External validation of a nomogram for predicting tubulointerstitial lesions in IgA nephropathy: a cross-regional study in ChinaAbstract
Background A diagnostic nomogram for predicting tubulointerstitial lesions (T1/2) in IgA Nephropathy (IgAN), developed in Guangzhou (GZ) using estimated Glomerular Filtration Rate (eGFR) and Urinary Protein Excretion (UPE), demonstrated high accuracy (AUC = 0.92) but lacked external validation.
Methods The authors externally validated the nomogram in an independent cohort from Kunming (KM, n = 387; median altitude: 1,891 m), including a high-altitude subgroup (> 2,000 m, n = 155). Model performance was evaluated using Receiver Operating Characteristic (ROC) curves, calibration plots, and Decision Curve Analysis (DCA). Both creatinine-based eGFR (eGFRcr) and Creatinine-Cystatin C-based eGFR (eGFRcr-cys), each combined with UPE, were assessed.
Results The nomogram achieved nearly identical AUCs of 0.80 in the overall KM cohort for both eGFRcr-UPE (95% CI: 0.75-0.85) and eGFRcr-cys-UPE (95% CI: 0.74-0.85), demonstrating strong generalizability across eGFR formulas. In the high-altitude subgroup, performance significantly improved (AUC = 0.89; 95% CI: 0.83-0.95) for both models.
Conclusion External validation demonstrates that the nomogram combining eGFR and UPE is feasible for non-invasive prediction of T1/2 lesions, with comparable performance between eGFRcr and eGFRcr-cys. The improved accuracy observed at high altitude suggests altitude-related influences that merit further study. Prospective studies are needed to determine its prognostic value for clinical outcomes such as CKD progression and treatment response.
Keywords
IgA nephropathy; Nomogram; External validation; Tubulointerstitial fibrosis; Altitude
Introduction
Immunoglobulin A Nephropathy (IgAN) accounts for approximately 12% of renal puncture biopsy cases in the United States, 25% in Europe, and up to 45% in China.1,2 Approximately 40% of patients with IgAN progress to End-Stage Renal Disease (ESRD).3 Oxford classification is the most widely applied IgAN pathology assessment System in clinical practice, and renal tubular atrophy or interstitial fibrotic lesions are not only important indicators of IgAN prognosis, but also guide clinical medication.4-8 The Validation in IgA (VALIGA) cohort showed no significant difference in the occurrence of renal endpoint events at 10-years after the combination of corticosteroids in IgAN patients with moderately active disease and significant tubulointerstitial lesions (T1/2) compared to the Renin-Angiotensin System Blockers (RASB) alone. In contrast, in IgAN patients without significant tubulointerstitial lesions (T0), RASB and corticosteroids significantly reduced the occurrence of renal endpoint events.9 Therefore, assessing tubulointerstitial lesions is crucial for optimizing treatment strategies in IgAN. Further research is required to refine personalized therapies based on pathological characteristics.
Renal tubulointerstitial lesions are a major tip-off for the use of corticosteroids, but renal puncture, as an invasive test, is difficult to repeat frequently. Qiongqiong Yang et al. developed a predictive model for tubulointerstitial lesions in IgAN based on estimated Glomerular Filtration Rate (eGFR) and Urinary Protein Excretion (UPE), and further constructed a diagnostic nomogram for interstitial lesions with a high predictive accuracy (AUC = 0.92, 95% CI: 0.90-0.95).10 However, the model distinguishes between the development set and the validation set by splitting the original cohort and does not perform external validation of the independent cohort. In external validation, AUC decreases of 0.1‒0.2 are common, up to 0.3 in extreme cases, especially when validating across regions or time.11 In this study, the validation cohort was based in Kunming (KM), Yunnan Province, while the development cohort of the original model was located in Guangzhou (GZ), Guangdong Province, with a straight-line distance of 1,160 km and an altitude difference of 1,880 m between the two cities. To evaluate the predictive performance and accuracy of the GZ cohort, the authors established the KM cohort by collecting data from 387 patients diagnosed with IgAN via nephropuncture and conducted external validation across regions.
Materials and methods
Study population
Cross-sectional data of patients diagnosed with IgAN by renal puncture from March 2023 to December 2024 were collected for this study. Systemic lupus erythematosus, purpura nephritis, hepatitis B virus-associated glomerulonephritis, and other secondary IgAN patients were excluded. Finally, 387 patients with IgAN were included in the study. This diagnostic/prognostic study was conducted and reported according to the STARD statement.
