Stereotactic radiosurgery for resected brain metastasis: Cavity dynamics and factors affecting its evolution

Abstract

Objective

To determine changes in post-surgical cavity volume for metastases based on time from surgery, pre-operative tumor dimensions and other predictors, in patients planned for post-operative stereotactic radiosurgery (SRS).

Methods

Patients with resected brain metastases from a primary solid tumor, treated with post-operative surgical cavity SRS from 2008 to 2014 were identified from an institutional prospective database. The segmented three-dimensional (3D) volume of the pre-operative tumor and post-operative surgical cavity were determined based on MRI and percent volume change was calculated. Patients were grouped according to early (<21 days), intermediate (22-42 days), and late (>42 days) intervals based on the number of days between the date of surgery and the treatment planning MRI. Potential predictive factors including tumor size, location, age, dural involvement, and degree of surgical resection were also analyzed.

Results

Sixty-one cavities in 59 patients were evaluated. Overall, a significant volume reduction (4cm³, p=0.03) was observed comparing tumor and cavity volumes. For larger tumors, an average volume reduction of 11.6% (p=0.01) was observed compared to an increase of 34.4% in smaller tumors (p=0.69). For both large and small tumors, cavities were larger in the early interval especially for smaller tumors. During the intermediate interval, a significant volume reduction was observed for larger tumors (28%, p=0.0007). Tumor size, dural involvement, age and time from surgery were significant predictors for volume change on univariate analysis. On multivariate analysis, tumor size, dural involvement and time from surgery were significant.

Conclusion

Tumor size (>3cm), dural involvement and longer time from surgery were significant predictors of cavity volume reduction. Caution must be taken when treating cavities in the early (<21 days) interval after surgery as it may lead to irradiating more normal tissue especially in small tumors.

Introduction

Surgical resection of brain metastases is usually indicated for patients with good performance status and a solitary metastasis, or large tumors (>3cm) causing neurologic dysfunction or refractory edema. Although traditional practice has been to treat with adjuvant whole brain radiation therapy (WBRT) [1], practice is evolving to treat these patients with focal single or fractionated stereotactic radiosurgery (SRS). The intent is to treat the surgical cavity to reduce the risk of local relapse, and spare patients from the adverse neurocognitive and quality of life effects associated with WBRT[2,3].

At present, several questions remain in the treatment of post-operative surgical cavities including optimal volume delineation, use of single or fractionated SRS, and timing of post-operative treatment planning. The latter is the focus of this investigation.

There are several implications associated with the timing of SRS to the post-operative surgical cavity. For example, if surgical cavities contract over time after surgery, then waiting may decrease the volume of irradiated normal brain tissue, which may reduce the risk of symptomatic radiation necrosis. If the volume increases or remains similar over time after surgery, then delays in radiation may translate into recurrent or progressive disease in the surgical cavity and delay systemic treatment. The objective of this study was to evaluate the changes in the volume of the surgical cavity as a function of pre-operative tumor dimensions, time from surgery and other factors that may correlate with volume reduction.

Methods

Consecutive patients with brain metastases from solid tumors treated with fractionated SRS (fSRS) to the surgical cavity, between 2008 to 2014, were identified from a prospective institutional database after local Institutional Research Ethics Board (IRB) approval. Categorical variables such as number of resected brain metastases, extent of resection, histology, location and dural involvement were summarized with counts and percentages. Continuous variables such as age, greatest tumor diameter, tumor volume and time between surgery and cavity SRS were summarized with means and standard deviation, or medians and ranges as appropriate.

All patients had a pre-operative (preop) brain MRI and a treatment planning MRI, with the later quantifying the extent of resection. Timing of cavity treatment was left up to the treating radiation oncologist. For each surgical cavity, the preop tumor volume and planning cavity volume were contoured using the axial T1-weighted post-gadolinium MRI sequence, as shown in Figure 1, and both the maximum diameter and segmented 3D volume were recorded. The outcome of interest was the relative change in the volume of the planning cavities compared to the preop tumor. Kolmogrov Smirnov’s test was assessed to check the normality assumption on the outcome of interest. We used three time intervals calculated from the date of the surgery; early (≤21 days), intermediate (22-42 days) and late (>42 days) to compare the relative change over these intervals. Factors that could potentially be predictive of change in the planning cavity volume were analyzed using Pearson/Spearman correlation coefficient for continuous variables and Kruskal Wallis or Analysis of Variance (as appropriate based on the normality test) for categorical variables. Potential predictors included: diameter of the intact tumor (≥3cm=large or <3cm=small), location (supra-tentorial vs. infra-tentorial), time from surgery, preop dural involvement, extent of resection (gross-total, GTR or sub-total resection, STR, based on the post-operative CT scan and operative report) and patient age. In addition to the continuous outcome, we also considered the same as a binary outcome of whether the cavity volume on the planning MRI was smaller or larger compared to the preop tumor volume to assess the association with potential predictors. Logistic regression analysis was finally implemented to assess the joint effect of potential predictors that were found significant on univariate analysis. All P-values were 2-sided and for the statistical analyses, p<0.05 was considered to indicate a statistically significant result. Statistical analysis was performed using version 9.4 of the SAS system for Windows, 2002-2012 SAS Institute, Inc., Cary, NC.

