Integrating urinary dual-gene methylation and VI-RADS score to predict residual tumors after TURBT in NMIBC: a nomogram-based model

Abstract

Background

Bladder cancer (BC) is the ninth most prevalent cancer globally, with approximately 75% of cases classified as non-muscle-invasive bladder cancer (NMIBC). Although transurethral resection of bladder tumor (TURBT) is the standard treatment for NMIBC, there remains a considerable risk of postoperative residual tumors. Accurately identifying patients who require repeat transurethral resection of bladder tumor (Re-TURBT) remains a significant clinical challenge.

Magnetic resonance imaging (MRI) exhibits high accuracy in detecting residual tumors after TURBT, aiding clinical decision-making. However, its relatively low specificity limits its clinical applicability. Urinary methylation markers, particularly those associated with H4 cluster protein 6 (H4C6) and Twist Family BHLH Transcription Factor 1 (TWIST1), have demonstrated high sensitivity and specificity in detecting bladder cancer, showing promising clinical performance. The combined detection of H4C6 and TWIST1 gene methylation provides a more reliable method for identifying residual tumors post-TURBT. Integrating the VI-RADS score with urine gene methylation analysis could complement each other, improving the accuracy of residual tumor detection and enhancing clinical applicability.

Objective

This study aimed to evaluate the predictive value of clinical/pathological factors, urinary H4C6 and TWIST1 gene methylation, and their combined performance in identifying postoperative residual tumors in NMIBC patients.

Methods

Morning urine samples from NMIBC patients at Zhongshan City People’s Hospital (June 2022–August 2024) were analyzed for H4C6 and TWIST1 methylation levels. Patients underwent MRI and Re-TURBT. Logistic regression was used to identify predictors of residual tumors, and a nomogram was developed. Model performance was evaluated using Receiver Operating Characteristic (ROC) curves and decision curve analysis (DCA).

Results

Among the 55 patients included in the study, the area under the ROC curve (AUC) for urinary H4C6 methylation was 0.698, whereas TWIST1 methylation had an AUC of 0.758. Combined dual-gene methylation yielded an AUC of 0.764. The AUC for VI-RADS scoring was 0.888. When dual-gene methylation and hematuria were combined with VI-RADS scores, the prediction model achieved an AUC of 0.972. DCA demonstrated that the prediction model offers substantial clinical benefit across a broad range of threshold probabilities. The combined prediction model of dual-gene methylation, hematuria, and VI-RADS scores offers a powerful non-invasive method to assist clinicians in identifying high-risk patients with residual tumors after TURBT, thereby reducing unnecessary Re-TURBT and improving patient management.

Conclusion

Urinary H4C6 and TWIST1 gene methylation, along with VI-RADS scoring, show high diagnostic performance in predicting residual tumors following TURBT in NMIBC patients. A combined prediction model incorporating dual-gene methylation, hematuria, and VI-RADS scores achieves superior discrimination, excellent calibration, and strong clinical utility.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12894-025-01857-w.

Introduction

Bladder cancer is the ninth most common cancer worldwide, with a higher incidence in males than in females. Globally, approximately 614,000 new cases and 220,000 deaths are reported annually [1]. Although the mortality rate for bladder cancer is relatively low, its high recurrence rate contributes significantly to the global disease burden [2, 3].

Diagnosis of BC primarily relies on TURBT. Approximately 75% of cases are classified as NMIBC, with the standard treatment consisting of TURBT followed by adjuvant therapy. Despite ongoing advancements in cystoscopy and TURBT techniques, residual tumors may still occur after the initial resection, necessitating Re-TURBT in some patients [4]. Furthermore, up to 42% of NMIBC patients experience tumor recurrence following TURBT [5], and high-risk patients require regular cystoscopic surveillance for at least five years.

