BACKGROUND AND PURPOSE: To develop and validate an adaptive intervention strategy for radiotherapy of head-and-neck cancer that accounts for systematic deformations by modifying the planning-CT (pCT) to the average misalignments in daily cone beam CT (CBCT) measured with deformable registration (DR). METHODS AND MATERIALS: Daily CBCT scans (808 scans) for 25 patients were retrospectively registered to the pCT with B-spline DR. The average deformation vector field () was used to deform the pCT for adaptive intervention. Two strategies were simulated: single intervention after 10 fractions and weekly intervention with an from the previous week. The model was geometrically validated with the residual misalignment of anatomical landmarks both on bony-anatomy (BA; automatically generated) and soft-tissue (ST; manually identified). RESULTS: Systematic deformations were 2.5/3.4mm vector length (BA/ST). Single intervention reduced deformations to 1.5/2.7 mm (BA/ST). Weekly intervention resulted in 1.0/2.2mm (BA/ST) and accounted better for progressive changes. 15 patients had average systematic deformations >2mm (BA): reductions were 1.1/1.9 mm (single/weekly BA). ST improvements were underestimated due to observer and registration variability. CONCLUSIONS: Adaptive intervention with a pCT modified to the average anatomy during treatment successfully reduces systematic deformations. The improved accuracy could possibly be exploited in margin reduction and/or dose escalation.