Clinical and laboratory data
Clinical indicators of patients were collected in this study, including gender, age, height, weight and blood pressure. Laboratory data included serum creatinine, serum albumin and UPE. Body Mass Index (BMI) was calculated as weight (kg)/height (m2). eGFR was calculated based on the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula for serum creatinine.12 All clinical and laboratory measurements were obtained within 1 week before renal biopsy.
Renal pathological evaluation
Renal tissue samples for IgAN were examined by light microscopy and immunofluorescence. The Oxford Classification scoring system was used to assess and classify histological lesions, including Mesangial cell hyperplasia (M), capillary Endothelial cell hyperplasia (E), Segmental glomerulosclerosis (S), mesangial fibrosis/Tubular atrophy (T), and Cellular/fibroblastoid Crescent (C).13 Histopathological manifestations were independently assessed by at least two pathologists, who were blinded to all clinical and laboratory data.
Statistical analyses
Continuous variables are expressed as mean ± standard deviation if they are normally distributed, and as median and interquartile range if they are not. Categorical variables were expressed as frequencies and percentages. Mann-Whitney or Chi-Square test was used to compare between-group differences for linked or categorical variables. Differences in Odds Ratios (ORs) between the two groups were assessed using the Z-test. Binary logistic regression analysis was used to assess risk factors for tubular atrophy/interstitial fibrosis. Receiver Operating Characteristic (ROC) curves and calibration curves were used to determine the predictive power and accuracy of the model. Clinically relevant operating thresholds were prespecified as 1) The Youden index-maximizing cutoff, 2) A high-sensitivity point (target sensitivity ≈90%), and 3) A high-specificity point (target specificity ≈90%). For each threshold, sensitivity, specificity, Positive Predictive Value (PPV), Negative Predictive Value (NPV), and corresponding confusion-matrix counts were calculated with 95% Confidence Intervals (95% CIs) derived from 1,000 bootstrap resamples, reported separately for the overall KM cohort and the high-altitude subgroup. R software 4.3.3 was used for statistical analysis; p < 0.05 was considered statistically significant.
Results
Comparison of clinical characteristics and tubulointerstitial injury between the KM and GZ cohorts
A total of 387 and 290 patients with biopsy-confirmed IgAN were included in the KM validation and GZ development cohorts, respectively (Table 1). Overall, patients in the KM cohort had a significantly higher proportion of males (42.89% vs. 32.07%, p < 0.05), higher body mass index (BMI; 23.42 vs. 22.42 kg/m², p = 0.03), and higher serum albumin levels (42.10 vs. 36.75 g/L, p < 0.05) compared to the GZ cohort. The KM cohort also had a greater proportion of patients with nephrotic-range proteinuria (UPE ≥ 3.5 g/d: 14.81% vs. 8.97%, p < 0.05), more frequent E1 (31.01% vs. 20.00%, p < 0.05) and S1 lesions (70.28% vs. 47.20%, p < 0.05), but a lower prevalence of M1 lesions (66.67% vs. 93.10%, p < 0.05). No significant inter-cohort differences were observed in age, hypertension, serum creatinine, estimated glomerular filtration rate (Both eGFRcr and eGFRcr-cys), T1/2 lesions, or C1/2 lesions (p > 0.05 for all). Although the difference was small, serum cystatin C was significantly higher in the KM cohort than in the GZ cohort (1.10 vs. 0.97 mg/L, p < 0.05).
Within the KM cohort, 136 patients (35.14%) exhibited T1/2 lesions, while 251 (64.86%) were classified as T0 (Supplementary Table 1). Compared with T0 patients, those with T1/2 lesions had significantly higher prevalence of hypertension (47.79% vs. 26.29%, p < 0.05), elevated serum creatinine (122 vs. 79 μmoL/L, p < 0.05), and increased serum cystatin C levels (1.51 vs. 0.99 mg/L, p < 0.05). These patients also had markedly lower eGFR, whether estimated by creatinine (54.35 vs. 97.00 mL/min/1.73 m²) or the combined creatinine-cystatin C formula (50.21 vs. 89.79 mL/min/1.73 m²), with all differences reaching statistical significance (p < 0.05). In addition, the T1/2 group showed lower serum albumin levels (40.10 vs. 42.90 g/L, p < 0.05) and a higher prevalence of nephrotic-range proteinuria (22.39% vs. 10.76%, p < 0.05). No significant differences were noted in age, sex, or BMI between the two groups (p > 0.05).
