The value of sound touch elastography in the noninvasive staging of fibrosis in IgA nephropathy Authors: Rong-Rong Chen, Zhong-Kai Lan, Wen-Jing Huang, Guo-Yi Wei, Xiao-Yan Meng, Xin-Yue Ge Published: 2026 ABSTRACT AbstractObjectiveTo investigate the value of Sound Touch Elastography (STE) in fibrosis grading of IgA Nephropathy (IgAN), focusing on the diagnostic performance of the mean elasticity (STEmean) alone and combined with conventional ultrasound parameters.MethodsA total of 129 biopsy-proven IgAN patients underwent renal conventional ultrasound and STE.Based on the Oxford MEST-C T-score(reflecting the percentage of tubular atrophy/interstitial fibrosis area), patients were classified as T0 (≤ 25%), T1 (26–50%), and T2 (> 50%). Non-parametric tests were used to compare intergroup differences, correlations and diagnostic performance were assessed by Spearman and receiver operating characteristic (ROC) analyses.ResultsSTEmean correlated significantly with T-score (r = 0.61, P < 0.001), eGFR (r = − 0.71, P < 0.001), and Scr (r = 0.67, P < 0.001). STEmean differed significantly across T0, T1, and T2 groups, as well as between S0 and S1 groups (all P < 0.001). Within-session measurement reliability was excellent (ICC = 0.990, 95% CI: 0.988–0.993). The combined model integrating STEmean and renal length (RL) achieved an AUC of 0.863 (95% CI: 0.795–0.932) for discriminating T0 from T1–2 and 0.826 (95% CI: 0.751–0.901) for discriminating T2 from T0–1, outperforming STEmean alone. This model provided high specificity (92.9%) and PPV (95.1%) for ruling in fibrosis, and high sensitivity (87.0%) and NPV (88.1%) for excluding severe fibrosis.ConclusionSTEmean, particularly when combined with renal length, demonstrated good diagnostic performance for the noninvasive assessment of renal fibrosis severity in IgA nephropathy, especially for identifying significant (≥ T1) and severe (T2) fibrosis. The distinction between adjacent moderate and severe fibrosis stages remains challenging, and the proposed thresholds require external validation. These findings support the potential role of STE as a noninvasive tool to complement fibrosis evaluation in IgAN.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12880-026-02466-1.Keywords: IgA nephropathy, Sound touch elastography, Renal fibrosis, Oxford classification, Noninvasive staging Introduction IgAN constitutes the most prevalent form of primary glomerulonephritis and a leading cause of end-stage renal disease on a global scale [1, 2]. The Oxford MEST-C classification provides a standardized pathological framework for IgAN, with tubular atrophy/interstitial fibrosis (T score) recognized as the strongest predictor of renal outcomes [3]. Compared with T0, patients with T1 or T2 lesions have a 2–3 times higher risk of renal function decline [3–5]. In clinical settings, T0–T1 lesions often respond well to immunosuppressive therapy, whereas progression to T2 typically indicates irreversible damage and a significantly worsened prognosis [4]. Accurate, noninvasive differentiation between early reversible and advanced irreversible fibrosis is therefore crucial for clinical management.Renal biopsy remains the diagnostic gold standard for pathological staging [6, 7], yet its invasiveness and sampling limitations restrict repeated or dynamic assessment [3, 8]. This underscores the urgent need for safe, reproducible, non-invasive tools capable of accurately quantifying fibrosis. STE is an emerging ultrasound-based technique that quantifies tissue stiffness. Preliminary studies have shown significant correlations between renal stiffness and global renal function markers such as Scr and eGFR [9, 10]. STE provides a reliable ultrasonographic approach for assessing chronic kidney disease progression and tubulointerstitial fibrosis. It demonstrates superior diagnostic performance for CKD staging compared to IFTA grading, supporting its clinical utility in non-invasive renal assessment [11].However, the relationship between STE-derived parameters and specific pathological components of the Oxford MEST-C classification, especially the T-score, remains unclear, and robust evidence supporting its role in accurate fibrosis staging is lacking [12]. Our previous work indicated that a nomogram incorporating two-dimensional ultrasonography, STE radiomics, and clinical characteristics enables noninvasive differentiation of renal fibrosis severity [13]. Nevertheless, the value of integrating STE parameters with conventional ultrasonographic morphological indicators for assessing pathological characteristics in IgAN remains underexplored. Given the proven stability and reliability of STE in assessing hepatic and thyroid fibrosis [14–16], and the close association between ultrasonographic morphology and renal histopathology [17, 18], this study aimed to quantify the relationship between STE-derived Emean and the Oxford MEST-C pathological scores in IgAN. We further evaluated the diagnostic performance of STEmean for T staging using ROC analysis and developed a combined model incorporating conventional ultrasound indicators to explore its potential for noninvasive assessment of fibrosis in IgAN. Materials and methods Study population This single-center, prospective study enrolled patients with biopsy-proven IgA nephropathy at our institution