2.1 Study design

This cross-sectional analytical study examined the discriminatory capacity of urinary podocalyxin for identifying diabetic nephropathy in adults with T2DM. This design was selected because it enables concurrent evaluation of the biomarker and the clinical outcome (albuminuria-based DN status) within a defined study population. We acknowledge at the outset that a cross-sectional design cannot establish causality or prognostic value.

2.2 Study setting

Participants were enrolled from both inpatient and outpatient services of the Department of Internal Medicine at Bangabandhu Sheikh Mujib Medical University (BSMMU), Dhaka, Bangladesh. Recruitment and data collection occurred between 25 March 2023 and 3 August 2024. All samples were collected and processed using BSMMU clinical laboratories and associated facilities. As a single-center study, the setting is reflected in the interpretation of generalizability.

2.3 Sample size and sampling

The total sample comprised 88 participants. Given the limited prior regional data on urinary podocalyxin and the exploratory nature of the study, a pragmatic recruitment target of approximately 30 participants in each major study category was adopted, consistent with previous pilot work [14]. Accordingly, we recruited 29 patients with diabetes without albuminuria, 30 patients with diabetes and albuminuria, and 29 healthy controls. The albuminuric diabetic cohort was subsequently stratified into microalbuminuria and macroalbuminuria subgroups. Participants were recruited using consecutive sampling: eligible patients attending BSMMU or affiliated primary health-care centers were approached sequentially, while healthy controls were recruited from hospital staff and relatives accompanying outpatients. Recruitment continued until the prespecified target was achieved.

2.4 Eligibility criteria

Eligible participants were adults aged ≥ 40 years. Individuals in the diabetic nephropathy group were adults with confirmed type 2 diabetes who showed signs of kidney damage according to KDIGO 2012 criteria, including raised serum creatinine, reduced eGFR, and albuminuria (uACR ≥ 30 mg/g). Individuals with T2DM but without DN were included if they were newly diagnosed, demonstrated no micro- or macroalbuminuria, and had no evidence of other renal disease. Healthy controls were individuals without diabetes, hypertension, or renal disease. Diabetes duration was recorded for all diabetic participants but was not applied as an eligibility threshold. All participants provided written informed consent prior to enrollment.

Exclusion criteria comprised the use of renin–angiotensin system (RAS) inhibitors, including angiotensin-converting enzyme inhibitors and angiotensin receptor blockers, as well as rate-limiting calcium channel blockers. Patients receiving any of these agents were not enrolled; consequently, no participant in the study was taking these medications, and no washout period was applied or required. Participants were additionally excluded if they had secondary causes of renal impairment (e.g., glomerulonephritis, urinary tract infection, pregnancy, uncontrolled hypertension, acute febrile illness, or alcoholism at the time of enrollment), or CKD attributable to non-diabetic etiologies. Common metabolic comorbidities such as hyperlipidemia and gout were not prespecified exclusion criteria; this is acknowledged as a potential source of residual confounding in the limitations.

2.5 Classification of participants

Diabetic participants were categorized according to uACR from a spot urine specimen, in accordance with KDIGO 2012 guidelines. Individuals with normoalbuminuria (uACR < 30 mg/g; n = 29) were classified as having no diabetic nephropathy, whereas those with microalbuminuria (uACR 30–300 mg/g; n = 15) and macroalbuminuria (uACR > 300 mg/g; n = 15) were grouped as having diabetic nephropathy. For analytical purposes, participants with micro- and macroalbuminuria were classified as having DN, while normoalbuminuric diabetics and healthy controls were categorized as non-DN.

2.6 Variables and data collection

We collected sociodemographic data (age, sex, religion, education, occupation) and clinical history (body mass index, duration of diabetes, blood pressure). Laboratory variables included fasting blood sugar, 2-hour post-prandial glucose, HbA1c, serum creatinine, serum urea, serum albumin, total protein, eGFR, urinary albumin, urinary creatinine, and u-PDX. All procedures were performed by trained nurses or laboratory technicians blinded to participant group.

2.7 Outcome assessment and diagnostic accuracy

Our primary outcome was the presence of DN, defined according to the KDIGO albuminuria categories. To evaluate diagnostic performance, we compared u-PDX concentrations across groups and plotted ROC curves to discriminate (a) healthy controls from diabetic patients without albuminuria, (b) diabetic patients without albuminuria from those with microalbuminuria, and (c) microalbuminuric from macroalbuminuric diabetic patients. We calculated areas under the ROC curve (AUC), sensitivity, specificity, and Youden’s J index to determine optimal cut-offs. To assess potential over-optimism, ROC analyses were internally validated using bootstrap optimism correction with 2,000 resamples, and optimism-corrected AUC estimates were calculated for each pairwise comparison.

