Clinical landscape of urachal carcinoma: a two-decade real-world study

Dear Editors, Urachal carcinoma (UrC) is an exceptionally rare and aggressive malignancy arising from urachal remnants, representing <1% of bladder cancers and occurring at an incidence on the order of ~1 per million per year[1,2]. Because of this extreme rarity, the evidence base for UrC has historically been dominated by case reports and small retrospective series, with limited ability to comprehensively describe demographic distributions, histopathologic spectrum, metastatic patterns, and real-world systemic treatment utilization. Even in contemporary literature, single-center cohorts commonly include less than 5 patients, and larger multi-institutional efforts remain uncommon[3–6]. This scarcity contributes directly to ongoing uncertainty in risk stratification and systemic management, since prospective trials are rarely feasible and standardized systemic regimens are not well established. Moreover, most published clinical datasets are derived from Western populations[2,7–9], limiting the generalizability of clinical atlases of UrC to Asian patients, where baseline characteristics, treatment patterns, and metastatic behavior may differ. To address these gaps, we present a comprehensive clinical landscape of UrC within a Chinese population, drawing on data from a cohort of 124 patients accrued over a span of twenty years (2005–2025) at the Sun Yat-sen University Cancer Center. This cohort represents one of the largest sample sizes reported for this rare malignancy. By systematically analyzing patient demographics, histopathological subtypes, metastatic profiles, and treatment patterns, this study offers crucial real-world insights into UrC, filling a significant knowledge void in this under-researched area and providing a foundation for future clinical investigations. The cohort was predominantly male (76.42%, n = 94), with a mean age at diagnosis of 49.00 years (±12.31). Of the patients, 73.17% (n = 90) presented with primary disease, while 26.83% (n = 33) had recurrent disease at diagnosis (Fig. 1A). Histopathological analysis identified adenocarcinoma as the most frequent subtype (68.29%, n = 84), followed by signet-ring cell carcinoma (3.25%, n = 4). Other less common histological types included fibroblastoma (2.44%, n = 3), PEComa (0.81%, n = 1), and squamous cell carcinoma (0.81%, n = 1) (Fig. 1A). In terms of tumor differentiation, high-grade tumors were observed in 16.26% (n = 20), low-grade in 32.52% (n = 40), and moderate differentiation in 34.15% (n = 42). Recurrence occurred in 30.89% (n = 38) of cases, while 66.67% (n = 82) remained recurrence-free. Metastatic involvement was noted in 51.22% (n = 63) of patients, with distant metastasis being most prevalent, underscoring the aggressive clinical behavior of this malignancy (Fig. 1A). Figure 1.: Clinical atlas and survival analysis of urachal carcinoma. (A) Baseline characteristics. (B) Kaplan–Meier survival analyses. (C) The time-dependent ROC curves of survival machine-learning approaches. Given the paucity of published research in urachal carcinoma, the prognostic relevance of routine clinicopathological variables remains uncertain[9]. We therefore performed a comprehensive set of Kaplan–Meier survival analyses to screen potential predictors of overall survival (OS) (Fig. 1B). The median OS for the entire study cohort was 39.1 months. Among the variables examined, recurrence status was associated with poorer OS (HR: 0.47, 95% CI: 0.26–0.83; P = 8.2 × 10−3) (Fig. 1B). Advanced Mayo stage (III/IV vs I/II) conferred markedly worse survival (HR: 0.21, 95% CI: 0.12–0.39; P = 4.6 × 10−8), as did advanced Sheldon stage (IV vs I–III) (HR: 0.21, 95% CI: 0.11–0.38; P = 2.7 × 10−8). Presence of metastatic disease was strongly associated with inferior survival (Fig. 1B). Among serum tumor markers, elevated CEA predicted worse outcomes (HR: 2.94, 95% CI: 1.13–7.62; P = 0.02), and elevated CA72-4 was similarly associated with poorer survival (HR: 3.35, 95% CI: 1.44–7.75; P = 2.8 × 10−3) (Fig. 1B). Having identified clinicopathological variables with prognostic signal on univariable survival analyses, we next sought to develop and internally validate prediction models for overall survival, aiming to transform the descriptive clinical atlas into an actionable risk stratification tool for this rare malignancy. Specifically, we applied eight survival machine-learning approaches – RSF, GBM, Lasso-Cox, CoxBoost, survival-SVM, XGBoost, SuperPC, and LSR-Cox[10] to develop and benchmark prognostic models for overall survival and to enable risk stratification in urachal carcinoma (Fig. 1C). Across prespecified time horizons, the time-dependent ROC curves showed consistently strong discrimination in the training cohort, indicating that these routinely available clinicopathological factors collectively provide robust prognostic signal for overall survival in urachal carcinoma, and can be integrated into a practical framework for risk stratification (Fig. 1C). In conclusion, our 20-year, single-center cohort of 124 Chinese patients with UrC delineates the largest and most detailed clinical-pathological profile of this rare malignancy in an Asian population to date. The data reveal a striking male preponderance, a mean age at diagnosis younger than that commonly reported for conventional bladder carcinomas, and a high incidence of metastatic disease at presentation. The histological predominance of adenocarcinoma aligns with prior series[7,9], but the inclusion of rare subtypes (e.g., PEComa, fibroblastoma) enriches the pathological spectrum of UrC. The frequent use of urachal excision and the emerging role of minimally invasive techniques reflect evolving surgical practice. Importantly, the high rates of advanced disease (Mayo stage IV: 24–39%; Sheldon stage III/IV: >90%) highlight the need for earlier detection and more effective systemic strategies. Eventually, eight survival machine-learning models demonstrated consistently good time-dependent discrimination, supporting the feasibility of integrating routinely available variables into practical risk stratification for UrC. Given the extreme paucity of UrC-specific evidence – particularly from non-Western populations – this comprehensive landscape offers an invaluable reference for clinical decision-making, trial design, and prognostic stratification. Future studies should validate these findings in multi-national cohorts and explore molecular drivers to guide targeted therapies.