In this retrospective analysis, we demonstrated clear relationships between specific dose metrics for healthy tissues and the risk of late complications following Ru-106 brachytherapy. Radiation dose dependence was found for some of the endpoints (visual acuity deterioration, maculopathy, optic neuropathy, vascular obliteration, and cataracts) but not for others (ocular hypertension and retinal detachment). Clinical factors (optic disc–tumour distance, age at treatment, largest base dimension, and pre-treatment visual acuity) also correlated with outcome for some endpoints. Detailed dose analyses were performed, utilising full dose area/volume histograms extracted from retrospectively recreated treatment plans with 3D image guidance.
We found a strong correlation between the risk of visual acuity deterioration and the area of retina receiving 10 Gy (retina A10Gy
), with a 50% risk of visual acuity deterioration when 20% of the retina received 10 Gy. Furthermore, the risk correlated with various macula dose metrics (macula A20Gy
and macula A80Gy
). These findings were in accordance with those of Aziz et al., who demonstrated total fovea dose as the most significant variable associated with an increased risk of visual acuity loss in 311 patients [10
]. Additionally, Heilemann et al. investigated visual acuity loss in 45 patients and found retina D2%
as the main risk factor [11
]. However, neither of them systematically explored a wide range of dose metrics.
Previous works have identified specific clinical variables as predictors for visual acuity loss, but reports are diverse and possibly contradictory, emphasising the complex mechanisms of visual acuity deterioration following Ru-106 treatments. Thus, the specific underlying causes remain not fully understood, and radiation-induced visual acuity deterioration might originate from several discrete pathophysiologies. Isager et al. investigated visual outcomes for 55 patients, and found tumour height and the largest base dimension as the most important risk factors for visual acuity deterioration, but they did not perform multivariate analyses, nor did they include dose in their analysis [12
]. According to Bergman et al., initial visual acuity was main risk factor along with the distance from the tumour to the fovea in a study with 579 patients. However, they did not consider dose either [13
]. Additionally, Damato et al. reported on 458 patients and also found initial visual acuity and tumour location as risk factors for visual acuity deterioration [5
]. These findings were in line with the present study, in which optic disc–tumour distance and pre-treatment visual acuity were important clinical predictors for visual acuity deterioration. However, dose was, contrary to our results, not significant in the multivariate analysis by Damato et al.
Risk factors for specific late complications have also been previously reported. Tagliaferri et al. recently found tumour location as the strongest risk factor for radiation-induced maculopathy based on a study with 197 patients [14
]. They did not find any significance of dose in multivariate analysis. However, the location of the tumour might be strongly correlated with the macular dose, and thereby have worked as a surrogate for the effect of dose in the multivariate analysis. Summanen et al. found a 30% five-year risk of developing radiation-induced maculopathy after Ru-106 treatments based on 100 patients [15
], while Naseripour et al. reported a five-year risk of maculopathy of 20% based on 51 patients [16
]. This is slightly less than the 45% five-year risk found in our work. This could possibly be explained by a lack of consensus in the reporting of maculopathy. Finger et al. have established guidelines for general strategies in the reporting of retinopathy [17
]. Extension to other radiation-induced late complications remains highly needed.
The reported risk of radiation-induced optic neuropathy varies considerable throughout the literature. According to Naseripour et al., the five-year risk was 40% (compared to 32% in our series), but no variables were significantly associated with the risk of developing optic neuropathy in their multivariate analysis. Summanen et al. reported a five-year risk of 12% and according to their study, the optic disc–tumour distance was the strongest predictor in radiation-induced optic neuropathy. This corresponds to the findings in our study. Furthermore, we found various optic disc metrics (optic disc V50Gy and V20Gy) as important risk factors for radiation-induced optic neuropathy.
Post-treatment vascular obliteration is to some extent expected after treatments with brachytherapy, especially for vessels near the tumour and thus near the plaque. However, only a limited number of studies describe predictors for vascular obliteration. Rouberol et al. found that the risk decreased with age and anterior location, but they used a prescription scheme that deviated from current standard regimens [18
]. We found optic disc V20Gy
as the only variable associated with increased risk.
