National survey of patient specific IMRT quality assurance in China

Four hundred and three centers responded to this survey, accounting for 56.92% of the centers that are currently practicing IMRT in China. Among the 403 centers, 152 centers have also implemented VMAT. A wide range of IMRT experience was reported, ranging from greater than ten years to less than three months. The 403 responding centers included 41 cancer hospitals and 362 general hospitals; or 100 academic hospitals and 303 non-academic hospitals. Three hundred five thousand patients were treated with IMRT per year in the responding centers (Fig. 1). The most commonly treated sites were lung, breast, cervical, nasopharyngeal, esophageal, and rectal cancer.

In all responding centers, medical physicists are reported to be responsible for the IMRT QA program. Radiotherapists and clinical engineers are also involved in IMRT QA under the supervision of medical physicists in 133 centers and 27 centers, respectively. Regardless of the QA performer, checking and approving of the IMRT QA results are performed by medical physicists. There are a total of 1599 medical physicists in the 403 responding centers (Fig. 2). The number of IMRT patients per physicist is shown in Fig. 3. More than 100 IMRT patients were treated per physicist in 272 centers.

Figure 4 shows the statistics of linear accelerators and treatment planning systems (TPS) used for IMRT. There are a total of 655 linacs and 819 TPS workstations in the responding centers. Some centers used linacs from more than one manufacturer. The number of centers that has 1, 2, 3, 4, and ≥ 5 linacs is 277, 92, 18, 13, and 17, respectively. The number of centers that has 1, 2, 3, 4, and ≥ 5 TPSs is 257, 71, 32, 10 and 33, respectively. The number of IMRT patients treated per linac is shown in Fig. 5. In more than half of the centers, the number of IMRT patients treated per linac was greater than 300.

Table 1 summarizes the verification techniques and methodologies used for IMRT QA. All responding centers used measurement-based verification method. Three hundred thirty one centers used point dose verification, of those 4 centers used point dose verification only without dose distribution verification. Ion chambers with different sizes (range: 0.01 cc to 0.6 cc) were used for point dose verification. For fixed gantry IMRT QA using 2D array devices, 107 centers used perpendicular field-by-field (PFF) verification to improve efficiency when the IMRT QA results were out-of-tolerance using the perpendicular composite (PC) method. All the responding centers that have implemented VMAT (n = 152) applied true composite (TC) method for VMAT QA using Delta4, ArcCheck, Octavious with dedicated phantom, film, Compass or EPID, etc.

235 (58.3%) centers performed measurement-based IMRT QA for all of their IMRT patients, while 168 centers (41.7%) only performed measurement-based IMRT QA for selected patients. The selection criteria varied between centers and included a wide range of factors such as tumor site, plan complexity, prescription dose, fractionation, normal tissue tolerances, treatment delivery technique, intent of treatment, availability of treatment machine, verification tools, physicist time, patient’s economic status, reimbursement, physician and/or patient’s preference, etc. Most of these centers performed patient-specific IMRT QA for all of their IMRT patients during the first year of IMRT implementation (or for the first 100 IMRT patients), after which they randomly selected patients for IMRT QA with a sampling rate ranging from 20 to 60%.

For patients that measurement-based IMRT QA were not performed, 16 centers performed calculation-based verification, but majority of the centers did not perform any other type of verification. All centers performed measurement-based IMRT QA for hypo-fractionation radiotherapy, SBRT and SRS plans.

The frequency of absolute dose calibration for diode/ion chamber arrays varied between centers. Three centers had never performed absolute dose calibration after commissioning. Measured dose was used as reference in 205 centers, calculated dose was used as reference in 282 centers, and interpolation of measured dose was used as reference in 195 centers. Pencil beam, convolution/and superposition and Monte Carlo based dose calculation algorithms were used in 108, 278 and 140 centers, respectively. Grid size of 1, 2, 2.5, 3, 4 and 5 mm was used for dose calculation in TPS in 39, 155, 50, 190, 67 and 2 centers. 1–4 mm was used in 75 centers based on different treatment planning system, delivery technique, or target size.

