Whole brain radiotherapy for the treatment of newly diagnosed multiple brain metastases

Summary of findings 1. Altered WBRT fractionation schedules compared with WBRT control for treatment of newly diagnosed multiple brain metastases

Outcomes	Relative effect (95% CI)	*Anticipated absolute effects (95% CI)**			Certainty (quality) of the evidence (GRADE)	What happens
Altered WBRT fractionation schedules compared with WBRT control for treatment of newly diagnosed multiple brain metastases
Patient or population: patients with newly diagnosed multiple brain metastases Setting: hospital Intervention: altered WBRT fractionation schedules Comparison: WBRT control
Outcomes	Relative effect (95% CI)	Without altered WBRT fractionation schedules	With altered WBRT fractionation schedules	Difference	Certainty (quality) of the evidence (GRADE)	What happens
Overall survival: lower‐dose WBRT vs control WBRT (3000 cGy/10 daily fractions) No. of participants: 705 (3 RCTs)	HR 1.21 (1.04 to 1.40)	Study population			⊕⊕⊕⊝ MODERATE^a
	HR 1.21 (1.04 to 1.40)	Not estimable as only HR data available	Not estimable as only HR data available	N/A	⊕⊕⊕⊝ MODERATE^a
Overall survival: higher‐dose WBRT vs control (3000 cGy/10 daily fractions) No. of participants: 846 (4 RCTs)	HR 0.97 (0.83 to 1.12)	Study population			⊕⊕⊕⊝ MODERATE^a
	HR 0.97 (0.83 to 1.12)	Not estimable as only HR data available	Not estimable as only HR data available	N/A	⊕⊕⊕⊝ MODERATE^a
Overall survival: WBRT 4000 cGy/20 fractions BID vs control WBRT (2000 cGy/4‐5 daily fractions) No. of participants: 203 (2 RCTs)	HR 1.18 (0.89 to 1.56)	Study population			⊕⊕⊕⊝ MODERATE^a
	HR 1.18 (0.89 to 1.56)	Not estimable as only HR data available	Not estimable as only HR data available	N/A	⊕⊕⊕⊝ MODERATE^a
Adverse effects No. of participants: 1754 (9 RCTs)	not pooled	Study population			⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 1754 (9 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊕ HIGH
Symptom control No. of participants: 2877 (7 RCTs)	not pooled	Study population			⊕⊕⊕⊝ MODERATE^b
Symptom control No. of participants: 2877 (7 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^b
Overall brain control No. of participants: 203 (2 RCTs)	not pooled	Study population			⊕⊕⊕⊝ MODERATE^c
Overall brain control No. of participants: 203 (2 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^c
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). BID = twice daily; CI = confidence interval; HR = hazard ratio; N/A = not applicable; OR = odds ratio; RCT = randomised controlled trial; RR = risk ratio; WBRT = whole brain radiation therapy.
GRADE Working Group grades of evidence. High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.
^aIncluded population heterogeneous with respect to survival owing to varying primary cancer diagnoses. Molecular subtypes important for prognoses may not have been balanced between groups. ^bOutcome of symptom control subject to attrition and reporting bias. ^cReporting bias.

Summary of findings 2. Altered WBRT fractionation schedules compared with WBRT control: neurological function improvement for treatment of newly diagnosed multiple brain metastases

Outcomes	Relative effect (95% CI)	*Anticipated absolute effects (95% CI)**			Certainty (quality) of the evidence (GRADE)	What happens
Altered WBRT fractionation schedules compared with WBRT control: neurological function improvement for treatment of newly diagnosed multiple brain metastases
Patient or population: patients with newly diagnosed multiple brain metastases Setting: hospital Intervention: altered WBRT fractionation schedules Comparison: WBRT control: neurological function improvement
Outcomes	Relative effect (95% CI)	Without altered WBRT fractionation schedules	With altered WBRT fractionation schedules	Difference	Certainty (quality) of the evidence (GRADE)	What happens
Neurological function improvement: lower‐dose WBRT vs control dose WBRT (3000 cGy/10 fractions) No. of participants: 1612 (5 RCTs)	OR 1.74 (1.06 to 2.84)	Study population			⊕⊕⊕⊝ MODERATE^a
	OR 1.74 (1.06 to 2.84)	49.6%	63.1% (51.0 to 73.6)	13.5% more (1.5 more to 24.1 more)	⊕⊕⊕⊝ MODERATE^a
Neurological function improvement: higher‐dose WBRT vs control dose WBRT (3000 cGy/10 fractions) No. of participants: 1480 (4 RCTs)	OR 1.14 (0.92 to 1.42)	Study population			⊕⊕⊕⊝ MODERATE^a
	OR 1.14 (0.92 to 1.42)	49.4%	52.7% (47.3 to 58.1)	3.3% more (2.1 fewer to 8.7 more)	⊕⊕⊕⊝ MODERATE^a
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI = confidence interval; OR = odds ratio; RCT = randomised controlled trial; RR = risk ratio; WBRT = whole body radiation therapy.
GRADE Working Group grades of evidence. High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.
^aUnclear whether groups were balanced for steroid use.

Summary of findings 3. WBRT with radiosensitisers (radiosen) compared with WBRT alone for treatment of newly diagnosed multiple brain metastases

Outcomes	Relative effect (95% CI)	*Anticipated absolute effects (95% CI)**			Certainty (quality) of the evidence (GRADE)	What happens
WBRT with radiosensitisers (radiosen) compared with WBRT alone for treatment of newly diagnosed multiple brain metastases
Patient or population: patients with newly diagnosed multiple brain metastases Setting: hospital Intervention: WBRT with radiosensitisers (radiosen) Comparison: WBRT alone
Outcomes	Relative effect (95% CI)	Without WBRT with radiosensitisers (radiosen)	With WBRT with radiosensitisers (radiosen)	Difference	Certainty (quality) of the evidence (GRADE)	What happens
Overall survival No. of participants: 2631 (8 RCTs)	HR 1.05 (0.99 to 1.12)	study population			⊕⊕⊕⊝ MODERATE^a
Overall survival No. of participants: 2631 (8 RCTs)	HR 1.05 (0.99 to 1.12)	not estimable as only HR data available	not estimable as only HR data available	N/A	⊕⊕⊕⊝ MODERATE^a
Brain tumour response rates: complete response (CR) and partial response (PR) combined No. of participants: 847 (6 RCTs)	OR 0.84 (0.63 to 1.11)	study population			⊕⊕⊕⊕ HIGH
	OR 0.84 (0.63 to 1.11)	62.2%	58.1% (50.9 to 64.6)	4.2% fewer (11.3 fewer to 2.4 more)	⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 2631 (8 RCTs)	not pooled	study population			⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 2631 (8 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊕ HIGH
Quality of life No. of participants: 966 (3 RCTs)	not pooled	study population			⊕⊕⊕⊝ MODERATE^b
Quality of life No. of participants: 966 (3 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^b
Symptom control No. of participants: 1814 (3 RCTs)	not pooled	study population			⊕⊕⊕⊝ MODERATE^b
Symptom control No. of participants: 1814 (3 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^b
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI = confidence interval; HR = hazard ratio; N/A = not applicable; OR = odds ratio; RCT = randomised controlled trial; RR = risk ratio; WBRT = whole body radiation therapy.
GRADE Working Group grades of evidence. High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.
^aHeterogeneous population owing to varying primary cancer histologies, and unclear whether groups were balanced for other prognostic variables such as molecular subtype. ^bAttrition and reporting bias.

Summary of findings 4. WBRT and radiosurgery compared with WBRT for treatment of newly diagnosed multiple brain metastases

Outcomes	Relative effect (95% CI)	*Anticipated absolute effects (95% CI)**			Certainty (quality) of the evidence (GRADE)	What happens
WBRT and radiosurgery compared with WBRT for treatment of newly diagnosed multiple brain metastases
Patient or population: patients with newly diagnosed multiple brain metastases Setting: hospital Intervention: WBRT and radiosurgery Comparison: WBRT
Outcomes	Relative effect (95% CI)	Without WBRT and radiosurgery	With WBRT and radiosurgery	Difference	Certainty (quality) of the evidence (GRADE)	What happens
Overall survival No. of participants: 358 (2 RCTs)	HR 0.61 (0.27 to 1.39)	study population			⊕⊕⊕⊝ MODERATE^a
Overall survival No. of participants: 358 (2 RCTs)	HR 0.61 (0.27 to 1.39)	not estimable as only HR data available	not estimable as only HR data available	N/A	⊕⊕⊕⊝ MODERATE^a
1‐Year overall brain control rates No. of participants: 400 (3 RCTs)	HR 0.39 (0.25 to 0.60)	study population			⊕⊕⊕⊕ HIGH
	HR 0.39 (0.25 to 0.60)	not estimable as only HR data available	not estimable as only HR data available	N/A	⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 400 (3 RCTs)	not pooled	study population			⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 400 (3 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊕ HIGH
Intracranial progression‐free duration No. of participants: 400 (3 RCTs)	not pooled	study population			⊕⊕⊕⊕ HIGH
	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊕ HIGH
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI = confidence interval; HR = hazard ratio; OR = odds ratio; RCT = randomised controlled trial; RR = risk ratio; WBRT = whole body radiation therapy.
GRADE Working Group grades of evidence. High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.
^aHeterogeneous population with varying primary cancer types; unclear whether groups were balanced by molecular subtype important for survival.

See: Summary of findings 1 Altered WBRT fractionation schedules compared with WBRT control for treatment of newly diagnosed multiple brain metastases; Summary of findings 2 Altered WBRT fractionation schedules compared with WBRT control: neurological function improvement for treatment of newly diagnosed multiple brain metastases; Summary of findings 3 WBRT with radiosensitisers (radiosen) compared with WBRT alone for treatment of newly diagnosed multiple brain metastases; Summary of findings 4 WBRT and radiosurgery compared with WBRT for treatment of newly diagnosed multiple brain metastases; Summary of findings 5 Radiosurgery alone compared with WBRT and radiosurgery for treatment of newly diagnosed multiple brain metastases

Summary of findings 5. Radiosurgery alone compared with WBRT and radiosurgery for treatment of newly diagnosed multiple brain metastases

Outcomes	Relative effect (95% CI)	*Anticipated absolute effects (95% CI)**			Certainty (quality) of the evidence (GRADE)	What happens
Radiosurgery alone compared with WBRT and radiosurgery for treatment of newly diagnosed multiple brain metastases
Patient or population: patients with newly diagnosed multiple brain metastases Setting: hospital Intervention: radiosurgery alone Comparison: WBRT and radiosurgery
Outcomes	Relative effect (95% CI)	Without radiosurgery alone	With radiosurgery alone	Difference	Certainty (quality) of the evidence (GRADE)	What happens
Overall survival No. of participants: 403 (3 RCTs)	HR 1.00 (0.80 to 1.25)	study population			⊕⊕⊕⊝ MODERATE^a
Overall survival No. of participants: 403 (3 RCTs)	HR 1.00 (0.80 to 1.25)	not estimable as only HR data available	not estimable as only HR data available	N/A	⊕⊕⊕⊝ MODERATE^a
1‐Year radiosurgery‐targeted lesion control No. of participants: 602 (4 RCTs)	HR 2.73 (1.87 to 3.99)	study population			⊕⊕⊕⊕ HIGH
	HR 2.73 (1.87 to 3.99)	not estimable as only HR data available	not estimable as only HR data available	N/A	⊕⊕⊕⊕ HIGH
1‐Year distant brain control No. of participants: 602 (4 RCTs)	HR 2.34 (1.73 to 3.18)	study population			⊕⊕⊕⊕ HIGH
	HR 2.34 (1.73 to 3.18)	not estimable as only HR data available	not estimable as only HR data available	N/A	⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 602 (4 RCTs)	not pooled	study population			⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 602 (4 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊕ HIGH
Neurocognition No. of participants: 403 (3 RCTs)	not pooled	study population			⊕⊕⊕⊝ MODERATE^b
Neurocognition No. of participants: 403 (3 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^b
Neurological function No. of participants: 544 (3 RCTs)	not pooled	study population			⊕⊕⊕⊝ MODERATE^b
Neurological function No. of participants: 544 (3 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^b
Quality of life No. of participants: 271 (2 RCTs)	not pooled	study population			⊕⊕⊕⊝ MODERATE^b
Quality of life No. of participants: 271 (2 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^b
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI = confidence interval; HR = hazard ratio; N/A = not applicable; OR = odds ratio; RCT = randomised controlled trial; RR: risk ratio; WBRT = whole body radiation therapy.
GRADE Working Group grades of evidence. High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.
^aHeterogeneous population with respect to survival. Prognostic variables such as primary cancer histologies and molecular cancer features may not have been balanced between groups. ^bAttrition bias.

Background

This review is an update of a review previously published in the Cochrane Database of Systematic Reviews (Tsao 2012b).

