Dose and clinical impact of delineation variability in brain metastases treated with stereotactic radiotherapy

Accurate delineation is critical for stereotactic radiotherapy (SRT) of brain metastases. We evaluated how inter-reader delineation variability affects target dose coverage and local relapse-free survival (LRFS).

We retrospectively analyzed 130 patients with 198 brain metastases treated with SRT. Four observers independently delineated gross tumor volumes (GTVs) on pre-treatment MRI; each GTV was expanded by 2 mm to create a planning target volume (PTV). Inter-observer variability was quantified using overlap- and distance-based similarity metrics. Dose coverage (V99%) for each observer-derived GTV/PTV was extracted from the delivered clinical dose distribution (the treatment plan created and delivered in routine practice), without replanning on alternative contours. Associations between variability, dose coverage and LRFS were evaluated using ROC analyses and Cox models.

Similarity metrics were significantly correlated with GTV volume. Higher inter-reader variability predicted lower GTV and PTV dose coverage (best AUC for Hausdorff distance: 0.92 for GTV coverage). However, neither reduced GTV/PTV dose coverage nor higher variability (HD or mean distance to agreement) was significantly associated with LRFS in univariate or multivariate analyses.

Inter-reader variability can substantially modify apparent GTV/PTV dose coverage when the delivered plan is evaluated on alternative contours. In this cohort, variability did not translate into a significant reduction in LRFS, likely reflecting prescription to the PTV and the use of a 2‑mm margin. Standardized contouring workflows and validated AI-assisted tools may further reduce variability and improve treatment robustness.