from January 2021 to August 2025.The study was approved by the institutional ethics committee (Approval No. KY2021146), and written informed consent was obtained from all participants in accordance with the Declaration of Helsinki. For any participant under the age of 16, written informed consent was obtained from their parents or legal guardians prior to their participation in the study. The inclusion criteria were: (1) pathological confirmation of IgAN, defined by immunofluorescence showing predominant IgA deposition; and (2) complete Oxford MEST-C scores and baseline clinical data.The exclusion criteria were: (1) poor-quality renal ultrasound images or failure to meet STE quality control standards, resulting in unstable or unreliable elasticity measurements; and (2) coexistence of other primary glomerular diseases (e.g., lupus nephritis, diabetic nephropathy) or renal injury secondary to systemic diseases. Pathological evaluation All renal biopsy specimens were independently evaluated by two experienced renal pathologists who were blinded to the clinical and ultrasonographic data. Pathological grading was performed in strict accordance with the latest Oxford MEST-C classification [3]. The key assessed parameters included: mesangial hypercellularity (M), endocapillary hypercellularity (E), segmental glomerulosclerosis (S), tubular atrophy/interstitial fibrosis (T), and cellular/fibrocellular crescents (C). Discrepancies between the two observers were resolved by consensus.This two-reader independent evaluation followed by consensus adjudication is a widely accepted reference standard in renal pathology research, combining blinded independent assessment with expert adjudication to minimize individual observer bias [19]. Patients were categorized into three groups based on the T score (Fig. 1d–f): T0 (tubular atrophy/interstitial fibrosis ≤ 25%), T1 (26–50%), and T2 (> 50%). Due to the small sample size in the C2 subgroup (n = 8),data from C1 and C2 were combined into a single crescentic lesion group (C_binary1) for comparison with the non-crescentic group (C_binary0) to ensure adequate statistical power. In addition, patients with endocapillary hypercellularity (E1) or crescent formation (C1/C2) were defined as the “inflammatory-active group,” while those with E0 and C0 were classified as the “non-inflammatory group.” Representative STE and histopathological images corresponding to Oxford MEST-C T-scores T0–T2. (a, d) STE Average = 10.88 kPa, T-score = 0.(b, e) STE Average = 12.86 kPa, T-score = 1.(c, f) STE Average = 16.63 kPa, T-score = 2 Conventional ultrasound and STE examination All examinations were performed using a Mindray Resona 7s ultrasound system (Mindray Bio-Medical Electronics Co., Ltd, Shenzhen, China) equipped with STE software and an SC5-1 convex array transducer (bandwidth frequency, 1–5 MHz). Patients were examined in the prone position and instructed to hold their breath at the end of deep inspiration to ensure optimal exposure of the kidneys and to minimize respiratory motion artifacts, as breath-hold standardization is known to significantly influence tissue stiffness measurements in elastography [20]. Measurements were taken from the kidney ipsilateral to the biopsy side (the right kidney in the majority of cases). When the biopsy-side kidney could not be adequately visualized, the contralateral kidney was used and this was recorded. A conventional two-dimensional ultrasound scan was first performed to measure and record morphological parameters, including RL, renal width(RW), and cortical thickness(CTh). CTh was selected over renal parenchymal thickness because cortical thinning directly reflects nephron loss and fibrosis in CKD, whereas parenchymal thickness is confounded by medullary variability [21, 22]. Renal cortical echogenicity was not included, as it is a subjective semi-quantitative measure with limited reproducibility and low specificity for fibrosis [23]. Subsequently, STE was conducted (Fig. 1a–c). The probe was placed gently on the body surface to obtain a longitudinal section through the mid-portion of the kidney. A standardized region of interest (ROI, 7 × 9 mm) was positioned within the renal cortex at the mid-portion level, carefully avoiding the renal sinus, medullary pyramids, major vessels, and cystic lesions, consistent with the cortical sampling methodology described in recent renal elastography studies [24]. The STE mode was then activated, and once a stable image was achieved—confirmed by a Motion-Stability (M-STB) Index ≥ 4 stars (corresponding to renal parenchymal displacement ≤ 3 mm/s within the ROI)—the frame was frozen and stored. Five independent measurements were obtained during separate breath-holds following this standardized protocol. The system automatically calculated the STEmean for analysis and the interquartile range-to-median ratio (IQR/Median) for quality control. An IQR/Median < 30% was used as the internal consistency criterion [25], reflecting the dispersion of the five measurements and ensuring reproducibility. Only cases meeting this quality standard were included, and their automatically calculated STEmean values were used for subsequent statistical analysis.This acquisition protocol is consistent with the methodology described in Dag et al. [24] and follows the respiratory