Description of the condition

Brain metastases represent a significant healthcare problem. It is estimated that 20% to 40% of people with cancer will develop metastatic cancer to the brain during the course of their illness (Loeffler 1997). The burden of brain metastases impacts quality and length of survival. Presenting symptoms include headache (49%), focal weakness (30%), mental disturbances (32%), gait ataxia (21%), seizures (18%), speech difficulty (12%), visual disturbance (6%), sensory disturbance (6%), and limb ataxia (6%) (Posner 1995).

Brain metastases may spread from any primary site. The most common primary site is the lung, followed by the breast, then gastrointestinal sites (Walker 1985). Eighty‐five per cent of brain metastases are found in the cerebral hemispheres, 10% to 15% in the cerebellum, and 1% to 3% in the brainstem (Arbit 1995).

Description of the intervention

The mainstay of treatment for brain metastases has been corticosteroids for treatment of peritumoural oedema, antiepileptic medication for treatment of seizures, whole brain radiotherapy (WBRT), and radiosurgery and surgery provided alone or in combination.

How the intervention might work

WBRT has been shown to improve neurological symptoms and function with minimal morbidity (Chao 1954; Posner 1977). Non‐randomised studies suggest that WBRT increases median survival up to three to six months (Chao 1954; Katz 1981; Posner 1977; Zimm 1981). Median survival of participants with symptomatic brain metastases was approximately one month without treatment and two months with corticosteroid use. The overall response rate to WBRT, which is dependent on the symptoms reported, ranged from 64% to 85% (Borgelt 1980a; Katz 1981; Sneed 1996). However, on the whole, studies have vaguely defined symptom response. In one study, 74% of participants showed improvement in neurological symptoms and 65% maintained this improvement for at least nine months (Cairncross 1980).

Why it is important to do this review

Even with treatment, brain metastases cause significant morbidity and mortality. Despite WBRT, up to one‐half of patients will die from intracranial progression (Borgelt 1980a; Chao 1954; Gelber 1981; Katz 1981; Noordijk 1994). As such, review authors evaluated strategies to improve outcomes involving WBRT.

In an attempt to improve outcomes, investigators have examined use of WBRT combined with systemic treatment (cytotoxic chemotherapy), molecular targeted therapy (agents that block growth of cancer cells by inhibiting specific molecules needed for carcinogenesis or tumour growth), or radiosensitisers (agents that enhance the radiation effect) and altered dose‐fractionation schedules.

Investigators have also studied neurocognitive protective strategies (such as drugs) and radiation strategies (such as hippocampal sparing methods) in an attempt to reduce the neurocognitive decline associated with WBRT.

Investigators have combined surgery with WBRT for single metastasis (Grant 2001) and with radiosurgery (a specialised, focused radiation technique) for selected people with brain metastases.

We have excluded from this review studies that examined use of surgery or WBRT, or both, for single brain metastases, as this was the topic of another Cochrane systematic review (Hart 2004).

Objectives

To assess the effectiveness and adverse effects of WBRT given alone or in combination with other therapies to adults with newly diagnosed multiple brain metastases.

Methods

Criteria for considering studies for this review

Types of studies

We included fully published phase 3 trials of adults with newly diagnosed multiple metastases to the brain who were randomised to treatment with WBRT or with other therapies discussed in this review.

Types of participants

We included adult participants (18 years and older) receiving WBRT for newly diagnosed multiple metastases to the brain from any primary cancer.

Types of interventions

Trials that compared the following interventions were eligible for inclusion.

Altered WBRT dose‐fractionation schedules versus conventional WBRT fractionation schedules (*3000 cGy in 10 fractions, or 2000 cGy in 4 or 5 daily fractions).
WBRT* and systemic therapy.
WBRT plus radiosensitisers versus WBRT*.
WBRT plus radiosurgery versus WBRT*.
Radiosurgery alone versus radiosurgery and WBRT*.
Steroids alone versus WBRT and steroids*.

We added three new categories to this updated review.

WBRT* and molecular targeted agents.
WBRT* and neurocognitive protective agents.
Hippocampal sparing WBRT versus WBRT*.

*Designated as a control arm in the trial.

We excluded trials of prophylactic WBRT in which WBRT was used without evidence of existing brain metastases. We also excluded studies that examined use of surgery or WBRT, or both, for single brain metastases, as this is the topic of another Cochrane systematic review (Hart 2004). We excluded intervention trials for people with recurrent brain metastases, as well as phase II randomised trials.

Types of outcome measures

We sought data for the following outcome measures.

Overall survival.
Intracranial progression‐free duration (defined as time from randomisation or entry into the trial until diagnosis of progressive brain disease (i.e. enlarging brain metastases or identification of new brain metastases based on contrast‐enhanced computed tomography (CT) or magnetic resonance imaging (MRI) scan)).
Brain response (reported as the percentage of participants achieving complete response (CR) or partial response (PR) of existing brain metastases to treatment). Complete response was defined as complete radiographic disappearance of brain metastases. Partial response was defined as a greater than 50% decrease in the size of brain metastases on CT or MRI.
Local brain control (reported as the percentage of participants with unchanged or improved serial post‐treatment CT or MRI scans judged as showing CR, PR, or stable disease (SD), with improving or stable neurological symptoms or neurological examination. SD was defined as a 0% to 50% decrease in the size of all lesions, with stabilisation of neurological symptoms or neurological examination and stable dexamethasone dose. Progressive disease was defined as an increase in the size of any lesion.
Distant brain control (defined as CR, PR, or SD of brain metastases not treated with focal therapy such as radiosurgery).
Quality of life assessed by any scale.
Symptom control.
Neurological function.
Proportion of participants who were able to reduce their daily dexamethasone dose and duration of reduced dexamethasone requirements.
Adverse effects.

Search methods for identification of studies

Electronic searches

Review update

We updated the searches in May 2017 ‐ Cochrane Central Register of Controlled Trials, CENTRAL Issue 4 May 2017, MEDLINE (Ovid) April week 4 2017, and Embase (Ovid) 2017 week 19. Refer to Appendix 1, Appendix 2, and Appendix 3, respectively.

Searching other resources

We handsearched the references of included studies to identify additional studies. We searched the National Cancer Institute Physicians Data Query (PDQ) for ongoing trials.

Data collection and analysis

Selection of studies

Two radiation oncologists (MNT and EC) assessed titles and abstracts retrieved by the search strategy. We obtained full published reports for all references deemed to meet the inclusion criteria. We retrieved articles if we believed that we should review article reference lists to look for additional relevant studies.

Assessment of full reports enabled identification of studies for inclusion in the review. We listed in the Characteristics of excluded studies table trials excluded at this stage along with reasons for exclusion.

Assessors were not blinded to author, institution, journal of publication, or trial results, as the review authors were familiar with most studies and with the typographical layout of journals.

Data extraction and management

Two review authors (MNT and EC) independently extracted data using standard data extraction forms. A third assessor (RW) resolved disagreements or discrepancies.

We extracted the following data items.

Study characteristics.
Participant characteristics.
Interventions.
Outcome data (Types of outcome measures).

We also extracted from trial reports participant characteristics based on age, Karnofsky Performance Status (KPS), and status of extracranial disease, along with treatment characteristics.

Assessment of risk of bias in included studies

Since the last update was published, we assessed risk of bias in included studies using the Cochrane tool for assessing bias (Higgins 2011). As none of the radiation trials were blinded, we excluded domains related to blinding and used the following domains.

Selection bias: random sequence generation and allocation concealment.
Attrition bias: incomplete outcome data.
Reporting bias: selective reporting of outcomes.

We summarised assessment results using the 'Risk of bias' graph (Figure 1) and the 'Risk of bias' summary (Figure 2).

Figure 1

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Since the last update, we have presented the overall certainty of evidence for each outcome in the meta‐analysis according to the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach, which takes into account issues related not only to internal validity (risk of bias, inconsistency, imprecision, publication bias) but also to external validity, such as directness of results. We used GRADEpro GDT (GRADEpro GDT 2015) to prepare 'Summary of findings' tables, according to the methods described in the Cochrane Handbook for Systematic Reviewsof Interventions (Higgins 2011).

Dealing with missing data

We abstracted study withdrawals, deaths, and losses to follow‐up. We performed all analyses on an intention‐to‐treat basis.

Assessment of heterogeneity

The Radiation Therapy Oncology Group (RTOG) analysed the database of 1200 participants from three consecutive RTOG trials that examined several dose‐fractionation schemes of WBRT and use of radiosensitisers with WBRT (Gaspar 1997). Using recursive partitioning analysis (RPA), we generated three prognostic classes based on survival: Class 1: KPS ≥ 70, < 65 years of age with controlled primary and no extracranial metastases; Class 3: KPS < 70; and Class 2: all others. These classes enable the classification of brain metastasis populations separated into homogeneous patient groups on the basis of survival.

Since the last Cochrane update, researchers have refined the prognostic classification of brain metastases (the Diagnosis‐Specific Graded Prognostic Assessment (DS‐GPA)) (Sperduto 2012). Many trials pre‐dated the DS‐GPA publication. In addition, many trials pre‐dated the recognition of molecular cancer subtypes important for prognoses. As such, it is unclear whether study groups in these older trials were balanced for prognostic variables based on the DS‐GPA, which also takes into account the molecular profile of certain cancers.

Patient characteristics predicting other outcomes of interest (intracranial progression‐free survival, quality of life, symptom control, neurological function, ability to taper down on dexamethasone dose) have not been defined.

Data synthesis

If pooling of data was deemed appropriate, we used the statistical package RevMan 5.3 (Review Manager 2014). In this updated review, we reported hazard ratios (HRs) with 95% confidence intervals (CIs) for overall survival, one‐year overall brain control, one‐year radiosurgery‐targeted lesion control, and one‐year distant brain control. We reported odds ratios (ORs) with 95% CIs using the random‐effects model for neurological function improvement (NFI) and brain tumour response.

We extracted data from full published trials. Data from abstracts as these reports lacked the detail to permit pooling of outcomes. We did not contact authors of abstracts to request further information.

This updated review includes the following statistical analyses.

For the pooled analysis of overall survival, we estimated one‐year overall brain control, one‐year radiosurgery‐targeted lesion control, and one‐year distant brain control by using hazard ratios (HRs) derived from the Hazard Ratio Meta‐analysis Box (Parmar 1998). We pooled HRs using the generic inverse variance method and the fixed‐effect model provided in RevMan 5.3
For the pooled analysis of brain tumour response, we abstracted from tables, figures, or text of published reports the proportion of participants with a complete or partial response. We determined tumour response using the proportion of participants achieving complete response (CR) or partial response (PR). We treated these data as dichotomous outcome measures. We pooled odds ratios (ORs) for brain tumour response using a random‐effects model.
We described neurological function improvement (NFI) as the proportion of participants with improved neurological function and treated this as a dichotomous outcome. We pooled ORs for NFI using random effects.
Owing to the heterogeneity of instruments used and differences in reporting, we described and did not pool quality of life, symptom control, and adverse effects outcomes.

Subgroup analysis and investigation of heterogeneity

We planned no subgroup analyses.

Sensitivity analysis

We performed no sensitivity analyses.

Results

Description of studies

Results of the search

We added to this updated review 10 new fully published randomised controlled trials (RCTs) involving 2457 participants (Brown 2013; Brown 2016; El Gantery 2014; El‐Hamamsy 2016; Fogarty 2015; Mehta 2009; Mulvenna 2016; Rapp 2015; Sperduto 2013; Zeng 2016).This updated review now includes a total of 45 fully published trials involving 11,152 participants (see Included studies). We identified four additional trials consisting of 746 participants that were published in abstract form (Antonadou 2003; Puthiyottil 2016; Ramlau 2013; Sturm 2015). Five additional publications described secondary outcomes, prognostic factors, or subgroup analyses for a subset of 1385 trial participants (Figure 3) (Aoyama 2015; Meyers 2004; Scott 2007; Sperduto 2014; Stea 2006).

Figure 3

Study flow diagram.

Included studies

Altered WBRT schedules

We identified a total of 10 published reports involved 4056 participants randomised to altered WBRT dose‐fractionation schedules compared with standard 3000 cGy in 10 daily fractions, or 2000 cGy in 4 or 5 daily fractions (Borgelt 1980a; Borgelt 1981a; Chatani 1985; Chatani 1994; Davey 2008; Graham 2010; Harwood 1977; Kurtz 1981; Murray 1997; Priestman 1996).

The published reports Borgelt 1980a and Borgelt 1980b presented the results of two sequential trials (Study 1: Borgelt 1980a; and Study 2: Borgelt 1980b). Investigators randomised participants to one of five WBRT schedules ranging from 4000 cGy/4 weeks to 2000 cGy/1 week. Study 1 randomised participants to one of four regimens (3000 cGy/2 weeks, 3000 cGy/3 weeks, 4000 cGy/3 weeks, or 4000 cGy/4 weeks) (Borgelt 1980a). Study 2 randomised participants to one of three regimens (2000 cGy/1 week, 3000 cGy/2 weeks, or 4000 cGy/3 weeks) (Borgelt 1980b).

