Effects of cone versus multi-leaf collimation on dosimetry and neurotoxicity in patients with small arteriovenous malformations treated by stereotactic radiosurgery

Abstract

Purpose/objective:

Linear accelerator (LINAC) based stereotactic radiosurgery (SRS) for arteriovenous malformations (AVMs) is delivered with cone or multileaf collimators (MLCs), and favorable dosimetry is associated with reduced radionecrosis in normal brain tissue. This study aims to determine whether cones or MLCs has better dosimetric characteristics, to predict differences in toxicity.

Methods:

All patients treated for AVMs using LINAC SRS from 2003-2017 were examined retrospectively. Demographic data, volumes of normal tissue exposed to 12Gy (V12Gy[cc]) and 4Gy (V4Gy[cc]), maximal dose, and dose gradient were analyzed. Univariate and multivariate analyses were used to evaluate relationships between collimator type, dosimetric parameters, and toxicity. Propensity score matching was used to adjust for AVM size.

Results:

Compared to MLC, cones were independently associated with reduced V12Gy[cc] after propensity score matching (p=0.008) and reduced neurotoxicity (p=0.016). Higher V12Gy[cc] (p=0.0008) and V4Gy[cc] (p=0.002) were associated with increased neurotoxicity.

Conclusions:

Treating AVMs with cone-based SRS over MLC-based SRS may improve dosimetry and reduce toxicities.

INTRODUCTION

Cerebral arteriovenous malformations (AVMs) are abnormal connections of arteries and veins in the brain and are a leading cause of nontraumatic intracerebral hemorrhage. Stereotactic radiosurgery (SRS) is often employed to treat AVMs, especially in patients that cannot be treated effectively or safely with microsurgery. SRS is generally preferred over microsurgery for AVMs that are classified as Spetzler Martin Grade III or above(1, 2). However, there are a select group of Spetzler Martin Grade I and II AVMs that are referred for SRS. Referrals for Grade I lesions are often made for poor surgical candidates or by patient preference, while grade II lesions may be non-operative based on eloquence of location and deep venous drainage. The goal of SRS is to give a high dose of ionizing radiation with steep dose gradients to obliterate the AVM while sparing normal tissue(3). SRS has been noted to be particularly effective for small volume (diameter ≤3cm) lesions(4-6), with a favorable complication profile(7). Photon-based SRS can be delivered using Gamma Knife, which utilizes cone-based delivery, or linear accelerators (LINAC) which have the option of multi-leaf collimation or cone-based delivery. LINACs tend to be more widely available and more commonly used than Gamma Knife and have been shown to have similar obliteration rates(8, 9), however the dosimetric advantage of Gamma Knife over LINAC-based radiosurgery is controversial and hotly debated(10), with each likely being advantageous in various clinical settings.

Optimizing the dose distribution of SRS for AVMs is an area of increasing study, as treatments must balance the efficacy of AVM obliteration with long term toxicity. The age of AVM patients at diagnosis tends to be relatively young(11), so minimization of long-term neurological sequelae is of critical importance. Overall, higher marginal doses and higher maximal doses have been shown to increase the obliteration rate of AVMs(12-14). Toxicity of SRS is dependent on the dose distribution, including the target treatment volume, as well as the collateral volume of normal tissue exposed to radiation. Thus, a steep dose drop-off outside of the target volume is strongly desired. The volume of normal brain tissue exposed to 12Gy of ionizing radiation (V12Gy[cc]) has been associated with increased risk of radionecrosis and can be used to predict permanent neurological sequelae(15, 16). While less established, there is preliminary evidence that the volume of brain tissue exposed to doses as low as 4Gy (V4Gy[cc]) can change neurological activity and decrease arousal in animal models(17). In humans, higher V4Gy[cc] may also be associated with an increase in radiation necrosis, T2W signal intensity on MRI, and duration of steroid use(18).

In LINAC-based SRS with the most updated version of iPlan (Brainlab, Munich, Germany), circular cones and/or multi leaf collimators (MLCs) can be used to shape treatment fields either independently or together. Cones are used predominantly for treating small lesions because larger lesions require multiple isocenters, which translates to prolonged treatment times(19). Cones are able to apply a more focused dose in a circular distribution, which may lead to a greater maximal dose and AVM obliteration rate(20). However, it is unknown whether cones offer dosimetric and clinical advantages compared to MLCs in terms of toxicity in the treatment of AVMs. MLCs have previously been shown to offer better conformality than cones for irregular or large volumes, while cones afford steeper dose gradients(21). On LINAC-based platforms, MLCs shorten treatment times and are easier to design treatments with when compared to cones, especially for larger or complex-shaped AVMs.