Figure 1.

An example of tumor and cavity contouring

MRI scans demonstrating reduction in postoperative cavity volume compared to preoperative tumor volume. (A) Tumor in the right frontal lobe contoured (red). (B) Postoperative planning MRI 21 days later shows a cavity (red contour) that is smaller than the preoperative tumor.

Figure 2.

Percent change in the planning cavity volume compared to the preop tumor volume over time.

Percent change (shown in quartiles) in the planning cavity volume compared to the preop tumor volume plotted against time interval from surgery

Results

Patients and tumor characteristics:

A total of 61 cavities in 59 patients were analyzed. A summary of the baseline patient and tumor characteristics is provided in Table 1. The median age was 61 years (range 26-86 years), 55 (90%) had GTR and the most common primary tumor was non-small cell lung cancer in 28 (46%) patients. Tumors were evenly divided between dural (i.e. abutting the dura without frank invasion) 29 (48%) and non-dural involvement 32 (52%). The mean preop maximum tumor diameter and volume were 3.2 cm (range, 1.2-7.5 cm) and 16 cm³ (range 1.6-52.1 cm³), respectively. The median and mean times between surgery and the treatment planning MRI were 28 and 35 days (9-109).

Table 1.

Patient and tumor characteristics

Characteristic	Value or Number
Number of patients	59
Number of resected brain metastases	61
Median age in years (range)	61 (range 26-86)
Extent of resection
Gross total	55 (90%)
Subtotal	6 (10%)
Histology
NSCLC	28 (46%)
Breast	15 (25%)
Colorectal	4 (7%)
Renal cell	4 (7%)
Melanoma	4 (7%)
Other	6 (10%)
Location
Supratentorial	44 (72%)
Infratentorial	17 (28%)
Mean greatest tumor diameter	3.2 cm (range 1.2-7.5)
Mean tumor volume	16 cm3 (range 1.6-52.1)
Number of brain metastases
1	39 (64%)
2	16 (26%)
>2	6 (10%)
Dural involvement
Yes	29 (48%)
No	32 (52%)

Post-operative cavity volume dynamics:

Kolmogorov-Smirnov normality test showed the relative change in tumor did not satisfy a normality assumption (P<0.01) and hence the sign test was used to assess reduction in tumor. There was a significant reduction of the planning cavity volume of 4 cm³ compared to the preop tumor volume, p=0.03. The mean cavity volume was 12.4 cm³ (range, 1.4-53.2 cm³) on the planning MRI. Overall, the volumes of 37 cavities (60.7%) were smaller and 24 (39.3%) were larger.

Factors predictive of cavity volume reduction:

For tumors with a preop diameter ≥3cm, 26/36 (72.2%) cavities were smaller than preop tumor volume and a significant reduction in the planning cavity volume was observed (11.6%, p=0.01) at a median of 28 days after surgery (range 10-109). For tumors <3cm, 14/25 (56%) cavities were larger than the preop tumor volume, and an increase (34.4%, p=0.69) in average volume was observed at the time of the planning MRI compared to preop tumor volume. Surgical cavities within the early interval (<21 days) were on average 58.6% larger than the preop tumor volume (p=0.27). Within the intermediate and late intervals, the cavity volume, however, decreased on average by 5% and 7%, respectively. Table 2A shows the relative cavity volume change in all the potential predictive factors collected. In the early interval, although the cavities for both large and small tumors were greater than the preop tumor volume, the cavities for small tumors were almost double their tumor volume (p=0.06). For large tumors, the cavity volume was 28% (p=0.0007) smaller at the intermediate interval as compared to the preop tumor volume. Table 2B shows the relative cavity volume change over time for tumors ≥3cm and <3cm.

