Executive summary from American Radium Society’s appropriate use criteria on neurocognition after stereotactic radiosurgery for multiple brain metastases

Abstract

Background

The American Radium Society (ARS) Appropriate Use Criteria brain malignancies panel systematically reviewed (PRISMA [Preferred Reporting Items for Systematic Reviews and Meta-Analyses]) published literature on neurocognitive outcomes after stereotactic radiosurgery (SRS) for patients with multiple brain metastases (BM) to generate consensus guidelines.

Methods

The panel developed 4 key questions (KQs) to guide systematic review. From 11 614 original articles, 12 were selected. The panel developed model cases addressing KQs and potentially controversial scenarios not addressed in the systematic review (which might inform future ARS projects). Based upon quality of evidence, the panel confidentially voted on treatment options using a 9-point scale of appropriateness.

Results

The panel agreed that SRS alone is usually appropriate for those with good performance status and 2–10 asymptomatic BM, and usually not appropriate for >20 BM. For 11–15 and 16–20 BM there was (between 2 case variants) agreement that SRS alone may be appropriate or disagreement on the appropriateness of SRS alone. There was no scenario (among 6 case variants) in which conventional whole-brain radiotherapy (WBRT) was considered usually appropriate by most panelists. There were several areas of disagreement, including: hippocampal sparing WBRT for 2–4 asymptomatic BM; WBRT for resected BM amenable to SRS; fractionated versus single-fraction SRS for resected BM, larger targets, and/or brainstem metastases; optimal treatment (WBRT, hippocampal sparing WBRT, SRS alone to all or select lesions) for patients with progressive extracranial disease, poor performance status, and no systemic options.

Conclusions

For patients with 2–10 BM, SRS alone is an appropriate treatment option for well-selected patients with good performance status. Future study is needed for those scenarios in which there was disagreement among panelists.

Key Points.

1. SRS alone is usually appropriate for 2–10 asymptomatic brain metastases.

2. SRS alone is usually not appropriate for >20 brain metastases.

Importance of the Study.

Emerging data show no detriment in overall survival with the omission of WBRT in patients with 2–4 BM treated with resection or SRS. Single-arm retrospective and prospective data show lesion number to have little or no effect on overall survival among those with ≤10 BM. The impact on neurocognitive function of SRS alone versus WBRT in patients with multiple BM is therefore of particular interest. This systematic review was based upon key questions focusing on neurocognitive outcomes in patients treated with SRS. An expert panel voted on treatment options for specific case variants. This methodology, used by the ARS Appropriate Use panels, facilitated the development of consensus guidelines that will be useful to practicing neuro-oncologists, radiation oncologists, neurosurgeons, and medical oncologists. The panel for brain malignancies utilized this process to identify areas of controversy that may benefit from future evidence-based guideline development.

Brain metastases (BM) are the most common intracranial adult tumor and an increasingly important cause of morbidity and mortality.¹ The incidence of BM depends on cancer type, histology, and stage.^1,2 In a US population-based study, 2% of patients with cancer, and 12% of those with metastatic disease, presented with BM at initial diagnosis.¹ Higher incidences are anticipated due to some patients not screened for BM with brain imaging and the potential for underreporting of BM to population-based registries. Many more patients (with rates dependent on the aforementioned factors) develop BM over their disease course. Recent estimates from the American Brain Tumor Association suggest ~200 000–300 000 cases annually.³ For patients with multiple BM, radiotherapy options include whole-brain radiotherapy (WBRT) and/or stereotactic radiosurgery (SRS) in single- or multifraction dose delivery.

SRS versus WBRT + SRS for 1–4 Brain Metastases: Survival and Intracranial Tumor Control

Five prospective studies^4-8 randomized patients with 1 BM (48–81% of patients) or 2–4 BM, treated with resection (n = 2 studies) or SRS (n = 5 studies), to receive or not receive WBRT (Table 1). These studies, recent meta-analyses,^9,10 and a systematic review¹¹ have shown no detriment in overall survival (OS) with omission of WBRT, though at the expense of greater risk of recurrence of treated BM (local recurrence) and development of new BM (distant-brain recurrence).

Table 1.

Summary of tumor control outcomes and need for salvage in four randomized controlled studies of SRS alone vs WBRT plus SRS

Study First Author	Patients	Randomization	1-Year Local Recurrence	1-Year Distant Recurrence	Rate of Salvage Therapy	Rate of Salvage WBRT	Median Survival (M)
Aoyama4	132 patients with	SRS	28%	64%	43%	16%	8.0
	1–4 metastases,		11% (P = 0.002)	41% (P = 0.003)	16% (P < 0.001)	0%	7.5 (P = 0.42)
	all ≤3 cm
	- 48% with 1 met.	WBRT+SRS
Chang5	58 patients with	SRS	23%	55%	87%	33%	15.2
	1–3 brain metastases		0% (P = 0.01)	27% (P = 0.02)	7%	0%	5.7 (P = 0.003)
	- 57% with 1 met.	WBRT+SRS
Kocher6	359 patients with	SRS or resection	~31%	~48%	} 51%	} 31%	} 10.9
			~57%	~42%
	1–3 brain metastases	WBRT + SRS or WBRT + resection	~19%	~33%	} 16%	} 3%	} 10.7 (NS)
			~27% (P < 0.003)*	~23% (P < 0.001)*
	- 81% with 1 met.
Brown7	213 patients with	SRS	23%	30%	32%	16%	10.4
	1–3 metastases,		10% (P = 0.003)	8% (P < 0.001)	8% (P < 0.001)	1%	7.4 (P = 0.92)
	all ≤3 cm,	WBRT+SRS
	non-brainstem
	- 52% with 1 met.
Hong8	207 patients with	SRS or resection	34%	51%	NR	NR	13.0
	1–3 metastases,	WBRT + SRS	20% (P = 0.03)	42% (P = 0..22)			16.5 (P = 0.86)
	all melanoma
	- 60% with 1 met.