Summanen et al. reported a five-year risk of radiation induced cataract of 37% [15
]. They found tumour height as the strongest risk predictor. Naseripour et al. demonstrated a five-year risk of 38%, but did not report any prognostic factors [16
]. The lens has previously been reported as a fairly radiosensitive structure [19
]. However, it should be kept in mind that cataracts are often a treatable condition, and sparing of the lens is thus less essential in the treatment optimisation.
The effect of dose on post-treatment retinal detachment has been demonstrated in a recent study by Heilemann et al. They found a strong correlation with both retina D2%
. However, our study did not find any support for this. Retinal detachment was the least frequent late complication in our series. Several patients had retinal detachment at the time of diagnosis as a result of tumour exudation. While exudation diminishes as the tumour shrinkages due to Ru-106 brachytherapy [20
], the retina fastens accordingly.
Published data elucidating the effect, if any, of dose rate on tumour control probability and late complications after Ru-106 brachytherapy is limited. According to Naseripour et al., a low dose rate did not have any significant negative effect on local tumour control [16
]. Furthermore, Fili et al. found that dose rate was not associated with the risk of secondary enucleation. However, they suggested that longer treatment times (and thus a lower dose rate) might be associated with increased ocular irritation [21
]. According to Mossböck et al., a high dose rate (as seen with shorter treatment times) might furthermore be associated with increased late complications and visual acuity deterioration [22
]. We included overall treatment time in the selection process to account for any dose rate effect, but we found no association with any of the late complications.
We only examined a single plaque source in the current study. However, compared to iodine-125 (I-125), Ru-106 generally causes fewer late complications [23
]. Miguel et al. found the type of plaque to be an important risk factor for optic neuropathy after I-125 brachytherapy, along with age, the largest base dimension of the tumour, and dose to the optic nerve [25
]. Furthermore, they found a 43% five-year risk of retinal detachment, with patient age at treatment and size of the plaque as important risk factors. Thus, plaque types utilising different isotopes might see different dose dependence than what has been observed here.
We estimated the dose delivered to the specific structures based on full dose area/volume histograms extracted from retrospectively recreated treatment plans. This approach used post-treatment retinographies to assess the actual plaque position (and hence the actually delivered dose) rather than an ideal pre-treatment plan. To the best of our knowledge, no previous studies have used 3D image guidance along with relevant clinical variables to perform normal tissue complication probabilities for choroidal melanoma patients.
Lasso statistics were used for variable selection, allowing for the consideration of a wide range of possible predictors; non-informative variables were excluded, while the appropriate predictors associated with the risk of each specific endpoint were included in the subsequent Cox regression analysis. This is a well-established method to manage large numbers of potential explanatory variables, but it is important to recognise that it tends to have difficulties when factors are strongly correlated [26
]. Even though the cohort size was relatively large, some of the late toxicities had a limited number of events (retinal detachment and ocular hypertension). This could potentially explain why the selection process was unsuccessful for these late complications. Furthermore, it is important to recognise that the retrospective nature of the treatment planning and the toxicity scoring could potentially introduce bias.
Prescription procedures for Ru-106 treatments remain controversial [20
]. Some centres use an apical prescribed dose of 85 Gy [1
], some use 100 Gy to the apex [2
], while others use at least 130 Gy to the apex along with restricted tumour base doses of at least 700 Gy [27
]. Eccentrically located plaques are also reported frequently in the literature, and studies have indicated that this strategy can spare healthy tissue in some cases [5
]. Further studies have suggested that eccentric positioning can lead to underdosage of the tumour [28
]. We recently reported that the minimum dose to the tumour is crucial for local control (Espensen et al., submitted manuscript(companion paper)), and compromises on tumour coverage should be made with caution and assessed relative to the potential benefits of visual acuity preservation and reduced probability for late complications.