Evaluation of interfraction setup variations for postmastectomy radiation therapy using EPID‐based in vivo dosimetry

Abstract Postmastectomy radiation therapy is technically difficult and can be considered one of the most complex techniques concerning patient setup reproducibility. Slight patient setup variations — particularly when high‐conformal treatment techniques are used — can adversely affect the accuracy of the delivered dose and the patient outcome. This research aims to investigate the inter‐fraction setup variations occurring in two different scenarios of clinical practice: at the reference and at the current patient setups, when an image‐guided system is used or not used, respectively. The results were used with the secondary aim of assessing the robustness of the patient setup procedure in use. Forty eight patients treated with volumetric modulated arc and intensity modulated therapies were included in this study. EPID‐based in vivo dosimetry (IVD) was performed at the reference setup concomitantly with the weekly cone beam computed tomography acquisition and during the daily current setup. Three indices were analyzed: the ratio R between the reconstructed and planned isocenter doses, γ% and the mean value of γ from a transit dosimetry based on a two‐dimensional γ‐analysis of the electronic portal images using 5% and 5 mm as dose difference and distance to agreement gamma criteria; they were considered in tolerance if R was within 5%, γ% > 90% and γmean < 0.4. One thousand and sixteen EPID‐based IVD were analyzed and 6.3% resulted out of the tolerance level. Setup errors represented the main cause of this off tolerance with an occurrence rate of 72.2%. The percentage of results out of tolerance obtained at the current setup was three times greater (9.5% vs 3.1%) than the one obtained at the reference setup, indicating weaknesses in the setup procedure. This study highlights an EPID‐based IVD system's utility in the radiotherapy routine as part of the patient’s treatment quality controls and to optimize (or confirm) the performed setup procedures’ accuracy.


| INTRODUCTION
Postmastectomy radiation therapy (PMRT) is technically difficult, given the complexity of the target volume and its proximity to critical structures, including the heart, lung, brachial plexus, and contralateral breast. [1][2][3] More advanced techniques like intensity modulated radiation therapy (IMRT) or volumetric modulated arc therapy (VMAT) can achieve highly conformal dose distributions with improved target volume coverage and sparing of normal tissues compared to conventional techniques. These techniques have the potential to improve treatment outcomes for PMRT and significantly reduce the dose to the heart and the ipsilateral lung. [4][5][6][7] Nevertheless, uncertainties related to interfraction positioning may lead to inaccuracies in the dose delivered; due to the steepness of the doseeffect curves, the efficacy of IMRT and VMAT can be limited and the patient outcomes for both local tumor control and normal tissue complications can be affected. As previously reported, [8][9][10] dose differences of breast treatment in the supine position can be correlated with the patient setup. These errors cannot be detected by pretreatment verification or through accurate quality control of the connected machines and medical devices. 11,12 The notion has gained ground that these techniques only benefit patients when a good imaging and patient positioning technique is available, suggesting a combined IMRT and image-guided radiation therapy approach. 13,14 Cone beam computed tomography (CBCT) scans can be considered a gold standard to assess interfraction uncertainties for many radiotherapy treatments 15,16 including treatments in the breast area 17 ; Jain et al. 18  Daily acquisition of the treatment images and immediate online correction can reduce the patient setup error's impact, but it increases department workload, with an unavoidable added dose to the patient that should be considered in the treatment plan. 19 Threedimensional (3D)-surface imaging systems are a valid tool able to control patient positioning throughout treatment delivery. They were found to be valuable for reducing errors when comparing with patient alignment from skin marks. 20 Moreover, their technical accuracy has been shown to be quite high, 21,22 nevertheless their suitability for clinical application is increased when combined with CBCT. 23 IVD is the last control within the radiotherapy workflow as it is performed during the treatment delivery; it can detect whether the dose delivered to the patient is within the tolerance level and whether the treatment is dosimetrically reproducible. These peculiarities in addition to IVD's capacity to avoid severe accidents distinguish it with respect to image-and surface-guided radiotherapy systems. For these reasons it has been recommended by several international organizations 24,25 and has become mandatory in some western countries. 26,27 Moreover, it is widely used in Europe to evaluate the interfractional variations in dose delivery and patient setup for many treatment regions; many studies have validated the technique 28,29 and presented the results obtained in the clinical practice. 8,30,31 Mijnheer et al. 32 , analyzed with EPID-based IVD more than 15 000 plans on different treatment sites, and found that more than 30% exceeded the alert criteria, attributing most of the errors to deviations from the routine clinical procedure and to anatomical changes. This study aims to investigate the robustness of the patient setup procedure in use for PMRT when the CBCT considered the gold standard for the patient setup, is not used daily. EPID-based IVD was used to evaluate inter-fraction setup variations occurring in two different scenarios of the radiotherapy routine identified as the reference setup, immediately after the CBCT, and the current setup performed without image guidance. The percentages of electronic portali maging device (EPID) -based IVD out of the tolerance level (OTL) registered, were compared and analyzed. The results obtained in the different scenarios were used to identify and adjust weak rings of the overall radiotherapy process. Six Megavolts photon beams of Synergy or Axesse linacs (Elekta, Stockholm, Sweden), available in the department, were used for the treatments. Experienced radiation oncologists conducted the target and organs at risk delineation according to the breast cancer atlas for the radiation therapy planning consensus definitions of the Radiation Therapy Oncology Group. 34,35 The plan consisted of one or two arcs for the VMAT technique, whereas five beams were delivered via a step-and-shoot technique for IMRT. A patient pretreatment verification was performed using an irradiating MatriXX Evolution two-dimensional (2D) array (IBA Dosimetry, Schwarzenbruck, Germany). Measured and calculated planar dose distributions were compared with the gamma index method using a γ passing rate greater than 90%, with 3 mm distance to agreement and 3% dose difference and a 10% dose threshold. 36,37 A CBCT was acquired for each patient at the first treatment fraction and then once a week.

