Statistical analysis of IMRT dosimetry quality assurance measurements for local delivery guideline

Purpose: To establish our institutional guideline for IMRT delivery, we statistically evaluated the results of dosimetry quality assurance (DQA) measurements and derived local confidence limits using the concept confidence limit of vertical bar mean vertical bar + 1.96 sigma. Materials and methods: From June 2006 to March 2009, 206 patients with head and neck cancer, prostate cancer, liver cancer, or brain tumor were treated using LINAC-based IMRT technique. In order to determine site specific DQA tolerances at a later stage, a hybrid plan with the same fluence maps as in the treatment plan was generated on CT images of a cylindrical phantom of acryl. Points of measurement using a 0.125 cm(3) ion-chamber were typically located in the region of high and uniform doses. The planar dose distributions perpendicular to the central axis were measured by using a diode array in solid water with all fields delivered, and assessed using gamma criteria of 3%/3 mm. The mean values and standard deviations were used to develop the local confidence and tolerance limits. The dose differences and gamma pass rates for the different treatment sites were also evaluated in terms of total monitor uints (MU), MU/cGy, and the number of PTV's pieces. Results: The mean values and standard deviations of ion-chamber dosimetry differences between calculated and measured doses were -1.6 +/- 1.2% for H&N cancer, -0.4 +/- 1.2% for prostate and abdominal cancer, and -0.6 +/- 1.5% for brain tumor. Most of measured doses (92.2%) agreed with the calculated doses within a tolerance limit of +/- 3% recommended in the literature. However, we found some systematic under-dosage for all treatment sites. The percentage of points passing the gamma criteria, averaged over all treatment sites was 97.3 +/- 3.7%. The gamma pass rate and the agreement of ion-chamber dosimetry generally decreased with increasing the number of PTV's pieces, the degree of modulation (MU/cGy), and the total MU beyond 700. Our local confidence limits were comparable to those of AAPM TG 119 and ESTRO guidelines that were provided as a practical baseline for center-to-center commissioning comparison. Thus, our institutional confidence and action limits for IMRT delivery were set into the same levels of those guidelines. Discussion and Conclusions: The systematic under-dosage were corrected by tuning up the MLC-related factors (dosimetric gap and transmission) in treatment planning system (TPS) and further by incorporating the tongue-and groove effect into TPS. Institutions that have performed IMRT DQA measurements over a certain period of time need to analyze their accrued DQA data. We confirmed the overall integrity of our IMRT system and established the IMRT delivery guideline during this procedure. Dosimetric corrections for the treatment plans outside of the action level can be suggested only with such rigorous DQA and statistical analysis.