Different evaluation metrics, tolerance and action limits were used for point dose verification and 2D dose verification. Seventy seven centers evaluated the concordance between the calculated and the measured dose distributions with different metrics and criteria in different dose gradient regions.

The gamma criteria implemented for evaluating IMRT QA varied between centers. Dose difference (DD) of 2–5% and dose to agreement (DTA) of 1–5 mm was used in these centers. The most commonly used DD/DTA value for gamma criteria was 3%/3 mm. For normalization methods, 207 centers used maximum dose point as the normalization point, 166 centers used isocenter and 80 centers used other points in the high dose plateau region. Global/local normalization was used in 306 and 78 centers, respectively. Dose analysis mode and threshold were also different between centers.

The most common causes for failed IMRT QA cases were over modulation, point dose measurement in a high dose gradient region, incorrect QA phantom setup, and MLC leaf position uncertainty. Other causes included TPS beam modeling error, QA planning error, small field or narrow long field, QA device error, linac output error, IMRT QA analysis error, and laser issues.

Most centers reported difficulties in analyzing root causes and providing solutions for failed IMRT QA cases. The methods of investigating failed IMRT QA included checking the verification plan, checking the verification system, planning system, and delivery system, repeating measurement, and repeating the entire IMRT QA verification. Three hundred forty nine centers checked the treatment machine consistency, verification system and treatment planning system performance, 195 centers edited the IMRT plan based on the results of IMRT QA, and 144 centers discussed with the radiation oncologist to make a clinical decision.

The frequency of MLC QA varied between centers. Thirty three centers did not perform MLC QA. The MLC tests, including leaf calibration and position accuracy, were performed using film, EPID, 2D array, graph papers or log files. One hundred sixty eight centers reported participation in external IMRT audits or inter-institution comparison, and 13 centers were credentialed for clinical trials with IMRT.

This survey revealed a significant issue of limited resources in physicist staffing, time, QA device, and treatment machine. In some centers, IMRT QA was voluntarily performed by medical physicists without salary compensation for working overtime. There was no reimbursement for IMRT QA in majority of the centers. One hundred thirty seven centers performed IMRT QA before the first treatment for hypo-fractionated treatments, SRS and SBRT. Two hundred fifty one centers performed IMRT QA during the first three fractions for conventional fractionated treatment. IMRT QA was performed during working hour in day time in 124 centers, in the evening in 177 centers, and on weekend in 283 centers.

The most concerning issues on IMRT QA devices and techniques included the optimal size of ion chamber for point dose measurement, the accuracy and comparability of various techniques and devices with different hardware and software. IMRT QA was largely considered time-consuming, complex and cumbersome in this survey. Easy-to-use devices with high resolution and high efficiency are highly desired. The clinical significance of IMRT QA results was unclear and it is difficult to appreciate for current 2D, phantom-based IMRT QA techniques. There was a lack of 3D anatomy-based IMRT QA devices and techniques.

For point dose verification, leakage current should be corrected when small volume ion chambers are used in point dose verification [8, 13]. The ion chamber with adequate spatial resolution should be selected and placed in a plateau dose region, considering of the dose gradient and positioning errors. In general, the dose gradient across the ion chamber for homogeneous dose distributions should be less than 5% of the mean dose to the chamber. For SBRT\SRS, it should be as close as possible to the criteria. The calculated dose to the chamber volume, instead of dose to the effective measurement point or middle of the chamber active volume should be compared with the measured dose.

For planar dose verification, if the angular dependence of 2D array is negligible or can be corrected accurately, the TC measurements can be used. Otherwise, the measurements should be performed using PFF method due to anisotropic dose response of the array detectors [14‐16]. The PC method should not be used due to the possibility of masking delivery errors. No significant correlation was observed between PFF 3%/3 mm DTA and the actual 3D dose differences [17‐20]. A confidence limit difference of 12.4 and 7% for TC and PFF was noted in the TG-119 report, respectively [3]. So, tools for patient anatomy-based 3D verification and specially designed VMAT QA are highly desirable [21‐25].

