Investigating the Correlation Between Pre-Treatment Imaging-Derived Body Composition, Chemotherapy Dose Adjustment, and Treatment Efficacy in Gynecological Cancer Patients "

Endometrial cancer ranks sixth in incidence among women worldwide and is rising, while ovarian cancer, though less common, remains the leading cause of gynecologic cancer mortality. Standard adjuvant chemotherapy for these malignancies includes paclitaxel (dose based on body surface area, BSA) and carboplatin (dose based on AUC). However, BSA-based dosing does not account for variations in body composition, which can significantly influence drug distribution, toxicity, and treatment efficacy. Retrospective analysis of 124 endometrial cancer patients at National Cheng Kung University Hospital demonstrated that CT-derived skeletal muscle parameters (SMA, SMI) are significantly associated with grade ≥3 hematologic toxicities and outperform traditional clinical judgment for risk prediction. Patients experiencing severe hematologic toxicity within the first two cycles are more likely to experience repeated toxicities in subsequent cycles. This trial integrates prospective randomized controlled trial (RCT) and retrospective target trial emulation. Prospective RCT: 294 patients will be stratified by cancer stage and randomized 1:1 to intervention vs. control groups. The intervention group will undergo AI-assisted risk stratification based on preoperative CT body composition analysis, with high-risk patients receiving a 15% paclitaxel dose reduction from cycle 1. The control group will receive standard-of-care dosing adjustments based on clinical judgment. Retrospective analysis: Patients diagnosed from 2012-2024 with available CT images and complete treatment records will be analyzed using propensity score methods to control for confounders. Primary endpoint: incidence of grade ≥3 leukopenia or neutropenia within the first two cycles. Secondary endpoints: chemotherapy cycle delay (\>7 days), hospitalization rate, HRQoL (CIPN, EORTC-QoL), two-year PFS and OS, cost-effectiveness (ICER, VBP). The trial hypothesizes that AI-guided dosing adjustments will reduce severe hematologic toxicity rates from 75% to 60% without compromising relative dose intensity (\>85%), thereby achieving both clinical and economic benefits.