Selective glomerular hypofiltration syndrome: a novel renal phenotype associated with cardiovascular disease and adverse outcomes in diabetic kidney disease

Abstract

Introduction

Selective glomerular hypofiltration syndrome (SGHS), defined by a low ratio of estimated glomerular filtration rate using cystatin C (eGFRcys) to estimated glomerular filtration rate using creatinine (eGFRcr), has emerged as a novel kidney dysfunction phenotype. Its role in diabetic kidney disease (DKD) remains unknown. This study aimed to investigate the association between SGHS and the prevalence of cardiovascular disease (CVD) at the time of renal biopsy-diagnosed DKD, as well as its association with long-term renal outcomes.

Research design and methods

This study included 182 patients with biopsy-proven DKD. SGHS was defined as eGFRcys/eGFRcr ratio <0.60. Predictive value of the SGHS for the prevalence of CVD was assessed using receiver operating characteristic (ROC) curves. Independent risk factors for CVD prevalence were identified by logistic regression and used to construct a nomogram. ROC curves, calibration curves, decision curve analysis, and clinical impact curves were used for assessment of model performance. The impact of SGHS on renal outcomes was assessed using Kaplan-Meier curves and Cox regression. Subgroup analyses stratified by gender and body mass index categories were performed. Sensitivity analysis was conducted by redefining SGHS as eGFRcys/eGFRcr ratio <0.70.

Results

Of 182 patients, 41 (22.5%) had SGHS. The eGFRcys/eGFRcr ratio better predicted CVD (area under the curve (AUC) 0.730, 95% CI 0.643 to 0.816) than eGFRcr (AUC 0.602, 95% CI 0.507 to 0.697), eGFRcys (AUC 0.701, 95% CI 0.615 to 0.786), estimated glomerular filtration rate based on creatinine and cystatin C (AUC 0.685, 95% CI 0.595 to 0.774), and urinary albumin-to-creatinine ratio (AUC 0.633, 95% CI 0.540 to 0.725). SGHS was independently associated with CVD after adjustment (OR 9.27, 95% CI 3.20 to 26.84; p<0.001). A nomogram incorporating SGHS demonstrated good discrimination (AUC 0.834; bias-corrected C-index 0.808) and calibration. Over a median follow-up of 30.1 months, patients with SGHS had higher cumulative DKD progression probability (log-rank test: χ²=11.79, p<0.001), and SGHS was an independent risk factor for DKD progression (HR 2.645, 95% CI 1.564 to 4.472; p<0.001). Subgroup analyses and sensitivity analysis confirmed the robustness of these findings.

Conclusions

SGHS is independently associated with the prevalence of CVD at the time of DKD diagnosis, and independently predicts adverse renal outcomes in patients with DKD. As an easily obtainable parameter, the eGFRcys/eGFRcr ratio may serve as a valuable risk assessment tool in patients with DKD.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• In patients with biopsy-proven DKD, SGHS independently predicts both CVD and adverse renal outcomes.

• The estimated glomerular filtration rate using cystatin C (eGFRcys) to estimated glomerular filtration rate using creatinine (eGFRcr) ratio outperforms traditional markers for CVD prediction, and a novel nomogram incorporating SGHS facilitates CVD risk assessment in patients with DKD.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The eGFRcys/eGFRcr ratio, an easily obtainable parameter, could be integrated into routine DKD evaluation to identify high-risk patients, guide personalized interventions, and potentially reduce cardiovascular and renal events.

Introduction

Diabetic kidney disease (DKD) is one of the most common and serious microvascular complications of diabetes and has become the leading cause of end-stage renal disease (ESRD) worldwide.¹ Cardiovascular disease (CVD) accounts for over 50% of mortality in patients with DKD.² Notably, up to 90% of patients with DKD die from cardiovascular causes before reaching kidney failure.³ Therefore, early identification of risk factors for CVD and appropriate intervention are crucial for reducing mortality and improving the prognosis of patients with DKD.

Selective glomerular hypofiltration syndrome (SGHS), first described as ‘shrunken pore syndrome’ by Grubb et al, who proposed that a low estimated glomerular filtration rate using cystatin C (eGFRcys)/estimated glomerular filtration rate using creatinine (eGFRcr) ratio reflects reduced glomerular pore size and impaired filtration of medium-sized molecules (5–30 kDa, such as cystatin C).⁴ This concept was later refined as impaired filtration of medium-sized molecules due to shrinking or elongation of glomerular filtration barrier pores,⁵ and renamed as SGHS. The absolute cut-off value of the eGFRcys/eGFRcr ratio for diagnosing SGHS has not yet been established. The first study on this syndrome used a ratio <0.60,⁴ while most current studies define SGHS using an eGFRcys/eGFRcr ratio of either 0.70 or 0.60.^6–9 The shrinking and elongation of the glomerular filtration barrier pores in patients with DKD has been confirmed by prior research, which revealed a negative association between glomerular basement membrane thickness and the eGFRcys/eGFRcr ratio in patients with DKD.¹⁰ The similar phenomenon has also been observed in DKD rat models.^{11 12} Recent studies have demonstrated a close association between SGHS and CVD, including the progression of atherosclerosis,¹³ myocardial infarction,¹⁴ heart failure,¹⁵ and increased carotid intima-media thickness.¹⁶ The most widely accepted mechanism is that selective hypofiltration in SGHS leads to the accumulation of 5–30 kDa plasma proteins, such as cytokines and inflammatory mediators, many of which are recognized as driver proteins of CVD.¹⁷

