Comment on Yu et al. “CT-based radiomics signature of visceral adipose tissue for prediction of early recurrence in patients with NMIBC: a multicentre cohort study”

To the Editor, Yu and colleagues present a clinically pragmatic concept: extracting CT-derived radiomics features from visceral adipose tissue (VAT) to predict early (≤12 months) recurrence after initial TURBT in non-muscle-invasive bladder cancer (NMIBC)[1]. Early recurrence is a meaningful endpoint because NMIBC recurrence remains common despite established clinicopathologic risk tools and intensive surveillance pathways[2,3]. This commentary is written in alignment with the TITAN 2025 Guidelines to support transparent and responsible reporting[4]. Key strengths and “take-home” findings Across three centers (n = 325), the VAT radiomics model (VAT-RM) showed stable discrimination (AUC ~ 0.81–0.85) and consistently outperformed the SAT-RM, while decision-curve analysis favored VAT-RM for net clinical benefit[1]. Importantly, combining VAT-RM with simple clinical factors (tumor number, tumor base morphology, immediate intravesical instillation) yielded a fusion model with excellent AUCs (~0.91–0.94) and superior performance versus EORTC/CUETO/EAU stratification[1]. This is a strong “proof-of-utility” example of radiomics as mineable quantitative data beyond visual interpretation[5,6]. Why VAT (and why it matters) The work shifts attention from BMI-like surrogates to adipose heterogeneity, arguing that VAT’s endocrine–inflammatory activity may better capture the host–tumor milieu relevant to recurrence biology[1]. From a translational standpoint, the VAT-RM provides a noninvasive readout that can be extracted from standard-of-care CT, potentially enabling earlier identification of patients who warrant intensified cystoscopic surveillance or earlier consideration of adjuvant strategies[1]. Points for caution and next steps Despite multicenter validation, the study remains retrospective with modest event counts, and imaging protocol heterogeneity (slice thickness, reconstruction kernels) can influence radiomics stability – underscoring the need for rigorous feature standardization and external calibration[5,6]. Clinically, the key unanswered question is incremental value: how much does the fusion model improve decision thresholds (e.g., surveillance intensity or early BCG escalation) beyond contemporary EAU risk groups, particularly in BCG-treated populations?[3,7] Prospective evaluation with prespecified imaging harmonization and impact-analysis endpoints (e.g., fewer missed early recurrences, acceptable false-positive burden) would accelerate clinical adoption.