2.8 Statistical analysis

Data were entered into a secure database and analyzed using IBM SPSS Statistics version 27 (IBM Corp., Armonk, NY, USA), with penalized regression performed in R. Continuous variables were summarized as mean ± standard deviation or median (interquartile range), as appropriate, and categorical variables as counts and percentages. Normality was assessed using the Shapiro–Wilk test. Group comparisons used one-way analysis of variance (ANOVA) or Kruskal–Wallis tests, followed by post-hoc comparisons with Bonferroni correction. Trends across ordered albuminuria categories were assessed using the Jonckheere–Terpstra test. Associations between u-PDX and continuous variables were assessed using Pearson’s correlation coefficient for variables meeting approximate linearity and normality assumptions, and Kendall’s tau-b as a rank-based, distribution-robust measure suitable for non-normally distributed and ordinal data; both are reported for transparency. The independent association between u-PDX and DN was examined using logistic regression adjusted for age, sex, duration of diabetes, and systolic blood pressure. Because the strong separation between groups produced unstable maximum-likelihood estimates (quasi-complete separation), the model was fitted using Firth’s penalized likelihood, and adjusted odds ratios (aOR) are reported per 1 ng/mL increase in u-PDX with 95% confidence intervals. Statistical significance was set at p < 0.05. Missing data were minimal (< 2%) and handled using listwise deletion.

2.9 Ethical considerations

The study was approved by the Institutional Review Board of BSMMU (BSMMU/2023/3056), which operates within Bangladesh’s national ethical framework for research involving human participants. All procedures complied with the Declaration of Helsinki. Participants provided written informed consent after receiving information about the study objectives, procedures, potential risks, and benefits; informed consent was not waived. To protect confidentiality, each participant and biological sample was assigned a coded identification number, with personal identifiers stored separately and excluded from the analysis dataset.

3.1 Participant characteristics

A total of 88 individuals were analyzed, comprising 59 participants with T2DM and 29 healthy participants. Among those with diabetes, 29 had normoalbuminuria, 15 had microalbuminuria, and 15 had macroalbuminuria, yielding 30 participants classified as having diabetic nephropathy (Table 1, Fig 1). The mean age of the cohort was 53.6 ± 10.6 years, with no significant age differences between groups (p > 0.05). The duration of type 2 diabetes was numerically higher with increasing albuminuria severity (normoalbuminuria 6.6 ± 5.3, microalbuminuria 8.3 ± 5.0, macroalbuminuria 9.3 ± 5.2 years; overall 7.7 ± 5.2 years among diabetic participants), although this trend did not reach statistical significance (p = 0.15). Males comprised 69.3% of the sample, with the highest proportion in the macroalbuminuric (80.0%) and normoalbuminuric (79.3%) groups and lower in the microalbuminuric group (46.7%); 65.5% of healthy controls were men. Education and occupation distributions are shown in Table 1, with no clear group differences.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Demographic characteristics of diabetic and healthy participants (n = 88).

https://doi.org/10.1371/journal.pone.0347975.t001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Flow diagram of study participants and classification.

Eighty-eight age- and sex-matched participants comprised 29 non-diabetic healthy controls and 59 participants with type 2 diabetes mellitus (T2DM), the latter divided into normoalbuminuria (n = 29) and albuminuria (n = 30); the albuminuric group was further stratified into microalbuminuria (n = 15) and macroalbuminuria (n = 15).

