KEY POINTS
- A portable 3D-printed head-and-neck phantom was tested at six proton therapy centers using their local pencil-beam-scanning infrastructure, including different treatment-planning, imaging and delivery systems. The phantom reproduced clinically relevant changes including weight loss, cavity filling changes and soft-tissue/bone interfaces.
- Three scenarios were tested: the original plan on baseline anatomy, the same plan delivered after an anatomical change without adaptation, and a newly optimized plan delivered on the altered anatomy. Point-dose testing used 2.2 Gy(RBE), while film-based two-dimensional testing used 6.6 Gy(RBE).
- Across all centers and scenarios, ion-chamber measurements agreed closely with treatment-planning calculations, with a median deviation of only −0.7% [IQR 1.2%]. There were no significant differences between institutions or scenarios, supporting reproducibility of the measurement framework.
- Delivering the original plan after the simulated anatomical change produced a median 13.9% target-dose decrease and a very large relative dose distortion in the buildup region. Replanning on the changed anatomy restored target dose to within 0.2% [IQR 0.8%] of baseline, with buildup-region differences no greater than 3.3%.
- Film measurements showed the same pattern. Median 2%/2-mm gamma pass rate was 98.3% for baseline delivery, fell to 85.9% when anatomy changed without adaptation, and increased to 89.8% after replanning; residual differences partly reflected different optimization weights between the original and adapted beam sets.
- Corrected CBCT supported accurate adaptive dose calculation, with median differences of 0.3% in the target and 0.0% in the buildup region. Log-file reconstruction differed from ion-chamber measurements by a median 0.1%, demonstrating compatibility with independent online QA approaches.
- Range-verification tools directly detected the induced anatomy change: prompt-gamma imaging measured a 5.9-mm range shift against a simulated 6.0-mm shift, while proton radiography measured 18 mm against an expected 17.8-mm water-equivalent shift. The deliberately simplified and often non-robust plans increase test sensitivity but do not reproduce every clinical adaptive scenario.
CLINICAL TAKEAWAY
Online adaptive proton therapy needs validation of the entire chain, not just the replanning algorithm. This multicenter framework provides a practical way to test whether a center can detect an anatomical change, quantify its dosimetric consequence, adapt correctly and independently verify the result before clinical deployment.