A Systematic Analysis of 2 Monoisocentric Techniques for the Treatment of Multiple Brain Metastases

Abstract

Background:

In this treatment planning study, we compare the plan quality and delivery parameters for the treatment of multiple brain metastases using 2 monoisocentric techniques: the Multiple Metastases Element from Brainlab and the RapidArc volumetric-modulated arc therapy from Varian Medical Systems.

Methods:

Eight patients who were treated in our institution for multiple metastases (3-7 lesions) were replanned with Multiple Metastases Element using noncoplanar dynamic conformal arcs. The same patients were replanned with the RapidArc technique in Eclipse using 4 noncoplanar arcs. Both techniques were designed using a single isocenter. Plan quality metrics (conformity index, homogeneity index, gradient index, and R_50%), monitor unit, and the planning time were recorded. Comparison of the Multiple Metastases Element and RapidArc plans was performed using Shapiro-Wilk test, paired Student t test, and Wilcoxon signed rank test.

Results:

A paired Wilcoxon signed rank test between Multiple Metastases Element and RapidArc showed comparable plan quality metrics and dose to brain. Mean ± standard deviation values of conformity index were 1.8 ± 0.7 and 1.7 ± 0.6, homogeneity index were 1.3 ± 0.1 and 1.3 ± 0.1, gradient index were 5.0 ± 1.8 and 5.1 ± 1.9, and R_50% were 4.9 ± 1.8 and 5.0 ± 1.9 for Multiple Metastases Element and RapidArc plans, respectively. Mean brain dose was 2.3 and 2.7 Gy for Multiple Metastases Element and RapidArc plans, respectively. The mean value of monitor units in Multiple Metastases Element plan was 7286 ± 1065, which is significantly lower than the RapidArc monitor units of 9966 ± 1533 (P < .05).

Conclusion:

For the planning of multiple brain lesions to be treated with stereotactic radiosurgery, Multiple Metastases Element planning software produced equivalent conformity, homogeneity, dose falloff, and brain V_{12 Gy} but required significantly lower monitor units, when compared to RapidArc plans.

Introduction

For several decades, stereotactic radiosurgery (SRS) has played a significant role in the management of patients with multiple metastases in the brain.^1,2 Traditionally, linear accelerator-based SRS techniques have utilized multi-isocentric plans for the treatment of patients with multiple brain metastases, whereby each lesion is associated with 1 isocenter that is placed near the geometric center of the lesion. The treatment delivery time would scale up with the number of lesions treated from approximately 20 to 30 minutes for a single lesion to over an hour for multiple lesions. Volumetric-modulated arc therapy (VMAT) was introduced by Otto³ as a novel plan optimization technique for efficient delivery of highly conformal plans using dynamically modulated arcs. The improved efficiency over the typical intensity-modulated radiotherapy plans was achieved by simultaneously varying the gantry rotation speed, dose rate, and multileaf collimator (MLC) field aperture shape.⁴ Although initially the VMAT technique was intended for conventionally fractionated cases, there have been recent publications reporting on the feasibility of using RapidArc (Eclipse treatment planning system version 8.9; Varian, Palo Alto, California) as an efficient delivery mechanism for multiple brain lesions^5,6 as compared to conventional SRS. Clark et al⁷ reported that RapidArc VMAT plans can be created using monoisocentric geometry with multiple noncoplanar dynamic-modulated conformal arcs. It has been shown that when treating multiple metastases, a monoisocentric VMAT plan can produce a highly conformal dose distribution with good plan quality and short treatment times.⁸

We have recently commissioned the Multiple Metastases Element (MME) software package from Brainlab as a planning system for brain metastases, which uses an optimized combination of dynamic conformal arcs to treat simultaneously multiple brain lesions. In this study, we compared the plan quality and the efficiency of planning and delivery of VMAT and MME plans for the treatment of multiple brain metastases.

Methods and Materials

Patient Population

Eight randomly selected patients (n = 8) were used for this planning study who were previously treated with multiple isocentric SRS plan at our institution for a minimum of 3 metastatic brain lesions. All patients had magnetic resonance imaging (MRI) scans that were reconstructed at 0.5 to 1.0 mm slice spacing. The high-resolution scans allowed for more accurate tumor localization and segmentation. Table 1 contains statistics for all patients in the study and their corresponding lesions.

