Quality assurance by systematic in vivo dosimetry: results on a large cohort of patients

Background: In vivo dosimetry is widely considered to be an important tool for quality assurance in external radiotherapy. Introduction: In this study we report on our experience over more than 4 years in systematic in vivo dosimetry with diodes. Materials and methods: From November '94 an in vivo entrance dosimetry check was performed for every new patient irradiated at one of our treatment units (Linac 6/100, 6 MV X-rays). Diodes were calibrated in terms of entrance dose; appropriate correction factors had been previously assessed (taking SSDs, field width, wedge, oblique incidence and blocking tray into account) and were individually applied to in vivo diode readings. The in vivo measured entrance dose was compared with the expected one, with a 5% action level; if a larger deviation was found, all treatment parameters were verified, and the in vivo dosimetry check was repeated. During the period November '94-May '99, 2824 measurements on 1433 patients were collected. Results: Nine out of 1433 (0.63%) serious systematic errors (leading to a 5% or more on the delivered dose to the PTV) were detected by in vivo dosimetry; four out of nine would produce a 10% or more error if not detected. The rate of serious systematic errors detected by an independent check of treatment chart and MU calculation was found to be 1.5%, showing that less than 1/3 of the errors escapes this check. One hundred and twelve out of 1433 (7.8%) patients had more than one check: the rate of second checks was significantly higher for breast patients (31/250, 12.4%) against non-breast patients (81/1183, 6.8%, P = 0.003). A number of patients demonstrated a persistent relatively large error even after two or more checks. For almost all patients the cause of the deviation was assessed; the most frequent cause was the difficulty in correctly positioning the patient and/or the diode. When analyzing the distribution of the deviations between measured and expected entrance doses (excluding first checks in the case of repetition of the in vivo dosimetry control) the mean deviation was 0.4% with a standard deviation equal to 3.0%. The rates of deviations larger than 5 and 7% were 9.9 and 2.6%, respectively. When considering the same data taking the average deviation in the case of opposed beams, the SD became 2.6% and the rates of deviations larger than 5 and 7%, respectively, 5.2 and 0.8%. When dividing the beams according to their orientation, significantly higher rates of large deviations (>5 and 7%) were found for oblique and posterior-anterior (PA) fields against lateral and anterior-posterior (AP) fields (P < 0.05). Similarly, higher rates of large deviations were found for wedged fields against unwedged fields (P < 0.03) and for blocked fields against unblocked fields (P < 0.01). When dividing the data according to the anatomical district, accuracy was worse for breast (mean deviation 0.1%, 1 SD: 3.5%) and neck AP-PA fields (mean deviation 1%, 1 SD: 3,4%). Better accuracy was found for vertebrae (0.1%, 1 SD 2.1%) and brain patients (-0.7%, 1 SD: 2.6%). During the considered period, in vivo dosimetry was also able to promptly detect a systematic error caused by a wrong resetting of the simulator height couch indicator, with a consequent error in the estimate of patient thickness of about 4 cm. Conclusions: In our experience, systematic in vivo dosimetry demonstrated to be a valid tool for quality assurance, both in detecting systematic errors which may escape the data transfer/MU calculation check and in giving an effective way of estimating the accuracy of treatment delivery. (C) 2000 Elsevier Science Ireland Ltd. All rights reserved.