Main Session
Sep 28
PQA 03 - Digital Health Innovation and Informatics, Patient Safety & Quality, and Radiation and Cancer Biology

2430 - Validation of a Population-Based GARD Tumor Control Model Using Published Carbon Ion Dose Escalation Outcomes in Stage I NSCLC

10:45am - 12:00pm ET
Poster Hall - Exhibit Hall A
Screen: 3
POSTER

Presenter(s)

Celina Chiodo, MD, MS Headshot
Celina Chiodo, MD, MS - Moffitt Cancer Center, Tampa, FL

C. Chiodo1, D. P. Calvin2, C. Hajj3, S. A. Eschrich4, J. G. Scott5, and J. F. Torres-Roca1; 1Department of Radiation Oncology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL, 2Department of Radiation Oncology, University of Mississippi Medical Center, Jackson, MS, 3Cleveland Clinic Abu Dhabi, Abu Dhabi, United Arab Emirates, 4Department of Integrated Bioinformatics and Biostatistics, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL, 5Department of Radiation Oncology, Taussig Cancer Institute, Cleveland Clinic Foundation, Cleveland, OH

Purpose/Objective(s): Radiation dose in non-small cell lung cancer (NSCLC) remains empirically prescribed despite substantial biological heterogeneity. We previously developed a population-level genomic-adjusted radiation dose (GARD)–based tumor control probability (TCP) model using 1,600 genomically profiled NSCLC patients. The model quantifies biologically optimal dose (RxRSI) and predicts population-level local control as a function of delivered dose. We evaluated whether model predictions align with outcomes from a published carbon ion dose-escalation study in stage I NSCLC.

Materials/Methods: The published NSCLC GARD-based TCP model was used to generate predicted population-level local control across 66–124 Gy (EQD2). For each dose level, the model estimates the proportion of patients achieving RxRSI and corresponding TCP. Predictions were compared to reported local control from a Japanese phase I/II carbon ion trial (Miyamoto et al., 2003) that escalated dose from 59.4–95.4 GyE (18 fractions) and 68.4–79.2 GyE (9 fractions). Carbon ion doses were converted to EQD2 to permit comparison with RxRSI-derived dose distributions. Pearson correlation between predicted and observed local control was calculated.

Results: The GARD-based TCP model predicted increasing local control with escalating dose, reflecting a greater proportion of patients achieving biologically optimal dosing at higher EQD2 levels. In the carbon ion study (n=81), dose-dependent improvements in local control were observed, with doses >86.4 GyE (18 fractions) and =72 GyE (9 fractions) achieving 90–95% local control. Across dose cohorts, predicted and observed local control were strongly correlated (R=0.91, p=0.00026) (see Table 1). Dose levels exceeding approximately 84 Gy EQD2 corresponded to substantial increases in predicted RxRSI achievement and reported clinical benefit.

Conclusion: Population-level GARD-based TCP modeling closely aligns with the dose–response observed in a published carbon ion dose-escalation study in stage I NSCLC. These findings support the biological validity of GARD-informed TCP modeling and suggest that delivery platforms capable of safely escalating effective dose may improve local control in biologically radioresistant NSCLC.

Table 1. Comparison of Predicted Population-Level Local Control from GARD-Based TCP Model and Observed Local Control in Carbon Ion Trial

EQD2 (Gy) Predicted Local Control (GARD-TCP) Observed Local Control (Japanese Trial)
66 0.45 0.40
74 0.55 0.57
84 0.70 0.75
95 0.85 0.66
100 0.90 0.80
106 0.95 0.95
108 0.97 0.98
112 0.98 1.00
124 0.99 1.00