Table 2A.

Relative change in volume for each potential predictive factor

Factor	Mean (SE) relative change in volume of cavity to tumor	P-value
Tumor Diameter
Small (<3 cm, n=25)	34.4% (20.7%)	0.69
Large (≥3.0 cm, n=36)	-11.6% (10.7%)	0.01
Location
Supra-tentorial	10.8% (12.8%)	0.17
Infra-tentorial	-1.9% (20.9%)	0.63
Time interval from surgery
<21 days	58.6% (28.5%)	0.27
22-42 days	-4.9% (13.5)	0.02
>42 days	-6.9% (18.5%)	0.58
Pre-op Dural Involvement
Yes	-15.3% (10.7%)	0.008
No	27.6% (17.7%)	0.86
Extent of Resection
GTR	8.3% (11.6%)	0.18
STR	-2.8% (32.2%)	0.69
Age
<60 years	-6.36 (12.2)	0.01
>60 years	20.4% (17.7%)	1

SE=standard error

Table 2B.

Relative Change in volume stratified by time from surgery for small and large tumors

Factor	Time interval from surgery	Mean relative change in volume of cavity to tumor	P-value
Small Tumor	<21 days	97.9%	0.06
(<3 cm, n=25)	22-42 days	24.1%	1
	>42 days	-22.3%	0.63
Large Tumor	<21 days	24.8%	1
(≥3.0 cm, n=36)	22-42 days	-27.8%	0.0007
	>42 days	0.03%	1

On univariate analysis, pre-operative tumor dimension, time from surgery, dural involvement and age were found to correlate with volume reduction (Table 3). More cavities were smaller than initial tumor volume for large vs. small tumors (26/36 vs. 11/25, p=0.02) and more cavities were smaller than initial tumor volume in the intermediate vs. early and late intervals (71% vs. 30% and 62%, respectively, p=0.04). A significantly greater proportion of cavities were smaller than the preop volume in tumors with dural involvement vs. no dural involvement (75.8% vs. 46.8%, respectively; p=0.02) and in patients <60 vs. >60 years (73.3% vs. 48.3%, respectively; p=0.04). On multivariable analysis, dural involvement (OR=4.13, 95% CI 1.16-14.69, p=0.03), maximum tumor diameter (OR=3.57, 95% CI 1.05-12.22, p=0.04) and time from surgery (p=0.04) were significant as shown in Table 4.

Table 3.

Univariate analysis of volume reduction by potential predictors

	Volume Reduction n (%)
Potential Predictor	Yes	No	P-value
Pre-op tumor diameter			0.03
<3.0 cm (n=25)	11 (44)	14 (56)
≥3.0 cm (n=36)	26 (72.2)	10 (27.8)
Location			0.86
Infra-tentorial (n=17)	10 (58.8)	7 (41.2)
Supra-tentorial (n=44)	27 (61.4)	17 (38.6)
Time from surgery			0.04
< 21 days (n=13)	4 (30.8)	9 (69.2)
22-42 days (n=34)	24 (70.6)	10 (29.4)
> 42 days (n=13)	8 (61.5)	5 (38.5)
Pre-op dural involvement			0.02
No (n=32)	15 (46.9)	17 (53.1)
Yes (n=29)	22 (75.9)	7 (24.1)
Extent of resection			1.00
GTR (n=55)	33 (60)	22 (40)
STR (n=6)	4 (66.7)	2 (33.3)
Age			0.046
≤60 Years (n=30)	22 (73.3)	8 (26.7)
>60 Years (n=31)	15 (48.4)	16 (51.6)

Table 4.

Multivariable analysis in modeling reduction in Tumor

Factor	Comparison	Odd Ratio (OR)	95% CI		P-value
			Lower	Upper
Tumor diameter	> 3cm vs < 3cm	3.57	1.04	12.22	0.04
Time interval from surgery	22-42 days vs < 21 days	7.74	1.59	37.48	0.04
	> 42 days vs < 21 days	3.34	0.56	19.63
Pre-op Dural Involvement	Yes vs No	4.13	1.16	14.68	0.03

Discussion

SRS to the post-operative cavity in brain metastasis has been increasing in clinical practice.¹¹ The recently presented N107/CEC3 trial has established the superiority of cavity SRS to WBRT in maintaining neurocognitive function[4]. However, many questions remain unanswered including optimal dose fractionation, timing of treatment and the appropriate volume coverage. This study aimed to evaluate how the planning cavity volume changes compared to the preop tumor and what factors influence this change.