*P-values comparing the addition vs omission of WBRT. NR = not reported; NS = not significant.

In an individual patient meta-analysis of 364 patients from 3 of the aforementioned randomized studies,^4–6 WBRT + SRS versus SRS alone reduced rates of distant-brain recurrences from 53% to 34% and local recurrences from 27% to 12%, with no detriment in OS.⁹ A 2018 Cochrane review of 5 studies reported similar quantitative findings (in terms of hazard ratios [HRs]).¹⁰ In the above-mentioned individual patient meta-analysis,⁹ 1 versus >1 BM was associated with significantly worse distant-brain control (HR = 0.63, 95% CI = 0.46–0.88) and OS (HR = 0.72, 95% CI = 0.57–0.90). Among those ≤50 years old, WBRT was associated with worse OS.

SRS for Multiple Brain Metastases: Impact of Number of Brain Metastases and Target Volume

Several studies^12–23 have shown that intracranial tumor burden (ie, net volume of BM) may be more prognostic for survival outcomes than number of BM (Table 2), with many of these studies^{12,13,18,20–23} including some patients (15–39% in 6 studies) who received SRS for salvage after prior WBRT; one set of analyses^12,13 also included 46% of patients who received upfront WBRT + SRS. The role of SRS alone for >4 BM is not well supported by randomized studies and is considered controversial²⁴; however, data (discussed below) are emerging.²⁵ Yamamoto and colleagues recently outlined the history of utilizing SRS for multiple (up to dozens) BM dating back to the 1990s.²⁶ While outcome data for patients with >10 BM selected to receive SRS^{12–14,19,20,22,23,26–40} date back decades (with some studies including patients with >20 BM^{20,22,23,27,29,31–35,39,40}), most of these studies included <100 patients with >10 BM, and most^{12,13,20,22,23,27–29,31–34,36,37,39} did not exclude patients who received SRS for salvage after prior WBRT or as a boost with upfront WBRT. Four studies^14,19,35,40 (discussed below), in which some or all patients underwent SRS alone for >10 BM, excluded patients with prior WBRT.

Table 2.

Select studies analyzing impact of net BM target/tumor volume on outcomes of patients treated with radiosurgery

Author Institution	Patients	Brain Metastases (various histologies)	WBRT	Impact of Number of Tumors and Target/Tumor Volume on Patient Outcomes
Bhatnagar12 U. Pittsburgh	N = 205€	n = 4–18 (median 5)	Prior: 38%	Multivariate analysis (MVA) for OS: number of BM (P = 0.33)
		Tumor: 0.6–51.0 (median 6.8) cc	w/SRS: 46%	total treatment volume (P = 0.002)
Serizawa14 Japan-multi	N = 2246£	Total n = 10 352	None	2–4 vs 5–10 lesions: no difference in rate of salvage tx
		Max tumor: 0.1–47.6 (median 2.8) cc		>15 cc total volume worse (P < 0.0001) vs ≤15 cc
		Total tumor: 0.1–50.0 (median 4.5) cc
Baschnagel15 William Beaumont	N = 251	n = 1–14	None	MVA for OS: number of BM (P = 0.098)
		Max tumor: 0.1–14.2 (median 1.2) cc		total treatment volume (HR = 1.04, P = 0.046)
		Total tumor: 0.1–14.2 (median 1.4) cc
Likhacheva16 MDACC	N = 251	n = 1–9 (median 2; total = 542)	None	MVA for OS: number of BM ≥4 (P = 0.17)
		Max tumor: 0.03–21.9 (median 0.66) cc		total tumor volume >2 cc (HR = 4.56, P < 0.001)
		Total tumor: 0.03–22.9 (median 0.89) cc
Yamamoto17 Japan-multi	N = 1194	n = 1–10	None	Univariate analysis (UVA) for OS: total treatment volume ≥1.9 cc (P < 0.0001)
		Total tumor: 0.01–14.96 (mean 2.84) cc		MVA for OS: 2–4 vs 5–10 lesions (P = 0.094)
				total treatment volume ≥1.9 cc (HR = 1.17, P = 0.24)
Limon20 Duke U.	N = 59	n = 4–23 (median 5)	Prior: 39%	UVA for OS: total PTV <10 vs ≥ 10 cc (P = 0.0001)
		Total tumor: 4.8 (0.7–28.8) cc		largest lesion <5 vs ≥5 cc (P = 0.0006)
				4–5 vs ≥6 lesion (NS)
				MVA for OS: PTV ≥10 cc (P = 0.0003)
Emery18 U. Viriginia	N = 300	n = 1‒3; total	Prior: 27%	UVA for OS: number of lesions—not significant.
				total treatment volume of supratentorial tumors (75%)
		n = 817 individual tumors: 0.01–65.01 cc		(HR = 1.44, P = 0.004)
Ali19 Japan & Australia-multi	N = 5750¶	n = 1‒10	None	MVA for OS: 2–4 vs 5–10 lesions (P = 0.69)
		Total tumor: 0.65–16.45 cc		5–10 vs >10 (P = 0.014) and 2–10 vs >10 (P = 0.002)
				total tumor volume (HR = 1.007, P =< .001)
Routman21 Mayo Clinic	N = 391	n = 2‒5	Prior: 15%	MVA for OS: 2–4 vs 5+ lesions (P = 0.19)
		Total tumor: 0.07–81.6 (mean 3.4) cc		total treatment volume <5 vs 5–10 cc (HR = 1.33, P = 0.068)
				total treatment volume <5 vs >10 cc (HR = 1.64, P = 0.0014)
Izard22 Macquarie U.	N = 180	n = 1–47 (median 5.5); total n = 1523	Prior: 18%	UVA for OS: 1–3 vs 4–9 vs >9 lesions (P = 0.2)
		Total tumor: 6.1-5441 (median 568.5) cc		<3.2 cc vs ≥3.2 cc (P < 0.01)
				MVA for OS: tumor volume (HR = 1.21, P < 0.01)
Hamel-Perreault23 U. Sherbrooke	N = 103§	n = 5–19 (mean 7)	Prior: 61%	UVA for OS: number of BM ≥4 (P = 0.21)
		Max tumor: 0.02–16 (mean 1.1) cc		treatment volume >6 mL (HR 2.54, P = 0.006)
		Total tumor: 0.06–28 (median 2.0) cc
Yamamoto26 Katsuta Hospital Mito Gamma House	N = 934*	N = 2–40+	Prior: 7%	UVA for OS: 2–9 vs ≥10 BM (P = 0.29)
		Total tumor: 0.06–106.7 (mean 10.1) cc		among patients with ≥10 BM, net volume ≥10 cc vs <10 cc
				for OS- UVA (HR = 1.39, P < 0.0001); MVA (P = 0.19)
Yamamoto38 Tokyo Women’s Medical U.	N = 2089	N = 2–15	Prior: 4%	UVA for OS: 5–15 vs 2–4 (HR = 1.169, P = 0.001)
		Total tumor: median 5.51 (IQR 1.84–13.38) cc		MVA for OS: total volume >10 cc not sign. for 2–4 or 5–15 BM
				subgroups (not analyzed over entire cohort)
Minniti40 U Siena	N = 40	N = 10–21 (median 13)	None	MVA for OS: total volume (HR = 1.6, P = 0.04)
		Total tumor: 2.0–12.3 (median 4.7) cc