The purpose of this study is to compare cones to MLCs with respect to dose gradient, maximal radiation dose, volume of normal tissue exposed to radiation, and toxicity. We hypothesized that dosimetric characteristics of cone-based plans, when elected for small AVMs, offers a steeper dose gradient advantage and may reduce the volume of normal tissue exposed to radiation, translating to lower toxicity.

METHODS

Patients

A database query of all patients treated with LINAC SRS for AVMs at our institution from 2003 to 2017 was conducted for this retrospective cohort study. Patients were included if they were at least 1 year of age, had an SRS plan available in Brainscan or iPlan software from which dosimetric data could be calculated, and had at least 1 year of follow-up after their initial treatment. Patients who received staged or boosted treatments, or patients with multiple intracranial AVMs, were excluded. This study was approved by the Institutional Review Board; informed consent was waived by the IRB due to determination of minimal risk.

Outcomes of interest

The primary outcome of this study was the incidence of post-SRS neurotoxicity. Neurotoxicity was defined as new or worsening neurological or radiation related symptoms not better explained by other causes. Asymptomatic radionecrosis or edema within the AVM target was not classified as neurotoxicity, as this is an intended effect of SRS. In contrast, asymptomatic radionecrosis outside of the target was classified as a toxicity. Symptomatic radionecrosis requiring treatment was classified as toxicity regardless of location. Toxicity was graded using the US Department of Health’s Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. Secondary outcomes were dosimetric parameters including volume exposed to 4Gy (V4Gy[cc]), volume exposed to 12Gy (V12Gy[cc]), maximal dosage (D_max) and ratio of volume exposed to 50% dosage to the planning target volume (R50). We collected clinical factors including age, sex, SM grade including size, eloquence, and venous drainage through medical record review. Treatment characteristics such as method of LINAC delivery, SRS dose, any salvage therapy and toxicity were recorded as well. Overall follow up was calculated from date of SRS to last follow up with MRI or MRA CT angiography.

Dosimetry and SRS Treatment Planning

The radiosurgical treatment techniques used in this study have been previously described(22). Briefly, SRS treatment planning consisted of fusion of brain MRI or MRA with planning CT scan by a radiation oncologist and vascular neurosurgeon. SRS was delivered using a Novalis Tx Linear accelerator (Varian Medical Systems, Palo Alto, California) with Brainlab ExacTrac (Brainlab, Munich, Germany) orthogonal kV imaging for anatomical alignment and high-definition MLC capabilities. Plans with either MLC or cone-based collimation were included, but not those utilizing both in a single session. The volume of the target lesion (GTV) and dosimetric parameters such as D_max, V4Gy[cc], V12Gy[cc], and R50 as a surrogate for dose gradient index were calculated using BrainLab software (Eaton 2018), with dose volume histograms for normal tissue being used to calculate V4Gy[cc] and V12Gy[cc].

Statistical analysis

Baseline characteristics and treatment details were summarized with descriptive statistics. The Shapiro-Wilk test was used to assess for normality of continuous variables. Covariates that lacked a normal distribution were described by median and inter-quartile range. For each cohort, dosimetric characteristics of treatment plans were compared by Wilcoxon rank-sum tests. Predictors of binary outcomes or continuous outcomes were evaluated by logistic regression or linear regression, respectively. Multivariable linear regression was performed to account for the effects of AVM volume on dosimetric endpoints. To account for the effects of AVM volume on the selection of SRS treatment techniques, propensity-score matching was performed followed by multivariable linear regression. Propensity scores were generated using a logistic model with SRS technique (cone versus MLC) as the dependent variable and AVM volume as the independent variable. Patients were matched 1:1 using optimal matching technique with a caliper distance limited to 25% of the pooled estimate of the common standard deviation of the logits of the propensity scores. Standardized mean difference was used to assess for covariate balance between the cohorts with a threshold of 0.1. All confidence intervals were reported at 95% and all p-values were reported as two-sided, with significance defined at a level of p<0.05. All statistical analysis was performed using SAS version 9.4 (Cary, NC).

