Boost delineation in breast radiation therapy: Isotropic versus anisotropic margin expansion

The purpose of this article is to compare isotropic and anisotropic margin expansion with regard to the size of the clinical target volume boost (CTV_boost) and the interobserver variability (IOV).

Lumpectomy cavities marked with 3 or more surgical clips were delineated by 6 radiation oncologists who specialized in breast radiation therapy. CTV_{boost anisotropic} was created by manually expanding the tumor bed with an anisotropic margin of 15 mm (20 mm in case of extensive intraductal component) minus the surgical free margins in 6 directions (anteroposterior, craniocaudal, and superoinferior). For the CTV_{boost isotropic}, the tumor bed was enlarged with an isotropic margin of 15 mm (20 mm in case of extensive intraductal component) minus the minimal surgical free margin. The volumes of the delineated CTV_boost (cm³) were measured. To assess the IOV, the Jaccard index (JI), defined as the intersection divided by the size of the union of the sample sets, was used (ideal value = 1). The JI was calculated for each case and each observer pair. Linear mixed models were used for all analyses.

A total of 444 delineated tumor beds were evaluated. The mean volume of the CTV_boost almost doubled by expanding the tumor bed with an isotropic margin compared with anisotropic margins (CTV_{boost isotropic} 94 mL [12.5-331.0] vs CTV_{boost anisotropic} 50 mL [3.2-332.7]; P = .0006). The IOV, assessed by the JI, significantly decreased by using isotropic versus anisotropic margin expansion (JI_{CTV boost isotropic} 0.73 [0.02-0.92] vs JI_{CTV boost anisotropic} 0.51 [0.0-0.8]; P< .0001). Because of the known positive correlation of the IOV and larger volumes, we corrected for CTV_boost volumes. With this correction, the difference in IOV remains highly significant (P < .0001) in favor of isotropic margin expansion.

The use of anisotropic margin expansion from tumorbed to CTV_{boost isotropic} significantly reduced the volume of the delineated CTV_boost with a factor of 1.9 compared with isotropic margin expansion, but it substantially increased the interobserver variability.