Abstract
Background/Aim: Radiotherapy (RT) outcomes are generally reported based on stage, patient background, and concomitant chemotherapy. This study aimed to investigate the effects of the prescribed dose to gross tumor volume (GTV) and the calculation algorithm on local control in definitive RT for head and neck (H&N) cancers using follow-up images after RT. Patients and Methods: This study included 154 patients with H&N cancers treated by Volumetric Modulated Arc Therapy at the Kobe City Medical Center General Hospital. Patients were classified into those receiving definitive RT (70 Gy of irradiation) and those not receiving it. Follow-up images were used to categorize the patients into the responders and non-responders groups. In the non-responders group, follow-up images were imported into the treatment planning system, and the contours of the residual or recurrent areas (local failure) were extracted and fused with computed tomography-simulated images for treatment planning. Dose evaluation parameters included maximum dose, dose administered to 1% of the volume, dose administered to 50% of the volume, dose administered to 99% of the volume (D99%), and minimum dose (Dmin) administered to the GTV. The doses to the GTV were compared between responders and non-responders. Results: D99% exhibited significant differences between local failure and responders and between local failure and non-responders. Dmin showed significant differences between responders and non-responders and between responders and local failure. Conclusion: This study emphasizes the importance of verifying dose distribution in all slices of treatment planning, highlighting the need for precise assessment of the dose to the GTV in head and neck cancers.
Radiotherapy (RT) aims to deprive cancer cells of their ability to grow and reproduce by delivering high radiation doses while minimizing doses to adjacent organs at risk (OARs). However, conventional RT delivers a constant dose from a fixed angle and irradiates the target and OARs, resulting in severe side effects. Particularly in head and neck (H&N) cancers, it is difficult to meet the general dose index because the distance between the target and normal tissue is small. Intensity Modulated Radiation Therapy (IMRT) (1, 2) delivers a high dose to the target while suppressing the dose to normal tissue and has become a common treatment modality. In addition, Volumetric Modulated Arc Therapy (VMAT) (3-5) is performed while the device is rotated, enabling irradiation in a shorter time.
RT that preserves function and morphology is recommended for H&N cancers because several such patients are anatomically difficult to operate on, and surgery reduces their quality of life (QOL) and makes them less cosmetically pleasing. Several H&N cancers are squamous cell carcinomas, which are radiosensitive and can be treated by RT alone or chemoradiotherapy. Epstein-Barr virus infection is generally thought to be involved in nasopharyngeal carcinoma (6), and although the nasopharynx is a difficult site to operate on, it is easily treated with anticancer agents and radiation. In recent years, the number of cases of oral human papillomavirus (HPV) infection has been increasing in the case of oropharyngeal carcinoma (7), and HPV-positive oropharyngeal carcinoma has a better treatment outcome than HPV-negative carcinoma. Hypopharyngeal carcinoma has a poor prognosis among H&N cancers because of its close association with alcohol consumption and smoking, and its location reduces QOL. Thus, RT is the mainstay of treatment for hypopharyngeal carcinoma.
The H&N regions are composed of sinuses, dental metal, and bone, where several steep density changes occur. RT planning systems utilize different dose calculation algorithms, including Acuros External Beam (AXB) and Analytical Anisotropic Algorithm (AAA). AAA (8) is a model-based algorithm used to reduce the accuracy of dose calculation because it cannot take composition into consideration. AXB is highly accurate for dose distribution evaluation in inhomogeneous regions (9); however, its use in the H&N regions has been scarcely reported (10, 11).
RT outcomes are generally reported based on clinical stage, patient characteristics, and concomitant chemotherapy. In addition, although RT treatment plans are currently created using dose constraints based on calculation algorithms with actual measurements, there are only a few reports (12-14) that consider heterogeneity correction, especially in the H&N regions. This study aimed to investigate the effects of the prescribed dose to the gross tumor volume (GTV) and the choice of calculation algorithm on local control in definitive RT for H&N cancers using VMAT and follow-up images for a comprehensive analysis after RT.