When focusing specifically on patients with T1/2 lesions, those in the KM cohort demonstrated more favorable clinical profiles than their counterparts in the GZ cohort (Table 2). KM patients with T1/2 lesions had significantly higher eGFRcr-cys (50.21 vs. 40.21 mL/min/1.73 m2, p < 0.05), higher serum albumin levels (40.10 vs. 34.60 g/L, p < 0.05), and a lower prevalence of hypertension (47.79% vs. 63.41%, p < 0.05). These findings suggest that, despite comparable pathological severity, patients in the KM cohort may present with relatively preserved renal function and better nutritional status at the time of biopsy, possibly reflecting differences in clinical presentation or timing of biopsy between the two cohorts.
Comparison of baseline characteristics between the KM and GZ cohorts stratified by oxford T lesions.
Risk factors for Oxford T lesions identified by logistic regression in the KM and GZ cohorts
Risk factors for tubular atrophy and interstitial fibrosis were assessed by multivariate logistic regression in the KM cohort. Two separate models were constructed to avoid collinearity between renal function indicators: one included eGFRcr and the other eGFRcr-cys. In both models, moderate-range proteinuria (UPE 1-3.5 g/d) and lower eGFR remained independent predictors of T-lesions (OR = 2.24 and 2.13 for UPE 1-3.5 g/d; OR = 0.97 for both eGFRcr and eGFRcr-cys; all p < 0.05). In contrast, other factors such as hypertension, serum creatinine, cystatin-C, and albumin were not statistically significant after adjustment. Notably, nephrotic-range proteinuria showed a positive association with T-lesions in the eGFRcr-cys model (OR = 1.51, p < 0.05), but not in the eGFRcr model. These findings underscore the combined importance of proteinuria and reduced kidney function in predicting tubulointerstitial injury in the KM cohort (Supplementary Table 2).
A cross-cohort comparison revealed notable differences in the strength of associations with Oxford T-lesions. In univariate analyses, both moderate-range (UPE 1-3.5 g/d) and nephrotic-range (UPE ≥ 3.5 g/d) proteinuria were more strongly associated with T1/2 lesions in the GZ cohort compared to the KM cohort (OR = 8.43 vs. 3.67 for 1-3.5 g/d, p = 0.046; OR = 26.03 vs. 4.39 for ≥ 3.5 g/d, p < 0.05). Similarly, reduced eGFR showed a stronger inverse association with tubulointerstitial damage in the GZ cohort, with lower odds ratios observed for both eGFRcr and eGFRcr-cys (0.94 vs. 0.97, p < 0.05). These inter-cohort differences were attenuated after multivariable adjustment. For moderate-range proteinuria, the between-cohort difference was no longer statistically significant in either model (Model 1: OR = 3.49 vs. 2.24, p = 0.398; Model 2: OR = 3.49 vs. 2.13, p = 0.343). The association between nephrotic-range proteinuria and T-lesions remained stronger in the GZ cohort across models (OR = 6.98 vs. 1.48 and 1.51; p = 0.088 and 0.094, respectively), though the inter-cohort difference did not reach significance. Importantly, lower eGFR remained an independent predictor of T-lesions in both cohorts in Model 1, with a statistically significant inter-cohort difference in effect size (p = 0.031), indicating a potentially stronger impact of renal dysfunction on tubulointerstitial injury in the GZ population. In Model 2, this difference approached but did not reach significance (p = 0.063), reinforcing the overall trend (Table 3).
Comparison of odds ratios for variables associated with oxford T-lesions between the KM and GZ cohorts.
In the high-altitude KM subgroup (altitude >2000 m, n = 155), stratified analysis revealed consistent trends. Among the 44 patients with T1/2 lesions (28.39%), serum creatinine (132.00 vs. 79.00 μmoL/L) and cystatin-C (1.64 vs. 1.02 mg/L) were significantly elevated, while eGFR was markedly reduced based on both creatinine (48.20 vs. 95.10 mL/min/1.73 m2) and creatinine-cystatin C equations (45.17 vs. 87.79 mL/min/1.73 m2) (all p < 0.05). Serum albumin was lower in the T1/2 group (40.40 vs. 43.10 g/L, p < 0.05), and a higher proportion had moderate-range proteinuria (56.82% vs. 29.73%, p < 0.05). Interestingly, hypertension was more prevalent in the T0 group than in the T1/2 group (74.77% vs. 52.27%, p < 0.05). No significant differences were observed in age, sex, or BMI. These findings further support the roles of reduced eGFR and moderate-range proteinuria as consistent predictors of tubulointerstitial injury, even within high-altitude settings (Table 4).
Baseline characteristics of patients in the KM cohort subgroup (> 2000 m altitude), stratified by oxford T-score.