Cystoscopy demonstrates high specificity (64%–100%) and sensitivity (87%–100%) for detecting BC; however, it is an invasive procedure, costly, and associated with surgical and anesthetic risks for patients [6, 7]. Compared to traditional white-light cystoscopy, flexible fluorescence cystoscopy significantly improves the detection rate of bladder cancer and shows higher sensitivity for identifying carcinoma in situ (CIS) [8, 9]. Nevertheless, flexible cystoscopy has not fully replaced rigid cystoscopy due to the latter’s greater cost-effectiveness, ease of operation, and widespread availability [10]. Urine cytology, while exhibiting specificity comparable to cystoscopy, is limited by low sensitivity (29.1%), particularly for detecting low-grade tumors [11, 12].

MRI plays a crucial role in diagnosing and staging bladder cancer. With advancements in multiparametric MRI and improved imaging techniques, its diagnostic accuracy has significantly increased. MRI demonstrates high sensitivity and specificity in evaluating residual lesions following TURBT [13, 14]. In clinical practice, the Vesical Imaging-Reporting and Data System (VI-RADS) [15] integrates multiparametric data, including T2-weighted imaging, diffusion-weighted imaging (DWI), and dynamic contrast-enhanced imaging (DCE), providing a reliable imaging foundation for assessing tumor residuals prior to Re-TURBT [16]. However, MRI has limitations in detecting small lesions and is affected by postoperative inflammation or bladder wall changes from intravesical therapy, potentially leading to false positives or negatives, thus reducing diagnostic accuracy.

Genetic testing has been widely used in the auxiliary diagnosis of various cancers. As a non-invasive source of proteins, DNA, and RNA, urine plays a vital role in bladder cancer detection and analysis [17, 18] and has become a focal point of research [19, 20]. Urinary molecular biomarkers provide crucial support for predicting residual tumors prior to Re-TURBT [21].

Currently identified biomarkers, such as NID2 and GDF15 methylation, nuclear matrix protein 22 (NMP22), and bladder tumor antigen (BTA), have been shown to possess good diagnostic efficacy in bladder cancer research. However, their relatively low methylation rates or insufficient specificity in bladder cancer limit their broader application [22–24]. Additionally, bladder cancer is tumor-specific, and a single tumor marker cannot fully capture the biological characteristics of the tumor [25, 26], making combined diagnostic techniques advantageous for improving diagnostic efficacy. Histones are core components of chromatin, and histone H4C6 plays a crucial role in histone synthesis. It has been explored as a potential biomarker for the early diagnosis of bladder cancer [27]. TWIST1, a transcription factor with a bHLH domain, is upregulated in tumors and can induce epithelial-mesenchymal transition (EMT), playing a pivotal role in tumor cell migration and invasion [28]. The detection of TWIST1 methylation in urine is a non-invasive method with potential for early bladder cancer diagnosis [29, 30].

However, studies exploring urinary DNA methylation for evaluating residual tumors before Re-TURBT remain limited, and its feasibility and clinical utility have yet to be fully validated. Currently, diagnostic companies have developed and clinically tested kits for detecting H4C6 and TWIST1 methylation in urine. These kits are user-friendly and have completed GCP clinical trials, but robust evidence supporting their clinical efficacy remains insufficient. We selected H4C6 and TWIST1 as the focus of our study due to their higher methylation rates in bladder cancer, which may provide improved sensitivity for detecting residual microlesions. Furthermore, urinary DNA methylation tests for both markers have been preliminarily validated for clinical feasibility. Combining urinary H4C6 and TWIST1 methylation with the VI-RADS score shows promise in enhancing diagnostic accuracy for residual tumors after TURBT and informing treatment decisions.

Materials and methods

General information

This study was not registered as a clinical trial due to its small sample size and its single-center, single-arm design. However, it adhered to Good Clinical Practice (GCP) guidelines and received ethical approval from the Clinical Research and Experimental Animal Ethics Committee of Zhongshan City People’s Hospital (Ethics Approval Number: 2024–128). Written informed consent was obtained from all participants. The data for all patients were collected at Zhongshan City People’s Hospital (June 2022–August 2024). The study flowchart is shown in Fig. 1. All patients underwent 3.0T bladder MRI and provided fresh early-morning urine samples prior to Re-TURBT. The methylation levels of H4C6 and TWIST1 in urine samples were measured, and both clinical and pathological data were collected. All patients underwent routine intravesical instillation therapy after TURBT and were followed up within three months, including Re-TURBT evaluation.