Borgelt 1981a reported on participants randomised to 1000 cGy/1 fraction in Study 1, and Borgelt 1981b reported on participants randomised to 1200 cGy/2 fractions in Study 2, as compared with participants treated on one of five schedules ranging from 4000 cGy/4 weeks to 2000 cGy/1 week. Thus, duplicate reports describe participants treated on one of five WBRT schedules ranging from 4000 cGy/4 weeks to 2000 cGy/1 week (Borgelt 1980a; Borgelt 1980b; Borgelt 1981a; Borgelt 1981b). It was not possible to separate the subgroup of participants duplicated in the comparison group treated on one of five schedules ranging from 4000 cGy/4 weeks to 2000 cGy/1 week. As such, we entered data from these reports to show completely separate groups of participants.

Chatani 1985 and Chatani 1994 reported on separate groups of participants who were enrolled sequentially into these trials. The other trials reported no duplication of participant data (Harwood 1977; Kurtz 1981; Murray 1997; Priestman 1996).

Regine 2001 subsequently reported neurocognitive outcomes in trial participants from Murray 1997.

Two RCTs examined the use of 4000 cGy in 20 twice‐daily fractions of WBRT versus 2000 cGy in 4 or 5 daily fractions of WBRT (Davey 2008; Graham 2010).

WBRT with or without radiosensitisers

Radiosensitisers are drugs that make cancer cells more sensitive to the treatment effects of radiation therapy. A total of nine fully published trials examined use of radiosensitisers in addition to WBRT (2712 participants) (DeAngelis 1989; El‐Hamamsy 2016; Eyre 1984; Komarnicky 1991; Mehta 2003; Mehta 2009; Phillips 1995; Suh 2006; Zeng 2016). Mehta 2009, El‐Hamamsy 2016, and Zeng 2016 are new to this update. Suh 2008, which included 368 participants, was published in abstract form.

The radiosensitisers used were metronidazole (Eyre 1984), lonidamine (DeAngelis 1989), misonidazole (Komarnicky 1991), bromodeoxyuridine (BrdU) (Phillips 1995), motexafin gadolinium (Mehta 2003; Mehta 2009), efaproxiral (Suh 2006). simvastatin (El‐Hamamsy 2016), and sodium glycididazole (Zeng 2016).

Suh 2006 reported on survival as the primary endpoint with response rate as the secondary endpoint among participants randomised to WBRT and efaproxiral versus WBRT alone. Stea 2006 explored prognostic significance based on radiographic response among participants in Suh 2006.

A published post hoc subgroup analysis examined participants with breast cancer treated with the radiosensitiser efaproxiral and WBRT versus WBRT (Scott 2007). Subsequently, Suh 2008 reported in abstract form a trial that included a priori only participants with breast cancer randomised to WBRT plus efaproxiral versus WBRT alone.

Mehta 2003 reported on survival and neurological outcomes in participants with a variety of primary cancers metastatic to brain who were randomised to WBRT and motexafin gadolinium as compared with WBRT alone. In a follow‐up report, Meyers 2004 reported specifically on neurocognitive outcomes among the same group of participants randomised in the motexafin gadolinium trial (Mehta 2003).

Subsequently, Mehta 2009 reported on a new group of participants with non‐small‐cell lung cancer brain metastases randomised to WBRT and motexafin gadolinium as compared with WBRT alone.

WBRT and chemotherapy

A total of nine fully published trials (1130 participants) reported on the use of WBRT and chemotherapy (Guerrieri 2004; Knisely 2008; Lee 2008; Mornex 2003; Neuhaus 2009; Postmus 2000; Robinet 2001; Sperduto 2013; Ushio 1991).

Four trials were reported in abstract form (Antonadou 2002; Antonadou 2003; Puthiyottil 2016; Ramlau 2013).

The Antonadou 2002 trial randomised participants to WBRT with or without temozolomide chemotherapy. Eighty‐two per cent of 134 eligible participants in Antonadou 2002 had lung primaries. These participants with lung cancer brain metastases were reported subsequently in abstract form (Antonadou 2003).

The Ramlau 2013 abstract reported on participants with non‐small‐cell lung cancer with brain metastases randomised to WBRT with or without oral topotecan.

Puthiyottil 2016 reported, in abstract form, on participants with non‐small‐cell lung cancer brain metastases who were randomised to WBRT with or without temozolomide chemotherapy.

The remaining fully published randomised trials studied a variety of other chemotherapy agents.

Guerrieri 2004 randomised participants with non‐small‐cell lung cancer metastatic to brain to WBRT with or without carboplatin chemotherapy.
Knisely 2008 examined use of WBRT with or without thalidomide in participants with brain metastases from a variety of primary cancers.
Lee 2008 randomised participants with metastatic non‐small‐cell lung cancer with brain metastases to WBRT with or without gemcitabine and vinorelbine chemotherapy.
Mornex 2003 randomised participants with metastatic malignant melanoma to brain to fotemustine and WBRT versus fotemustine alone.
Neuhaus 2009 randomised participants with brain metastases from non‐small‐cell and small‐cell lung cancer to WBRT with or without topotecan chemotherapy.
Postmus 2000 examined participants with metastatic small‐cell lung cancer to brain and randomised these participants to teniposide versus teniposide and WBRT.
Robinet 2001 reported on participants with non‐small‐cell lung cancer metastatic to brain randomised to early versus delayed WBRT with concurrent cisplatin and vinorelbine chemotherapy.
Sperduto 2013 reported on participants with metastatic non‐small‐cell lung cancer to brain randomised to one of three groups: WBRT and radiosurgery; WBRT, radiosurgery, and temozolomide chemotherapy; or WBRT, radiosurgery, and the targeted agent, erlotinib.
Ushio 1991 randomised participants with metastatic lung cancer to brain to one of three groups: WBRT alone; WBRT and chloroethyl nitrosoureas; or WBRT, chloroethyl nitrosoureas, and tegafur.

WBRT with or without radiosurgery boost

Three trials examined use of WBRT with or without radiosurgery boost for up to four brain metastases (464 participants in total) (Andrews 2004; Chougule 2000; Kondziolka 1999). Two trials have been fully published (Andrews 2004; Kondziolka 1999). Kondziolka 1999 included participants with two to four brain metastases (all ≤ 25 mm in diameter). Andrews 2004 reported on participants with one to three brain metastases, with a maximum diameter of 4 cm for the largest lesion and additional lesions not exceeding 3 cm in diameter. Chougule 2000 has been published in abstract form. This trial included participants with one to three brain metastases, tumour volume of 30 cc or less, and minimum life expectancy of three months.

We added to this update one new fully published trial (El Gantery 2014), which examined use of WBRT alone versus WBRT and radiosurgery versus radiosurgery alone. This trial included 60 participants, each with one to three brain metastases.

Since the last published Cochrane Review, Sperduto PW 2014 has reported a secondary analysis of the Andrews 2004 trial, stratifying overall survival by the Diagnosis‐Specific Graded Prognostic Assessment (DS‐GPA) (Sperduto 2012),

Radiosurgery alone versus radiosurgery and WBRT

Aoyama 2006 randomised 160 participants, each with one to four brain metastases (≤ 3 cm) to radiosurgery alone or radiosurgery and WBRT. A secondary analysis of Aoyama 2006 trial data explored overall survival and control rates of brain metastases based on the DS‐GPA (determined by significant prognostic factors such as primary cancer type, performance status, age, presence of extracranial disease, and number of brain metastases) (Aoyama 2015).

Chang 2009 randomised 58 participants, each with one to three brain metastases, to radiosurgery alone versus WBRT and radiosurgery.

Kocher 2011 randomised 359 participants, each with one to three brain metastases, to one of the following groups: radiosurgery alone, surgery alone, radiosurgery and WBRT, or surgery and WBRT.

Since the last update, three additional RCTs have been fully published (Brown 2016; El Gantery 2014; Fogarty 2015).

El Gantery 2014 randomised 60 participants, each with one to three brain metastases, to radiosurgery alone versus WBRT versus both.
Fogarty 2015 reported on an interim analysis of 100 metastatic melanoma participants, each with one to three brain metastases treated with radiosurgery or surgery, who were then randomised to adjuvant WBRT or observation.
Brown 2016 randomised 213 participants, each with one to three brain metastases, to radiosurgery alone versus radiosurgery and WBRT.

Steroids with or without WBRT

One trial (48 participants) reported on use of oral prednisone with or without WBRT (Horton 1971).

Since the last update, one RCT (538 participants) has reported on use of WBRT and optimal supportive care versus optimal supportive care alone in participants with multiple brain metastases from non‐small‐cell lung cancer (Mulvenna 2016).

We added the following three categories since the last update.

WBRT and molecular targeted agents

WBRT with or without neurocognitive sparing agents

Brown 2013 randomised participants with brain metastases to memantine versus placebo. Memantine is a potentially neurocognitive protective agent.

Rapp 2015 randomised participants with brain metastases and primary brain tumour who had undergone partial brain radiotherapy or WBRT to donepezil (also a potential neurocognitive protective agent) versus placebo.

Hippocampal sparing WBRT versus WBRT

No fully published phase III randomised trials have examined use of hippocampal sparing WBRT versus WBRT. However, phase 3 trials are ongoing (NRG‐CC001; Sturm 2015). It is hypothesised that minimising radiation dose to the hippocampus might be associated with less cognitive decline after WBRT.

Excluded studies

We excluded one study because the trial design did not include a standard WBRT dose‐fractionation group (3000 cGy in 10 fractions, or 2000 cGy in 5 fractions) (Haie‐Meder 1993).

We excluded another trial as the radiotherapy group used three‐dimensional conformal radiation therapy (3D CRT) rather than WBRT (Wang 2015).

Risk of bias in included studies

See Figure 1 and Figure 2.

Allocation

We assessed the method used to generate the allocation sequence as conferring low risk of bias when investigators used any truly random process, and when treatment allocation was protected before and until assignment.

Incomplete outcome data

We defined risk as low when less than 10% of participants did not complete the outcome assessment. Not all studies described the percentage of missing data in sufficient detail to allow a judgement (classified as unclear risk).

Selective reporting

We deemed overall survival as not subject to reporting bias. We determined that other outcomes reported in included trials such as progression‐free survival, quality of life, and adverse effects may have been subject to possible selective outcomes reporting bias.

Effects of interventions

1. Altered WBRT dose‐fractionation schedules versus conventional WBRT fractionation schedule (control: 3000 cGy in 10 daily fractions, or 2000 cGy in 4 or 5 daily fractions)

Ten reports examining the effectiveness of different dose‐fractionation schedules of WBRT provided data for statistical analysis (Borgelt 1980a; Borgelt 1981a; Chatani 1985; Chatani 1994; Davey 2008; Graham 2010; Harwood 1977; Kurtz 1981; Murray 1997; Priestman 1996).

Regine 2001 reported on neurocognitive outcomes in participants with brain metastases treated in the Murray 1997 trial of accelerated fractionation versus accelerated hyperfractionated WBRT.

Dose response was the primary outcome for this comparison. To evaluate dose response, investigators compared many different dose‐fractionation schedules. The most commonly employed 'control' regimen was 3000 cGy in 10 daily fractions. Investigators used the concept of 'biological equivalent dose' (BED) to facilitate comparison between different dose‐fractionation regimens.

BED can be calculated using the equation BED = nd [1+d/(alpha/beta)], where n = number of fractions, d = dose per fraction, and alpha/beta = 10 for tumour (Hall 2000). For the purpose of assessing dose response, review authors divided studies into those comparing lower biological doses versus 3000 cGy in 10 daily fractions and those comparing higher biological doses versus 3000 cGy in 10 daily fractions. To explore whether a dose response relationship was present, we used relative biological effectiveness (RBE) = 39 Gy for the fractionation scheme of 3000 cGy in 10 daily fractions as a control. We compared outcomes between RBE < 39 Gy versus 39 Gy and 39 Gy versus > 39 Gy.

With regard to the outcomes of interest, none of the included trials reported on:

intracranial progression‐free duration;
local brain control (CR + PR + SD);
local brain tumour response (CR + PR);
quality of life; or
the proportion of participants able to reduce their dexamethasone dose.

Overall survival

Data for this outcome were available from eight trials (Chatani 1985; Chatani 1994; Davey 2008; Graham 2010; Harwood 1977; Kurtz 1981; Murray 1997; Priestman 1996).

Three trials compared lower‐dose radiation with RBE < 39 Gy (2000 cGy in 5 fractions, 1000 cGy in a single fraction, or 1200 cGy in 2 fractions) versus standard‐dose WBRT with RBE = 39 Gy (3000 cGy in 10 fractions) (Chatani 1994; Harwood 1977; Priestman 1996). Combining these data in a meta‐analysis revealed an HR of 1.21 (95% CI 1.04 to 1.40; P = 0.01, favouring 3000 cGy in 10 fractions; moderate‐certainty evidence) (Analysis 1.1; summary of findings Table 1).