2.A | Patients and treatment workflow
The couch was moved into the correct position after the CBCT alignment process; however, the maximum accepted displacement was ±5 mm in any one of the x, y, or z directions 38,39 ; if higher displacements were required, the patient was aligned again and the CBCT repeated; in case of a persisting error, a consultation with the radiation oncologist for the management of this treatment was scheduled. This study was reviewed and approved by the Ethics Committee of Sichuan Cancer Hospital in April 2017.
The daily patient setup, named current setup, comprises the patient positioning on the immobilization device, alignment of the lasers with patient's CT reference tattoos, and translation of the treatment couch following the treatment planning indications, to align the machine and the treatment isocenters. The daily patient setup followed by the CBCT scan and successive positioning adjustments was named reference setup.

2.B | EPID-based IVD
An EPID-based IVD was scheduled for each patient twice a week: during the reference and the current setup. A portal image was acquired for each beam with the portal imaging system iViewGT a-Si panels (Elekta, Crawley, United Kingdom), and it was imported into SOFT-DISO version 1.24 EPID-based IVD software (Best Medical Italy, Chianciano, Italy). 40,41 Considering five beams for an IMRT treatment and one or two arcs for a VMAT treatment, at the end of the course of radiotherapy, 50 EPID images for each IMRT patient and 10-20 EPID images for each VMAT patient were acquired. SOFTDISO uses a dosimetric method and provides for each IVD test: the ratio R between the reconstructed (D iso ) and planned (D tps ) isocenter doses (R = D iso / D tps ) and the γ-analysis obtained comparing the signal between the first EPID image (reference image obtained at the reference fraction) and the subsequent images acquired during the treatment course by the SOFTDISO are reported in literature. [40][41][42] The ratio R represents the accuracy of the dose delivered at the isocenter point. The gamma analysis supplies a transit dosimetry to verify the treatment reproducibility, which can be affected by the patient setup, linac output factor variations, beam interruptions, dose calculations, and the presence of patient morphological changes. The ratio R between the reconstructed and planned isocenter doses is considered in tolerance when 0.95 ≤ R ≤ 1.05 considering the difficulty in chest wall dosimetry and the statistical propagation of the errors for R (uncertainties of D iso estimated in 4% in inhomogeneous tissues, uncertainties in D tps within 3%). 43,44 The global γ-analysis adopted two gamma criteria: (a) the EPID percentage signal agreement, ΔS%, and (b) the distance to agreement, Δd (mm). We adopted as pass criteria 5% and 5 mm. The current choice of pass-fail criteria aligns with previous literature data 30,45 and is based on our experience with 2D in vivo dose verification of IMRT-VMAT using gamma evaluation since the start of routine clinical implementation in 2016. 9 Δd = 5 mm is also the maximum displacement value acceptable in clinical practice by the radiation oncologists, while the ΔS was defined by considering the presence of dose gradients (interface lung-PTV) and mobility of the irradiated organs (breath). Two tolerance levels were fixed: (a) the percentage γ-index, γ% ≥ 90.0% (i.e., the number of points with γ<1 must be greater than 90.0%), and (b) the mean γ value, γ mean ≤ 0.4. Therefore, within the EPID irradiated area, a maximum of 10% of the points in disagreement was considered acceptable; moreover, the distribution of the γ values characterized by a mean γ < 0.4 is an indicator of the weight of the discrepancy. An IVD test warning started when even one of the three indices resulted OTL. IVD tests out of tolerancecaused by acquisition errors or due to a lack of sensitivity of the system (predicted dose lower than 5 cGy)were excluded from the results.