Overall, there was a significant reduction of cavity volumes (mean 12.4 cm³) compared to preop tumor volume (mean 16 cm³), p=0.03. A priori, we decided to evaluate cavity volume reduction as a function of preop tumor diameter. A cut-off of 3cm was used based on previous studies showing significant differences in volume reduction with this cut-off. Shah et al. showed tumors >3cm had a greater volume reduction (16.3%) compared to tumors <3cm (5.7%) [5]. Similarly, Jarvis et al found a 47.2% decrease in volume for tumors >3cm, 25.8% for tumors 2-3 cm and an increase in volume for tumors <2cm by 56.4% [6]. In addition, Atalar et al, used a volume cut-off of 4.2 cm³, and found a similar pattern with a decrease of 35% for larger tumors compared to an increase by 46% in smaller tumors [7]. Our results are consistent with these studies with an overall significant volume reduction for larger tumors (11.6%, p=0.01). This is an important finding, as the amount of radiated normal tissue in larger tumors can be quite significant and may be associated with a high risk of radiation necrosis. The observation that larger tumors have a significant reduction in the cavity volume is reassuring and adds to the importance of resection of large tumors. The corollary is that smaller tumors on average increase in volume after surgery and therefore the volume of normal tissue radiated may be larger. In these patients, unless there is an indication for surgery, radiosurgery alone may be considered, especially given the comparable local control rates to surgical resection [8].

Knowing that the planning cavity volume reduces in large tumors compared to the preop tumor volume, we sought to identify if there was a time dependent relationship. We divided the cavities into 3 intervals (early, intermediate and late) depending on when they had their planning MRI. Overall, cavities were larger than their tumor volume in the early interval by a mean of 59% and smaller than the tumor in the intermediate and late intervals (-4.9%, 6.8%, respectively). The increase in the early interval was greater for smaller tumors (97%) versus large tumors (25%). For larger tumors (≥3cm), there was a significant relative volume reduction observed in the intermediate interval (p=0.0007). In contrast, cavities of smaller tumors show relative volume reduction only in the late time interval. This observation suggests that a longer time may be required for reduction in volume of cavities from smaller tumors as they tend to be larger than initial tumor volume especially within the first 21 days after surgery and become smaller than tumor volume only in the late time interval (-22%, p=0.6). This time-dependent relative change is similar to that observed in the study by Shah et al. which found that longer time (>1 month) after surgery was associated with greater cavity shrinkage (<1 month)[9]. Similarly, another smaller study in abstract form showed volume reduction only after day 21 from surgery[5]. In contrast, Jarvis et al. found no further reduction in the cavity size at the planning MRI compared to immediate post-operative MRI[6]. Yet another study by Atalar et al. found no association between the number of days post-resection and cavity volume change[7]. The differences between the findings of these studies is difficult to explain, but may be related to patient and tumor factors such as age, or histology. Additionally, an inherit bias resulting from the tendency to wait longer for some cavities that are expected to shrink may also have contributed to the time-dependent shrinkage observed in some studies. Table 5 compares the relevant findings in our study with that of previous studies.

Table 5.

Summary of reported studies

Study	No. of tumors	Mean tumor volume	Mean volume at immediate post-operative MRI/Planning MRI	Days between surgery and planning MRI (range)	Cavity volume change at planning MRI	Relevant findings
Current study	61	15.5 cm3	N/A 12.2 cm3	Median 28 (10-110)	-21%*	Change based on time interval: ≤21 days: 59% 22-41 days: 5% >42 days: 11% Change based on tumor size: <3cm: 5% ≥3cm: 28.6%
Jarvis et al6	43	14.23 cm3	8.53 cm3 8.77 cm3	Mean 29.8 (8-111)	No.of smaller cavities: 10 No.of larger: 13 No.of stable cavities:20φ	Change based on tumor size: ≤2cm: 56.4% 2.1-3 cm: 25.8% > 3cm: 47.2%
Atalar et al7	68	14.5 cm3 (Median )	10.1 cm3 (Median)# NR	Median 16 (2-43)	-29%*	Change based on tumor volume: >4.2 cm3 : 35% <4.2 cm3 : +46% No association between post-resection day number and volume change
Shah et al9	21Ψ	NR	7.8 cm3 4.5 cm3	Median 39 (11-77)‡	-43%¥	Change based on time interval: -13% when interval‡ <1 month -61% when interval‡ ≥1 month
Abstract Wald et5 ASTRO 2016	24Ψ	17.89 cm3	20.1 cm3 16.9 cm3	Mean 22.9 (10-49)	-10.6%¥	Change based on tumor size: <3cm: 5.7% >3cm: 16.3% Change based on tumor volume: <12 cm3: 10.5% >12 cm3: 10.7% Change based on time interval: <12 days: + 7.36% >21 days: 25.8%

