Uncertainty analysis shows equivalence of PTV-based VMAT and robust IMPT for model-based selection
Jesus Rojo Santiago,
The Netherlands
OC-0449
Abstract
Uncertainty analysis shows equivalence of PTV-based VMAT and robust IMPT for model-based selection
Authors: Jesus Rojo Santiago1,2, Erik Korevaar3, Zoltán Perkó4, Stefan Both3, Steven J.M. Habraken1,2, Mischa S. Hoogeman1,2
1Erasmus MC Cancer Institute, Radiotherapy, Rotterdam, The Netherlands; 2HollandPTC, Medical Physics & Informatics, Delft, The Netherlands; 3University Medical Center Groningen, Radiation Oncology, Groningen, The Netherlands; 4Delft University of Technology, Radiation Science, Delft, The Netherlands
Show Affiliations
Hide Affiliations
Purpose or Objective
In the Netherlands, head-and-neck cancer (HNC)
patients are referred for proton therapy through model-based selection. For each
patient, a photon (VMAT) and a proton (IMPT) plan are made. The plans are
compared in terms of normal tissue complication probabilities (NTCPs) for grade
II and III xerostomia and dysphagia. However, the differences between the
modalities may impact CTV dose and NTCPs. Our aim is to assess: (i) the
consistency and robustness of CTV dose for VMAT and IMPT and (ii) the
sensitivity of NTCPs to beam and patient alignment errors (geometric) and,
for IMPT, stopping-power prediction errors (range).
Material and Methods
Thirty oropharyngeal HNC patients, treated to 70
Gy(RBE) and 54.25 Gy(RBE) for the primary and elective CTVs respectively, were included. Clinical
VMAT and IMPT plans for all patients were available from the plan comparison. A 3mm PTV margin and 3mm/3%
geometric and range robustness settings were used for VMAT and IMPT planning, respectively. For VMAT, dose was
prescribed to the PTV-D98%,PTV≥95%Dpres, while, for IMPT, it was
prescribed to the voxel-wise minimum dose of the 28 clinical robustness evaluation scenarios: VWmin-D98%,CTV≥94%Dpres.
Polynomial chaos expansion (PCE) was applied to generate a fast patient- and plan-specific model of voxel doses. PCE enabled a robustness evaluation of 100,000
error scenarios for each plan. Systematic and random geometric errors were
sampled from Gaussian distributions with errors (1SD) of Σ = 0.94mm and σ = 1.14mm, consistent with a M = 2.5Σ+0.7σ = 3mm margin based on van Herk's recipe. A systematic range error of 1.5% (1SD) from literature was used, in line with the 3% range
setting. For each patient and each plan, the PCE model was used to calculate the
median and the 5th and 95th percentiles of the D98% to both CTVs and the
NTCPs. Zero baseline toxicities were assumed for all patients.
Results
Figure 1 shows a correlation plot of the D98%
to both CTVs with VMAT and IMPT planning. For IMPT, the population median D98% was
69.1 (range 68.5-69.4) GyRBE and 53.6 (range 53.3-53.8) GyRBE for the primary
and elective CTVs. For VMAT, values of 68.7 (68.4-68.9) Gy and 53.2 (52.9-53.3) Gy were found respectively. Figure 2 shows
similar NTCP spreads for the grade II and
III xerostomia and dysphagia in both modalities. A median NTCP spread of 2.39
(1.16-2.93) p.p. and 2.47 (0.28-3.86) p.p. for IMPT and 1.86 (1.28-2.33) p.p. and 2.37 (0.90-3.46) p.p. for VMAT were found for grade
II xerostomia and dysphagia. For grade III xerostomia and dysphagia, values of 0.73 (0.29-1.02) p.p.
and 0.37 (0.03-2.11) p.p. for IMPT and 0.62
(0.33-0.90) p.p. and 0.73 (0.09-2.07) p.p. for VMAT were obtained respectively.
Conclusion
Despite the differences in photon and proton
planning, the comparison between PTV-based VMAT vs. robustly optimized IMPT is
consistent, with a slightly higher CTV dose in IMPT. Geometric and range errors
have a moderate impact on the NTCP, indicating that the nominal plan is a good
estimator to select patients for IMPT.