Performance of nomogram models in predicting Oxford T lesions in the KM cohort and high-altitude subgroup
In the KM cohort, the predictive model based solely on UPE showed modest discrimination in the unadjusted analysis, with an AUC of 0.67 (95% CI: 0.61-0.72) (Supplementary Fig. 1A). Upon adding eGFR, the model’s discriminative ability improved substantially. The nomogram incorporating eGFRcr and UPE yielded an AUC of 0.80 (95% CI: 0.75-0.85), while the model using eGFRcr-cys and UPE achieved a comparable AUC of 0.80 (95% CI: 0.74-0.85) (Fig. 1A and 1C). These results remained consistent after 1,000 bootstrap resamples, with corresponding AUCs of 0.79 (95% CI: 0.74-0.84) and 0.79 (95% CI: 0.73-0.84), as shown in Supplementary Fig. 1B and 1C, respectively, confirming model robustness. Calibration plots demonstrated good agreement between predicted and observed probabilities, with Hosmer-Lemeshow test p-values of 0.595 and 0.227, respectively (Fig. 1B and 1D).
Receiver operating characteristic (ROC) and calibration curves for the nomogram based on eGFR and UPE in the KM cohort and high-altitude subgroup. (A) ROC curve of the nomogram based on eGFRcr and Urinary Protein Excretion (UPE) in the KM validation cohort. (B) Calibration curve assessing the agreement between predicted and observed probabilities using eGFRcr and UPE in the KM validation cohort. (C) ROC curve of the nomogram based on eGFRcr-cys and UPE in the KM validation cohort. (D) Calibration curve assessing the agreement between predicted and observed probabilities using eGFRcr-cys and UPE in the KM validation cohort. (E) ROC curve of the nomogram based on eGFRcr and UPE in the high-altitude KM subgroup (altitude >2000 m). (F) Calibration curve assessing the agreement between predicted and observed probabilities using eGFRcr and UPE in the high-altitude KM subgroup. (G) ROC curve of the nomogram based on eGFRcr-cys and UPE in the high-altitude KM subgroup (altitude >2000 m). (H) Calibration curve assessing the agreement between predicted and observed probabilities using eGFRcr-cys and UPE in the high-altitude KM subgroup. In all calibration plots, the dashed diagonal line represents perfect prediction. The solid orange line indicates the apparent model performance, while the blue line represents the bias-corrected curve. Calibration was evaluated using the Hosmer-Lemeshow goodness-of-fit test, with the corresponding p-values shown in each panel.
In the high-altitude subgroup (altitude > 2000 m), both nomograms demonstrated enhanced predictive performance. The models combining UPE with either eGFRcr or eGFRcr-cys each achieved an AUC of 0.89 (95% CI: 0.83-0.95), as shown in Fig. 1E and 1G. These results remained consistent after 1,000 bootstrap resamples, with identical AUCs of 0.89 (95% CI: 0.83-0.95) in Supplementary Fig. 1D and 1E, confirming model robustness. Calibration was generally accceptable, with Hosmer-Lemeshow p-values of 0.005 and 0.131, respectively (Fig. 1F and 1H). At prespecified clinical operating thresholds, consolidated diagnostic metrics, including sensitivity, specificity, PPV, NPV, and confusion-matrix counts, are provided for both models in the overall KM cohort and the high-altitude subgroup (Supplementary Table 3). These threshold-level summaries complement AUC and calibration by facilitating clinical interpretation and deployment.
Decision curve analysis further supported the clinical utility of these models (Fig. 2). In both the KM cohort and the high-altitude subgroup, models incorporating eGFR and UPE demonstrated superior net benefit across a wide range of threshold probabilities (∼10% to 75%), compared with default strategies of assuming all or no patients have tubulointerstitial lesions. The net benefit remained consistently favorable in the high-altitude subgroup, highlighting the models’ value for individualized risk prediction of tubulointerstitial injury in IgAN.
Decision curve analysis (DCA) of Nomogram Models in the KM Cohort and High-Altitude Subgroup. (A) DCA for the KM validation cohort based on eGFRcr and Urinary Protein Excretion (UPE). (B) DCA for the KM cohort using eGFRcr-cys and UPE. (C) DCA for the high-altitude subgroup of the KM cohort (altitude >2000 m) based on eGFRcr and UPE. (D) DCA for the high-altitude subgroup using eGFRcr-cys and UPE. Decision curve analysis was performed to evaluate the clinical usefulness of the prediction models by calculating the net benefit across a range of risk thresholds. The net benefit was compared between using the nomogram and alternative strategies across clinically relevant probability thresholds.
Together, these findings highlight the added predictive value of eGFR when combined with UPE, particularly in high-altitude patients, and support the use of these nomograms for clinical risk stratification across geographically distinct IgAN populations.
