Resection cavity radiosurgery for intracranial metastases: a review of the literature

Abstract

Purpose

This study aims to perform a systematic review of published literature addressing outcomes related to postoperative stereotactic radiosurgery (SRS) delivered to the cavity of resected intracranial metastases.

Methods

A thorough literature search was performed on all reports published in English of SRS to the resection cavity after surgical resection, in patients who did not receive immediate whole brain radiotherapy (WBRT).

Results

15 single-institution publications were identified which fit the search criteria. In 11 of the 13 studies that reported on number of lesions treated (85%), more than 50% of the patients had a single lesion. 11 publications (73%) reported the percentage of gross total resections (GTR) , which ranged from 68-100%. The predominant histology was non-small cell lung cancer. Modalities used included GammaKnife, CyberKnife, and linac-based radiosurgical platforms. Nine institutions (60%) added a margin to the post-surgical cavity. One year local control ranged between 74-91.5%. Distant brain recurrences occurred at a median of 53.8% of the time at a median of 7.8 months. Very few (<10%) patients developed symptomatic necrosis. Leptomeningeal disease incidence at recurrence was reported in four studies ranging from 4.2% to 25%, with 44.4% to 50% occurring in the posterior fossa. Salvage therapy included WBRT used 19-47% of the time at a median of 8months.

Conclusion

Postsurgical SRS is a safe and effective modality that can be used to limit recurrences in the postoperative cavity when postoperative WBRT is omitted but does not address distant intracranial recurrences. Further investigation of its efficacy and toxicity is ongoing in a randomized control trial.

1. INTRODUCTION

With an estimated 170,000 new cases of brain metastases each year that continues to rise, the management of brain metastases is not only a significant neurological complication of concern for patients and physicians alike, but also an increasingly complex problem to manage [¹]. Two landmark trials in the 1990s dictated the treatment paradigm for these patients. The first trial, published by Patchell and colleagues in 1990, and showed not only a benefit in local control at the site of metastasis with the addition of surgery to whole brain radiotherapy (WBRT), but also an improvement in overall survival from 15 to 40 weeks [²]. This study established the critical role of surgical resection of brain metastases. A second trial, also by Patchell and colleagues, published in 1998, showed improvement in both local control of disease in the resection cavity, as well as reduction of new incidence of metastases elsewhere in the brain, in patients who received whole brain radiotherapy (WBRT) in addition to surgical resection of brain metastases, compared to those who received surgical resection alone [³]. However, this trial did not show a difference in overall survival, despite showing a 30% absolute decrease in deaths from neurologic causes, from 44% to 14%, with the addition of WBRT [³]. These two trials prompted surgical resection when possible for single brain metastases, followed by the routine use of WBRT postoperatively, with the hope that as control of systemic disease becomes better controlled, the improvement in deaths from neurologic causes would translate into improvement in overall survival. However, WBRT is not without its drawbacks, including the potential for long term cognitive deficits as well as the known acute toxicities including alopecia, skin irritation, and fatigue [⁴, ⁵, ⁶]. This must be counterbalanced by the fact that surgical resection alone without radiation results in an unacceptably high rate of recurrence in the surgical bed, shown to be 59% at two years in a recent EORTC study [⁷]. To minimize these potential side effects while still providing local control in the surgical bed, postoperative radiosurgery (SRS) has increasingly gained popularity in use. However, as there continues to be a significant risk of recurrences elsewhere in the unirradiated brain (37% in the Patchell study [³] and 42% in the EORTC study [⁷]), observation of the resection cavity is not acceptable, and thus, an alternative postoperative treatment using radiosurgery has been utilized with the caveat that these patients need to be followed closely so that these recurrences can be treated.

To date, there have been no direct comparisons of outcomes after the use of WBRT compared to SRS in the postoperative setting, although one is currently accruing [⁸]. This paper aims to 1) review the current literature that is currently limited to single-institution experiences and 2) to begin to address some of the nuances in the use of radiosurgery in the postoperative setting for brain metastases.