^*Comparing cavity mean volume at planning MRI to mean preop tumor volume

^¥Comparing cavity mean volume at planning MRI to immediate post-operative MRI

^#Only 31 patients had immediate post-operative MRI

^‡Interval between immediate post-operative MRI and planning MRI

^ΨNumber of patients as number of brain lesions was not reported.

^φOnly proportions were reported. Stable cavity was defined as a change of ≤ 2cm³

N/A: not applicable

NR: not reported

We also analyzed other factors that may be associated with changes in the planning volume relative to the preop tumour. These included: age, pre-operative tumor diameter, dural involvement, time from surgery, extent of resection and tumor location. Age, pre-operative tumor diameter, dural involvement, and time from surgery were found to correlate with volume reduction on univariate analysis. All of these factors were significant on multivariable analysis except age. Post-operative cavities in younger patients (≤60 years) were more likely to be smaller than the preop tumor volume. This might be explained by the fact that there is more brain tissue in younger patients and after surgery the brain tissue may expand into the tumor bed. Similarly, resection of tumors with dural involvement was more likely to result in a smaller cavity. In contrast, Shah et al found no association between dural abutment and cavity volume shrinkage[9]. This may be attributed to the different degrees of dural involvement that was assessed in this study compared to the current study or the relative subjectivity in assessing radiographic dural involvement. When we looked at tumors with dural contact without frank invasion, we found no difference in volume change compared to tumors without dural contact.

A theoretical concern with delaying cavity SRS is that tumor regrowth may occur and impact local control. Jarvis et al, found 6/41 patients had local progression during the interval between the first post-operative and the planning MRI mostly among patients with STR[6]. Atalar et al. reported 7/68 cases of post-operative tumor progression before SRS, however, number of days between surgery and SRS did not affect local control[7]. Shah et al. found a much higher rate of 52%; however, no impact on local recurrence rate was observed in patients who had SRS <1 months after resection compared to > 1month[9]. In our study, no interaction between local control and timing of SRS was observed.

In addition to the inherit biases normally associated with retrospective studies, limitations of this study included the absence of an immediate post-operative MRI for all patients. Without an immediate postoperative MRI, it is not clear what changes might have occurred in the interim before the planning MRI. While previous studies assessed cavity volume change over time compared to the immediate post-operative MRI, we looked exclusively at the change from the preop MRI to the planning MRI. Other limitations in this study include heterogeneous tumour histology, and a limited number of patients that reduces the statistical power of the study. Finally, the timing of SRS was left to the discretion of the treating physician; factors such as the initial size of the metastases, when patients were referred after surgery and the functional status of patients may have affected when SRS was delivered and biased results.

Conclusion

This study found that cavities with the greatest volume reduction were those of larger tumors (≥3cm). Most of this reduction occurred 22-42 days post-operatively. Caution must be taken when treating cavities in the early post-operative phase as it may lead to irradiating more normal tissue especially in small tumors. Younger patient age and tumors with dural involvement have a higher likelihood of cavity shrinkage.

Abbreviations

SRS: stereotactic radiosurgery

WBRT: whole brain radiotherapy

fSRS: fractionated stereotactic radiosurgery

IRB: Institutional Research Ethics Board

Preop: pre-operative

GTR: gross total resection

STR: sub-total resection

Declaration

- Ethics was institutionally approved by our Institutional Research Ethics Board at Odette Cancer Center/Sunnybrook health sciences/University of Toronto, Toronto, Ontario, Canada.

- This research does not include individual patient’s material like video or images.

- All data supporting this research results are present upon request.

- This research was not externally funded.

Acknowledgements

The authors acknowledge CNS oncology team at Odette Cancer Center for their ongoing support and Al Baha university for sponsoring Dr. Alghamdi during his training.