Patients and Methods
Patient selection and characteristics. This study comprised 154 patients with H&N cancers treated by VMAT at the Kobe City Medical Center General Hospital from January 2013 to December 2019. All the patients were classified into two groups: patients receiving definitive RT (70 Gy of irradiation) and patients not receiving definitive RT. The latter included those who were treated by definitive RT but the total dose was <70 Gy, those who received RT postoperatively, those who were interrupted owing to severe side effects, and those who were interrupted at their request.
The study overview is shown in Figure 1. Seventy-seven patients were irradiated with 70 Gy/35 fractions over 7 weeks. Follow-up images, such as positron emission tomography (PET) (15) or magnetic resonance imaging (MRI), were periodically obtained after RT by referring to electronic medical records. The GTV was the sum of the primary tumor and the lymph nodes. Patients with complete or partial response on post-treatment follow-up images as of March 2023 were defined as responders, and those with progressive disease or stable disease were defined as non-responders. Patient characteristics are presented in Table I.
Study overview. The 154 patients were classified into two groups: patients who received definitive radiotherapy (70 Gy of irradiation) and those who did not receive definitive radiotherapy. Based on follow-up images, such as positron emission tomography (PET) and magnetic resonance imaging (MRI) after radiotherapy, the patients were classified into responders and residual or recurrent disease (non-responders). The areas of residual or recurrent (local failure) were extracted using PET and MRI images.
Characteristics of patients receiving definitive radiotherapy (n=77).
Treatment specifications. All the patients were immobilized in a thermoplastic head-neck-shoulder mask in the supine position. Computed tomography (CT) scans covered the area from the vertex to the whole thoracic regions using a 16-slice CT scanner (GE Healthcare, Milwaukee, WI, USA), and the scans were reconstructed with a 2.5-mm-layer spacing. Contours were extracted after the fusion of CT scan images with contrast CT, PET, and MRI images.
Clinical target volume (CTV) comprised the entire lymph node area and lymph nodes that needed to be included, with CTV 70 for the area to be irradiated with 70 Gy, CTV 63 for the area to be irradiated with 63 Gy, and CTV 56 for the area to be irradiated with 56 Gy. Planning target volume (PTV) 70/63/56 was defined as CTV with a 5-mm margin added to CTV. The OARs were selected according to the irradiation range and purpose, such as the brain, brainstem, spinal cord, eyes, lens, optic nerves, optic chiasm, submandibular glands, and parotid glands, and planning organ at risk volume (PRV) margins were set for each of them. The area of streak artifact caused by metals used in the dental treatment was replaced with water after region of interest (ROI) extraction (16).
The treatment planning system (TPS) used AXB and AAA computational algorithms on Eclipse version 13.6 (Varian Medical Systems, Inc., Palo Alto, CA, USA). Dose evaluations were recalculated using AXB and AAA, and modified to actual irradiated monitor units (MU). VMAT mainly uses the Simultaneous Integrated Boost (SIB) method (17, 18), aimed at D50%=70 Gy for PTV70. Double-arc VMAT plans were achieved by radiation oncologists using progressive resolution optimization in the TPS. All the patients were treated with 6-MV photon beams generated from a Clinac iX linear accelerator (Varian Medical Systems, Inc.). The anterior beam, affected by the dental metal artifacts area with water displacement, was planned to avoid irradiation because of uncertain calculation accuracy.
For position matching, kV imaging in two directions was performed, aligning with the bone structure. Cone-beam CT (CBCT) imaging (19), performed after bone matching for the initial three fractions, ensured accurate bone structure alignment. If discrepancies arose in bone structure and tumor position, CBCT imaging was performed each time; otherwise, it was performed weekly. Changes in tumor size or weight prompted additional CT scans (20), which were evaluated in consultation with the radiation oncologists. If re-planning was necessary, plans were promptly modified and verified in advance.
Comparison methods. In the non-responders group, follow-up images [PET (21) or MRI] were imported into the TPS. Contours of residual/recurrent areas (local failure) were extracted and fused with CT images used for treatment planning.
