CBCT provides a promising method to quantify inter-fraction and intra-fraction errors, and allows a significant reduction in inter-fraction errors. To our knowledge, this is the first report on intra-fraction errors during VMAT in NPC patients, and it provides a guide for designing PTV margins in NPC patients during VMAT.
When applying new treatment techniques to malignancies at other sites, it is necessary to determine setup errors to acquire appropriate PTV margins. The inter-fraction error observed in this study is similar to the inter-fraction error reported by other researches, including the study of Wang et al. in which CBCT analysis indicated that the systematic deviations and random errors during set-up in NPC patients were 1.1–1.3 mm . Similarly, Velec et al. analyzed daily CBCT images acquired during the treatment of 20 head and neck cancer patients, and obtained systematic deviations ranging from 0.8 mm to 1.1 mm and random translational deviations of less than 2 mm . However, a review carried out by Hurkmans et al. focusing on set-up verification in head and neck cancer patients using portal imaging, concluded that the systematic and random deviations varied by 1.6–4.6 mm and 1.1–2.5 mm, respectively , which are higher than the values in this study. These differences may be due to the fact that we automatically matched a smaller region of interest confined to the nasopharynx and upper neck, as there is evidence that positional variation is greater in the lower neck than in the upper neck . In addition, our study used rigorous immobilization devices, such as the thermoplastic mask covering the head, neck and shoulders, and such devices have previously been used in IMRT delivery for head and neck cancer to improve the reproducibility and stability of patient position .
In our study, CBCT effectively improved the accuracy of VMAT in NPC patients because the inter-fraction systematic error could be decreased from 1.0–1.4 mm to 0.4–0.5 mm, which was similar to the results reported by Wang et al. and Dionisi et al.[5, 17]. While the uncertainties in delivering RT can be reduced with CBCT systems, it is important to understand the uncertainties arising from this process. One of the important factors is the uncertainties in image registration. It is current practice in head and neck cancer to use rigid registrations of bony anatomies with the planning CT, to obtain setup errors. The accuracy of these registrations can be affected by the image dose, image resolution, region of interest used for registration etc.[18–20]. Many research studies have reported bone alignment with an accuracy of nearly 1 mm for the translational displacement of a head phantom [21, 22]. However, head and neck cancer patients experience significant deformation, shrinkage and rotation, all of which could also affect the efficacy of a bone match. Recent studies have shown that due to the considerable and frequent shape and posture changes in patients with head and neck cancer, not all structures within a single large region of interest can be simultaneously aligned using bone registration [23, 24]. Therefore, misregistration can occur and the misalignment can persist due to shape or posture changes. It is encouraging that new registration and correction methods to reduce such misalignment and deformation have been proposed [25, 26]. In future, software updates will be aimed at automating these methods to allow better quantification of the setup error.
Furthermore, due to the limited availability of conventional treatment tables, not all users of the bone registration algorithm can adjust the rotational setup of the patient to minimize setup error [9, 27]. There is a possibility of significant misalignment when only the translational part of a six-dimensional (6D) registration is applied to the patient’s position in a completely general 6D registration framework. Fortunately, strategies for avoiding this misalignment have been proposed by Murphy ; these consist of putting the landmark/treatment isocenter at or very near the origin of the rotation axes or calculating only a 3D registration using landmarks near the treatment site. Therefore, it is important to realize the limits of bone registration and deal carefully with the practical application of bone registration tools and patient setup practices during imaging.
Generally, intra-fraction error was mainly influenced by the immobilization device and delivery time. Theoretically, the magnitude and probability of patient intra-fraction movement will most likely increase when fraction times are extended. Hoogeman et al. concluded that intra-fraction systematic geometric error increases with time . However, our study showed that there was no significant correlation between the delivery time and intra-fraction error. One reason for this result was possibly the narrow range of delivery time (5.6–9.4 min), which did not allow statistically significant results to be obtained. Another reason was that limited data were available for analysis, as our study included only 15 patients and six CBCT scans for each patient. In addition, in our study, the intra-fraction systematic error ranged from 0.2 mm to 0.4 mm during the 5–9 min VMAT time. Velec et al. reported an intra-fraction systematic error of 0.3–0.7 mm in patients fixed with a thermoplastic mask covering the head, neck and shoulders during an approximately 15-min IMRT time . A comparison study of IMRT and VMAT in the same institution is required to determine whether VMAT is associated with reduced intra-fraction motion.
Patient intra-treatment movement can be assessed by several methods along a continuum of imaging frequencies. The most commonly used method of assessing intra-treatment motion is pre- and post-treatment imaging [8, 9], which we used in this study. The other approaches are intermittent imaging, acquired as frequently as every 0.5–2 min, or continuous, real-time tracking of the tumor target during radiation delivery with technologies such as CyberKnife or electromagnetic localization [7, 29]. However, as motion may be sustained during the entire course of radiation delivery, differences in the measurement and acquisition schedule could potentially lead to discrepancies between the measured motion and actual motion at radiation delivery. Further studies using continuous imaging should be performed to evaluate the intra-fraction motion during radiation delivery in NPC patients undergoing VMAT.
Gradual increase in displacement as a function of treatment time
Treatment accuracy during fractionated radiotherapy may decrease with time due to tumor regression or weight loss . Den et al. reported that PTV margins in the last 3 weeks were significantly larger than those in the first 3 weeks . We obtained a similar result, as patient inter-fraction and intra-fraction 3D displacements increased gradually as a function of time. Adaptive re-planning strategy is an effective method to account for significant dosimetric variation during radiotherapy, which was mainly caused by setup errors and anatomical changes. Currently, the optimal timing of re-planning remains to be determined.
Appropriate PTV margins
Reasonable designs of PTV margins are the key point for local control and normal tissue protection. Narrow margins tend to be associated with local recurrence, while wide margins result in excessive treatment. Decreasing the PTV margins can theoretically improve the therapeutic gains; such a benefit was illustrated by van Asselen et al., who observed that narrow PTV margins improved parotid sparing and decreased the probability of normal-tissue complications such as xerostomia. However, the application of narrow margins must be based on the premise of excellent quality-control measures such as daily CBCT online correction. In general, narrow margins are not widely used in clinical practice, and are reserved for special cases such as locally advanced tumors that invade tissues adjacent to the brain stem or spinal cord. In this study, we have discussed the PTV margins accounting for setup errors, and these margins were 1.7–2.2 mm and 3.4–4.1 mm with and without correction, respectively. However, it should be noted that PTV margins are not the only component of the setup margin (inter- and intra-fraction motion), and other process-related components such as image registration, treatment planning uncertainties and transfer errors from the planning CT to the simulator should be accounted for. In our institution, therefore, we recommend a PTV margin of 5 mm for NPC patients undergoing VMAT without CBCT and 3 mm for those treated with rigorous daily CBCT scans. Our results provide a theoretical basis for the appropriate design of PTV margins for VMAT in NPC patients.