standardization principles established by Kaya & Gürün et al. [20]. All examinations were performed by a single sonographer with more than eight years of experience in renal ultrasound and elastography, who was blinded to the patients’ clinical and pathological results. Statistical analysis Categorical variables were expressed as counts and percentages. The normality of continuous variables was tested using the Kolmogorov–Smirnov test. Normally distributed data were presented as mean ± standard deviation(SD), and non-normally distributed data as median (interquartile range).Group comparisons were performed using one-way ANOVA, the Kruskal–Wallis H test, or the Mann–Whitney U test, as appropriate. Spearman correlation was used to assess relationships among STE parameters, conventional ultrasound variables, pathological scores, and clinical indicators. Diagnostic performance for fibrosis staging was evaluated using ROC curves and the corresponding AUC. A multivariate logistic regression model was used to develop the combined diagnostic model. To optimize performance and avoid overfitting, only conventional ultrasound parameters showing moderate or stronger correlations with the T score (|r| ≥ 0.4, P < 0.05) were included [13, 26]. For the combined logistic regression model, the predicted probabilities were calculated as P = 1 / [1 + e^−(b₀ + b₁×STEmean + b₂×RL)], and the optimal probability threshold was determined by maximizing the Youden index (sensitivity + specificity − 1). To facilitate direct clinical application without requiring probability calculation, the logistic equation was algebraically rearranged into a linear predictor (LP = b₁×STEmean + b₂×RL), and the corresponding LP decision thresholds were derived.Bonferroni correction was applied for multiple comparisons. Statistical analyses were conducted using R software (version 4.5.1), and P < 0.05 was considered statistically significant. Results Patient enrollment and measurement reliability A total of 136 patients were initially enrolled. Following quality control screening, 7 cases (5.1%) were excluded: 4 due to STE parameter instability (IQR/Median ≥ 30%), 2 with concomitant renal diseases, and 1 lacking Oxford classification data. Consequently, 129 patients were included in the final analysis (Fig. 2). All 129 cases completed five valid STE measurements and passed the quality control. The IQR/Median ratio was < 30% in all cases, with a mean value of 20.09% ± 6.36% (range: 6.40–29.90%), demonstrating excellent internal consistency. To further assess measurement reliability, within-session test-retest analysis was performed on two independent STE acquisitions obtained from all 129 patients. The intraclass correlation coefficient for absolute agreement (ICC [A,1]) was 0.990 (95% CI: 0.988–0.993, P < 0.001), confirming excellent repeatability under the standardized acquisition protocol. Prospective patient enrollment flowchart eGFR estimated glomerular filtration rate, Scr serum creatinine, UPCR urinary protein-to-creatinine ratio, BUN blood urea nitrogen; 24h_Ualp 24h urinary protein, BMI body mass index, STE sound touch elastography, STEmean, Mean elastic modulus measured by sound touch elastography; RL, renal lenght; RW, renal width; CTh, cortexThickness Baseline characteristics Among the 129 patients with IgAN, 71 were male and 58 were female, with a mean age of 46.31 ± 13.54 years (range, 14–81 years). Table 1 summarizes the clinical, ultrasonographic, and pathological characteristics of patients stratified by the Oxford MEST-C T-score. Table 1Patient’s characteristics at baselineCharacteristicTotalT0T1T2p valueNo. of patients12942(32.5%)33(25.6%)54(41.9%)-Age (years)46.31 ± 13.5441.4 ± 13.649.2 ± 13.948.4 ± 12.40.014Gender(female/male)58/71(45%/55%)21/21(50%/50%)11/22(33.3%/66.7%)26/28(48.1%/51.9%)0.293Renal length(mm)94.98 ± 11.37101.71 ± 9.4593.91 ± 9.5390.39 ± 11.42<0.001Renal width (mm)45.43 ± 6.3047.07 ± 5.9545.67 ± 5.1744.0 ± 6.930.057CortexThickness(mm)11.0(9.0–13.0)12.0(11.0–14.0)11.0(10.0–13.0)10.0(9.0–12.0)<0.001STEmean(kPa)12.00(11.09–14.28)11.00(9.88–11.46)11.99(11.19–14.12)14.20(12.88–16.14)<0.001eGFR(ml/min/1.73 m²)43.00(13.80–74.50)92.50(70.00-107.00)47.60(31.00-55.70)12.55(8.50–21.00)<0.001Scr(µmol/L)144.0(93.0-358.0)79.5(64.0-102.0)142.0(116–209)376.5(252.0-582.0)<0.001BUN(mmol/L)8.0(5.0–16.0)4.75(4.20–5.9)8.60(6.9–10.2)16.3(10.8–20.6)<0.001UPCR(mg/mmol)855.48(237.59-2405.66)486.69(108.10-1459.85)471.20(196.26-1815.07)1657.09(390.84-2916.63)0.008albumin(g/L)37.6(33.6–42.3)38.75(33.70–43.10)39.00(34.50–42.60)36.2(32.30–38.90)0.08724h_Ualp(mg/24 h)1451.0(485.0-2872.0)959.5(286.0-2019.0)1069.0(423.8–2386.0)2208.3(1049.0-3226.5)0.007BMI (kg/m²)24.03(21.90–27.60)23.85(22.00-28.60)24.70(23.07–26.40)23.8(20.7–28.4)0.541Oxford classificationMesangial hypercellularity(M0/M1)22/107(17.1%/82.9%)15/27(35.7%/64.3%)3/30(9.1%/90.9%)4/50(7.4%/92.6%)<0.001Endocapillary hypercellularity(E0/E1)110/19(85.3%/14.7%)36/6(85.7%/14.3%)28/5(84.8%/15.2%)46/8(85.2%/14.8%)0.994Segmental glomerular sclerosis(S0/S1)38/91(29.5%/70.5%)24/18(57.1%/42.9%)3/30(9.1%/90.9%)11/43(20.4%/79.6%)<0.001Crescents(C0/C1/C2)85/36/8(65.9%/27.9%/6.2%)32/10/0(76.2%/23.8%/0%)18/12/3(54.5%/36.4%/9.1%)35/14/5(64.8%/25.9%/9.3%)0.173CKD_Stage (1/2/3/4/5)24/20/29/23/33(18.6%/15.5%/22.5%/17.8%/25.6%)22/15/4/0/1(52.4%/35.7%/9.5%/0%/2.4%)1/5/19/6/2(3.0%/15.1%/57.6%/18.2%/6.1%)1/0/6/17/30(1.9%/0%/11.1%/31.5%/55.5%)<0.001STEmean, Mean elastic modulus measured by sound touch elastography; eGFR, estimated glomerular filtration rate; Scr, serum creatinine; UPCR, urinary protein-to-creatinine ratio; BUN, blood urea nitrogen; 24 h Ualp, 24 h urinary protein; BMI, body mass index; CKD_Stage,chronic kidney disease stages. Data are presented as mean±standard error, median (interquartile range), or count (percentage) Patient’s characteristics at baseline STEmean, Mean elastic modulus measured by sound touch elastography; eGFR, estimated glomerular filtration rate; Scr, serum creatinine; UPCR, urinary protein-to-creatinine ratio; BUN, blood urea nitrogen; 24 h Ualp, 24 h urinary protein; BMI, body mass index; CKD_Stage,chronic kidney disease stages. Data are presented as mean±standard error, median (interquartile range), or count (percentage) Comparison of STE parameters across MEST-C scores STEmean differed significantly among T0, T1, and T2 groups, as well as between S0 and S1 groups (all P < 0.001). STEmean increased progressively with the severity of tubular atrophy/interstitial fibrosis: T0: 11.00 (9.88–11.46) kPa, T1: 11.99 (11.19–14.12) kPa, T2: 14.20 (12.88–16.14) kPa (P < 0.001; Fig. 3a).Similarly, STEmean was higher in S1 patients [12.92 (11.20–15.03) kPa] than in S0 patients [11.17 (10.54–11.98) kPa] (P < 0.001; Fig. 3b).Partial correlation analysis showed a positive S–T correlation (r = 0.317, P < 0.001). The STEmean–S partial correlation after controlling for T was 0.096 (P = 0.283), while the STEmean–T partial correlation after controlling for S was 0.517 (P < 0.001) (Supplementary Table S1). In contrast, no significant differences in STEmean were observed across groups defined by mesangial hypercellularity (M; P = 0.067), endocapillary hypercellularity (E; P = 0.382), crescent formation (C_binary; P = 0.986), or inflammatory status (inflame; P = 0.956) (Fig. 3c–f). Violin plots showed significant differences in STEmean across T(a) and S(b) groups(***p < 0.001), but not in the M(c), E(d), C_binary(e), or inflame(f) groups Correlation of STE and conventional ultrasound parameters with pathological and clinical indicators Spearman correlation analysis revealed that STEmean was significantly correlated with the T-score (r = 0.61, P < 0.001), eGFR (r = − 0.71, P < 0.001), and Scr (r = 0.67, P < 0.001). No significant correlations were observed between STEmean and other Oxford pathological components (M, E, S, or C).Among conventional ultrasound parameters, RL showed a moderate negative correlation with the T-score (r = − 0.44, P < 0.001), while RW (r = − 0.21, P = 0.016) and CTh (r = − 0.38, P < 0.001) showed weaker associations (|r| < 0.4). Based on the predefined inclusion criterion (|r| ≥ 0.4), RL was identified as the ultrasound variable most closely associated with renal fibrosis.Therefore, RL was selected as the representative conventional ultrasound parameter to construct the combined “STEmean + RL” diagnostic model for ROC analysis, while RW and CTh were excluded due to weaker correlations. Figure 4 illustrates the Spearman correlation heatmap of key variables, highlighting only significant associations with |r| > 0.4 and P < 0.05. Correlation heatmap of key ultrasonographic parameters and clinical indicators. The color scale represents the Spearman correlation coefficient (r) value. To highlight the primary associations, only correlations with |r| > 0.4 are displayed in the heatmap; significance markers are based on Spearman correlation test: ***p < 0.001 Diagnostic performance of STEmean and combined models for assessing renal fibrosis (T classification) To evaluate the efficacy of STEmean and combined models in identifying different stages of renal fibrosis, we performed ROC analysis. The detailed diagnostic performance metrics for each model across all T-stage comparisons are presented in Table 2. Figure 5a displays the ROC curves for discriminating no/mild fibrosis (T0) from moderate-to-severe fibrosis (T1-T2). The AUC for STEmean was 0.844 (95%CI:0.771–0.916;diagnostic threshold: 12.145 kPa; sensitivity 66.7%; specificity 92.9%; PPV 95.1%; NPV 57.4%). The “STEmean + RL” combined model achieved an AUC of 0.863 (95%CI:0.795–0.932;diagnostic probability value: 0.637; derived by maximizing the Youden index; the corresponding linear predictor for clinical use is provided in the Table 2 footnote; sensitivity 81.6%; specificity 83.3%; PPV 91.0%; NPV 68.6%). Figure 5b shows the ROC curves for distinguishing severe fibrosis (T2) from non-severe fibrosis (T0-T1). STEmean yielded an AUC of 0.812 (95%CI:0.734–0.890;diagnostic threshold: 12.6 kPa; sensitivity 79.6%; specificity 78.7%; PPV 72.9%; NPV 84.3%). The “STEmean + RL” combined model yielded an AUC of 0.826 (95%CI:0.751–0.901;diagnostic probability value: 0.312; sensitivity 87.0%; specificity 69.3%; PPV 67.1%; NPV 88.1%). Table 2Diagnostic performance for differentiating renal fibrosis stagesIndicatorAUC(95CI)ThresholdSensitivitySpecificityPPVNPVP_Value T0 vs. T1-2 STEmean(kPa)0.8 44(0.771–0.916)12.14566.70%92.90%95.10%57.40%< 0.001 RL(mm)0.763(0.677–0.850)95.566.70%81.00%87.90%54.00%< 0.001 STEmean + RL0.863(0.795–0.932)0.63781.60%83.30%91.00%68.60%< 0.001 T2 vs. T0-1 STEmean(kPa)0.812(0.734–0.890)12.679.60%78.70%72.90%84.30%< 