Fig. 1.

Research flowchart

Urinary H4C6 and TWIST1 methylation detection

Morning urine samples were collected after at least six hours of bladder retention to ensure sufficient human cell content. A 30–50 mL aliquot of morning urine was transferred to a 50 mL storage tube containing 1 mL of a specialized preservation solution (Guangdong Bright-Innovation BioMed Co., Ltd., Shunde, China). The preservation solution’s protective effect ensures the stability of cfDNA, thus omitting the urine pre-processing step. This solution prevents DNA degradation in urine for up to one month at room temperature. The H4C6 and TWIST1 gene methylation detection kits, developed and manufactured by Guangdong Bright-Innovation BioMed, were used for analysis. According to the kit protocol, 3.5 mL of preserved urine was used for nucleic acid extraction. Due to the relatively small urine volume, DNA extraction and bisulfite modification were performed on the automated AutoPure-12 instrument (Guangdong Bright-Innovation BioMed Co., Ltd.).

The entire procedure, including DNA extraction, bisulfite treatment, and purification/recovery of bisDNA, was completed automatically within 3 h, yielding 100 µL of bisDNA solution. The methylation fluorescence quantitative kit for H4C6 and TWIST1 was a three-channel assay, including each target gene along with the internal control ACTB gene. The ACTB gene served as the reference gene, and samples with a cycle threshold (Ct) value for ACTB above 35 were considered “low quality” and excluded. Ct values for H4C6 and TWIST1 below 40 were deemed positive.

Sample size estimation

Based on previous literature [31], the AUC for urinary DNA methylation in predicting tumor residuals is approximately 0.87. Using PASS software and the Confidence Intervals for the Area Under an ROC Curve function, we set the following parameters: an expected ROC AUC of 0.8, a confidence interval width not exceeding 0.25, and a significance level (α) of 0.05. The calculation determined that a minimum of 50 patients undergoing Re-TURBT would be required.

Multiparametric MRI and VI-RADS scoring

Bladder MRI examinations were performed using the SIGNA™ Premier 3.0T MRI scanner (GE Healthcare, USA). All patients underwent multiparametric MRI within 2 weeks prior to Re-TURBT. Patients were instructed to drink 500–1000 mL of water 30 min before the examination to ensure adequate bladder filling. For the dynamic contrast-enhanced (DCE) sequence, a low-molecular-weight gadolinium-based contrast agent was administered at a dosage of 0.1 mmol/kg body weight, injected at a rate of 2 mL/s over 10 s.

VI-RADS scoring was performed according to the 2018 Panebianco et al. criteria [15]. MRI images were independently reviewed and scored by two senior radiologists specializing in MRI imaging, who were blinded to the patients’ clinical and pathological information. In the event of a discrepancy between the two radiologists, the final VI-RADS score was determined through discussion and consensus.

Inclusion and exclusion criteria

The inclusion criteria were: patients with NMIBC confirmed by pathology following initial TURBT, who underwent Re-TURBT within 1 to 3 months, including cases with T1 stage, high-grade tumors, or absence of bladder muscle in the pathological findings. The exclusion criteria were: patients with other tumors or severe inflammatory infections, allergy to contrast agents, or poor MRI image quality that affected interpretation.

Follow-up and Re-TURBT

All patients underwent regular follow-up after TURBT, including biweekly telephone calls to assess hematuria and urinary irritation symptoms. High-risk patients underwent Re-TURBT within 4–6 weeks postoperatively, while intermediate- and low-risk patients underwent Re-TURBT within 1–3 months. All procedures were performed by the same senior urologist. Visible tumors were deeply resected into the muscular layer. For patients without visible tumors, tissue beneath the scar at the initial resection site was resected, extending 1 cm beyond the original margins. All patients received immediate postoperative intravesical instillations of either 50 mg mitomycin or 50 mg epirubicin.

