S2340
Physics - Quality assurance and auditing
ESTRO 2026
(major; three-character; subsite). Mapping: C15 → Chest; C69/C73 → Head & Neck; C77 followed the recorded primary. TL/AL were defined per TG-218. Differences across ICD-10 groups and, within the same three-character code, among PORT/Def-RT/Def-CCRT were tested with a permutation-based Kruskal–Wallis test (n ≥ 10/group). Results: At 2%/2 mm, within-major variability was significant for Chest and Others (both p ≈ 0.001), while Head & Neck and Pelvis were not significant (Fig. 1). Representative TL contrasts were clinically meaningful: within Chest, C50 vs C15 showed ≈ +5–6% in TL (and ≈ +6% in AL); within Others, C79 vs C77 showed ≈ +4– 5% in TL/AL. Chest with TL/AL overlays is shown (Fig. 2). Subsite analysis provided additional separation when group sizes were adequate (e.g., within C79, C79.3 vs C79.5 yielded ≈ +3.3% in TL). Selected Head & Neck and Pelvis codes showed signed contrasts typically around ±3–4% (TL) and ±3% (AL). At 3%/3 mm, contrasts contracted due to ceiling effects; 2%/2 mm offered superior discrimination. Within the same ICD- 10 code, no consistent TL/AL differences were detected among PORT/Def-RT/Def-CCRT; however, at 2%/2 mm several codes exhibited small downward shifts of the median and lower tail for Def-CCRT versus Def-RT, indicating greater susceptibility to low outliers despite similar averages.
and Monitor Chamber Output Constancy exhibited particularly strong stability. A small subset of parameters – including J1mm and J7mm Peak-to-Peak Fluence Variation and Maximum Absolute Leaf Projection Error – showed that while lower Cpmk values of approximately 0.3 were observed, Cpk results remained well above 1, suggesting that the applied target values may have been unrealistically strict, although the process still performs within specification limits with low inherent variability. The calculated LCL and UCL values provide valuable reference thresholds for ongoing monitoring, allowing early identification of deviations or emerging process drifts. Conclusion: The combination of specification limits (defining pass/fail criteria) and control limits (defining process behavior) established a comprehensive framework for quality management. This dual-limit approach enables early detection of performance drifts, facilitating timely corrective action before clinical impact and ensuring long-term stability of the TomoTherapy delivery system. Keywords: TomoTherapy, Quality Assurance, SPC Digital Poster 1202 ICD-10 stratified integration of clinical data and patient-specific quality assurance: multi- institutional analysis for improved prediction Taichi Wada 1,2 , Kazunori Miyaura 1 1 Graduate School of Health Sciences, Division of Medical Technology, Showa Medical University, Tokyo, Japan. 2 Department of Radiology, Kanto Rosai Hospital, Japan Organization of Occupational Health and Safety, Kawasaki, Japan Purpose/Objective: Patient-specific quality assurance (PSQA) is essential for safe IMRT, yet clinical variables are often siloed. Using a four-hospital database where clinical and PSQA data are jointly recorded and analysed, we tested whether ICD-10 stratification (major; three- character; subsite = three-character plus one decimal) explains variability in gamma ( γ ) pass rates and informs tolerance/action levels (TL/AL) per AAPM TG- 218. We also examined modality effects (PORT, Def-RT, Def-CCRT), focusing on lower-tail shifts. Material/Methods: We analysed γ pass rates for 2%/2 mm and 3%/3 mm (threshold 10%). Cases were stratified by ICD-10
Figure 1. ICD-10 three-character variability at γ 2%/2 mm (TH=10%; n ≥ 10). Greater dispersion in Chest/Others supports ICD-based stratification.
Figure 2. Chest three-character at γ 2%/2 mm (TH=10%; n ≥ 10). Dashed/dotted lines show Chest TL/AL (Kruskal–Wallis p ≈ 0.001). Conclusion: ICD-10–based stratification—especially at the subsite level—explains PSQA variability masked by coarse site
Made with FlippingBook - Share PDF online