Performance of novel biomarkers for prediction of diabetic kidney disease in patients with diabetes mellitus

Abstract

Introduction

Diabetic kidney disease (DKD) is a common and serious complication in patients with diabetes mellitus (DM). This study was aimed to reveal the validity of seven emerging novel biomarkers of angiopoietin-like-4 (ANGPTL4), neutrophil gelatinase-associated lipocalin (NGAL), monocyte chemoattractant protein-1 (MCP-1), growth differentiation factor-15 (GDF15), fibroblast growth factor-23 (FGF23), n-terminal osteopontin (ntOPN) and pyruvate kinase muscle isozyme M2 (PKM2) in detecting DM patients at high risk of DKD and establish prediction models for DKD onset in DM patients.

Methods

This was a cross-sectional study of 348 adult patients with Type 1 DM for at least 5 years, or Type 2 DM, followed by a prospective observational cohort of 141 adult DM patients without renal involvement at baseline and follow-up for at least 2 years. We performed logistic regression analysis to analyze the relationship between the variables and the risk of DKD occurrence, and receiver operator characteristic (ROC) analysis to assess the predictive ability of multi-biomarker panels for DKD onset.

Results

In the cross-sectional cohort, the seven urinary biomarkers were all elevated in DKD patients, of which the high levels of urinary ntOPN, GDF15, NGAL, MCP-1 and FGF23 significantly increased the risk of DKD diagnosis; the urinary MCP-1 alone performed best in DKD detection with the largest area under the ROC curve (AUC). In the prospective cohort, the high levels of urinary GDF15, MCP-1, ANGPTL4 and FGF23 significantly increased the risk of DKD development, and the model constructed based on the above four biomarkers had the largest AUC (0.873) for predicting the 2-year risk of DKD occurrence.

Conclusion

Our study demonstrated that the four-biomarker model performed the best in predicting DKD, which could provide more accurate tools for DKD risk prediction, thereby improving the prognosis in DM patients.

Introduction

Diabetes mellitus (DM) is a global health challenge. Its prevalence was approximately 11% in the adult population, with 589 million adult DM patients in 2024, and the number is predicted to rise to 853 million by 2050 [1]. Diabetic kidney disease (DKD), chronic kidney disease (CKD) attributed to DM, develops in 20%–40% of DM patients and has become the leading cause of CKD worldwide [2]. DKD progression varies widely and depends on multiple factors; therefore, novel kidney biomarkers may reflect the DKD pathogenesis with crucial clinical implications. Biomarker-guided strategies should be promising for early DKD prediction, diagnosis and intervention. Based on the variability of mechanisms of DKD development, the precision and personalized management might thereby improve the renal and cardiac prognosis in DM patients [3].

Therefore, constructing forecast models using panels of biomarkers was more promising than those with a single biomarker. Rational clinical implementation might provide a non-invasive and accurate tool for early DKD detection, and thereby delay DKD development and its complications. Therefore, we selected seven potential biomarkers, which had been reported to involve in DKD pathophysiological processes, including glomerular injury [angiopoietin-like-4 (ANGPTL4) [4]], renal tubule injury [neutrophil gelatinase-associated lipocalin (NGAL) [5]], inflammation and immunity [monocyte chemoattractant protein-1 (MCP-1) [5]], proliferation and fibrosis [growth differentiation factor-15 (GDF15) [6], fibroblast growth factor-23 (FGF23) [7] and n-terminal osteopontin (ntOPN) [8]], as well as energy metabolism [pyruvate kinase muscle isozyme M2 (PKM2) [9]].

We investigated their variations in DM populations with renal impairment. Subsequently, the biomarkers that exhibited significant association with DKD were selected to construct combined forecast models. Given that DKD is a major complication of DM, urine albumin-to-creatinine ratio (UACR) and estimated glomerular filtration rate (eGFR) are the final consequences of DKD; we hypothesized that these novel biomarkers could be more promising for DKD prediction.

Materials and methods

Study design and participant selection

This investigation employed a two-phase design, beginning with a cross-sectional analysis followed by a prospective observational cohort study. During the initial cross-sectional phase, the participants with DM were enrolled from the endocrinology and nephrology departments of Xuzhou Central Hospital (Xuzhou, China). Inclusion criteria: (1) patients diagnosed with T1DM ≥5 years, or those with T2DM [2,10,11]; (2) age ≥18 years. Exclusion criteria: (1) insufficient or incomplete clinical records; (2) presence of confirmed non-DKD, malignancies, active infections, systemic conditions (such as autoimmune disorders), significant cardiovascular or cerebrovascular disease, or malnutrition at initial assessment; (3) pregnant or breastfeeding women; (4) unwillingness to provide informed consent.

In the subsequent prospective phase, participants from the cross-sectional cohort were further screened, excluding those who: (1) had diagnosed DKD at baseline [2,10,11]; (2) exhibited persistent albuminuria and/or impaired renal function upon initial evaluation; (3) developed confirmed non-DKD during the follow-up period; (4) lacked complete data over the 2-year follow-up period. The prospective cohort was initiated in June 2021, with follow-up until the occurrence of DKD or the end of the study in December 2023 (Figure S1).

Ethics approval and consent to participate

Written informed consent was obtained from participants in the study. This study adhered to the International Conference on Harmonization guidelines for Good Clinical Practice and was conducted by the Declaration of Helsinki. Since this study is a continuation of our previous study [8], the ethics approval number remains the same. The protocol was approved by the Ethics Committee of Xuzhou Central Hospital (approval no. XZXY-LJ-20201030-055).

Sample size estimation

The formula for sample size calculation for the prospective cohort was as follows:

$$N=Z_{1-\frac{\alpha }{2}}^{2}p(1-p)/d^2$$

Assuming an α value of 0.05, Z_1-α/2 value of 1.96, d value of 0.1, and p is set as 0.5, therefore, N = 96. Considering the design effect as 1–3 and the drop-out rate as 10%, the sample size is from 107 (if design effect = 1) to 320 (if design effect = 3).

Data acquisition and measurements

Baseline demographic and clinical data were documented, covering age, sex, body mass index (BMI) and pre-existing medical conditions. Blood samples were obtained after an overnight fast, along with first-morning urine specimens. All laboratory analyses were conducted at the certified laboratory of Xuzhou Central Hospital. The plasma and urine samples were processed by centrifugation (10 min at 3500 rpm, 4 °C) and subsequently preserved at −80 °C until biomarker quantification was performed.

Enzyme-linked immunosorbent assay

The baseline plasma and urine samples of all participants were analyzed using competitive enzyme-linked immunosorbent assays (ELISAs). The selected proteins were ANGPTL4 (Cat. JL15325), NGAL (Cat. JL11263), MCP-1 (Cat. JL19334), GDF15 (Cat. JL19786), FGF23 (Cat. JL10444), ntOPN (Cat. JL20845) and PKM2 (Cat. JL48224). These were measured using ELISA test kits (Jianglai Bio, Shanghai, China), with a coefficient of variation (CV) < 10% in both intra- and inter-assay precision assessments. The urinary levels of the selected proteins were normalized using urinary creatinine measurements, with the results expressed as the protein-to-creatinine ratio.