Discussion
This study presents the first external validation of a diagnostic nomogram for tubulointerstitial lesions (T1/2) in IgAN across geographically distinct Chinese cohorts, including explicit altitude stratification (median altitude of the KM cohort: 1,891 m; high-altitude subgroup: > 2,000 m). The nomogram, originally developed in the low-altitude GZ cohort using eGFRcr-cys combined with UPE (AUC = 0.92; 95% CI: 0.90-0.95), demonstrated robust predictive performance in the KM validation cohort (AUC = 0.80 for both eGFRcr and eGFRcr-cys combined with UPE; 95% CI: 0.75-0.85). The comparable AUCs across eGFR formulas support their potential interchangeability in clinical settings where cystatin-C may not be routinely available. Notably, model performance was further enhanced in the high-altitude KM subgroup (AUC = 0.89; 95% CI: 0.83-0.95), offsetting the typical 0.1-0.2 AUC decline often observed in cross-regional validations due to methodological or population heterogeneity.11 These findings support the nomogram’s generalizability and utility for risk stratification of tubulointerstitial injury in diverse IgAN populations.
Key factors influencing model performance
Marked OR disparities between the KM and GZ cohorts ‒ such as the stronger association between UPE and T-lesions in GZ (OR = 8.43 vs. 3.67 for UPE 1-3.5 g/d) ‒ suggest regional differences in disease severity and response to risk factors. Although multivariable adjustment reduced these differences, residual gaps remained, particularly for nephrotic-range proteinuria and eGFR.
These differences likely reflect underlying clinical and pathological variation. The KM cohort had more males, higher BMI, and more S1 lesions, while the GZ cohort showed higher M1 scores and lower eGFR and albumin at biopsy ‒ indicating more advanced disease.14 Environmental and genetic influences may also contribute; proteomic studies have identified pro-inflammatory proteins associated with IgAN progression, which may vary by region.15 Additionally, disparities in healthcare access could delay diagnosis in some areas, amplifying the observed effect sizes.16 Together, these findings highlight the need to validate predictive models across diverse populations to ensure robustness and clinical relevance.
Generalizability and practicality of creatinine-based eGFR
The authors first applied the eGFRcr formula for external validation of the nomogram, whereas the GZ development cohort used the eGFRcr-cys. Although serum creatinine is widely accessible and commonly used in clinical practice, it is influenced by non-GFR determinants such as muscle mass, age, sex, and dietary protein intake.12 These factors vary across regions and populations, potentially introducing bias into eGFR estimation. For example, the KM cohort had a higher mean BMI and a greater proportion of male patients than the GZ cohort, which may have elevated serum creatinine levels independent of true renal function, leading to misclassification of tubular injury severity. This concern is particularly relevant in early-stage CKD, where cystatin C, a biomarker less affected by muscle mass may provide a more accurate estimation of GFR and stronger correlation with histological damage.17,18 Indeed, previous studies have shown that creatinine-based eGFR tends to overestimate renal function in individuals with low muscle mass (e.g., the elderly or malnourished) and underestimate it in those with higher muscle mass or obesity.18-20
Nevertheless, in the present study, incorporating cystatin C into the eGFR calculation did not meaningfully improve model performance within the KM cohort. Both nomograms, based on eGFRcr with UPE and eGFRcr-cys with UPE, yielded identical AUC values of 0.80, suggesting equivalent predictive power. This finding supports the generalizability of the nomogram and indicates that creatinine-based eGFR may be sufficient for identifying tubulointerstitial lesions when combined with proteinuria. Moreover, the reliance on creatinine offers practical advantages, particularly in regions where cystatin C testing is not routinely available. Serum creatinine remains one of the most accessible and cost-effective renal biomarkers worldwide, especially in low-resource settings and primary care facilities lacking advanced laboratory infrastructure.21,22 The widespread use of creatinine-based equations such as CKD-EPI also ensures cross-study comparability and aligns with current clinical practice guidelines emphasizing simplicity, feasibility, and scalability.23-25 In such settings, cystatin C testing may not be economically feasible, further underscoring the relevance of creatinine-based tools. The structure of the present nomogram, based on creatinine and UPE, helps bridge this diagnostic gap and enables effective risk stratification even when advanced assays are unavailable. This aligns with the World Health Organization’s task-shifting strategy, which promotes the use of accessible technologies to improve healthcare delivery in underserved populations.26
Enhanced nomogram performance at high altitude (>2000 m)
The enhanced performance of the nomogram at high altitude (AUC = 0.89 vs. 0.80 in the overall KM cohort) likely stems from hypoxia-driven amplification of renal injury signals and attenuation of metabolic confounders. Chronic exposure to high-altitude hypoxia has been shown to impair renal function, even in healthy individuals, resulting in reduced eGFR, increased proteinuria, and altered hemodynamics independent of comorbidities.27 The kidney’s high oxygen demand and blood flow make it particularly vulnerable to hypobaric hypoxia.28 In the present study, patients with T1/2 lesions at high altitude exhibited more pronounced declines in eGFR and a higher prevalence of moderate-range proteinuria, enhancing the discriminative power of these markers. Meanwhile, traditional confounders such as hypertension were less prevalent in this subgroup, and BMI remained comparable, allowing a more direct correlation between renal dysfunction and histopathological injury.
