A meta-analysis of dose-volume parameters and treatment efficiency comparing O-ring and C-arm accelerator systems for craniospinal irradiations

Graphical abstract

Highlights

• •A systematic study of three treatment modalities in the craniospinal irradiation.
• •Dose-volume comparisons of O-ring and C-arm linacs included 12 studies (87 patients).
• •O-ring linacs reduced dose to eyes and esophagus by 28.6% and 10.5%.
• •Dual-layer multi-leaf collimator radiotherapy shortened the treatment time by 50%

Abstract

Background and Purpose

Craniospinal irradiation (CSI) is a critical treatment modality for central nervous system (CNS) tumors. This study aimed to compare the dose-volume parameters and treatment efficiency between O-ring and C-arm linear accelerators (linacs) for CSI.

Materials and Methods

A systematic search of English and Chinese databases was conducted from January 2005 to January 2026. Studies were evaluated using the PICOS framework. Data extraction and meta-analysis were performed using Stata 18.0 software encompassing subgroup analysis.

Results

Twelve studies involving 87 patients were included. No significant differences were found in dose to the planning target volume (PTV) between O-ring and C-arm linacs (D_max: standardized mean difference (SMD) = -0.14, 95% confidence interval (CI): −0.61 to 0.32; D_mean: SMD = −0.53, 95% CI: −1.57 to 0.50). However, O-ring linacs, including dual-layer multi-leaf collimator linacs (DLM-linac) and helical radiotherapy (HR), delivered significantly lower maximum doses (D_max) to the heart and liver (heart D_max: SMD = -1.53, 95% CI: −2.53 to −0.52, p = 0.003; liver D_max: SMD = -1.33, 95% CI: −2.29 to −0.37, p = 0.007). Subgroup analysis revealed that DLM-linac significantly reduced D_max to the eyes and esophagus compared to C-arm linacs (eyes D_max: SMD = -0.62, 95% CI: −1.22 to −0.01, p = 0.045; esophagus, D_max: SMD = -0.45, 95% CI: −0.89 to −0.01, p = 0.044). Additionally, HR had the highest monitor units and the longest treatment time among all linac types.

Conclusions

Both O-ring and C-arm linacs are effective for CSI. No significant difference in PTV dose was revealed when the doses to the lungs and thyroid are comparable. However, O-ring linacs, particularly DLM-linac, provide superior protection for critical organs such as eyes and esophagus without extending treatment time. Considering both dose advantage and treatment efficiency, DLM-linac is a suitable radiotherapy treatment modality for CSI.

1. Introduction

Craniospinal Irradiation (CSI) has emerged as a crucial treatment modality for central nervous system (CNS) tumors, particularly for malignancies such as medulloblastoma, germ cell tumors, ependymoma, and CNS leukemia, which are prone to disseminating through the cerebrospinal fluid [1]. The increasing incidence of CNS disorders poses significant health risks, resulting in considerable societal burdens, while effective diagnostic and therapeutic measures remain limited [2]. Advancements in medical technology have enhanced patient survival rates and reduced recurrence rates through CSI [3]. CSI involves irradiating the entire brain and spinal cord, presenting challenges due to its large target volume and extensive immobilization requirements. Traditional radiotherapy methods, such as three-dimensional conformal radiation therapy (3D-CRT), often result in non-uniform dose distribution due to field junction issues, thereby increasing the treatment risk.

Nowadays, modern radiotherapy technologies, including intensity modulated radiation therapy (IMRT), volumetric modulated arc therapy (VMAT), helical radiotherapy (HR), and proton beam scanning, have significantly enhanced dose conformity and uniformity while reducing radiation exposure to critical organs [4], [5], [6], [7]. For example, HR enables single-setup treatment for the whole body, avoiding dose non-uniformity caused by field junctions in conventional radiotherapy [8]. Additionally, proton beam scanning further minimizes radiation exposure to normal tissues, making it particularly suitable for pediatric patients. CSI tumors constitute 25% of pediatric tumors [9]. Despite these advancements, efficacy and safety in CSI still face challenges [10].