All studies are retrospective aside from studies by Yamamoto et al (2013) and Izard (2019).

Multi = multi-institutional.

N = number of patients; n = number of tumors/lesions; tx = treatment/therapy.

^€ N = 6 (3%) with >10 BM

^£ N = 369 (16%) with >10 BM

^¥Not significant for tumors within the brainstem or cerebellum

^¶ N = 981 (17.1%) with >10 BM

^§ N = 22 (21%) with ≥10 BM

*934 case-matched (using the propensity score-matching method) patients among 2371 patients with 2+ BM (740 with ≥10 BM); some overlap with other studies by Yamamoto et al.

SRS Alone for Multiple Brain Metastases—Overall Survival

Selected studies of SRS alone for multiple BM are shown in Table 3; the largest series are from the prospective Japanese JLGK0901 study¹⁷ and retrospective analyses^14,41–43 performed to evaluate potential inclusion factors for JLGK0901. A 2010 retrospective study (2 institutions) analyzed 1508 patients treated with SRS for 1–10 BM, with a net tumor volume <15 cc and largest BM <10 cc.⁴² On multivariate analyses, OS was significantly better in patients with 1 versus 2–4 BM, though not for those with 2–4 versus 5–10 BM. In a 2012 retrospective study of 2246 patients,¹⁴ >10 BM was associated with a significantly greater need for salvage therapy compared with 5–10 BM (P = 0.023), and served as a cutoff for the JLGK0901 trial. Tumor volume was a significant factor for local control, and net tumor volume >15 cc was significantly (P < 0.0001) associated with worse OS, neurologic survival (survival without death from neurologic cause), and qualitative survival (survival without impaired activities of daily living). A study that compared outcomes after SRS alone for 1–4 versus 5–89 BM showed no significant difference between these groups in OS, death from progression of BM, neurologic deaths, or receipt of salvage therapy.⁴³

Table 3.

Select studies analyzing impact of number brain metastases on overall survival and need for salvage therapy after radiosurgery-alone

Study Country	# Brain Metastases	# Patients: Grouped by # Brain Metastases						Outcomes Analyzed by Number of Metastases
		All	1	2–4	5–10	5–15+	>10
Serizawa41Japan	1–10	1508	565 (37%)	577 (38%)	366 (24%)	NA	NA	OS: 1 vs 2–4 (HR = 0.70; P < 0.0001)
								2–4 vs 5–10 (HR = 1.12; P = 0.10)
Serizawa42Japan	1–10+ §	2246	982 (44%)	766 (34%)	498 (33%)	NR	369 (16%)	need for salvage therapy: >10 vs 5–10 metastases (P = 0.023)
Yamamoto43Japan	1–89 *	2553	1553 (61%)		1,000 (39%)			For 1–4 vs
								5+ metastases-OS: worse in 5+ group (HR 1.18, P = 0.01)
								-neurologic survival (without neurologic death): similar (P = 0.77)
								salvage therapy: similar in both groups
JLGK090117,44Japan	1–10	1194	455 (38%)	531 (44%)	208 (17%)	NA	NA	OS: 1 vs 2–4 (HR = 0.76; P = 0.0004 for all patients)
								1 vs 2–4 (HR = 0.81; P = 0.0015 for patients aged ≥65)
								2–4 vs 5–10 (HR = 0.97; P = 0.78)
								2–4 vs 5–10 (HR = 0.97; P = 0.77 for patients aged ≥65)
Ali19Japan & Australia	1–10+ §	5750†	1,753 (30%)	1,872 (33%)	1,144 (20%)	NR	981 (17%)	OS: 1 vs 2–4 (HR = 0.92; P = 0.010)
								2–4 vs 5–10 (HR = 0.95; P = 0.17)
								5–10 vs >10 (HR = 0.91; P = 0.025)
								2–10 vs >10 (HR = 0.88; P < 0.001)
Hughes358 US institutions	1–15	2089	995 (48%)	882 (42%)	NR	212 (10%)	NR	OS: 1 vs 2–4 (HR = 0.73; P < 0.01)
								2–4 vs 5–15 (HR = 1.11; P = 0.25)

NR = not reported; NA = not applicable.

*This study included a separate case-matched comparison (n = 1096 patients) for patients with 1–4 vs 5–51 brain metastases.

†Included brain metastases from lung (n = 3745), breast (n = 710), gastrointestinal (n = 692), and renal (n = 321) cancers as well as melanoma (n = 282).