A decline in eGFRcr is observed in only a few patients with early-stage DKD. However, the shrinking or elongation of the glomerular filtration barrier pores can impair the clearance of medium-sized molecules such as cystatin C, causing a reduction in eGFRcys.^{10 18} Therefore, SGHS may indicate early glomerular filtration dysfunction in DKD. Currently, no study has linked SGHS to the incidence of CVD and long-term renal outcomes in patients with DKD. This study aims to clarify the above associations, providing new insights into reducing CVD risk and improving prognosis in patients with DKD.

Materials and methods

This study consisted of two analytical components. First, a cross-sectional analysis was performed to evaluate the association between SGHS and the prevalence of CVD at baseline (ie, at the time of renal biopsy-confirmed DKD diagnosis). Second, a longitudinal cohort analysis was conducted to investigate the association between SGHS and long-term renal outcomes during follow-up.

Study population

The study included patients diagnosed with DKD by renal biopsy from June 1, 2015 to June 30, 2025 at the First Affiliated Hospital of Fujian Medical University. All patients were >18 years of age. The exclusion criteria were as follows: (1) combined with other primary or secondary glomerular diseases diagnosed by renal biopsy (n=16); (2) ESRD at renal biopsy, which is defined as baseline eGFRcr <15 mL/min/1.73 m² or have already entered renal replacement therapy (n=6); (3) combined with malignant tumor (n=3); (4) combined with severe infections (n=4); (5) pregnant at renal biopsy (n=0); (6) treated with high doses of corticosteroids (n=1); (7) combined with thyroid disease (n=2); (8) baseline data or follow-up data are missing too much (n=8). Finally, 182 patients were included. Included patients were divided into two groups according to the presence (n=41) or absence (n=141) of SGHS and followed up by December 31, 2025.

Data collection

Baseline demographic characteristics, comorbidities, clinical risk factors, and clinical treatment data were collected from medical records. Laboratory data measured at admission included cystatin C, serum creatinine (Scr), blood urea nitrogen (BUN), urinary albumin-to-creatinine ratio (UACR), 24-hour urinary protein, fasting plasma glucose (FPG), glycosylated hemoglobin (HbA1c), total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), hemoglobin (Hb), serum albumin, uric acid, fibrinogen (FIB), and parathyroid hormone (PTH). All patients underwent renal biopsy and renal histopathological examination. The total number of glomeruli, the number of glomeruli with sclerosis, and the presence of Kimmelstiel-Wilson (K-W) nodules were recorded. DKD was classified into grade I–IV according to RPS criteria.¹⁹ Follow-up indicators, including the last follow-up time, last reassessment of Scr and eGFRcr, whether endpoint events occurred, and the time of endpoint event occurrence, were obtained through outpatient consultation or telephone interviews.

Outcomes

Baseline was defined as the date of DKD diagnosis through renal biopsy. All patients were followed up by December 31, 2025. The primary outcome was DKD progression, defined as serum creatinine doubling or eGFRcr decline exceeding 50% from baseline, or initiation of renal replacement therapy (kidney transplantation, hemodialysis, or peritoneal dialysis) during follow-up.

Definitions

Diabetes mellitus was defined by WHO 1999 criteria²⁰ and DKD was defined by American Diabetes Association 2023 criteria.²¹ CVD was defined as the presence of any of the following conditions, based on either participants’ self-report of having been diagnosed by a physician or a review of previous medical records: (1) heart failure: a clinical syndrome caused by structural/functional cardiac abnormality, with elevated natriuretic peptides or evidence of congestion²²; (2) coronary heart disease: chronic cardiac or vascular disease due to reduced coronary blood flow, including angina pectoris, previous myocardial infarction, or coronary revascularization procedures²³; (3) angina pectoris: transient chest discomfort from reversible myocardial ischemia due to increased workload in fixed coronary stenosis²⁴; (4) myocardial infarction: detection of a rise/fall in cardiac biomarkers (preferably troponin) above the 99th percentile, with symptoms, electrocardiographic changes, or imaging evidence of new myocardial loss or wall motion abnormality²⁵; (5) stroke: rapidly developing focal/global cerebral dysfunction of vascular origin, with symptoms lasting ≥24 hours or leading to death.²⁶ eGFRcys was calculated using the 2012 Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equations.²⁷ The eGFRcr and estimated glomerular filtration rate based on creatinine and cystatin C (eGFRcr-cys) were calculated using the 2021 CKD-EPI equations.²⁸ In our study, an eGFRcys/eGFRcr ratio <0.60 was used to diagnose SGHS, as this ratio was associated with high specificity and good sensitivity for predicting mortality and morbidity.¹⁷