Currently, most medical linear accelerators (linacs) are divided into two types: the traditional C-arm and the O-ring gantry linacs [11], [12], [13]. The C-arm linac, a prevalent radiotherapy device, offers high flexibility, allowing non-coplanar irradiation and partial rotation around the patient to improve dose distribution. Furthermore, the C-arm linac demonstrates broad applicability across diverse radiotherapy contexts, encompassing orthopedic and vascular interventions. The O-ring linacs can be categorized into two types: 1) HR type for single-setup irradiation and 2) a ring-gantry linac with a dual-layer MLC and integrated kV CBCT imaging type allowing multi-isocenter treatment. HR is a dedicated device for helical IMRT that allows image guidance using megavolt-computed tomography [14]. The gantry of HR enables 360-degree rotational irradiation, providing a more uniform dose distribution for deep-seated tumors. The compact design of the O-ring gantry reduces spatial requirements in treatment rooms. However, notable disadvantages include higher equipment and maintenance costs, as well as limited patient applicability. Nevertheless, a dual-layer MLC linear accelerator (DLM-linac) was a clinical alternative to HR. DLM-linac aims to provide rapid, accurate and safe treatment [15], [16]. The DLM-linac integrates the benefits of the O-ring linac and C-arm linac, making it highly desirable among medical oncologists and physicists.

This study aimed to thoroughly compare the dose distributions and treatment efficiency between O-ring and C-arm linacs and provide clinical advice in future CSI treatment.

2. Material and methods

2.1. Searching methodology

A systematic search was conducted across English and Chinese databases, including PubMed, Web of Science, the Cochrane Library, China National Knowledge Infrastructure (CNKI) and Wanfang Data, for studies published from January 2005 to January 2026. The search strategy incorporated the terms “dose comparison,” “linear accelerator,” and “CSI,” along with their respective free-text alternatives. After initially removing duplicate studies, the titles and abstracts of the remaining records were reviewed to exclude other types of literature such as reviews, case reports, and news. The full texts of the selected articles were reviewed, and their references were also examined to ensure completeness. Data extraction and evaluation were independently conducted by two researchers (KHH and LLL), with any discrepancies resolved through discussion. If discrepancies still existed after discussion, a third researcher must evaluate independently.

2.2. Inclusion and exclusion criteria

In this study, all included studies were guided by the PICOS (Participants, Intervention, Comparison, Outcomes, and Study Design) framework, as follows: (1) Participants (P): The studies enrolled patients undergoing CSI who received radiotherapy with pathological confirmation of their diagnosis. (2) Intervention (I): The experimental group received radiotherapy using O-ring linacs. (3) Control group (C): The control group received radiotherapy using C-arm linac. (4) Primary outcome (O): The primary outcomes assessed were dose to specific regions. These included mean dose (D_mean) and D_max of PTV and D_max values of organs at risk (OARs). (5) Study Design: The studies incorporated both randomized controlled trials (RCTs) and observational designs.

Studies were excluded based on the subsequent criteria: (1) Review articles, case reports, abstracts and correspondence; (2) Insufficient research quality or considerable risk of bias; (3) Lacking appropriate data for aggregation.

2.3. Quality evaluation

The Newcastle-Ottawa Scale (NOS) was used to evaluate bias in nonrandomized research, addressing three essential aspects: selection, comparability, and outcomes [17]. The assessment employed a scoring system with a maximum of 9 points, distributed as follows: 4 points for selection criteria, 2 points for comparability, and 3 points for outcomes. Studies scoring 6 points or higher were considered to be of good quality [18].

2.4. Statistical analysis

A paired meta-analysis was performed using Stata 18.0 to compare studies on similar treatment pairs. The standardized mean difference (SMD) which was explained in Equation 1 and 95% confidence interval (CI) were used to evaluate continuous outcomes. The Cochrane q-test and the I² statistic were used to assess heterogeneity, indicating the extent of total variation attributed to heterogeneity instead of random error. A fixed-effect model (FEM) was applied if the p-value for the q-test was greater than 0.10 and the I² statistic was less than 50%. Alternatively, a random-effects model (REM) was used for data with substantial heterogeneity. p < 0.05 was considered statistically significant [19], [20]. The subgroup meta-analysis was also performed by dividing O-ring linacs into HR and DLM-linac.

$$\mathit{SMD}=\frac{\overline{X_1}-\overline{X_2}}{{\mathit{SD}}_{\mathit{pooled}}}$$ (1)

Where $\overline{X_1}$ and $\overline{X_2}$ represent the means of the two groups being compared, and ${\mathit{SD}}_{\mathit{pooled}}$ is the pooled standard deviation of both groups.

