On the insensitivity of single field planar dosimetry to IMRT inaccuracies.

PURPOSE: To report on the sensitivity of single field planar measurements in identifying IMRT plans with poor calculational accuracy.
METHODS: Three IMRT plans for head and neck cancer were subjected to extensive quality assurance. The plans were recalculated on a cylindrical phantom and between eight and 18 low gradient points were measured in each plan with an ion chamber. Every point measured in these plans agreed to within 4% of the dose predicted by the planning system and the plans were judged acceptable for clinical use. Each plan was then reoptimized with aggressive dose constraints so that the new treatment fields were more highly modulated than the ones from the original plans. Very complex fields can be calculated less accurately and ion chamber measurements of these plans in the cylindrical phantom confirmed significant dosimetric errors--Several of the measured points in each plan differed from the calculated dose by more than 4%, with a maximum single deviation of 10.6%. These three plans were judged unacceptable for clinical use. All six plans (three acceptable, three unacceptable) were then analyzed with two means of individual field planar dosimetry: Portal imaging with an electronic portal imaging device (EPID) and an ion chamber array. Gamma analysis was performed on each set of planar measurements with 2%/2 mm distance to agreement (DTA) and 3%/3 mm DTA criteria to try to determine a gamma analysis threshold which would differentiate the flawed plans from the acceptable ones.
RESULTS: With the EPID and 2%/2 mm DTA criteria, between 88.2% and 92.8% of pixels from the acceptable IMRT plans passed the gamma analysis, and between 87.5% and 91.9% passed for the unacceptable IMRT plans. With the ion chamber array and 2%/2 mm DTA criteria, between 92.4% and 94.9% of points in the acceptable plans passed the gamma analysis, while 86.8% to 98.3% of the points in the unacceptable plans passed the gamma analysis. The difference between acceptable and unacceptable plans was diminished further when gamma criteria were expanded to 3%/3 mm DTA. A fraction of pixels passing the gamma analysis was found to be a poor predictor of dosimetric accuracy with both planar dosimeters, as well as both sets of gamma criteria.
CONCLUSIONS: Deconstruction of an IMRT plan for field-by-field QA requires complex analysis methods such as the gamma function. Distance to agreement, a component of the gamma function, has clinical relevance in a composite plan but when applied to individual, highly modulated fields, it can mask important dosimetric errors. While single field planar dosimetry may comprise one facet of an effective QA protocol, gamma analysis of single field measurements is insensitive to important dosimetric inaccuracies of the overall plan.