Clinical outcome in the prospective study

Glomerular filtration rate (GFR) was assessed using: (1) the CKD-EPI equation incorporating both serum creatinine and cystatin-C (denoted as eGFRcr-cys) [5] and (2) the CKD-EPI equation based solely on serum creatinine (denoted as eGFRcr) [6]. Both eGFRs were reported in units of ml/min/1.73 m². The clinical outcome was the DKD onset, determined as: established a cause–effect relationship between prior DM and UACR/eGFR changes, excluding alternative glomerular and systemic diseases, and meeting one of the following criteria [10]: (1) repeated UACR measurements ≥30 mg/g in at least two of three tests over 3–6 months; (2) prolonged eGFR < 60 ml/min/1.73m² (>3 months); (3) kidney biopsy findings confirming DKD.

Statistical analysis

Baseline characteristics were presented as percentages for categorical data, and either median (IQR) or mean (±SD) for continuous variables, with normality evaluated using the Shapiro–Wilk test. Group comparisons were made using appropriate statistical tests (t-test, Wilcoxon test, chi-square, or Fisher’s exact test) based on data distribution patterns. The logistic regression analyses were conducted to evaluate the association between baseline biomarkers and the 2-year risk of DKD. The predictive performance of the established models for clinical outcomes was evaluated using receiver operating characteristic (ROC) curve analysis. To evaluate the improvement in predictive performance of the multivariate Cox proportional hazard models, the net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were employed as key evaluation metrics. Statistical significance was set at a two-sided P-value < 0.05. All analyses were conducted using SAS 9.4 (SAS Institute Inc., USA) and R version 4.5 (https://www.r-project.org).

Results

Baseline characteristics of the participants in the cross-sectional cohort

According to the inclusion and exclusion criteria, a total of 348 DM patients were enrolled in the cross-sectional phase, of which 121 patients had verified DKD, with a median age of 55 (46, 62) years and 39.9% of females. Compared with the DM patients without kidney injury, all seven selected biomarkers were elevated in urine in DKD patients (all p < 0.05, Table 1). In addition, the logistic regression analyses showed that the high levels of urinary ntOPN, GDF15, NGAL, MCP-1 and FGF23 significantly increase the risk of DKD diagnosis (OR > 1 and p < 0.05), while elevated plasma ANGPTL4 might decrease the risk of DKD diagnosis (OR < 1 and p < 0.001), indicating that the former five urinary biomarkers could be risk factors for DKD development, and the plasma ANGPTL4 might be a protective factor for DKD (Table 2).

Table 1.

Baseline characteristics of the enrolled 348 DM patients in the cross-sectional cohort.

	Total	DM	DKD
	N = 348	N = 227	N = 121	p overall
Sociodemographics:				.
Age	55.0 [46.0;62.0]	54.0 [46.0;61.0]	55.0 [47.0;65.0]	0.292
Female, n (%)	139 (39.9%)	89 (39.2%)	50 (41.3%)	0.788
BMI (kg/m2)	24.5 [22.1;27.2]	24.5 [22.1;27.1]	24.6 [21.8;27.6]	0.847
Smoking, n (%)	111 (31.9%)	73 (32.2%)	38 (31.4%)	0.982
Alcohol, n (%)	91 (26.1%)	64 (28.2%)	27 (22.3%)	0.289
Comorbid illnesses:				.
Hypertension, n (%)	146 (42.0%)	81 (35.7%)	65 (53.7%)	0.002
Hyperlipemia, n (%)	165 (47.4%)	112 (49.3%)	53 (43.8%)	0.383
Hyperuricemia, n (%)	55 (15.8%)	26 (11.5%)	29 (24.0%)	0.004
Medication:	348 (100%)	227 (100%)	121 (100%)	.
RAS inhibitor, n (%)	70 (20.1%)	40 (17.6%)	30 (24.8%)	0.147
SGLT-2 inhibitor, n (%)	106 (30.5%)	78 (34.4%)	28 (23.1%)	0.041
GLP-1RA, n (%)	0.00 [0.00;0.00]	0.00 [0.00;0.00]	0.00 [0.00;0.00]	0.148
Metformin, n (%)	156 (44.8%)	115 (50.7%)	41 (33.9%)	0.004
Laboratory data:				.
Plasma ANGPTL4 (pg/mL)	6.92 [5.29;8.53]	7.34 [5.83;8.95]	6.29 [3.87;7.81]	<0.001
Urinary ANGPTL4 (ng/mmol)	0.20 [0.16;0.32]	0.19 [0.15;0.31]	0.23 [0.16;0.33]	0.017
Plasma NGAL (ng/mL)	538 [418;698]	516 [412;590]	644 [429;811]	<0.001
Urinary NGAL (μg/mmol)	2.70 [1.39;6.54]	2.15 [1.24;4.04]	5.83 [2.18;23.4]	<0.001
Plasma PKM2 (pg/mL)	675 [378;1127]	635 [342;947]	850 [478;1724]	<0.001
Urinary PKM2 (ng/mmol)	13.1 [9.46;19.5]	12.4 [8.88;19.2]	15.3 [10.8;21.5]	0.010
Plasma MCP-1 (pg/mL)	334 [252;485]	334 [247;452]	333 [268;514]	0.197
Urinary MCP-1 (ng/mmol)	34.3 [18.6;74.8]	28.4 [17.2;42.7]	60.8 [33.9;139]	<0.001
Plasma GDF15 (pg/mL)	213 [124;379]	165 [113;285]	291 [141;606]	<0.001
Urinary GDF15 (ng/mmol)	382 [283;568]	352 [263;476]	478 [319;725]	<0.001
Plasma ntOPN (ng/mL)	168 [124;245]	167 [124;236]	172 [124;296]	0.267
Urinary ntOPN (μg/mmol)	38.8 [23.3;110]	36.0 [23.4;81.1]	50.7 [22.2;230]	0.060
Plasma FGF23 (pg/mL)	48.8 [43.3;57.0]	47.7 [42.8;53.8]	52.1 [44.4;67.4]	0.001
Urinary FGF23 (ng/mmol)	16.4 [11.7;24.7]	15.3 [11.4;23.0]	18.7 [12.3;28.5]	0.016
UACR (mg/g)	1.81 [0.97;5.04]	1.16 [0.73;1.76]	15.9 [4.30;41.2]	<0.001
TCO2 (mmol/L)	25.8 [23.8;27.5]	25.9 [23.9;27.6]	25.6 [23.5;27.5]	0.477
Albumin (g/L)	43.4 [39.8;45.6]	43.8 [41.4;46.2]	41.4 [35.0;45.0]	<0.001
Globulin (g/L)	25.9 [22.7;28.6]	26.0 [22.9;28.8]	25.9 [22.5;28.6]	0.790
Pre-albumin (mg/L)	241 [191;281]	244 [204;278]	228 [168;287]	0.178
GA%	16.9 [12.8;24.6]	17.0 [13.1;25.5]	16.7 [11.6;21.9]	0.080
α-HBDH (U/L)	143 [123;161]	137 [121;152]	153 [132;181]	<0.001
ALP (U/L)	88.0 [71.0;112]	88.0 [71.2;107]	94.0 [71.0;119]	0.128
CK (U/L)	77.0 [51.0;104]	76.0 [49.5;95.5]	82.0 [52.0;109]	0.353
BUN (mmol/L)	5.88 [4.81;7.46]	5.84 [4.90;6.94]	6.03 [4.65;8.67]	0.140
Cystatin C (mg/L)	0.88 [0.79;1.07]	0.86 [0.78;0.98]	1.02 [0.82;1.39]	<0.001
UA (μmol/L)	292 [241;372]	285 [234;348]	307 [254;419]	0.001
Creatinine (μmol/L)	48.1 [39.0;58.6]	47.0 [38.7;55.5]	53.0 [39.9;79.4]	0.001
eGFR (ml/min/1.73m2)	114 [100;127]	116 [107;128]	107 [89.2;123]	<0.001
eGFRcr_cys (ml/min/1.73m2)	103 [85.2;118]	106 [95.9;121]	93.5 [67.6;112]	<0.001
Fasting glucose (mmol/l)	8.52 [7.01;11.9]	8.24 [6.76;11.1]	9.28 [7.63;13.0]	0.068
HbA1c (%)	9.01 [7.34;11.0]	8.93 [7.34;10.9]	9.30 [7.40;11.3]	0.271
TCH (mmol/L)	4.62 [3.81;5.64]	4.53 [3.79;5.29]	5.05 [3.92;6.48]	0.002
TG (mmol/L)	1.39 [0.92;2.21]	1.32 [0.86;1.98]	1.69 [1.06;2.89]	0.001
HDL-C (mmol/L)	1.19 [1.01;1.39]	1.19 [1.02;1.38]	1.17 [0.97;1.40]	0.668
LDL-C (mmol/L)	2.79 [2.24;3.52]	2.68 [2.23;3.39]	3.09 [2.31;3.84]	0.004
TCH/HDL-C ratio	3.81 [3.26;4.78]	3.69 [3.16;4.50]	4.30 [3.38;5.42]	<0.001
APOB (g/L)	0.94 [0.76;1.21]	0.90 [0.76;1.08]	1.08 [0.81;1.36]	<0.001
APOA1 (g/L)	1.24 [1.06;1.43]	1.25 [1.08;1.42]	1.22 [1.03;1.47]	0.395
APOB/APOA1 ratio	0.77 [0.60;0.97]	0.73 [0.56;0.92]	0.88 [0.65;1.12]	<0.001
Lipoprotein a (mg/L)	205 [89.2;387]	192 [83.5;323]	228 [112;507]	0.014
Haemoglobin (g/L)	142 [126;152]	143 [130;154]	137 [118;148]	0.001
Neutrophils (%)	60.2 [52.8;70.2]	57.8 [51.5;64.3]	67.5 [56.2;89.1]	<0.001
NLR	2.02 [1.52;3.12]	1.93 [1.44;2.60]	2.35 [1.65;4.27]	0.002