The International Commission on Radiation Units (ICRU) reference point was difficult to define for IMRT owing to the high inhomogeneity of dose distribution within the target. The ICRU Report 83 (22) suggested the use of a dose-volume-histogram based dose index. Dosimetric parameters included maximum dose (Dmax), dose administered to 1% of the volume (D1%), dose administered to 50% of the volume (D50%), dose administered to 99% of the volume (D99%), and minimum dose (Dmin) to the GTV calculated using AXB and AAA. To assess the influence of AXB and AAA, doses to GTV were compared between the responders and non-responders groups. Furthermore, doses to responders, non-responders, and local failure, calculated using AXB and AAA, were compared for comprehensive analysis.
Results
A comparison of the influence of the calculation algorithm on the prescribed dose to the GTV is shown in Figure 2. Doses are presented as averages±standard deviations, and significant differences were calculated by the Student’s t-test. When doses were normalized by MU, the specified dose was higher by AAA than by AXB (23). D1% and D50% showed a large variation, with a significant difference between AXB and AAA (*p<0.05, **p<0.01).
A comparison of the influence of the calculation algorithm on the prescribed dose to gross tumor volume. The prescription dose is the average, and the error bar is the standard deviation. *p<0.05, **p<0.01.
The prescription doses to responders, non-responders, and local failure were compared. The dose calculated with AXB is shown in Figure 3. Dmax values were 74.8±1.6 Gy, 75.3±1.9 Gy, and 74.3±2.0 Gy for responders, non-responders, and local failure, respectively. D1% values were 73.6±1.4 Gy, 74.0±1.8 Gy, and 73.5±1.9 Gy for responders, non-responders, and local failure, respectively. D50% values were 71.3±1.7 Gy, 71.6±1.5 Gy, and 71.1±1.6 Gy for responders, non-responders, and local failure, respectively. D99% values were 68.9±1.4 Gy, 68.3±1.8 Gy, and 64.1±4.9 Gy for responders, non-responders, and local failure, respectively. Dmin values were 66.7±3.7 Gy, 62.9±11.1 Gy, and 60.3±9.2 Gy for responders, non-responders, and local failure, respectively.
The dose was calculated using Acuros External Beam. Responders was defined as the response group, non-responders as the residual or recurrent group, and local failure as the residual or recurrent area in the follow-up images. The prescription dose is the average, and the error bar is the standard deviation. *p<0.05; **p<0.01.
The dose calculated with AAA results is shown in Figure 4. Dmax values were 75.3±1.8 Gy, 75.6±1.9 Gy, and 75.0±2.0 Gy for responders, non-responders, and local failure, respectively. D1% values were 74.3±1.6 Gy, 74.7±1.8 Gy, and 74.2±1.8 Gy for responders, non-responders, and local failure, respectively. D50% values were 71.9±1.4 Gy, 72.2±1.4 Gy, and 71.8±1.5 Gy for responders, non-responders, and local failure, respectively. D99% values were 69.1±1.6 Gy, 68.7±1.9 Gy, and 64.7±5.2 Gy for responders, non-responders, and local failure, respectively. Dmin values were 66.6±4.3 Gy, 62.7±11.0 Gy, and 60.5±10.6 Gy for responders, non-responders, and local failure, respectively.
The dose was calculated using Analytical Anisotropic Algorithm. Responders was defined as the response group, non-responders as the residual or recurrent group, and local failure as the residual or recurrent area in the follow-up images. The prescription dose was the average, and the error bar was the standard deviation. *p<0.05; **p<0.01.
D99% showed significant differences between local failure and responders and between local failure and non-responders (p<0.05). Dmin showed significant differences between responders and non-responders and between responders and local failure (p<0.05).
Discussion
In the comparison between AXB and AAA, a significant difference was observed at D1% and D50%, highlighting the importance of selecting an appropriate algorithm to ensure accurate radiation dose delivery. In the H&N regions, several areas, such as the trachea and sinuses, have significant density changes, which caused dose differences between AXB and AAA. In recent years, the calculation accuracy of AXB has been considered superior to that of AAA, especially in complex regions such as the thoracic regions, which have several heterogeneities. Therefore, AXB should also be used for dose evaluation in the H&N regions.