Statistical analysis

Statistical analyses were performed using IBM SPSS software, and ROC curves were generated with GraphPad Prism. Categorical data were compared using the Chi-square test or the corrected Chi-square test when the expected frequency was < 5. VI-RADS scores were expressed as the mode, and group differences were compared using the Mann–Whitney U test. The inter-observer consistency of the VI-RADS scoring interpretation results was assessed using the kappa statistic. McNemar’s test was used to evaluate the consistency between urinary DNA methylation results and pathological findings. Variables with P < 0.05 in univariate analysis (performed using R 4.4.1) were included in multivariate logistic regression with backward elimination to construct the nomogram prediction model. Model performance was assessed using ROC curves and DCA. The ROC differences between the models were compared using the DeLong test.

Results

Baseline characteristics and comparison between tumor residual and non-residual groups

A total of 86 NMIBC patients were initially enrolled in the study, with 55 patients included in the final analysis after screening. The inter-observer consistency of the VI-RADS scoring had a kappa value of 0.827, P < 0.001. Baseline comparisons revealed that in patients with tumor residuals, the positive rate of H4C6 methylation was 52.17% (12/23), and the positive rate of TWIST1 methylation was 60.87% (14/23). Both rates were significantly higher than those in patients without tumor residuals (χ² = 10.211, P = 0.001; χ² = 16.616, P < 0.001). When either H4C6 or TWIST1 methylation positivity in urine was defined as a combined diagnostic positive result, the positive rate in the tumor residual group was 65.22% (15/23), compared to 12.50% (4/32, with all cases pathologically confirmed as bladder inflammation) in the non-residual group. The difference was statistically significant (χ² = 16.447, P < 0.001), as shown in Table 1.

Table 1.

Baseline characteristics and between-group comparison of 55 patients

Indicator		Tumor (n = 23)	Normal (n = 32)	Statistical value	P value
Age/year	≥ 65	11	19	0.720	0.396
	< 65	12	13
T staging of initial TURBT	T1	21	26	0.430	0.512
	Ta	2	6
Histological grading of initial TURBT	HG	18	19	2.168	0.141
	LG	5	13
Location of initial TURBT	sidewall	13	19	0.045	0.832
	other	10	13
Number of tumors undergoing initial TURBT	single	10	17	0.498	0.480
	multiple	13	15
Characteristics of initial TURBT	stemmed	5	10	0.610	0.435
	broad-based	18	22
H4C6 methylation before Re-TURBT	positive	12	4	10.211	0.001
	negatives	11	28
TWIST1 methylation before Re-TURBT	positive	14	3	16.616	< 0.001
	negatives	9	29
H4C6-TWIST1	positive	15	4	16.447	< 0.001
	negatives	8	28
Hematuria before Re-TURBT	yes	15	7	10.474	0.001
	no	8	25
VI-RADS score before Re-TURBT		1	4	67.000	< 0.001

HG represents High Grade, and LG represents Low Grade

Consistency between urinary H4C6, TWIST1 methylation and pathological results

Using Re-TURBT pathology as the gold standard, McNemar’s test was applied to assess the consistency between urinary H4C6 and TWIST1 methylation, their combination, and cystoscopy for detecting bladder cancer. The results showed no statistically significant difference in the positive detection rates between the urinary methylation markers (H4C6, TWIST1, and their combination) and cystoscopy (Supplementary Table 1).

Diagnostic performance of urinary H4C6, TWIST1 methylation, and VI-RADS scoring for detecting tumors before Re-TURBT in NMIBC patients

The analysis of urinary H4C6 and TWIST1 methylation demonstrated moderate diagnostic performance, with AUCs of 0.698 and 0.758, respectively. Combining H4C6 and TWIST1 methylation improved sensitivity to 65.22% and achieved an AUC of 0.764. Additionally, the VI-RADS scoring system, when dichotomized at the established cutoff of 2 (score 1 vs ≥ 2), demonstrated excellent diagnostic performance for residual tumor detection with 86.96% sensitivity and 90.63% specificity, yielding an AUC of 0.888. Detailed results are presented in Table 2 and Fig. 2A.