Abbreviations: BMI, body mass index; DM, diabetes mellitus; DKD, diabetic kidney disease; RASI, renin–angiotensin system inhibitor; SGLT-2i, sodium-dependent glucose transporters 2 inhibitor; GLP-1RA, glucagon like peptide-1 receptor agonist; ANGPTL4, angiopoietin-like-4; NGAL, neutrophil gelatinase-associated lipocalin; PKM2, pyruvate kinase muscle isozyme M2; MCP-1, monocyte chemoattractant protein-1; GDF15, growth differentiation factor-15; ntOPN, n-terminal osteopontin; FGF23, fibroblast growth factor-23; UACR, urinary albumin to creatinine ratio; TCO2, total carbon dioxide; GA, glycosylated albumin; α-HBDH, α-hydroxybutyrate dehydrogenase; ALP, alkaline phosphatase; CK, creatine phosphokinase; BUN, blood urea nitrogen; UA, uric acid; eGFR, estimated glomerular filtration rate; TCH, total cholesterol; TG, triglyceride; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; APOA1, apolipoprotein A1; APOB, apolipoprotein B; NLR, neutrophil-to-lymphocyte ratio.

Table 2.

Potential DKD risk factors identified by univariate logistic regression analysis in the cross-sectional cohort.

Variable	OR	95%CI	P value
Age (years)	1.011	0.994–1.027	0.204
Female	1.092	0.697–1.711	0.701
BMI (kg/m2)	1.004	0.951–1.061	0.881
Smoking	0.966	0.601–1.552	0.886
Alcohol	0.732	0.436–1.226	0.236
Hypertension	2.092	1.336–3.277	0.001
Hyperlipemia	0.800	0.514–1.247	0.325
Hyperuricemia	2.437	1.359–4.37	0.003
RAS inhibitor	1.541	0.902–2.633	0.113
SGLT-2 inhibitor	0.575	0.348–0.952	0.031
GLP-1RA	0.547	0.24–1.249	0.152
Metformin	0.499	0.316–0.789	0.003
Plasma ANGPTL4 (pg/mL)	0.830	0.762–0.904	<0.001
Urinary ANGPTL4 (ng/mmol)	1.952	0.739–5.158	0.177
Plasma NGAL (ng/100 mL)	1.325	1.187–1.48	<0.001
Urinary NGAL (μg/mmol)	1.050	1.025–1.076	<0.001
Plasma PKM2 (pg/100 mL)	1.045	1.024–1.067	<0.001
Urinary PKM2 (ng/100 mmol)	1.098	0.731–1.648	0.653
Plasma MCP-1 (pg/100 mL)	1.077	0.989–1.172	0.088
Urinary MCP-1 (ng/100 mmol)	1.418	1.168–1.723	<0.001
Plasma GDF15 (pg/100 mL)	1.236	1.139–1.342	<0.001
Urinary GDF15 (ng/100 mmol)	1.202	1.107–1.305	<0.001
Plasma ntOPN (ng/100 mL)	1.179	0.992–1.4	0.062
Urinary ntOPN (μg/100 mmol)	2.106	1.471–3.015	<0.001
Plasma FGF23 (pg/100 mL)	4.675	1.608–13.592	0.005
Urinary FGF23 (ng/100 mmol)	3.925	1.325–11.623	0.014
UACR (mg/g)	1.454	1.269–1.665	<0.001
TCO2 (mmol/L)	0.972	0.908–1.04	0.411
Albumin (g/L)	0.894	0.859–0.93	<0.001
Globulin (g/L)	0.987	0.948–1.027	0.516
Pre-albumin (mg/L)	0.998	0.995–1.001	0.207
GA%	0.978	0.953–1.003	0.086
α-HBDH (U/L)	1.013	1.006–1.02	<0.001
ALP (U/L)	1.004	0.999–1.009	0.105
CK (U/L)	1.001	0.997–1.004	0.655
BUN (mmol/L)	1.100	1.025–1.18	0.008
Cystatin C (10 mg/L)	1.193	1.118–1.272	<0.001
UA (μmol/L)	1.004	1.002–1.006	<0.001
Creatinine (μmol/L)	1.026	1.015–1.037	<0.001
eGFR (ml/min/1.73m2)	0.979	0.97–0.989	<0.001
eGFRcr_cys (ml/min/1.73m2)	0.975	0.967–0.984	<0.001
Fasting glucose (mmol/l)	1.063	1.002–1.129	0.044
HbA1c (%)	1.068	0.977–1.168	0.147
TCH (mmol/L)	1.335	1.174–1.519	<0.001
TG (mmol/L)	1.271	1.089–1.484	0.002
HDL-C (mmol/L)	1.077	0.589–1.968	0.809
LDL-C (mmol/L)	1.385	1.156–1.66	<0.001
TCH/HDL-C ratio	1.390	1.185–1.632	<0.001
APOB (g/L)	3.808	2.059–7.043	<0.001
APOA1 (g/L)	0.932	0.467–1.859	0.841
APOB/APOA1 ratio	4.049	2.067–7.93	<0.001
Lipoprotein a (mg/L)	1.001	1–1.002	0.002
Haemoglobin (g/L)	0.975	0.962–0.988	<0.001
Neutrophils (%)	1.023	1.014–1.032	<0.001
NLR	1.133	1.045–1.229	0.002