Table 1.

Demographics of the 8 Patient Plans With at Least 3 Lesions.

	No. Lesions/Patient	Volume/Lesion (cc)	Rx (Gy)
Mean	5	0.7	23.9
Median	5	0.4	24
Range	3-7	0.3-4.3	18-24

Planning Techniques

The gross tumor volume was segmented by the physician using the contrast-enhanced T1-weighted MRI images for each patients. The contours were then remapped on the respective simulation computed tomography (CT) image set for each patient. For the VMAT plans, we used the Eclipse treatment planning system that has the proprietary Varian RapidArc implementation. All patients were planned with an arrangement of 4 noncoplanar arcs using a single isocenter (monoisocentric VMAT technique) based on the reported optimal arc arrangement that was proposed by Thomas et al.⁹ The 4-arc arrangement consisted of a 358° axial arc (couch angle of 0°) and 3 additional arcs spanning 170° at couch angles of 45°, 90°, and 135°, respectively. This beam arrangement is shown in Figure 1. A second plan was developed for each patient using the Brainlab (Munich, Germany) MME plan planning tool. The MME software has a fixed geometry of a maximum of 5 bidirectional arcs that can be used for treatment. During a preoptimization phase, the software will eliminate the arcs that do not have significant dose contribution and will select the optimal number of arcs for treatment based on the number and the location of the lesions treated. The MME performs pencil beam algorithm-based dose computation on simulation CT images. Once the prescription dose and the dose–volume criteria are specified, the dose optimization is performed with minimal user interference. The plans for both planning systems were calculated with the 6 MV SRS photon beam mode (dose rate = 1000 motor unit [MU]/min) with a Varian Novalis Tx linac with HD120 MLC. For the RapidArc-based VMAT plans, we added user-defined planning structures (such as rings and avoidance areas) in an effort to improve conformity, dose gradients, and to minimize low-dose spill. In RapidArc, minimum dose to the planning target volume (PTV) was set to the prescription dose and weights were continuously altered as a part of the optimization such that dose–volume criteria were met. During RapidArc optimization, the normal tissue objective (NTO) was also set to produce a high-gradient falloff from the targets. The NTO started at 0.1 mm from each target and set to achieve 100% to 20% falloff of the dose within 1 cm. Furthermore, a lower objective (50% of the volume receives 125% of the prescribed dose) for each of the targets was required to achieve the necessary hot spot. When the targets are close to each other, bridging dose can be lowered using an avoidance structure in the intervening region. Prescription dose was selected based on the volume of the target following the RTOG 9005 guidelines, recognizing that those guidelines are largely based on single-target SRS.¹⁰ Plans were normalized such that a minimum of 99% of the target volume was covered by the prescription dose. Physician and physicist reviews were performed to ensure clinical fitness of all the plans considering the target coverage as well as the dose to organs at risk (OARs). Critical structure doses were limited such that they abided by the normal tissue tolerance doses set forth by AAPM TG-101.¹¹

Figure 1.

Four-arc geometry utilized for the monoisocentric RapidArc-based SRS planning. SRS indicates stereotactic radiosurgery.

Plan Quality Evaluation

Clinical evaluation of SRS plans included general plan overview and evaluation of dosimetric indices including the conformity index (CI), homogeneity index (HI), and the dose falloff index (gradient index [GI]). Dose conformity was based on the inverse Paddick CI given as follows¹²:

$$\text{CI}=\frac{\text{TV} \times \text{ PV}}{{({\text{TV}}_{\text{PV}})}^2}.$$

A perfectly conformal plan would have CI = 1.0 and less conformal plans would have CI <1 or >1 depending on whether the target volume was over- or undercovered by the prescription isodose volume.

Uniformity of dose distribution within a PTV was estimated using the HI defined based on the dose coverage to 2% (D_2%T) and 98% of the target volume (D_98%T):

$$\text{HI}=\frac{{\text{D}}_{2\text{\%T}}}{{\text{D}}_{98\text{\%T}}}.$$

Values of HI closer to 1.0 indicate greater dose homogeneity, whereas values larger than 1 indicate more heterogeneous dose distribution.

