Main Session
Sep
28
PQA 03 - Digital Health Innovation and Informatics, Patient Safety & Quality, and Radiation and Cancer Biology
2500 - A Mechanistic In-Silico Virtual Clinical Trial System for Optimizing RT-Chemotherapy-Immunotherapy in Stage III NSCLC
Presenter(s)
Daisuke Kawahara, PhD - Hiroshima University, Hiroshima, Hiroshima
D. Kawahara1, Y. Murakami2, M. Kishi3, T. Wada4, and Y. Kadooka5; 1Department of Radiation Oncology, Hiroshima University Hospital, Hiroshima, Japan, 2Department of Radiation Oncology, Graduate School of Biomedical Health Sciences, Hiroshima University, Hiroshima, Japan, 3Hiroshima University, Hiroshima, Japan, 4Graduate School of Biomedical and Health Sciences, Hiroshima University, Hiroshima, Hiroshima, Japan, 5Graduate School of Biomedical and Health Sciences, Hiroshima University, Hiroshima, Japan
Purpose/Objective(s):
To develop an in-silico virtual clinical trial (VCT) platform calibrated to PACIFIC-like pneumonitis rates and to generate safety-aware CRT–immunotherapy regimen hypotheses under explicit pneumonitis (PRP) constraints.Materials/Methods:
PRP was modeled as an event probability combining RT mean lung dose response with immunotherapy-associated pneumonitis driven by dosing and schedule, calibrated to PACIFIC symptomatic PRP (Grade =2: 19.8%; Grade =3: 4.6%). Stage-1 isotoxic screening randomly sampled 600 regimens, simulating 1,200 virtual patients per regimen. Variables included immunotherapy start delay (0–70 days after CRT), duration (3–24 months), dosing interval (every 2–4 weeks), relative immunotherapy intensity, relative chemotherapy intensity, and RT plan-quality assumptions represented as approximately 0–20% reductions in mean lung dose. Endpoints were tumor control probability (TCP) and RMST at t=548 days derived from simulated PFS Kaplan–Meier curves. Regimens meeting mean PRP constraints were considered feasible, and Pareto analysis was used to visualize efficacy–toxicity tradeoffs across the feasible design space.Results:
411/600 regimens were feasible. The best TCP-feasible regimen achieved TCP=0.633 with PRP(G=2)=0.134 and PRP(G=3)=0.038 and RMST(548)=407.8 days, using immunotherapy start ~10 weeks after CRT, 24-month duration, dosing every 4 weeks, ~20% higher immunotherapy intensity, ~10% lower chemotherapy intensity, and an assumed ~20% lower mean lung dose. The best RMST-feasible regimen achieved RMST(548)=409.2 days with TCP=0.628 using start ~2 weeks, 18-month duration, dosing every 4 weeks, ~30% lower immunotherapy intensity, standard chemotherapy intensity, and an assumed ~20% lower mean lung dose. Relative to the feasible median (TCP=0.542; RMST(548)=392.4 days), improvements were +0.091 TCP and +16.8 RMST days. Clinically, higher-ranked feasible regimens tended to combine improved lung dose metrics (“toxicity headroom”) with longer immunotherapy exposure (18–24 months) and less frequent dosing schedules (every 4 weeks) while remaining within the calibrated PRP envelope.Conclusion:
We developed a mechanistic in-silico VCT system enabling quantitative screening of CRT–immunotherapy regimens under PACIFIC-anchored pneumonitis constraints. In this hypothesis-generating Stage-1 screen, feasible candidates improved simulated TCP/RMST without exceeding mean PRP targets, and Pareto analysis highlighted a clinically interpretable pattern favoring improved plan quality, extended immunotherapy duration, and less frequent dosing (every 4 weeks). Ongoing Stage-2 refinement with larger virtual cohorts will assess stability of recommended regimens and incorporate additional timing-toxicity tradeoffs to support rational design of future CRT–immunotherapy trials.