Statistical analysis

R software, V.4.4.1 and SPSS software, V.25.0 were used for statistical analysis. Missing continuous data were handled by mean imputation. Baseline characteristics between groups were compared using Student’s t-test (normally distributed data, mean±SD), Mann-Whitney U test (non-normal data, median (IQR)), and χ² test (categorical variables). The predictive value of the eGFRcys/eGFRcr ratio, eGFRcr, eGFRcys, eGFRcr-cys, and UACR for CVD prevalence at baseline was analyzed using receiver operating characteristic (ROC) curves. Logistic regression was performed to identify risk factors of CVD prevalence at baseline. Variables with p<0.05 in univariate analysis or those considered clinically important were included in the multivariate logistic regression models. Model 1 was adjusted for demographic characteristics, model 2 was adjusted for the variables in model 1 and medication use, model 3 was adjusted for the variables in model 2 and some significant laboratory indicators. The statistically significant variables from model 3 were used to construct a nomogram, validated by ROC curve, calibration curve, decision curve analysis curve, and clinical impact curve.

Survival analysis was performed using the Kaplan-Meier method and the log-rank test was used to compare the cumulative probability of DKD progression between patients with and without SGHS. Univariate and multivariate Cox regression were used to identify factors affecting DKD progression. Collinearity was assessed using variance inflation factor.

Subgroup analyses stratified by gender and BMI were performed to evaluate the interaction effects, and the results are presented in the form of forest plots.

Sensitivity analysis was conducted by redefining SGHS as eGFRcys/eGFRcr ratio <0.70. P value <0.05 was considered statistically significant.

Results

Comparison of baseline characteristics between patients with DKD with and without SGHS

Baseline characteristics of patients are shown in table 1. Among 182 patients with DKD, 41 (22.5%) had SGHS. A total of 45 patients (24.7%) had CVD, including 23 cases (16.3%) in the group without SGHS and 22 cases (53.7%) in the group with SGHS. Among the cohort, 129 (70.9%) were male. The median age at the time of renal biopsy was 55.00 (48.00, 61.00) years, and the median duration of diabetes was 10.00 (5.00, 17.00) years. Patients with SGHS had a higher prevalence of hypertension and diuretic use. In this cohort, patients with SGHS had significantly higher cystatin C levels, resulting in a markedly lower eGFRcys, whereas no significant differences were observed between the two groups in Scr levels and eGFRcr. Patients with SGHS also had higher levels of UACR and lower levels of serum albumin. The pathological grade of patients with DKD in this cohort was mainly grade III, accounting for 87 cases (47.8%). The glomerulosclerosis rate, presence of K-W nodules, and distribution of pathological grades did not differ significantly between the two groups.

Table 1. Comparison of baseline characteristics in patients with/without SGHS.