Table 2.

Analysis of the diagnostic performance of urinary H4C6 and TWIST1 methylation for detecting residual tumor in NMIBC

	Sensitivity, % (95% CI)	Specificity, % (95% CI)	AUC(95%CI)
H4C6 Methylation	52.17(32.96, 70.76)	87.50(71.93, 95.03)	0.698(0.551, 0.845)
TWIST1 Methylation	60.87(40.79, 77.84)	90.63(75.78, 96.76)	0.758(0.619, 0.896)
H4C6-TWIST1	65.22(44.89, 81.19)	87.50(71.93, 95.03)	0.764(0.628, 0.900)
VI-RADS score	86.96(67.87, 95.46)	90.63(75.78, 96.76)	0.888(0.788, 0.988)

Fig. 2.

Diagnostic performance of individual indicators and the combined predictive model for detecting residual tumors before Re-TURBT. A ROC curves showing the diagnostic performance of individual indicators: urinary H4C6 and TWIST1 methylation combined (AUC = 0.764, P < 0.001), VI-RADS (AUC = 0.888, P < 0.001), and the combined predictive model (AUC = 0.972, P < 0.001). At the optimal cut-off value of 0.637, the combined model achieved a sensitivity of 82.61% and a specificity of 96.88%. B Nomogram of the predictive model incorporating dual-gene methylation, hematuria, and VI-RADS score. Each factor is assigned points based on its contribution, and the total points correspond to the probability of residual tumor diagnosis. For example, when a patient with positive urinary dual-gene methylation and no hematuria has a VI-RADS score of 3, the total score is approximately 67.5 + 0 + 100 = 167.5, which corresponds to a residual tumor probability of over 90%

Nomogram prediction model for tumor residuals before Re-TURBT in NMIBC patients

A nomogram prediction model was developed using logistic regression analysis, incorporating combined urinary H4C6 and TWIST1 methylation markers, pre-Re-TURBT hematuria, and VI-RADS scores. The detailed logistic regression results are provided in Supplementary Table 2. The model achieved an AUC of 0.972 (95% CI: 0.937–1.000). At a cut-off value of 0.637, the sensitivity and specificity were 82.61% and 96.88%, respectively, as shown in Fig. 2A. A nomogram was constructed to visualize the contribution of each factor, as illustrated in Fig. 2B. The DeLong test demonstrated statistically significant differences in AUROC values between the developed predictive model and individual variables (P < 0.05).

Model calibration was assessed using the calibration curve demonstrated a Brier score of 0.065 (Fig. 3), indicating good agreement between predicted and observed outcomes. Additionally, decision curve analysis confirmed that the model provided net clinical benefit across a wide range of threshold probabilities, as illustrated in Fig. 4.

Fig. 3.

Calibration curve for the predictive model. The x-axis represents the predicted risk, and the y-axis shows the observed frequency. The diagonal gray line indicates perfect prediction, while the red line represents the actual performance of the model. The closer the red line aligns with the diagonal, the better the calibration of the model. The AUC is 97.2 (95% CI: 93.8–100.0), and the Brier score is 0.065, indicating excellent model performance. Bootstrap resampling with 1000 iterations (Bootstrap = 1000) was applied to enhance result stability, and the data were divided into 10 groups (M = 10) for calibration analysis

Fig. 4.

Decision curve analysis for the model. The x-axis represents the high-risk threshold, and the y-axis shows the net benefit. The red line indicates the net benefit of the predictive model, while the gray and black lines represent the strategies of treating all patients (“All”) and treating none (“None”), respectively. The model demonstrates a higher net benefit across a wide range of threshold probabilities, supporting its clinical utility for risk stratification and decision-making

Discussion

Despite advancements in TURBT techniques, NMIBC patients continue to exhibit a high rate of tumor residuals following the initial resection [4, 32]. Re-TURBT can remove residual tumors, clarify staging, and improve prognosis for NMIBC patients [33–35]. However, as an invasive procedure, Re-TURBT carries risks related to anesthesia and surgery, along with additional financial burdens. Potential complications include urinary tract infections, bleeding, and urethral strictures. Therefore, there is an urgent need for a non-invasive, cost-effective, and widely applicable method to monitor tumor residuals in NMIBC patients prior to Re-TURBT.