These findings may also reflect underlying disparities in socioeconomic status and healthcare accessibility in high-altitude regions. In Yunnan province, for instance, income and medical infrastructure are relatively limited in many mountainous and rural areas. As a result, most renal biopsies in the province are performed in tertiary centers located in Kunming, the provincial capital. This referral pattern may lead to a concentration of more severe or advanced cases in the KM cohort. Such disparities are consistent with broader evidence from Low- and Lower-Middle-Income Countries (LLMICs), where chronic kidney disease is often diagnosed late, poorly managed, and associated with higher morbidity and mortality due to limited access to early intervention and kidney replacement therapy.29,30 Furthermore, in these settings, systematic CKD screening is frequently infeasible due to resource constraints, making targeted case identification and effective downstream management all the more critical.31
Decision curve analysis further confirmed superior net benefit in the high-altitude subgroup, and bootstrap validation demonstrated excellent model stability (ΔAUC ≤0.01). Together, these findings underscore the nomogram’s clinical utility in high-altitude, resource-constrained environments, where hypoxia amplifies biomarker fidelity and socioeconomic realities necessitate practical, accessible risk assessment tools.
Limitations and future directions
This study has several limitations. First, its cross-sectional design precludes evaluation of longitudinal outcomes such as eGFR decline, progression to ESRD, or treatment response.
Second, although both cohorts used the eGFRcr-cys formula, unmeasured factors such as regional variation in muscle mass, diet, or ethnicity may still influence biomarker levels and risk classification. Residual confounding also remains possible, since dietary composition and protein intake can modulate glomerular hemodynamics and proteinuria, while genetic background may interact with dietary exposures to alter kidney function trajectories, as shown in population-based studies.32 High-altitude investigations also demonstrated that both creatinine- and cystatin C-based eGFR estimates decline with increasing elevation, partly due to hematocrit changes and hypoxia-related physiology.33 Moreover, recent evidence indicates that creatinine- and cystatin C-based eGFR, though highly correlated, often yield discrepant CKD staging in clinical practice, reflecting the influence of unmeasured factors such as age, body composition, or inflammation.34
Third, important variables, including body composition, genetic or proteomic data, were not collected, limiting mechanistic interpretation, particularly regarding altitude-related effects. In addition, Socioeconomic Status (SES), a known driver of CKD progression, is not captured in the present dataset. In regions like Yunnan, where income and healthcare access vary widely, future studies should incorporate SES indicators such as household income or insurance status to better evaluate disparities in disease burden and care access. Recent biopsy data from Western China also highlight regional variation in glomerular disease patterns across age, sex, and ethnicity,35 further supporting the need for geographic model recalibration.
Finally, this study was limited to Chinese cohorts, and the performance of the nomogram in non-East Asian populations remains unknown. Prospective studies are warranted to validate the nomogram in multiethnic cohorts and to assess its prognostic value over time.
Conclusion
This external validation confirms that the nomogram combining eGFR and UPE achieves moderate discriminative performance (AUC = 0.80) for predicting tubulointerstitial lesions in geographically distinct IgAN populations, supporting its feasibility as a non-invasive risk stratification tool. The comparable performance between creatinine-based and creatinine-cystatin C-based eGFR equations highlights the model’s flexibility and practical utility, particularly in resource-limited settings where cystatin-C is not routinely available. The significantly enhanced accuracy observed in high-altitude subgroups (AUC = 0.89) further suggests altitude-related amplification of renal injury signals, reinforcing the model’s clinical relevance and prompting further investigation into altitude-modulated pathophysiological mechanisms. Future longitudinal studies are warranted to determine whether the nomogram can predict clinically meaningful outcomes, such as CKD progression or treatment response.
Declarations
ChatGPT (OpenAI) was used solely to enhance the linguistic clarity and readability of the manuscript during the writing process. No AI tools were used for data analysis, interpretation of results, or generation of scientific content. All scientific analyses and conclusions were conducted and validated by the authors.
Availability of data and materials
The datasets generated and/or analyzed during the current study are publicly available in the Zenodo repository at: https://doi.org/10.5281/zenodo.15663338.