Abbreviations: BMI, body mass index; OR, odds ratio; CI, confidence interval; DKD, diabetic kidney disease; RASI, renin–angiotensin system inhibitor; SGLT-2i, sodium-dependent glucose transporters 2 inhibitor; GLP-1RA, glucagon like peptide-1 receptor agonist; ANGPTL4, angiopoietin-like-4; NGAL, neutrophil gelatinase-associated lipocalin; PKM2, pyruvate kinase muscle isozyme M2; MCP-1, monocyte chemoattractant protein-1; GDF15, growth differentiation factor-15; ntOPN, n-terminal osteopontin; FGF23, fibroblast growth factor-23; UACR, urinary albumin to creatinine ratio; TCO2, total carbon dioxide; GA, glycosylated albumin; α-HBDH, α-hydroxybutyrate dehydrogenase; ALP, alkaline phosphatase; CK, creatine phosphokinase; BUN, blood urea nitrogen; UA, uric acid; eGFR, estimated glomerular filtration rate; TCH, total cholesterol; TG, triglyceride; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; APOA1, apolipoprotein A1; APOB, apolipoprotein B; NLR, neutrophil-to-lymphocyte ratio.

Validity of the seven biomarkers for DKD diagnosis in the cross-sectional cohort

To investigate the ability of the selected biomarkers to detect DKD from DM patients, we depicted their ROC curves for DKD diagnosis using a univariate logistic regression model, of which the urinary MCP-1 performed the best with the largest area under the ROC curve (AUC) (Figure 1A). Furthermore, after the adjustment for the clinical variables of age, eGFRcr-cys, BMI and systolic blood pressure (SBP), the urinary GDF15 performed the best in detecting DKD from DM patients, with the largest AUC using multivariable logistic regression model (Figure 1B), the NRI and IDI analysis revealed that the models of GDF15 combined clinical variables also performed better than the model of clinical variables (Table S1).

Figure 1.

The receiver operator characteristic (ROC) curves of models for DKD diagnosis in the cross-sectional cohort. (A) The ROC curves of seven urinary biomarkers for DKD diagnosis. (B) The ROC curves of seven urinary biomarkers for DKD diagnosis after the adjustment for the clinical variables of age, eGFRcr-cys, BMI and SBP. DKD, diabetic kidney disease; eGFR, estimated glomerular filtration rate; BMI, body mass index; SBP, systolic blood pressure; uGDF15, urinary growth differentiation factor-15; untOPN, urinary n-terminal osteopontin; uNGAL, urinary neutrophil gelatinase-associated lipocalin; uANGPTL4, urinary angiopoietin-like-4; uFGF23, urinary fibroblast growth factor-23; uMCP-1, urinary monocyte chemoattractant protein-1; uPKM2, urinary pyruvate kinase muscle isozyme M2.

Predictors for DKD occurrence in the prospective cohort

Based on enrolment and exclusion criteria, we recruited 141 DM patients without evidence of kidney injury into the prospective cohort; the median age was 53 (44, 61) years, with 36.9% of females. By the end of 2-year follow-up, a total of 51 DM patients were diagnosed with DKD. Compared with the patients without kidney involvement, the seven urinary biomarkers were all increased in the patients developing DKD during 2-year follow-up (all p < 0.001, Table 3). Moreover, the logistic regression analyses showed that the high levels of urinary GDF15, MCP-1, ANGPTL4 and FGF23 significantly increased the risk of DKD development (OR > 1 and p < 0.01), indicating that these four urinary biomarkers could be predictors for DKD development (Table 4).

Table 3.

Baseline characteristics stratified by DKD onset in the enrolled 141 DM patients in the prospective cohort.

