Evaluation of Concordance between Radiation Therapists and Medical Physicists in MR‐Guided Online Adaptive Planning

Abstract Purpose Magnetic resonance (MR) ‐guided online adaptive radiotherapy requires skill in online adaptive treatment planning. The aim of this study was to demonstrate that radiation therapists, with appropriate training, can consistently produce high‐quality adaptive plans concordant with those created by medical physicists for treatment on a 1.5 Tesla MR‐Linac (MRL). Methods We evaluated plans generated by two experienced MRL radiation therapists and two MR‐Linac medical physicists using an adapt‐to‐position (ATP) workflow to treat 10 patients with head and neck cancer; this workflow includes an iso‐center shift and plan re‐optimization to account for setup discrepancy between simulation and treatment. For each adaptive session, CT–MR images were fused and isocenter translational shifts recorded. Differences among the three roles (Therapist, Physicist 1, Physicist 2) were assessed by repeated‐measures analysis of variance (RM‐ANOVA) and Bonferroni‐corrected post‐hoc paired t tests, reported as secondary, exploratory analyses. Agreement on shifts and dosimetric endpoints was quantified with Lin's concordance correlation coefficient (CCC), with bias‐corrected and accelerated bootstrap 95% confidence intervals (CIs), as the primary analysis. Results No significant differences were observed among roles for X‐ or Y‐axis shifts ( p = 0.804 and p = 0.950). A statistically significant difference was observed for Z‐axis shifts ( p = 0.005), driven by Therapist versus Physicist 1 (Bonferroni‐adjusted p = 0.015; mean paired difference 0.10 cm, 95% CI 0.04–0.17 cm), a systematic difference of approximately 1 mm that is small relative to clinical planning margins. CCC point estimates were 0.866–0.970 for translational shifts and 0.972–0.997 for target coverage and maximum dose, with no point estimate in the poor‐agreement range (CCC < 0.80). Bootstrap 95% CIs were wide for the shifts, as expected at n = 10, but therapist–physicist agreement was comparable to physicist–physicist agreement for every endpoint, and CI lower bounds for the dosimetric endpoints were ≥0.75 (maximum dose ≥0.97). Conclusion Our results showed good concordance between adaptive plans created by radiation therapists and medical physicists in the ATP workflow for head and neck cancer, suggesting that appropriately trained therapists can produce high‐quality adaptive plans in this specific context. These conclusions are limited to the ATP workflow and to head and neck cancer; extension to other sites or to the adapt‐to‐shape (ATS) workflow requires independent investigation.