RESULTS

Patient Characteristics

There were 63 patients who met inclusion criteria, with 22.2% (N=14) of patients treated using cone-based SRS, while 77.8% (N=49) of patients were treated using MLC-based SRS (Table 1). Median follow-up for the entire cohort was 970 days; median follow-up for patients who received cone-based SRS was 1357 days (range 354-4670) and median follow-up for patients who received MLC-based SRS was 919 days (range 123-4079) (p=0.4874). Patients were relatively young, with a median age of 38 years (range 4-75). The median AVM volume was 2.60 cm³. AVMs treated by cones had a median volume of 0.87 cm³, while AVMs treated by MLC had a significantly higher median volume of 3.03 cm³ (p=0.0002). The median marginal dose was 17Gy, and not significantly different between cohorts treated with cones and MLCs with median marginal doses of 16Gy and 18Gy respectively (p=0.6273).

Table 1.

Descriptive statistics of patients treated for AVMs with LINAC SRS

	Cone (N=14)	MLC (N=49)	P value	Overall (N=63)
Gender	7 female 7 male	26 female 23 male	0.8397	33 female 30 male
Age (years)	32 (IQR 13-48)	38 (IQR 21-53)	0.3340	38 (IQR 8-52)
Prescribed Dose (Gy)	16 (IQR 16-19)	18 (IQR 16-18)	0.6273	17 (IQR 16-18)
AVM Volume (cm3)	0.87 (IQR 0.17-1.7)	3.03 (IQR 2.0-7.6)	0.0002	2.6 (IQR 1.0-5.7)

Quantitative variables are reported as Median (IQR)

Single Predictor Analysis

The median D_max of all treated patients was 23Gy, and the median maximum dose was 22.8Gy for cones and 23.2Gy for MLCs (p=0.5462) (Table 2). The V4Gy[cc] of cone-based SRS was significantly lower than the V4Gy[cc] of MLC-based SRS (median 14.7 cm³ versus 58.5 cm³, p=0.0001). Likewise, the V12Gy[cc] for patients treated by cones was significantly lower compared to those treated with MLC-based LINAC radiosurgery (median 2.4 cm³ versus 8.9 cm³, p<0.0001). In addition, AVM volume was also significantly correlated with both V4Gy[cc] (unadjusted Beta 11.11, p<0.0001) and V12Gy[cc] (unadjusted Beta 1.52, p<0.0001). However, R50 was not significantly different between the cone cohort compared to the MLC cohort (median 2.9 versus 3.3, p=0.10). Additionally, conformity index was not significantly different between the cone cohort compared to the MLC cohort (median 1.34 versus 1.27, p=.5594).

Table 2.

Single predictor comparison of dosimetric parameters for cones vs MLCs

	Cone (N=14) Median (IQR)	MLC (N=49) Median (IQR)	P value
V4Gy[cc] (cm3)	14.7 (IQR 6.5-28.9)	58.5 (IQR 30.9-109.6)	0.0001
V12Gy[cc] (cm3)	2.4 (IQR 1.2-4.1)	8.9 (IQR 5.2-15.3)	<0.0001
R50	2.9 (IQR 2.5-3.5)	3.3 (IQR 3.0-3.7)	0.1034
Maximum Dose (Gy)	22.8 (IQR 21.4-28.0)	23.2 (IQR 21.3-24.6)	0.5462
Conformity Index	1.34 (IQR 1.12-1.53)	1.27 (IQR 1.16-1.38)	0.5594

Multivariable Analysis

Due to the unadjusted correlations detected between AVM volume and V4Gy[cc] and V12Gy[cc], multivariable linear regressions were performed to assess whether treatment type was independently associated with dosimetry despite confounding effects from AVM volume (Table 3). In the V12Gy[cc] multivariable model, cone-based treatments predicted a smaller V12Gy[cc] compared with MLC (adjusted Beta -2.92, p=0.01) independent of AVM volume which also predicted V12Gy[cc] (adjusted Beta 1.47, p<0.0001). However, in the V4Gy[cc] model, cone-based treatments were no longer predictive of V4Gy[cc] (adjusted Beta -12.16, p=0.26) when accounting for volume (adjusted Beta 10.99, p<0.0001).

Table 3.

Multivariable linear regression with and without propensity-score matching for AVM volume.