2. MATERIALS AND METHODS

A thorough literature search of published manuscripts in the English literature via MEDLINE/Pubmed was conducted. Key words included “radiosurgery,” “resection,” “brain,” “metastasis,” and “postoperative.” Date of publication was limited to between January 1^st, 1990 and September 31^st, 2013. Fifty-three articles were identified. The goal of the search was to identify reports of postoperative radiosurgery delivered to the cavity of the resected intracranial metastases. Fifteen articles fit these criteria. The remaining articles were excluded because of either 1) publication in a language other than English, 2) described radiosurgery for benign diseases, 3) described radiosurgery to extracranial diseases, or 4) described radiosurgery for cranial metastases that did not undergo surgical resection. Abstracts from national meetings of major medical societies were not included.

3. RESULTS

Compiled results of the single institution studies are shown in Table 1-3. All of the reports are retrospective in nature from single institutions, published between 2008 and 2013. Median follow-up ranged from 9.3 months [⁹] to 24 months [¹⁰], at a median of 12.9 months. Numbers of patients included varied widely as well, and ranged from as low as 15 [¹¹] to as many as 112 [¹²], with a median of 47 patients. All of the institutions were from within the United States except for one from Japan [¹³]. The median ages of the patients ranged from 55 [¹⁴] to 64 years old [⁹], with a median age of 60 years old.

Table 1.

Patient characteristics of single institution publications on postoperative stereotactic radiosurgery for intracranial metastases

Institution	Patients (n)	Median Follow-up (month)	Median Age	% single lesion	% GTR	Predominant histology	2nd most histology	3rd most histology	Median KPS	% RPA class I/II
Osaka 13	21	21.6	61	76	86	NSCLC	Colorectal	Other	88	NR/NR
UC Irvine 15	30	NR	61.5	43.3	NR	NSCLC	Breast	Melanoma	6/23
Pittsburgh/ Sherbrooke 20	40	13	59.5	67.5	80	NSCLC	Melanoma	Colorectal	80	22.5/67.5
Allegheny 22	52	13	61	55.7	92.3	NSCLC	Breast	Unknown	36.1/49.2
Virginia 23	47	10	61	28	100	NSCLC	Melanoma	RCC, Breast	88	NR/NR
WUSTL 11	15	20	56.8	NR	80	NSCLC	Breast	RCC	NR	53.3/40
Dartmouth 9	47	9.3	64	70	76	NSCLC	Melanoma	Breast	80	NR/NR
Barrow 16	68	13.2	60	100	NR	NSCLC	Breast	Melanoma	90	NR/NR
Tufts 18	25	NR	57	NR	95	NSCLC	Breast	Other	NR	NR/NR
Stanford 12	112	11	61	63	90	NSCLC	Melanoma	Breast	21/71
Dana Farber 24	17	12.7	61.8	70.6	NR	NSCLC	Melanoma	Other	80	24/76
Henry Ford 19	85	11.2	58	62.4	68	NSCLC	Breast	Melanoma, RCC	80	16/72
Emory 14	62	9.7	55	71	81	NSCLC	Melanoma	Breast	NR	24/68
Northwestern10	56	24	58.5	100	NR	NSCLC	Breast	RCC	NR	63/37
Wake Forest17	106	NR	56.1	57.5	96.4	NSCLC	Breast	GI	NR	NR/NR
Median	47	12.9	60	67.5	86	n/a	n/a	n/a	80	24/67.5

NSCLC – nonsmall cell lung cancer. NR – not reported. GTR – Gross total resection. RPA – recursive partitioning analysis classification. PF – posterior fossa. LMD – leptomeningeal disease. KPS- Karnofsky performance status.

3.1 Number of lesions

Of the 15 series shown in Table 1, 13 of the publications presented data on the number of lesions for each patient. In 11 of the 13 series (85%), over 50% of the cases were single brain metastasis. In the series from University of California at Irvine , 43% of the patients had one lesion, 33.3% of patients had two lesions, 16.7% of patients had three lesions, and 6.7% of patients had four lesions – most of the patients had a single lesion ¹⁵. For the series from University of Virginia, only 28% of the patients had one lesion, and no details were given regarding percentage of patients with additional lesions. The median percentage of patients with 1 lesion across all 13 studies with number of lesions reported was 68%.