§Upper limit not reported.

The phase II JLGK0901 study¹⁷ enrolled 1194 patients with 1–10 BM (net tumor volume <15 cc and largest BM <10 cc). Similar to retrospective studies from the JLGK0901 authors, OS was significantly better in patients with 1 versus 2–4 BM, though not for those with 2–4 versus 5–10 BM,¹⁷ even with analyses restricted to patients (n = 693) aged ≥65 years.⁴⁴ Maximum BM diameter (≥1.6 vs <1.6 cm) and cumulative tumor volume (≥1.9 vs <1.9 mL) were adverse factors for OS on univariate but not multivariate analyses. Local recurrence of treated BM was similar across the 1, 2–4, and 5–10 BM subgroups (ranging from 10–16%, P = 0.78), while distant-brain recurrence occurred in 48% of those treated for 1 BM and >60% for those treated for 2–4 and 5–10 BM. With longer follow-up (median 46 mo among censored observations) and using case-matched patients, local recurrence rates did not significantly differ between patients treated for 1 versus 2–4 versus 5–10 BM.⁴⁵

In a Japanese/Australian study¹⁹ of 5750 patients treated with SRS alone for BM, significant albeit “modest” differences in OS were seen for 1 versus 2–10 (HR = 1.110, P = 0.001) and 2–10 versus >10 (HR = 1.128, P = 0.002) BM. With exclusion of breast cancer (for which BM number did not significantly impact OS), each increment of 6–7 lesions was associated with a ~4% increase in hazard of death. Cumulative intracranial volume was associated with an HR for survival of 1.015 per cc (P < 0.001). From a multi-institutional study of 2089 patients treated with SRS alone,³⁵ for 2–4 versus 5–15 BM, OS was not significantly different (P = 0.25), while distant-brain progression (P = 0.01) and developing new BM over time (6.1 vs 11.7 BM per year) was significantly better.

A systematic review published in 2017 summarized 10 studies that included only patients with ≥4 BM treated with SRS alone; an additional 5 studies were discussed, though excluded for not reporting distant-brain control.²⁵ The authors described the wide range of reported outcomes, attributable to heterogeneity between studies with respect to histology, extracranial disease status, and definition of outcomes (such as local tumor control). They conclude that data from phase III studies (of which some are accruing) are needed.

Complicating interpretation of the aforementioned analyses are the many variables known to impact OS and recent development of novel systemic therapies with potential efficacy against BM (a topic for future American Radium Society [ARS] guidelines). Adverse factors for OS (other than number and volume of BM) in the aforementioned (and other) studies include older age, poorer performance status, male sex, presence of neurologic symptoms, higher Radiation Therapy Oncology Group (RTOG) recursive partitioning analysis (RPA) class, uncontrolled primary cancer, presence and extent of extracranial metastases, and lung cancer primary.^{17,19,26,35,38,42–44} The diagnosis-specific graded prognostic assessment (DS-GPA) incorporates many of these factors,^46,47 as well as molecular factors.^48–50 For many cancer types, DS-GPA considers number of BM, but only 1 versus 2–3 versus >3. Brain metastases velocity, reflecting the rapidity of development of new metastases and relevant in consideration of salvage therapy after initial SRS, is another important factor correlated with OS.^51–53

Omission of WBRT for Brain Metastases—Neurocognition

Accepting that for >1 BM, the number of BM has a modest impact on OS after SRS alone, more clinically significant outcomes in the decision of SRS versus WBRT are neurocognitive effects and quality of life (QoL).⁵⁴ From survey data of 104 patients treated at the University of Pittsburgh with SRS for BM, the addition of WBRT was associated with greater patient-reported risks of problems with short-term memory (P < 0.0001), long-term memory (P = 0.03), concentration (P = 0.0007), and mood (P = 0.0008).⁵⁵ The authors recognized “limitations of this kind of survey” and reported that “treatment selection was fairly random with a tendency to use WBRT with increasing numbers of tumors,” though they did not specify the number of BM treated in this cohort. Cognitive effects can worsen or develop shortly after starting WBRT, particularly verbal memory, which has been a primary endpoint in several studies discussed below.^56,57 Patients with BM tend to have compromised neurocognitive function at baseline, perhaps related to effects from BM (which would be location dependent), prior chemotherapy, and/or general functional decline.^57–60 A recent study from the University of Siena prospectively analyzed neurocognitive outcomes and functional capacity in 40 patients treated with SRS alone for ≥10 BM; a significant deterioration in immediate recall, delayed recall, and/or recognition developed in <20% of patients, with maintenance of baseline function in the majority.⁴⁰

In order to better characterize the impact of treatment choice on neurocognitive outcomes in patients with multiple BM, the ARS Appropriate Use Criteria group for brain malignancies sought to review published literature on neurocognitive outcomes after SRS for patients with multiple BM. Hippocampal avoidance with WBRT, via modulated radiotherapy beams, is another strategy used to lower risks of neurocognitive decline (specifically memory and recall),^61,62 though it was not included in this literature search, which focused on SRS.

Methods

The panel was composed of radiation oncologists, neuro-oncologists, and neurosurgeons (nominated by the committee chair and approved by the President-elect of ARS) with expertise in management of BM. Two panel members identified themselves as private practice/community radiation oncologists, while the others were associated with academic centers (a distribution which may have impacted voting). The panel developed 4 key questions (KQs) to guide the literature search and selection of studies for this report:

• • KQ1: Is there an advantage or detriment, with respect to neurocognitive and/or QoL outcomes, to treating 2–4 BM with SRS alone versus WBRT alone?

• • KQ2: Is there an advantage or detriment, with respect to neurocognitive and/or QoL outcomes, to treating ≥5 BM with SRS alone versus WBRT alone?

• • KQ3: After SRS alone for multiple BM, what is the impact of BM number on neurocognitive function?

• • KQ4: After SRS alone for multiple BM, what is the impact of net BM volume on neurocognitive function?

