Alternative Criteria to Address Major Flaws of Common Evaluation Metrics for Dose Distributions Through...

Purpose This study aimed to reveal major flaws in common evaluation metrics for dose distributions of stereotactic radiosurgery (SRS) for brain metastases (BMs) and to present alternative methods that are more objective and more relevant to treatment outcomes through characterization of dose distributions with volumetric-modulated arcs (VMA) by a simple optimization method using the High Definition Dynamic Radiosurgery (HDRS) platform (Elekta AB, Stockholm, Sweden). Materials and methods The subjects were 37 lesions with the gross tumor volume (GTV) ranging from 0.04 cc to 48.09 cc (median: 7.32 cc), which were treated as solitary lesions. The main constituents of the HDRS are the 5-mm leaf-width, 160-leaf collimator Agility^® (Elekta AB) and the planning system Monaco^® (Elekta AB). The VMA-based SRS plan for each GTV was optimized by applying only three cost functions with prioritizing the steepness of dose falloff outside the GTV boundary. The prescription dose was uniformly assigned to the GTV D _{V - 0.01 cc}, minimum dose of GTV minus 0.01 cc (D _>95%), for GTV >0.20 cc and the D _95% for the GTV ≤0.20 cc. An average gradient distance (AGD) was devised as a more quantitative evaluation measure of the dose gradient, which was defined as the difference between the radii of spheres equivalent to irradiated isodose volumes (IIDVs) of 50% and 100% of a reference dose. The AGDs were calculated from the isodose surfaces (IDSs) of the prescription dose and the GTV D _{eIIV (equivalent IIDV)}, minimum dose of IIDV equivalent to GTV. Results The GTV dose heterogeneity and the dose 2 mm inside the GTV boundary were significantly correlated with the GTV, being the most inhomogeneous and highest, respectively, at the GTV of 1.71 cc. The dose 2 mm outside the GTV boundary and the AGDs were significantly correlated with the GTV, being steepest and shortest, respectively, at the GTV of 0.72 cc. All the minimum volumes outside the GTV, receiving the GTV D _eIIV, the prescription dose, and 50% of the prescription dose, increased significantly with increasing the GTV. However, both the conformity and gradient indices (CIs, GIs) showed significantly better values with increasing the GTV, regardless of the reference doses. The CIs showed significantly better values as the GTV coverage by the reference IDS decreased: the D _{V - 0.01 cc} (median: D _99.88%) to the D _98%, whereas the GIs showed significantly better values as the GTV coverage by the reference IDS increased and further became over-covered. The interlesion differences in the CIs and the GIs varied substantially depending on the definitions. Conclusions In the VMA-based SRS with the HDRS for single BMs, the dose gradients outside and inside the GTV boundary were steepest at the GTVs of 0.72 cc and 1.71 cc, respectively, and correlated significantly with the GTV, peaking at the threshold volumes. As the GTV increased, both the CIs and GIs showed significantly better values, leading to less clinical significance of the differences in the values. The GIs also showed significantly better values as the target coverage by a reference IDS increased. The CI values ​​were significantly affected by the target coverage and the definition. In evaluating dose distributions of SRS for BMs, the absolute values such as the IIDVs should be prioritized over the common indices based on relative ratios.