0.001 RL(mm)0.713(0.621–0.805)91.561.10%76.00%64.70%73.10%< 0.001 STEmean + RL0.826(0.751–0.901)0.31287.00%69.30%67.10%88.10%< 0.001 T0 vs. T1 STEmean(kPa)0.773(0.664–0.883)11.5866.70%78.60%71.00%75.00%< 0.001 RL(mm)0.720(0.603–0.837)95.560.60%81.00%71.40%72.30%< 0.001 STEmean + RL0.801(0.696–0.906)0.47972.70%83.30%77.40%79.50%< 0.001 T1 vs. T2 STEmean(kPa)0.717(0.607–0.826)12.679.60%60.60%76.80%64.50%< 0.001 RL(mm)0.615(0.493–0.737)89.555.60%69.70%75.00%48.90%0.037 STEmean + RL0.724(0.616–0.833)0.53879.60%60.60%76.80%64.50%< 0.001 T0 vs. T2 STEmean(kPa)0.886(0.813–0.960)12.43579.60%92.90%93.50%78.00%< 0.001 RL(mm)0.790(0.697–0.882)9570.40%81.00%82.60%68.00%< 0.001 STEmean + RL0.903(0.834–0.972)0.55383.30%90.50%91.80%80.90%< 0.001Abbreviations: AUC, area under the receiver operating characteristic curve; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value; RL, renal length; T0, tubular atrophy/interstitial fibrosis ≤ 25%; T1, 26–50%; T2, > 50%STEmean Mean elastic modulus measured by sound touch elastography, RL renal lengh, AUC area under the curve, PPV positive predictive value, NPV negative predictive value, p value: compared with a random classifier (AUC = 0.5)¹Probability thresholds for combined models correspond to the Youden index maximum from logistic regression. Regression equations and linear predictors for clinical use:T0 vs. T1–2: logit(P) = − 0.177 + 0.515 × STEmean − 0.055 × RL; LP = 0.515 × STEmean − 0.055 × RL; classify as ≥T1 if LP ≥ 0.739T2 vs. T0–1: logit(P) = − 2.492 + 0.426 × STEmean − 0.035 × RL; LP = 0.426 × STEmean − 0.035 × RL; classify as T2 if LP ≥ 1.701Complete regression coefficients are provided in Supplementary Table S3²All diagnostic thresholds are derived from a single-center prospective cohort and require external validation before clinical implementation Diagnostic performance for differentiating renal fibrosis stages Abbreviations: AUC, area under the receiver operating characteristic curve; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value; RL, renal length; T0, tubular atrophy/interstitial fibrosis ≤ 25%; T1, 26–50%; T2, > 50% STEmean Mean elastic modulus measured by sound touch elastography, RL renal lengh, AUC area under the curve, PPV positive predictive value, NPV negative predictive value, p value: compared with a random classifier (AUC = 0.5) ¹Probability thresholds for combined models correspond to the Youden index maximum from logistic regression. Regression equations and linear predictors for clinical use: T0 vs. T1–2: logit(P) = − 0.177 + 0.515 × STEmean − 0.055 × RL; LP = 0.515 × STEmean − 0.055 × RL; classify as ≥T1 if LP ≥ 0.739 T2 vs. T0–1: logit(P) = − 2.492 + 0.426 × STEmean − 0.035 × RL; LP = 0.426 × STEmean − 0.035 × RL; classify as T2 if LP ≥ 1.701 Complete regression coefficients are provided in Supplementary Table S3 ²All diagnostic thresholds are derived from a single-center prospective cohort and require external validation before clinical implementation Diagnostic performance of STEmean and combined models for differentiating renal fibrosis stages. (a) Discrimination between no/mild fibrosis (T0) and moderate-severe fibrosis (T1-2); (b) Discrimination between severe fibrosis (T2) and non-severe fibrosis (T0-1) To further assess the discriminatory ability of the models at critical junctures of disease progression (adjacent T stages), we conducted ROC analyses for adjacent-stage comparisons. The results demonstrated that both STEmean and the “STEmean + RL” combined model exhibited favorable performance in discriminating between various adjacent and spanning fibrosis stages. For distinguishing T0 from T1, the AUCs for STEmean and the combined model were 0.773(95%CI:0.664–0.883) and 0.801(95%CI:0.696–0.906), respectively. For distinguishing T1 from T2, the AUCs for STEmean and the combined model were 0.717(95%CI:0.607–0.826) and 0.724(95%CI:0.616–0.833), respectively. The diagnostic performance was optimal for distinguishing T0 from T2, with AUCs reaching 0.886(95%CI:0.813–0.960) and 0.903 (95%CI:0.834–0.972)for STEmean and the combined model, respectively. The corresponding ROC curves for the different stage comparisons are shown in Fig. 6.To formally assess whether the combined model was significantly superior to STEmean alone, DeLong tests were performed across all five T-stage comparisons. Although the combined model achieved numerically higher AUCs than STEmean alone in all comparisons, the differences did not reach statistical significance (all DeLong P > 0.05; Supplementary Table S7). Diagnostic performance of STEmean and combined models for distinguishing adjacent T-stages. (a) Discrimination between no/mild fibrosis (T0) and moderate fibrosis (T1); (b) Discrimination between moderate fibrosis (T1) and severe fibrosis (T2); (c) Discrimination between no/mild fibrosis (T0) and severe fibrosis (T2) To determine whether adding RW and CTh to the STEmean + RL model improved diagnostic performance, we compared the two nested models across all five T-stage comparisons using DeLong tests. No significant difference in AUC was observed for any comparison (all P > 0.05; Supplementary Table S2). In the multivariate regression for T0 vs. T1–2, neither CTh (β = − 0.188, OR = 0.829, 95% CI: 0.658–1.043, P = 0.110) nor RW (β = 0.073, OR = 1.075, 95% CI: 