3. Result

3.1. Study selection and characteristics

After removing duplicates, initial searches identified 323 primary studies. After excluding letters, reviews, and conference proceedings, 155 articles were screened based on titles and abstracts, reducing the number to 28 studies. Comprehensive full-text evaluations led to 16 studies excluded due to the exclusion criteria, resulting in the inclusion of 12 peer-reviewed studies. The selection process is shown in Fig. 1.

Fig. 1.

Flow chart of the search process for the meta-analysis.

This meta-analysis included 12 studies involving 87 patients who had received CSI using different treatment modalities. All included studies were deemed high quality according to the NOS. Table 1 provides a detailed overview of the 12 studies included in this analysis. If a singular study reported multiple data sets, each set must be subjected to independent analysis.

Table 1.

Characteristics of inclusive articles.

Study	Journal	Linac types C (C-arm linac) H (helical radiotherapy) D (DLM-linac)	Year	Dose (Gy)	Monitor units	Treatment time(min)	NOS scale score
N.Papanikolao [33]	Radiother Oncol	C H	2005	36	/	9 27	8
C. Kunos [34]	Technol Cancer Res Treat	C H	2008	23.4	/	/	7
A. B.Luque [35]	Rep Pract Oncol Radiother	C H	2015	35.2	/	4.7 20	8
D. SHARMA [36]	Br J Radiol	C H	2009	35	/	6.5 21.8	8
Ö. Şenkesen [31]	Radiat Oncol	C H	2024	36	/	/	7
E. Seravalli [37]	Acta Oncol	C H	2018	36	/	/	7
S. Biswa [38]	J Cancer Res Ther	C D	2023	23.4	1328.1 1319.3	24 20	9
T.Stroubinis [39]	Adv Radiat Oncol	C D	2023	36	1327.3 1581.7	4.1 2	9
Nan Jiang [40]	Chin J Med Instrum	C H	2018	36	1128 7381.2	/	8
Helong Wang [41]	J Pract Oncol	C H	2017	36	875.1 12,868	8.6 15.1	9
Meiling Yang [42]	Chin J Clin Oncol	C H	2014	36	/	10 30	8
B. Sarkar [25]	Sci Rep	C D	2023	23.4	/	/	7

3.2. Dose comparisons of PTV and OARs between different linac types

D_mean and D_max values of PTV for O-ring and C-arm linacs demonstrated significant heterogeneity (p < 0.10, I² > 50%). The REM analysis revealed that the SMD of D_mean was −0.53 (95% CI: −1.57 to 0.50, p > 0.05) and the SMD of D_max was −0.14 (95% CI: −0.61 to 0.32, p > 0.05). These findings suggest no significant differences between the two types of treatment modalities.

D_max values of OARs for O-ring and C-arm linacs are shown in Table 2, except for the lung and thyroid, other OARs all demonstrate significance between different paired-analysis. Forest plots of D_max to eye and heart are shown in Fig. 2A and Fig. 2B, respectively. Significant lower heterogeneity can be found in the subgroup of DLM-linac compared to the subgroup of HR from Fig. 2. And other OARs such as lung (Fig. S3), thyroid (Fig. S4), kidney (Fig. S5), liver (Fig. S6) and esophagus (Fig. S7) are all shown in the supplementary file. Significant difference was revealed in the D_max values of heart (Fig. 2B) and liver (Fig. S6) between O-ring and C-arm linacs. However, no significant difference was revealed in the subgroup analysis of DLM-linac vs C-arm linac which meant comparing HR to C-arm, tremendous advantage emerged in lower dose to heart and liver (Table 2). Similarly, in the subgroup analysis, D_max values of the eye and esophagus (DLM-linac vs C-arm linac); heart, liver and kidney (HR vs C-arm linac) revealed significant difference. From the above results, different subgroup classifications were the main origin of heterogeneity.

Table 2.

Summary of Meta-Analysis Data for Organs at Risk.