	All patients (n=182)	With/Without SGHS		t/Z/χ2 value	P value
		Without SGHS (n=141)	With SGHS (n=41)
General data
Age (years)	55.00 (48.00, 61.00)	54.00 (47.00, 60.50)	57.00 (49.00, 63.50)	1.441	0.150
Gender, male, n (%)	129 (70.9%)	98 (69.5%)	31 (75.6%)	0.574	0.449
DM duration (years)	10.00 (5.00, 17.00)	10.00 (5.00, 16.00)	10.00 (4.50, 18.00)	0.708	0.479
Weight (kg)	66.00 (57.08, 73.50)	66.00 (57.60, 73.10)	64.00 (57.00, 75.00)	−0.029	0.977
BMI (kg/m2)	23.88 (21.86, 25.94)	23.95 (22.06, 25.94)	23.44 (21.39, 25.78)	−1.186	0.236
SBP (mm Hg)	150.01±23.60	148.94±24.62	153.66±19.53	0.993	0.584
DBP (mm Hg)	83.56±13.81	83.58±13.80	83.49±14.04	0.987	0.089
Smoking, n (%)	63 (34.6%)	46 (32.6%)	17 (41.5%)	1.097	0.295
CVD, n (%)	45 (24.7%)	23 (16.3%)	22 (53.7%)	23.803	<0.001*
Hypertension, n (%)	162 (89.0%)	122 (86.5%)	40 (97.6%)	3.955	0.047*
Gout, n (%)	21 (11.5%)	19 (13.5%)	2 (4.9%)	2.300	0.129
Use of SGLT2 inhibitors, n (%)	56 (30.8%)	43 (30.5%)	13 (31.7%)	0.022	0.882
Use of ACEI/ARB, n (%)	125 (68.7%)	94 (66.7%)	31 (75.6%)	1.181	0.277
Use of GLP-1 RA, n (%)	7 (3.8%)	6 (4.3%)	1 (2.4%)	0.283	0.595
Use of metformin, n (%)	78 (42.9%)	63 (44.7%)	15 (36.6%)	0.850	0.357
Use of statins, n (%)	80 (44.0%)	61 (43.3%)	19 (46.3%)	0.122	0.727
Use of diuretics, n (%)	43 (23.6%)	27 (19.1%)	16 (39.0%)	6.954	0.008*
Laboratory data
FPG (mmol/L)	7.14 (5.13, 9.50)	7.10 (5.03, 9.38)	7.52 (5.77, 10.56)	1.074	0.283
HbA1c (%)	7.65 (6.50, 9.10)	7.80 (6.60, 9.10)	7.30 (6.35, 8.95)	−0.830	0.406
TC (mmol/L)	5.28 (4.05, 6.65)	5.28 (4.03, 6.40)	5.15 (4.16, 7.73)	0.716	0.474
TG (mmol/L)	1.73 (1.09, 2.52)	1.72 (1.07, 2.54)	1.78 (1.14, 2.47)	0.141	0.888
HDL-C (mmol/L)	1.09 (0.89, 1.37)	1.09 (0.88, 1.36)	1.13 (0.91, 1.42)	0.754	0.451
LDL-C (mmol/L)	3.47 (2.31, 4.50)	3.40 (2.30, 4.36)	3.63 (2.47, 4.97)	1.209	0.227
Scr (μmol/L)	127.65 (91.75, 171.08)	129.00 (93.50, 180.65)	121.80 (87.60, 160.50)	−1.206	0.228
eGFRcr (mL/min/1.73 m2)	52.75 (38.38, 76.70)	51.00 (37.55, 75.75)	56.70 (40.80, 86.70)	1.303	0.192
eGFRcys (mL/min/1.73 m2)	40.95 (27.80, 59.68)	47.00 (38.80, 63.60)	29.20 (20.25, 39.35)	−4.791	<0.001*
Cystatin C (mg/L)	1.62 (1.24, 2.13)	1.44 (1.20, 2.07)	2.08 (1.67, 2.83)	4.704	<0.001*
UACR (mg/g)	3243.67 (1633.10, 5455.07)	2916.53 (1392.36, 4901.46)	4789.84 (2457.03, 7460.08)	3.134	0.002*
CRP (mg/L)	2.53 (1.26, 5.30)	2.54 (1.26, 5.31)	2.26 (1.24, 5.16)	0.128	0.898
Hb (g/L)	110.22±23.49	109.83±23.57	111.56±23.44	0.988	0.109
BUN (mmol/L)	8.96 (6.49, 13.01)	9.34 (6.75, 13.09)	7.74 (6.15, 12.11)	−1.415	0.157
Uric acid (μmol/L)	370.50 (317.38, 427.93)	369.00 (318.25, 424.50)	384.00 (316.35, 433.00)	0.618	0.537
Serum albumin (g/L)	30.60 (25.98, 35.80)	31.30 (26.50, 35.95)	28.10 (23.75, 33.90)	−2.129	0.033*
Urinary protein in 24 hours (g/24 hours)	3.86 (2.00, 7.33)	3.65 (1.77, 7.25)	4.65 (2.99, 8.54)	1.729	0.084
Calcium (mmol/L)	2.08 (1.99, 2.20)	2.10 (1.99, 2.22)	2.06 (2.00, 2.16)	−0.938	0.348
Phosphorus (mmol/L)	1.27 (1.12, 1.42)	1.30 (1.13, 1.43)	1.25 (1.03, 1.35)	−1.273	0.203
Potassium (mmol/L)	4.17 (3.68, 4.56)	4.20 (3.71, 4.60)	3.88 (3.61, 4.41)	−1.581	0.114
FIB (g/L)	4.61 (3.79, 5.74)	4.57 (3.82, 5.55)	4.93 (3.73, 6.28)	0.823	0.410
PTH (pmol/L)	5.38 (2.92, 6.51)	5.09 (2.84, 6.28)	5.73 (3.71, 7.29)	1.444	0.149
Pathological feature
Glomerulosclerosis rate (%)	40.00 (14.30, 66.70)	40.00 (15.85, 66.65)	37.50 (13.30, 69.05)	−0.280	0.780
K-W nodules, n (%)	126 (69.2%)	94 (66.7%)	32 (78.0%)	1.932	0.165
Pathological classification, n (%)				7.812	0.099
I	4 (2.2%)	2 (1.4%)	2 (4.9%)
IIa	23 (12.6%)	19 (13.5%)	4 (9.8%)
IIb	22 (12.1%)	21 (14.9%)	1 (2.4%)
III	87 (47.8%)	67 (47.5%)	20 (48.8%)
IV	46 (25.3%)	32 (22.7%)	14 (34.1%)

Data are expressed as means±SD, medians (IQR), or counts (%).

^*P<0.05.

ACEI, ACE inhibitors; ARB, angiotensin receptor antagonist; BMI, body mass index; BUN, blood urea nitrogen; CRP, C reactive protein; CVD, cardiovascular disease; DBP, diastolic blood pressure; DM, diabetes mellitus; eGFRcr, estimated glomerular filtration rate using creatinine; eGFRcys, estimated glomerular filtration rate using cystatin C; FIB, fibrinogen; FPG, fasting blood glucose; GLP-1 RA, glucagon-like peptide-1 receptor agonist; Hb, hemoglobin; HbA1c, glycosylated hemoglobin, type A1c; HDL-C, high-density lipoprotein cholesterol; K-W, Kimmelstiel-Wilson; LDL-C, low-density lipoprotein cholesterol; PTH, parathyroid hormone; SBP, systolic blood pressure; Scr, serum creatinine; SGHS, selective glomerular hypofiltration syndrome; SGLT2, sodium-dependent glucose transporters 2; TC, total cholesterol; TG, triglycerides; UACR, urinary albumin to-creatinine ratio.