The search strategy (outlined in Appendix 1) used both keywords and controlled vocabulary combining terms for radiotherapy, brain metastatic tumor, and neurocognition. Databases searched on May 20, 2018 included OVID Medline, OVID Embase, Web of Science, PubMed, and Cochrane.

The search yielded 18 391 records; after removing duplicates, there were 11 614 original articles. After initial screening, 48 articles were selected to be reviewed in-depth by 2 panel members (divided among the group), who evaluated quality of study design and relevance to the KQs. After final screening, 12 articles were included in the evidence table (Supplementary Table 1) used for this guideline.

The panel then developed variant cases mirroring those cases commonly seen in their clinics and that addressed the KQs. Based upon quality of evidence, the panel confidentially voted on appropriateness of each treatment option, using a 9-point scale (Supplementary Table 2). Agreement versus disagreement was determined by the relative number of outliers (ie, for 11–13 votes, 4+ outliers are considered disagreement). After each of 2 consecutive preliminary voting rounds, those clinical scenarios for which there was disagreement were discussed among the group, followed by a third confidential voting round.

Additional information on ARS Appropriate Use Criteria methodology is provided online (http://www.americanradiumsociety.org/page/aucmethodology).

Results

Appendix 2 shows each of the 6 case variants and the voting for each of several treatment options that were selected for voting.

For patients with ≤0.5 cm, asymptomatic BM from non-small-cell lung carcinoma (NSCLC; without targetable mutations): for 2–4 or 5–10 BM the panel agreed that SRS alone was appropriate, whether BM were diagnosed at initial NSCLC diagnosis (case variant 1) or at time of extracranial progression while on systemic therapy (case variant 2). For 11–15 BM, most panelists (11 of 13 votes) felt that SRS alone for case variant 2 may be appropriate, while there was disagreement on SRS alone (7 of 12 voting for may be appropriate) for case variant 1. For 16–20 BM, most on the panel (10 of 13 votes) felt that SRS alone for case variant 1 may be appropriate, while there was disagreement on SRS alone for case variant 2 (7 of 12 voting for may be appropriate). For >20 BM, most considered SRS alone usually not appropriate for case variants 1 and 2. Some panelists routinely offer SRS to select patients with >20 BM, though all concurred (with no votes of SRS alone being usually appropriate) that more studies are needed. Table 4 summarizes the voting for these case variants.

Table 4.

Panelists voting on appropriateness of SRS alone

# Brain Metastases*	Timing of Brain Metastases†	SRS Alone
		Panel Voting	Recommendation
2–4	At initial diagnosis of NSCLC	agreement	Usually appropriate§
2–4	At time of extracranial progression	agreement	Usually appropriate§
5–10	At initial diagnosis of NSCLC	agreement	Usually appropriate
5–10	At time of extracranial progression	agreement	Usually appropriate
11–15	At initial diagnosis of NSCLC	disagreement	Usually/may be appropriate¥
11–15	At time of extracranial progression	agreement	May be appropriate
16–20	At initial diagnosis of NSCLC	agreement	May be appropriate
16–20	At time of extracranial progression	disagreement	Usually not/may be appropriate‡
>20	At initial diagnosis of NSCLC	agreement	Usually inappropriate
>20	At time of extracranial progression	agreement	Usually inappropriate

*Asymptomatic, <5 mm BM from non-small-cell lung cancer (NSCLC) without targetable mutation.

†2 general case variants of: 1) patient diagnosed with brain metastases at time of NSCLC diagnosis; and 2) patient diagnosed with brain metastases at time of progression of NSCLC while on systemic therapy (and amenable to further systemic therapy). The separate grouping, based upon number of metastases, represented case variants within the broader grouping of timing of brain metastases diagnosis.

§All panelists’ votes fell within this category.

¥All votes fell within “usually appropriate” (5 of 12 votes) or “may be appropriate” (7 of 12 votes) categories.

‡Most votes fell within “may be appropriate” (7 of 12 votes) or “usually not appropriate” (4 of 12 votes) categories.

For 2 asymptomatic BM (1.3 cm pontine and 0.4 cm left frontal) from NSCLC (case variant 3), the panel mostly agreed that SRS alone, either single- or multifraction,^63–66 is usually appropriate, and all agreed that multifraction SRS for the brainstem lesion and single-fraction SRS to the frontal BM is usually appropriate. The panel felt that WBRT alone⁶⁷ is usually inappropriate (9 of 11 votes) for this case, and there was disagreement on appropriateness of hippocampal sparing⁶¹ WBRT alone or with SRS boost.

For 3 asymptomatic (largest 2.2 cm) BM from melanoma (case variant 4), SRS alone to all sites or resection of the largest BM followed by SRS to all sites^68–74 was considered usually appropriate by all and most voters, respectively. WBRT alone⁶⁷ or WBRT + SRS boost was considered usually not appropriate by most (11 of 12) panelists; there was disagreement on appropriateness of resection followed by WBRT,^72–74 which was considered usually not appropriate (6 of 12) or may be appropriate (6 of 12) by all panelists, with no panelists considering this option as usually appropriate.

For a patient with newly diagnosed NSCLC, with 13 BM (≤2.5 cm) and new-onset right-sided hemiparesis (from largest lesion) that partially responded to corticosteroids (case variant 5), the panel mostly agreed that resection of the largest BM (in premotor area) and SRS to all sites or fractionated SRS^{63–65,75–77} to premotor cortex BM and single-fraction SRS to all other sites were usually appropriate. There was disagreement on the use of WBRT alone or WBRT after resection, single-fraction SRS to all sites or multifraction SRS to all sites.

For a patient with Karnofsky performance status (KPS) of 60 and 6 asymptomatic BM (0.2–1.5 cm) from NSCLC (case variant 6), the panel mostly (9 of 12) agreed that hospice care was usually appropriate; there was disagreement on appropriateness of standard WBRT alone, hippocampal sparing WBRT alone, SRS for all BM, or SRS for selected (ie, largest or critically located) BM.

