Purpose: To investigate sources of systemic and random errors in in-vivo OSLD measurements during regular clinical use, by comparing adjacently placed pairs of OSLDs in total body irradiation (TBI) and total skin electron therapy (TSET) under the assumption of equal dose delivered to both OSLDs.
Methods: Directly adjacent pairs of NanoDot OSLDs were placed on patients with a medical tape prior to TBI and TSET, with or without 1.5cm of bolus respectively, at each anatomical site monitored. Each OSLD was read 5 times (standard error of mean σ=0.2%). Resulting raw readings were divided by vendor-provided sensitivity (mean 0.85±0.02) to obtain corrected readings (CR). Used OSLDs were bleached between each use until 1500cGy was accumulated. OSLD database was queried January 2019 to November 2020 to obtain 622 paired samples possibly used for 200cGy prescription TBI and TSET. Pairs of 101 TBI and 110 TSET measurements were verified to be present within clinical records. Relative difference in corrected counts (∆CRᵣ = pair-difference/mean) was fitted linearly versus sensitivity, bleaching history, cumulative dose, and age differences between paired OSLDs.
Results: OSLDs were bleached 8 times (0~28), accumulated 942cGy (60~3015), and were used for 238days (0~819) on average before termination. ∆CRᵣ was normally distributed (σ=±5.9%). ∆CRᵣ for TBI and TSET subsets deviated σ=±4.7% and ±6.3% respectively, despite similar predictor distributions. OSLD with higher sensitivity yielded -0.68% less per +0.01 than the other (p=4e-17). Partial multiple linear models taking sensitivity into account found little contributions from bleaching (p=0.89) or cumulative dose (p=0.047). TBI and TSET samples yielded RMSE=±4.8% after sensitivity correction.
Conclusion: Our findings suggest vendor-recommended reading correction via sensitivity division results in a small but significant systemic error. OSLD sensitivities should be considered during reading-to-dose calibration and clinical measurements. Bleaching and cumulative dose likely minimally affect measurements. Random errors were higher for TSET than TBI.
Funding Support, Disclosures, and Conflict of Interest: None of the authors received funding for this abstract.