KEY POINTS
- The study included 192 consecutive patients with esophageal cancer treated with intensity-modulated proton therapy from 2019-2025. 85 patients underwent adaptive replanning and 107 remained non-adaptive; treatment consisted of either neoadjuvant 23 × 1.8 Gy or definitive 28 × 1.8 Gy chemoradiotherapy.
- Adaptation was triggered by ITV D98% falling below 94% of prescription in the voxel-wise minimum robustness evaluation, an increase in mean heart dose of ≥1.5 Gy, or both. Of the 85 adapted patients, 40 were adapted for target underdose, 33 for cardiac overdose, and 12 for both triggers.
- Using deformable dose accumulation, the authors compared the clinically delivered multi-plan adaptive course with a simulated single-plan course using only the initial plan. Across adapted patients, ITV D98% increased from 95.8% to 97.1% (p=0.028), GTV D98% from 98.9% to 99.7% (p<0.001), and ITV Dmax decreased from 108.4% to 107.4% (p<0.001).
- In patients adapted specifically for target underdose, ITV D98% improved from 95.0% to 96.9%, a 1.9 percentage-point gain, while GTV D98% increased from 98.9% to 99.7%. Failure of the ITV coverage criterion fell from 10 of 40 patients (25%) without adaptation to 5 of 40 (12.5%) after adaptation.
- Heart-triggered adaptation produced a different benefit: mean heart dose decreased from 12.4 to 11.4 Gy, and heart V40Gy fell from 9.1% to 7.4%. The predicted 2-year post-treatment mortality NTCP decreased from 33.9% to 32.4%, an absolute reduction of 1.5 percentage points.
- Adaptation did not produce a meaningful lung-dose penalty. Across all adapted patients, mean lung dose remained 5.9 Gy with and without adaptation (p=0.969), supporting the feasibility of cardiac sparing without trading against pulmonary exposure.
- Adaptive therapy did not eliminate all cumulative undercoverage: 17 of 85 patients (20.0%) failed the coverage goal in the simulated non-adaptive course versus 12 of 85 (14.1%) after adaptation. Interpretation of small cardiac-dose changes also requires caution because deformable registration uncertainty for mean heart dose was approximately 0.62 Gy, and the mortality effect was derived from an NTCP model rather than observed survival.
CLINICAL TAKEAWAY
For esophageal IMPT, a structured adaptive workflow that monitors both target coverage and cardiac dose appears more informative than relying on target underdose alone. The data support explicit cardiac triggers for replanning, but this remains workflow-level rather than survival evidence because the study is retrospective, depends on deformable dose accumulation, and uses modeled NTCP.