Ethical approval
This study was approved by the Ethics Committee of The First People's Hospital of Yunnan Province (Approval no. KHLL2025-KY117). For all participants, written informed consent was obtained prior to inclusion.
-
Funding
This research was supported by the Science and Technology Plan Project of the First People's Hospital of Yunnan Province (No. KHBS-2024-014 and KHBS-2024-012), the National Natural Science Foundation of China (No. 82460150), the Yunnan Fundamental Research Projects (No. 202501AU070020), and the Ten Thousand Talent Plan for Medical Expert of Yunnan Province (No. YNWR-MY-2018-019).
Data availability statement
The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.
Acknowledgements
The authors sincerely thank the patients for providing consent to the use of their clinical data for this study.
Supplementary materials
Supplementary material 1
Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.clinsp.2026.100860.
References
- 1 Rajasekaran A, Julian BA, Rizk DV. IgA nephropathy: an interesting autoimmune kidney disease. Am J Med Sci. 2021;361(2):176-94.
- 2 Li LS, Liu ZH. Epidemiologic data of renal diseases from a single unit in China: analysis based on 13,519 renal biopsies. Kidney Int. 2004;66(3):920-3.
- 3 D'Amico G. Natural history of idiopathic IgA nephropathy and factors predictive of disease outcome. Semin Nephrol. 2004;24(3):179-96.
- 4 Chen T, Li X, Li Y, Xia E, Qin Y, Liang S, et al. Prediction and risk stratification of kidney outcomes in IgA nephropathy. Am J Kidney Dis. 2019;74(3):300-9.
- 5 Zhu X, Li H, Liu Y, You J, Qu Z, Yuan S, et al. Tubular atrophy/interstitial fibrosis scores of Oxford classification combined with proteinuria level at biopsy provides earlier risk prediction in lgA nephropathy. Sci Rep. 2017;7(1):1100.
- 6 Alamartine E, Sauron C, Laurent B, Sury A, Seffert A, Mariat C. The use of the Oxford classification of IgA nephropathy to predict renal survival. Clin J Am Soc Nephrol. 2011;6(10):2384-8.
- 7 Bellur SS, Roberts ISD, Troyanov S, Royal V, Coppo R, Cook HT, et al. Reproducibility of the Oxford classification of immunoglobulin A nephropathy, impact of biopsy scoring on treatment allocation and clinical relevance of disagreements: evidence from the VALidation of IGA study cohort. Nephrol Dial Transpl. 2019;34(10):1681-90.
- 8 Zhang W, Zhou Q, Hong L, Chen W, Yang S, Yang Q, et al. Clinical outcomes of IgA nephropathy patients with different proportions of crescents. Med (Baltim). 2017;96(11):e6190.
- 9 Cambier A, Troyanov S, Tesar V, Coppo R. Indication for corticosteroids in IgA nephropathy: validation in the European VALIGA cohort of a treatment score based on the Oxford classification. Nephrol Dial Transpl. 2022;37(6):1195-7.
- 10 Gan Y, Cai Y, Li J, Wu J, Zhang R, Han Q, et al. Development and validation of a diagnostic nomogram to evaluate tubular atrophy/interstitial fibrosis of IgA nephropathy. Int J Med Sci. 2024;21(4):674-80.
- 11 Moons KG, Kengne AP, Grobbee DE, Royston P, Vergouwe Y, Altman DG, et al. Risk prediction models: II. External validation, model updating, and impact assessment. Heart. 2012;98(9):691-8.
- 12 Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF, 3rd, Feldman HI, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150(9):604-12.
- 13 Trimarchi H, Barratt J, Cattran DC, Cook HT, Coppo R, Haas M, et al. Oxford Classification of IgA nephropathy 2016: an update from the IgA nephropathy classification working group. Kidney Int. 2017;91(5):1014-21.
- 14 Liu Y, Meng XH, Wu C, Su KJ, Liu A, Tian Q, et al. Variability in performance of genetic-enhanced DXA-BMD prediction models across diverse ethnic and geographic populations: A risk prediction study. PLoS Med. 2024;21(8):e1004451.
- 15 Paunas FTI, Finne K, Leh S, Marti HP, Berven F, Vikse BE. Proteomic signature of tubulointerstitial tissue predicts prognosis in IgAN. BMC Nephrol. 2022;23(1):118.
- 16 Okabayashi Y, Tsuboi N, Amano H, Miyazaki Y, Kawamura T, Ogura M, et al. Distribution of nephrologists and regional variation in the clinical severity of IgA nephropathy at biopsy diagnosis in Japan: a cross-sectional study. BMJ Open. 2018;8(10):e024317.