Abnormal DNA methylation is closely associated with various human diseases [36]. In recent years, non-invasive urinary DNA methylation testing has demonstrated high efficacy in the early diagnosis of urothelial carcinoma, providing important support for personalized treatment strategies [37, 38]. Previous studies have shown that H4C6 exhibits significantly elevated methylation levels in 17 types of cancer tissues, highlighting its potential as a multi-cancer biomarker [39]. It has also been identified as a diagnostic marker for bladder cancer [27]. Similarly, TWIST1, a key regulatory factor in tumor biology, promotes EMT and facilitates bladder cancer progression [40]. TWIST1 methylation has been reported in 85.71% to 98.2% of bladder cancer tissues, emphasizing its role as a potential diagnostic biomarker [22, 29, 40, 41].

In this study, we assessed the methylation levels of H4C6 and TWIST1 in early-morning urine samples to concentrate methylation substances, thus improving detection sensitivity. Compared to previous studies on urinary H4C6 and TWIST1 methylation for diagnosing bladder cancer [29, 30, 42, 43], the methylation positivity rate in this study was somewhat lower. This discrepancy may be attributed to smaller residual tumor size and reduced exfoliation of tumor cells into urine in patients undergoing Re-TURBT within three months of initial TURBT. Although the combined analysis of H4C6 and TWIST1 improved the AUC, its relatively low sensitivity may result in higher false-negative rates, limiting its standalone diagnostic utility.

In subsequent studies, we integrated dual-gene methylation, hematuria, and VI-RADS scores to develop a nomogram supporting clinical decision-making. The model exhibited high diagnostic accuracy, achieving an AUC of 0.972 (95% CI: 0.937–1.000). The calibration curve showed excellent predictive performance, with a Brier score of 0.065. Additionally, DCA highlighted the clinical utility of the model. The DeLong test demonstrated that the predictive model outperforms the individual factors H4C6-TWIST1 and VI-RADS in terms of predictive power, suggesting that the integration of urinary DNA methylation and imaging can effectively complement each other.

The early-developed urine cytology test (UroVysion) detects genetic abnormalities in urothelial cells with a specificity of over 85%, but its low sensitivity limits its clinical application [11, 44]. Other tumor markers, such as Nuclear Matrix Protein 22 (NMP22) and bladder tumor antigen (BTA), also cannot fully replace cystoscopy [25, 26]. Several previous studies have developed tools based on urinary DNA methylation for predicting tumor residuals after TURBT in NMIBC patients, with results similar to those of our study. For example, a model developed by Wei Ouyang et al. based on urinary NRN1 showed an AUC of 0.97 [31]. Xu Chen et al. developed the utMeMa model, which showed a sensitivity of 93.3% and specificity of 87.5% for detecting tumor residue [45]. This study pioneered a predictive model integrating urinary dual-gene methylation and VI-RADS, offering a non-invasive adjunct to cystoscopy for detecting post-TURBT tumor residuals in NMIBC patients. This approach addresses a critical diagnostic gap and may reduce the need for repeated TURBT procedures. The assay kit employed in this study features a user-friendly protocol, requires no invasive procedures, and demonstrates lower overall costs compared to cystoscopy. Future work will include formal cost-effectiveness analyses to validate its clinical utility and facilitate broader adoption.

In our study, despite the high predictive ability of the VI-RADS scoring system, its accuracy may be insufficient in cases with low scores, small residual lesions, or local postoperative inflammatory changes in the bladder. Urinary DNA methylation has high specificity and sensitivity in detecting trace amounts of tumor information, effectively compensating for the limitations of MRI. Obtaining accurate urinary DNA methylation testing and MRI results prior to Re-TURBT is crucial for Re-TURBT decision-making, as these results provide strong guidance on whether Re-TURBT is necessary. Additionally, the interval between the first TURBT and testing should exceed two weeks to minimize the interference of local bladder inflammation on the test results, ensuring accuracy.

