Uveal melanoma (UM) is the predominant primary intraocular tumor among adults, with an incidence rate of approximately six cases per million individuals [1]. Approximately 90% of UMs primarily affect the choroid. UM is significantly rarer in Asians, with an incidence rate of approximately 0.25 to 0.64 per million individuals annually, compared to 5 to 6 per million individuals in Western countries [2,3]. Phenotypic traits, such as blue or grey eyes and fair skin in the Caucasian population, predispose individuals to UM. However, the impact of these traits on survival rates and distant metastasis remains unclear [4,5].

Diverse therapeutic interventions are available, encompassing enucleation and approaches aimed at preserving the eye [6,7]. Historically, enucleation has been considered the gold standard treatment for UM. Nevertheless, a shift in perspective has occurred following the findings of the Collaborative Ocular Melanoma Study (COMS), which demonstrated no discernible survival advantage or discrepancy in the rate of distant metastasis between enucleation and radiotherapy (RT) [8,9].

Various RT techniques are available for the preservation of the eye in the management of UM. Plaque brachytherapy is deemed appropriate for small tumors with heights ≤ 6 mm when employing ruthenium-106 [7,10]. This necessitates invasive surgical procedures, frequently resulting in complications arising from the impact of radiation on neighboring tissues. Proton beam therapy (PBT) has been frequently used for both small and large tumors due to its advantages of sharp dose fall-off and little radiation exposure to adjacent tissues. However, surgical intervention is unavoidable, as a radiopaque target must be inserted onto the surface of the eye to achieve precise beam targeting. Moreover, considering the limited availability of PBT facilities than that of photon RT, it is imperative to acknowledge that accessibility to PBT is compromised, presenting a potential impediment to its universal availability on demand.

CyberKnife (Accuray), a modality for photon stereotactic surgery (SRS)/stereotactic body radiation therapy (SBRT), requires relatively lower installation and operational costs and can be more easily integrated into existing radiation facilities. It is capable of treating UM without the need for surgical intervention and reduces radiation exposure to adjacent tissues. Given the rarity of UM in Asian populations, which has led to a scarcity of region-specific data on photon SRS/SBRT, this investigation aimed to examine the clinical efficacy of CyberKnife treatment, emphasizing parameters such as survival rates, occurrence of distant metastasis, and local tumor control in Asian populations. Additionally, this study evaluated the rates of ocular complications and visual outcomes, recognizing the preservation of visual acuity without complications as the primary objective of eye-preserving therapy for UM.

We conducted a retrospective analysis of 77 consecutive patients with UM who underwent SBRT using CyberKnife at Yonsei Cancer Center from September 2015 to October 2021. Larger UMs with a height > 6 mm are treated using CyberKnife in our institution, whereas small UMs with a height of ≤ 6 mm are treated using plaque brachytherapy [10]. The inclusion criteria comprised individuals meeting the following conditions: (1) patients with UM who had undergone SBRT using CyberKnife and (2) those with an Eastern Cooperative Oncology Group performance status ≤ 2. Patients were excluded if they met any of the following criteria: (1) the presence of distant metastasis at the initiation of RT, (2) a history of prior or concurrent malignancies, (3) a record of radiation to the orbital area, (4) incomplete RT (biologically effective dose < 40 Gy), and (5) absence of any follow-up after RT. Based on these criteria, 13 patients were excluded and 64 patients were included in the study.

Each patient was positioned supine and immobilized using a thermoplastic mask, while maintaining a fixed-straight gaze. An acrylic frame was attached to the mask to secure the camera. During this time, the patient’s gaze was directed towards the center of the camera lens, and a gaze-tracking program, which was developed at our institution, was activated. This program tracks the position of the iris in real-time and returns the central coordinates of the iris, which can be continuously monitored through the camera. If the tracking proved stable, the simulation computed tomography (CT) scan was conducted accordingly. The CT images were obtained with a 1-mm slice thickness. Radiation oncology experts performed three-dimensional segmentation of the UM using MIM Software ver. 6.5.8 (MIM Software Inc.). The gross tumor volume (GTV) was defined based on magnetic resonance imaging (MRI) findings, while bilateral organs at risk (OAR)—optic chiasm, lacrimal glands, lenses, optic nerves, and orbits—were delineated on CT images. A 2 mm margin was added to the GTV to determine the planning target volume (PTV), and a total dose of 60 Gy in four fractions was prescribed to the 75% isodose line. Our planning approach focused on ensuring adequate PTV coverage, according to the prescribed dose, rather than on reducing the dose to the OAR. Treatment plans were derived using a non-isocentric inverse algorithm and the treatment was administered every other day. During the actual treatment, the stability of the gaze tracking was reassessed. Once the patient’s iris movement was determined to be minimal and stable, a 2 mm margin was created based on the current coordinates as a threshold for deviation of iris movement. Iris movement was tracked real-time and if the patient’s gaze deviated outside the predetermined 2 mm margin, the beam was manually interrupted with a delay of approximately 2 seconds due to the manual operation. Ophthalmic examinations were performed at baseline, 3 months after RT, and every 6 months thereafter. Visual acuity assessments, routine ophthalmic examinations, and measurement of tumor base dimensions and height using standardized B-scan ultrasonography or MRI were performed at each follow-up session. The treatment response was evaluated using the Response Evaluation Criteria in Solid Tumors ver. 1.1, and the best response was documented after treatment.

