Process control analysis of IMRT QA: implications for clinical trials.

The purpose of this study is two-fold: first is to investigate the process of IMRT QA using control charts and second is to compare control chart limits to limits calculated using the standard deviation (sigma). Head and neck and prostate IMRT QA cases from seven institutions in both academic and community settings are considered. The percent difference between the point dose measurement in phantom and the corresponding result from the treatment planning system (TPS) is used for analysis. The average of the percent difference calculations defines the accuracy of the process and is called the process target. This represents the degree to which the process meets the clinical goal of 0% difference between the measurements and TPS. IMRT QA process ability defines the ability of the process to meet clinical specifications (e.g. 5% difference between the measurement and TPS). The process ability is defined in two ways: (1) the half-width of the control chart limits, and (2) the half-width of +/-3sigma limits. Process performance is characterized as being in one of four possible states that describes the stability of the process and its ability to meet clinical specifications. For the head and neck cases, the average process target across institutions was 0.3% (range: -1.5% to 2.9%). The average process ability using control chart limits was 7.2% (range: 5.3% to 9.8%) compared to 6.7% (range: 5.3% to 8.2%) using standard deviation limits. For the prostate cases, the average process target across the institutions was 0.2% (range: -1.8% to 1.4%). The average process ability using control chart limits was 4.4% (range: 1.3% to 9.4%) compared to 5.3% (range: 2.3% to 9.8%) using standard deviation limits. Using the standard deviation to characterize IMRT QA process performance resulted in processes being preferentially placed in one of the four states. This is in contrast to using control charts for process characterization where the IMRT QA processes were spread over three of the four states with none of the processes in the ideal state. Control charts may be used for IMRT QA in clinical trials to categorize process performance, minimize protocol variation and guide process improvements. For the duration of an institution's participation in a protocol, updated control charts can be periodically sent to the protocol QA center to document continued process performance to protocol specifications.