For gamma index analysis, the evaluated dose distribution should have the same or higher resolution than the reference distribution [26]. For the 282 centers that used calculated dose as the reference distribution, the comparison accuracy was compromised if no interpolation of measured dose distribution was used. When measured dose distribution is used as reference, small calculation grid size should be used. Dose calculation grid size larger than 3 mm is not appropriate for IMRT QA.

Dose difference, DTA, dose profiles and isodose distribution should be reviewed in addition to the gamma pass rate. Furthermore, not only the failure percentage, but also the maximum, average gamma value, and gamma distribution should be reviewed [11]. It is difficult to establish the acceptance limits for IMRT QA because different delivery systems, planning systems, and verification devices are used [27, 28]. Analyzing gamma pass rate with different dose difference/DTA criteria is useful to find the sources and judge the impact of discrepancies. Stricter criteria of 3%/2 mm, even 2%/2 mm should be used, as the experience and confidence increase in IMRT QA. Some centers used locally defined limits varying with cancer site and plan complexity [29]. 10% action limit for point dose verification, 5 mm DTA limits, and 80% action limit for gamma analysis is not acceptable. The European Society for Radiotherapy and Oncology (ESTRO) recommended tolerance and action limits of 3 and 5% for ion chamber measurements [30]. AAPM recommended that tolerance and action limits should be within ≤2% and ≤ 3%, respectively [11]. Planar dose verification using a 2D array with the detector spacing of 7 mm could not detect MLC leaf position errors smaller than 2 mm, with 3%/3 mm criteria and a 90% gamma passing rate [31]. The average deviations in D_95% of target, D_0.1cc of the spinal cord reached 8 and 12%, respectively, in complex head and neck plans for systematic leaf position errors of 1 mm [32]. Therefore, tighter tolerances should be used together with accelerator and MLC QA [33, 34].

In addition, absolute dose mode should be used for IMRT QA analysis because considerable differences may go undetected using relative dose mode. The absolute dose calibration of the ion chamber or diode arrays should be performed before each measurement, in order to rule out the influence of detector response and accelerator output variation. Global normalization should be used due to its clinical relevance. The normalization point should be placed in a high dose, low gradient region, often the maximum dose point, not necessarily isocenter of the plan, especially when the isocenter is located in the low dose or high gradient region [35].

If IMRT QA failed, medical physicist should systematically review the dose difference, DTA, gamma index, isodose distribution, dose profile, structure specific dose distribution and DVH when available, to determine if the dose deviations are clinically acceptable. A comprehensive root causes analysis should be performed to determine the reasons for these discrepancies and find the solution to them. It may be necessary to check the clinical plan, QA plan, QA device, setup, and/or measure with a different measurement device or different geometry. Medical physicists should understand the characteristics and performance of their QA tool, implementation details, and test its accuracy. If the modulation of the failed plan is much more complex than usual, planning with less complex intensity patterns should be considered. If the gamma passing rate is systematically lower than the recommended action limits, then the dose differences should be thoroughly reviewed, using local normalization and tighter criteria, to find subtle regional discrepancies.

The IMRT workflow should be thoroughly investigated, including the TPS beam modeling and commissioning, QA planning, QA device and linac performance testing, and end-to-end tests. Gamma passing rates should be tracked among patients for the same sites, to differentiate if the errors are specific for a treatment site, or delivery equipment. In addition, patient-specific verification QA for previous cases, the multiple center comparison or independent validation tests can also be performed to help identify the sources of errors.

The survey also revealed issues that need to be further investigated and discussed. For example, tools such as EPID and log file that can simplify the set-up, measurement or analysis of IMRT QA should be encouraged [36]. While there are controversies on the value and methods of patient-specific IMRT QA [37‐40], especially whether computation can replace measurements, a significant increase is expected in calculation-based verification. For plans that the measurement-based IMRT QA was not performed, calculation-based verification should be considered, together with systematic QAs of linac and TPS [41]. Furthermore, reimbursement of IMRT QA should be addressed and made known to physicians, department directors and hospitals, as well as to the administration and public.