Data analysis was conducted using SPSS ver. 25.0 (IBM Corp.). Survival outcomes, including the local failure-free rate (LFFR), distant metastasis-free rate (DMFR), progression-free survival rate (PFS), and overall survival rate (OS), were assessed using the Kaplan-Meier method and log-rank test. All rates were defined as the duration from the date of diagnosis to event occurrence. The Cox proportional hazards model was used for the univariate and multivariate analyses of independent prognostic clinical factors for each survival type. Variables significantly associated with survival rates in univariate analysis were selected as candidates for multivariate analysis. The Kaplan-Meier method and chi-squared test were used to investigate factors influencing visual outcomes, and Cox regression analysis was performed to identify risk factors associated with vision loss. Radiation-induced toxicities were scored according to the Common Terminology Criteria for Adverse Events ver. 5.0.

The primary goal for treating UM is local control of the tumor and prevention of recurrence and distant metastasis. The COMS results demonstrated no significant difference in the OS and risk of distant metastasis when comparing the outcomes of enucleation with those of various eye-preserving RT methods [11,12]. In a meta-analysis of 49 articles [13], local treatment failure rates for plaque brachytherapy, photon-based SBRT, and charged particle radiation therapy were 9.5%, 7.9%, and 4.2%, respectively. These findings have promoted a shift towards eye-sparing treatments, resulting in a discernible increase in the number of institutions equipped to provide such therapies, especially particle therapies over time [14,15]. While many patients with UM undergo PBT, geographical disparities in the availability of PBT facilities,—along with variations in insurance coverage for this costly treatment,—highlight the practical necessity for treatment with and research on more accessible options like plaque brachytherapy, photon SRS/SBRT using Gamma Knife, CyberKnife and other photon-based RT. As previously mentioned, UM is significantly rarer in Asia than in Western countries [2,3,16] and reports on photon SRS/SBRT of UM in Asians are scarce. This study is particularly important because it focuses on large UMs where brachytherapy cannot be performed, enhancing its importance due to the rarity of such cases. Therefore, our research fills a critical gap by providing valuable data on photon SRS/SBRT for large UMs in Asian populations.

Plaque brachytherapy is generally suitable for treating small tumors, but it tends to be less effective for large tumors. As the tumor size increases, the technical difficulties of plaque insertion escalate, leading to an increased risk of inadequate radiation coverage of the tumor periphery [7,17-19]. PBT, recognized for its efficacy across varying tumor sizes, provides several therapeutic advantages over photon therapy. Beyond their higher relative biological effectiveness compared with that of photons, the Bragg peak effect allows proton beams to exhibit a narrower penumbra, enabling the delivery of higher energy with greater precision and fewer complications [7,20-22]. Defining the penumbra for photon beams can be challenging, as it varies with field size, depth, and photon energy while generally tending to be broader compared to that of proton beams. The photon treatment planning system focuses on optimizing and generating SBRT plans with a steep dose fall-off gradient outside PTV. To achieve this goal, the maximum dose inside the PTV was defined to be greater than 120% as usual, so that the dose fall-off gradient can be steep. The CyberKnife system utilizes a robotic arm that offers a high degree of flexibility in beam delivery, allowing for non-coplanar and highly conformal treatment plans. These non-coplanar and non-isocentric beam configurations can further promote steepness of the dose gradient, contributing to the successful treatment of UM.

The outcomes and complications of UM treated with PBT and photon SRS/SBRT, as documented in the literature, are summarized in Table 5. Local control rates for PBT range from 86% at 2 years to 96.8% at 5 years, whereas photon SRS/SBRT demonstrates a broader range, with 5-year local control rates varying between 73% and 96%. Our data showed a 3-year LFFR of 89.5%, which falls within the range when compared to those reported in the studies summarized in Table 5. This could be attributed to the larger tumor dimensions in our dataset compared to previous data. In the study by Muller et al. [23], the median tumor height was 6.0 mm, compared to 8.4 mm in our study, with LFFR of 96% and 89.5%, respectively. Furthermore, Dunavoelgyi et al. [24] documented a median tumor volume of 0.27 cm³, compared to 0.95 cm³ reported in the current study, with an LFFR of 95.9% and 89.5%, respectively. Tumor size is recognized as a critical determinant of local tumor control across various investigations. Nonetheless, we only applied SBRT to large tumors, thus no significant results pertaining to tumor size were observed.