https://doi.org/10.1371/journal.pone.0347975.g001

3.2 Clinical and biochemical parameters according to albuminuria status

Table 2 summarizes clinical and biochemical parameters across groups. Body mass index, systolic blood pressure, and diastolic blood pressure did not differ significantly among groups (p > 0.05). Glycemic and renal indices showed a graded pattern across albuminuria categories. Fasting blood glucose and 2-hour post-glucose levels were markedly higher in all diabetic subgroups than in controls (p < 0.001). HbA1c followed a similar pattern (9.41 ± 1.55% in normoalbuminuria vs. 8.65 ± 2.83% in macroalbuminuria; p < 0.001). Serum creatinine and urea increased progressively with albuminuria severity: creatinine rose from 0.97 ± 0.19 mg/dL in normoalbuminuric diabetics to 1.64 ± 1.08 mg/dL in macroalbuminuric patients, and urea from 30.70 ± 12.74 to 47.70 ± 24.67 mg/dL (both p < 0.001). Serum albumin decreased from 4.47 ± 0.30 to 3.76 ± 0.60 g/dL (p < 0.001), and eGFR declined from 85.69 ± 18.40 to 62.61 ± 33.18 mL/min/1.73 m² (p < 0.001). Mean uACR increased sharply from 8.39 ± 4.28 to 1,561.01 ± 1,157.76 mg/g, and u-PDX concentrations increased from 1.16 ± 0.39 to 7.28 ± 2.05 ng/mL across the same categories (p < 0.001). Fig 2 illustrates the progressive increase in u-PDX with worsening albuminuria.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Clinical and laboratory data among T2DM subgroups by ACR and healthy subjects.

https://doi.org/10.1371/journal.pone.0347975.t002

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Urinary podocalyxin (u-PDX) concentrations across albuminuria categories.

Box plots show the median, interquartile range, and outliers for healthy controls and T2DM patients with normo-, micro-, and macroalbuminuria; values therefore differ in presentation from the mean ± SD reported in Table 2. u-PDX increased progressively with worsening albuminuria (p < 0.001).

https://doi.org/10.1371/journal.pone.0347975.g002

3.3 Comparison of diabetics with and without diabetic nephropathy

When participants with diabetes were reclassified as with or without DN, similar patterns were observed (Table 3). Serum creatinine, urea, and uACR were significantly higher in albuminuric diabetics (1.38 ± 0.89 mg/dL, 40.84 ± 20.77 mg/dL, and 825.91 ± 1,098.82 mg/g, respectively) than in non-albuminuric diabetics (0.97 ± 0.19 mg/dL, 30.70 ± 12.74 mg/dL, and 8.39 ± 4.28 mg/g). The eGFR was correspondingly lower in albuminuric patients (70.27 ± 36.41 vs. 85.69 ± 18.40 mL/min/1.73 m²; p < 0.001). The mean u-PDX concentration was nearly five-fold higher in patients with albuminuria (5.48 ± 2.89 ng/mL) than in those without (1.16 ± 0.39 ng/mL; p < 0.001).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Clinical and laboratory data by diagnosed DN status and healthy subjects.

https://doi.org/10.1371/journal.pone.0347975.t003

3.4 Correlation of u-PDX with clinical and laboratory variables

u-PDX was strongly and positively associated with multiple glycemic and renal markers (Table 4). Pearson’s coefficients showed moderate-to-strong positive correlations between u-PDX and fasting blood glucose (r = 0.49), post-glucose plasma glucose (r = 0.40), HbA1c (r = 0.42), serum creatinine (r = 0.50), serum urea (r = 0.52), and, most notably, uACR (r = 0.79), all with p < 0.001. u-PDX showed strong inverse associations with serum albumin (r = −0.62) and eGFR (r = −0.51), both p < 0.001. Age showed a borderline positive correlation (r = 0.22, p = 0.05). No significant correlations were observed with BMI or blood pressure. Kendall’s tau-b coefficients were concordant in direction and significance with the Pearson estimates (Table 4). In the Firth penalized logistic regression adjusted for age, sex, duration of diabetes, and systolic blood pressure, each 1 ng/mL increase in u-PDX was independently associated with diabetic nephropathy (aOR 22.67, 95% CI 2.47–208.09; p = 0.006); no other covariate was independently associated with the outcome (S1 Table).

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Correlation of u-PDX with clinical and laboratory variables.

https://doi.org/10.1371/journal.pone.0347975.t004

3.5 Diagnostic performance of u-PDX

Two complementary ROC analyses were performed. First, in pairwise stage comparisons, a u-PDX cutoff of 0.80 ng/mL discriminated healthy controls from diabetic patients without albuminuria with 100.0% sensitivity and 96.4% specificity; a cutoff of 1.60 ng/mL discriminated diabetic patients without albuminuria from those with microalbuminuria with 86.7% sensitivity and 93.1% specificity; and a cutoff of 5.51 ng/mL discriminated microalbuminuric from macroalbuminuric diabetic patients with 86.7% sensitivity and 93.3% specificity (Fig 3). Bootstrap internal validation demonstrated minimal optimism, with optimism-corrected AUCs remaining very close to the apparent AUCs across all pairwise comparisons (S2 Table).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Receiver operating characteristic (ROC) curves for pairwise discrimination by urinary podocalyxin: healthy controls versus diabetes without albuminuria (AUC 0.989), diabetes without albuminuria versus microalbuminuria (AUC 0.918), and microalbuminuria versus macroalbuminuria (AUC 0.893).

