LET optimization reduces biological dose model disagreement in carbon-ion pelvic osteosarcoma plans

In 15 pelvic osteosarcoma plans, LET optimization reduced LEM-I versus mMKM target dose discrepancies by up to 9% of prescription, with modest robustness penalties.

KEY POINTS

  • This retrospective planning study from the Heidelberg Ion-Beam Therapy Center analyzed 15 patients with unresectable pelvic osteosarcoma previously treated under the OSCAR protocol. The clinical treatment consisted of 54 Gy(RBE) proton therapy in 27 fractions, followed by an 18 Gy(RBE) carbon-ion boost in six fractions. Only the carbon-ion boost plans were reoptimized for the present investigation; median GTV was 411.3 cm³ (IQR 193.3–686.1).
  • Original carbon-ion plans were optimized using the Local Effect Model I (LEM-I). Investigators introduced explicit dose-averaged LET objectives of >45 keV/μm in the CTV, >50 keV/μm in the GTV, and >55 keV/μm in the GTV excluding its outer 10 mm, while maintaining target dose and OAR constraints. Biological dose distributions were then recalculated with the modified microdosimetric kinetic model (mMKM) to quantify model-dependent differences.
  • LET optimization substantially increased target LET while maintaining LEM-I-based target coverage. Median GTV LET95 increased from 31 keV/μm (IQR 28–33) in dose-optimized plans to 49 keV/μm (IQR 47–50) after LET optimization (p<0.001). The proportion of GTV receiving LET ≥50 keV/μm increased from approximately 15% to 88%, while CTV coverage at LET ≥45 keV/μm increased from 39% to 97%.
  • The greatest benefit was reduced disagreement between biological dose models. After mMKM recalculation, median CTV D50 increased from 16.5 to 17.3 Gy(RBE), approaching the 18-Gy(RBE) LEM-I prescription. GTV D95 improved from approximately 86% to 94% of prescription. LET optimization reduced the median LEM-I–mMKM dose discrepancy by 9.0 percentage points of prescription for GTV D95 and 6.3 points for GTV D50 (both p<0.001).
  • An additional beam was required in 12/15 cases to achieve improved LET distributions, so comparisons were also performed against dose-only plans with the same number of beams. This helped distinguish LET optimization effects from changes in beam geometry. Most optimized plans still did not fully satisfy the specified minimum LET objectives, particularly in the central target, demonstrating practical limitations when balancing high LET against homogeneous dose coverage and OAR constraints.
  • OAR doses remained largely unaffected by the optimization strategy or biological dose model. However, LET optimization modestly worsened robustness under anatomical, setup, and range perturbations. In particular, ±3% density uncertainty increased the dispersion of target dose metrics by approximately 0.7–2.3 percentage points of prescription, while changes assessed on repeated control CT scans were generally smaller. These effects were substantially below the reduction in model-to-model uncertainty but remained statistically and technically relevant.
  • The analysis was entirely dosimetric and involved 15 retrospectively reoptimized carbon-ion boost plans, not prospectively delivered LET-optimized treatment. Moreover, the carbon-ion boost represented only part of the combined proton/carbon-ion prescription, and biological dose predictions depend on the selected LEM-I and mMKM implementations. No tumor-control, toxicity, or validated clinical outcome endpoint established that increasing target LET improves treatment effectiveness.

CLINICAL TAKEAWAY

LET-guided optimization may reduce important differences between LEM-I and mMKM biological dose estimates for large pelvic osteosarcomas, potentially improving the consistency of carbon-ion treatment planning. However, the modest loss of robustness, frequent requirement for additional beams, and absence of clinical validation mean that this strategy remains a planning optimization concept rather than a proven method of improving tumor control.

SOURCE

Physics and Imaging in Radiation Oncology

Browse more research Suggest a correction