	V4Gy[cc]	V12Gy[cc]
	Beta, (95% CI), p-value	Beta, (95% CI), p-value
Propensity Matched Model (N=28)	R2 = 0.93	R2 = 0.97
AVM Volume	12.49, (11.07, 13.92), <0.0001	1.69, (1.57, 1.82), <0.0001
Cone treatments (versus MLC)	1.03, (-6.58, 8.65), 0.78	-0.94, (-1.62, 0.27), 0.008
Clinical Model Without Matching (N=63)	R2 = 0.83	R2 = 0.89
AVM Volume	10.99, (9.56, 12.27), <0.0001	1.47, (1.33, 1.62), <0.0001
Cone treatments (versus MLC)	−12.16, (-33.63, 9.32), 0.26	−2.92, (-5.24, -0.60), 0.01

Propensity-score Matched Subset Analysis

Furthermore, because AVMs treated by MLCs were significantly larger than AVMs treated with cone-based SRS (p=0.0002) and may have influenced the selection of treatment technique, propensity-score matched cohorts were created followed by multivariable linear regression for further validation. Fourteen patients treated by cone-based SRS were matched with fourteen patients treated by MLC-based SRS (Table 3). Within this cohort, single predictor analysis showed no significant differences between cone-based SRS and MLC-based SRS with regard to V4Gy[cc], V12Gy[cc], R50, maximum dose, and conformity index. However, in the V12Gy[cc] matched model, cone-based treatments continued to significantly predict for V12Gy[cc] (adjusted Beta -0.94, p=0.008) independently of volume (adjusted Beta 1.69, p<0.0001) (Table 3). However, in the V4Gy[cc] matched model, cone-based treatments were not associated with V4Gy[cc] (adjusted Beta 1.03, p=0.78) (Table 3).

Neurotoxicity

The incidence and grade of each toxicity observed after treatment are given in Table 4. 49.2% (N=31) of patients experienced at least one toxicity, with 41 toxicity events in total. The vast majority of toxicities were neurological, with non-neurological toxicity including just 1 case of alopecia. The most common toxicities were headache, contributing 13 cases, and clinically significant radionecrosis, contributing 8 cases. 30 toxicities were classified as grade 1, 14 toxicities were classified as grade 2, and 6 toxicities were classified grade 3. Grade 3 toxicities were only found in patients treated with MLCs-based SRS and included severe radionecrosis and seizures. Cone-based SRS was associated with a lower probability of neurotoxicity compared to MLC-based SRS (OR 0.205, 95% CI=0.051-0.826, p=0.016). Increasing V12Gy[cc] was correlated with a higher probability of neurotoxicity (OR 1.203; 95% CI=1.080-1.341; p=0.0008), and higher V4Gy[cc] was also correlated with a higher probability of neurotoxicity (OR 1.022; 95% CI=1.008-1.036; p= 0.002). Maximal dose, however, was not found to correlate with neurotoxicity (OR 0.955; 95% CI=0.809-1.126; p=0.58).

Table 4.

CTCAE toxicities of cone and MLC based LINAC SRS

	Cones			MLCs
Toxicity	Grade 1	Grade 2	Grade 3	Grade 1	Grade 2	Grade 3
Radionecrosis	N/A	0	0	N/A	5 (63%)	3 (37%)
Headache	1 (8%)	1 (8%)	0	8 (62%)	3 (23%)	0
Seizure	0	0	0	1 (25%)	0	3 (75%)
Weakness	0	0	0	0	3 (100%)	0
Paresthesia	0	0	0	1 (100%)	0	0
Tremor	0	0	0	1 (100%)	0	0
Confusion	0	0	0	1 (100%)	0	0
Amnesia	0	0	0	2 (100%)	0	0
Dizziness	1 (50%)	0	0	1 (50%)	0	0
Dysgeusia	0	0	0	1 (100%)	0	0
Vision changes	0	0	0	1 (33%)	2 (67%)	0
Hallucination	0	0	0	1 (100%)	0	0
Alopecia	1 (100%)	0	0	0	0	0

DISCUSSION

In cerebral AVMs treated with radiosurgery, cone-based SRS was associated with superior dosimetry and lower neurotoxicity compared to MLC-based SRS when corrected for size of the AVM. These novel findings suggest that cone-based SRS reduces the amount of normal brain tissue exposed to ionizing radiation when compared to MLC-based SRS and therefore correspondingly reduces neurotoxicity. Thus, in the setting of small AVMs that can be readily treated either by cone-based SRS or MLC-based SRS, this study indicates that cone-based SRS is preferable compared to MLC-based SRS. For larger lesions, cone-based dosimetric planning on linear accelerators is more complex and may require longer treatment times with multiple isocenters, consideration for cone-based radiosurgery using Gamma Knife or other modalities may be reasonable given that MLC-based radiosurgery may have dosimetric disadvantages.