Of note, Chatani 1994 reported median and one‐year survival as 3.4 months versus 2.4 months and 6% versus 4% (P = 0.943) for participants treated with 3000 cGy in 10 daily fractions versus 2000 cGy in 5 daily fractions of WBRT, respectively.

Four trials compared higher‐dose WBRT with RBE > 39 Gy (5000 cGy in 20 fractions, or 5440 cGy in 34 fractions twice daily) versus standard‐dose WBRT with RBE = 39 Gy (3000 cGy in 10 fractions) (Chatani 1985; Chatani 1994; Kurtz 1981; Murray 1997). Pooling of data from these four trials revealed an HR for survival of 0.97 (95% CI 0.83 to 1.12; P = 0.65; moderate‐certainty evidence) (Analysis 1.2; summary of findings Table 1).

We could not obtain overall survival data from published reports of Borgelt 1980a, Borgelt 1980b, Borgelt 1981a, and Borgelt 1981b.

Two published RCTs examined participants treated with WBRT 4000 cGy in 20 fractions twice daily (BID) versus standard WBRT 2000 cGy in 4 or 5 daily fractions. Davey 2008 randomised participants to 4000 cGy in 20 fractions BID WBRT versus 2000 cGy in 5 daily fractions WBRT. Graham 2010 randomised participants to 4000 cGy in 20 fractions BID WBRT versus 2000 cGy in 4 daily fractions WBRT. Pooling of data from these two trials revealed an HR for survival of 1.18 (95% CI 0.89 to 1.56; P = 0.25; moderate‐certainty evidence) (Analysis 1.3; summary of findings Table 1).

Symptom control

Seven published reports assessed symptom control (Borgelt 1980a; Borgelt 1981a; Chatani 1985; Chatani 1994; Harwood 1977; Kurtz 1981; Priestman 1996). However, investigators used a variety of scales (e.g. neurological symptom relief, palliative index, performance status), making statistical pooling of data not meaningful. Considered individually, none of these trials showed a difference in symptom control with altered dose‐fractionation schedules as compared with control (3000 cGy in 10 fractions).

Overall brain control

A post hoc subset analysis from the Davey 2008 trial showed that time to retreatment for intracranial relapse was 14 weeks in the control WBRT group (2000 cGy in 5 daily fractions) as compared with 32 weeks in the 4000 cGy in 20 BID (twice‐daily) fractions group (P = 0.03). The primary endpoint for Graham 2010 was the proportion of participants with overall brain progression. Brain progression was worse for participants treated with 2000 cGy in 4 fractions as compared with those given 4000 cGy in 20 BID fractions (64% vs 44%; P = 0.03, respectively).

Neurological function

We pooled neurological function outcomes from seven reports (Borgelt 1980a; Borgelt 1981a; Chatani 1985; Chatani 1994; Harwood 1977; Kurtz 1981; Priestman 1996). Neurological function grading scales utilised were similar across trials and typically ranged from one (minimal interference) to four (comatose or requiring constant nursing care).

Three trials reported NFI only for participants with a baseline neurological function of grade 2 or 3 (Borgelt 1980a; Borgelt 1981a; Kurtz 1981). Within this limitation, the OR was 1.14 (95% CI 0.92 to 1.42; P = 0.23; moderate‐certainty evidence) for those treated with a biologically higher dose versus those given the control dose (Analysis 2.2).

The OR was 1.74 (95% CI 1.06 to 2.84; P = 0.027; moderate‐certainty evidence) for NFI, favouring the control WBRT dose of 3000 cGy in 10 daily fractions as compared with a lower dose (Analysis 2.1; summary of findings Table 2). The duration of any improvement was not consistently reported.

Regine 2001 reported on neurocognitive outcomes in participants with brain metastases treated in the Murray 1997 trial of accelerated fractionation versus accelerated hyperfractionated WBRT. Mini‐Mental State Examination (MMSE) scores showed that neurocognitive function outcomes between the two treatment groups of the trial were similar. However, at three months, participants with uncontrolled brain metastases showed a significant drop in MMSE scores as compared with participants with radiographically controlled brain metastases.

Adverse effects

Included trials inconsistently reported adverse effects among participants in terms of incidence and grade. Borgelt 1981a described no differences in treatment morbidity (defined as worsening of neurological symptoms or the appearance of new symptoms) among treatment groups.

Chatani 1994 reported that no participant experienced grade 3 or higher acute toxicity. Chatani 1985 did not report adverse effects.

In the Harwood 1977 study, 40% (of 51 participants) treated with a single dose of WBRT developed acute complications (increased headache, nausea and vomiting, neurological deficit; or a fall in the level of consciousness) compared with 27% (of 50 participants) treated with the fractionated course. This difference did not reach statistical significance (P = 0.254).

Murray's study of accelerated hyperfractionation versus 3000 cGy in 10 fractions over 10 days resulted in one grade 4 ototoxicity and one grade 5 toxicity (death) due to cerebral oedema in the accelerated hyperfractionation group (N = 216 participants) (Murray 1997). Data show no differences in the incidence of acute grade 3 toxicity nor in the incidence of grade 3 or 4 late toxicity.

In Priestman 1996, 22 (of 274) participants treated with 1200 cGy in 2 fractions of WBRT developed drowsiness or lethargy, headache, nausea or vomiting, dizziness or ataxia, cerebral haemorrhage, blurred vision, or fits, as compared with 13 (of 270) participants treated with 3000 cGy in 10 fractions.

Davey 2008 reported similar central nervous system late effects of normal tissues (LENT)/subjective, objective, management, analytical (SOMA) scores between the two groups.

In the Graham 2010 trial, late significant toxicity was uncommon. Grade 2 or higher toxicity occurred in one participant in each group.

2. WBRT plus radiosensitisers versus WBRT

We did not pool results from Scott 2007, as this report described a post hoc subset analysis of participants with breast cancer randomised to WBRT and efaproxiral versus WBRT alone (Suh 2006). A larger confirmatory trial, which included participants with breast cancer only (a priori), was subsequently published in abstract form (Suh 2008). We could not pool results from Suh 2008 as the abstract provided insufficient details.

Mehta 2003 reported on 401 participants with a variety of primary cancers metastatic to brain randomised to WBRT with or without motexafin gadolinium.

Mehta 2009 reported on 554 participants with non‐small‐cell lung cancer as primary cancer metastatic to brain randomised between December 2002 and March 2005 to WBRT with or without motexafin gadolinium.

These publications did not describe whether duplicate participants may have been reported by Mehta 2003 and Mehta 2009.

Nine fully published RCTs examined use of radiosensitisers in addition to WBRT (DeAngelis 1989; El‐Hamamsy 2016; Eyre 1984; Komarnicky 1991; Mehta 2003; Mehta 2009; Phillips 1995; Suh 2006; Zeng 2016).

With regard to the following outcomes of interest, none of the included trials reported on:

neurological function; or
proportions of participants who were able to reduce their daily dexamethasone dose and the duration of reduced dexamethasone requirements.

Overall survival

We pooled data on survival from eight trials (El‐Hamamsy 2016; Eyre 1984; Komarnicky 1991; Mehta 2003; Mehta 2009; Phillips 1995; Suh 2006; Zeng 2016). When these data were combined, the HR was 1.05 (95% CI 0.99 to 1.12; P = 0.12; moderate‐certainty evidence) for overall survival (summary of findings Table 3).

DeAngelis 1989 examined use of WBRT with or without lonidamine. Median survival was 165 days for the WBRT group versus 120 days for the lonidamine group (P = 0.42). Details were insufficient to generate a hazard ratio for survival from this study, and as such we could not pool survival data.

Local brain tumour response

Six trials reported on local brain tumour response rates ‐ complete response (CR) or partial response (PR) (DeAngelis 1989; El‐Hamamsy 2016; Eyre 1984; Phillips 1995; Suh 2006; Zeng 2016). The OR was 0.84 (95% CI 0.63 to 1.11; P = 0.22; high‐certainty evidence) for response rate between participants receiving only WBRT and those receiving treatment with WBRT and radiosensitisers (Analysis 3.2; summary of findings Table 3).

Quality of life

Mehta 2003 reported quality of life outcomes for 401 participants with lung cancer, breast cancer, melanoma, and other cancers randomised to motexafin gadolinium and WBRT or WBRT alone. Data show similar time to progression for brain‐specific quality of life assessment (Functional Assessment of Cancer Therapy‐Brain (FACT‐BR)) in both treatment groups.

Mehta 2009 did not report quality of life outcomes using validated quality of life instruments.

Suh 2006 reported on Spitzer Quality of Life Index and KPS outcomes for participants randomised to efaproxiral and WBRT versus WBRT alone. A higher proportion of participants in the efaproxiral group had stable or improving quality of life scores and KPS ratings (OR 1.21 and 1.38, respectively; P = 0.008) as compared with participants treated with WBRT alone.

El‐Hamamsy 2016 reported on quality of life using European Organization for Research and Treatment of Cancer core quality of life questionnaires (EORTC QLQ‐C30). However, the number of participants (N = 29) at four weeks was too small to allow any meaningful comparisons.

Symptom control

For participants receiving WBRT with or without the radiosensitiser misonidazole (Komarnicky 1991), data show no differences between treatment groups in the percentage of total survival time spent in an improved or stable KPS, nor in median time to deterioration of KPS. The percentage of participants who spent 90% to 100% of their survival time in an improved or stable neurological state was also similar among treatment and control groups.

Meyers 2004 reported on participants in the Mehta 2003 trial of WBRT with or without motexafin gadolinium. The HR for all eight neurocognitive tests was 0.66 (P = 0.114). Participants with lung cancer (but not other types of cancer) who were treated with motexafin gadolinium in addition to WBRT tended to have improved memory and executive function (P = 0.062) along with improved neurological function as assessed by a blinded events review committee (P = 0.048).

In Mehta 2009, the interval to neurological progression (based on symptoms and radiographic findings) for participants treated with WBRT and motexafin gadolinium as compared with those treated with WBRT alone was 15 months versus 10 months (HR 0.78; P = 0.12). We concluded that motexafin gadolinium did not confer an overall advantage in survival or time to neurological progression for the entire cohort. Subgroup analyses suggest that participants with lung cancer may benefit (in the light of improved neurological function, memory, and executive function). A randomised trial examining this specific subgroup has been published (Mehta 2009).

Adverse effects

All nine fully published trials that assessed the addition of radiosensitisers to WBRT reported serious adverse effects.

Participants who received WBRT and metronidazole in the Eyre 1984 study reported a 51% (of 57 participants) incidence of nausea and vomiting compared with 3.2% (of 54 participants) in the WBRT‐alone group.

DeAngelis 1989 found that the most common side effects of lonidamine and WBRT were myalgia (68%, or 21/31), testicular pain (42%, or 8/19 of men), anorexia (26%, or 8/31), ototoxicity (26%, or 8/31), malaise or fatigue (26%, or 8/31), and nausea and vomiting (19%, or 6/31). Investigators noted no acute or subacute radiation‐related neurotoxicity in either treatment group.

Komarnicky 1991 reported that misonidazole administration with WBRT was well tolerated and produced no grade 3 neurotoxicity or ototoxicity. However, investigators noted several (number not reported) grade 3 symptoms of nausea and vomiting (defined as occurring one to three times daily).

Phillips 1995 reported three fatal toxicities in 34 participants randomised to WBRT with administration of the radiosensitiser BrdU. One death resulted from a severe Stevens‐Johnson skin reaction (a rare and severe adverse reaction to sulphonamides involving skin and mucous membranes of the eyes, mouth, nose, and genitals with ulceration and loss of epithelium) and two other deaths were due to neutropenia and infection. Trial authors noted no increase in rate of radiation skin reactions or central nervous system injuries in the BrdU group of this study.

Mehta 2003 reported grade 3 and 4 adverse events ‐ hypotension (5.8%), asthaenia (2.6%), hyponatraemia (2.1%), leukopaenia (2.1%), hyperglycaemia (1.6%), and vomiting (1.6%) ‐ among 193 participants randomised to the WBRT and motexafin gadolinium group.

In the Suh 2006 trial, 28% of participants in the efaproxiral group experienced grade 3 or 4 treatment‐emergent adverse events. The most common efaproxiral severe adverse effect was hypoxaemia (11%; 29 of 266 participants).

Mehta 2009 reported that the most frequent motexafin gadolinium side effects (all grades) were green discolouration (68%), chromaturia (35%), asthaenic conditions (51%), nausea/vomiting (44%), pain and paraesthesia (28%), finger blisters (20%), hypertension (20%), diarrhoea (22%), headache (30%), liver function abnormalities (11%), dizziness (14%), rash (10%), arthralgia (10%), and gastrointestinal and abdominal pain (9%).