Predictive value of SGHS for CVD in patients with DKD

ROC curve analysis was used to compare the predictive value of the eGFRcys/eGFRcr ratio, eGFRcr, eGFRcys, eGFRcr-cys, and UACR for CVD in patients with DKD. The eGFRcys/eGFRcr ratio showed the best predictive value for CVD prevalence. It had an area under the curve (AUC) of 0.730 (95% CI 0.643 to 0.816; p<0.001). The best cut-off was 0.685. In comparison, eGFRcr alone had an AUC of 0.602 (95% CI 0.507 to 0.697; p=0.040), eGFRcys alone had an AUC of 0.701 (95% CI 0.615 to 0.786; p<0.001), eGFRcr-cys had an AUC of 0.685 (95% CI 0.595 to 0.774; p<0.001), UACR had an AUC of 0.633 (95% CI 0.540 to 0.725; p=0.008) (figure 1). Furthermore, 24-hour urinary protein showed limited discriminative ability for CVD prevalence, with an AUC of 0.515 (95% CI 0.419 to 0.610; p=0.770) (online supplemental figure S1).

Figure 1. ROC curves of eGFRcys/eGFRcr, eGFRcr, eGFRcys, eGFRcr-cys, and UACR for CVD in patients with DKD. CVD, cardiovascular disease; DKD, diabetic kidney disease; eGFRcr, estimated glomerular filtration rate using creatinine; eGFRcr-cys, estimated glomerular filtration rate based on creatinine and cystatin C; eGFRcys, estimated glomerular filtration rate using cystatin C; ROC, receiver operating characteristic; UACR, urinary albumin-to-creatinine ratio.

Risk prediction model of CVD for patients with DKD

Univariate and multivariate logistic regression analysis were used to examine the association between SGHS and CVD in patients with DKD. SGHS was strongly associated with CVD in the unadjusted model (OR 5.94, 95% CI 2.78 to 12.69; p<0.001). The association stayed significant after adjusting for demographic factors in model 1 (OR 7.37, 95% CI 3.14 to 17.33; p<0.001). It did not change after further adjusting for medication use in model 2 (OR 7.23, 95% CI 2.94 to 17.76; p<0.001). In model 3, we added significant laboratory indicators and other variables that were statistically significant in univariate analysis. SGHS remained a strong predictor of CVD (OR 9.27, 95% CI 3.20 to 26.84; p<0.001) (online supplemental table S1, table 2).

Table 2. Logistic regression analysis for the relationship of SGHS and CVD.

	Unadjusted		Model 1		Model 2		Model 3
	OR (95% CI)	P value	OR (95% CI)	P value	OR (95% CI)	P value	OR (95% CI)	P value
SGHS	5.94 (2.78 to 12.69)	<0.001*	7.37 (3.14 to 17.33)	<0.001*	7.23 (2.94 to 17.76)	<0.001*	9.27 (3.20 to 26.84)	<0.001*
Age			1.02 (0.99 to 1.06)	0.224	1.02 (0.99 to 1.06)	0.212	1.02 (0.98 to 1.07)	0.323
Male			0.32 (0.12 to 0.87)	0.026*	0.30 (0.11 to 0.85)	0.023*	0.23 (0.07 to 0.80)	0.021*
BMI			1.16 (1.05 to 1.29)	0.005*	1.15 (1.04 to 1.28)	0.006*	1.16 (1.04 to 1.29)	0.007*
Smoking			4.40 (1.73 to 11.20)	<0.001*	5.04 (1.90 to 13.38)	0.001*	6.04 (2.04 to 17.86)	0.001*
Use of SGLT2 inhibitors					1.18 (0.50 to 2.78)	0.706	1.42 (0.54 to 3.72)	0.473
Use of ACEI/ARB					0.55 (0.23 to 1.27)	0.161	0.84 (0.29 to 2.43)	0.746
Use of statins					1.18 (0.53 to 2.65)	0.682	1.36 (0.52 to 3.53)	0.528
Use of diuretics					2.08 (0.85 to 5.09)	0.111	1.82 (0.64 to 5.21)	0.261
LDL-C							1.26 (1.02 to 1.56)	0.035*
TG							0.61 (0.36 to 1.02)	0.057
eGFRcr							0.98 (0.95 to 1.00)	0.079
UACR							1.00 (1.00 to 1.00)	0.941
BUN							0.94 (0.84 to 1.07)	0.352
Uric acid							1.00 (0.99 to 1.01)	0.683
Phosphorus							1.51 (0.21 to 10.59)	0.681
PTH							1.12 (1.02 to 1.24)	0.022*
Hypoproteinemia							1.30 (0.36 to 4.66)	0.685

^*P<0.05.

ACEI, ACE inhibitors; ARB, angiotensin receptor antagonist; BMI, body mass index; BUN, blood urea nitrogen; CVD, cardiovascular disease; eGFRcr, estimated glomerular filtration rate using creatinine; LDL-C, low-density lipoprotein cholesterol; PTH, parathyroid hormone; SGHS, selective glomerular hypofiltration syndrome; SGLT2, sodium-dependent glucose transporters 2; TG, triglycerides; UACR, urinary albumin-to-creatinine ratio.

