KEY POINTS
- Within the multicentre STRONG II phase II trial, six centres completed a common benchmark case for unresectable perihilar cholangiocarcinoma, first delineating targets and OARs and then generating SBRT plans on reference contours.
- Initial agreement was good for most OARs, with median Dice similarity coefficients above 0.8, but substantially worse for the target, central biliary tract, and cystic duct: 0.57, 0.63, and 0.00, respectively.
- After multidisciplinary review and radiologist-informed updates to the contouring guidance, median Dice improved to 0.74 for the target, 0.75 for the central biliary tract, and 0.42 for the cystic duct; the cystic-duct improvement was statistically significant (p=0.04).
- Target boundary agreement also improved: mean distance to agreement decreased from 5.2 to 2.5 mm and HD95 from 17.1 to 7.4 mm (both p=0.05). For the cystic duct, the corresponding reductions were 12.0 to 2.5 mm and 20.0 to 5.7 mm (both p=0.04).
- No major protocol violations were identified in treatment planning, but 5 of 6 centres had a minor PTV-undercoverage violation because target coverage had to be compromised to respect stomach and duodenal constraints. Median PTV V60 Gy was 87.7% and V67.5 Gy 41.5%.
- The duodenum was the principal dose-limiting structure: median Dmax was 55.5 Gy against a 57.0-Gy constraint, while median V41 Gy was 3.3 cm³ against a 5.0-cm³ constraint.
CLINICAL TAKEAWAY
Iterative peer review and imaging-based guidance can meaningfully reduce contouring variation in perihilar cholangiocarcinoma SBRT. However, persistent target variability and frequent OAR-driven undercoverage show why rigorous trial QA remains particularly important for this anatomically difficult disease.