3.2 Histology

As shown in Table 1, in every report, the most frequently reported histology was non-small cell lung cancer. The most frequently reported 2^nd most common histology was breast cancer. The 3^rd most frequently reported histology was renal cell carcinoma. Other histologies included melanoma, colorectal, breast, and other GI cancers.

3.3 Technique – dose, modality, volume, margin

Radiosurgery techniques varied greatly, depending on the specific modality available at each institution. As shown in Table 2, 8 of the 15 institutions (53%) utilized the GammaKnife® stereotactic radiosurgery platform (GKRS). GKRS treatments were performed in a single fraction, with doses ranging from 15 [¹⁶] to 24Gy [¹¹], at a median of 17Gy, prescribed to the 50% isodose line. Median target volume ranged from 6.4 [¹⁰] to 12.65 cubic centimeters [¹⁷], with a median of 10.5cc. Three of the 8 Gamma Knife institutions added a clinical target volume (CTV) margin of one to three millimeters. The other 5 Gamma Knife institutions either did not add a margin (n= 3), or did not specify the use of a margin (n= 2).

Table 2.

Various radiosurgery modalities, target volume and dose employed each institution

Institution	Technique	Med Time Surgery to SRS (days)	Med Volume (cc)	Med Margin Dose (cGy)	Med IDL (%)	Margins added (mm)
Osaka 13	GKRS	NR	10.7	1700	50	NR
UC Irvine 15	Linac	NR	NR	1500-1800 in 1 fx, 2200-2750 in 4-6fx	NR	1-3
Pittsburgh/ Sherbrooke 20	GKRS	28	11	1600	50	1
Allegheny 22	Linac	41	3.85	1500	85	0-2
Virginia 23	GKRS	15	10.5	1900	50	2-3
WUSTL 11	GKRS	NR	NR	1600-2400	45-50	0
Dartmouth 9	Linac	23	11.1	1000	67	2
Barrow 16	GKRS	15.5	10.3	1500	50	0
Tufts 18	GKRS	27.6	NR	1500-2000	50	NR
Stanford 12	CKRS	NR	8.7	2000	79	0-2
Dana Farber 24	Linac	NR	3.49	1800	68	0
Henry Ford 19	Linac	18	13.96	1600	90	2-3
Emory 14	Linac	31.5	8. 5 cavity, 13.9 PTV	1800	88	0-2
Northwestern10	GKRS	NR	6.4	1710	50	1-2
Wake Forest17	GKRS	24	12.65	1700	50	0
Median for LINAC-based (n= 6 studies, 293 pts )		27.5	11.1	1500	85	2 (range: 0-3)
Median for GKRS (n= 8 studies, 378 pts )		24	10.5	1700	50	1 (range: 0-3)
Median for CKRS (n= 1 studies, 112 pts )		NR	8.7	2000	79	Range: 0-2

IDL – prescription isodose line. Fx fraction

The other seven institutions utilized either linac-based (n=6) or CyberKnife® platforms (n=1). These ranged from a single fraction, used by the majority of institutions, to as many as six fractions [¹⁵]. The doses ranged from 10 [⁹] to 20Gy [¹⁸] when prescribed in single fraction, and up to 27.5Gy [¹⁵] when prescribed in a fractionated manner. The median dose prescribed for linac-based platforms was 15Gy and the single CKRS report cited a median dose of 20Gy [¹²] over one to five fractions. For linac-based platforms, the doses were prescribed to isodose lines ranging from 67% [⁹] to the 90% [¹⁹], with a median of 85%. The prescription dose was delivered to a median of 79% isodose line for patients treated in the single CKRS report [¹²]. Six of these seven (86%) non-Gamma knife institutions utilized an additional margin of 1 to 3 millimeters.

3.4 Performance status

As seen in Table 1, performance status of patients in these reported series were high. Median Karnofsky Performance Status (KPS) was reported in seven of the 15 series (47%), and ranged from 80 to 90, with a median of 80. Nine of the institutions (60%) reported on patient’s Recursive Partitioning Analysis (RPA), of which KPS is a component. In seven of these nine institutions (78%), the majority of the patients had RPA class II classification. The median percentage of patients with RPA class II classification for these nine institutions was 67.5%. For the other two institutions (22%), the majority of the patients had RPA class I classification (53.3% at WUStL[¹¹] and 63% at Northwestern [¹⁰]).

