Head and neck experts show major elective target variation despite identical gross tumor volumes

Across 32 experts, agreement fell to Dice 0.45 for elective nasopharyngeal targets despite identical starting GTV and OAR contours.

KEY POINTS

  • This international study, conducted through an ASTRO International Committee initiative, evaluated interobserver variation in target delineation and treatment planning for two standardized head and neck cancer scenarios: HPV-associated T1N2 oropharyngeal carcinoma and EBV-associated T1N2 nasopharyngeal carcinoma. Thirty-two international experts submitted target contours, with 25 providing complete, deliverable IMRT/VMAT plans. Participating centers represented 11 countries, and reported clinical experience ranged from five to 30 years.
  • All participants received identical gross tumor volume and organ-at-risk contours, allowing the study to isolate differences arising from clinical target-volume expansion, elective nodal coverage, and planning priorities. Experts generated high-dose CTV1, intermediate-dose CTV2, optional elective CTV3, and corresponding PTVs according to their institutional practices, without additional contouring instructions. Prescriptions generally followed 70/(63–60)/(57–54) Gy in 33–35 fractions or 64/60/50 Gy in 25 fractions. A 50% expert iso-agreement contour served as the consensus reference rather than an independently validated anatomical ground truth.
  • Agreement on high-dose target volumes was moderate but far from uniform. Mean Dice similarity coefficient (DSC) for CTV1 was 0.72 ± 0.19 in oropharyngeal cancer and 0.73 ± 0.19 in nasopharyngeal cancer, despite the identical starting GTV. The consensus high-dose volume approximated a 5-mm GTV expansion, but some experts used untrimmed geometric margins, others adjusted margins around anatomical barriers, and some included additional anatomical compartments. Nasopharyngeal approaches included whole-nasopharynx coverage and, in a minority of cases, no expansion around the supplied GTV.
  • Variation increased substantially for intermediate- and elective-dose targets. Mean CTV2 DSC declined to 0.66 ± 0.17 for oropharyngeal cancer and 0.67 ± 0.17 for nasopharyngeal cancer. The greatest disagreement occurred in the elective low-dose nasopharyngeal PTV3, with DSC 0.45 ± 0.21. Relative to the consensus, extra PTV3 volume had a median of 119.5 cm³ and mean of 245.5 cm³ (SD 310.5 cm³), with center-of-mass displacement reaching approximately 17 mm laterally and 26 mm craniocaudally. Disagreements primarily involved elective lower-neck and retropharyngeal nodal coverage.
  • Target dose coverage was relatively consistent despite large differences in contour geometry. Among 25 deliverable plans, mean high-dose PTV1 D95% was 69.39 ± 1.42 Gy for oropharyngeal cancer and 68.74 ± 1.99 Gy for nasopharyngeal cancer. Intermediate- and elective-dose PTV coverage was also generally close to prescription. These findings indicate that participating planning teams largely achieved the coverage objectives for the targets they had individually defined, even when the underlying treatment volumes differed markedly.
  • OAR exposure varied substantially between plans considered clinically acceptable. In the nasopharyngeal case, mean optic chiasm dose metric was 15.66 Gy (SD 12.45) across expert plans versus 9.57 Gy in the consensus plan. Corresponding values were 32.44 versus 18.68 Gy for the left optic nerve and 48.20 versus 41.74 Gy for the brainstem. However, the consensus plan delivered more dose to the supraglottic larynx (D66 40.97 versus 34.94 Gy). In oropharyngeal cancer, the supraglottic larynx was also among the most variable structures (mean D66 33.74 Gy; SD 8.45), while parotid mean doses were comparatively consistent.
  • The study identifies variation but does not establish which expert strategy produces better clinical outcomes. Only two standardized cases were analyzed, the consensus atlas was derived from the participating experts themselves, and no recurrence, dysphagia, visual toxicity, or other clinical outcomes were measured. The authors propose structured peer review and potentially AI-assisted contouring or planning as strategies to reduce disagreement, but neither intervention was tested. Importantly, greater conformity to a consensus contour is not necessarily equivalent to superior anatomical or oncological accuracy.

CLINICAL TAKEAWAY

Even when radiation oncologists agree on the gross extent of a head and neck tumor, substantial differences remain in clinical target-volume expansion, elective nodal coverage, and normal-tissue dose prioritization. The greatest variation occurred in elective treatment volumes, highlighting an important opportunity for standardized contouring guidance and structured peer review. However, this geometric and dosimetric study cannot determine which planning choices provide the best balance between tumor control and toxicity.

SOURCE

Practical Radiation Oncology

Browse more research Suggest a correction