Discussion

Compelling single-institution and multi-institution single-arm studies suggest that SRS alone for multiple BM results in low risks of neurocognitive decline, albeit with limited follow-up due to poor OS after treatment for BM. Randomized comparisons of SRS ± WBRT demonstrate superior neurocognitive outcomes with omission of WBRT, though studies include a large proportion of patients with a single BM and exclude patients with more than a few (n = 3–4) BM. Data on neurocognitive outcomes after SRS alone for >4 BM are currently lacking though being assessed in ongoing prospective studies.

Single-Arm Studies of Neurocognition After SRS Alone for Multiple Brain Metastases

A few studies have analyzed factors affecting neurocognitive and QoL outcomes after SRS for BM.^58,59,78,79 In a Dutch study of 97 patients treated with fractionated SRS alone for 1–4 BM (56% with >1 BM), neurocognitive testing was performed before SRS and at regular intervals following SRS.⁵⁹ Net tumor volume >12.6 cc was associated with worse information processing speed (P = 0.069) and verbal memory (P = 0.077) at baseline and significantly (P = 0.02) worse information processing speed over time (3–6 mo). QoL physical functioning and fatigue measures also significantly declined. Self-perceived cognitive function (a QoL measure) correlated with KPS (P = 0.029) and tumor volume (P = 0.045) but not number of BM. The authors concluded that while “patients with large tumor volumes were impaired in some aspects of health-related QoL and neurocognitive functioning, most scores did not differ significantly from those of patients with smaller tumors.”

In a Norwegian study of 97 patients with 1–6 BM treated with SRS, QoL was measured over 12 months using the brain cancer subscale (BRCS) of Functional Assessment of Cancer Therapy–Brain (FACT-Br).⁷⁸ Overall QoL scores did not appreciably change over 12 months after SRS. Pre-SRS/baseline factors associated with improved QoL after SRS included asymptomatic BM (P = 0.001), higher KPS score (P = 0.017), lower RPA class (P = 0.049), lack of seizures (P = 0.040), and absence of cognitive impairment (P = 0.033); number of BM, total volume of BM, tumor location, primary cancer type, and prior WBRT (n = 14 patients) were not significant factors. After SRS, QoL scores were better with local control of BM (P = 0.018), improved or stable symptoms from BM (P = 0.005), lack of steroid treatment (P = 0.003) and absence of extracranial progression (P = 0.001).

This same Norwegian group assessed QoL changes (again with BRCS) in 44 patients (some with prior WBRT) treated with SRS for 1–6 BM from lung cancer.⁷⁹ Net tumor volume was the only factor that significantly predicted the trajectory of QoL after SRS. The BRCS score improved for 32 patients (72.3%) with total BM volume ≤5 cc, while it declined in patients with greater BM burden (P = 0.04). Notably, presence of pre-SRS symptoms, rate of symptom improvement after SRS, and local control rates were similar between patients with net volume ≤5 versus >5 cc. The authors postulated that less post-SRS steroid administration and/or less extracranial progression in patients with smaller intracranial tumor burden might account for differences between post-SRS QoL scores between these 2 groups.

Neurocognition After SRS With versus Without WBRT for Multiple Brain Metastases

Phase III studies (Table 1), which randomized patients with ≤3–4 BM (from any primary site) to receive or not receive WBRT, also analyzed neurocognitive, neurologic, and/or QoL measures and allowed for comparison between WBRT and no WBRT groups (Table 5).

Table 5.

Summary of neurocognitive and functional outcomes in randomized controlled studies of SRS-alone vs. WBRT plus SRS

Study First Author	Impact of use vs omission of whole brain radiotherapy
Aoyama (Japan)	- MMSE: average duration until deterioration 16.5 vs 7.6 M (P = 0.05; longer with WBRT + SRS)
	- Freedom from MMSE decline @1, 2, 3 years: 76.1%, 68.5%, and 14.7% for WBRT+SRS vs 59.3%, 51.9%, and 51.9% for SRS-alone (P = 0.79)
	- KPS: preserved at 1Y in 34% vs 27% (P = 0.53)
	- Neurologic function: preserved in 72 vs 70% (P = 0.99)
Chang (MDACC)	- HVLT–R: drop in total recall (64% vs 20%) at 4 M (96% confidence in difference) worse with WBRT
	- Battery of othercognitive* assessmentsperformed as well
	- QoL (FACT-BR) assessed, though associated with wide confidence intervals
Kocher; Soffietti (EORTC)	- WHO–PS decline (to a score of >2): 9.5 vs 10 M (P = 0.71)
	- QoL – multiple domains significantly worse with WBRT
	• Global health at 9 M (P = 0.015)
	• Physical functioning at 2 M (P = 0.007)
	• Cognitive functioning at 2 and 12 M (P =0.026 and 0.049, respectively)
	• Role functioning at 2 M (P = 0.049)
	• Fatigue at 2 and 3 M (P < 0.001 and 0.044, respectively)
Brown (NCCTG)	- cognitive function (battery of tests†): deterioration at 3 M 92% vs 63% (P < 0.001) worse with WBRT
	- WBRT associated with inferior cognitive outcomes on all of the assessments, significant for:
	• HVLT–R immediate and delayed recall
	• COWA (verbal fluency).
	- QoL (FACT-Br): worse at 3 M (P = 0.002) with WBRT
	(functional well-being index): worse at 3 M (P = 0.03) with WBRT
	- ADLs (Barthel Index of Activities of Daily Living scale) at 3 M: similar
Hong (multicenter)	- ECOG PS: median time to deterioration 5.3 v. 6.0 M (P = 0.32)

Assessments that were used are bolded; significant P-values are also bolded.

