KEY POINTS
- This retrospective single-center analysis included 74 patients with oropharyngeal cancer treated with definitive sequential-phase IMRT and planned offline adaptive radiotherapy. A resimulation CT around week 3 was used to quantify the actual dosimetric benefit obtained from adaptation.
- Adaptive benefit was defined from a six-dimensional outcome incorporating change in xerostomia normal-tissue complication probability plus mean-dose changes to the bilateral parotid glands, bilateral submandibular glands and oral cavity. Unsupervised clustering classified 31 patients as high-benefit and 43 as low-benefit.
- For prediction before resimulation, the investigators recalculated the original treatment plan on the week-3 CBCT after deformable registration, field-of-view completion using planning CT data and intensity normalization. The resulting mean-dose changes to five normal tissues were entered into a multivariable logistic model.
- The CBCT dose-based model achieved a leave-one-out cross-validated AUC of 0.78, compared with only 0.54 for a model based on normal-tissue volume changes. This suggests that actual dose drift contains substantially more useful information for selecting replanning candidates than anatomical shrinkage alone.
- CBCT-measured dose changes showed significant monotonic associations with later adaptive benefit for every evaluated normal tissue: Spearman correlations ranged from −0.76 for the ipsilateral parotid to −0.43, with all trend-test P values <.01. Volume changes showed no significant corresponding associations, with all |ρ|≤0.25 and P≥.10.
- The high-benefit and low-benefit groups differed most strongly in ipsilateral parotid benefit, by approximately 27 percentage points; differences were about 16 points for oral cavity, 15 for contralateral parotid and 14 for ipsilateral submandibular gland. The ipsilateral parotid was the strongest standalone predictor.
- The model remains a decision-support proof of concept rather than a validated clinical trigger. It was developed in only 74 patients at one institution, lacks external validation and predicts a constructed dosimetric/NTCP benefit label rather than observed xerostomia or tumor-control outcomes; its performance also depends on CBCT calibration, deformable registration and dose-recalculation accuracy.
CLINICAL TAKEAWAY
For head and neck adaptive radiotherapy, dose change may be a better trigger than anatomy change: week-3 CBCT recalculation distinguished patients likely to gain dosimetrically from replanning far better than simple volume metrics. The approach could make selective ART more efficient, but it needs multicenter validation against meaningful clinical outcomes before implementation as a treatment trigger.