3044 - Retrospective Weekly Photon Re-Optimization on QA Scans to Assess Convergence toward Proton Dosimetry during Treatment
Presenter(s)
Y. Kumar1, D. N. Stanley1, U. S. Ram1, R. Djavadian2, S. Chen1, C. D. Willey1, J. A. Bonner1, R. A. Cardan1, J. A. Pogue1, J. Harms3, M. B. Yusuf1, C. Stanley1, N. Viscariello1, and C. Cardenas1; 1University of Alabama at Birmingham, Birmingham, AL, 2University of Alabama at Birmingham Department of Radiation Oncology, Birmingham, AL, 3WashU Medicine, Department of Radiation Oncology, St. Louis, MO
Purpose/Objective(s): Proton therapy (PT) has been associated with improved overall survival in head and neck cancer but is costly and resource-limited. We hypothesized that weekly online photon adaptation could achieve dosimetry comparable to PT by continuously accounting for anatomic changes during treatment.
Materials/Methods: Ten previously delivered head and neck proton cases with weekly QA scans (45 total scans) were retrospectively analyzed. For each QA scan, DVH metrics were extracted for (1) the clinically delivered proton plan recalculated on that week’s anatomy and (2) weekly adapted photon plans generated on the same QA anatomy. Reference photon plans were retrospectively created and physician reviewed/approved; adaptive plans were generated in a CBCT-guided adaptive emulator using the weekly QA CT scans. Analysis was limited to the shared window (weeks 1–4) and endpoints present for all patients: CTV_High D95% and D98%, spinal cord/brainstem (and PRVs) D0.03cc, parotid mean and V30, esophagus mean and V50, oral cavity mean, and mandible D0.03cc. Repeated measures were analyzed with patient-correlated models (Value ~ Modality + Week + Modality×Week; random patient intercept). The Modality×Week interaction tested convergence (decreasing adaptive-photon – proton differences over time). Per-patient drift was summarized by slope differences (adaptive-photon – proton) with Wilcoxon testing.
Results: Weekly-adapted photon plans’ target coverage was slightly higher than proton at baseline: CTV_High D95% +1.24 Gy and D98% +2.07 Gy (both p<0.001), with no evidence of convergence over time (Modality×Week p=0.79). For OARs, photon – proton differences were frequently positive, including spinal cord D0.03cc +12.0 Gy, brainstem D0.03cc +12.4 Gy, and oral cavity mean +12.1 Gy; mandible D0.03cc was lower with photons (-3.1 Gy). Photon – proton differences did not consistently decrease across weeks; only oral cavity mean showed a small negative drift (-0.36 Gy/week; p=0.03) and esophagus V50 showed a negative slope difference (-0.94%/week; p=0.036), neither remaining significant after multiple-testing correction.
Conclusion: In this cohort, retrospectively generated weekly adaptive photon plans did not converge toward recalculated proton dosimetry. Modality-dependent differences were observed while weekly-adapted photon plans maintained (and slightly increased) target coverage. Larger, coverage-constrained studies are warranted to determine whether QA-guided adaptation can yield clinically meaningful convergence or define practical adaptation triggers.