OARs	O-ring vs C-arm			DLM-linac vs C-arm			HR vs C-arm
	SMD	95%CI	p	SMD	95%CI	p	SMD	95%CI	p
eye	−0.86	(−2.30, 0.59)	> 0.05	−0.61	(−1.22, −0.01)	0.045	−1.86	(−5.93, 2.20)	> 0.05
heart	−1.53	(−2.53, −0.52)	0.003	−0.40	(−0.83, 0.03)	> 0.05	−2.73	(−4.47, −0.98)	0.002
lung	−0.38	(−0.83, 0.07)	> 0.05	−0.26	(−0.70, 0.18)	> 0.05	−0.62	(−1.59, 0.36)	> 0.05
thyroid	−0.02	(−1.01, 1.05)	> 0.05	0.54	(−0.79, 1.86)	> 0.05	−0.99	(−2.12, 0.15)	> 0.05
kidney	−0.88	(−2.00, 0.23)	> 0.05	0.22	(−0.72, 1.16)	> 0.05	−2.10	(−4.12, −0.08)	0.042
liver	−1.33	(−2.29, −0.37)	0.007	−0.39	(−0.85, 0.04)	> 0.05	−2.51	(−3.53, −1.49)	< 0.001
esophagus	−0.53	(−1.59, 0.52)	> 0.05	−0.45	(−0.89, −0.01)	0.044	−1.99	(−9.11, 5.13)	> 0.05

Fig. 2.

Forest plot of D_max for the Eye(A), Heart(B) from stata 18.0 software.

3.3. Treatment time and monitor unit (MU)/ prescription dose

Fig. 3A illustrates the treatment time for different linac types. The average treatment time for C-arm linac was 3–5 min per treatment isocenter, while HR treatments typically 20–30 min. DLM-linac had the shortest treatment time, with a mean value of 2 min per treatment isocenter.

Fig. 3.

Treatment time(A) and MU/ prescription dose (B) for different linacs.

Only several enrolled studies reported MU across different treatment types. Given the diverse prescription dose used, the relative MU, namely MU/ prescription dose was selected for comparison. The value of MU/ prescription dose for HR was 6 ∼ 10 times higher than that in C-arm linac, whereas DLM-linac exhibited a lower MU/ prescription dose compared to C-arm linac, as shown in Fig. 3B.

4. Discussion

This study provided valuable empirical references by systematically analyzing and comparing the dose difference of C-arm and O-ring (including DLM-linac and HR) linacs in CSI treatment. Our findings indicated significant differences among various types of linacs in terms of dose distribution, treatment time, and dose delivered to OARs. Firstly, this meta-analysis showed no significant difference between O-ring and C-arm linacs for dose to the PTV. Although different prescription dose could affect the treatment outcomes which were proved by a previous preclinical model of CSI [21]. The key reason was that the relative prescription for PTV was identical, and a more uniform dose distribution for PTV was preferred in clinic which was consistent with previous studies, highlighting the improvements in dose uniformity offered by modern radiotherapy techniques [22], [23].

Furthermore, the O-ring linac groups demonstrated significant advantages in terms of lower doses to OARs. Specifically, the significant advantages of lower OARs dose were shown in O-ring linac groups which could bring clinical benefits to the patients. D_max values of the heart and liver were significantly lower in the O-ring group compared to the C-arm group. In the subgroup meta-analysis of the O-ring group, the D_max values of the eye and esophagus in DLM-linac were significantly lower than those in the C-arm group with statistical significance. Nevertheless, the D_max values of the heart, liver, and kidney were lower in HR compared to the C-arm group with statistical significance. These distinct dose benefits of DLM-linac and HR for different OARs may be attributed to their differing dose modulation techniques. Compared to the C-arm linac, DLM-linac achieved better dose distribution through a faster speed (5 cm/s vs. 2.5 cm/s) and lower leakage of the new version dual-layer multi-leaf collimators. In contrast, HR achieved this through continuous couch movement in a helical delivery during treatment. As a result, HR exhibits significantly higher MU and longer treatment time. These extended treatment durations required greater patient compliance and reduced treatment efficiency. Additionally, treatment plans with higher MU were associated with a greater likelihood of secondary carcinogenesis [15], [16]. Despite its effectiveness for CSI treatment, these disadvantages were notable in HR. In brief, more significant dose advantages were demonstrated in HR whereas higher treatment efficiency was demonstrated in DLM-linac. Other study reported that higher beam-on time revealed in HR (median beam-on time: HR = 11 min, VMAT = 5.49 min) [24]. Low heterogeneity was observed in certain OARs of DLM-linac following subgroup meta-analysis, suggesting a more stable treatment plan quality compared to HR [25], [26]. In this study, the dose delivered to six of the seven OARs demonstrated reduced heterogeneity compared to HR. However, high heterogeneity remained in the dose delivered to the kidney and thyroid, potentially attributable to the limited number of cases and studies available for analysis. Integrated with a golden-beam-data model and a fast converged optimization algorithm, DLM-linac significantly reduced dependence on the planner's experience [27], [28]. In other words, treatment plans based on this system demonstrated greater consistency.

