KEY POINTS
- The retrospective cohort included 50 patients receiving curative head and neck VMAT with simultaneous integrated boost and 271 weekly cone-beam computed tomography scans. Primary sites included the larynx and nasopharynx in 20 patients each, with smaller hypopharyngeal, nasal-cavity and oropharyngeal groups.
- Patients were immobilized with three-point thermoplastic masks and treated with a 3 mm clinical target volume-to-planning target volume margin. Daily planar imaging was used for setup correction, while cone-beam computed tomography was acquired at the first fraction and approximately weekly thereafter.
- Population systematic errors were 2.0 mm superior-inferior, 2.0 mm left-right and 1.7 mm anterior-posterior. Random errors were larger at 3.4, 2.9 and 3.8 mm, respectively.
- Applying the van Herk formula produced calculated margins of 7.4 mm superior-inferior, 7.0 mm left-right and 6.9 mm anterior-posterior, substantially larger than the institutional 3 mm margin.
- The population systematic shift was simulated in 46 recalculable plans without reoptimization. A planning target volume D95% reduction of at least 3% occurred in 18 patients (39.1%), and 7 patients (15.2%) fell below 95% of prescription in at least one planning target volume. Clinical target volume coverage remained acceptable on review.
- In the organ-at-risk cohort, the simulated shift increased brainstem maximum dose by 1.23 Gy (p = 0.004) and left parotid mean dose by 3.64 Gy (p < 0.001). Changes in spinal-cord maximum dose and right-parotid mean dose were not significant.
- Mean early three-dimensional shifts were similar in flagged and non-flagged patients. Receiver operating characteristic areas under the curve were 0.49 using the first three scans and 0.53 using the first four, with sensitivity of 61–67% and specificity of 50–54%.
CLINICAL TAKEAWAY
Routine early translational shifts should not be used alone to trigger adaptive replanning because their individual predictive performance was essentially random. The study still demonstrates sensitivity of highly conformal plans to systematic offsets, but the simulated rigid-shift method omitted anatomical deformation, daily correction and accumulated delivered dose; composite triggers incorporating weight loss and tumour or parotid shrinkage are more defensible.