2957 - Setup Uncertainty in Pediatric Craniospinal Irradiation Proton Therapy: A Prospective Study

Purpose/Objective(s):

Craniospinal irradiation (CSI) is a key modality in pediatric oncology and a major indication for proton radiotherapy. To guide CSI planning margins, a prospective study was conducted to quantify setup uncertainty and intrafraction motion in a large pediatric cohort in the setting of daily volumetric imaging and robotic patient positioning.

Materials/Methods:

From 2020–2024, 100 patients (2-22 years; median age 10; 80% anesthesia) receiving proton CSI were prospectively enrolled. Daily pre-correction CBCT ensured alignment, and post-correction and post-treatment CBCTs assessed residual and intrafraction setup uncertainties, respectively. Euclidian displacement was estimated using a 6-DoF variance-based stochastic model with Monte Carlo propagation of translational and rotational uncertainties. Stratified nonparametric bootstrap resampling yielded the 95th-percentile displacement (q95). q95 and two-sided unstudentized bootstrap p-values were evaluated at 150 mm from isocenter, a conservative surrogate for cranial (cribriform) and spinal (superior/inferior) distances. Site-specific optimization margins were time-weighted using imaging-to-treatment latency at each isocenter to integrate post-correction and post-treatment deviations into final estimates.

Results:

Across all isocenters, q95 decreased from 9.2 mm on pre-treatment CBCT to 2.2 mm after correction (p < 0.001), reflecting the benefit of daily volumetric imaging. Intrafraction drift increased q95 to 3.0 mm on post-treatment CBCT (p < 0.001). By isocenter, post-treatment q95 values were (? = post-treatment - post-correction): brain 2.9 mm (? = 0.9 mm; p < 0.001), upper spine 2.9 mm (? = 0.7 mm; p = 0.02), lower spine 3.1 mm (? = 0.9 mm; p = 0.01), and single-isocenter spine 2.5 mm (? = 0.5 mm; p = 0.15).

Anesthesia reduced intrafraction motion: post-treatment q95 was 2.8 mm vs 3.7 mm without anesthesia (p = 0.003). The largest effects occurred in the brain (? = 0.8 mm; p = 0.03), with no significant difference in the upper spine (? = 0.6 mm; p = 0.20) and the lower spine (? = 0.5 mm; p = 0.50).

After applying a 1 mm isocenter setup tolerance and temporal weighting, final setup offsets were 3.7 mm (brain), 3.6 mm (upper spine), 3.6 mm (lower spine), and 3.2 mm (single-isocenter spine). Offsets were larger in awake patients: brain 3.5 ? 4.2 mm, upper spine 3.5 ? 3.9 mm, and lower spine 3.5 ? 3.8 mm (anesthesia ? no anesthesia). These results support a 3.5 mm cranial margin with anesthesia and suggest larger margins for awake patients.

Conclusion:

Despite thorough patient positioning and verification, clinically significant intrafraction drift persists in pediatric patients undergoing CSI. A 4 mm optimization margin for patient setup is adequate as a treatment planning parameter and may serve as a practical guideline for similar clinical scenarios.