3.5 Extent of resection

The percentage of tumors that underwent a gross total resection (GTR), shown in Table 1, was reported in eleven of the 15 studies (73%). In these 11 studies, the percentage of patients with GTR was very high, ranging from 68% [¹⁴] to 100% [²³], with a median percentage of patients achieving GTR across studies of 86%.

3.6 Time interval between dates of surgery and postoperative radiosurgery

Shown in Table 2, the time elapsed between craniotomy and radiosurgery was reported in nine of the 15 (60%) studies. The time elapsed was measured in days, and ranged from as low as 15.5 days [¹⁶] to as long as 41 days [²²]. The median elapsed number of days was 27.5 days for linac-based radiosurgery series, and 24 days for GKRS series, but was not reported for the single CKRS series.

3.7 Local control/recurrence and survival

Shown in Table 3, local control rates by Kaplan-Meier analysis in the postoperative bed was consistently in the range of 74% [²⁰] to 91.5% [¹²] at one year, with a median local control of 81.7%. All institutions except Stanford (death as competing risk) and University of Pittsburgh (actuarial) used Kaplan-Meier analysis to generate their local control rates. Only three institutions reported results at 2 years, and they remained high at 66.9% [⁹] to 88.4% [¹²], with a median 2-year local control rate of 75.7%. Overall, crude local recurrences in the tumor bed occurred at a median rate of 13.2%, and occurred at a median time of 7 months.

Table 3.

Treatment outcomes reported in each single institution study

Institution	1/2yr LC %	Median OS (mth)	1/2yr OS %	Crude Distant Brain Recur %	Med Time to Distant Brain Recur (month)	Crude Tumor Bed Recur %	Med Time to Tumor Bed Recur (month)	Necrosis rate	Salvage WBRT (%)	Med Time to Salvage WBRT (month)	%LMD	%PF w. LMD
Osaka 13	82/NR	20	NR/NR	48	NR	24	7	NR	NR	NR	25	44.4
UC Irvine 15	82/NR	NR	51/NR	63	NR	13.3	NR	6.6	47	NR	NR	NR
Pittsburgh/ Sherbrooke 20	74/NR1	13	57/26	54	7	27	6	5	16	4	NR
Allegheny 22	NR/NR	15	50/NR	44	16	7.7	not reached	NR	30.7	8.7	NR	NR
Virginia 23	NR/NR	10	NR/NR	NR	NR	6	2	NR	21.3	NR	NR	NR
WUSTL 11	NR/NR	20	NR/NR	60	8	26.7	11	NR	40	8	NR	NR
Dartmouth 9	85.5/66.9	NR	52.5/31.7	63	NR	16	NR	NR	45	NR	4.2	NR
Barrow 16	NR/NR	13.2	40/NR	19.1	10.6	20.6	NR	NR	NR	NR	NR	NR
Tufts 18	NR/NR	15	NR/NR	28	NR	0	not reached	NR	NR	NR	NR	NR
Stanford 12	91.5/88.42	17	62/NR	58.9	6	10.8	6	NR	28	7	NR	NR
Dana Farber 24	NR/NR	NR	93/NR	35.3	NR	11.1	NR	prophylactic steroids	23.5	NR	NR	NR
Henry Ford 19	81.4/75.7	12.1	51.5/NR	55	5.6	9	NR	8	35	NR	8.2	NR
Emory 14	78/NR	13.4	53/NR	NR	NR	17	NR	NR	19.4	NR	NR	NR
Northwestern10	NR/NR	20.5	NR/NR	37.5	10.4	8.9	15.9	NR	14.3	NR	NR	NR
Wake Forest17	80.3/NR	10.9	46.8/NR	53.8	6.9	13.2	not reached	2.8	36.8	12.6	7.5	50
All studies (median)	81.7/75.7	14.2	52/29	53.8	7.8	13.2	7	5.8	29.4	8	7.9	47.2

LC – local control. OS – overall survival. WBRT- whole brain radiation therapy. All statistical analysis done by Kaplan-Meier unless otherwise indicated.