https://doi.org/10.1371/journal.pone.0347975.g003

Second, we evaluated overall discrimination of albuminuria-defined diabetic nephropathy from normoalbuminuric diabetes among participants with T2DM (Table 5). Because DN status in this study was defined using uACR thresholds, the ROC performance of uACR reflects incorporation of the reference-defining variable into the outcome definition and should be interpreted as an internal classification reference rather than as an independent diagnostic comparator. As expected, uACR showed perfect discrimination, with an AUC of 1.00 and 100.0% sensitivity and specificity at a cutoff of 34.06 mg/g. Against this albuminuria-defined outcome, u-PDX demonstrated excellent discriminatory performance, with an AUC of 0.96 (95% CI 0.90–1.00). The optimal u-PDX cutoff was 2.92 ng/mL, yielding 86.7% sensitivity, 100.0% specificity, 100.0% positive predictive value, 89.9% negative predictive value, and 93.2% overall accuracy. These findings indicate that u-PDX is strongly associated with albuminuria-defined diabetic nephropathy, although its diagnostic performance should be validated against longitudinal renal outcomes or independently adjudicated diabetic kidney disease in future studies.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Diagnostic performance of uACR and u-PDX for diabetic nephropathy among participants with T2DM (n = 59).

https://doi.org/10.1371/journal.pone.0347975.t005

4.1 Comparison with previous studies

Hara et al. first established a highly sensitive ELISA for urinary podocalyxin and reported that u-PDX was above the predefined threshold in 53.8%, 64.7%, and 66.7% of normoalbuminuric, microalbuminuric, and macroalbuminuric diabetic patients, respectively [11]. In their cohort, u-PDX correlated with HbA1c and tubular injury markers but not with serum creatinine or eGFR [11]. Our findings likewise demonstrate that u-PDX increases before overt albuminuria and is more closely linked to glomerular pathology than to overall renal function; the moderate inverse correlation with eGFR observed here may reflect the later stages of podocyte loss captured in our cohort. Kostovska et al. reported a stepwise increase in u-PDX with advancing DN severity and a modest correlation with uACR (r = 0.227, p = 0.002), with a threshold of 43.8 ng/mL achieving 73.3% sensitivity and 93.3% specificity [12].

The absolute u-PDX thresholds in our cohort were lower than those previously reported and appeared stage-dependent rather than defined by a single universal cut-off. This likely reflects differences between populations, including possible ethnic and genetic variation in podocyte turnover and baseline shedding, together with assay-specific calibration, variation in ELISA methodology, and differences in the operational definition of albuminuria. In our pairwise analysis, the optimal threshold separating normoalbuminuric diabetes from microalbuminuria was 1.60 ng/mL, whereas the optimal threshold for discriminating overall albuminuria-defined DN among diabetic patients was 2.92 ng/mL (Table 5). These two values correspond to two distinct analyses (stage transition versus overall DN classification) and should not be conflated; the difference highlights that a single fixed cut-off is unlikely to be appropriate across all stages and populations.

The largest study to date involving biopsy-proven diabetic kidney disease evaluated both urinary and intrarenal podocalyxin. Zeng et al. found that urinary podocalyxin correlated with corresponding mRNA expression and with the extent of proteinuria, but urinary levels alone did not predict renal outcomes; intrarenal podocalyxin expression was an independent predictor of dialysis-free survival and more accurately reflected glomerulosclerosis [15]. Our cross-sectional design precludes assessment of prognostic value; however, the positive relationship between u-PDX and albuminuria supports the interpretation that apical podocyte injury is captured by urinary podocalyxin.

In addition to podocalyxin, other podocyte-derived markers such as nephrin and podocin have been investigated. Kostovska et al. found nephrinuria in all patients with macroalbuminuria, 88% of those with microalbuminuria, and 82% of normoalbuminuric individuals, with a strong inverse correlation with eGFR (r = −0.54) and excellent diagnostic performance (ROC-AUC 0.96) [12]. Tubular and stress-related markers, including neutrophil gelatinase-associated lipocalin (NGAL), kidney injury molecule-1 (KIM-1), N-acetyl-β-D-glucosaminidase, and transferrin, also show promise for early detection [2]. Compared with these markers, podocalyxin offers the advantage of directly reflecting podocyte-specific apical injury, whereas NGAL and KIM-1 predominantly reflect tubular rather than glomerular damage and are less specific to the podocytopathy that characterizes early diabetic glomerulopathy; conversely, many of these alternatives lack specificity or require complex assays, which limits routine use, and head-to-head comparative data remain scarce. More broadly, urinary and tissue biomarkers are increasingly applied across diverse causes of kidney injury, including environmental and drug-induced nephrotoxicity, frequently using integrative computational and single-cell approaches [16,17]; these developments underscore both the promise and the methodological challenges of biomarker-based nephrology.