The independent risk factors selected from model 3 were used to construct a risk prediction model by nomogram (figure 2A). The model demonstrated good discrimination, with an AUC of 0.834 (95% CI 0.769 to 0.898) (figure 2B) and a bias-corrected C-index of 0.808 (bootstrap, 1000 resamples). The calibration curve showed a good fit with the reference line, suggesting that the model is well-calibrated (figure 2C). Decision curve and clinical impact curves confirmed its net clinical benefit across a range of thresholds (figure 2D–2E).

Figure 2. Risk prediction model for CVD in patients with DKD. (A) Nomogram for predicting risk of CVD in patients with DKD. (B) ROC curve of the prediction model. (C) Calibration curve of the predictive model. (D) Decision curve analysis of the prediction model. (E) Clinical impact curve of the prediction model. AUC, area under the curve; BMI, body mass index; CVD, cardiovascular disease; DKD, diabetic kidney disease; LDL, low-density lipoprotein; PTH, parathyroid hormone; ROC, receiver operating characteristic; SGHS, selective glomerular hypofiltration syndrome.

Long-term renal outcomes in patients with DKD with SGHS

During a median follow-up of 30.1 (25.55, 34.65) months, 91 patients (50.0%) reached renal endpoints. Kaplan-Meier analysis showed higher cumulative incidence of renal progression in patients with DKD with SGHS (log-rank test: χ²=11.79, p<0.001) (figure 3).

Figure 3. Comparison of cumulative probability of DKD progression in patients with/without SGHS. DKD, diabetic kidney disease; SGHS, selective glomerular hypofiltration syndrome.

Univariate Cox regression analysis was conducted to find the risk factors for DKD progression (online supplemental table S2). Variables with p<0.05 were entered into multivariate models. Collinearity diagnostics showed the multicollinearity among the variables eGFRcr, eGFRcys, and eGFRcr-cys, while the eGFRcys/eGFRcr ratio was not collinear with any of them (online supplemental figure S2). Given that eGFRcr is the most commonly used clinical metric for CKD risk stratification, we selected it for multivariate Cox regression analysis to construct the primary model. The results showed that SGHS, CVD, use of diuretics, TG, urinary protein in 24 hours, eGFRcr, Scr, and Hb were independent risk factors for DKD progression (p<0.05) (figure 4, online supplemental table S3). To further compare the predictive value of eGFRcys and eGFRcr-cys with that of eGFRcys/eGFRcr for DKD progression, we constructed another two multivariate Cox regression models. Model A included eGFRcys and model B included eGFRcr-cys. Both models were adjusted for the same confounders identified in the univariate analysis. The results showed that eGFRcys in model A (HR 1.010, 95% CI 0.988 to 1.031, p=0.373) and eGFRcr-cys in model B (HR 1.009, 95% CI 0.987 to 1.032, p=0.427) were not independent risk factors for DKD progression. In contrast, SGHS remained a significant and independent risk factor in both models (model A: HR 3.289, 95% CI 1.736 to 6.231, p<0.001; model B: HR 3.097, 95% CI 1.728 to 5.552, p<0.001) (online supplemental table S4). To compare the independent predictive value of each filtration metric, we constructed four separate multivariate Cox regression models, each containing only one of the core predictors (SGHS, eGFRcr, eGFRcys, or eGFRcr-cys), all adjusted for the same confounders identified in the univariate analysis. As shown in online supplemental table S5, SGHS (HR 2.771, 95% CI 1.689 to 4.548, p<0.001) and eGFRcr (HR 0.978, 95% CI 0.961 to 0.996, p=0.015) were significantly associated with DKD progression, whereas eGFRcys and eGFRcr-cys were not (p>0.05). These findings confirm that eGFRcr remains a robust predictor of DKD progression, consistent with the established literature, and further support the independent prognostic value of SGHS beyond conventional filtration metrics.

Figure 4. Multivariate Cox regression analysis of factors affecting the renal prognosis in patients with DKD. CVD, cardiovascular disease; DKD, diabetic kidney disease; eGFRcr, estimated glomerular filtration rate using creatinine; Hb, hemoglobin; Scr, serum creatinine; SGHS, selective glomerular hypofiltration syndrome; TG, triglycerides.

Subgroup analyses

Since serum creatinine is influenced by muscle mass, we performed the subgroup analyses stratified by gender and BMI categories to assess the robustness of our findings. The subgroup analysis based on multivariate logistic regression models showed that SGHS remained independently associated with an increased CVD prevalence in all the subgroups (p<0.05) (online supplemental figure S3). The subgroup analysis based on multivariate Cox regression models showed that SGHS remained the independent risk factors for DKD progression irrespective of gender and BMI categories (p<0.05) (online supplemental figure S4). There were no significant interaction effects observed among the various subgroups (p for interaction >0.05).

Sensitivity analysis

There are no standardized cut-off values that define SGHS currently. In most studies, an eGFRcys/eGFRcr ratio <0.60 or <0.70 was used to diagnose SGHS. Therefore, we performed a sensitivity analysis to assess the reliability of our results by redefining SGHS as an eGFRcys/eGFRcr ratio <0.70. The findings showed that SGHS remained significantly associated with CVD and long-term renal prognosis in patients with DKD, regardless of whether the cut-off was set at 0.6 or 0.7 (online supplemental table S6 and online supplemental figure S5).