This study compared the doses to responders, non-responders, and local failure, providing valuable insights into the relationship between prescribed doses and treatment outcomes. The differences in D99% and Dmin values among these groups indicate their potential effects on local failure. Lower doses in certain regions (D99% and Dmin) for non-responders and local failure cases could be associated with decreased local control rates, underscoring the need for precise dose planning and delivery.
The utilization of the SIB methods posed challenges in uniformly administering the target dose to adjacent areas with varying prescribed doses, leading to a nonuniform dose distribution. This nonuniformity might have contributed to dose differences. Additionally, there were instances where the GTV did not receive sufficient doses. Particularly for GTV near the body surface (24-26), the ROI had to be modified inward by 2 to 3 mm to ensure adequate dosing. In some cases, it was not possible to deliver sufficient doses. Even when targets were near OARs, there were instances where the dose to the PTV could not meet the constraints of the OARs. IMRT adjusts the dose intensity based on the extracted ROI. Therefore, an uneven ROI shape results in a corresponding dose distribution, making it challenging to deliver an adequate dose.
The dose distribution in the three cases was calculated using AXB in Figure 5. In the responders case, the 70 Gy (100%) dose line covered the GTV, whereas, in the two local failure cases, the 66.5 Gy (95%) dose line of the GTV was not covered. Although the residual or recurrent areas may not fall within this dose range, the uneven dose distribution within the GTV suggests a potential effect on control. Dose distribution per slice should always be assessed to confirm the absence of hotspots or low-dose areas within the GTV. The H&N regions are clinically complex and require formulation of an optimal treatment plan (27-30). This study’s results suggest that personalized treatment planning is key to ensuring appropriate dose distribution according to patient characteristics and nature of the tumor.
Example images of responder and local failure cases. The first row shows the gross tumor volume of the treatment plan, the second row shows the follow-up images from which residual or recurrent areas were extracted, and the third row shows the dose distribution calculated using Acuros External Beam.
Study limitations. First, we solely focused on dosimetric parameters and did not assess clinical factors based on staging or patient characteristics. This study used a retrospective design, relying on data collected from patient records and follow-up images. Second, owing to the single-hospital setting, a relatively few numbers of patients, and limited patient follow-up, there was a lack of information regarding long-term treatment efficacy and recurrence rates. Third, although we compared different calculation algorithms, we did not compare them with other TPSs (31, 32). Finally, although the treatment plans used in this study were based on certain protocols, we did not consider detailed differences in treatment plans, and changes made were based on the radiation oncologists’ judgment.
Conclusion
There is no significant difference in trend between AXB and AAA in the treatment of H&N cancers. However, the H&N regions contain several heterogeneous areas; thus, the more accurate AXB should be used for the TPS algorithm. Dosimetric parameters, such as D99% and Dmin to GTV and PTV, should be evaluated in future studies.
Acknowledgements
The Authors thank Hirokazu Mizuno, Ph.D., and Yutaka Takahashi, Ph.D., for their valuable suggestions and conception.
Footnotes
Authors’ Contributions
Mikiko Yamashita: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Resources; Validation; Visualization; Roles/Writing – original draft; and Writing – review & editing. Shingo Ohira: Conceptualization, Data curation, Investigation, Methodology, Project administration, Visualization, Roles/Writing – original draft, and Writing – review & editing. Hiroaki Tanabe: Data curation, Formal analysis, Investigation, Resources, Validation, Roles/Writing – original draft, and Writing – review & editing. Masaki Kokubo: Project administration, Resources, Supervision, Roles/Writing – original draft, and Writing – review & editing. Masahiko Koizumi: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Supervision, Visualization, Roles/Writing – original draft, and Writing – review & editing.
Funding
This study was supported by JSPS KAKENHI Grant [Grant-in-Aid for Scientific Research (C) 21K08797].
Conflicts of Interest
The Authors have no conflicts of interest to declare in relation to this study.
- Received November 2, 2023.
- Revision received December 6, 2023.
- Accepted December 7, 2023.
- Copyright © 2024 The Author(s). Published by the International Institute of Anticancer Research.
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY-NC-ND) 4.0 international license (https://creativecommons.org/licenses/by-nc-nd/4.0).