	Total	DM	DKD onset
	N = 141	N = 90	N = 51	p overall
Sociodemographics:				.
Age (years)	53.0 [44.0;61.0]	53.0 [44.2;59.0]	55.0 [43.5;62.5]	0.240
Female, n (%):	52 (36.9%)	33 (36.7%)	19 (37.3%)	1.000
1	89 (63.1%)	57 (63.3%)	32 (62.7%)
2	52 (36.9%)	33 (36.7%)	19 (37.3%)
BMI (kg/m2)	24.8 [22.8;27.7]	25.0 [22.9;28.2]	24.7 [22.7;27.1]	0.384
Smoking, n (%):	45 (31.9%)	23 (25.6%)	22 (43.1%)	0.050
Alcohol, n (%):	37 (26.2%)	22 (24.4%)	15 (29.4%)	0.711
Comorbid illnesses:				.
Hypertension, n (%):	55 (39.0%)	33 (36.7%)	22 (43.1%)	0.564
Hyperlipemia, n (%):	61 (43.3%)	35 (38.9%)	26 (51.0%)	0.224
Hyperuricemia, n (%):	18 (12.8%)	13 (14.4%)	5 (9.80%)	0.596
Medication:	141 (100%)	90 (100%)	51 (100%)	.
RAS inhibitor, n (%):	27 (19.1%)	18 (20.0%)	9 (17.6%)	0.906
SGLT-2 inhibitor, n (%):	45 (31.9%)	24 (26.7%)	21 (41.2%)	0.112
GLP-1RA, n (%):	20 (14.2%)	13 (14.4%)	7 (13.7%)	1.000
Metformin, n (%):	63 (44.7%)	41 (45.6%)	22 (43.1%)	0.919
Laboratory data:				.
Plasma ANGPTL4 (pg/mL)	6.96 [5.39;8.60]	7.34 [5.71;8.87]	6.39 [5.02;8.16]	0.065
Urinary ANGPTL4 (ng/mmol)	2.04 [1.55;3.28]	1.77 [1.52;2.67]	3.15 [1.92;4.44]	<0.001
Plasma NGAL (ng/mL)	524 [411;656]	522 [420;684]	551 [377;632]	0.417
Urinary NGAL (μg/mmol)	2.79 [1.28;6.47]	2.16 [1.08;4.60]	4.61 [2.07;9.67]	<0.001
Plasma PKM2 (pg/mL)	635 [350;947]	635 [324;895]	671 [359;961]	0.538
Urinary PKM2 (ng/mmol)	13.6 [8.75;19.5]	11.4 [8.19;16.6]	18.9 [13.3;28.8]	<0.001
Plasma MCP-1 (pg/mL)	340 [242;463]	345 [251;452]	334 [218;471]	0.827
Urinary MCP-1 (ng/mmol)	31.5 [17.3;69.0]	20.9 [14.3;35.4]	69.0 [30.0;115]	<0.001
Plasma GDF15 (pg/mL)	164 [112;276]	154 [112;240]	232 [115;406]	0.118
Urinary GDF15 (ng/mmol)	360 [257;517]	337 [240;425]	499 [319;771]	<0.001
Plasma ntOPN (ng/mL)	170 [120;231]	178 [131;239]	159 [110;223]	0.112
Urinary ntOPN (μg/mmol)	5.47 [2.68;15.7]	3.55 [2.51;7.67]	11.6 [5.86;34.2]	<0.001
Plasma FGF23 (pg/mL)	47.9 [42.6;53.4]	46.8 [42.5;52.3]	50.3 [43.8;54.2]	0.087
Urinary FGF23 (ng/mmol)	15.4 [11.4;24.1]	13.5 [10.9;20.1]	21.4 [14.9;34.1]	<0.001
UACR (mg/g)	1.39 [0.91;2.00]	1.19 [0.76;1.48]	2.04 [1.67;2.34]	<0.001
TCO2 (mmol/L)	25.8 [23.9;27.3]	25.8 [23.9;27.0]	25.8 [23.9;27.8]	0.509
Albumin (g/L)	43.9 [40.8;46.1]	43.9 [41.5;46.0]	43.7 [40.2;46.2]	0.822
Globulin (g/L)	25.6 [22.5;28.9]	25.4 [22.5;28.8]	25.8 [22.8;28.8]	0.663
Pre-albumin (mg/L)	242 [197;277]	244 [208;280]	225 [186;264]	0.066
GA%	16.7 [13.1;23.5]	16.7 [13.3;24.0]	16.0 [12.6;22.8]	0.673
α-HBDH (U/L)	140 [123;155]	137 [124;152]	141 [119;170]	0.391
ALP (U/L)	85.3 [70.8;103]	85.6 [68.2;108]	85.0 [73.0;96.5]	0.749
CK (U/L)	77.0 [55.0;106]	74.5 [48.0;95.8]	85.0 [60.0;113]	0.040
BUN (mmol/L)	5.83 [4.84;6.93]	5.82 [4.79;6.75]	5.83 [4.99;7.30]	0.310
Cystatin C (mg/L)	0.88 [0.78;0.99]	0.88 [0.77;0.98]	0.88 [0.79;1.02]	0.601
UA (μmol/L)	297 [243;367]	312 [250;371]	275 [234;352]	0.129
Creatinine (μmol/L)	47.3 [39.3;56.7]	48.3 [40.2;57.6]	45.0 [37.6;55.2]	0.213
eGFR (ml/min/1.73m2)	117 [108;128]	117 [109;124]	117 [101;133]	0.632
eGFRcr_cys (ml/min/1.73m2)	105 [93.6;121]	105 [96.9;118]	112 [83.6;123]	0.827
HbA1c (%)	8.90 [7.30;11.0]	9.00 [7.85;11.0]	8.09 [7.10;11.2]	0.376
TCH (mmol/L)	4.57 [3.77;5.25]	4.50 [3.76;5.12]	4.58 [3.80;5.55]	0.267
TG (mmol/L)	1.30 [0.91;2.01]	1.29 [0.91;1.96]	1.32 [0.92;2.26]	0.707
HDL-C (mmol/L)	1.17 [1.00;1.37]	1.15 [0.99;1.33]	1.25 [1.04;1.40]	0.267
LDL-C (mmol/L)	2.68 [2.24;3.21]	2.70 [2.31;3.15]	2.67 [2.21;3.51]	0.817
TCH/HDL-C ratio	3.69 [3.18;4.52]	3.68 [3.15;4.50]	3.78 [3.30;4.69]	0.524
APOB (g/L)	0.90 [0.75;1.07]	0.90 [0.73;1.07]	0.90 [0.76;1.10]	0.696
APOA1 (g/L)	1.22 [1.07;1.42]	1.22 [1.06;1.42]	1.26 [1.08;1.42]	0.986
APOB/APOA1 ratio	0.72 [0.58;0.93]	0.70 [0.58;0.92]	0.76 [0.58;0.96]	0.565
Lipoprotein a (mg/L)	197 [76.2;307]	206 [77.4;312]	158 [74.5;296]	0.659
Neutrophils (%)	59.5 [52.0;69.7]	59.3 [51.5;64.7]	61.5 [52.8;73.6]	0.137
NLR	1.90 [1.42;2.66]	1.90 [1.42;2.52]	1.90 [1.43;3.36]	0.560

Abbreviations: BMI, body mass index; DM, diabetes mellitus; DKD, diabetic kidney disease; RASI, renin–angiotensin system inhibitor; SGLT-2i, sodium-dependent glucose transporters 2 inhibitor; GLP-1RA, glucagon like peptide-1 receptor agonist; ANGPTL4, angiopoietin-like-4; NGAL, neutrophil gelatinase-associated lipocalin; PKM2, pyruvate kinase muscle isozyme M2; MCP-1, monocyte chemoattractant protein-1; GDF15, growth differentiation factor-15; ntOPN, n-terminal osteopontin; FGF23, fibroblast growth factor-23; UACR, urinary albumin to creatinine ratio; TCO2, total carbon dioxide; GA, glycosylated albumin; α-HBDH, α-hydroxybutyrate dehydrogenase; ALP, alkaline phosphatase; CK, creatine phosphokinase; BUN, blood urea nitrogen; UA, uric acid; eGFR, estimated glomerular filtration rate; TCH, total cholesterol; TG, triglyceride; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; APOA1, apolipoprotein A1; APOB, apolipoprotein B; NLR, neutrophil-to-lymphocyte ratio.

