Multimodal Prognostication for Locally Advanced Cervical Cancer: Development and Internal Validation With Exploratory External Testing
International Journal of Clinical Practice · 2026-01 · 2026 (1)
DOI: 10.1155/ijcp/1177726 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Clinical Practice, online erschienen am 01.07.2026. Verfasst von Xiaojie Ma und 4 weiteren Autor:innen. Die Arbeit nennt 18 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Background
Outcomes for locally advanced cervical cancer (LACC) post–concurrent chemoradiotherapy (CCRT) exhibit substantial heterogeneity. This study developed and internally validated a multimodal prognostic system integrating clinicopathological variables and multisequence MRI radiomics, with exploratory bioinformatics analysis of genomic signatures from an independent public dataset to predict overall survival (OS) and progression‐free survival (PFS).
Methods
A development cohort of 446 LACC patients (2014–2021) was randomly partitioned (8:2) into training n = 356 and internal validation n = 90 sets. An independent temporal cohort n = 116, 2022 − 2023 provided exploratory external testing primarily for PFS. OS was modeled via stepwise Cox regression. For PFS, nine machine learning architectures were evaluated within a Fine–Gray competing risk framework to account for mortality as a competing event.
Results
The clinical OS model achieved a C‐index of 0.767, effectively stratifying patients into high‐risk (5‐year survival: 62%) and low‐risk (94%; p < 0.0001) groups. Integration of triple‐sequence MRI radiomics significantly enhanced accuracy, with the fusion model achieving a C‐index of 0.863 and a peak AUC of 0.954. For PFS prediction, neural network B emerged as the optimal architecture, demonstrating superior discriminative power in the external testing set (AUC = 0.868 and sensitivity = 0.936) and refined probability estimation (Brier score = 0.230). Exploratory analysis of a four‐gene signature derived from TCGA‐CESC independently corroborated risk stratification with a 4‐year AUC of 0.925, though validation in the institutional cohort was not performed.
Conclusions
While internally validated, the OS model’s transportability to external settings requires confirmation in larger multicenter cohorts given the exploratory nature of external testing. The PFS model demonstrated robust discrimination but exhibited temporal calibration drift, warranting prospective validation. This multidimensional framework nonetheless provides a prognostic tool with superior discrimination compared with traditional FIGO staging. By incorporating an optimized neural network within a competing risk framework, the system offers a methodologically rigorous assessment of recurrence risk, facilitating personalized therapeutic escalation post‐CCRT.
Zitieren
APA
Ma, X., Tang, N., Zhang, H., Zou, Q., & Han, Y. (2026). Multimodal Prognostication for Locally Advanced Cervical Cancer: Development and Internal Validation With Exploratory External Testing. International Journal of Clinical Practice, 2026(1). https://doi.org/10.1155/ijcp/1177726
Vancouver
Ma X, Tang N, Zhang H, Zou Q, Han Y. Multimodal Prognostication for Locally Advanced Cervical Cancer: Development and Internal Validation With Exploratory External Testing. International Journal of Clinical Practice. 2026;2026(1). doi:10.1155/ijcp/1177726
BibTeX
@article{Ma2026,
author = {Ma, Xiaojie and Tang, Ni and Zhang, Hanjing and Zou, Qian and Han, Yunwei},
title = {Multimodal Prognostication for Locally Advanced Cervical Cancer: Development and Internal Validation With Exploratory External Testing},
journal = {International Journal of Clinical Practice},
year = {2026},
volume = {2026},
number = {1},
doi = {10.1155/ijcp/1177726}
}
EUVIMED