S422
Clinical - Gynaecological
ESTRO 2026
achieved AUC = 0.64 ± 0.15, indicating moderate discrimination, and random-effect variance ( σ² = 358) confirmed relevant inter- patient variability. Among ML models, Random Forest achieved the best balance (sensitivity 0.66, precision 0.77 for late toxicity), confirming a measurable, but limited predictive ability of dosimetric and anatomic variables. Conclusion: SBRT-induced toxicity in gynecologic cancers is mainly driven by anatomical and biological factors, while traditional dosimetric metrics have limited predictive power. Both mixed- effects and machine-learning models showed that dose patterns modestly but consistently affect toxicity prediction. These results highlight the value of multilevel modeling and suggest that combining organ-at-risk dosimetry, radiomics, and patient-specific biology could improve prediction and personalize SBRT planning. References: 1.Macchia G et al. doi: 10.1016/j.ijrobp.2023.04.025. .2.Macchia G et al. doi: 10.1136/ijgc-2021-003237. 3.Macchia G et al. doi: 10.1634/theoncologist.2019- 0309. Keywords: A.I. predicition model; SBRT; toxicity Real-World Outcomes of Induction Chemotherapy in Locally Advanced Cervical Cancer Alexander Pawsey 1 , Simran Vaja 2 , Anam Habib 1 , Guihong Liu 3 , Asma Sarwar 1 , Mary McCormack 1 , Gemma Eminowicz 1 1 Oncology department, University College London Hospital, London, United Kingdom. 2 Cancer Research UK Cancer Trials Centre, University College London, London, United Kingdom. 3 Oncology department, Dongguan Tungwah Hopsital, Dongguan, China Digital Poster 1431
Induction chemotherapy (IC) with carboplatin+paclitaxel before
chemoradiotherapy (CRT) improves survival in locally advanced cervical cancer (LACC)(1). IC has been used at our UK-centre since 2015 for this population. We aim to evaluate real- world outcomes of patients treated with IC over a decade. Material/Methods:
All patients with LACC treated with IC followed by CRT were retrospectively identified on electronic records. Data included FIGO-stage, histology,
chemotherapy details, maximum-tumour- diameter (MTD) before treatment and at brachytherapy (BT), overall CRT treatment time (OTT), 12-week MRI response, and long- term outcomes (until 23/10/2025). PFS and OS were analysed using Kaplan–Meier methods. Results: 56 patients were identified with complete data for 52 (table 1); median follow-up 80.4 months. Median age was 50 years (range 27– 78). FIGO(2009) stage: IB1 1 (1.9%), IIA2 2 (3.8%), IIB 29 (55.8%), IIIB 14 (26.9%), IVA 6 (11.5%); 33 (63.5%) were node-positive, 10 (19.2%) para-aortic node-positive. Squamous histology was present in 40 (76.9%). Median MTD at diagnosis was 5.5cm (range 0–10.5). 48 (92.3%) completed six IC cycles, 40 (76.9%) completed ≥ 4 cisplatin cycles.
Purpose/Objective:
Made with FlippingBook - Share PDF online