*Additional testing included the HVLT-R delayed recall and delayed recognition, Wechsler Adult Intelligence Scale-III (WAIS-III) digit span and digit symbol, Trail Making Test Parts A and B, Multilingual Aphasia Examination Controlled Oral Word Association (COWA), and Lafayette Grooved Pegboard

†Validated testing included the HVLT–R immediate and delayed recall, HVLT–R Recognition, Grooved Pegboard Test, Controlled Oral Word Association Test, and Trail Making Test Parts A and B.

HVLT-R= Hopkins Verbal Learning Test‒Recall (test of recall)

FACT-Br = Functional Assessment of Cancer Therapy–Brain

ADL = Activities of Daily Living

MMSE = Mini-Mental State Examination

KPS = Karnofsky performance scale

ECOG PS = Eastern Cooperative Oncology Group performance scale

WHO PS = World Health Organization performance score

QoL = quality of life

M = months

In the study by Aoyama et al of SRS ± WBRT, Mini-Mental State Examination (MMSE) was an optional endpoint, with 92 of 132 having baseline and follow-up exams.⁴ The average duration until deterioration of MMSE score was longer in the group that received WBRT, attributed to adverse effects on cognition from intracranial recurrence, which was more likely after SRS alone. However, beyond 1–2 years, there was an appreciably greater likelihood of preserving MMSE after SRS alone.

In The MD Anderson Cancer Center’s (MDACC) study of SRS ± WBRT, the primary endpoint was neurocognitive function objectively measured by Hopkins Verbal Learning Test–Revised (HVLT–R) total recall at 4 months.⁵ A battery of other tests (listed in Table 5 footnote) were also analyzed. With interim analysis, after 58 patients were enrolled, the WBRT + SRS group was significantly more likely to have had a drop in HVLT–R total recall; delayed recall and delayed recognition scores were also more compromised with addition of WBRT.

In the European Organisation for Research and Treatment of Cancer (EORTC) study of SRS or resection ± WBRT, the primary endpoint was deterioration of World Health Organization (WHO) performance scale (PS) to a score of >2.⁶ Median time to deterioration was similar in both study groups. Health-related QoL was analyzed as a secondary endpoint, with 88% compliance at baseline and 45% compliance at 1 year. Multiple QoL domains (listed in Table 5 footnote) were significantly worse among the group that received WBRT.⁸⁰

In the North Central Cancer Treatment Group (NCCTG) N0574 study of SRS ± WBRT,⁷ the primary endpoint was cognitive deterioration (>1 SD from baseline) on at least 1 cognitive test (listed in Table 5 footnote) at 3 months. WBRT was associated with a greater risk of cognitive deterioration (92% vs 63%, P < 0.001), inferior cognitive outcomes on all assessments, worse QoL using the FACT-Br scale (P = 0.002) and functional well-being (P = 0.03). In the multicenter NCCTG N107C/CEC.3 phase III trial of postoperative SRS (a scenario not included in the literature search) versus WBRT for 1–4 BM (n = 194 patients), survival rates were similar between the 2 groups, while survival free of cognitive deterioration (from 6 standardized tests) at 6 months was significantly better in the group not receiving WBRT.⁷²

Neurocognition After SRS Alone for >4 Brain Metastases

In the JLGK0901 2017 update, the rate of maintaining MMSE score did not significantly differ (by competing risk analysis) between patients with 1, 2–4, or 5–10 BM.⁸¹ None of the randomized studies discussed above separately analyzed neurocognitive outcomes for patients with multiple versus single BM, though the results may apply equally to both subgroups. As those randomized studies excluded patients with >4 BM, the impact on neurocognitive function of SRS alone versus WBRT ± SRS is of particular interest. Ongoing studies of SRS alone for multiple BM are summarized in Table 6.

Table 6.

Summary of selected ongoing studies analyzing radiosurgery alone in patients with multiple brain metastases

Study Group	Study NCT# / Design	# Patients	Eligibility	Study Arms	Primary Outcomes
MD Anderson	NCT01592968	100	4–15 metastases	- WBRT	Local control
Cancer Center	RCT		all <3.5 cm	- SRS	Cognition (HVLT-R)
MDACC	NCT01644591	49	4+ metastases from melanoma	- SRS	Local control
	Phase II				Cognition
NAGKC 12-01	NCT01731704	closed before accrual	≥5 metastases	- WBRT	Cognition
	RCT		All ≤10 cc	- SRS	QoL
			net GTV ≤15 cc
Netherlands	NCT02353000	260	4–10 metastases	- WBRT	QOL
	RCT		all <2.5 cm	- SRS
			net GTV ≤30 cc
Duke U.	NCT02886572	40	4+ metastases	-SRS	OS
Netherlands	NCT02953717	46	11–20 metastases	- WBRT	Cognition (HVLT-R)
CAR-Study B	RCT		net GTV ≤30 cc	- SRS
Dana Farber	NCT03075072	196	5–20 metastases	- WBRT (with or without hippocampal sparing)	QOL
Cancer Inst.	RCT			- SRS
UT Southwestern	NCT03508752	45		- SRS	Cognition (HVLT-R)
	Phase I/II		6+ metastases all <4 cm
Canadian Cancer	NCT03550391	206	5–15 metastases	- WBRT (hippocampal sparing) memantine	OS
Trials Group/NRG	RCT		all <2.5 cm	+ - SRS	Neuro-cogn. PFS
Sunnybrook Health Sciences Centre	NCT03775330	125	5–20 metastases*	- SRS + WBRT	Cognition (HVLT-R)
	RCT			- SRS

Other Recent Guidelines

Recent systematic reviews and evidence-based guidelines from the International Stereotactic Radiosurgery Society,¹¹ the Congress of Neurologic Surgeons,^82–84 and the American Society for Radiation Oncology Choosing Wisely campaign⁸⁵ advocate against routinely administering WBRT in conjunction with SRS. The 2019 Congress of Neurologic Surgeons recommends local therapy (SRS or surgery) without WBRT for patients with ≤4–15 BM, unless cumulative volume exceeds 7 cc or BM are not amenable to local therapy (based on size or location).^82–84 The American Society for Radiation Oncology model policy for SRS⁸⁶ offers no limits based upon number of BM, and emphasizes necessity of stable extracranial disease, good PS, and “reasonable survival expectations.” National Comprehensive Cancer Network guidelines state that SRS can be considered for patients with “limited” BM for whom SRS is “equally effective and offers significant cognitive protection compared with WBRT”; ⁸⁷ the guidelines describe “limited” as reflecting the number and intracranial volume of BM, though conceding that the relevance of these factors can depend on the clinical situation.