Table 4.

Risk factors for DKD occurrence identified by univariate logistic regression analysis in the prospective cohort.

Variable	OR	95%CI	P value
Age (years)	1.016	0.991–1.042	0.217
Female	1.026	0.504–2.089	0.945
BMI (kg/m2)	0.963	0.887–1.045	0.365
Smoking	2.210	1.066–4.581	0.033
Alcohol	1.349	0.621–2.932	0.449
Hypertension	1.310	0.65–2.64	0.45
Hyperlipemia	1.634	0.817–3.27	0.165
Hyperuricemia	0.644	0.216–1.923	0.43
RAS inhibitor	0.857	0.353–2.079	0.733
SGLT-2 inhibitor	1.925	0.93–3.985	0.078
GLP-1RA	0.942	0.35–2.538	0.906
Metformin	0.907	0.454–1.812	0.781
Plasma ANGPTL4 (pg/mL)	0.880	0.761–1.018	0.086
Urinary ANGPTL4 (ng/mmol)	1.435	1.142–1.803	0.002
Plasma NGAL (ng/100 mL)	0.905	0.749–1.094	0.302
Urinary NGAL (μg/mmol)	1.000	0.989–1.011	0.946
Plasma PKM2 (pg/100 mL)	1.029	0.993–1.067	0.121
Urinary PKM2 (ng/100 mmol)	1.062	0.58–1.943	0.847
Plasma MCP-1 (pg/100 mL)	1.016	0.872–1.183	0.843
Urinary MCP-1 (ng/100 mmol)	4.433	1.903–10.329	0.001
Plasma GDF15 (pg/100 mL)	1.161	0.978–1.379	0.089
Urinary GDF15 (ng/100 mmol)	1.416	1.199–1.672	<0.001
Plasma ntOPN (ng/100 mL)	0.843	0.606–1.173	0.311
Urinary ntOPN (μg/100 mmol)	3.630	0.997–13.215	0.051
Plasma FGF23 (pg/100 mL)	1.155	0.252–5.294	0.852
Urinary FGF23 (ng/100 mmol)	34.426	2.643–448.465	0.007
UACR (mg/g)	1.261	1.165–1.365	<0.001
TCO2 (mmol/L)	0.976	0.879–1.085	0.656
Albumin (g/L)	0.972	0.9–1.051	0.479
Globulin (g/L)	1.035	0.985–1.088	0.175
Pre-albumin (mg/L)	0.996	0.99–1.001	0.095
GA%	0.996	0.953–1.04	0.838
α-HBDH (U/L)	1.009	0.999–1.019	0.065
ALP (U/L)	0.996	0.987–1.006	0.449
CK (U/L)	1.007	1.001–1.013	0.015
BUN (mmol/L)	1.034	0.916–1.166	0.591
Cystatin C (10 mg/L)	1.049	0.933–1.179	0.424
UA (μmol/L)	0.998	0.994–1.001	0.21
Creatinine (μmol/L)	1.001	0.982–1.021	0.902
eGFR (ml/min/1.73m2)	1.001	0.983–1.019	0.917
eGFRcr_cys (ml/min/1.73m2)	0.996	0.981–1.012	0.629
HbA1c (%)	0.962	0.828–1.117	0.61
TCH (mmol/L)	1.211	0.924–1.588	0.165
TG (mmol/L)	1.126	0.87–1.458	0.365
HDL-C (mmol/L)	1.315	0.481–3.593	0.593
LDL-C (mmol/L)	1.050	0.713–1.545	0.806
TCH/HDL-C ratio	1.188	0.886–1.592	0.251
APOB (g/L)	1.465	0.463–4.633	0.516
APOA1 (g/L)	0.732	0.205–2.617	0.632
APOB/APOA1 ratio	1.656	0.455–6.034	0.444
Lipoprotein a (mg/L)	1.000	0.999–1.002	0.492
Neutrophils (%)	1.004	0.996–1.013	0.329
NLR	1.175	1.001–1.378	0.048

Abbreviations: BMI, body mass index; OR, odds ratio; CI, confidence interval; DKD, diabetic kidney disease; RASI, renin–angiotensin system inhibitor; SGLT-2i, sodium-dependent glucose transporters 2 inhibitor; GLP-1RA, glucagon like peptide-1 receptor agonist; ANGPTL4, angiopoietin-like-4; NGAL, neutrophil gelatinase-associated lipocalin; PKM2, pyruvate kinase muscle isozyme M2; MCP-1, monocyte chemoattractant protein-1; GDF15, growth differentiation factor-15; ntOPN, n-terminal osteopontin; FGF23, fibroblast growth factor-23; UACR, urinary albumin to creatinine ratio; TCO2, total carbon dioxide; GA, glycosylated albumin; α-HBDH, α-hydroxybutyrate dehydrogenase; ALP, alkaline phosphatase; CK, creatine phosphokinase; BUN, blood urea nitrogen; UA, uric acid; eGFR, estimated glomerular filtration rate; TCH, total cholesterol; TG, triglyceride; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; APOA1, apolipoprotein A1; APOB, apolipoprotein B; NLR, neutrophil-to-lymphocyte ratio.

Validity of the seven urinary biomarkers for predicting 2-year risk of incident DKD in DM patients

To investigate the ability of the seven biomarkers to detect DKD from DM patients, we depicted their ROC curves for DKD prediction using a univariate logistic regression model, of which, both urinary MCP-1 alone and combined with clinical variables performed the best with the largest AUC (Figure S2, Figure 2A and Table S2). Using the potential predictors selected by logistic regression analyses, whose OR > 1 and p < 0.05, a series of biomarker panels were set up for prediction models of 2-year risk of DKD occurrence. After the adjustment for the clinical variables of age, eGFRcr-cys, UACR, BMI and SBP, the model based on four biomarkers of GDF15, MCP-1, ANGPTL4 and FGF23 had the significant largest AUC (0.873) for predicting the 2-year risk of DKD occurrence, indicating that this four-biomarker panel might play an important role in the clinical prediction of DKD (Figure 2B and Table S3).

Figure 2.

The receiver operator characteristic (ROC) curves of prediction models for DKD onset in the prospective cohort. (A) The ROC curves of seven urinary biomarkers for DKD prediction. (B) The ROC curves of combined urinary biomarkers for DKD prediction after the adjustment for the clinical variables of age, eGFRcr-cys, BMI and SBP. DKD, diabetic kidney disease; eGFR, estimated glomerular filtration rate; BMI, body mass index; SBP, systolic blood pressure; uGDF15, urinary growth differentiation factor-15; untOPN, urinary n-terminal osteopontin; uNGAL, urinary neutrophil gelatinase-associated lipocalin; uANGPTL4, urinary angiopoietin-like-4; uFGF23, urinary fibroblast growth factor-23; uMCP-1, urinary monocyte chemoattractant protein-1; uPKM2, urinary pyruvate kinase muscle isozyme M2.

