Regional Lymph Node Delineation variability and its Dosimetric Impact in Breast Cancer Radiotherapy

Aims

To quantify the interobserver variability of regional lymph node delineation for breast cancer radiotherapy (RT) and establish whether a relationship exists between contouring variations and dosimetry using the FAST-Forward (FF) pre-trial RT quality assurance (QA) benchmark cases.

Materials and methods

As part of the pre-trial RT QA, local site principal investigators (PIs) were asked to complete a single outlining QA benchmark case involving the delineation of axillary lymph node clinical target volumes (LNCTVs) levels 1–4. These contours were evaluated for concordance against an expert defined consensus gold standard (GS) volume using various conformity indices (CIs): discordance index (DI), geographical miss index (GMI), Jaccard index (JCI), mean distance to conformity (MDC) for both over- and under- contouring. Descriptive statistics including interquartile range (IQR) was used to evaluate interobserver variation. Wilcoxon signed-rank tests were used to establish if there were any statistically significant differences in the dosimetric parameters between plans conforming to GS volume and the volumes from the individual PI.

Results

Pre-trial outlining QA benchmark cases from 29/33 PIs were assessed. The median submitted LNCTV volume was 131.4 cc (IQR: 112.4 – 145.3) compared with the GS volume of 105.46 cc. For conformity indices, the median DI was 0.37 (IQR: 0.31 – 0.40), the median GMI was 0.21 (IQR: 0.13 – 0.28), the median JCI was 0.53 (IQR: 0.49 – 0.56), MDC_under was -0.43 (IQR: -0.64 - -0.32) and MDC_over was 0.46 (IQR: 0.43 – 0.53). A dosimetric analysis showed all plans met the mandatory planning dose constraints but not the optimal objectives for target volumes as required in the trial protocol. Statistically significant differences were found in 7/13 organs at risk dosimetric parameters between plans conformed to individual PI volumes and the GS volume.

Conclusion

Analysis of the FF pre-trial QA outlining benchmark case highlights the interobserver variation that exists in axillary nodal CTV (levels 1–4) delineation. Conformity indices demonstrated moderate agreement with a median Jaccard conformity index of 0.53, with both under- and over-contouring observed. All QA submissions achieved the mandatory planning dose constraints but not all optimal dose objectives of the FF trial despite the interobserver variation in target volume contouring.