Author: Fre'Etta Brooks, Catharine H Clark, Patricia Diez, Peter B. Greer, Hideaki Hirashima, Rebecca M. Howell, Mohammad Hussein, Stephen F. Kry, Joerg Lehmann, Jessica Lye, Mitsuhiro Nakamura, Christopher L. Nelson, Rushil Patel, Christine Peterson, Julianne M. Pollard-Larkin π¨βπ¬
Affiliation: Kyoto University, The University of Texas MD Anderson Cancer Center, Radiotherapy Trials Quality Assurance Group, Department of Radiation Physics, The University of Texas MD Anderson Cancer Center, Department of Medical Physics & Biomedical Engineering, University College London, Metrology for Medical Physics, National Physical Laboratory, Austin Health, University of Washington Medical Center, University of Newcastle π
Purpose: To assess the consistency of end-to-end dosimetry audits currently used by different international clinical trial QA organizations with the goal of harmonizing audits and reducing audit overlap in multinational trials while maintaining quality.
Methods: Established international dosimetry audits for IMRT from six Global Quality Assurance of Radiation Therapy Clinical Trials Harmonization Group (GHG) member organizations were included in this work. Sixteen previously used reference plans based on established quality assurance geometries were employed. Realistic plan modifications, developed from reported variations in clinical practice, were introduced into nine copies of each plan, generating 144 modified plans. These plans were categorized as acceptable (should pass) or unacceptable (should fail) using different assessment tolerances on three evaluation metrics: CTVmean, CTVD95, and OARD0.03cc.
The impact for each evaluation metric was highly comparable, regardless of tolerance. The measured CTVmean dose in the error-free reference plans was compared to the calculated dose of the modified plans. Using the audit systemsβ established tolerances, outcomes were expressed as pass/fail, which were then compared to the βshould pass/should failβ benchmark to determine sensitivity and specificity of each system. Furthermore, an optimal CTVmean tolerance was determined for each system to achieve a common sensitivity across all audit systems.
Results: All audit systems reliably identified failing plans (sensitivity 0.92β1.0) with a plan assessment tolerance of 5% ΞCTVmean. Specificity demonstrated greater range (0.49-0.91). Adjusting the tolerance revealed that each audit system had been tuned to detect different error thresholds (3.3%β5.7%). Changing audit system tolerances increased specificity from 0.49 to at least 0.60 (0.75 in most cases), while preserving system sensitivity.
Conclusion: The established comparable high sensitivity of the investigated audit systems could allow harmonizing of dosimetry audits for clinical trials on an international scale based on sensitivity alone. Further work harmonizing their specificity would help streamline credentialing for international clinical trials.