Paddick’s GI¹³ is defined as follows:

$$\text{GI }=\frac{{\text{V}}_{50\text{\% Rx}}}{{\text{V}}_{100\text{\% Rx}}}.$$

Gradient index is an indication of low-dose spillage, with lower GI values indicating greater dose falloff and better dose conformity outside the target volume. Although for single-target SRS plan GI value of ≤ 3.0 is considered ideal, data are not available regarding optimal GI values for plans consisting of multiple targets. The GI values are expected to be higher when multiple targets are in close proximity with overlapping 50% isodose lines.

In addition, medium- to high-dose falloff could be studied using R_50% which is defined as:

$${\text{R}}_{50\%} = \frac{{\text{V}}_{50\% \text{Rx}}}{\text{PTV}}.$$

Being a ratio of V_50% normalized to the volume of PTV, R_50% helps in comparing V_50% across tumors. This metric is useful in scenarios where comparison between 2 tumors with similar V_50% values but dissimilar V_100% values could be tricky and even prone to error. Similar to GI, low values of R_50% indicate steep dose falloff and vice versa. It should be noted that the acceptable values of R_50% would be based on the volume of PTV. It is worth mentioning here that while GI is calculated for the patient plan, R_50% is evaluated for each tumor.

With regard to the dose to the OARs, brain dose has been studied in detail considering the probability of complications from radionecrosis as outlined in the QUANTEC report.¹⁴ Dosimetric quantities evaluated for brain included dose to 2% and 98% of the brain volume (D_2%B and D_98%B, respectively), mean brain dose (MBD), and volume of brain irradiated by the 12 Gy isodose line (VB_{12 Gy}).

The resulting MUs required to deliver the prescribed dose by the 2 planning techniques, MME and RapidArc, were also compared. Finally, the differences in planning time between the 2 different modalities were also studied.

Statistical Analysis

Test for normality was performed using Shapiro-Wilk test in the R statistical package.¹⁵ Statistical significance was tested using a paired Student t test for normally distributed data and a Wilcoxon signed-rank test for those distributions that failed the normality test. Test of statistical significance was compared against a threshold P value less than .05.

Results

Figure 1 shows a diagrammatic representation of the beam geometry used for the VMAT plans including couch angles and arc angular ranges. Based on the target volume (shown in Table 1), prescription dose ranged between 18 and 24 Gy with a median value 24 Gy. Baring a single lesion that received 18 Gy, the remaining lesions received 24 Gy.

The CI, HI, and GI were evaluated for each plan, and the mean, median values, and range are tabulated in Table 2. The MME and RapidArc plans had very similar CI values with respective mean ± standard deviation (SD) of 1.8 ± 0.7 and 1.7 ± 0.6. The HI values between the MME and RapidArc plan were also identical with a mean ± SD value of 1.3 ± 0.1. The mean ± SD value of GI was 5 ± 1.8 and 5.1 ± 1.9 in the MME and RapidArc plans, respectively. The mean ± SD value of R_50% was 4.9 ± 1.8 and 5.0 ± 1.9 in the MME and RapidArc plans, respectively. A Shapiro-Wilk test on the CI, HI, GI, and R_50% values led to rejection of null hypothesis that the data are normally distributed. A paired Wilcoxon signed rank test showed that the differences in values of CI, HI, GI, and R_50% were not significant (respective P values are .65, .57, .96, and .71).

Table 2.

Comparison Statistics of CI, HI, GI, and R_50% of the 8 Treatment Plans From MME and RapidArc.

	CI		HI		GI		R50%
	MME	RapidArc	MME	RapidArc	MME	RapidArc	MME	RapidArc
Mean	1.8	1.7	1.3	1.3	5.0	5.1	4.9	5.0
Median	1.5	1.5	1.3	1.3	4.5	4.7	4.4	4.5
Range	1.2-4.5	1.2-4.6	1.2-1.6	1.1-1.5	2.6-6.0	3.5-5.9	2.6-6.0	3.4-5.9

Abbreviations: CI, conformity index; HI, homogeneity index; GI, gradient index; MME, Multiple Metastases Element.

Table 3 lists some of the dosimetric quantities estimated for the brain tissue including D_2%B, D_98%B, MBD, and VB_{12 Gy}. The mean ± SD of VB_{12 Gy} was 20.7 ± 9.4 cm³ and 20.8 ± 8.1 cm³ for MME and RapidArc plans, respectively. Although the average value of MBD was marginally higher and the volume of hot spots was smaller for the RapidArc plan in comparison to the MME plan, the differences were not significant.

Table 3.

Comparison Statistics of D_2%B, D_98%B, MBD, and VB_{12 Gy} of the MME and RapidArc Plans.

	D98%B		D2%B		MBD (Gy)		VB12 Gy
	MME	RapidArc	MME	RapidArc	MME	RapidArc	MME	RapidArc
Mean	0.5	0.6	11.4	10.2	2.3	2.7	20.7	20.8
Median	0.3	0.5	12.0	9.7	2.5	2.8	22.3	21.2
Range	0.1-1.1	0.1-1.4	7.2-16.2	7.8-13.0	1.3-3.1	1.3-3.9	9-33.1	10.6-33.5

Abbreviations: D_98%B, dose to and 98% of the brain volume; D_2%B, dose to and 2% of the brain volume; MBD, mean brain dose; MME, Multiple Metastases Element; VB_{12 Gy}, volume of brain irradiated by the 12 Gy isodose line.

The dose distribution of the MME and RapidArc for one of the patient plans is shown in Figure 2. Notice the difference in hot spots and the low-dose areas in the healthy brain tissue between the 2 plans. Figure 3 illustrates the comparative dose–volume histogram of PTVs and normal brain in a case with 5 metastases.

Figure 2.

Dose distribution for the (A) MME and (B) RapidArc plans for a patient with 5 metastatic lesions. Notice the difference in the hot spots (28 Gy) and the low-dose (8, 12 Gy) regions in the brain. MME indicates Multiple Metastases Element.

Figure 3.

Comparison of cumulative DVH of the 5 PTVs and brain between MME and RapidArc. DVH indicates dose–volume histogram; PTV, planning target volume; MME, Multiple Metastases Element.

Table 4 tabulates statistics of the target, arc arrangement, and prescription dose for each patient plan. The mean and median number of arcs in MME plans were 5.5 ± 1.2 and 5.3 ± 1.2, respectively. The MME plans required on average of 7286 ± 1065 MUs against 9966 ± 1533 MUs in RapidArc plans for achieving adequate target coverage and fulfilling the dose prescription. This translates to 27% lower MUs and a corresponding savings in treatment time for the MME plans. The total MUs per plan were evaluated to be normally distributed in a Shapiro-Wilk test. The MU differences were significant according to a 2-tailed paired Student t test with P < .05. It can be noticed that a high-dose rate linac would taper the slight temporal advantage with lower MU treatment. Overall treatment time would depend on the setup time, which goes up with the number or arcs (mean of 5.3 in MME vs 4 in RapidArc plans).

Table 4.

Number of Targets, Prescription Dose per Target, and Number of Arcs for the 8 Patients Along With the Mean, Median, and Standard Deviation (SD) of the Values.

Patient	No. Targets	Rx/target (Gy)	No. Arcs in MME	MUs in MME	MUs in RapidArc
1	6	All at 24 Gy	6	7028	9268
2	4	All at 24 Gy	6	7210	9681
3	5	All at 24 Gy	4	7614	9135
4	5	All at 24 Gy	4	6654	11 962
5	4	All at 24 Gy	4	5663	9642
6	7	All at 24 Gy	7	8997	12 523
7	3	18, 24, 24 Gy	5	6657	7850
8	6	All at 24 Gy	6	8468	9668
Mean	5	23.9	5.3	7286	9966
Median	5	24	5.5	7119	9655
SD	1.3	0.9	1.2	1065	1533

Abbreviations: MME, Multiple Metastases Element; MU, monitor unit.

A substantial savings in planning time was also observed for the MME plans, which took 3 ± 1 minutes to generate on a computer with Intel i7 processor and 32 GB memory. On the other hand, RapidArc plans took 65 ± 10 minutes using Analytical Anisotropic Algorithm (AAA) calculation engine and 15 ± 6 minutes using the Acuros XB algorithm on a desktop with Intel Xeon E5 processor and 8 GB memory. All calculations for both planning systems were done with a 1 × 1 × 1 mm³ dose grid resolution, and the treatment planning time was recorded after the segmentation was completed.

Discussion

The outcome of this study indicates equivalence in plan quality between the RapidArc and MME techniques based on the similarities in the resulting values of CI, HI, and GI for the patient population studied. There was, however, a significant reduction in planning time and delivery time (based on the MUs calculated) for the MME plans over the RapidArc plans using the 4 noncoplanar arc configuration that we selected.

Achieving rapid dose falloff from the target volume can be challenging, especially when the targeted lesions are in close proximity to each other. We used the method of specifying lower dose ceilings with increasing diameter rings around the target as an effective way in obtaining conformal dose distribution with the RapidArc plans. For RapidArc, alternate planning techniques using more arcs could be explored in an effort to further improve plan quality at the cost of higher planning time as noted by Thomas et al⁹ and Fogliata et al.¹⁶

The choice of R_50% that was calculated for each lesion provides the planner with a perspective of dose falloff outside the lesion. However, challenges arose in the estimation of GI when the 50% isodose lines from multiple lesions overlap with one another. Further study is needed to validate GI and R_50% as meaningful dosimetric quantities and evaluation of optimal range of values of these metrics in scenarios where multiple lesions are in close proximity to one another. Different metrics may be needed to determine the quality of a plan created for treating multiple cranial metastases, as suggested by McDonald et al.¹⁷

One of the significant differences between the 2 planning techniques that we used in this study is that although the MME algorithm uses a pencil beam dose algorithm and a hybrid dynamic conformal arc for the MLC segments, the RapidArc algorithm uses the AAA dose algorithm and does full MLC modulation with variable gantry and MLC leaf speed as well as dose rate modulation. In this preliminary study, we found that the RapidArc and the MME techniques produce plans that are dosimetrically equivalent. It was recognized, however, that especially for the smaller targets, the control points (MLC segments) that are generated for some of the RapidArc plans are very small where we typically observe nonlinearity dose effects, and the output factors for such segments are difficult to characterize. On the other hand, the MME algorithm uses primarily a dynamic conformal arc approach for the optimization of dose and the CI and as such produced MLC segments that were typically of the same size as the tumors treated.

Although pencil beam-based algorithms are not as accurate as those that are convolution based^18,19 (such as the AAA used in Eclipse), the differences are not relevant in the context of SRS where we typically use a large number of beams or arcs and the brain is of uniform density.

The outcome of our initial experiences of monoisocentric treatment with 2 planning techniques holds good for patients with multiple metastasis. A detailed analysis of the suitability of a treatment plan shall be dealt by the physician on a case-by-case basis. The details about the tumor shape, location, and proximity to OAR were not a part of the study. Some of the features of treatment planning systems in dealing with complexities from proximity of metastasis to OARs including optical apparatus and brain stem were beyond the scope of this initial study.

The study primarily illustrates the increased efficiency of the monoisocentric technique in a linac, whether MME or VMAT, and does not deal with other treatment techniques like GammaKnife or CyberKnife. A detailed comparison of the monoisocentric treatment with GammaKnife has been reported by Thomas et al.⁹

Conclusion

In our study of 8 patients with multiple brain metastases, we found that the monoisocentric cranial MME planning system produces SRS plans of comparable quality to the RapidArc-based VMAT plans. The added value of the MME is the significantly reduced planning in addition to the reduced complexity of the plan.

Abbreviations

CT

computed tomography

CI

conformity index

GI

gradient index

HI

homogeneity index

MBD

mean brain dose

MLC

multileaf collimator

MME

Multiple Metastases Element

MRI

magnetic resonance imaging

NTO

normal tissue objective

OAR

organ at risk

PTV

planning target volume

SD

standard deviation

SRS

stereotactic radiosurgery

VMAT

volumetric-modulated arc therapy