In addition, we conducted an extensive internal comparative analysis evaluating the performance of the models of four biomarkers combined with clinical variables using seven machine learning algorithms. These algorithms were trained and tested on the same standardized dataset using identical preprocessing (including normalization) and a consistent 7:3 train–test split protocol. The detailed performance metrics for all models on the independent testing dataset are presented in Table 5. Among them, LightGBM and Support Vector Machine (SVM) demonstrated superior performance. LightGBM achieved the highest sensitivity (0.770) and F1 score (0.739), making it ideal for high-recall scenarios (e.g. reducing missed diagnoses). In contrast, SVM led in AUC (0.864) and positive predictive value (0.779), making it better suited to high-precision tasks (e.g. confirming diagnosis). Naive Bayes demonstrated exceptional specificity (0.912), but critically low sensitivity (0.515), which increased the risk of missing cases. The Decision Tree and Random Forest/XGBoost models offered balanced but unremarkable accuracy (0.787). These results demonstrated that model selection should be guided by clinical priorities, whether emphasizing recall or precision, and validated for generalizability beyond any single metric.

Table 5.

Performance of DKD forecasting models in the testing dataset.

	Logistic regression	Decision tree	Support vector machine	Random forest	Naive Bayes	XGBoost	LightGBM
TPR	0.685	0.606	0.725	0.671	0.515	0.671	0.770
TNR	0.869	0.874	0.857	0.848	0.912	0.856	0.833
PPV	0.738	0.720	0.779	0.744	0.774	0.739	0.737
NPV	0.830	0.814	0.836	0.811	0.768	0.817	0.866
ACC	0.802	0.780	0.808	0.787	0.766	0.787	0.808
F1 score	0.708	0.631	0.732	0.701	0.593	0.696	0.739
AUC	0.817	0.811	0.864	0.837	0.841	0.775	0.861

Abbreviations: TPR, true positive rate (sensitivity/recall); TNR, true negative rate (specificity); PPV, positive predictive value (precision); NPV, negative predictive value; ACC, accuracy; AUC, area under curve.

Discussion

DKD is a major contributor to end-stage renal disease (ESRD) worldwide. Our previous research revealed that the 5-year survival rate for ESRD patients is <65.7% [5], with annual treatment costs reaching approximately US$15,000 per patient in China [6]. The rising incidence of ESRD imposes substantial economic and healthcare burdens. According to a decision model assessing DKD management strategies, early screening combined with renin–angiotensin system (RAS) inhibitor therapy proved more cost-effective and yielded better health outcomes than no intervention in individuals with T2DM [7]. Given that DKD onset is often silent and multifactorial, a DKD prediction model could shift the management of DKD from a ‘passive response’ to ‘proactive prevention’. It enables early identification of high-risk patients and facilitates personalized treatment decisions, thereby improving the overall prognosis of DM patients.

The identification of biomarkers associated with the development of DKD has been a research hotspot, with substantial literature emerging on this topic [12,13]. Based on different pathways of DKD pathogenesis, this study screened seven emerging biomarkers in recent years and compared their roles in the diagnosis and prediction of DKD.

MCP-1, also known as CC chemokine ligand 2 (CCL2), belongs to the chemokine family. MCP-1 binds to CC-chemokine receptor-2 (CCR2) and recruits monocyte/macrophages infiltration to promote inflammation and serves as a mediator of inflammation and innate immunity [14]. MCP-1 is overexpressed in the glomeruli and renal tubular epithelial cells, inducing local inflammation by promoting the infiltration and migration of macrophages, T cells and dendritic cells to the diabetic kidney. For instance, MCP-1 induces nephrin loss and apoptosis in podocytes via TGF-β1 and enhances the production of fibronectin and collagen in mesangial cells [15]. The serum levels of MCP-1 were higher in DM patients with microalbumin than those without microalbumin, which were closely related to renal injury and poor prognosis in DKD patients [16]. Urinary MCP-1 was correlated with interstitial fibrosis surface (IFS) from the very early phases of DKD [5]. Compared with healthy controls, urinary MCP-1 was elevated in DM patients and even higher in patients with DKD. The urinary MCP-1 could be a promising inflammatory biomarker and independent predictor for diagnosing and predicting early DKD [17] and other diabetic microvascular complications [14]. In this study, our findings revealed that urinary MCP-1 outperformed the other six biomarkers in both diagnosing DKD and predicting incident DKD. MCP-1 could be the most promising biomarker among the seven selected proteins, indicating that inflammation and immunity should play important roles in very early kidney injury in DM patients. Moreover, the models of urinary MCP-1 combined with clinical variables further enhanced the predictive accuracy for DKD prediction in patients with DM.

GDF15, a member of the transforming growth factor β (TGF-β) superfamily, known as macrophage inhibitory cytokine-1 (MIC-1), is a stress-inducing cytokine. GDF15 is involved in various pathological conditions through regulating metabolism, stress, proliferation and fibrosis. Increased circulating GDF15 was related to high risk of T2DM complications, especially DKD [18]. Further, urinary GDF15 was significantly higher in patients with histologic diagnosis of DKD than in diabetic patients without DKD [19]. Urinary GDF-15 could serve as a prognostic biomarker for DKD progression [6]. FGF23, a bone-secreted phosphaturic hormone, is a key factor participating in regulating vitamin D and calcium-phosphate homeostasis [20]. In DM patients, the serum FGF23 increases at early stage of CKD progression (eGFR < 75 ml/min/1.73m²), even before the increase of serum phosphate. The elevated FGF23 enhanced phosphate excretion in proximal tubules and helped to maintain stable levels of serum phosphorus [21]. Serum FGF23 level was inversely correlated with the eGFR and was an independent predictor for ESRD in the patients with DKD [22]. Moreover, baseline serum FGF23 increased in the early stage of DKD and could independently predict DKD incidence [7]. However, the roles of urinary FGF23 in DKD development and whether it can also predict DKD onset and progression in DM patients remain unclear. Our results revealed that elevated urinary levels of GDF15 and FGF23 were associated with both DKD diagnosis and DKD development, indicating that proliferation and fibrosis had already manifested in the kidneys of DM patients before the onset of clinical symptoms of DKD.

Osteopontin (OPN) is also a profibrotic adhesion phosphoprotein, which participates in cell chemotaxis, adhesion, migration and proliferation, as well as extracellular matrix (ECM) hyperplasia [23], and is involved in multiple inflammatory and autoimmune processes [24]. OPN has important functions in DM patients: (1) promotes atherosclerotic plaque formation to aggravate its vascular complications, such as cardiovascular diseases and DKD; (2) increases macrophage accumulation in adipose tissue to cause insulin resistance and inflammation; (3) promotes cell proliferation and growth in glomeruli. Therefore, decreased plasma OPN may help to prevent diabetic complications [23]. Previous studies reported that the plasma OPN was independently associated with the incipient DKD and all-cause mortality in DM patients [25]. OPN structurally contains two terminal zones, the N-terminal and C-terminal. The ntOPN has a stronger profibrotic adhesion effect than the full-length OPN and plays an important role in mediating inflammation [26]. Our previous study reported that urinary ntOPN is an independent predictor for DKD occurrence and progression [8]. The results of this study were consistent with our previous report.