0.982–1.177, P = 0.116) was an independent predictor after adjusting for STEmean and RL. Thus, adding RW and CTh did not meaningfully improve diagnostic efficiency, supporting the more parsimonious STEmean + RL model. To evaluate the risk of overfitting, bootstrap internal validation (1,000 resamples with replacement) was performed for the STEmean + RL model across all five T-stage comparisons. The optimism-corrected AUCs and calibration slopes are summarized in Supplementary Table S4, and the calibration curves are shown in Supplementary Figs. S1–S5. The optimism values were modest across comparisons, and calibration slopes were close to 1 (range: 0.934–0.981), indicating minimal overfitting and good model calibration. To evaluate the independent contribution of STEmean beyond established clinical predictors, we performed multivariate logistic regression with a baseline clinical model (age, eGFR, and UAER) and an extended model incorporating STEmean. CKD stage and Scr were omitted to avoid collinearity with eGFR. For T0 vs. T1–2, eGFR was the dominant independent predictor (OR = 0.926, P < 0.001), while STEmean did not reach independent significance after adjustment (OR = 1.014, P = 0.941). This finding is attributable to the strong collinearity between STEmean and eGFR (Spearman r = − 0.706, P < 0.001): both parameters share substantial variance in predicting T-score, and eGFR, with the stronger individual association, dominates the regression model. Across all five T-stage comparisons, the addition of STEmean to the clinical model did not significantly improve the AUC (all DeLong P > 0.05; Supplementary Table S6). Full regression output is provided in Supplementary Table S5. Discussion This study systematically evaluated the value of STE in the non-invasive assessment of renal fibrosis in IgAN. The core findings confirmed that STEmean exhibits a significant correlation with the T score (tubular atrophy/interstitial fibrosis) of the Oxford MEST-C classification and shows significant correlations with clinical parameters reflecting renal function (eGFR, Scr). More importantly, the combined diagnostic model constructed from STEmean and conventional ultrasound parameters (RL) showed superior diagnostic performance in discriminating between no/mild fibrosis (T0) and severe fibrosis (T2). Our findings provide evidence supporting the “pathophysiological specificity” of STE in evaluating renal fibrosis. We found that STEmean values demonstrated significant differences across T and S grade subgroups, but showed no statistically significant variations among M grade, C grade and inflammatory activity groups. This distinct pattern clearly indicates that the tissue stiffness changes detected by STE are more likely attributable to “extracellular matrix deposition and tissue structure remodeling caused by fibrosis” rather than transient inflammatory cell infiltration or mesangial cell proliferation. This observation aligns with the consensus in literature that elastographic parameters primarily reflect chronic, structural fibrotic changes [27–30], while demonstrating lower sensitivity to acute inflammatory responses [28]. This specific response to fibrotic pathological alterations positions STE as a promising modality for noninvasively distinguishing reversible pathologies from irreversible damage, thereby providing clinicians with structural information that extends beyond conventional serological markers.We also observed a univariate association between STEmean and S-score. Partial correlation analysis revealed: (i) a positive S–T correlation (r = 0.317, P < 0.001); (ii) after controlling for T, the STEmean–S partial correlation decreased to 0.096 (P = 0.283), becoming non-significant; (iii) after controlling for S, the STEmean–T partial correlation remained robust (r = 0.517, P < 0.001). This indicates that the STEmean–S relationship is largely mediated by T. Pathophysiologically, segmental glomerulosclerosis often coexists with tubulointerstitial fibrosis and both are independent predictors of renal outcome in IgA nephropathy [31]. Sclerotic glomeruli drive interstitial matrix deposition and tissue stiffening via hyperfiltration and profibrotic cytokines such as transforming growth factor-β1 (TGF-β1) [32, 33], which STE detects. Thus, elevated STE in S1 patients may signal underlying significant interstitial fibrosis (higher T score), carrying direct prognostic implications and warranting closer follow-up. Another important contribution of this study lies in delineating the “clinical application scenarios and boundaries” for STE. ROC analysis demonstrated that the combined model integrating STEmean and RL consistently achieved superior diagnostic performance across various grouping paradigms, with an AUC between 0.801 and 0.903. This suggests that the combined model, incorporating both tissue stiffness (STEmean) and structural characteristics (RL), provides more comprehensive and accurate fibrosis information than any single parameter, aligning with the established trend in non-invasive liver fibrosis diagnosis that integrates disparate biological markers [34]. Compared with prior elastography studies, Ruan et al. (STE, 60 proteinuria patients) reported AUCs of 0.718–0.744 for nephropathy detection but did not provide Oxford T-score staging or a combined