2.C | Effectiveness of EPID-based IVD in detecting errors
The effectiveness of EPID-based IVD in detecting setup errors was verified for simple geometries with a thorax phantom. CT dataset with a slice thickness of 3 mm of an inhomogeneous anthropomorphic (Alderson) phantom was acquired and imported into Pinnacle

3.A | Clinical practice
For each pre-treatment plan verification, the gamma analysis of the measured and calculated planar dose distributions was acceptable with a mean γ% of 94.0% (range 93.1%-100%).
One thousand one hundred and ninety EPID-based IVD checks were scheduled and acquired. However, due to errors during the acquisition process, around 15% of these tests were excluded from the results. Lack of synchronization between SOFTDISO and the beam delivery and wrong (and not realignable) positioning of the EPID were the leading causes of these errors. Table 1 shows the results of EPID-based IVD for IMRT and VMAT treatments and the current and reference setups. From the results obtained 6.3% (corresponding to a total of 64) of the overall EPID-based IVD analyzed, resulted OTL. The percentage of OTL registered during the current setup was three times greater (9.5% vs 3.1%) than the one registered at the reference setup. The errors that can contribute to an EPID-based IVD OTL, and the frequency of their occurrence for IMRT, VMAT, current and reference setups, are reported in Table 2.  Fig. 2(a)], the inline and crossline signal profiles of the EPID images acquired for different fractions [ Fig. 2(b)], the R ratio [ Fig. 2(c)]     The percentage of patients (P%) ending the treatment with the mean values of the indices R, γ%, γ mean within the tolerance level and the percentage of the tests (T%) with R, γ%, γ mean indices within the tolerance level are displayed in Table 3. P% was found to be between 96.0% and 100%, while T% varied between 92.0% and 95.8%.

| DISCUSSION
Setup errors are a major source of deviations during EPID-based IVD verification programs. 8,9,30,32,45,46 If the setup of a patient is irreproducible, large variations in the dose reconstruction will be observed especially when high-conformal techniques are utilized. Online setup verification using CBCT information improves the results, but the added radiation dose to the patient and the increase of the department workload contribute to the decision to not program it daily.
EPID-based IVD is complementary to CBCT use, and can provide useful information in the procedure optimization process. From the analysis of the EPID-based IVD, the OTL registered at the current setup (i.e. when CBCT was not performed before the delivery) were three times higher than the ones obtained during the reference setup and were mainly due to setup errors (occurrence rate 80.3%).
This increase of inter fraction setup variation registered during the current setup is an important result because it highlights a lack of robustness of the PMRT setup procedure in use in the department.
In this case, the different steps of the radiotherapy path (from the scanning CT to the dose delivery) should be the object of an accurate evaluation to identify the weak link in the radiotherapy chain and minimize possible errors.
Our study identified patient positioning errors due to the follow-   refine and optimize the current setup procedure in use in the clinical practice. With the certainty of the consolidated setup procedure for PMRT, the authors intend to go further with an accurate analysis investigating errors highlighted by more stringent thresholds and gamma criteria, 37,51 and extending the study to other types of radiotherapy treatments.

| CONCLUSIONS
Interfraction setup variations occurring at current setup in the daily clinical routine of PMRT can be higher than the ones registered in concomitance with the CBCT scheduled along the radiotherapy course, pointing out weaknesses of the setup procedure in use.
The study highlights the feasibility and utility of an EPID-based IVD system in the radiotherapy routine as part of the patient's treatment quality controls and to optimize (or confirm) the accuracy and reproducibility of the procedures performed.

CONF LICT OF I NTEREST
There is no conflict of interest declared in this article.