El‐Hamamsy 2016 reported that use of simvastatin added to WBRT was tolerated. Data show a significant difference within the simvastatin group with respect to serum alanine aminotransferase (ALT) at baseline median 24 IU/L (95% CI 16 to 29) and after WBRT 36 IU/L (95% CI 17 to 59) (P = 0.035).

Zeng 2016 reported no significant differences in toxicity between WBRT with or without sodium gycididazole and no grade 4 or 5 toxicities.

3. WBRT and systemic therapy

A total of nine fully published trials have provided information on comparison of the effectiveness of WBRT and systemic therapy (Guerrieri 2004; Knisely 2008; Lee 2008; Mornex 2003; Neuhaus 2009; Postmus 2000; Robinet 2001; Sperduto 2013; Ushio 1991). However, we could not pool results from the trials of systemic therapy and WBRT owing to the heterogeneity of study interventions. Therefore, we report the results for each trial separately.

In the randomised controlled trial by Ushio 1991, investigators randomised 100 participants to one of three treatment groups and reported brain tumour regression (more than 50%) as follows.

WBRT alone (36%, or 5/14).
WBRT plus chloroethyl nitrosoureas (methyl‐CCNU or ACNU) (69%, or 11/16).
WBRT plus chloroethyl nitrosoureas plus tegafur (74%, or 14/19).

Response rates were 36% (5/14) for WBRT alone and 74% (14/19) for WBRT + chloroethyl nitrosoureas + tegafur (P < 0.05). For median survival times, however, three groups reported P > 0.05.

WBRT alone, 27 weeks.
WBRT plus chloroethyl nitrosoureas (methyl‐CCNU or ACNU), 29 weeks.
WBRT plus chloroethyl nitrosoureas plus tegafur, 30.5 weeks.

Investigators did not provide information regarding intracranial progression‐free duration, quality of life, symptom control, neurological function, and ability to taper down on dexamethasone dose. Two participants died of probable adverse effects of chemotherapy.

Postmus 2000 conducted an RCT on use of teniposide chemotherapy with or without WBRT in participants with small‐cell lung cancer metastatic to brain. Data show a 57% response rate in the combined modality group and a 22% response rate in the teniposide‐alone group (P < 0.001). Time to brain progression was longer in the combined group (P = 0.005). Response outside the brain was no different. Median survival was 3.5 months for the combined group versus 3.2 months for the teniposide‐only group (P = 0.87). Forty‐five teniposide‐only participants and 43 combined modality participants had a neurological function score both at baseline and after cycle two. Among participants in the teniposide‐only group, eight improved and 34 remained stable compared with 12 and 23. respectively. in the combined modality group. Trial authors did not report quality of life outcomes and did not report the proportions of participants able to reduce their dexamethasone dose. Toxicities were mild, and the most common toxicity was haematological, resulting in dose delays in 45 participants and dose reductions in eight participants. Non‐haematological toxicities occurred infrequently, for example, World Health Organization (WHO) grade 3 or 4 nausea and vomiting occurred in 11% of 60 participants treated with combined modality, and in 5% of 60 participants treated with teniposide only. WHO grade 3 or 4 infection occurred in 4% of 60 participants treated with combined modality and in 6% of 60 participants treated with teniposide only. Less than 3% of 60 participants in each group experienced mucositis, headache, or cutaneous reactions not exceeding WHO grade 3 toxicity.

Robinet 2001 examined early versus delayed WBRT with concurrent cisplatin and vinorelbine chemotherapy in participants with metastatic non‐small‐cell lung cancer. WBRT was given on a delayed (after two to six cycles of chemotherapy for intracranial non‐responders) or early basis (on days 1 to 12 during the first cycle of chemotherapy). Median survival duration was 24 weeks in the delayed radiotherapy group and 21 weeks in the early radiotherapy group (P = 0.12). Investigators did not report intracranial progression‐free duration, quality of life, symptom control, neurological function, or ability to reduce dexamethasone dose but did report a high number of deaths:

as a result of toxicity ‐ 13 participants (six in the delayed radiotherapy group (6.9%) and seven in the early radiotherapy group (8.2%));
as a result of sepsis during severe neutropenia ‐ 10 participants;
from pneumonia (without neutropaenia) ‐ one participant in each study group following the second cycle of chemotherapy; and
from renal failure ‐ one participant in the delayed chemotherapy group after the first cycle.

Mornex 2003 reported on use of fotemustine combined with WBRT (37 participants) versus fotemustine alone (39 participants) for metastatic melanoma to brain. Median survival was 86 days in the fotemustine‐alone group versus 105 days in the combined treatment group (P = 0.73). Objective cerebral response at day 50 was 7.4% in the fotemustine‐alone group versus 10% in the combined treatment group (P = 0.73). Investigators did not report quality of life, symptom control, neurological function, or ability to reduce dexamethasone dose. Myelosuppression was the most severe adverse event. Delayed grade 3 to 4 neutropaenia occurred in 46% of 39 participants in the fotemustine‐alone group and in 35% of 37 participants in the combined treatment group. Delayed grade 3 to 4 thrombocytopaenia occurred in 44% of 39 participants in the fotemustine‐alone group and in 38% of 37 participants in the combined treatment group. Severe anaemia occurred in 5% of 39 participants in the fotemustine‐alone group and in 11% of 37 participants in the combined treatment group. One participant suffered a cerebral haemorrhage, and three participants died of pneumonia.

Guerrieri 2004 randomised participants with brain metastases from non‐small‐cell lung cancer to WBRT alone versus WBRT and carboplatin chemotherapy. The planned accrual target was 300 participants. However, the study was closed early owing to poor accrual, and investigators reported the results of 42 participants entered into the study. As such, no firm conclusions could be drawn. From this small sample, median survival was 4.4 months in the WBRT group versus 3.7 months in the combined group (P = 0.64). Objective brain response rates were 10% in the WBRT group and 29% in the combined treatment group (P = 0.24).

Knisely 2008 compared WBRT versus WBRT and thalidomide. Median survival was 3.9 months for both groups. Trialists noted no novel toxicities but reported that thalidomide was not tolerated. Forty‐eight per cent of participants discontinued thalidomide because of side effects. Time to brain progression curves was similar (P = 0.097) between the two treatment groups.

Lee 2008 included participants with metastatic non‐small‐cell lung cancer to the brain treated with primary chemotherapy (gemcitabine and vinorelbine) followed by WBRT versus WBRT first then followed by the same chemotherapy. Progression‐free survival was 3.6 months versus 4.4 months (P = 0.62), and overall survival was 9.1 months versus 9.9 months (P = 0.61), for primary chemotherapy versus WBRT first groups, respectively. In the WBRT first group, grade 3 or 4 neutropaenia was more frequent (79% vs 40%) during chemotherapy.

Neuhaus 2009 randomised participants with brain metastases from lung cancer (small‐cell lung cancer or non‐small‐cell lung cancer) to WBRT and topotecan chemotherapy versus WBRT alone. The trial was stopped early because of poor accrual (total of 96 participants compared with a planned 320 participants). Data show no significant differences in overall survival (HR 1.32, 95% CI 0.83 to 2.10; P = 0.43) or progression‐free survival (HR 1.28, 95% CI 0.73 to 2.43; P = 0.89). Haematological events occurred mainly in the combined treatment group.

Sperduto 2013 randomised participants with brain metastases with non‐small‐cell lung cancer to WBRT and radiosurgery versus WBRT and radiosurgery plus temozolomide chemotherapy or the molecular targeted agent erlotinib. The trial was stopped early owing to poor accrual (total 126 participants enrolled of a planned 381 participants). The addition of temozolomide or erlotinib to WBRT and radiosurgery was associated with worse survival and toxicity. Median survivals for WBRT and radiosurgery versus WBRT and radiosurgery and temozolomide versus WBRT and radiosurgery and erlotinib were 13.4 (95% CI 6.5 to 20,8), 6.3 (95% CI 3.4 to 10.1), and 6.1 (95% CI 3.6 to 12.1) months, respectively. Grade 3 to 5 toxicities were 11%, 41%, and 49%, respectively (P< 0.001).

4. WBRT plus radiosurgery versus WBRT

Four trials provided information on the effectiveness of WBRT alone compared with WBRT plus radiosurgery (Andrews 2004; Chougule 2000; El Gantery 2014; Kondziolka 1999). The Chougule 2000 trial was published in abstract form and the full report has never been published.

No trials reported the following outcomes of interest.

Quality of life.
Symptom control.

Overall survival

(See comparison Analysis 4.1.)

We pooled results of the two fully published reports (Andrews 2004; Kondziolka 1999). The HR for overall survival was 0.61 (95% CI 0.27 to 1.39; P = 0.24; moderate‐certainty evidence) (summary of findings Table 4) between participants with multiple brain metastases treated with WBRT and radiosurgery boost as compared with WBRT alone.

Andrews 2004 reported improved survival (P = 0.0393) for the subset of participants with surgically unresectable single brain metastasis treated with WBRT and radiosurgery boost as compared with WBRT alone (median survival 6.5 months vs 4.9 months).

We could not pool survival data from the El Gantery 2014 trial as an HR could not be calculated owing to lack of detail in the publication. Chougule 2000 (reported in abstract form) also did not find median survival differences (P value not stated) among the randomised groups (radiosurgery alone: seven months; radiosurgery and WBRT: five months; WBRT alone: nine months). We could not pool results of the Chougule 2000 trial as insufficient detail was available from the published abstract.

One‐year overall brain tumour control

(See comparison Analysis 4.2.)

Local control was defined as unchanged or improved serial post‐treatment MRI scans judged as showing complete response, partial response, or stable disease. Progressive disease was defined as an increase in size of any brain lesion, development of new brain lesions, or stable disease with neurological deterioration. We pooled data for one‐year overall brain control provided by the Kondziolka 1999, Andrews 2004, and El Gantery 2014 trials, and found that overall, the HR was 0.39 (95% CI 0.25 to 0.60; P < 0.0001; high‐certainty evidence) for local brain control favouring the WBRT and radiosurgery boost group (summary of findings Table 4). Data on one‐year local brain control were not provided in the Chougule 2000 abstract.

Intracranial progression‐free duration

In the Kondziolka 1999 trial, median time to local brain failure was six months after WBRT alone (95% CI 3.5 to 8.5) in comparison with 36 months after WBRT and radiosurgery boost (95% CI 15.6 to 57). In the Andrews 2004 trial, the P value was 0.1278 with respect to overall time to intracranial tumour progression or neurological death rates between those treated with WBRT and radiosurgery boost as compared with WBRT alone. The Chougule 2000 abstract and the El Gantery 2014 report did not describe intracranial progression‐free duration. We could not pool these results owing to differences in reporting of this outcome.

Neurological function

Kondziolka 1999, Chougule 2000, and El Gantery 2014 did not report neurological function outcomes. Andrews 2004 reported that the KPS was improved at six months in 10/79 or 13% of participants treated with WBRT and radiosurgery boost as compared with 3/75 or 4% of participants treated with WBRT alone (P = 0.0331). However, mental status as measured by the MMSE was similar between the two groups.

Proportion of participants who were able to reduce their daily dexamethasone dose and duration

Andrews 2004 reported that a higher proportion of participants (41/79) in the WBRT and radiosurgery group had decreased steroid requirements at six months as compared with 25/75 participants in the WBRT‐alone arm (P = 0.0158). The other trials did not report dexamethasone use.

Adverse effects

Kondziolka 1999 reported no neurological or systemic morbidity related to stereotactic radiosurgery. After WBRT, participants expectedly developed mild scalp erythema and hair loss. In the Andrews 2004 trial, early and late toxicities did not differ greatly between the two treatment arms. However, investigators included more participants with acute grade 3 and 4 toxicity in the WBRT and radiosurgery boost arm (4/160 participants) than in the WBRT‐alone arm (0/166 participants). Also, they reported more late grade 3 and 4 toxicity in the combined arm (6/160 participants) than in the WBRT‐alone arm (3/166 participants). El Gantery 2014 reported no differences in acute nor late toxicities among randomised groups.

5. Radiosurgery alone versus radiosurgery and WBRT

For this updated review, we identified two new fully published trials (on radiosurgery alone versus radiosurgery and WBRT (Brown 2016; Fogarty 2015).

Fogarty 2015 randomised participants with metastatic melanoma brain metastases to local therapy (neurosurgical resection or radiosurgery) followed by WBRT or observation.

None of the trials examining use of radiosurgery alone versus radiosurgery and WBRT reported on response rates of brain metastases (CR + PR) nor on the percentage of participants able to reduce the dexamethasone dose or the duration of a reduced dexamethasone dose (Aoyama 2006; Brown 2016; Chang 2009; Fogarty 2015; Kocher 2011).

Overall survival

The HR for the pooled overall survival analysis for three RCTs (Aoyama 2006; Brown 2016; Chang 2009) was 1.0 (95% CI 0.80 to 1.25, P = 0.99; moderate‐certainty evidence) (Analysis 5.1; summary of findings Table 5).

