KEY POINTS
- This two-year single-centre implementation study included 30 patients with cervical cancer and 744 CBCT-guided online adaptive fractions delivered on the Ethos platform between 2023 and 2025.
- Treatment generally followed EMBRACE II with 45 Gy in 25 fractions; ten patients received nodal SIB to 55 or 57.5 Gy. Implementation expanded progressively from simpler pelvic cases to nodal boosts, para-aortic treatment and two-isocentre workflows.
- After structured training and supervised sessions, treatment transitioned to an RTT-led workflow using a traffic-light escalation protocol with radiation oncologist, physicist and technical-physician support available when required.
- Median total in-room session duration was 33.0 minutes. Importantly, the first five supervised fractions and subsequent RTT-led fractions took essentially the same time: 32.5 versus 32.0 minutes.
- Contour review was the main workflow bottleneck. Influencer and target review required a median 10 minutes and correlated strongly with total treatment time (ρ=0.73). AI-generated contours required manual correction in every patient.
- Daily adaptation maintained CTV D98 ≥95% across all fractions, whereas recalculating the original scheduled plan on daily anatomy produced substantially more variable coverage and underdosage in several patients.
- Intrafraction motion remained relevant despite adaptation. The uterus extended outside the PTV on pretreatment verification CBCT in 8.1% of sessions and on post-treatment CBCT in 11.2%, generally involving small fundal regions. Bladder emptying, rectal venting and manual position adjustments were sometimes required.
- This was a 30-patient implementation study rather than a clinical-outcome comparison. More complex two-isocentre cases were underrepresented, intrafraction motion was assessed mainly qualitatively, and concurrent AI/software upgrades prevent clean separation of learning and technology effects.
CLINICAL TAKEAWAY
Daily CBCT-guided adaptive RT for cervical cancer can be operationalized as an RTT-led workflow without sacrificing target coverage, provided that structured training and rapid expert backup are available. The biggest remaining efficiency problem is not optimization or beam delivery but manual contour review, making better AI segmentation a key target for future workflow improvement.
SOURCE
Technical Innovations & Patient Support in Radiation Oncology