INTRODUCTION: The aims of the study were to evaluate interobserver variability in contouring the brachial plexus (BP) using the Radiation Therapy Oncology Group (RTOG)-approved protocol and to analyse BP dosimetries. METHODS: Seven outliners independently contoured the BPs of 15 consecutive patients. Interobserver variability was reviewed qualitatively (visually by using planning axial computed-tomography images and anteroposterior digitally reconstructed radiographs) and quantitatively (by volumetric and statistical analyses). Dose-volume histograms of BPs were calculated and compared. RESULTS: We found significant interobserver variability among outliners in both qualitative and quantitative analyses. These were most pronounced for the T1 nerve roots on visual inspection and for the BP volume on statistical analysis. The BP volumes were smaller than those described in the RTOG atlas paper, with a mean volume of 20.8 cc (range 11-40.7 cc) compared with 33 ± 4 cc (25.1-39.4 cc). The average values of mean dose, maximum dose, V60Gy, V66Gy and V70Gy for patients treated with conventional radiotherapy and IMRT were 42.2 Gy versus 44.8 Gy, 64.5 Gy versus 68.5 Gy, 6.1% versus 7.6%, 2.9% versus 2.4% and 0.6% versus 0.3%, respectively. CONCLUSION: This is the first independent external evaluation of the published protocol. We have identified several issues, including significant interobserver variation. Although radiation oncologists should contour BPs to avoid dose dumping, especially when using IMRT, the RTOG atlas should be used with caution. Because BPs are largely radiologically occult on CT, we propose the term brachial-plexus regions (BPRs) to represent regions where BPs are likely to be present. Consequently, BPRs should in principle be contoured generously.