In comparison to C-arm linacs, the dose advantages observed in the DLM-linac subgroup are less pronounced than those of HR. Nevertheless, two significant advantages of the DLM-linac merit attention. First, it offers shorter treatment time. Second, auto-feathering algorithm installed in the new version of treatment planning system for dose calculation which enhances the robustness of complicated treatment plans. These benefits stem from a more rapid treatment workflow, faster gantry speed, and an auto-shifting couch (two isocenters within 10.5 cm), which collectively enhance treatment efficiency. A key concern in CSI treatment using C-arm linac is the ambiguous dose at each field border. To ensure accurate and reliable dose distribution at these borders, medical physicists typically designed a uniform dose overlap extending 15 cm in the axial direction. In contrast, the new auto-feathering algorithm automatically accounts for dose distribution across multi-isocenter fields, effectively mitigating dose discrepancy across the inter-fraction treatment.

More and more clinical trials of proton CSI versus photon CSI demonstrated that proton CSI was superior in respect of progression-free survival and overall survival [29]. Proton CSI should be considered when available [30]. Another study revealed proton CSI stood out as the most efficient approach, closely followed by HR in terms of achieving superior target coverage and OARs protection [31]. In view of insufficient proton equipment, VMAT CSI represents a feasible treatment with more widespread availability [32].There are the following limitations in this study: (1) The number of studies included is limited. Only 12 studies were enrolled. Many factors with heterogeneous designs (different prescriptions, treatment planning systems, and endpoints) could induce potential bias. (2) Since DLM-linac was introduced clinically in 2017, only a few comparative studies have been reported. Some studies based on HR were performed much earlier than DLM-linac, which could also induce a bias. (3) Although dose advantages were discussed in this study, the long-term clinical outcomes and toxicity of these patients have not been thoroughly investigated and clinical benefits were not established. It is interesting to investigate the long-term clinical outcomes in future research. (4) Different types of VMAT were not discussed in depth. There are two prevailing techniques: full arc versus partial arc delivery. These approaches result in significantly different dose distributions and are highly dependent on the arc length used in the spinal PTV. There were reported that partial arc plans could not offer a significant dose reduction for delineated organs compared to full arc plans, except for bilateral kidneys. (5) Various statistical methods for treatment time of different linac types may lead to potential bias. Some studies only counted treatment time including Beam-on time; conversely, other studies counted treatment time from patient setup to Beam-off.

In conclusion, both O-ring and C-arm linacs are effective modalities for CSI treatment, with no significant differences in dose to the PTV when comparable doses are delivered to the lungs, thyroid, and esophagus. However, O-ring linacs, particularly the DLM-linac, offer better protection for critical organs such as the eyes and esophagus, in situation of reduced treatment time and robust dose distribution. In summary, clinicians should select the most suitable modality for CSI patients when balancing dose advantage and treatment efficiency.

Data availability

All datasets can be available from the corresponding author on reasonable request.

Funding

The present study was supported by the Guangdong Province Basic and Applied Basic Research Foundation Regional Joint Fund Regional Cultivation Project (Yuehui Joint Fund, Project No : 2023A1515140145), the Science and Technology Innovation Team Project of Huizhou Science and Technology Bureau (Project No : 2023EQ050012) and the Strong Foundation of Scientific Research Program of Guangzhou Medical University (Project No. 2024SRP215).

Ethics declarations

Ethics approval and consent to participate.

This study, which used de-identified, publicly available data, was exempt from additional institutional review board approval and did not require additional consent.

Consent for publication

Not applicable.

CRediT authorship contribution statement

Yaotao Li: Writing – original draft, Software, Formal analysis, Data curation, Conceptualization. Guozi Yang: Methodology, Investigation, Funding acquisition. Liangliang Lei: Software, Resources, Methodology. Kunhai Huang: Visualization, Software, Resources. Yanling Liao: Methodology, Investigation, Data curation. Hongzhi Liu: Software, Formal analysis, Data curation. Ran Tang: Visualization, Software, Resources. Anyan Gu: Writing – original draft, Supervision, Software. Yuanyuan Li: Visualization, Software, Methodology. Suyan Bi: Validation, Software, Methodology. Zhuocheng Li: Resources, Formal analysis. Yu Wu: Writing – review & editing, Writing – original draft, Visualization. Zhenyu Pan: Writing – review & editing, Writing – original draft, Supervision, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Xingru Sun: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: