Dose Homogeneity Analysis of Adjuvant Radiation Treatment in Surgically Resected Brain Metastases: Comparison of IORT, SRS, and IMRT indices

Abstract

Background:

While surgery remains a treatment option for symptomatic brain metastases, the need for adjuvant radiation following surgery is widely accepted as standard. Despite a multitude of randomized trials aimed at identifying the ideal radiation treatment plan for surgically resected metastases, the development of new delivery regiments necessitate a periodic re-evaluation of dosimetric performance/outcome. Here, we compare the homogeneity index across three platforms: single-session stereotactic radiosurgery, multi-session stereotactic radiotherapy, and intra-operative radiotherapy.

Methods:

Patients treated with IORT following surgical resection of BMs were identified and dosimetric parameters collected from the dose-volume histograms based on the development of conformal plans for adjuvant radiation using Gamma Knife-SRS (GK-SRS), LINAC-IMRT, and IORT. HIs were calculated using four established methods and compared across platforms within the patient cohort. Statistical analyses were performed using ANOVA.

Results:

The mean maximal doses for the GK-SRS and IMRT plans were 30Gy and 29Gy with margin prescription doses of 16Gy and 24Gy, respectively. The IORT dose was 30Gy to the applicator surface. HIs varied based on calculation methods, but maintained consistency when comparing across platforms with IORT having the lower mean HI value (0.56; 95%CI 0.55–0.60) in single-fraction treatment, compared to GK-SRS (0.77; 95%CI 0.76–0.80). The mean multi-session IMRT HI was lower than both single fraction treatment modalities at 0.41 (95%CI 0.40–0.42).

Conclusion:

When using HI as the primary dosimetric parameter for adjuvant radiation plans following surgical resection of BMs, IORT offers improved dose homogeneity compared to GK-SRS in single fraction treatment, while fractionated LINAC-based IMRT was superior with respect to HI in comparison among all three methods.

Introduction

Brain metastases remain the most common brain tumors in adults. Annually in the United States, approximately 170,000 new brain metastases are diagnosed with nearly 25% of cancer patients developing intracranial metastases (1,2). Among those, only about 8–10% are reportedly symptomatic (3). Treatment and management have become increasingly individualized as surgical and radiation techniques continue to evolve, and systemic therapies offers greater potential for systemic and intracranial disease control in many cancer types. Several factors contribute to treatment decisions including metastatic tumor burden, tumor size and location, degree of mass effect and edema, presence or absence of symptoms, status or extent of systemic disease and patient performance status. Surgical resection is often reserved and favored for those with large tumors, presence of symptom(s) or with an uncertainty regarding the histologic diagnosis or radiographic findings exists. Following surgical resection, it is now adopted as standard of care for patients to undergo adjuvant postoperative radiation given the local recurrence rates of 50–60% with surgery alone (4).

Determining the modality and technique for optimal adjuvant radiation delivery continues to be a subject of debate. Unique technical challenges exist in the treatment of tumor resection cavities, including the target size, location relative to eloquent tissue, and clear target delineation in the setting of recent surgery. The use of stereotactic radiosurgery (SRS) in this setting continues to expand with obvious limitations based on cavity size and the potential development of radiation necrosis (5). An alternative to SRS is multi-fraction stereotactic therapy (SRT), delivered over 3–5 fractions with the possibility of reducing the incidence of radionecrosis (RN) risks compared to SRS (6). Recently, intraoperative radiotherapy (IORT) has been to be viable modality for delivery of adjuvant radiation, offering elimination of both delayed time to initiation of therapy and the need for target contouring needed for SRS and SRT. To this end, a recent retrospective international study has demonstrated local control and RN rates equal if not superior to the SRS single fraction data (7).

Across each of these treatment platforms (SRS, SRT, and IORT) are unique advantages and disadvantages based on dosing, risk profiles, and patient compliance. The present study specifically compares the dosimetry and homogeneity indices between these various modalities. Our data support the concept that the dose-volume parameters, including homogeneity index, for IORT are superior to SRS for single fraction resection cavity treatment and comparable to SRT using modern conformal planning technique.

Methods

Patient population.

Following approval from the IRB (1809291320A001), fourteen consecutive patients were identified that had undergone IORT for adjuvant radiation in the setting of surgically resected intracranial metastases at a single institution. Pre-operative and postoperative gadolinium-enhanced MRIs has been acquired for the standard of care management. Pre-operative volumetric CT scans were utilized for target delineation in the IORT, while postoperative MRI and CT scans were used for the Gamma Knife and IMRT planning, respectively. The pre-operative CT was used for the IORT planning based on the need for CT formatted imaging for use with the planning software along with the fact that the pre-operative tumor volume most closely correlated with applicator volume needed to fill the resection cavity at the time of surgery.

Dosimetric data acquisition.

Post-resection cavity treatment plans were developed using Gamma Plan v10.1 (Elekta, Stockholm, Sweden) and Eclipse 11.0 (Varian Medical Systems, Palo Alto, CA). Contouring within Gamma Plan included the surgical resection cavity wall without the addition of a margin. The IMRT plan added a 2 mm expansion on the cavity wall contour. IORT contouring was not factored into the planning as the resection cavity remodeling often resulted in the use of an applicator smaller than tumor on pre-operative CT (data not shown). Prescription doses for GK plans were 16Gy to the margin at the 50% isodose while IMRT plans were developed for a prescription dose of 24Gy at the 80% isodose delivered in 3 fractions using a non-coplanar method. The prescription dose for the IORT treated patients was 30Gy at the applicator surface. The dose-volume histograms for the GK and IMRT plans were utilized for collection of the volumes of the targets receiving 98% (D98), 95% (D95), 50% (D50), 5% (D5), and 2% (D2) of the total dose. The same dose parameters were defined for each of the 8 spherical applicators (1.5 cm – 5.0 cm diameter) and the corresponding values used based on the applicator size during the IORT procedure.

Homogeneity Index (HI) calculations.

Four independent methods of HI calculation were used for analysis of the dosimetric data, designated H1 through H4, as follows:

• H1 = (D2 – D98) / D50

• H2=(D5–D95)/ D50

• H3 = (D2 – D98) / D_max (designated as D2)

• H4 = D5/D95

Descriptive statistics of the individual dosimetric values and HIs were reported as the mean, standard error of the mean (SEM), and minimum/maximum values. 95% confidence intervals (CI) of the mean values of HIs were determined and one-way ANOVA performed for each calculation method across all treatment modalities. All statistical analyses were performed using SPSS v.26 (IBM, Armonk, NY) and GraphPad v.8.0.2 (GraphPad Software, San Diego, CA).

Results

Dosimetric calculations across three platforms.

The dose-volume histograms were utilized for calculating the doses for 98%, 95%, 50%, 5% and 2% of each volume of target as defined in Gamma Plan or Eclipse and designated as D98, D95, D50, D5, and D2 (Table 1). The doses for the same volume parameters were calculated for each fixed volume spherical IORT applicator based on the actual applicator size used during the intra-operative radiotherapy session. Therefore, the patients treated with the same size applicator were all presumed to have the same dosimetric values, with variations only based on the use of different applicator sizes to optimally fit the resection cavity. The mean maximal dose (D2 ± SEM) for the GK and IMRT treatment plans were 30.0±0.02Gy and 29.3±0.1Gy, respectively, with prescription doses for GK being 16Gy to the cavity margin in a single fraction and 24Gy to the margin in 3 fractions with IMRT. The mean maximal dose (D2 ± SEM) for the IORT treatments was 42.1±0.5Gy in a single fraction where the prescription dose was 30Gy to the applicator surface. A graphical representation of the color wash of the dose distribution was created for a single target across all three platforms (Figure 1).

Table 1.

Individual value calculations from DVH. The D₂ through D₉₈ values for all patients across all three platforms are displayed along with mean, SEM, minimum and maximum values.

	Gamma Knife					IORT					IMRT
	D98	D50	D2	D5	D95	D98	D50	D2	D5	D95	D98	D50	D2	D5	D95
	17.8	23.6	29.6	28.8	18.8	30.0	36.0	45.4	44.4	30.6	23.9	28.0	29.0	28.8	24.4
	16.4	22.6	29.8	28.8	17.4	30.0	34.0	40.0	39.4	30.4	24.1	27.8	28.7	28.5	24.7
	16.4	24	30.4	29.8	17.8	30.0	35.5	44.3	43.4	30.6	25.0	28.5	29.3	29.1	25.4
	16.6	23.8	30	29	18.2	30.0	34.7	40.4	39.8	30.4	24.4	28.1	29.2	29.0	25.0
	15.2	21.2	27.8	27.8	16.4	30.0	34.7	41.9	41.2	30.5	24.4	28.2	29.0	28.9	25.0
	16.2	24	30.4	29.6	17.6	30.0	34.0	40.0	39.4	30.4	23.8	28.2	29.2	29.0	24.5
	17.4	24.4	30.8	30.2	18.2	30.0	34.7	41.9	41.2	30.5	24.2	27.9	28.7	28.6	24.8
	16.6	23.8	30	29	17.8	30.0	35.5	44.3	43.4	30.6	24.1	27.8	28.8	28.6	24.7
	16.6	22	28.6	28	17.2	30.0	36.0	45.4	44.4	30.6	24.2	28.0	29.2	29.0	24.7
	17.4	24	30	29	18.6	30.0	34.0	40.0	39.4	30.4	23.5	27.8	28.7	28.5	24.1
	17.4	25.6	30.8	30	18.8	30.0	34.7	41.9	41.2	30.5	24.3	28.0	29.1	28.9	24.7
	16.6	23.8	29.8	28.8	18	30.0	34.7	41.9	41.2	30.5	24.2	28.1	29.2	29.0	24.8
	16.6	24	31.2	30.6	17.6	30.0	34.7	41.9	41.2	30.5	24.0	27.8	28.8	28.6	24.5
	17.4	24.8	30.4	29.6	18.4	30.0	34.0	40.0	39.4	30.4	23.7	27.8	28.7	28.5	24.4
Mean	16.8	23.7	30.0	29.2	17.9	30.0	34.8	42.1	41.4	30.5	24.1	28.0	29.0	28.8	24.7
SEM	0.2	0.3	0.2	0.2	0.2	0.0	0.2	0.5	0.5	0.0	0.1	0.1	0.1	0.1	0.1
Min	15.2	21.2	27.8	27.8	16.4	30.0	34.0	40.0	39.4	30.4	23.5	27.8	28.7	28.5	24.1
Max	17.8	25.6	31.2	30.6	18.8	30.0	36.0	45.4	44.4	30.6	25.0	28.5	29.3	29.1	25.4
	Gamma Knife				IORT				IMRT
	H1	H2	H3	H4	H1	H2	H3	H4	H1	H2	H3	H4
	0.5	0.42	0.4	1.53	0.43	0.38	0.34	1.45	0.18	0.16	0.18	1.18
	0.59	0.5	0.45	1.66	0.3	0.27	0.25	1.3	0.16	0.14	0.16	1.15
	0.58	0.5	0.46	1.67	0.4	0.36	0.32	1.42	0.15	0.13	0.14	1.15
	0.56	0.45	0.45	1.59	0.3	0.27	0.26	1.31	0.17	0.14	0.16	1.16
	0.59	0.54	0.45	1.7	0.34	0.31	0.28	1.35	0.17	0.14	0.16	1.16
	0.59	0.5	0.47	1.68	0.3	0.27	0.25	1.3	0.19	0.16	0.18	1.18
	0.55	0.49	0.44	1.66	0.34	0.31	0.28	1.35	0.16	0.13	0.16	1.15
	0.56	0.47	0.45	1.63	0.4	0.36	0.32	1.42	0.17	0.14	0.16	1.16
	0.55	0.49	0.42	1.63	0.43	0.38	0.34	1.45	0.18	0.15	0.17	1.17
	0.53	0.43	0.42	1.56	0.3	0.27	0.25	1.3	0.19	0.16	0.18	1.18
	0.52	0.44	0.44	1.6	0.34	0.31	0.28	1.35	0.17	0.15	0.16	1.17
	0.55	0.45	0.44	1.6	0.34	0.31	0.28	1.35	0.18	0.15	0.17	1.17
	0.61	0.54	0.47	1.74	0.34	0.31	0.28	1.35	0.17	0.15	0.17	1.17
	0.52	0.45	0.43	1.61	0.3	0.27	0.25	1.3	0.18	0.15	0.17	1.17
Mean	0.56	0.48	0.44	1.63	0.35	0.31	0.29	1.36	0.17	0.15	0.17	1.17
SEM	0.009	0.010	0.005	0.015	0.013	0.012	0.009	0.015	0.003	0.003	0.003	0.003
Min	0.50	0.42	0.40	1.53	0.30	0.27	0.25	1.30	0.15	0.13	0.14	1.15
Max	0.61	0.54	0.47	1.74	0.43	0.38	0.34	1.45	0.19	0.16	0.18	1.18

Figure 1.

Planning dose across treatment platforms. A representative dose color wash comparison between the GK-SRS (top), IORT (middle), and three-fraction IMRT (bottom) plans with the corresponding dose scale legend (right).

Homogeneity Index (HI) calculations.

Based on four separate formulae, the HI was calculated for each patient in each treatment modality (Table 2a). The mean HIs ranged from 0.15 for formula H2 (D5-D95/D50) for the 3 fraction IMRT plan to 1.74 using formula H4 (D5/D95) in the GK plan. Consistently, H2 provided the lowest ratio of HI (0.13–0.42) while H4 resulted in the highest ratios (1.18–1.74) across all three treatment platforms. A composite analysis of all HIs calculated for all methods in all treatments demonstrated a clear differentiation in means based on platform, as demonstrated in Figure 2 (GK:0.77, 95%CI 0.76–0.80; IORT:0.56, 95%CI 0.55–0.60; IMRT:0.41, 95%CI 0.40–0.42). A one-way ANOVA was conducted to compare the impact of the treatment modality on HI for each of the four calculation methods. There was a significant effect of the type of treatment on HI for all calculation methods (Table 2b; p<0.001). Post-hoc comparison using Tukey HSD test indicated that the means of all three treatment HIs were significantly different for each calculation method (p<0.001).

Table 2.

Homogeneity Index calculations. Using four separate methods, HIs were calculated for all patients based on dosimetry values (upper panel). Comparisons across platforms within the same calculation method were all significantly different based on ANOVA testing (lower panel).

ANOVA (One-Way)
	F	df (between, within)	p
H1	440.8	(2, 39)	<0.001
H2	346.3	(2, 39)	<0.001
H3	506.8	(2,39)	<0.001
H4	359.7	(2,39)	<0.001

Figure 2.

Box plot comparison of composite HIs. Homogeneity indices were calculated according to four separate formulae for individual patients and compared via box plot (5–95 percentile). ANOVA testing demonstrated significance in differences across groups (p<0.001).

Discussion

The importance of radiation treatment in achieving local control of brain metastases both as a definitive therapy and in the adjuvant setting following surgery has been well established (4,8,9). While the use of SRS has increased relative to WBRT based on the reduction of neurocognitive side effects, the limited utility of single session SRS for large lesions has been clearly demonstrated by studies following from RTOG 90–05 (10,11). While RN rates have been shown to be up to an order of magnitude higher for metastases >2 cm treated with single session SRS, some retrospective data have suggested that fractionated SRS plans have safety advantages with respect to RN over single fraction treatments (11,12). Meanwhile, local control in SRS treated lesions has been poor, with additional concern that surgical resection may be associated with increased risk of leptomeningeal failure (13,14). Given the competing objectives of achieving adequate local control and reducing the incidence of RN, SRT has emerged as an alternative means of dose escalation without significantly increased toxicity (15,16). Similarly, IORT has been shown as another potential alternative therapy, though less widely explored in the randomized setting with respect to intracranial applications. In the context of breast cancer, local therapy with IORT has been clearly shown to safe and efficacious (17,18).

Cross platform comparisons of SRS have been the subject of numerous studies, with the development of a myriad of measurement techniques (19). Although the biological endpoints, such as local control, distant failure and, of course, overall survival, are critical, several indices have emerged that allow for dose plan optimization by the treatment team prior to initiation of therapy (20,21). Minimization of adverse effects, such as RN, while maximizing of the biologically effect dose to the target represent the ultimate goals of the dose planning process. In the current study, our cross platform analysis of surgically resected brain metastases utilized the homogeneity index (HI) the primary metric.

Although the calculation methodology of determining the HI varies based on the literature references, our use of the four most widely used formulae followed the same trend as other studies, expressing the HI as a measure of distribution of the D_max to the D_min. In the setting of single fraction therapy, IORT demonstrated improved, or lower, HI values from all methods compared to Gamma Knife SRS. A visual comparison of the dose distributions confirms this finding (Figure 1) and also identifies the fundamental difference between IORT and Gamma Knife SRS with regard to the manner in which dose is delivered. In order to reach the desired margin dose, the GKRS plan has the development of “hot spots” within the cavity, which is largely devoid of target eligible tissue, while the IORT applicator fills the cavity void and allows for a more homogeneous dose to the resection cavity wall.

The addition of a multi-fraction plan, as shown with the IMRT data, introduced even greater dose homogeneity than that seen in IORT based on HI calculation. Once again, the spatial relationship of the IORT applicator versus the IMRT planin the dose map demonstrates that the majority of the dose delivered in the central portion of the resection cavity is not high risk for local recurrence compared to the cavity wall. Moreover, the current data does not take into consideration the integral dose to the brain received by IMRT, which, in other studies, has been shown to be greater than even the SRS plans achieved by GKRS (22–24).

As previously mentioned, the development of objective metrics for comparisons of platform dose deliveries is desirable from the standpoint of optimizing dose and minimizing risk, but these cannot be the only considerations in the treatment of patients with large brain metastases. Standard delays in time to initiation (TTI) of GKRS or IMRT create a relative risk for local tumor growth in the setting of subtotal resection and tumor recurrence with prolonged delays due to extended rehabilitation or convalescence (25,26). In such cases, IORT offers the complete elimination of TTI delays by consolation of care into a single episode. Although some degree of TTI delay exists for GKRS, the single additional episode care may be more acceptable to patients and caregivers than 3 to 5 fractions delivered by IMRT. Travel distance, loss of wages, and associated costs can all factor into developing an optimal plan for management with respect to the “best” modality of adjuvant radiation delivery. Finally, the obstacle of target contouring in the context of surgically resected brain metastases is a concern for dose delivery in GKRS and IMRT, while the target delineation in IORT is negligible in the setting of a gross tumor resection and a direct apposition of the applicator to the cavity wall. Several groups have noted the challenge of target identification for surgical cavities, with the recent formation of expert consensus guidelines still awaiting further validation studies (27,28).

As retrospective study of the theoretical dose plans of GKRS and IMRT for patient who underwent IORT, one limitation of this study involves the use of the preoperative CT scan for the IORT dosimetric data acquisition rather than intra-operative imaging, which was not available. Post-operative MRIs were used for the GKRS and IMRT plans to provide the most likely representative cavity size for planning, often smaller than the original tumor size, and universally less in volume than the pre-operative CT scan. Additionally, the difference in dosimetry does not address other differences between these modalities with regards to radiobiology. The relative biological effectiveness (RBE) of the 50kV photon system used for IORT has been shown to be greater than the 6MV systems with data from the use of IORT in breast sparing lumpectomy that suggest an additional effect of low-kV radiation within the tumor microenvironment (29).

Overall, these data offer additional insight into the dosimetry across treatment platforms used for the delivery of adjuvant radiation to the resection cavities of intracranial metastases. Future prospective trials will be more effective in determining the optimal approach and/or treatment modality, remaining cognizant that patient care preferences may factor heavily into the treatment of choice.