- 17 Ferguson TW, Komenda P, Tangri N. Cystatin C as a biomarker for estimating glomerular filtration rate. Curr Opin Nephrol Hypertens. 2015;24(3):295-300.
- 18 Groothof D, Post A, Polinder-Bos HA, Erler NS, Flores-Guerrero JL, Kootstra-Ros JE, et al. Muscle mass and estimates of renal function: a longitudinal cohort study. J Cachexia Sarcopenia Muscle. 2022;13(4):2031-43.
- 19 Stämmler F, Grassi M, Meeusen JW, Lieske JC, Dasari S, Dubourg L, et al. Estimating glomerular filtration rate from serum myo-inositol, valine, creatinine and cystatin C. Diagnostics. 2021;11(12):2291.
- 20 Nankivell BJ, Nankivell LFJ, Elder GJ, Gruenewald SM. How unmeasured muscle mass affects estimated GFR and diagnostic inaccuracy. EClinicalMedicine. 2020;29-30:100662.
- 21 Hill NR, Fatoba ST, Oke JL, Hirst JA, O'Callaghan CA, Lasserson DS, et al. Global prevalence of chronic kidney disease - a systematic review and meta-analysis. PLoS One. 2016;11(7):e0158765.
- 22 Myers GL, Miller WG, Coresh J, Fleming J, Greenberg N, Greene T, et al. Recommendations for improving serum creatinine measurement: a report from the laboratory working group of the national kidney disease education program. Clin Chem. 2006;52(1):5-18.
- 23 George C, Mogueo A, Okpechi I, Echouffo-Tcheugui JB, Kengne AP. Chronic kidney disease in low-income to middle-income countries: the case for increased screening. BMJ Glob Health. 2017;2(2):e000256.
- 24 Omuse G, Maina D, Sokwala A. The new creatinine-based chronic kidney disease epidemiology collaboration (CKD-EPI) 2021 equation: potential impact on screening for chronic kidney disease in an asymptomatic black African population. J Appl Lab Med. 2024;9(3):502-11.
- 25 Kidney disease: improving global outcomes (KDIGO) CKD work group. KDIGO 2024 clinical practice guideline for the evaluation and management of chronic kidney disease. Kidney Int. 2024;105(4S):S117-S314.
- 26 Mbouamba Yankam B, Adeagbo O, Amu H, Dowou RK, Nyamen BGM, Ubechu SC, et al. Task shifting and task sharing in the health sector in sub-Saharan Africa: evidence, success indicators, challenges, and opportunities. Pan Afr Med J. 2023;46:11.
- 27 Hurtado-Arestegui A, Plata-Cornejo R, Cornejo A, Mas G, Carbajal L, Sharma S, et al. Higher prevalence of unrecognized kidney disease at high altitude. J Nephrol. 2018;31(2):263-9.
- 28 Wang SY, Gao J, Zhao JH. Effects of high altitude on renal physiology and kidney diseases. Front Physiol. 2022;13:969456.
- 29 Tannor EK, Chika OU, Okpechi IG. The impact of low socioeconomic status on progression of chronic kidney disease in low- and lower middle-income countries. Semin Nephrol. 2022;42(5):151338.
- 30 Talbot B, Athavale A, Jha V, Gallagher M. Data challenges in addressing chronic kidney disease in low- and lower-middle-income countries. Kidney Int Rep. 2021;6(6):1503-12.
- 31 Tonelli M, Dickinson JA. Early Detection of CKD: implications for low-income, middle-income, and high-income countries. J Am Soc Nephrol. 2020;31(9):1931-40.
- 32 Jang MJ, Tan LJ, Park MY, Shin S, Kim JM. Identification of interactions between genetic risk scores and dietary patterns for personalized prevention of kidney dysfunction in a population-based cohort. Nutr Diabetes. 2024;14(1):62.
- 33 Pichler J, Risch L, Hefti U, Merz TM, Turk AJ, Bloch KE, et al. Glomerular filtration rate estimates decrease during high altitude expedition but increase with Lake Louise acute mountain sickness scores. Acta Physiol. 2008;192(3):443-50.
- 34 Gottlieb ER, Estiverne C, Tolan NV, Melanson SEF, Mendu ML. Estimated GFR with cystatin c and creatinine in clinical practice: a retrospective cohort study. Kidney Med. 2023;5(3):100600.
- 35 Han Q, Xu H, Li L, Lei S, Yang M. Demographic distribution analysis of different glomerular diseases in Southwest China from 2008 to 2022. Int Urol Nephrol. 2024;56(6):2011-20.
Edited by
-
Edited by
José Maria Soares Junior