1. Actuarial data

2. Data using death as competing risk

Overall survival can be influenced by many factors in a heterogeneous group of patients, with metastatic cancer survival that varies greatly depending on histology and systemic burden of disease. RPA and GPA have been developed to account for important prognostic variables that compose these indices. Across studies, median survival ranged from 10 months [²³] to 20.5 months [¹⁰], with a median of 14.2 months. One year overall survival across studies ranged from 40% [¹⁶] to 93% [²⁴], with a median of 52%. The median 2-year survival was only reported in two of 16 (12.6%) series: 26% from Pittsburgh [²⁰] and 31.7% from Dartmouth [⁹]. All survival analyses were performed using Kaplan-Meier.

Crude rates of recurrences elsewhere in the brain were also reported in 13 of 15 (87%) series. These varied greatly from as low as 19% [¹⁶] to as high as 63% [⁹, ¹⁵], with a median distant recurrence rate of 53.8%. These distant recurrences occurred between 5.6 months [¹⁹] and 16 months [²²], at a median time of 7.8 months.

3.8 Complications

As shown in Table 3, toxicity in general and radiation necrosis was reported in 9 of 15 (60%) series. The radiation necrosis rate was measured and reported radiographically in four series, and ranged from as low as 5% at 6 months [¹⁴], to as high as 11% [²³]. The median necrosis rate assessed radiographically was 5.8%. In the other five series, toxicity was reported in terms of clinical symptomatology. This ranged from one series with no new reported neurological deficits or grade 3 or higher toxicity [²⁴] to 26.6% with grade 2 reported toxicity in another series [¹⁵]. These results must be interpreted in the context that necrosis is regarded as difficult to diagnose and ascribe to radiosurgery [²⁵]. Wake Forest was the only institution to report details on the surgical re-operation rate, of 7/106 (6.6%) for potential toxicities related to radiation¹⁷.

3.9 Leptomeningeal disease

The incidence of leptomeningeal disease (LMD) at time of recurrence was reported in four series (26.7%). One series from Osaka, revealed an LMD rate of 44.4% for posterior fossa tumor resections, and an overall LMD rate of 25% in all tumors that were resected and treated with postoperative radiosurgery[¹⁰]. The series from Wake Forest found an overall LMD rate of 7.5%, with 50% of them in the cerebellum [¹⁷]. The other two revealed LMD failure rates of 4.2% [⁹] and 8.2% [¹⁹].

<tab3/>

3.10 Salvage therapy

The use of salvage WBRT was reported in twelve of the 15 series (80%). Utilization ranged from 19% [¹⁴] to 47% [¹⁵], with a median of 29.4%, between 4 months [²⁰] to 12.6 months[¹⁷] (median of 8 months) after SRS.

4. DISCUSSION

The management of the resection cavity of a metastatic lesion after surgical resection remains controversial. The Patchell study in the postoperative management of brain metastases serves as a benchmark for comparing results from other studies. It showed a substantial decrease in local recurrence at the site of surgery following WBRT, from 46% to 10%, as well as an equally marked decrease in recurrence at other sites in the brain, from 37% to 14% [³]. While there was no difference in overall survival between patients treated with surgery alone and surgery followed by WBRT, the incidence of neurologic deaths was lower in the latter group, 44 vs 14%, suggesting a benefit in disease control in the brain by radiation limited by poor systemic control. A recent study conducted by the EORTC showed similar results, with 2-year local recurrence in the surgical cavity dramatically reduced from 59% without postoperative whole brain radiation to 27% with addition of postoperative whole brain radiation [⁷].

Since that landmark study was published, there have been significant advances in both radiation oncology, such as the increased adoption of SRS, as well as in medical oncology, such as significant number of new chemotherapeutic agents that have prolonged survival in patients with a variety of cancers. As such, the postoperative management of metastatic lesions in the brain has become more complicated due to additional considerations that must be taken into account.