4.2 Biological interpretation

Podocytes form the visceral component of the glomerular filtration barrier, anchoring to the glomerular basement membrane and maintaining selective permeability through slit-diaphragm complexes and an apical glycocalyx [18]. Chronic hyperglycemia and associated metabolic stressors induce podocyte hypertrophy, foot-process effacement, detachment, and apoptosis, ultimately producing podocytopenia and glomerulosclerosis [19]. Because podocytes reside on the urinary aspect of the basement membrane, apical injury releases viable cells or membrane-derived fragments containing podocyte-specific proteins, including podocalyxin and nephrin, into the urine, potentially before albuminuria onset [9,20]. Our finding that u-PDX correlates strongly with albuminuria but only modestly with eGFR suggests that u-PDX reflects ongoing podocyte injury rather than reduced filtering capacity. The inverse association with eGFR may arise from reduced nephron number and compensatory hyperfiltration, which increases mechanical stress on remaining podocytes. These mechanisms support the biological plausibility of u-PDX as an early DN marker.

4.3 Clinical implications

Our results suggest that u-PDX can complement existing clinical markers of DN. In our cohort, u-PDX appeared most informative in the earlier stages, particularly for distinguishing healthy controls from normoalbuminuric diabetic patients and normoalbuminuric patients from those with microalbuminuria, supporting the concept that u-PDX may capture subclinical podocyte injury before albuminuria-based staging fully reflects disease progression. Routine measurement could enable detection of high-risk individuals before microalbuminuria, facilitating earlier intervention—including tighter glycemic control, optimal blood-pressure management, dietary and lifestyle measures, and timely initiation of renin–angiotensin–aldosterone system inhibitors or sodium–glucose co-transporter-2 inhibitors, which are known to slow progression of diabetic kidney disease. Identifying such individuals earlier also creates an opportunity for nutritional and “food as medicine” strategies that may support metabolic and renal health [13]. u-PDX measurement is technically straightforward using ELISA, although standardization of assay protocols and reference values across laboratories is needed for clinical adoption.

4.4 Strengths and limitations

This study is among the first to evaluate u-PDX in a South Asian population and includes internal validation of diagnostic performance. Several limitations should be considered. First, the cross-sectional design precludes causal inference and assessment of prognostic value; u-PDX cannot be claimed to predict progression of diabetic nephropathy, and longitudinal follow-up is required to determine whether elevated u-PDX predicts future eGFR decline or progression to macroalbuminuria. Second, this was a single-center study with a relatively modest overall and subgroup sample size, which reduces the precision of sensitivity, specificity, and effect-size estimates; the wide confidence interval around the adjusted odds ratio reflects this and should be regarded as exploratory. Third, our findings may not generalize to other ethnicities, as previous studies report different baseline u-PDX levels and cut-offs (11, 12). Fourth, lipid profile, smoking, and alcohol use were not systematically captured, and metabolic comorbidities were not prespecified exclusions, leaving the possibility of residual confounding. Fifth, the ROC-derived thresholds were obtained from the same dataset in which discrimination was assessed; although bootstrap optimism correction showed little attenuation of AUC estimates (S1 Table), the reported AUCs, sensitivities, and specificities should be interpreted as exploratory and may overestimate performance in external populations, so external validation in an independent cohort is required before clinical application. Also, RAS inhibitors were an exclusion criterion to remove their confounding effects on albuminuria and podocyte dynamics, which was necessary to establish baseline diagnostic performance; because RAS inhibitors are standard of care for diabetic nephropathy, this exclusion limits the immediate real-world applicability of our findings, and future studies should include treated patients or adjust for treatment. Finally, We did not directly measure other podocyte or tubular markers, which limits the assessment of u-PDX’s relative performance; head-to-head comparison is needed in future work.

Generative AI statement

The authors affirm that Paperpal was used solely for language refinement and grammatical editing in this manuscript. All suggested revisions were reviewed and approved by the authors.