This study showed that increasing V4Gy[cc] and V12Gy[cc] were both associated with the presence of new-onset neurotoxicity after radiosurgery, and that there were significantly fewer toxicities associated with cone-based treatments compared to MLC-based radiosurgery. The vast majority of toxicities were grade 1 or grade 2, although notably, the only grade 3 toxicities in this study were found in patients treated with MLCs. We considered that MLCs are generally used to treat larger AVMs, however, propensity score matching for AVM size followed by multivariable analysis showed that even when accounting for size, V12Gy[cc] is significantly higher for MLC-based radiosurgery. This close relationship we observed between dosimetry and toxicity has been validated in previous studies. For example, Flickinger et al found that symptomatic post-radiosurgery injury for AVM patients was significantly correlated with V12Gy[cc](16). Under QUANTEC guidelines, V12Gy[cc] in SRS should be kept under 5-10cc, when feasible, as the risk of radionecrosis rises over 20% for greater V12Gy[cc] volumes(15). Therefore, cone-based SRS may improve toxicity relative to MLC based SRS via reduction of V12Gy[cc].

Theoretically, circular cones offer a more focused dosage with a steeper gradient than MLCs (23). R50, being the ratio of the 50% isodose volume to the planning target volume, reflects the dose falloff outside the treated area. More rapid dose falloff has been associated with less histologic change in brain tissue surrounding the target(24) and would thus be expected to reduced toxicity. It is interesting to note that no significant difference in R50 between circular cones and MLCs was observed. This lack of significance may be due in part to an underpowered sample including only 14 patients treated with cone-based SRS. A larger sample size with prospective validation, therefore, is warranted to better delineate the differences, if any, in dose gradient between cones and MLCs.

Multileaf collimators are thought to more closely cover the field of irradiation than circular cones, and have been preferred for treating, larger, more geometrically complex lesions(25). While the results of this study support the use of cone-based SRS over MLC-based SRS to improve dosimetry and toxicity, it is important to note that cone-based SRS may be limited by the size of the stereotactic cones. Cones at our institution had a maximum diameter of 15mm, which limited the volume that could be efficiently treated with a single cone. Larger lesions would require multiple isocenters to achieve obliteration and may prolong treatment times beyond what is clinically feasible, in addition to being challenging to design dosimetrically with linear accelerators.

In the setting of larger lesions not amenable to cone-based LINAC SRS, Gamma Knife SRS presents a possible alternative solution, as it has been shown to compare favorably to LINAC with respect to both conformity index and dose falloff in treating brain metastases(26). Gamma Knife utilizes cones exclusively, affords excellent dose drop off and conformity using multiple individual fixed beams (27), and potentially provides shorter treatment times for large volumes. However, Gamma Knife SRS is not as readily available as LINAC SRS. In addition, its cobalt source actively decays, leading to a variable dose rate that may possibly prolong treatment time (28). Gamma Knife usually requires a rigid stereotactic frame and skull penetration with pins to ensure its accuracy(27), which is less comfortable for patients. Given the possible advantages and disadvantages of adopting Gamma Knife over linear accelerators for large AVMs, further studies will be needed to further compare the modalities clinically. Decisions to transition from LINAC-based radiosurgery to Gamma Knife for larger lesions should not be made prematurely without further evidence of a clear clinical benefit.

Chief limitations to this study include its retrospective nature, limited follow-up, and small sample size with only 14 of the 63 patients being treated with cones. Although we attempted to account for selection bias and confounding effects with multivariable analysis and propensity score matching, small sample size limited the number of predictors that could be included in regression models. In addition, late neurotoxicity can occur many years following SRS, and this study had a minimum follow-up of less than 1 year (29)<.

Thus, prospective well-powered studies with long-term follow-up are needed for further validation of the findings reported here.

CONCLUSION

In summary, this study indicates that the dosimetric advantages conferred by cone-based SRS over MLC-based SRS for cerebral AVMs may result in improved dosimetry and fewer toxicities. With consideration to the limitations of this study, these findings justify increasing the use of circular cones compared to MLCs in treating cerebral AVMs when feasible, such as for small AVMs. Further exploration is necessary to determine the best radiosurgical platform for large AVMs.