There are a number of factors that must be weighed when considering the use of postsurgical SRS. One category of such factors pertains to the features to the lesion themselves specifically, the histology, number of lesions, and residual volume. The major histology reported in studies included in this review article was non-small cell lung cancer, due to its relative high prevalence overall. However, for more radio-resistant histologies such as melanoma or renal cell carcinoma, the benefit of a hypofractionated regimen such as that employed in SRS would be theoretically greater. These histologies respond poorly to traditional fractionation schedules employed in WBRT and in these instances SRS has been shown to be a more effective option [²⁶].

In most of the included single institution studies, the majority of the patients had a single lesion. Single resection beds are not only easier to target and plan, but also suggests a lower burden of disease elsewhere in the brain that does not necessitate the preemptive treatment that would be obtained with whole brain radiation therapy. While it is difficult to delineate a concrete cut-off for maximum number of lesions, historically, the maximal number of lesions recommended for SRS treatment was three to four. However, with recent advances in SRS technique and technology, there has been an increase in the number of lesions that are treated [²⁷]. In the same vein, extent of resection is also important in that it reflects the extent of the burden of disease in the rest of the brain [²⁸]. Of the studies in this series that reported the number of patients with a gross total resection (GTR), all but two studies had at least 80% of the patients treated with a gross total resection. Delaying prophylactic radiation elsewhere in the brain is a strategy pursued in these studies, and would seem appropriate in the setting of a low burden of disease, as would be evidenced by a single lesion and/or GTR.

The other category of prognostic variables to consider lies with patient specific factors. The theoretical advantage in delaying whole brain radiation by using stereotactic radiosurgery is to delay the potential cognitive side effects of whole brain radiation for as long as possible. This is most beneficial in those patients whose projected survival exceeds the expected timeframe for manifestations of symptoms of whole brain radiotherapy. Most of the publications in this series presented either median Karnofsky Performance Score (KPS) or percentage of patients in each Recursive Partitioning Analysis (RPA) class [²⁹]. The majority of patients treated with SRS in these series had excellent performance status according to these well validated prognostic indicators. Another factor also reflective of expected survival would be control of primary disease. This is difficult to characterize and has not been recorded often in publications. However, both of these factors would be strong predictors of the potential benefit of treating the postoperative bed with SRS and are included in prognostic tool such as the RPA [²⁹].

Many of the previously discussed tumor and patient characteristics have been studied and stratified in an attempt to predict expected survival. One of the most recent assessments is the Graded Prognostic Assessment (GPA) based on multiple brain metastasis databases of the Radiation Therapy Oncology Group (RTOG). Taking histology into consideration, this has been validated in a multi-institution study to provide a relatively accurate prediction of survival [³⁰]. Similarly, this diagnosis specific GPA was validated in a retrospective study published from MD Anderson [³¹]. These tools can be utilized to give the best estimate of patient prognosis and balance the risks and benefits of SRS compared to WBRT in the postoperative setting.

Once the decision to offer SRS is made, it is unclear what the optimum length of time that should allowed to elapse between surgery and SRS. The data from this series is varied, with median time as short as 15 days to as long as 41 days, with a median of 24 days for GKRS and 27.5 days for linac-based platforms. The theoretical disadvantage to instituting treatment prematurely is that it may lead to a resection cavity that has not reached its maximal involution and lead to higher volume of normal brain volume treated accompanied by a higher risk of development of radionecrosis [³²]. Two recent publications suggest that tumor cavities do not involute significantly after postoperative day three, and that there is no benefit to delaying SRS. However, because there is always potential for the target volume to change, it is recommended to obtain the planning MRI as close to SRS delivery as possible [³³, ³⁴].