In this study, the 3-year DMFR was 70.5%. This result is slightly lower than the findings of PBT or other photon-based studies (Table 5), regardless of the success of local control [25-27]. Of the predictive factors affecting treatment outcomes, a larger tumor diameter was associated with a significantly worse prognosis for DMFR and OS in the multivariate analysis.

Visual outcomes in patients with UM are generally poor. In our study, only 6.3% (n=4) of patients maintained visual acuity ≥ 20/200 after treatment among the 41 patients with visual acuity ≥ 20/200 at baseline. Previous studies reported final visual acuity ≥ 20/200 in 15.9%-70.0% of patients treated with PBT and 8.8%-40.0% with photon-based SRS/SBRT (Table 5). The relatively inferior visual outcomes may have been due to the selected population of large tumors in our study. In our analysis, better initial visual acuity (≥ 20/40) was an independent significant factor associated with retaining visual acuity ≥ 20/200. This suggests that patients with better baseline vision have better visual outcomes after treatment. These findings highlight the importance of early UM detection to maximize the chance of preserving visual function, a key goal of eye-preserving treatments.

Our analyses of the complications associated with RT for UM were consistent with those reported by previous literature [28], with enucleation rate ranging from 4.1%-26.0% with PBT, and 9.3%-27.0% with photon SRS/SBRT, respectively (Table 5). Notably, although the enucleation rate in this study was 20.3% (n=13), the majority of enucleations were performed due to complications (15.6%, n=10), with only 4.7% (n=3) attributed to tumor progression. This implies that if access to PBT or carbon ion therapy were to improve, it could potentially reduce complications and thereby significantly decrease the enucleation rate. In addition, in the univariate analysis of the enucleation-free rate in S5 Table, GTV and PTV were significant predictive factors of enucleation. This suggests that the enucleation rate may be influenced by the tumor size at the time of diagnosis, underlining the importance of early detection of UM.

This study has several limitations. One notable limitation pertains to the treatment planning process. During the planning process, only the optic chiasm, lacrimal gland, and optic nerve were contoured, while structures such as the ciliary body, macula, optic disc, and retina were not delineated. This was because our approach focused on ensuring adequate PTV coverage in accordance with the prescribed dose, rather than reducing the dose to these OARs. Table 5 describes the incidence of normal tissue toxicities among modalities. As the table shows, the incidence of neovascular glaucoma in this study was slightly higher than that of other studies. We may achieve better toxicity outcomes in the future if we delineate the ciliary body, macula, disc, and retina. Among the treated patients, two received a relatively low dose radiation because the tumor was located anteriorly; one received 56 Gy in 4 fx, and the other received 48 Gy in 4 fx. These patients were at a substantial risk of developing neovascular glaucoma; however, the occurrence of neovascular glaucoma was averted by administering a dose lower than the prescribed amount. The patient who received 56 Gy in 4 fx developed serous retinal detachment, vitreous hemorrhage, and cataract, whereas the patient who received 48 Gy in 4 fx developed only vitreous hemorrhage. Another major limitation attributable to this study’s retrospective design is the possible effect of selection bias. For instance, excluding patients who lack post-RT follow-up data could artificially increase the survival rates observed in the cohorts. Additionally, toxicity evaluations were possibly conducted less rigorously in this study than in prospective studies. This discrepancy could have led to the underestimation or inconsistent reporting of adverse effects associated with treatment, further complicating the interpretation of the study’s outcomes. Nevertheless, despite its retrospective nature, the study was meticulously conducted with thorough ophthalmic examinations, and the patients were treated and monitored following a uniform protocol at a single center. Another limitation is the rarity of the disease and the confinement of data collection to a single center, which resulted in a restricted patient sample, thereby imposing constraints on the study. This may have reduced the statistical robustness of the analysis. To strengthen the robustness of our findings, it would be beneficial to conduct future research with a larger sample size and longer follow-up, potentially through multicenter studies. This would allow for more comprehensive clinical data and enhance the validity of our conclusions.

This study demonstrates the potential of SBRT to achieve excellent local control rates in patients with large UMs. However, limitations, such as distant metastasis and adverse effects leading to enucleation, have been identified. Future research should focus on investigating novel treatment approaches aimed at reducing distant metastases and minimizing side effects.