Fogarty 2015 did not report on survival outcomes.

In the Kocher 2011 trial, overall survival for radiosurgery alone versus WBRT and radiosurgery boost could not be isolated. Nevertheless, median survival was 10.9 months (among participants treated with combined therapy) versus 10.7 months (in those treated with surgery or radiosurgery alone) (P = 0.89).

One‐year radiosurgery‐targeted lesion control

Data from four trials show an HR of 2.73 (95% CI 1.87 to 3.99; P < 0.00001) for one‐year radiosurgery‐targeted lesion control (Analysis 5.2), with high‐certainty evidence favouring WBRT and radiosurgery versus radiosurgery alone (summary of findings Table 5) (Aoyama 2006; Brown 2016; Chang 2009; Kocher 2011).

Fogarty 2015 did not report local brain control outcomes by randomised study group.

One‐year distant brain control

The HR from four trials was 2.34 (95% CI 1.73 to 3.18; P < 0.00001) for one‐year distant brain control (Analysis 5.3), with high‐certainty evidence favouring WBRT and radiosurgery versus radiosurgery alone (summary of findings Table 5) (Aoyama 2006; Brown 2016; Chang 2009; Kocher 2011).

Fogarty 2015 did not report distant brain control outcomes by randomised study group.

Neurological function

Aoyama 2006 reported systemic functional preservation rates (KPS ≥ 70) at 12 months of 33.9% in the WBRT and radiosurgery arm as compared with 26.9% in the radiosurgery‐alone arm (P = 0.53). The neurological preservation rate at 12 months was 72.1% in the WBRT and radiosurgery arm, similar to 70.3% in the radiosurgery alone arm (P = 0.99).

Chang 2009 did not describe neurological function outcomes.

In Kocher 2011, the duration of functional independence was the primary endpoint (defined as the date of randomisations to the first report of WHO performance status decline > 2). Median time to WHO performance status greater than 2 (survival with functional independence) was 10 months in the surgery or radiosurgery‐alone arm versus 9.5 months in the surgery and WBRT or radiosurgery and WBRT arms (P = 0.71).

Brown 2016 reported that Barthel activity of daily living scores were not significantly different between the two randomised groups.

Fogarty 2015 did not report neurological function outcomes.

Neurocognition

The Aoyama 2006 trial used a crude measure of neurocognition ‐ the MMSE. Aoyama 2006 reported that average duration until MMSE deterioration was 16.5 months in the WBRT and radiosurgery group versus 7.6 months in the radiosurgery‐alone group (P = 0.05).

Neurocognition was the primary outcome in Chang 2009, which used a formal battery of neurocognitive tests. After 58 participants were entered, the trial was stopped early based on early stopping rules indicating the high probability (96%) that participants randomised to radiosurgery and WBRT were significantly more likely to show a decline in learning and memory function (mean posterior probability of decline 52%) at four months compared with participants treated with radiosurgery alone (mean posterior probability of decline 24%).

Kocher 2011 did not report neurocognitive outcomes.

The primary endpoint in the Brown 2016 trial was cognitive deterioration, which was less frequent in the radiosurgery‐alone group than in the radiosurgery plus WBRT group (63.5% vs 91.7%; P < 0.001).

Fogarty 2015 did not report neurocognitive outcomes.

Quality of life

Aoyama 2006 did not report quality of life outcomes.

The Chang 2009 trial used the validated quality of life instrument FACT‐BR. The FACT‐BR mean difference between randomised groups at four months compared with baseline was 2.8 (95% CI ‐26 to 21; P = 0.76). Trial authors indicated, however, that the wide confidence interval implied that the quality of life results were inconclusive.

Kocher 2011 indicated that quality of life was a secondary endpoint but has not yet reported trial results related to quality of life.

Brown 2016 reported better overall quality of life at three months among participants randomised to radiosurgery alone compared with radiosurgery and WBRT (P = 0.001).

Fogarty 2015 did not report quality of life outcomes.

Side effects

Aoyama 2006 observed symptomatic acute neurological toxicity in four (of 65) participants receiving WBRT and radiosurgery and in eight (of 67) participants receiving radiosurgery alone (P = 0.36). Investigators noted symptomatic late neurological toxic effects in seven participants in the WBRT and radiosurgery group versus three participants in the radiosurgery‐alone group (P = 0.20).

Chang 2009 reported one case of grade 3 toxicity in the radiosurgery and WBRT group and one case of grade 3 toxicity in the radiosurgery‐alone group that were attributable to radiation. Data also show two cases of grade 4 toxicity (pathologically proven radiation necrosis) in the radiosurgery‐alone group.

Kocher 2011 reported 16 serious acute toxicities (13 participants in the surgery and WBRT or radiosurgery and WBRT group and three participants in the surgery or radiosurgery‐alone group). Trialists noted symptomatic radionecrosis in 8% (7/90 participants) after radiosurgery alone and in 13% (12/95 participants) after treatment with radiosurgery and WBRT.

Brown 2016 reported no differences in grade 3 or higher adverse effects between the two randomised groups.

Fogarty 2015 reported minimal adverse events and no serious adverse events related to the trial.

6. Steroids alone versus steroids and WBRT

One randomised trial examined use of prednisone with or without WBRT (Horton 1971). This small trial (48 recruited participants) was conducted at a time when computed tomography (CT) scanning was not available. Participants were enrolled in the study if they had a histologically proven cancer and clinical symptoms and signs of brain metastases, such as abnormalities identified by radioisotope brain scans, electroencephalograms (EEGs), angiograms, and spinal fluid chemistry and cytology. The proportion of participants with improved performance status was similar in the prednisone‐alone and WBRT plus prednisone groups (63% of 19 participants and 61% of 28 participants, respectively). Median survival time in the prednisone‐alone group was 10 weeks as compared with 14 weeks in the combined treatment group (P value not stated). Participants were not stratified for other known prognostic factors such as age, performance status, and extent of extracranial disease. Investigators did not report other outcomes such as intracranial progression‐free duration, quality of life, symptom control, neurological function, toxicities, and ability to taper steroid doses.

Mulvenna 2016 published a multi‐centre randomised controlled trial of optimal supportive care (including use of dexamethasone) and WBRT versus optimal supportive care alone in 538 non‐small‐cell lung cancer participants with brain metastases not eligible for surgery or radiosurgery. The HR for overall survival was 1.06 (95% CI 0.90 to 1.26; P = 0.8084) between the two randomised groups. As well, overall quality of life and dexamethasone use were similar between the two groups. The primary endpoint was quality‐adjusted life‐years, which was 46.4 days, with a standard deviation (SD) of 3.66 days for those assigned to optimal supportive care and WBRT versus 41.7 days (SD 3.23) for participants randomised to optimal supportive care alone. The mean difference was only 4.7 days in favour of optimal supportive care and WBRT. Trial authors reported no differences in symptoms or side effects and no performance status changes in both groups. They did not report brain control and response.

We could not pool data from Horton 1971 and Mulvenna 2016, as Horton 1971 did not publish results in sufficient detail.

7. WBRT alone or with molecular targeted agents

Sperduto 2013 examined both chemotherapy and molecular targeted therapy. Investigators randomised participants with metastatic non‐small‐cell lung cancer to brain to one of three groups: WBRT and radiosurgery; WBRT, radiosurgery, and temozolomide chemotherapy; or WBRT, radiosurgery, and the targeted agent, erlotinib. The trial was stopped early owing to poor accrual (total 126 participants enrolled of a planned 381 participants). The addition of temozolomide or erlotinib to WBRT and radiosurgery was associated with worse survival and toxicity. Median survival for WBRT and radiosurgery versus WBRT and radiosurgery and temozolomide versus WBRT and radiosurgery and erlotinib was 13.4 (95% CI 6.5 to 20.8), 6.3 (95% CI 3.4 to 10.1), and 6.1 (95% CI 3.6 to 12.1) months, respectively. Grade 3 to 5 toxicities were 11%, 41%, and 49%, respectively (P< 0.001).

Yang 2015 outlined in abstract form an ongoing phase 3 randomised trial of WBRT with or without erlotinib for metastatic non‐small‐cell lung cancer to brain. Trialists did not report results in this abstract.

8. WBRT alone or with neurocognitive protective agents

The Brown 2013 double‐blind trial randomised participants with brain metastases to memantine versus placebo. The primary outcome was median decline in delayed recall at 24 weeks, which was 0 in the memantine group and ‐0.90 in the placebo group (P = 0.059). However, owing to attrition, Investigators analysed only 149 (of 554 randomised) participants at that time point. For the secondary endpoint ‐ time to cognitive failure ‐ trial authors reported that participants in the memantine group had longer time to cognitive decline (HR 0.78, 95% CI 0.62 to 0.99; P = 0.01). The memantine group was favoured for executive function at 8 weeks (P = 0.008) and at 16 weeks (P = 0.0041). In addition, the memantine group was better for processing speed (P = 0.0137) and delayed recognition (P = 0.0149) at 24 weeks. Grade 3 and 4 side effects were the same (28%) in both groups. The most common side effects were fatigue, alopecia, nausea, and headache, with no differences reported between the two groups. Investigators did not report quality of life.

Rapp 2015 was a double‐blind trial that randomised participants with brain metastases and primary brain tumour who had received partial brain radiotherapy or WBRT to donepezil versus placebo. At 24 weeks, mean composite cognitive score was 0.22 versus 0.19 in donepezil and control groups, respectively (P = 0.484). For the subset analysis of memory recognition (P = 0.027), memory discrimination (P = 0.007), and motor speed/dexterity (P = 0.016), use of donepezil was favoured. Trial authors did not report survival, progression‐free survival, quality of life, dexamethasone use, neurological function, or side effects.

Owing to the heterogeneity of included participants (brain metastasis and primary brain tumour) in the Rapp 2015 trial, we did not pool these results. The subset of participants with brain metastases could not be separated in the Rapp 2015 publication.

9. Hippocampal sparing WBRT versus WBRT

No fully published phase 3 randomised trials have examined use of hippocampal sparing WBRT versus WBRT. However, phase 3 trials are ongoing (NRG‐CC001; Sturm 2015). The Sturm 2015 abstract described the planned hippocampal sparing trial but presented no results.

Discussion

Summary of main results

Altered whole body radiation therapy (WBRT) dose‐fractionation schedules versus conventional WBRT fractionation schedule

In summary, none of the included randomised controlled trials (RCTs) found benefit (in terms of overall survival, neurological function, or symptom control) with higher biological altered dose‐fractionation schedules as compared with standard doses (3000 cGy in 10 fractions, or 2000 cGy in 4 or 5 daily fractions). Two studies provided data on comparison of two fractionation schedules commonly employed in Canada (2000 cGy in 5 fractions, or 3000 cGy in 10 fractions) (Borgelt 1980a; Chatani 1994). Overall survival and neurological function were similar between these two fractionation schemes.

However, trials with lower biological altered dose‐fractionation schedules reported worse survival and neurological function improvement (NFI) as compared with standard care.

Two trials reported on 4000 cGy in 20 twice‐daily (BID) fractions WBRT versus 2000 cGy in 4 or 5 daily fractions WBRT (Davey 2008; Graham 2010). The hazard ratio (HR) for overall survival was 1.18 (95% confidence interval (CI) 0.89 to 1.56; P = 0.25; moderate‐certainty evidence).

A post hoc subset analysis in the Davey 2008 trial showed that time to retreatment for intracranial relapse was 14 weeks in the control WBRT group (2000 cGy in 5 daily fractions) as compared with 32 weeks in the 4000 cGy in 20 BID (twice‐daily) fractions group (P = 0.03). The primary endpoint for the Graham 2010 trial was proportion of participants with overall brain progression. Brain progression was worse for participants treated with 2000 cGy in 4 fractions as compared with 4000 cGy in 20 BID fractions (64% vs 44%; P = 0.03).

WBRT plus radiosensitisers versus WBRT

In an attempt to improve local brain control, clinicians have added radiosensitisers to WBRT. However, none of the included RCTs showed benefit in terms of overall survival or brain response (complete response (CR) + partial response (PR)). Use of radiosensitisers in these trials (lonidamine, metronidazole, misonidazole, bromodeoxyuridine (BrdU), motexafin gadolinium, efaproxiral) was associated with toxicity. Use of radiosensitisers with WBRT remains experimental.

WBRT and systemic therapy

All fully published phase 3 trials examining use of WBRTand chemotherapy versus WBRT alone reported no survival benefit with the addition of chemotherapy to WBRT. These trials showed an increase in toxicity when chemotherapy was added to WBRT.

Use of chemotherapy with WBRT remains experimental.

