further inquiries can be directed to the corresponding author.

Declarations

Ethics approval and consent to participate

This study protocol was reviewed and approved by the Research Ethics Review Committee at National Taiwan University Hospital (202308085RINC), and was performed in accordance with the principles of the Declaration of Helsinki. This study used de-identified data from a clinical trial (NCT05064904) and was granted an exemption from requiring written informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.