While this prospective study provides novel insights, several limitations warrant consideration. First, the relatively small sample size increases overfitting risks, potentially compromising the model’s robustness and generalizability. Subsequent validation through large-scale, multi-institutional cohorts—with stratified analyses of clinically relevant subgroups (e.g., low-grade tumors, Ta-stage disease)—is essential to establish broader applicability. Second, inherent selection bias may exist, as patients with negative MRI findings showed higher refusal rates for repeat TURBT, potentially introducing spectrum bias. Third, the absence of external validation and multicenter replication limits immediate clinical translation; collaborative efforts across diverse populations are needed to confirm utility. Fourth, confounding factors such as inflammatory-driven methylation or benign urothelial changes could generate false-positive signals [46], while reliance on predefined commercial gene panels restricts exploration of alternative epigenetic biomarkers. Importantly, the short-term follow-up (1–3 months) precludes assessment of longitudinal residual tumor dynamics, necessitating extended surveillance studies. Additionally, standardized implementation of VI-RADS requires rigorous interobserver concordance testing across imaging centers. Future directions should prioritize: 1) expanded biomarker panels to enhance specificity, 2) health-economic evaluations of Re-TURBT reduction potential, and 3) artificial intelligence-assisted VI-RADS interpretation to minimize diagnostic variability.

Conclusion

This study highlights the potential of VI-RADS scoring and urinary H4C6 and TWIST1 methylation as a combined tool for detecting tumor residuals before Re-TURBT in NMIBC patients. Despite the relatively small sample size, our findings emphasize the strong diagnostic potential of these biomarkers. By integrating molecular markers, hematuria, and imaging assessments into a multimodal diagnostic model, we present a novel strategy for accurate risk stratification and personalized treatment of NMIBC patients. The non-invasive model proposed in this study holds significant clinical application potential for routine NMIBC monitoring. It can effectively complement cystoscopy by providing a more accurate identification of high-risk patients through non-invasive means, thereby reducing unnecessary Re-TURBT procedures, optimizing patient management, and lowering healthcare costs. Consequently, we recommend incorporating this model into standard NMIBC monitoring protocols as a valuable auxiliary tool.

Future research should focus on externally validating this model, particularly through larger, multi-center cohort studies to further assess its accuracy. Moreover, efforts to identify new methylation biomarkers and refine diagnostic models are essential for enhancing both sensitivity and generalizability. Such advancements could reduce reliance on invasive procedures, improve patient compliance, minimize surgical risks, and lower financial burdens. Ultimately, these developments will contribute to a more scientific and patient-centered approach in the management and treatment of NMIBC.

Authors’ contributions

Huang Yaqiang: Proposed the main concept and design of the study, supervised the entire research process, and served as the corresponding author. Wang Qi: Conducted the data collection, performed statistical analysis, and wrote the initial draft of the manuscript. Lin Qisheng: Contributed to patient enrollment, clinical data acquisition, and sample preparation for analysis. Wei Weiyang: Participated in urine sample analysis, performed laboratory experiments, and contributed to data interpretation. Liu Yuan: Participated in data analysis, conducted imaging statistical analysis, and contributed to manuscript revision. Li Linfeng: Assisted with laboratory experiments, data organization, and manuscript revisions. Huang Yaqiang: Reviewed and critically revised the manuscript for important intellectual content and approved the final version for submission.

Funding

This work was supported by the Science and Technology Project of Zhongshan Municipality (No. 2024B1032, 2024B1056).

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author, Huang Yaqiang (E-mail:hyq128@126.com), on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Clinical Research and Experimental Animal Ethics Committee of Zhongshan City People’s Hospital (Ethics Approval Number: 2024–128). The study was conducted in accordance with the principles outlined in the Declaration of Helsinki. Informed consent was obtained from all patients and their families prior to participation.

Informed consent was obtained from all individual patients and their families included in the study.

Consent for publication

The authors affirm that informed consent for publication was obtained from all patients and their families.

Competing interests

The authors declare no competing interests.