model [9]. Ge et al. (radiomics nomogram, 150 CKD patients) achieved a validation AUC of 0.85 for IFTA grading, yet the model relies on complex post-processing [13]. Our study is the first to target biopsy-proven IgA nephropathy using the Oxford MEST-C T-score. The STEmean + renal length model attained an AUC of 0.863 with only two routine parameters, prospectively collected under predefined quality control (IQR/Median < 30%). This disease-specific, clinically simple model complements prior work and provides a readily translatable fibrosis staging tool. In the multivariate analysis adjusted for age, eGFR, and proteinuria, eGFR emerged as the dominant independent predictor of T‑score (P < 0.001 for all comparisons), while STEmean did not retain independent significance. This finding is attributable to the strong collinearity between STEmean and eGFR (Spearman r = − 0.706, P < 0.001): both parameters share substantial variance in predicting fibrosis severity, and eGFR, with the stronger univariate association with the T‑score, dominates the regression model. Importantly, this statistical observation does not diminish the clinical value of STE. eGFR reflects global glomerular filtration—a functional parameter influenced by hemodynamics, acute injury, and compensatory hypertrophy [35]—whereas STEmean directly quantifies tissue stiffness resulting from extracellular matrix deposition and fibrotic remodeling [33]. The strong STEmean–eGFR correlation itself is clinically meaningful: it demonstrates that tissue stiffness tracks closely with renal functional decline, consistent with prior reports [13, 27]. The unique role of STE lies in scenarios where eGFR may be insufficient or unreliable, such as early fibrosis with preserved renal function [35] or when structural damage assessment independent of hemodynamic fluctuations is needed [27]. Future studies evaluating STE in patient subgroups with preserved eGFR are warranted to better isolate its independent diagnostic contribution beyond functional biomarkers. Regarding clinical translation, the diagnostic thresholds established in this study correspond to distinct decision-making scenarios: First is the “exclude” and “confirm” scenario: The conventional group comparison (T0 vs. T1-2) established a clinically valuable rule-in threshold. A STEmean value greater than 12.145 kPa identified the presence of any fibrosis (≥ T1) with high specificity (92.9%) and positive predictive value (95.1%), thus enabling a definitive diagnosis when the threshold is exceeded.Conversely, the “STEmean + RL” combined model exhibited high sensitivity (87.0%) and NPV(88.1%) for detecting severe fibrosis (T2), making it an effective “exclude” tool for excluding the risk of severe fibrosis. Second, Disease Progression Monitoring Scenarios: Adjacent-group analysis provided refined tools for disease monitoring. The STEmean threshold of 11.58 kPa serves as a potential cut-off for distinguishing no/mild from moderate fibrosis (T0 vs. T1). More importantly, the 12.6 kPa STEmean threshold demonstrated optimal diagnostic performance in two distinct group comparisons (severe fibrosis T2 vs. non-severe T0-T1; moderate T1 vs. severe T2). This finding suggests that a STEmean value ≥ 12.6 kPa may indicate a higher likelihood of severe fibrosis (T2), warranting closer monitoring and further evaluation. However, this threshold is derived from a single-center cohort with a modest sample size and should not be regarded as a standalone clinical decision point until externally validated in multicenter prospective studies. During clinical follow-up, progression of STEmean values from below to above this 12.6 kPa threshold may signal potential advancement to severe fibrosis, prompting consideration of further diagnostic workup rather than immediate treatment modification. However, this study also revealed the current technological limitations. Diagnostic performance in adjacent-group analyses, especially the distinction between T1 and T2, was suboptimal (AUC = 0.615–0.724), and the sample size is relatively limited. Several factors may contribute to this phenomenon: the frequent focal, patchy distribution of moderate fibrosis increases measurement variability due to tissue heterogeneity; a potential “threshold effect” in fibrotic progression might mean that changes in tissue stiffness remain insufficient for reliable detection by current technology until a certain cumulative degree is reached. This finding is consistent with other research reports in the field of chronic kidney disease elastography [13, 36], suggesting that reliable non-invasive diagnosis of moderate fibrosis requires future exploration, potentially through integration with serum biomarkers or more sophisticated radiomics models. Despite these limitations, our findings remain highly insightful. Compared to renal biopsy, which is invasive, prone to sampling error, and unsuitable for dynamic monitoring [3, 8], STE provides a safe and reproducible quantitative assessment method. In conclusion, the combined model integrating STEmean and renal length demonstrated good diagnostic performance for identifying significant fibrosis (T0 vs. T1–2, AUC = 0.863) and severe fibrosis (T2 vs. T0–1, AUC = 0.826) in patients with IgA nephropathy. However, the distinction between adjacent moderate and