WBRT plus radiosurgery versus WBRT

We pooled results of the two fully published reports (Andrews 2004; Kondziolka 1999). The HR for overall survival was 0.61 (95% CI 0.27 to 1.39; P = 0.24) between participants with multiple brain metastases treated with WBRT and radiosurgery boost as compared with those given WBRT alone. The Kondziolka 1999 trial was methodologically weaker (small trial size and early trial closure) as compared with the Andrews 2004 trial. Andrews 2004 included participants with one to three brain metastases. Although overall there was no improvement in survival with the addition of radiosurgery boost in the Andrews 2004 trial, survival was improved among participants with surgically unresectable single brain metastasis. This trial was designed with an a priori sample size calculation to test the hypothesis of improved survival in participants with single brain metastasis. Median survival time was 6.5 months for those with single brain metastasis treated with WBRT and radiosurgery boost versus 4.9 months for those treated with WBRT alone (P = 0.0393). The Kondziolka 1999 trial excluded participants with single brain metastasis (two to four brain metastases included).

We pooled data for one‐year overall brain control in the Kondziolka 1999, Andrews 2004, and El Gantery 2014 trials. Overall, the HR was 0.39 (95% CI 0.25 to 0.60; P < 0.0001; high‐certainty evidence) for local brain control, favouring the WBRT and radiosurgery boost group.

Thus, although overall brain control favours the addition of radiosurgery boost, radiosurgery boost added to WBRT did not improve survival among participants with two to four brain metastases. One trial reported that survival was increased among selected participants with surgically unresectable single brain metastasis treated with WBRT and radiosurgery boost as compared with WBRT alone.

Radiosurgery alone versus WBRT and radiosurgery

The HR for the pooled overall survival analysis for the three randomised controlled trials comparing radiosurgery alone versus WBRT and radiosurgery was 1.0 (95% CI 0.80 to 1.25; P = 0.99) (Aoyama 2006; Chang 2009; Brown 2016).

Again, these trials pre‐dated the recognition of significant prognostic factors identified by the Diagnosis‐Specific Graded Prognostic Assessment. As well, molecular markers are now recognised as important for survival of certain cancers. It is unclear whether specific groups of participants with brain metastases may derive survival benefit with WBRT and radiosurgery as compared with radiosurgery alone, as trials evaluated heterogeneous participants with brain metastases.

Although local and distant control of brain metastases is favoured with WBRT and radiosurgery as compared with radiosurgery alone, two trials reported worse neurocognitive function after WBRT and radiosurgery as compared with radiosurgery alone (Brown 2016; Chang 2009). Brown 2016 reported better quality of life outcomes at three months among participants randomised to radiosurgery alone versus WBRT and radiosurgery.

Although overall survival is not different among the strategies of radiosurgery alone, radiosurgery and WBRT, and WBRT alone for selected patients with brain metastases, the strategy of radiosurgery alone spares neurocognitive decline associated with WBRT and has been associated with better quality of life as compared with WBRT.

Steroids versus steroids and WBRT

The older Horton 1971 trial could not be interpreted as the study was small, diagnostic criteria used for brain metastases were outdated, and no statistical analysis was performed.

The modern multi‐centre trial of Mulvenna 2016 reported no differences in quality‐adjusted life‐years for participants with brain metastases from non‐small‐cell lung cancer randomised to optimal supportive care and WBRT versus WBRT alone. However Mulvenna 2016 spanned the time from March 2007 to August 2014. During this time, molecular drivers for non‐small‐cell lung cancer such as epidermal growth factor receptor (EGFR), anaplastic lymphoma kinase (ALK), and programmed death‐ligand 1 (PD‐L1) were becoming increasingly important for prognostic classification and molecular targeted therapy. It is not known whether the two randomised groups were balanced for molecular features. Furthermore, survival was notably short, with a median survival of 9.2 weeks (95% CI 7.2 to 11.1) for those who received optimal supportive care and WBRT versus 8.5 weeks (95% CI 7.1 to 9.9) for those who received optimal supportive care alone. Participants in this trial were not suitable for resection or radiosurgery and likely were selected for enrolment in this trial on the basis of estimated poor prognoses.

More RCTs are needed to help guide practitioners as to which subsets of patients with brain metastases may benefit (in terms of survival, symptom control, or quality of life) from WBRT, and which subsets may not.

WBRT and molecular targeted agents

Sperduto 2013, which was closed early owing to poor accrual, reported worse survival and increased toxicity in participants with non‐small‐cell lung cancer with brain metastases randomised to WBRT and radiosurgery and erlotinib as compared with WBRT and radiosurgery. However, participants were not tested for EGFR mutation; thus any possible beneficial effect of erlotinib may have been diluted by participants with non‐EGFR mutated non‐small‐cell lung cancer. Additional trials are needed to assess use of molecular targeted agents in subsets of patients with brain metastases, based on molecular cancer profiles.

WBRT and neurocognitive protective agents

Although the Brown 2013 and Rapp 2015 trials were negative for the primary outcome of delayed recall or composite cognitive scores with use of neuroprotective agents (memantine and donepezil, respectively), certain subsets of cognitive function may be improved with use of neuroprotective agents combined with WBRT for brain metastases.

Hippocampal sparing WBRT versus WBRT

Full reports from phase III randomised trials are awaited to determine whether hippocampal sparing WBRT preserves cognitive function as compared with WBRT.

Overall completeness and applicability of evidence

These trials pre‐date the recognition of significant prognostic factors identified by the Diagnosis‐Specific Graded Prognostic Assessment (Sperduto 2012). As well, molecular markers are increasingly important for therapeutic intervention and for prognoses in certain cancers. It is unclear whether groups in these older trials were balanced for these prognostic variables.

In addition, results of these RCTs may not be applicable to certain subgroups of patients with brain metastases (e.g. based on molecular tumour subtypes, or the Diagnosis‐Specific Graded Prognostic Assessment), as trials evaluated heterogeneous patients with brain metastases. Most participants in these studies had lung, breast, or colorectal cancer primaries. Participants with small‐cell lung cancer, germ cell tumours, and haematological malignancies were often excluded from trials.

Many of these trials lacked quality of life outcomes, neurocognitive outcomes, and information regarding the proportion of participants in each group who were able to reduce steroid dose. In cases in which overall survival was similar among trial groups, palliative endpoints such as quality of life, neurocognition, and steroid use are important.

Quality of the evidence

Evidence ranged from moderate to high certainty. Depending on the comparisons, downgrading of evidence was based on one or more of the following factors.

The included population was heterogeneous with respect to survival owing to varying primary cancer diagnoses and varying molecular subtypes.
Attriton bias
Reporting bias
Trial groups may not have been balanced for steroid use.

Potential biases in the review process

This meta‐analysis is biased toward older trials, which are flawed by the inclusion of heterogeneous populations with brain metastases. Older trials are also limited by the lack of neurocognitive outcomes, which have become increasingly important in later trials examining radiosurgery and neurocognitive protective strategies.

Agreements and disagreements with other studies or reviews

Findings of this meta‐analysis are consistent with those of other published meta‐analyses examining interventions using WBRT (Patil 2017; Tsao 2012a).

Figure 1

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Figure 3

Study flow diagram.

$Comparison 1: Altered WBRT fractionation schedules versus WBRT control, Outcome 1: Overall survival: lower‐dose WBRT vs control WBRT (3000 cGy/10 daily fractions)$

Analysis 1.1

Comparison 1: Altered WBRT fractionation schedules versus WBRT control, Outcome 1: Overall survival: lower‐dose WBRT vs control WBRT (3000 cGy/10 daily fractions)

$Comparison 1: Altered WBRT fractionation schedules versus WBRT control, Outcome 2: Overall survival: higher‐dose WBRT vs control (3000 cGy/10 daily fractions)$

Analysis 1.2

Comparison 1: Altered WBRT fractionation schedules versus WBRT control, Outcome 2: Overall survival: higher‐dose WBRT vs control (3000 cGy/10 daily fractions)

$Comparison 1: Altered WBRT fractionation schedules versus WBRT control, Outcome 3: Overall survival: WBRT 4000 cGy/20 fractions BID vs control WBRT (2000 cGy/4‐5 daily fractions)$

Analysis 1.3

Comparison 1: Altered WBRT fractionation schedules versus WBRT control, Outcome 3: Overall survival: WBRT 4000 cGy/20 fractions BID vs control WBRT (2000 cGy/4‐5 daily fractions)

$Comparison 2: Altered WBRT fractionation schedules versus WBRT control: neurological function improvement, Outcome 1: Neurological function improvement: lower‐dose WBRT vs control dose WBRT (3000 cGy/10 fractions)$

Analysis 2.1

Comparison 2: Altered WBRT fractionation schedules versus WBRT control: neurological function improvement, Outcome 1: Neurological function improvement: lower‐dose WBRT vs control dose WBRT (3000 cGy/10 fractions)

$Comparison 2: Altered WBRT fractionation schedules versus WBRT control: neurological function improvement, Outcome 2: Neurological function improvement: higher‐dose WBRT vs control dose WBRT (3000 cGy/10 fractions)$

Analysis 2.2

Comparison 2: Altered WBRT fractionation schedules versus WBRT control: neurological function improvement, Outcome 2: Neurological function improvement: higher‐dose WBRT vs control dose WBRT (3000 cGy/10 fractions)

Analysis 3.1

Comparison 3: WBRT with radiosensitisers (radiosen) versus WBRT alone, Outcome 1: Overall survival

Analysis 3.2

Comparison 3: WBRT with radiosensitisers (radiosen) versus WBRT alone, Outcome 2: Brain tumour response rates: complete response (CR) and partial response (PR) combined

Analysis 4.1

Comparison 4: WBRT and radiosurgery versus WBRT, Outcome 1: Overall survival

Analysis 4.2

Comparison 4: WBRT and radiosurgery versus WBRT, Outcome 2: 1‐Year overall brain control rates

Analysis 5.1

Comparison 5: Radiosurgery alone versus WBRT and radiosurgery, Outcome 1: Overall survival

Analysis 5.2

Comparison 5: Radiosurgery alone versus WBRT and radiosurgery, Outcome 2: 1‐Year radiosurgery‐targeted lesion control

Analysis 5.3

Comparison 5: Radiosurgery alone versus WBRT and radiosurgery, Outcome 3: 1‐Year distant brain control

Summary of findings 1. Altered WBRT fractionation schedules compared with WBRT control for treatment of newly diagnosed multiple brain metastases

Outcomes	Relative effect (95% CI)	*Anticipated absolute effects (95% CI)**			Certainty (quality) of the evidence (GRADE)	What happens
Altered WBRT fractionation schedules compared with WBRT control for treatment of newly diagnosed multiple brain metastases
Patient or population: patients with newly diagnosed multiple brain metastases Setting: hospital Intervention: altered WBRT fractionation schedules Comparison: WBRT control
Outcomes	Relative effect (95% CI)	Without altered WBRT fractionation schedules	With altered WBRT fractionation schedules	Difference	Certainty (quality) of the evidence (GRADE)	What happens
Overall survival: lower‐dose WBRT vs control WBRT (3000 cGy/10 daily fractions) No. of participants: 705 (3 RCTs)	HR 1.21 (1.04 to 1.40)	Study population			⊕⊕⊕⊝ MODERATE^a
	HR 1.21 (1.04 to 1.40)	Not estimable as only HR data available	Not estimable as only HR data available	N/A	⊕⊕⊕⊝ MODERATE^a
Overall survival: higher‐dose WBRT vs control (3000 cGy/10 daily fractions) No. of participants: 846 (4 RCTs)	HR 0.97 (0.83 to 1.12)	Study population			⊕⊕⊕⊝ MODERATE^a
	HR 0.97 (0.83 to 1.12)	Not estimable as only HR data available	Not estimable as only HR data available	N/A	⊕⊕⊕⊝ MODERATE^a
Overall survival: WBRT 4000 cGy/20 fractions BID vs control WBRT (2000 cGy/4‐5 daily fractions) No. of participants: 203 (2 RCTs)	HR 1.18 (0.89 to 1.56)	Study population			⊕⊕⊕⊝ MODERATE^a
	HR 1.18 (0.89 to 1.56)	Not estimable as only HR data available	Not estimable as only HR data available	N/A	⊕⊕⊕⊝ MODERATE^a
Adverse effects No. of participants: 1754 (9 RCTs)	not pooled	Study population			⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 1754 (9 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊕ HIGH
Symptom control No. of participants: 2877 (7 RCTs)	not pooled	Study population			⊕⊕⊕⊝ MODERATE^b
Symptom control No. of participants: 2877 (7 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^b
Overall brain control No. of participants: 203 (2 RCTs)	not pooled	Study population			⊕⊕⊕⊝ MODERATE^c
Overall brain control No. of participants: 203 (2 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^c
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). BID = twice daily; CI = confidence interval; HR = hazard ratio; N/A = not applicable; OR = odds ratio; RCT = randomised controlled trial; RR = risk ratio; WBRT = whole brain radiation therapy.
GRADE Working Group grades of evidence. High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.
^aIncluded population heterogeneous with respect to survival owing to varying primary cancer diagnoses. Molecular subtypes important for prognoses may not have been balanced between groups. ^bOutcome of symptom control subject to attrition and reporting bias. ^cReporting bias.