Another variable is the actual delivery system of radiosurgery. While historically the most established method is via GammaKnife®, the use of CyberKnife® and other linac-based modalities has seen rapid adoption [²¹, ³⁵]. These differing technologies employ different immobilization strategies, which have been shown to have different magnitudes of intrafraction movement [³⁶, ³⁷]. While still within the range appropriate for stereotactic treatment, this may sway some clinicians to include an additional PTV margin. Margin in the form of clinical target volume (CTV) has also been recently incorporated by the Stanford group to account for microscopic residual disease [¹²]. For those published studies with linac-based or CyberKnife modalities, there was a much higher use of a PTV margin (6/7) compared to the studies using GammaKnife (3/8). These margins increase the volume of treatment, which directly impacts the dose delivered. The RTOG 90-05 dose escalation study showed that the maximal safe dose was inversely related to diameter of the target [³⁸]. Immobilization variations also affect fractionation: GammaKnife, with its frame-based immobilization system, is much less conducive to multiple treatments (unless modified with the Elekta Extend System for relocatable non-invasive frame placement) compared to nonframe-based systems of linac-based and CyberKnife platforms. Even more importantly, the inherent dosimetry of the systems can lead to differences in dose prescription. Whereas doses are often prescribed to the 70-80% isodose line in linac-based and CyberKnife radiosurgical platforms (i.e. the maximal dose in the center of the target is 1.25-1.4 times that of the periphery), it is usually to the 50% isodose line in the GammaKnife platform (i.e. the maximal dose in the center of the target is 2 times that of the periphery) [³⁵]. This makes direct comparisons of outcome between different studies utilizing different modalities difficult.

Outcomes of SRS in the postsurgical setting, however, are collectively excellent. In this review of single institution studies, the local control rates remained consistently high at 74-91.5% at one year. Only three studies published local control rates at two years, with range of 66.9 to 88.4%. When tumor bed recurrences do happen, they typically occurred within one year except for one institution reporting on a case at 16 months [²²]. What ultimately drives utilization of salvage WBRT, however, is failure at distant sites in the brain. Given the current paucity of data, it is difficult to conclude the ultimate success rate of using salvage SRS for distant brain failures. The results of this series shows salvage WBRT being used in 14.3 to 47% of instances. What is unclear in the studies reviewed for this article, however, is if any salvage SRS was attempted because of lack of reporting on salvage SRS. While not routinely reported, it is done routinely at our own institution.

Leptomeningeal disease (LMD) carries a grave prognosis along the spectrum of CNS metastatic disease. Therefore, SRS would not be appropriate in the setting of leptomeningeal disease (LMD). Conversely, the incidence or development of LMD following postop SRS to the cavity of resected brain metastasis has been reported. A recent study from Stanford showed slightly increased risk for LMD in patients with breast cancer histology after several indices were examined including posterior fossa location [³⁹]. It has been postulated by other studies that the surgical technique, especially when employed in tumor located in the posterior fossa, with closer proximity to the cerebellar cisterns, may lead to an increased risk of LMD [¹⁷, ¹⁹]. Two of the four reports in this series suggest a higher correlation of posterior fossa location tumors and LMD recurrence, with rates of 44.4% [¹³] and 50% [¹⁷]. However, without more evidence, it is difficult to conclude specific criteria which would suggest a high enough risk of LMD such that radiosurgery would be inappropriate for treating resected posterior fossa brain metastasis.

Overall, SRS was well tolerated. Symptomatic rates of radiation injury or necrosis remained low. The highest reported rate was 26% grade 2 toxicity [¹⁵]. Every other study had less than 10% toxicity, with a median of 5.8%, and a much lower percentage of symptomatic toxicity requiring significant intervention. Only one study reported the incidence (6.6%) of re-operation for symptoms stemming from SRS [¹⁷]. Furthermore, radionecrosis is notoriously difficult to diagnose radiographically as tumoral edema can have a similar appearance, even with more advanced imaging techniques, such as MR spectroscopy and diffusion-weighted imaging [²⁵].