severe fibrosis (T1 vs. T2) remains challenging (AUC = 0.724), reflecting the inherent difficulty in capturing early fibrotic transition with current elastography technology. These findings support the potential role of STE as a non-invasive screening tool for ruling in or ruling out significant fibrosis, while highlighting the need for multi-parameter strategies to improve moderate-stage discrimination. External validation in larger multicenter cohorts is required before clinical implementation. This study confirms that integrating functional information reflecting tissue stiffness (STEmean) with morphological information reflecting organ atrophy (RL) yields diagnostic efficacy superior to either parameter alone. This lays a solid foundation for developing multimodal ultrasound-based non-invasive follow-up protocols for IgA nephropathy, pending external validation of the proposed thresholds. This study has several limitations. First, as a single-center prospective study, the limited sample size may restrict generalizability and underpower certain subgroup analyses (e.g., by E or C grade). Second, while within-session intra-observer reliability was excellent (ICC = 0.990, 95% CI: 0.988–0.993), all STE measurements were completed by a single operator on one device; inter-observer and inter-device reliability remain to be established. Although bootstrap internal validation indicated minimal overfitting, external validation in independent multicenter cohorts is required before the proposed diagnostic thresholds can be adopted clinically.Future multi-center studies incorporating multiple operators and device types are needed to confirm the generalizability of the proposed diagnostic thresholds. Third, although RW and CTh showed weak correlations with fibrosis (|r| < 0.4), we prioritized RL for model inclusion due to its stronger biological rationale and association with fibrotic progression, aiming to enhance model robustness and interpretability. Future studies should explore automated feature selection or nonlinear modeling in larger cohorts. Fourth, although two experienced pathologists independently and blindly scored all biopsies using the Oxford MEST-C classification, the pre-consensus scores were not archived. Consequently, inter-observer Kappa statistics could not be calculated. While the dual-reader consensus approach minimizes individual bias, the lack of quantified agreement remains a limitation. Future studies should prospectively archive independent scores to enable Kappa reporting.Moreover, the cross-sectional design cannot evaluate the prognostic utility of STE parameters, warranting longitudinal validation. Finally, the suboptimal performance in distinguishing T1 fibrosis underscores the need for multi-parameter strategies to improve diagnostic accuracy. Conclusion This prospective study demonstrates that STEmean, alone and in combination with renal length, provides good diagnostic performance for identifying significant (T1–2) and severe (T2) fibrosis in IgA nephropathy, while the distinction between T1 and T2 remains limited. The proposed thresholds may serve as reference values for non-invasive fibrosis assessment, pending external validation in larger multicenter cohorts. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Acknowledgements Not applicable. Abbreviations Sound Touch Elastography IgA nephropathy Mean elastic modulus measured by sound touch elastography Receiver operating characteristic Area under the curve estimated Glomerular filtration rate Serum creatinine Urinary protein-to-creatinine ratio Blood urea nitrogen 24 h Urinary protein Body mass index Chronic kidney disease stages Renal length Renal width Cortical thickness Standardized region of interest Interquartile range-to-median Motion-Stability Standard deviation Author contributions Rong-Rong Chen: Conceptualization, Data curation, Formal analysis, Methodology, Supervision, Validation, Writing-original draft. Zhong-Kai Lan: Conceptualization, Supervision, Visualization, Writing-review & editing. Wen-Jing Huang: Conceptualization, Formal analysis, Investigation. Guo-Yi Wei: Conceptualization, Formal analysis, Investigation. Xiao-Yan Meng: Formal analysis, Investigation, Data curation, Methodology. Xin-Yue Ge: Formal analysis, Investigation, Methodology, Project administration. Funding This work was supported by the Project of Liuzhou Scientific Research and Technology Development Plan [2024YB0101A013]; and Guangxi Young Elite Scientist Sponsorship Program [GXYESS2025195]. Data availability The datasets generated and analysed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The study was approved by the Institutional Ethics Committee of the Fourth Affiliated Hospital of Guangxi Medical University(Approval No. KY2021146). and written informed consent was obtained from all participants in accordance with the Declaration of Helsinki. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Rong-Rong Chen and Zhong-Kai Lan contributed equally to this work. Xiao-Yan Meng and Xin-Yue Ge also contributed equally to this work. Contributor Information Xiao-Yan Meng, Email: mengxiaoyanlzgryy@163.com. Xin-Yue Ge, Email: Galaxyge1991@163.com. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Data Availability Statement The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.