Summary of findings 1. Altered WBRT fractionation schedules compared with WBRT control for treatment of newly diagnosed multiple brain metastases

Summary of findings 2. Altered WBRT fractionation schedules compared with WBRT control: neurological function improvement for treatment of newly diagnosed multiple brain metastases

Outcomes	Relative effect (95% CI)	*Anticipated absolute effects (95% CI)**			Certainty (quality) of the evidence (GRADE)	What happens
Altered WBRT fractionation schedules compared with WBRT control: neurological function improvement for treatment of newly diagnosed multiple brain metastases
Patient or population: patients with newly diagnosed multiple brain metastases Setting: hospital Intervention: altered WBRT fractionation schedules Comparison: WBRT control: neurological function improvement
Outcomes	Relative effect (95% CI)	Without altered WBRT fractionation schedules	With altered WBRT fractionation schedules	Difference	Certainty (quality) of the evidence (GRADE)	What happens
Neurological function improvement: lower‐dose WBRT vs control dose WBRT (3000 cGy/10 fractions) No. of participants: 1612 (5 RCTs)	OR 1.74 (1.06 to 2.84)	Study population			⊕⊕⊕⊝ MODERATE^a
	OR 1.74 (1.06 to 2.84)	49.6%	63.1% (51.0 to 73.6)	13.5% more (1.5 more to 24.1 more)	⊕⊕⊕⊝ MODERATE^a
Neurological function improvement: higher‐dose WBRT vs control dose WBRT (3000 cGy/10 fractions) No. of participants: 1480 (4 RCTs)	OR 1.14 (0.92 to 1.42)	Study population			⊕⊕⊕⊝ MODERATE^a
	OR 1.14 (0.92 to 1.42)	49.4%	52.7% (47.3 to 58.1)	3.3% more (2.1 fewer to 8.7 more)	⊕⊕⊕⊝ MODERATE^a
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI = confidence interval; OR = odds ratio; RCT = randomised controlled trial; RR = risk ratio; WBRT = whole body radiation therapy.
GRADE Working Group grades of evidence. High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.
^aUnclear whether groups were balanced for steroid use.

Summary of findings 2. Altered WBRT fractionation schedules compared with WBRT control: neurological function improvement for treatment of newly diagnosed multiple brain metastases

Summary of findings 3. WBRT with radiosensitisers (radiosen) compared with WBRT alone for treatment of newly diagnosed multiple brain metastases

Outcomes	Relative effect (95% CI)	*Anticipated absolute effects (95% CI)**			Certainty (quality) of the evidence (GRADE)	What happens
WBRT with radiosensitisers (radiosen) compared with WBRT alone for treatment of newly diagnosed multiple brain metastases
Patient or population: patients with newly diagnosed multiple brain metastases Setting: hospital Intervention: WBRT with radiosensitisers (radiosen) Comparison: WBRT alone
Outcomes	Relative effect (95% CI)	Without WBRT with radiosensitisers (radiosen)	With WBRT with radiosensitisers (radiosen)	Difference	Certainty (quality) of the evidence (GRADE)	What happens
Overall survival No. of participants: 2631 (8 RCTs)	HR 1.05 (0.99 to 1.12)	study population			⊕⊕⊕⊝ MODERATE^a
Overall survival No. of participants: 2631 (8 RCTs)	HR 1.05 (0.99 to 1.12)	not estimable as only HR data available	not estimable as only HR data available	N/A	⊕⊕⊕⊝ MODERATE^a
Brain tumour response rates: complete response (CR) and partial response (PR) combined No. of participants: 847 (6 RCTs)	OR 0.84 (0.63 to 1.11)	study population			⊕⊕⊕⊕ HIGH
	OR 0.84 (0.63 to 1.11)	62.2%	58.1% (50.9 to 64.6)	4.2% fewer (11.3 fewer to 2.4 more)	⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 2631 (8 RCTs)	not pooled	study population			⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 2631 (8 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊕ HIGH
Quality of life No. of participants: 966 (3 RCTs)	not pooled	study population			⊕⊕⊕⊝ MODERATE^b
Quality of life No. of participants: 966 (3 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^b
Symptom control No. of participants: 1814 (3 RCTs)	not pooled	study population			⊕⊕⊕⊝ MODERATE^b
Symptom control No. of participants: 1814 (3 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^b
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI = confidence interval; HR = hazard ratio; N/A = not applicable; OR = odds ratio; RCT = randomised controlled trial; RR = risk ratio; WBRT = whole body radiation therapy.
GRADE Working Group grades of evidence. High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.
^aHeterogeneous population owing to varying primary cancer histologies, and unclear whether groups were balanced for other prognostic variables such as molecular subtype. ^bAttrition and reporting bias.

Summary of findings 3. WBRT with radiosensitisers (radiosen) compared with WBRT alone for treatment of newly diagnosed multiple brain metastases

Summary of findings 4. WBRT and radiosurgery compared with WBRT for treatment of newly diagnosed multiple brain metastases

Outcomes	Relative effect (95% CI)	*Anticipated absolute effects (95% CI)**			Certainty (quality) of the evidence (GRADE)	What happens
WBRT and radiosurgery compared with WBRT for treatment of newly diagnosed multiple brain metastases
Patient or population: patients with newly diagnosed multiple brain metastases Setting: hospital Intervention: WBRT and radiosurgery Comparison: WBRT
Outcomes	Relative effect (95% CI)	Without WBRT and radiosurgery	With WBRT and radiosurgery	Difference	Certainty (quality) of the evidence (GRADE)	What happens
Overall survival No. of participants: 358 (2 RCTs)	HR 0.61 (0.27 to 1.39)	study population			⊕⊕⊕⊝ MODERATE^a
Overall survival No. of participants: 358 (2 RCTs)	HR 0.61 (0.27 to 1.39)	not estimable as only HR data available	not estimable as only HR data available	N/A	⊕⊕⊕⊝ MODERATE^a
1‐Year overall brain control rates No. of participants: 400 (3 RCTs)	HR 0.39 (0.25 to 0.60)	study population			⊕⊕⊕⊕ HIGH
	HR 0.39 (0.25 to 0.60)	not estimable as only HR data available	not estimable as only HR data available	N/A	⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 400 (3 RCTs)	not pooled	study population			⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 400 (3 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊕ HIGH
Intracranial progression‐free duration No. of participants: 400 (3 RCTs)	not pooled	study population			⊕⊕⊕⊕ HIGH
	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊕ HIGH
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI = confidence interval; HR = hazard ratio; OR = odds ratio; RCT = randomised controlled trial; RR = risk ratio; WBRT = whole body radiation therapy.
GRADE Working Group grades of evidence. High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.
^aHeterogeneous population with varying primary cancer types; unclear whether groups were balanced by molecular subtype important for survival.

Summary of findings 4. WBRT and radiosurgery compared with WBRT for treatment of newly diagnosed multiple brain metastases

Summary of findings 5. Radiosurgery alone compared with WBRT and radiosurgery for treatment of newly diagnosed multiple brain metastases

Outcomes	Relative effect (95% CI)	*Anticipated absolute effects (95% CI)**			Certainty (quality) of the evidence (GRADE)	What happens
Radiosurgery alone compared with WBRT and radiosurgery for treatment of newly diagnosed multiple brain metastases
Patient or population: patients with newly diagnosed multiple brain metastases Setting: hospital Intervention: radiosurgery alone Comparison: WBRT and radiosurgery
Outcomes	Relative effect (95% CI)	Without radiosurgery alone	With radiosurgery alone	Difference	Certainty (quality) of the evidence (GRADE)	What happens
Overall survival No. of participants: 403 (3 RCTs)	HR 1.00 (0.80 to 1.25)	study population			⊕⊕⊕⊝ MODERATE^a
Overall survival No. of participants: 403 (3 RCTs)	HR 1.00 (0.80 to 1.25)	not estimable as only HR data available	not estimable as only HR data available	N/A	⊕⊕⊕⊝ MODERATE^a
1‐Year radiosurgery‐targeted lesion control No. of participants: 602 (4 RCTs)	HR 2.73 (1.87 to 3.99)	study population			⊕⊕⊕⊕ HIGH
	HR 2.73 (1.87 to 3.99)	not estimable as only HR data available	not estimable as only HR data available	N/A	⊕⊕⊕⊕ HIGH
1‐Year distant brain control No. of participants: 602 (4 RCTs)	HR 2.34 (1.73 to 3.18)	study population			⊕⊕⊕⊕ HIGH
	HR 2.34 (1.73 to 3.18)	not estimable as only HR data available	not estimable as only HR data available	N/A	⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 602 (4 RCTs)	not pooled	study population			⊕⊕⊕⊕ HIGH
Adverse effects No. of participants: 602 (4 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊕ HIGH
Neurocognition No. of participants: 403 (3 RCTs)	not pooled	study population			⊕⊕⊕⊝ MODERATE^b
Neurocognition No. of participants: 403 (3 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^b
Neurological function No. of participants: 544 (3 RCTs)	not pooled	study population			⊕⊕⊕⊝ MODERATE^b
Neurological function No. of participants: 544 (3 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^b
Quality of life No. of participants: 271 (2 RCTs)	not pooled	study population			⊕⊕⊕⊝ MODERATE^b
Quality of life No. of participants: 271 (2 RCTs)	not pooled	not pooled	not pooled	not pooled	⊕⊕⊕⊝ MODERATE^b
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI = confidence interval; HR = hazard ratio; N/A = not applicable; OR = odds ratio; RCT = randomised controlled trial; RR: risk ratio; WBRT = whole body radiation therapy.
GRADE Working Group grades of evidence. High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.
^aHeterogeneous population with respect to survival. Prognostic variables such as primary cancer histologies and molecular cancer features may not have been balanced between groups. ^bAttrition bias.

Summary of findings 5. Radiosurgery alone compared with WBRT and radiosurgery for treatment of newly diagnosed multiple brain metastases

Comparison 1. Altered WBRT fractionation schedules versus WBRT control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Overall survival: lower‐dose WBRT vs control WBRT (3000 cGy/10 daily fractions) Show forest plot	3		Hazard Ratio (IV, Fixed, 95% CI)	1.21 [1.04, 1.40]

1.2 Overall survival: higher‐dose WBRT vs control (3000 cGy/10 daily fractions) Show forest plot	4		Hazard Ratio (IV, Fixed, 95% CI)	0.97 [0.83, 1.12]

1.3 Overall survival: WBRT 4000 cGy/20 fractions BID vs control WBRT (2000 cGy/4‐5 daily fractions) Show forest plot	2		Hazard Ratio (IV, Fixed, 95% CI)	1.18 [0.89, 1.56]

Comparison 1. Altered WBRT fractionation schedules versus WBRT control

Comparison 2. Altered WBRT fractionation schedules versus WBRT control: neurological function improvement

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Neurological function improvement: lower‐dose WBRT vs control dose WBRT (3000 cGy/10 fractions) Show forest plot	5	1612	Odds Ratio (M‐H, Random, 95% CI)	1.74 [1.06, 2.84]

2.2 Neurological function improvement: higher‐dose WBRT vs control dose WBRT (3000 cGy/10 fractions) Show forest plot	4	1480	Odds Ratio (M‐H, Random, 95% CI)	1.14 [0.92, 1.42]

Comparison 2. Altered WBRT fractionation schedules versus WBRT control: neurological function improvement

Comparison 3. WBRT with radiosensitisers (radiosen) versus WBRT alone

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Overall survival Show forest plot	8		Hazard Ratio (IV, Fixed, 95% CI)	1.05 [0.99, 1.12]

3.2 Brain tumour response rates: complete response (CR) and partial response (PR) combined Show forest plot	6	847	Odds Ratio (M‐H, Random, 95% CI)	0.84 [0.63, 1.11]

Comparison 3. WBRT with radiosensitisers (radiosen) versus WBRT alone

Comparison 4. WBRT and radiosurgery versus WBRT

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 Overall survival Show forest plot	2		Hazard Ratio (IV, Fixed, 95% CI)	0.61 [0.27, 1.39]

4.2 1‐Year overall brain control rates Show forest plot	3		Hazard Ratio (IV, Fixed, 95% CI)	0.39 [0.25, 0.60]

Comparison 4. WBRT and radiosurgery versus WBRT

Comparison 5. Radiosurgery alone versus WBRT and radiosurgery

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
5.1 Overall survival Show forest plot	3		Hazard Ratio (IV, Fixed, 95% CI)	1.00 [0.80, 1.25]

5.2 1‐Year radiosurgery‐targeted lesion control Show forest plot	4		Hazard Ratio (IV, Fixed, 95% CI)	2.73 [1.87, 3.99]

5.3 1‐Year distant brain control Show forest plot	4		Hazard Ratio (IV, Fixed, 95% CI)	2.34 [1.73, 3.18]

Comparison 5. Radiosurgery alone versus WBRT and radiosurgery