Also relatively absent is the reporting of neurocognitive outcomes for patients who delay WBRT by utilizing SRS in the postoperative setting. The neurocognitive deficit experienced by patients who undergo WBRT is well documented. This was demonstrated in a phase III trial which showed worse memory function at four months in those patients who underwent whole brain radiotherapy in addition to postoperative radiosurgery [⁴]. While this was not a postsurgical study, it demonstrated a significant neurologic deficit in similar patients due to whole brain radiotherapy, using a battery of well validated assessments. This has also been demonstrated in an EORTC phase III trial comparing WBRT versus observation after surgical resection or SRS of brain metastases [⁵]. This study demonstrated that the risk of neurologic deficits was higher for WBRT when compared to the risk from tumor. Conversely, a prospective RTOG trial examined baseline and serial Mini-Mental Status Exams (MMSE) after whole brain radiotherapy, and found only a significant drop with uncontrolled metastases [⁴⁰]. This was confirmed by a Japanese Radiation Oncology Study Group 99-1 trial, which also found that WBRT after SRS doubled the duration of neurocognitive stability [⁶]. These trials using MMSE may not have been sensitive to detect more subtle neurocognitive changes. A phase III RTOG trial (0614) was recently published which demonstrated a statistically significant improvement in cognitive decline using Memantine, an N-methyl-Daspartate (NMDA) receptor antagonist [⁴¹].

One potential benefit of postoperative SRS which has not been investigated is the hypothetical ability to institute chemotherapy without delay. While there have been studies on patients with brain metastases and significant systemic visceral disease who undergo primary chemotherapy before whole brain radiotherapy [⁴²], few clinicians offer this. However, with radiosurgery, the treatment volume is minimized, potentially allowing for the use of concurrent chemotherapy and SRS without risk of worse neurologic toxicity associated with juxtaposition of WBRT and chemotherapy.

Given the rather inconsistent nature of reporting in this setting, and to address the issues discussed above, we propose a standardized set of reporting parameters for future publications. They are broken down by categories pertaining to the patient, lesions, treatment, outcome, and complications, shown in Table 4. To further standardize reporting of results, we recommend reporting of local control and survival at 1 year using Kaplan-Meier analysis. Given the heterogeneous nature of this group of patients, this would allow better tailoring of therapy for the individual patient. For example, recommendations on whether or not SRS should be utilized after surgical resection of intracranial metastases could be vastly different depending on if the patient had poor KPS and five relatively radiosensitive NSCLC brain metastases, versus a healthy patient with a single relatively radioresistant melanoma brain metastasis. Such a standardized set of reporting parameters would not only provide a consistent reporting mechanism for future publications and easy comparison across studies, but also help answer these lingering questions and guide future treatment decisions.

Table 4.

Proposed reporting parameters for patients undergoing SRS after resection of intracranial metastases.

Patient Characteristics
Performance Status	KPS
Prognostic Index	RPA or GPA
Lesion Characteristics
Histology
Number of lesions
Resection status	GTR, STR, or biopsy only
Cystic vs. solid
Post-resection dimensions (AP x CC x TR)
Pre SRS dimensions (AP x CC x TR)
Median time from surgery to SRS (days)
Prior WBRT or SRS
Treatment Parameters
Modality	GKRS, CKRS, or linac-based
Total prescribed dose at periphery (cGy)
Number of fractions
Prescribed isodose line (%)
Additional margin (mm)
Treatment volume (cc)
Outcome
1 year local failure at tumor bed by Kaplan-Meier (%)
Median time to local failure (months)
1 year distant failure elsewhere in the brain by Kaplan-Meier (%)
Median time to distant failure (months)
1 year overall survival by Kaplan-Meier (%)
Incidence of LMD at time of failure by Kaplan-Meier (%)
Location of LMD failure	Supratentorial, posterior fossa, or diffuse
Rate of WBRT salvage (%)
Complications
Radionecrosis	Radiologic or pathologic diagnosis
Symptoms of radionecrosis	Yes or No
Neurologic function by standardized metric	Improved, stable, or diminished

Currently, there is also an ongoing phase III clinical trial, Intergroup N107C, seeking to answer these unanswered questions [⁸]. In this trial, patients with a resected brain metastasis and up to three small unresected metastases (up to four total) are randomized to receive either postoperative SRS or WBRT. In addition to evaluating overall survival, it also seeks to analyze domains involving memory, fluency, executive function, and motor dexterity as primary endpoints. This will allow collection of level one data and permit a direct comparison of WBRT and SRS after surgical resection of brain metastases, evaluating critical endpoints of survival as well as neuropsychometric domains.