Discussion

To our knowledge, this is the first study revealing the association between SGHS and both CVD prevalence and long-term renal prognosis in patients with DKD. The study suggested that SGHS was closely linked to CVD prevalence and renal prognosis. SGHS showed better predictive value for CVD than eGFRcr, eGFRcys, eGFRcr-cys, and UACR in patients with DKD. After adjusting for confounders, SGHS, gender, smoking history, BMI, LDL-C, and PTH were independent risk factors of CVD and were used to build a risk prediction model. Moreover, SGHS was independently associated with DKD progression. Subgroup analyses stratified by gender and BMI categories confirmed that the predictive value of SGHS for CVD prevalence and DKD progression was not affected by muscle mass-related variations in serum creatinine. A sensitivity analysis confirmed the robustness of these findings across different SGHS cut-off values.

SGHS may reflect early glomerular filtration barrier injury, characterized by endothelial cell swelling, which narrows the endothelial pores, and by glomerular basement membrane thickening, which elongates the pores.^{5 18} Current KDIGO criteria for kidney dysfunction rely on eGFR and albuminuria.²⁹ However, studies in healthy populations have shown that the morbidity of kidney disease significantly rises when the eGFRcys/eGFRcr ratio is <1.0,^{7 30} and SGHS was proven to be a hallmark of premature kidney aging in a population free from chronic kidney disease and diabetes.³¹ A systematic review and meta-analysis showed that a lower eGFRcys relative to eGFRcr was consistently associated with higher mortality and elevated risks of cardiovascular and renal outcomes.³² These findings suggested that a significant number of patients with serious kidney disease would be missed if assessment were based solely on creatinine-based GFR estimates. A longitudinal observational study of type 2 diabetes found that a low eGFRcys/eGFRcr ratio predicts mortality independently, including in people without conventional DKD markers.³³ Therefore, adding SGHS assessment in the population with diabetes may facilitate the early detection of renal impairment and improve outcomes.

In our study, 22.5% of patients with biopsy-proven DKD met the SGHS criteria using a cut-off of <0.60, consistent with previous reports in various populations. For example, two large cohort studies reported a prevalence of 27% and 36% using a cut-off of <0.60 in patients undergoing elective coronary artery bypass grafting⁶ and in a population with measured GFR,⁷ respectively. Patients with SGHS had a higher prevalence of hypertension and diuretic use, suggesting greater fluid retention or cardiovascular comorbidity, and supporting a potential link between SGHS and CVD.

Both albuminuria and eGFRcr are established predictors of cardiorenal outcomes in type 2 diabetes.³⁴ A recent prospective study showed that patients with DKD with reduced eGFRcr or both reduced eGFRcr and elevated UACR had a significantly higher risk of CVD, whereas those with elevated UACR alone did not have a heightened risk.³⁵ Conversely, data from NHANES and the UK Biobank showed that the albuminuria-only phenotype was associated with a higher risk of CVD mortality than the reduced eGFRcr-only phenotype.³⁶ Our study found that the eGFRcys/eGFRcr ratio had a significantly higher predictive value for CVD than either eGFRcr or UACR in patients with DKD. This is consistent with the report by Ito et al, who showed that eGFRcys outperformed eGFRcr and UACR for coronary heart disease risk prediction in patients with obesity and diabetes.³⁷ However, our findings further suggest that the eGFRcys/eGFRcr ratio provides better CVD predictive value than either eGFRcys alone or eGFRcr-cys in patients with DKD. This indicates that the ratio captures a unique phenotype (SGHS) that cannot be identified by a single biomarker, thereby offering better predictive ability for CVD prevalence. Consistent with prior reports linking a low ratio to increased CVD risk,^{6 8 13–16 38} our study extends these findings to the DKD population and further confirms that SGHS is an independent risk factor of CVD after adjustment for traditional cardiovascular risk factors and laboratory variables. Therefore, we suggest adding the eGFRcys/eGFRcr ratio to routine DKD assessment for better CVD risk stratification.

The mechanism linking SGHS to CVD remains unclear but may involve proteomic alterations driven by low eGFRcys, leading to accumulation of pro-inflammatory molecules and promoting systemic inflammation, atherosclerosis, and endothelial dysfunction.³⁹ Proteomic analyses identified ≥66 elevated plasma proteins in patients with SGHS, 22 of which are recognized driver proteins of CVD.^{13 40} For example, elevated levels of soluble tumor necrosis factor receptor superfamily member 1 (TNFR1) and soluble interleukin-2 receptor subunit alpha have been observed in patients with SGHS.⁴⁰ Both proteins were associated with a higher risk of heart failure in a multi-ethnic cohort of 2867 participants without previous history of CVD.⁴¹ The accumulation of these proatherogenic and pro-inflammatory proteins may contribute to the increased cardiovascular burden in patients with DKD with SGHS.

Our study constructed a CVD risk prediction model for patients with DKD using the independent risk factors identified from the adjusted logistic regression model, namely SGHS, gender, smoking history, BMI, LDL-C, and PTH. A nomogram was developed to facilitate individualized risk assessment and guide personalized intervention strategies. To our knowledge, no previous study has developed a CVD risk prediction model that incorporates SGHS as a variable in patients with DKD. Our nomogram represents a novel approach to risk stratification in this high-risk population.