NGAL, a 25-kDa protein in the lipocalin family, is usually up-regulated in injured renal epithelial cells and excreted in the urine [5]. NGAL levels have histological correlation with progressive renal lesions, particularly in proximal and distal tubular lesions. Therefore, the serum and urinary NGAL could be an innovative and independent marker for renal tubular injury of DKD [5]. Urinary NGAL was up-regulated before/without the presence of albuminuria, and it could be an early marker of tubular injury in both T2DM patients [27] and T1DM patients [28]. We previously reported that elevated urinary NGAL was a risk factor for DKD development. The combination of urinary NGAL and ntOPN could further improve the predictive ability for DKD progression compared with classical biomarkers of serum creatinine plus UACR [8]. This study found that elevated urinary NGAL was associated with DKD diagnosis, while it might not be valuable for DKD prediction.

ANGPTL4, a 50-kDa glycoprotein that belongs to the family of angiopoietin-like proteins, is secreted by a variety of cells, including adipocytes, cardiomyocytes, endothelial cells and macrophages, which inhibits lipoprotein lipase, and therefore modulates lipid metabolism, angiogenesis, vascular permeability and inflammation [29]. Previous studies reported that circulating ANGPTL4 significantly increased in DKD patients and could be a predictive indicator for early nephropathy in T2DM patients [30]. Podocyte- and tubule-derived ANGPTL4 induces podocyte and endothelial cell damage, as well as vascular endothelial dysfunction, thereby leading to albuminuria and promoting DKD progression [4]. Our study revealed that urinary ANGPTL4 was elevated in both the DKD population and those at high risk of DKD onset; urinary ANGPTL4 could be a predictor for DKD development, while increased plasma ANGPTL4 might be a protective factor against DKD.

PKM2, a critical glycolytic enzyme converting phosphoenolpyruvate to pyruvate, plays an important role in regulating energy metabolism, including glycolytic flux, and tricarboxylic acid cycle/mitochondrial pathways [31]. PKM2 and its post-translational modifications play multifaceted roles in CKD progression by regulating metabolic processes, fibroblast proliferation, podocyte injury, as well as macrophage polarization. Intriguingly, both PKM2 activators have demonstrated therapeutic potential for DKD, indicating its pleiotropic effects in kidney disease [32]. Additionally, podocyte-specific deletion of PKM2 aggravated podocyte with foot process effacement, which might be due to the subsequently compromised glycolytic flux and insufficient energy supply to foot process, leading to its cytoskeletal remodelling and apoptosis[33]; thus, restoring PKM2 activation might prevent and even reverse the DKD pathogenesis [9]. Our results revealed that although urinary PKM2 was elevated in both the DKD population and those at high risk of DKD, it was not a significant risk factor for neither DKD diagnosis nor DKD development, suggesting that energy metabolism may not be the key initiating factor in the development of DKD.

This study exhibits several strengths. First, few literatures compared the role of the selected seven biomarkers in early screening of DKD and predicting renal outcomes in DM population. This research stands as the first endeavour to address and compare their clinical implementation in the diagnosis and prediction of DKD, which could help to establish accurate and non-invasive models to detect and predict DKD. Second, our study demonstrated that all seven urinary biomarkers were associated with DKD development, while the urinary MCP-1 seemed to be the best predictor. This observation also suggested that urinary MCP-1 levels might hold promise for diagnosing and predicting DKD in DM patients.

This study has limitations. First, this was a single-centre study, and the selection of patients could have introduced bias. Second, the sample size was relatively small, and the follow-up duration was relatively short. Third, the current KDIGO guideline [34] recommended initiating RAS inhibitors, SGLT-2 inhibitors and metformin as first-line therapies in patients with DKD. Nevertheless, these medications were prescribed at very low rates among our participants. As a result, no significant association was observed between the use of these agents and reduced incidence of DKD in the prospective cohort, and the suboptimal treatment may partly explain the relatively accelerated renal outcomes in this study. Therefore, further research is needed that includes a larger sample size, a longer-term follow-up, external validation and a more dedicated design for medications, lifestyles and dietary variables.

In conclusion, our findings showed that urinary MCP-1 performed the best in diagnosing and predicting DKD in DM patients. Compared with the model based on classical biomarkers of eGFR combined with UACR, our four-biomarker model performed better in predicting DKD onset, which could further improve the predictive ability of models and provide more accurate tools for DKD assessment. Our results might be useful for early screening and intervention of DKD, thereby helping to prevent and delay DKD progression in DM patients.

Funding Statement

This research was funded by the Major Project of Philosophy and Social Science Research in Colleges and Universities of Jiangsu Province (grant no. 2024SJZD062), the Research Project of Jiangsu Commission of Health (grant no. ZD2022044), the Clinical Observership in NHS Hospitals London U.K Project of Jiangsu Commission of Health, the Preventive Medicine Research Project of Jiangsu Commission of Health (grant no. Y12023008), 333 High-level Personnel Cultivation Project in Jiangsu Province (grant no. (2022)3-12-151) and the Science and Technology Foundation of the Xuzhou Commission of Health (grant no. XWKYHT20240026).

Ethics approval and consent to participate

Written informed consent was obtained from participants in the study. This study adhered to the International Conference on Harmonization guidelines for Good Clinical Practice and was conducted by the Declaration of Helsinki. Since this study is a continuation of our previous study, the ethics approval number remains the same. The protocol was approved by the Ethics Committee of Xuzhou Central Hospital (approval no. XZXY-LJ-20201030-055).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Table S1. The IDI and NRI for multivariate Cox proportional hazard models in Figure 1B.

Table S2. The IDI and NRI for multivariate Cox proportional hazard models in Figure 2 A.

Table S3. The IDI and NRI for multivariate Cox proportional hazard models in Figure 2B.

Figure S1. Overview of study design and cohort participants.

Figure S2. The receiver operator characteristic (ROC) curves of prediction models for DKD onset in the prospective cohort, using seven urinary biomarkers. uGDF15, urinary growth differentiation factor-15; untOPN, urinary n-terminal osteopontin; uNGAL, urinary neutrophil gelatinase-associated lipocalin; uANGPTL4, urinary angiopoietin-like-4; uFGF23, urinary fibroblast growth factor-23; uMCP-1, urinary monocyte chemoattractant protein-1; uPKM2, urinary pyruvate kinase muscle isozyme M2.

Data availability statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.