Summary and Conclusions

Based upon a systematic literature review stemming from KQs on the use of SRS for multiple BM, evaluating evidence from systematic review, and voting on appropriateness of select treatment options for specified case variants, the ARS Appropriate Use Criteria group for brain malignancies’ voting determined that:

• • SRS alone is usually appropriate for patients with good PS and 2–10 asymptomatic BM, and usually not appropriate for those with >20 BM. For 11–20 BM there was (between 2 case variants) agreement that SRS alone may be appropriate or disagreement among panelists on appropriateness of SRS alone (Table 4).

• • Conventional WBRT alone is usually not appropriate for those with 2–4 asymptomatic BM and good PS.

While the focus of this review and guidelines report was on the impact of radiotherapy for BM on neurocognition, we recognize the limitation of panelists’ voting being influenced by other outcomes and endpoints, including relative impact on quality of life, local and distant intracranial control, OS, risk of adverse events, and treatment logistics; additionally cost of treatment and a specific region’s payor/government criteria for payment could also have impacted voting.

There were several areas of disagreement that were not addressed by the literature search but were presented as treatment options in the clinical scenarios to inform future ARS guidelines. The disagreement of panel members reflects uncertainty in appropriateness as a result of lack of compelling data. These areas include:

• • Use of hippocampal sparing WBRT for patients with good PS and 2–4 asymptomatic BM (for whom SRS alone would be a treatment option).

• • Use of WBRT after resection of a BM in a patient with good PS whose disease is amenable to SRS.

• • Use of fractionated versus single-fraction SRS for resected BM, larger targets, and/or brainstem BM.

• • Optimal treatment (WBRT, hippocampal sparing WBRT, SRS to all BM, SRS to select BM) for patients with multiple BM, progressive extracranial disease, and poor PS from extracranial disease, for which systemic therapy is not an option.

SRS dose fractionation and the role of SRS with resection of BM are 2 (of several) topics that will be analyzed in the forthcoming Agency for Healthcare Research and Quality (AHRQ) systematic review and evidence report, commissioned by the Patient-Centered Outcomes Research Institute.⁸⁸ These AHRQ analyses will focus on survival and cancer control outcomes, symptom burden (including quality of life and neurocognition), and adverse events. With respect to SRS dose fractionation, all of these endpoints are considered by treating physicians, though any impact of fractionation on neurocognition is unknown.

Neurocognition reflects multiple complex outcome measures, and patients with BM are a highly heterogeneous group with respect to age, baseline function, cancer type, extracranial disease status, size and anatomic distribution of BM, other treatment (ie, resection) for BM, response to treatment, susceptibility to treatment toxicity, and many other factors. The impact of these factors on neurocognition after cranial radiosurgery could not be addressed in this report due to paucity of data. As such, further study is needed to quantify risks of neurocognitive decline, and factors that affect those risks. Patients at risk of developing distant-brain recurrence shortly after SRS alone may benefit from WBRT; though in those patients with uncontrolled extracranial disease with potential to seed new BM, upfront WBRT may be disadvantageous, as its use as salvage therapy would be compromised. Thus, a better understanding of risk and patterns of cancer recurrence, perhaps using tumor and/or host genomics in addition to other clinical and pathologic factors, and a better understanding of factors that affect (in a clinically meaningful way) neurocognition are needed to optimally guide treatment decision making.

Funding

None. The ARS provided administrative support and facilitated teleconferences and live conferences at the annual ARS meeting.

Conflict of interest statement. MTM: Galera Therapeutics (Honorarium); Wolters Kluwer (Royalties). VLSC: BrainLab (Speaker/Honoraria); Monteris Medical Inc (Consultant/Advisory Board); MRInterventions (Consultant/Advisory Board). SGS: Inovio Pharmaceuticals (Consultant), Novocure (Research Support). TJW: AbbVie (Consultant/Advisory Board, Research Support); AstraZeneca (Consultant/Advisory Board); Cancer Panels (Consultant/Advisory Board, Speaker/Honoraria); Doximity (Consultant/Advisory Board, Ownership Interest); Elekta (Consultant/Advisory Board, Speaker/Honoraria); Merck (Research Support); Novocure (Consultant/Advisory Board, Speaker/Honoraria); RTOG Foundation (Research Support); Wolters Kluwer (Royalties). SSL: Elekta AB (Member, Elekta Gamma Knife ICON Expert Group; Research Support). LMH: Abbvie (Research Grant). SC: Varian Medical Systems (Speaker/Honoraria). AS: Advisor/consultant with Abbvie, Merck, Roche, Varian (Medical Advisory Group), Elekta (Gamma Knife Icon), BrainLAB, and VieCure (Medical Advisory Board). Board Member: International Stereotactic Radiosurgery Society (ISRS). Past educational seminars with Elekta AB, Varian (CNS Teaching Faculty), BrainLAB, Medtronic Kyphon. Research grant with Elekta AB. Travel accommodations/expenses by Elekta, Varian, BrainLAB. Belongs to Elekta MR LINAC Research Consortium, Elekta Spine, Oligometastases and LINAC-Based SRS Consortia.

Authorship statement. Experimental design and implementation: all authors (following ARS guidelines). Interpretation of data: all authors. Writing and revising of manuscript: all authors. Approval of final version of manuscript: all authors. List of any unpublished papers cited: none.