Beyond CVD, our study also demonstrated that SGHS was strongly associated with long-term renal prognosis in DKD. Patients with SGHS had a significantly higher cumulative probability of DKD progression, and SGHS remained an independent risk factor for DKD progression after adjusting for well-established prognostic factors such as Scr, eGFRcr, proteinuria, and other variables. Our study also indicated that eGFRcys/eGFRcr provides an additional advantage in independently predicting DKD prognosis compared with GFRcys or eGFRcr-cys. While SGHS has been associated with increased risk of acute kidney injury^42–45 and all-cause mortality^{6 7 9} in several clinical settings, its relationship with renal outcomes in DKD remains unexplored. Our findings suggest that SGHS has prognostic value for renal endpoints in DKD.

We observed that UACR was not independently associated with DKD progression in the multivariate models, whereas 24-hour urinary protein retained its independent association. This discrepancy may be attributable to the greater intra-individual variability of UACR compared with 24-hour urinary protein, as UACR is influenced by hydration status, urine collection timing, and variations in urinary creatinine excretion. In contrast, 24-hour urinary protein provides a more stable quantitative measure of total protein excretion, which may explain its retained significance in our models. However, the significance of 24-hour urinary protein became borderline when eGFRcys or eGFRcr-cys replaced eGFRcr, likely because cystatin C-based markers have stronger associations with proteinuria than eGFRcr, as supported by previous studies demonstrating that cystatin C reflects albuminuria trends more accurately than serum creatinine,⁴⁶ and that eGFRcys, but not eGFRcr, is independently associated with moderate albuminuria.⁴⁷ Therefore, the contribution of proteinuria would be diluted when included together in the model with cystatin C-based markers, which are more strongly associated with proteinuria. Notably, the majority of patients in our cohort had significant proteinuria, which placed them at high risk for DKD progression but limited the discriminatory ability of proteinuria measures within this cohort. In this context, SGHS provided additional prognostic information beyond proteinuria, suggesting its potential value as a complementary marker for risk stratification in DKD populations where proteinuria levels are uniformly elevated.

Identification of SGHS and its associated plasma proteomic changes may open new therapeutic avenues in DKD. For example, monoclonal antibodies or receptor antagonists targeting CVD driver proteins identified in patients with SGHS may help reduce cardiovascular burden.⁵ In the Canagliflozin Cardiovascular Assessment Study (CANVAS) trial, the SGLT2 inhibitor canagliflozin attenuated the rise in TNFR1 and TNFR2, with the greatest cardiovascular and renal risk reductions observed in patients with the highest baseline levels of these biomarkers.⁴⁸ A prospective study has shown that SGLT2 inhibitor-induced albuminuria reduction correlated with decreased serum TNFR levels.⁴⁹ These findings suggest that by lowering levels of TNFRs and other potentially SGHS-related pro-inflammatory proteins, SGLT2 inhibitors may partially offset the inflammatory consequences of SGHS-mediated protein accumulation, thereby reducing CVD risk. Additionally, animal studies have shown that SGLT2 inhibitors ameliorated glomerular basement membrane (GBM) thickening and podocyte injury in DKD models.^{50 51} By mitigating GBM thickening, SGLT2 inhibitors may structurally improve the selective filtration capacity of the glomerular barrier for medium-sized molecules. Additionally, adequate dietary protein intake, particularly from non-red meat sources, has been associated with reduced all-cause and cardiovascular mortality in individuals with SGHS.⁵²

Several limitations should be acknowledged. First, this retrospective, single-center study with a moderate sample size may limit the generalizability of our findings. External validation in larger, multicenter cohorts is warranted. Second, as CVD status was assessed at baseline rather than during follow-up, we were unable to evaluate the association between SGHS and incident CVD events. Prospective studies with longer follow-up are warranted. Third, the SGHS cut-off value remains non-standardized. Although sensitivity analysis using a cut-off of <0.70 supported our findings, the optimal threshold needs further validation. Fourth, GLP-1 RAs have become a prevalent therapy for type 2 diabetes and have been shown to improve cardiovascular and renal outcomes. However, the low utilization of GLP-1 RAs in our cohort limited our ability to assess whether GLP-1 RAs modify the association between SGHS and outcomes, warranting future studies with larger sample sizes. Finally, due to the limited sample size, we could not analyze the association between SGHS and distinct manifestations of CVD, such as heart failure and myocardial infarction. We commit to expanding the cohort to further investigate the predictive value of SGHS for these specific CVD manifestations.

Conclusion

In this study, we demonstrated for the first time that SGHS, an early manifestation of glomerular filtration dysfunction, was independently associated with the prevalence of CVD at the time of DKD diagnosis and independently predicted adverse renal outcomes in patients with DKD. A CVD risk prediction model including SGHS was developed to improve the identification of high-risk patients and to facilitate individualized risk assessment. The eGFRcys/eGFRcr ratio is an easily obtainable parameter that can be incorporated into routine DKD evaluation. Future prospective studies are needed to validate these findings and to explore whether early intervention in patients with SGHS can improve cardiovascular and renal outcomes in DKD.

Data availability statement

Data are available on reasonable request.